Jump to content

eGPU experiences [version 2.0]


Tech Inferno Fan

Recommended Posts

 

THIS PAGE IS A RUNNING SUMMARY OF THIS LONG THREAD. >> SEARCH this eGPU forum.
AcqKcth.png
experiences [2.0]

N
E
W
S

 
j3junyB.png KPvnBEd.png 4EoyAFp.png tCFq0Ji.png ID1Nyjh.png uK8H5u6.png bRuy5In.png
 
GETTING STARTED >>>
Introduction A brief introduction with the story of the solution.
Pros and Cons A summary of the pros and cons of a DIY eGPU
Pre purchase FAQ Answers to questions before you commit
First steps Parts/software/info for a successful installation
Performance: AMD versus NVIDIA benchmark results Compares AMD and NVIDIA perf on the same sys
Performance: Scaling Analysis Compare performance over x1, x1E, x1.Opt, x2, x2E links
Implementations: Examples and Leaderboard PC and Macbook eGPU implementations by users
Implementations: older Macbook Implementations on older core2duo Macs
Implementations: CUDA/OpenCL Implementations using the GPU for applications+computation
Implementations: non-video Non-video implementations: soundcards atm
Non English eGPU forums eGPU forums in non-English speaking locales
Archived news Misc other articles

 

 



 

INTRODUCTION >>>


This thread serves as a roadmap for those who wish to attach an external Graphics Processing Unit (eGPU) to a notebook/laptop computer. Connection may be achieved through an ExpressCard, mPCIe slot, or Thunderbolt port.This thread is a place for users to share installation and performance details for others to duplicate or reference.

 

 

 

logo_masthead.png
* SimoxTav's Video game benchmarking* Graw/TWE/MafiaII12" Len.X201t+HD5770* StarcraftII13"_Dell_M1330+GTX460* Star Trek Online15"_Macbook Pro+GT240
el08.th.png on8j.th.jpg 8m54.th.jpg yv23.th.jpg
Gaming videos running on an eGPU implementation US$70 PE4L-EC2C (expresscard) or PE4L-PM3N (mPCIe) products $91 PE4H-EC2C (expresscard) or PE4H-PM3N (mPCIe) products HP 2560P GTX560Ti @x1.2Opt DIY eGPU impl playing NFS Shift on the internal LCD HP 2510P: [mPCIe] Swap wifi card via underside covers

Note: For those running Windows 7, you can install the eGPU without rebooting your system. Simply and the turn the system to standby system and attach the PE4L. Your system will return to its desktop presets once it detects the PE4L. You will immediately see a significant upgrade in the performance.

For example. a Inspiron 1440 using the x1E tweak saw 3dmark06 go from 866 to 9551. RE5 benchmark went from an unplayable 4.1FPS to a very playabe 60.9FPS. The HP 2560P's x1.2Opt link saw 3dmark06 increase over 4 times to 18340 and RE5 gave an average of 148.3FPS(!!), even allowing the internal LCD to be driven by the eGPU.

 

 

 

 

 
PROS AND CONS >>>

 

 

Pros

  • Inexpensive: US$92-shipped PE4L with HD5770+120W=$200, GTS450+120W=$160, GTX460-1GB+430W=$210, GT430+72W=$130, HD5670+72W=$140. Add $66 for a x2 capable kit using a PE4H.

  • can be used on any system with an accessible expresscard OR mPCIe slot

  • has inexpensive eGPU  enclosure examples if you want an enclosure

  • 10 times or more increase in video performance

  • can have Internal LCD-only setup: NVIDIA Optimus or Ultramon/Chung-Gun method.

  • provides additional HDMI video+audio, s-video and VGA out

  • provides dual-link DVI connection for 2560x1600 LCDs

  • allows multi-monitor connectivity:
    • AMD Eyefinity to provide up to 3 LCDs attached to any HD54xx or better card.

    • Drive up to 6 LCDs to using a HD7xxx or better card supporting Eyefinity 2.0

    • AMD FirePro 2450, 2450x1, and 2460 to drive up to 4 monitors from a low power card

    • NVIDIA Surround can connect up to 4 simultaneous LCDs.







    •  
    •  
    •  
    •  
    •  
    •  
  • easy, on-demand attachment without a reboot if running Win7

  • can be reused after a system upgrade, or shared amongst multiple systems

  • see an extra 30-60% in performance when attached to an expresscard 2.0 system, or do a x2 1.0 setup

  • provides an additional USB port when connected to expresscard slot

  • Time-share your desktop video card b/w your desktop and notebook system.

  • perform CUDA/OpenCL processing for applications that support it like Adobe CS5/Premiere [NVIDIA graphics only]

  • ability to use physX/hybrid physX & game detail

  • components can become a stepping stone to a full deskop system if want higher performance

Cons

 
  • requires a 12V PSU to drive the desktop eGPU

  • requires an external LCD/CRT/TV to connect to display at highest FPS  

  • may require booting DIY eGPU Setup 1.x from a disk image or USB stick to overcome error 12 issues

  • restricted bandwidth: a x1/x2/x4 link represents 1/16, 1/8 or 1/4 the bandwidth of a x16 slot.

 

PRE PURCHASE FAQ >>>

-> moved to http://forum.techinferno.com/diy-e-gpu-projects/7969-pre-purchase-faq.html#post109073

 

 

 

FIRST STEPS >>>

-> moved to http://forum.techinferno.com/diy-e-gpu-projects/7969-pre-purchase-faq.html#post109073

 

 

 

 

PERFORMANCE >>>


AMD versus NVIDIA back-to-back benchmark results

GTX560Ti vs GTX660 @x1.2Opt (int vs ext) (i5-2540M 2.6, bjorm)
HD7770@x1 vs x1E (P8400-2.26, damir001 inc PCISpeedTest)
HD7870 vs GTX660 at all expresscard/mPCIe bandwidths available (i5-3320M 2.6, Tech Inferno Fan)
HD6970@x4 2.0 vs GTX460@x4 2.0 vs x16 2.0 (i5-2400 3.1, tomshardware)
[email protected] vs GTX560Ti (i7-2820QM 2.3, carage)
[email protected] vs [email protected] (i7-2630QM 2.0, timohour)
NVS4200M vs [email protected] (i7-2670QM, Tom_ger)
[email protected] vs x1.1Opt and GTX580 versus [email protected] (i5-2520M 2.5, arsenal243)
GTX480 versus GTX560Ti@x1Opt (i7-620M 2.66, Mikjoa)
GTX460-1GB versus GTX560Ti @x1.1Opt (i5-2520M 2.5, Warwickeo)
GTS450 versus GTX460 @x1.Opt (i7-640M 2.80, docusyn)
HD5750 versus GTX470 on x1, x1.Opt and x2 links (i5-430M 2.26, Tech Inferno Fan)
HD5750 versus GTX460 on x1 and x1E links (T9300-2.5, ithildin)
HD5750 versus GTX470 on x1, x1E, x1.Opt and x2 links (T6600-2.2/T2050-1.6/U7600-1.2, Tech Inferno Fan)

Scaling Analysis

(07-05-2014) [Performance] Desktop vs eGPU - GTX770 + R9 270x
HD5870@x2 1.0 vs x16 2.0 (i7-720QM/i7-3820, kizwan)
HD7970 versus GTX680: pci-e scaling x4, x8, x16 @pci-e 1.0, 2.0 and 3.0
HD7970: pci-e 3.0 x2, x4, x8, x16 scaling
GTX580: x1.2Opt vs x1.1Opt +30% overclock & (i7-2920XM 2.4, Khenglish)
GTX560Ti: x1.2Opt vs x1.1Opt [internal LCD vs external LCD] (i5-2540M 2.6, Tech Inferno Fan)
HD5850: x16, x4, x2, x1 (~i7-2820QM, timohour)
HD5870: x2 versus x1 (i7-720QM 1.6, kizwan)
GTX580+GTX460: x2 vs x1 vs GTX260M (P8700-2.53, mangos47)
[email protected] scaling analysis internal vs external LCD, +25% PCIe, +40% GPU core O/C ([email protected], Tech Inferno Fan)
GTX460@x2 scaling analysis +20% cpu, +20% PCIe, ~20% GPU (T9400-2.53, khenglish)
HD5750: x1 versus x1E (T9300-2.5, ithildin)

 

 

 

 

Older MACBOOK >>>


04-03-2011 MacBook 17 T7600-2.33 with X1600 + HD6850 implemented with Win7/32
11-10-2010 Mac Mini T7200-2.0 + HD5450 implemented with Win7.
11-05-2010 MBP 17" T2600-2.16 with X1600 + GT240M result with Win7.
9-27-2010 MBP 15" Unibody T7500-2.2, 6GB RAM with 8600M GT+HD4650/HD5770-7LCDs impl. & & & with Win7.
8-11-2010 MBP 15" Unibody T7500-2.2 with 8600M GT + HD4870 implemented with Win7 and MacOS 10.6

For MacOS DIY eGPU implementations see DIY eGPU MacOS Experiences.

 

 

 

 

CUDA/OPENCL APPLICATIONS >>>

Octane Render: 10x faster than CPU-based rendering.
Badaboom: CUDA based video transcoding

 

 

 

 

NON-VIDEO PCI-E CARDS >>>

Creative SB X-Fi Titanium desktop sound card via the ExpressCard slot. See details.

 

 

 

 

NON ENGLISH eGPU FORUMS >>>

 

 

 

 

 

 

ARCHIVED NEWS >>>

 

 

 

 


Acknowledgements Many thank yous to all the members who have shared their DIY eGPU experiences for others to be able to learn from and duplicate. Special thanks to SimoxTav for the artistic touches.

Project saboteurs of this community-driven project who halted/derailed the creative opportunities that pluggable, affordable eGPU graphics/processing provides.

 

 

 

  • Thumbs Up 47
Link to comment
Share on other sites

@ nando If bonding is possible, is it possible to connect two mpcie and one express card to have 3x pcie? Furthermore, after researching pm55 chipset it is pcie 2.0 compliant as well as the CPU. I do not know what would happen if I changed the CPU to Intel® Core™ i5-580M but I believe the interconnects for the Integrated GPU are not connected however I believe that the GPU would still appear and allow the pcie data compression to be enabled. i7-640M and the i5-580M are both compatible with the PM55 however with the loss of two cores but a massive bump in core speed. Thoughts?

Unfortunately your system is x1 1.0 capable only, so there is nothing more that can be done to extract better performance. Consider:

* PM55 is pci-e 2.0 compliant but only for power management. The link speed is still pci-e 1.0. You need a Series-6 chipsets or newer for pci-e 2.0.

* Last time I checked, the port listing of a DV7 wasn't capable to support x2. There would need [port1+2], [port3+4], [port5+6] or [port7+8] accessible as expresscard+mPCIe or mPCIe+mPCIe). To answer your question, x3 isn't possible.

* Putting an Arrandale CPU like a i5-580M into your system won't enable the iGPU to be the primary bootup video device. Sure, you can enable it using Setup 1.1x but the NVidia driver will only engage pci-e compression if the iGPU is the primary bootup video device. So no x1.Opt would be possible.

My suggestion is to offload the system for an inexpensive Sandy Bridge system like a 17" Dell Vostro 3750. Then you'll be able to enjoy x1.2Opt performance, approx 4 times more bandwidth than the best your DV7 can do (x1 1.0).

Link to comment
Share on other sites

[Posted 6-7-2011 in DIY eGPU experiences (NBR). Relocated to T|I due to being banned by Lenovo fan ZaZ]

<a name=bplus-updates></A><span style="font-size: large">Thunderbolt, MXM, USB 3.0, PCIe 2.0 eGPU update (Series-6+ chipsets)</span>

<span style="font-size:large">Updates from BPlus</span>

Future x2 2.0 eGPUs via M.2/NGFF (mSATA) slot

I wrote to BPlus earlier this week:

> Further to my request to get a socketted mPCIe Gen2 solution, consider too that Apple's 2013 MBA has a x2 2.0 pcie SSD slot.

> If other manufacturers do the same AND give a second drive (2.5" 7mm/9.5mm), then that x2 2.0 slot could be used in the same

> way for eGPU connectivity as mPCIe has been used in the past. 10Gbps would match current Thunderbolt speed.

>

> Critical in getting a workable solution there would be a socketted Gen2 cable for a PE4L/PE4H. At least that it is socketed on one

> end (the mPCIe end). Please do your best to try to make it happen!

>

Masaharu @ Bplus replied:

Silicon Power T20 is Thunderbolt to SSD solution, soon be released;

Silicon Power Presents Superior Thunderbolt Series at IDF 2013 | techPowerUp

T20 may have PCIe SSD, such as SanDisk A110 PCIe SSD;

SanDisk A110 PCIe Solid State Drive

In this case, PCIe X16 graphic card might be replaced from PCIe SSD by using M.2 / NGFF to PCIe adapter.

6-15-2013: Masaharu about the new Gen2 compliant PCIEMM-060B detachable cables

> I see you guys have updated at

PE4L ( PCIe passive adapter ver1.5 )

> with a new "Gen2" capable PCIEMM-060B cable that's different to thePCIEMM-060A one.

> Users have desired detachable Gen2 compliant cables for a while now, especially

> for mPCIe use, so are very excited about this new product.

Yes, now we have PCIEMM-060B.

The cable is narrow, 2.5mm diameter, very high quality with Sumitomo cable inside.

> Can you confirm this does pass Gen2 signals OK?

> We had a fiasco a few months back with the PCIEMM-060A claiming

> as such then users finding it could only pass Gen1.

We tested PCIEMM-060B many times and confirmed

Gen2 / 5Gbps support and much better than 060A.

> I'm happy to test one for you with a Series-7 (3rd Gen i5) 2570P + GTX660Ti I have.

Ok, we will provide two of PCIEMM-060B to you.

Please them with you PC to confirm.

03-27-2013: Masaharu's reply requesting MXM3 product

Dear Mr. Nando,

Thank you for your information.

Roger,

Please refer MXM3 information for new product.

Best Regards,

------

Hi Masaharu,

Please monitor the development of this: EXP GDC 笔记本Expresscard to MXM3外置独立显卡 标准套装

We've recently had a result of a user attaching to the BPlus eGPU gear to a htpc pci-e 3.0 port getting x1 3.0 results:

http://forum.techinferno.com/diy-e-gpu-projects/3094-egpu-desktop-htpc.html#post44463

Now notebooks have expresscard slots off the Southbridge which is limited to pci-e 2.0. The Northbridge on a Series-7 notebook is pci-e 3.0, however, no manufacturer is attaching the expresscard slot off that port. Meaning, you could route the MXM slot out as an expresscard 3.0 slot using an adapter to a PE4L 2.1b and get x1 3.0 (inc Optimus) performance using an eGPU. Then again, you could route it as a full external x16 3.0 slot too :)

Something to keep an eye on.

Nando

12-14-2012: Masaharu's reply to x1 jumper request on next TH05 revision

> One other request if I may. For the next TH05 version, could you put a jumper that disables the second lane?

> We've discovered that the NVidia driver engages pci-e compression on a Intel HD iGPU systems only when running x1

> (single lane). This gives slightly better performance for DirectX 9 titles, but is worse for DX10/DX11. So such a jumper

> would mean users wouldn't need to cellophane tape lane2 to get the faster performance.

Thank you for suggestion. Yes, we can understand the situation.

PortRidge has ROM setting to choose x1, but it will fix to x1, so it is not good solution.

Regarding the jumper for 2nd PCIe lane is very interesting, and we have to reduce impact of the jumper for the PCIe

lane. We will think about the jumper and other way to disable 2nd PCIe lane, when we have chance to revise the board.

12-14-2012: Masaharu on new product that mixes best of current PE4L 2.1 and PE4H 2.4 feature set

> It occurred to me that there could be another product that could be the best seller from the current

> PE4L/PE4H releases. I really liked the PE4H 2.4 form factor, it's x2/x4 ability and it's US$91

> pricepoint but it was missing the following features:

>

> - pcie 2.0 on the first lane with options for the other 2-4 lanes.

> - 12V input power

>

> So I propose you consider making a PE4xx with the following modifications to PE4H 2.4. This may tide us

> over until 20Gbps Thunderbolt is made available in 2014.

>

> 1. solder lane1 directly with PM3N or EC2C to give it pci-e 2.0 capability. Then it would be a

> PE4H-EC060A, PE4H-PM060A. This would cater to Series-6/7 systems that are pci-e 2.0 capable as well as

> older Series-5 systems using just a single lane.

>

> 2. factory solder lane2-4 as mHDMI connectors so as to give x2/x4 pci-e 1.x capability. This would cater

> to Series-5 or older systems that are x2/x4 pci-e 1.x capable.

>

> 3. spread out lane2 - lane4 as solderable pads. Meaning, the factory provided mHDMI connectors in (2)

> could be removed and BPlus or a user could solder 7 wires directly onto the pads to form pci-e 2.0

> additional lanes. This would allow x2/x4 pci-e 2.0 development work to be done using the board.

> Enthusiastic users may want to try to change the flash descriptor of Series6-7 systems to gain x2/x4 pcie

> 2.0 link support. This hardware would then be available to them to do that.

>

> 4. change the DC input plug to accept 12V rather than 19V. Add voltage regulators to provide 5V/3.3V when

> running that way. I prefer the PE4L direct 12V input and would like the PE4H to be able to do the same.

Thank you for your suggestion. The product is placed between PE4L and PE4H v2.4. 12V DC Jack sounds good

to reduce on-board components. Also, we may use PCIe x4 connector for smaller PCB.

We discuss this product for 2013 Q1 development.

11-25-2012: Masaharu on x4 2.0 Thunderbolt device ETA

> When will the x4 2.0 TH05 be available?

We hope we can release it Jan 2013 for engineering sample.

However, the schedule is unstable.

Intel is going to release new 6 layer reference design,

much cheaper than previous 10 layer one.

We saw the PCB directly in their TBT seminar in Taipei,

the design is very mature and very small, so we would like to

use the design with almost no change.

However, the PCB design is not released yet, because

the reference board is still under certification process.

As the schematic and the BOM are released, we are

ready to kick off, and will start to design as soon as the

PCB design is released.

Intel said the PCB design will be released Nov/E, and

the schedule is kept, we can also release our TBT-PCIeX4

in Jan 2013.

10-01-2012: Masaharu on x4 2.0 Thunderbolt TH05 device

> Lastly, is there any way you could make a prototype x4 2.0 capable

> Thunderbolt-to-pcie adapter within a relatively short timeframe?

> I'm guessing it would not be too difficult to extend what was done with the TH05.

> That is, a bigger board to accomodate a presumably larger

> Thunderbolt chip + 2 extra lanes.

Thunderbolt x4 2.0 product itself is not difficult for us at present,

because we already finished TH05 as you mentioned.

However, we still hesitate to develop it because of some reasons;

1. Thunderbolt is just started this year so politically unstable

between Apple and intel and may take very long time

to release logo'd products.

2. Due to 1., we do not have many product lines.

At present, we use "PortRidge" only, the simplest and the cheapest

Thunderbolt controller and the external components are not so many.

Thunderbolt (interface) - Wikipedia, the free encyclopedia

If we make x4 products, we have to use expensive CactusRidge and

lots of external components including Display Port ones.

So, until issue 1. is not solved, we would like to develop

PortRidge related products.

9-26-2012: Masaharu on PE4L-L060A being pci-e 1.x only!

> Q: Is PE4L-L060A pci-e 2.0 capable?

> I know you wrote me saying it is, but I had another user write

> that it's only pci-e 1.0 capable, quoting an email he received from BPlus.

> Can you advise once again if PE4L-L060A with a EC2C or PM3N is pci-e 2.0 capable?

Basically, L060A can not support pci-e 2.0.

On some desktop system, L060A can work as pcie-e 2.0, however the behavior is not stable.

Besides, L060A can not work as pci-e2.0 on some laptop PCs on which PM060A or EC060A can works as pci-e2.0.

So, for pci-e 2.0 tests, we recommend not to use L060A, but to use PE4L-EC060A 2.1b and PE4L-PM060A 2.1b.

7-19-2012: Masaharu on PE4L-L060A, mPCIe double-adapter and Thunderbolt notebook availability

> PE4L-060A details pls

> 1. RE: PE4L-L060A product. Can you advise if this can negotiate

> pci-e 2.0 successfully? If so, I can see no real reason to use/recommend

> the PE4L-PM060A/PE4L-EC060A as that product offers the same

> convenience previously seen in the PE4L 1.5 or PE4H 2.4 of matching

> the end connector to the system (mPCIe/expresscard, etc).

Yes, it was successfully connected with PCIe 2.0. Basically, before we develop EC060A, we made L060A type sample and direct cable soldering to PCB worked fine.

> mPCIe double-adapter?

> The aforementioned mPCIe double-adapter (like MM3U) could be

> extended to encompass many more uses. Such a double-adapter

> would need clever design as below I propose fitting two of them

> in the same cavity. Consider:

> i. to allow eGPU to be switch through place of the wifi (pci-e),

> where it would route pci-e signals to the eGPU when attached.

We are very sorry but we are not able to imagine the product well. Also, we think to switch PCIe devices might be more difficult.

> Thunderbolt development machine for low $$?

> Some users in the DIY eGPU experiences thread have requested I be more

> involved with the Thunderbolt eGPU development. Given the small amount

> of Thunderbolt systems available I'm wondering if any vendors have provided

> you some sample/demo machines to work with? I personally wouldn't buy a

> Apple notebook myself but in this case is probably the best system to use

> for testing and development (eg: 13" Macbook Pro). Can you get one cheaply?

To get the Thunderbolt products are headaches for us too.

Unfortunately, Thunderbolt group does not offer group discount or such kind of thing to us.

Now, it is the best way to use my nephew's name to get academic discount from apple. I heard there is Thunderbolt plugfest in Sep or Oct, and we will meet a lot of Thunderbolt vendors. So, we will ask them if they have discount.

7-07-2012: Masaharu on TH05 TB progress (x2 2.0!!, no DP-to-internal LCD), KZ-B21/KZ-B26 mPCIe products

> The best person to get a sample would be the Sony Z2 owner Mikjoa that

> has just used a PE4L-PM060A to do the first mPCIe 2.0 DIY eGPU

> implementation here.

> I can ask if he'd like to do testing and writeups with this KZ-B26 but

> I'd only ask if it was pci-e 2,0 compliant.

Unfortunately, PCIe 2.0 is not successful in our lab, probably because

using PCIe-mPCIe adapter on desktop PC.

So, when KZ-B26 is directly attached to laptop mPCIe slot, it may

be still chance to work properly.

> KZ-B26 questions

> 1. Is it pci-e 2.0 compliant end-to-end?

So far, it is not successful.

However, similar product KZ-B21 using equivalent cable works

fine with PCIe 2.0, so we suppose KZ-B26 should work fine with PCIe2.0.

KZ-B21 (Flexible x1 PCI Express Extender)

Also, shorter cables (10cm or 20cm) have much more chance

for PCIe2.0 than present (30cm/12inch) cable.

> 2. I can see the ZIF cable is flexible. Would the cable be durable enough

> to be threaded out of the chassis, be wrapped under the chassis and

> taped down when not in use? Then when required just undo tape

> and connect to the ZIF-expresscard connector?

We suppose KZ-B26 can work the kind of use.

However, it's a kind of "drip injection", it can not work well hard mobile work.

> 3. Would it be possible to go one step further and create some form of

> double-adapter allowing a wifi card and the ZIF-PM3A adapter to occupy

> the same mPCIe slot, similar to your MM3U? So when the ZIF cable was

> attached to the expresscard-ZIF adapter, it would detect it and disengage

> the wifi card from working?

It's interesting idea!

We will think about it in the next version.

> Thunderbolt details with TH05 questions

> 1. It is mentioned on your facebook page that it is 2 channel pci-e 2.0 link.

> That is x2 2.0.

Yes, TH05 support PCIe2.0 x2.

> The concern me since Sandy/Ivy Bridge expresscard is x1 2.0 which

> approximates to x2 2.0 with Optimus' pci-e compression (requires iGPU +

> NVidia Fermi/Keplar card and a x1 link). So x2 2.0 Thunderbolt isn't

> offering us any more performance.

We can understand the situation, PCIe2.0 x2 is not large leap

from PCIe 2.0 x1.

> What I'm wondering is whether x4 2.0 and/or x4 3.0 Thunderbolt will

> be available with the TH05? We have very little detail on the Thunderbolt

> controllers other than the Series-6 chipset has a x4 2.0 link for the

> Thunderbolt controller and the Series-7 chipset is x4 3.0 capable.

Unfortunately, x4 can not be supported by TH05, because we

use IC with Thunderbolt x1 channel and PCIe 2.0 x2.

In 2012, and early 2013, we will keep using the IC, though

there are TBTx2+PCIe2.0x2 ICs.

First we will develop all product with TBT IC with TBTx1+PCIe2.0x2.

After we get experience of TBT, we will use TBTx2+PCIe2.0x4 for

higher performance.

Step by step and tick-tock.

> So that link needs the right matched components (Thunderbolt

> mux/demux controllers and cabling) to get full performance out of it.

> Finding the right candidate machine may require figuring out if they

> have x2 2.0 or x4 2.0 or x2 3.0 or x4 3.0 TB links.

> THe manufacturers of course are not telling us any of that detail.

We suppose PCIe2.0x4 (5Gbps x4) is reasonable, because

TBT I/F is 10Gbps x2.

So, TBT I/F itself will be bottle neck for PCIe3.0x4.

PCIe3.0x2 may possible, however, we do not see PCIe3.0 on

intel road-map so far.

> 2. Are you going to re-route the eGPU's Displayport output back via

> Thunderbolt to the system? That would mean that if the system had

> a muxed internal LCD it could display output to the internal LCD

> without incurring any bandwidth penalties as we are seeing with

> NVidia Optimus. It would require a mDP connector on the TH05

> to attach to the video card's HDMI/DP connector.

So far, we do not have plan to add DisplayPort to our products,

because TBT IC we use does not have the option.

After we use another IC with PCIe2.0 x4, we will add DP port.

> 3. Obviously people want to get their hands on these TH05

> Thunderbolt-to-pcie adapters. How far are we realistically away

> from seeing a publically available product? Is there any NDA

> preventing beta/engineering samples being released?

Yes, intel's NDA is very strict.

We suppose intel wants to control TBT market strictly

because it's very high speed and the market should not

be messy with low quality products.

So only intel certified products can be released, and

hopefully TH05 should be certified in the end of August.

After that we can release samples for limited quantity.

6-07-2012 BPlus Computex 2012 Thunderbolt/eGPU products

6-05-2012 NEW! US$92-shipped PE4L-PM060A (mPCIe 2.0)

5-29-2012: Masaharu on PE4L 2.1b product listing and mPCIe PE4L-PM060A 2.1 availability

> Can you clearly label you are selling PE4L 2.1b now as have confused forum members . . .

We are very sorry about this, and thank you for your proposals. We know PE4L v2.1a issue because we got RMA from customer and confirmed CLKREQ# pin needs rework for some of released v2.1a.

There are some ways to tell v2.1a or v2.1b;

- Shipped after March 23rd.

- Version Silk on PCB

We will update our web site of PE4L v2.1b to see this information.

Also, MiniCard version of PE4L 2.1 (pci-e 2.0). We got samples of PE4L-PM060A. Please refer attached jpg file for the product image.

clip_image001.jpg

5-22-2012: Masaharu Thunderbolt, mPCIe products and pci-e 3.0 compliant cables

> 1. details of the gerbered Thunderbolt products?

One is TH03, ExpressCard 54 Thunderbolt Adapter. TH03 ( ExpressCard 54 Thunderbolt Adapter)

The other one is TH05L/M. PCIe X1/X2 - Thunderbolt adapter. It looks like PE4L. Sorry no photo at present.

> 2. Any pics/photos of the pci-e 2.0 mPCIe product with the detachable cable extender. This would help the users visualize the solution.

Unfortunately, sample is not arrived yet. When we got them, we will send the photo to you.

> Series-7 chipset (Ivy Bridge) is a x4 3.0 capable Thunderbolt interface. Meaning I'd recommend ensuring pci-e 3.0 compliance of your cabling solution as soon as possible.

So far, PE4L v2.1b achieved PCIe3.0. As mPCIe-PCIe solution will be same configuration as PE4L v2.1b, we hope it can also support PCIe3.0.

Thank you for your information. We will develop other product lines, such as enhanced PE4H or mPCIe solutions as you suggested.

5-18-2012: Masaharu - mPCIe pci-e 2.0 product with detachable cable available

Regarding mPCIe product, Gerry developed mPCIe to mPCIe extender with FFC type flat cable. It is like PCIe-PCIe type KZ-B21;

KZ-B21 (Flexible x1 PCI Express Extender)

and front-end changed to mPCIe. So, we should make PCIe-mPCIe type product, and we think it probably satisfy your request.

As KZ-B21-30 (30cm) achieved PCIe 2.0, we hope PCIe-mPCIe one also will be PCIe2.0.

5-02-2012: Masaharu - mPCIe pci-e 2.0 product sample available

> Apart from the 13" Sony Z2, a 11.6" Alienware M11xR3 (with a HUGE userbase)

> could also benefit from a PE4L-PM3N 2.1.

Regarding that kind of product, we already developed samples, after we asked my boss.

The product name is "PE4L-PM060A V2.1b". PE4L portion is same as PE4L-EC2C v2.1b?and PM060A stands for Pcie Minicard (Half size) and 60cm cable.

We hope we can test in the middle of May.

3-11-2012: Masaharu responds to PCI-E 3.0 compliance request

> Appears Intel's upcoming Ivy Bridge CPU will have PCIe 3.0 compliance.

> Means the dGPU link into the CPU (x16 port), that's used by Thunderbolt

> by at x4 only, will be able to transmit at 8GT/s.

I am also curious for PCIe 3.0. We would like to buy PCIe3.0 M/B and video card to try them.

> Can you make the cabling/connectors on the PE4L/PE4H to be PCIe 3.0 compliant?

We will test first with PE4L v2.1a can be capable for PCIe3.0. Also, not only 60cm cable but also longer and shorter cable how the PE4L can do with PCIe3.0

3-7-2012: Masaharu on PE4L 2.1 manufacturing improvements

Dear Mr. Nando,

> One confirms he too ran his PE4L and video card in a perpendicular

> arrangement, so there was 'slight' twisting around the EC2C point.

> He, like me, did believe the device was resilient enough to handle

> that minor force.

First 50 units run with enhancement of current design. Double of heat shrink tube and glue inside of EC2C point. Please attachment.

> Regarding bottom plate for PE4L v2.1, we will have similar one as PE4L v1.5.

Regarding bottom plate for PE4L v2.1, please refer attachment.

clip_image002.jpg

clip_image003.jpg

3-2-2012: Masaharu RE EC2C end reinforcement of PE4L 2.1/PE4H 3.0

Dear Mr. Nando,

> Hmm.. a bit distress this news. I had hoped some interim workaround could be

> put in place to secure the EC2C point in place.

We found enclosure below, it's for ethernet ExpressCard and generic to obtain easily. mHDMI cable can go through the ethernet connector, so we can use it somehow and production will have running change if it is available.

clip_image004.jpg

> I'll give another warning about RMA/return units.

> There is also at least two other users who are having problems

> with their first-release PE4L 2.1a units detecting their video card.

> One confirms he too ran his PE4L and video card in a perpendicular

> arrangement, so there was 'slight' twisting around the EC2C point.

> He, like me, did believe the device was resilient enough to handle

> that minor force.

Thank you for your advice. At first, we enhanced power line in mHDMI cable soldered at ExpressCard to apply heat shrinkable tube so there is little possibility to have short circuit to contact to ExpressCard enclosure.

> All usual troubleshooting has been exhausted to diagnose the problem.

> Which leads me to believe those users too may have faulty PE4L 2.1 units.

> Should they be advised to seek a RMA for their 'faulty' units?

It is the way the engineering sample is. At present, they may have same issue with the present product version, because it is a little bit improved from your version.

So, if they need RMA, we recommend after this issue is comprehensively solved and become mature.

3-1-2012: Masaharu RE socketted mPCIe pci-e 2.0 product desired

Dear Mr. Nando,

> You can see an excellent Lenovo Z570 Series-6 mPCIe DIY eGPU implementation at

> DIY eGPU experiences - Page 867

> The flat cable allowing easy insertion/removal from the PM3N when has a nice cutout like shown.

This rework is great.

> Now the upcoming PE4L-PM3N 2.1 won't be able to be inserted in that same way since it's a

> PE4L+mHDMI cable + PM3N all soldered together.

We suppose cutout still work for soldered PE4L-mHDMI-PM3N, and also cutout area will be smaller because fat connector portion will be removed.

> Is there no way you can investigate other connectors to try to achieve pci-e 2.0 signalling

> so that at least the EC2C/PM3N end can be socketted? Inserting the whole PM3N is not

> such a practical solution and causes wear on the mPCIe socket.

We also wondering if there were connector to realize it.

We suppose any connectors do not have so much difference such as

about contact impedance. At this time, we have to think about

cable characteristics, such as impedance 100ohm and maximum frequency.

After that, we have a few options, such as HDMI, miniDisplayPort or Thunderbolt cables.

At present, mHDMI is the most reasonable finally, so we are not able to

switch to other cables.

2-19-2012: Masaharu RE mPCIe products and PE4H 3.0+enclosure

> Can you comment on plans/ETA for a pci-e 2.0 mPCIe-based PE4L/PE4H per my last message?

Yes, we have a plan to have mPCIe-PE4Lv2.1, maybe in April. Probably we can release mPCIe-PE4H v3.0 after releasing ExperssCard version.

> Would be useful to know ETA of the PE4H 3.0 and enclosure products as well.

We will release PCB version of PE4H v3.0 first. Enclosure version will be after a few month later of the release, so we hope we can release it in May or June.

2-10-2012: Masaharu RE Pictures of sample PE4H 3.0

> Now did you have any sample photos of the PE4H 3.0 and perhaps the enclosure to go along with it?

We just got the PE4H v3.0 sample. Please refer attached JPG pictures. We hope you like it.

clip_image005.jpg clip_image006.jpg

2-2-2012: Masaharu RE PE4H 3.0 features, REQUEST higher power PE4L jack (no go)

> Excellent news on he PE4H 3.0 is on the way.

> Does that mean you can provide some photos of it?

When we got samples from Shanghai, we can provide them.

> As to the features list.

> - Direct ATX power connection: 300W available

> Do you mean there is a 24-pin ATX connector on the PE4H 3.0

> so can plug directly into it?

Yes, 24pin female connector is on PE4H v3.0.

> While that might be handy consider that the a pci-e 1.x slot

> only needs 75W. A pci-e 2.0 slot is rated up to 150W

> but I've never seen one uses that much power..

> I believe they still adhere to 75W.

Power supply from PCIe slot is 75W. 300W is available 6pin and 8pin PCIe power connectors are connected from ATX power to VGA board.

> The other benefits of the ATX connector on the PE4H 3.0 would be:

> 1) you can add circuitry to power on/off the ATX PSU via the PS_ON pin.

> 2) many ATX PSUs are dual rail, though there are single rail ones.

> Having the ATX connector allows access to the one 12V rail.

> The molex/pci-e power connectors tend to be one another 12V rail.

PE4H v3.0 power is controlled by the hardware power switch and ExpressCard 3.3V power. Power from ATX power and 19V AC adapter are controlled by ExpressCard 3.3V power, external PS_ON is not necessary.

> - PCIe2.0 x1 capable

> - Enclosure available

> Pictures/pricing of the enclosure please.

> We have a new eGPU logo you can create as a sticker: See here.

Unfortunately, we are not able to see the picture even though we registered. We will check it again later.

> - Power & Reset timing switches

> If I am correct about the ATX connector then can have

> an ON/OFF/AUTO power switch where AUTO detects

> the pci-e signal lines and sends an active low to the

> ATX connector's PS_ON when a signal is detected to power on the PSU?

Unfortunately, PE4H v3.0 is not so intelligent. PE4H v3.0 just check 3.3V from Expresscard, it means, ExpressCard slot portion of PE4H v3.0 is inserted to ExpressCard slot or not. If ExpressCard 3.3V is provided, main power will be on.

> Comments RE: PE4L strengthened tracks to support say 200W via DC jack

> Also, can you comment in my last emai?. There I requested the PE4L 2.1's

> 12V/GND PE4L tracks and DC jack to be strengthened to allow say 200W

> to pass via the DC jack and so can then be split to pass out via the floppy

> molex connector as inputs into the pci-e power connector. Some users like

> the small profile of the PE4L and would like to power a card such as a

> GTX560/560Ti using an AC adapter in a portable fashion. Splitting power

> in that way makes it possible.

Regarding PE4L v2.1, 12V/5V could be maximum. We checked PE4L layout. For DC jack connector to floppy power cable, 12V/3A (5A at most) should be capable. So, additional jumper is required as you suggested.

Even though we tried to have additional power plane for 12V, we were not able to find it on present PCB surface budget. So, we also suggest to add external jumper to enhance 12V and GND between DC jack connector and floppy power connector to use 150W AC adapter for PCIe 6/8pin power

<span style="background-color:yellow">01-21-2012: PE4L 2.1 (pci-e 2.0) will be available on Feb-20-2012 as a US$70+shipping engineering sample. Detailed photos of one are here.</span>

1-20-2012: Masaharu from BPlus response to 12V bypass for PE4H DC jack + PE4H 3.0 arrival

> One other thing I know folks wanted in the PE4H was a 12V DC-jack passthru to supply the 75W (max) to drive the slot.

> The idea being they could set a jumper or switch that bypassed the 15-19V->12V regulator and just use the DC jack with a

> 12V AC adapter.

It's good idea, we would like to add it next revision.

Regarding PE4H v3.0, first sample was arrived at Shanghai. Some of brief tests are finished, and PCIe2.0 is good at 100cm cable. We hope we can have additional evaluation after Chinese New Year.

A happy Chinese New Year!

1-19-2012: Masaharu from BPlus response on cable length and some other requests

>Upon asking the DIY eGPU community for cable length we get all sorts of replies b/w 45cm to 200cm.

At present we consider 60cm for product release which is same length as cable length with v1.5 package. At this time at least 100cm must be achieved for the 60cm product release. For internal testing, the target is 200cm, we may have 100cm if the 200cm is ok.

> 1. You can see that the original idea of a socketed cable was a good one. Perhaps continue to test different sockets

> and even keep the ReDriver idea in mind with the goal of being able to have a variable length socketted cable

> (NOT soldered).

Regarding this kind of configuration, we already tested and failed many times. So, we came to conclusion that mHDMI cable must be soldered at the both ends of the cable.

> 2. Can add a new PS_ON output connector on the PE4L/PE4L? This was requested some time ago.

As v2.1 was derived from v1.5 so the PS_ON is not applied for v2.1. We will have the PS_ON and other small fixes on next version, because it can not be on time for v2.1 release.

> 3. Ensure you have 12V->3.3V voltage regulation @ 3A (10W) so the user doesn't need an auxilery 5V/3.3V input.

> That is the pci-e spec for maximum 3.3V input power.

12v-3.3v regulator can not be applied for PE4L because the regulator exceeds thermal envelope of PE4L. Instead, we will apply another way that put 5V from USB port for 5v-3.3v regulator to achieve 3A/10W for 3.3v power plane;

3.3V from ExpressCard: 1.3A

5V-3.3V on PE4L: 2A

Total 3.3V current: 3.3A

> 4. For the PE4H 3.0, please add some more perspex height to the base. If I installed my video card on previous

> versions then the slot cover plate was slightly higher resulting in the card lifting itself out of the x16 socket

> so I had to put it on another base to give it some more height.

This time, we have different type of PCIe x16 slot connector so that your current issue will not occur. Please refer picture below. There is 4mm space on the rear side of PCIe boards so the most of VGA can be used.

clip_image007.jpg

* This picture is different product's prototype. PCIe x16 slot arrangement is same.

1-17-2012: Masaharu from BPlus finalizes the pci-e 2.0 compliant PE4L 2.1 pricing

My boss Gerry Chen quite respects you and he said we have to hear voice of you and market especially in this sector.

So finally we decided to put $70 tag for PE4L v2.1.

Thank you for your suggestion to give us a chance to reconsider the price before making mistake in the market.

1-17-2012: Tech Inferno Fan response requesting lower PE4L 2.1 price for next 6 months

I am yet again not sure why BPlus wish to charge $105+shipping for the PE4L 2.1, regardless of it being a engineering sample or final product. You advised me on 11-11-2011 about pricing:

price is not unknown yet either. However, we would like to apply reasonable price compared with current PE4H products.
Based on those comments, I know DIY eGPU users have avoided purchase of competitor hardware and put their faith in BPlus delivering a cost effective solution. A $105+shipping price is a nasty surprise after a 1 year wait. The comment above was also made when the PE4L cost $55. So I consider x2/1.5x of then/now pricing as being unreasonable. A $105 pricepoint is 3 times more than the original project guideline of $35.

Please advise if you can guarantee a price of US$71 for an engineering sample or final PE4L 2.1 product for the next 6 months. That is 1.3x the previous $55 price of a PE4L which is reasonable. It would be a terrible loss for everyone if BPlus was no longer the preferred source of the DIY eGPU hardware because of unreasonable pricing.

1-17-2012: Masaharu from BPlus clarified PE4L 2.1 pricing and requests optimized cable length

Thank you for your suggestion, and I am very sorry that my explanation in previous mail was not sufficient.

PE4L v2.1 at $105 is ES sample price. There is some duration period before MP, and we have some pilot run ES samples at that time for evaluation.

If somebody need to obtain earlier even the ES samples, we can provide them at the ES price with limited quantity.

MP price is not fixed yet, however, we can provide more reasonable price for it as you suggested.

Now, I need a help. Regarding cable length, we would like to know the optimized length, because the cable cannot be changed. The length of PE4L v2.1 sample we sent you is 25cm.

We suppose it might not be long enough, but should not be too long for handling. So, we would like to know how you feel the length and the optimized length.

1-16-2012: Reply from Tech Inferno Fan to Masaharu at BPlus last message below about pricing

]I have serious reservations recommending the PE4L 2.1 if the price will be $105 + shipping (1.5x of PE4L 1.5 price). This is because:

(1) the original PE4L was set at $35 so I consider it to be gross profiteering and no longer aligned with the community spirit of the project

(2) The Thunderbolt-based MSI GUS II with enclosure is rumored to cost US$150 so a basic PE4L 2.1 for $105 is not at all competitive for older technology. It's not far off the cost of an enclosure-based pci-e 2.0 compliant ViDock 2/3/4/4+.

(3) The original request for a pci-e 2.0 compliant product was exactly 1 year ago. The ETA has been excruciating slow with no real reason for users who held out to purchase an overpriced product if it's going to be $105. It's a terrible surprise after a long wait.

(4) Technical users who think the pricing is excessive may consider making a true DIY version like this or a $55-delivered PE3A-based unit.

If this is to be the price then I'll be requesting the DIY eGPU community organize an alternative manufacturer to deliver a pci-e 2.0 compliant PE4L/PE4H clone at the US$71 pricepoint.

Your comments are much desired on this important aspect.

1-16-2012: Update from Masaharu at BPlus

> Has the product been tested sufficiently such that you can now sell it?

We well tested PE4L v2.1 on PC as attached BPLUS-PC.txt and AMD 4670 VGA. In our lab, the results were pretty good, and achieved PCIe 2.0 speed. At "Resident Evil 5" benchmark test, v2.1 had twice performance of v1.5's.

PE4L v1.5: 23FPS

PE4L v2.1: 46FPS

So, we are confident that v2.1 achieved PCIe 2.0.

However, it's still engineering sample, because we still need to test longer cable (yours is 25cm, another 35cm one was also ok, we would like to test 55cm) and tested only one PC system.

Also, we have some small modification (which not related to PCIe portion). So, next version will come out after Chinese new year holiday season. We hope we can release it by the end of Feb.

> If so, perhaps you can give another update I can add to the Thunderbolt, USB 3.0,

> PCIe 2.0 eGPU update so that Series-6 chipset users can now buy it?

Regarding Thunderbolt, we requested intel to provide permission to use the chip, and they will reply the results in beginning of Feb. Regarding PCIe 2.0, PE4H v3.0 will arrive in this week. We consider to sell the module (= without enclosure) first.

> Is there a product link to the new PE4L 2.1?

> What will be it's price? Have you soldered the PE4L+cable+EC2C all together now?

We are very sorry but product site is still under construction. We consider the price is approx x1.5 of PE4L v1.5, but not fixed yet. PE4L v2.1 is PE4L+cable+EC2C all together as you recommended;

clip_image008.png

I forgot to say one thing. Joints portion of the both ends of cable is not strong enough now. Please take care of them carefully.

1-10-2012: Update from Masaharu at BPlus

> It's now January 9th so wondering if you'll be releasing the PE4L v2.0

> in the next couple of days?

Good timing.

Today, we got PCBs in Shanghai and started SMT.

We hope they will arrive at Taipei in weekend and be tested.

If it is go, we will ship a piece to you.

> If not, can you give a revised ETA for products to be added to the

> Thunderolt, USB 3.0, PCIe 2.0 eGPU update.

Regarding Thunderbolt, it is not so easy to develop

so we ask Intel to provide chip and spec.

We hope we can receive them soon and resume developing.

12-22-2011: Update from Masaharu at BPlus

> Yes, I would advise getting the PCB out as soon as conveniently possible

> as that is all the most people will require. The enclosure is a nice option

> for those that want a tidy solution similar to what Villagetronic/Magma offers.

Only PE4H v3.0 will have enclosure. PE4L v2.0 will have same style as before.

> Is the PE4L 2.0 scheduled to be released at the same time as the PE4H 3.0?

> Some might prefer it's smaller footprint and price.

We would have PE4L 2.0 earlier, hopefully Jan/2011. The footprint will be almost same as before.

Regarding PE4H 3.0, it will be larger for enclosure, and quite different from before. It will be released after enclosure is finished so probably Feb or March/2012.

> Consider that CES 2011 will likely see mainstream Thunderbolt products being

> released. Those will compete with expresscard products but deliver more bandwidth.

> Ie: x1 2.0 Optimus is approx x2 2.0 in bandwidth due to pci-e compression, or half

> of Thunderbolt when used on an external LCD.

We already ordered a Thunderbolt product, however, the arrival will be delayed due to large back order.

We will also develop Thunderbolt products after PE4x to shift to Thunderbolt.

12-19-2011: Update from Masaharu at BPlus

> Can you give an further ETA on when the PE4H 3.0 will be available for purchase?

> Maybe those unwilling to wait will buy and modify a PE3A instead if it's going

> to still be a few months away.

Regarding PE3A, we decided not to have updated product any more.

Regarding PE4H v3.0, it is easy just for PCB, and now we are under discussion for enclosure so that it takes some more time to fix the spec. We hope we can fix the spec within this year and release it on January at least for PCB portion.

12-07-2011: Update from Masaharu at BPlus

> Thunderbolt cable Products:

> PE5H v1.0: ThunderboltX2-PCIe2.0x4, for PCIe2.0x16 video card w/ enclosure, Nov/E

> HP4C v1.0: PCIe2.0x4-ThunderboltX2, Install to Desktop PCIe slot, Nov/E

We got the samples in the beginning of this week. However, they do not work well at present, so they are now pending for a while including other TB products such as EC3C.

> mHDMI cable Products (Revise):

> EC2C v2.0: ExpressCard(PCIe2.0x1)-mHDMI with ReDriver, Install to ExC slot, Dec/B

> PE4L v2.0: mHDMI-PCIe2.0x1, for PCIe2.0x16 video card, Dec/M

> PE4H v3.0: mHDMIx4-PCIe2.0x4, for PCIe2.0x16 video card, Dec/M

Regarding mHDMI products, PCIe2.0 was achieved with combination of mHDMI-cable-direct-connected-ADP and PE4L v1.5 (without ReDriver). ADP: ADP (Mini PCI-E / PCI-E adapter ver1.0). So, we decided to develop PCIe2.0 products without ReDriver.

Now, we plan to develop PE4H ver3.0 which has direct connected to EC2C with mHDMI cable for both sides so that there is no mHDMI connector between PE4H and EC2C.

By the way, we also developed mHDMI-cable-direct-connected-EC2C (sample).

11-18-2011: Update from Masaharu at BPlus

Now, we focus PE5H (TB-PCIEx4) and HP4C (PCIEx4-TB).

First one will be quite dirty with jumpers to reserved pads to confirm Thunderbolt specification (Actually Thunderbolt cable pin-out).

EC3C will be developed after the pin-out become clear, or apply same way as HP4C.

PE4Lv2.0+EC2Cv2.0 can be developed parallel with PE5H.

PE4Hv3.0(ReDriver) will be after PE4Lv2.0 issue become clear.

When Gerry come back from Shanghai next Monday, I will discuss detail with him.

11-11-2011: Masaharu about products inc USB 3.0 no go

I am Masaharu Adachi R&D engineer, just started to work in Bplus. Now, we have these products to develop;

Thunderbolt cable Products:

PE5H v1.0: Thunderbolt-PCIe2.0x4, for PCIe2.0x16 video card w/ enclosure, Nov/E

HP1C v1.0: PCIe2.0x4-Thunderbolt, Install to Desktop PCIe slot, Nov/E

HP4C v1.0: PCIeX4 edge-ThunderboltX2 adapter

EC3C v1.0: ExpressCard(PCIe2.0x1)-Thunderbolt, Install to ExC slot, Dec/B

mHDMI cable Products (Revise):

EC2C v2.0: ExpressCard(PCIe2.0x1)-mHDMI with ReDriver, Install to ExC slot, Dec/B

PE4L v2.0: mHDMI-PCIe2.0x1, for PCIe2.0x16 video card, Dec/M

PE4H v3.0: mHDMIx4-PCIe2.0x4, for PCIe2.0x16 video card, Dec/M

At present, we concentrate to overcome PCIe2.0, because ReDriver IC or Thunderbolt cables are compliant with PCIe2.0. After that, if pin-compatible ReDriver IC for PCIe3.0 or PCIe3.0 Thunderbolt cable is developed, the products above can be used for PCIe3.0. If not, we will develop PCIe3.0 in the future somehow, of course.

Regarding Thunderbolt and enclosure product, PE4H v3.0 is it.

Our target ETA is Nov/E, and we hope evaluation can be finished until Dec/B. As development is not finished nor components are not fixed yet, the price is not unknown yet either. However, we would like to apply reasonable price compared with current PE4H products.

Regarding USB3.0 enclosure

We have already developed it and the product number is "USB3380". However, it is not suitable for PC consumers but for embedded system, because installed video card can not be used directly and lots of device driver modification is required, due to bus architecture difference between PCIe and USB3.0.

The USB3380 is available at PLX or PLX's distributors, however we still recommended except for embedded systems.

10-12-2011: Aaron Yang (source)

The PE4H(support PCIe 2.0) with housing is developing now and will be available in middle of November. In addition, the EC2C(support PCIe 2.0) is also developing and will be available in begin of November. I think you can buy PE4H-EC2C(support PCIe 2.0) in middle of November if the production is smooth.

9-21-2011: GerryChen

Subject: RE: Request: modified PE3A for pci-e 2.0 testing

New version of PE4L-EC2C add ReDriver IC design now. An engineering sample will be available next month.

9-3-2011: GerryChen summary: using a active transmitter/receiver pair for reliable pci-e 2.0 transmission over copper.

Unfortunately, USB3.0 cable test fail for PCIe interface. The next stage. I will try ReDriver IC. Please check attached file for datasheet.

PI3EQX5701.pdf

8-4-2011: new flat cables are now available. BPlus allude to the mini HDMI cable as being the culprit preventing pci-e 2.0 signalling.

Thu, Aug 4, 2011 at 5:50 PM

subject RE: pci-e 2.0 compliance of hwtools' PE4L/PE4H kit

[ highlights that the email node has changed ]

Please check attached file for new cable spec. The new cable available in my warehouse. It will launch web-shop this week.

New cable quality is better than old cable. But new cable still can’t work for PCIe2.0 speed. Let me know your test result. Thank you.

PCIEMM comparison_2011Aug04.pdf

PCIEMM_A.pdf

Who has pci-e 2.0 compliant cabling?

<br>

· BPlus: PE4L 2.1 (avail 20th Feb) and PE4H 3.0 are pci-e 2.0 compliant.

· BPlus: a PE3A+riser is pci-e compliant, as is a modded PE3A. The latter is a cableless solution with the eGPU butted right up against the expresscard slot.

<br>

· bplus are now developing an active transmitter/receiver pair for reliable pci-e 2.0 transmission over copper. Their previous attempt at a new flat cable only allowed pci-e 1.0 over a longer length as shown. It uses a mini HDMI connector. <strike>Interested parties could try a $15 NXG active HDMI cable via two HDMI<->mHDMI adapters, noting that pin18 provides power with HDMI transmitting 5V and the EC2C/PM3N transmitting 3.3V which might be an issue.</strike> <-- doesn't work. <strike>Khenglish is in the process of obtaining a short mini HDMI to mini HDMI adapter to try instead of the cable.</strike> <- doesn't work.

<br>

· Netstor's TurboBox-mini has pci-e 2.0 compliant cabling.

<br>

· Apple's $50 Thunderbolt cable is the only one certified to do pci-e 2.0 signalling on copper due to being active in nature. It uses a mini Displayport connector

<a name=TB></A><span style="font-size:large">Thunderbolt Products</span> Updated 4-22-2012

MSI GUS II: availability/pricing is unknown but rumored to be $199 with 150W PSU.

US$400 MLogic MLink Thunderbolt pci-e expansion: Available in July 2012 . Red Rocket version starting at $800. NAB 2012 - M Logic on Vimeo

OWC Thunderbolt PCIe x4 Expander Box ~ single slot chassis at 4x. OWC will be Nabbing Some Attention at NAB. | Other World Computing Blog

Sonnet Thunderbolt Echo Express: Available in June 2012. $800 for 150W chassis, $600 for 100W chassis

US$979 Magma Thunderbolt product - ExpressBox 3T: Pre-order on the product pre-order page.

Netstor's Thunderbolt product - TurboBox : Available soon according to the discussion page.

Villagetronic Thunderbolt product: Is being developed. External Thunderbolt PCI Expansion Chassis and Hub in Development - Mac Rumors. Do NOTE, their CEO indicates on their facebook page that Intel would not provide Villagetronic a development kit.

Which systems have or will have Thunderbolt ports?

* Apple MBA/MBP 2011

* Sony Z2 -> it's a proprietory Lightpeak port

* Lenovo S430 : the first PC notebook announced to be released with a Thunderbolt port

* Asus and Acer will have TB port systems in Q1-Q2 2012

* Intel Working on Standardized Thunderbolt Docking System for PCs - Mac Rumors

<A name=usb3></A><span style="font-size:large">USB 3.0 Products</span> - PLXtech's USB3.0 to pci-e adapter has been demonstrated working with a desktop video card attached to a USB 3.0 port here.

11-11-2011: Masaharu- USB 3.0 is no go

Regarding USB3.0 enclosure

We have already developed it and the product number is "USB3380". However, it is not suitable for PC consumers but for embedded system, because installed video card can not be used directly and lots of device driver modification is required, due to bus architecture difference between PCIe and USB3.0.

The USB3380 is available at PLX or PLX's distributors, however we still recommended except for embedded systems.

10-14-2011: GerryChen

Subject: USB 3.0 PE4L/PE4H development?

We still co-work with PLXtech to develop video card to USB3.0 solution this year. Modify video card driver is needed. Eva-Kit already released Oct/2011.

plxdev1.jpgplxdev2.jpg

  • Thumbs Up 4
Link to comment
Share on other sites

Thank you for letting users know. The patch for Setup 1.10b5 is an interim fix and is now included in the new page http://forum.techinferno.com/diy-e-gpu-projects/2123-diy-egpu-setup-1-x.html#post27337 . Those fixes are rolled into the development version of Setup 1.11a with a stack of others. It's release is pending more fixes/updates to compact whose author will be working with an iGPU+dGPU+eGPU rig soon so will be seeing long outstanding bugs and features requested firsthand. Trying also to organize a way to get a 2011+ Macbook too so could tweak the software to work on it too. Looks like Thunderbolt systems still encounter error 12.

Can you remind me of the system in question here? Are you using expresscard/mPCIe, PE4H/PE4L version? The method I've found was always successful in detecting the eGPU being:

1. Disconnect and power off the eGPU from the notebook.

2. Boot Setup 1.1x

3. Power on the eGPU.

4. Hotplug the EC2C (expresscard) or PM3N (mPCIe). A variation of this if using PE4H 2.4 or PE4L 1.5 is to leave the EC2C in the expresscard slot throughout steps 1-3 above but just attach the mHDMI cable to it with the powered eGPU to get it detected.

Some systems do not boot with the eGPU powered and attached. That's a great troubleshooting step to confirm the hardware itself works as the bios sees it and doesn't like it. I believe the Troubleshooting FAQ will be up in the next day or two. It has pretty much the same steps as I list above for getting an eGPU detected.

This is actually another system. My dgpu in my last system totally crashed and i was no longer able to see any video output...

this system is a fujitsu lifebook n3430 that i got from a friend.

Win XP Media Center Edition SP3

Intel T2300 @ 1.66 GHz

512 RAM (i know it sucks - is it even enough RAM)

integrated intel video

egpu is GTX 460SE, connected through PE4L-EC2C with Corsair CX430 psu

Actually, i have tried all of the methods you have listed (in setup 1.x help menus) including the one above. nothing has worked. before it actually worked when the egpu was plugged in before i turned on the laptop, but i'm not sure what happened. do i need to get more memory? the laptop can support up to 2gb.

  • Thumbs Up 1
Link to comment
Share on other sites

egpu is GTX 460SE, connected through PE4L-EC2C with Corsair CX430 psu

Actually, i have tried all of the methods you have listed (in setup 1.x help menus) including the one above. nothing has worked. before it actually worked when the egpu was plugged in before i turned on the laptop, but i'm not sure what happened. do i need to get more memory? the laptop can support up to 2gb.

Memory won't be an issue until after the eGPU is detected if you have error 12 appearing. If none of the methods listed got the eGPU detected then it would likely be a hardware problem. Either the PE4L, video card, mHDMI cable or power/PSU or expresscard. If you have another system with an expresscard slot handy I'd connect it up and see if it's detected. Another highly likely issue is the mHDMI cable if you have a PE4L 1.5 or older. I had a few mHDMI get faulty after repeated uses/bends. Swapping at least the ends might get some results but better would be swapping with a complete new mHDMI cable.

Link to comment
Share on other sites

Memory won't be an issue until after the eGPU is detected if you have error 12 appearing. If none of the methods listed got the eGPU detected then it would likely be a hardware problem. Either the PE4L, video card, mHDMI cable or power/PSU or expresscard. If you have another system with an expresscard slot handy I'd connect it up and see if it's detected. Another highly likely issue is the mHDMI cable if you have a PE4L 1.5 or older. I had a few mHDMI get faulty after repeated uses/bends. Swapping at least the ends might get some results but better would be swapping with a complete new mHDMI cable.

i have a PE4L 2.4a (i think). swapped the ends of the mHDMI. it worked. i can't believe it was so simple. should have tried that, siince that was the issue for my last rig! sorry and thanks again!!!

Link to comment
Share on other sites

PCI-Express to Thunderbolt Adapter is released :Banane20:

TH05 ( PCI-Express to Thunderbolt? Adapter) .

Has anyone already tried it?

2.0 x 2 is 1.1 x 4. If Nvidia Optimus technology work on Thunderbolt Adapter,it's the best egpu solution !!!

We found that Nvidia Optimus had two benefits on a expresscard solution. Firstly, it engaged pci-e compression when a x1 link was detected accelerating output 20-300% and secondly, it could display accelerated output to the internal LCD. This was however dependent on the hosting machine booting up with a Intel 4500MHD or HD iGPU as the active video card.

With Thunderbolt's x2/x4 link, the pci-e compression will no longer be available. So if you don't need the internal LCD mode then an ATiAMD card may in some cases give better performance. On a x1 link (with no pci-e compression) I found that DX9 was faster on an ATi card and DX10 was faster on a NVidia card.

BPlus' TH05 is far more affordable but at x2 2.0, it's going to be pretty much on par to a x1.2Opt link in performance when using an expresscard slot + NVidia Fermi/Keplar card on an iGPU-equipped notebook. The PE4L 2.1b expresscard solution costs US$71 with a lot of notebook options, while a TH05 costs US$180 with few TB notebook options, except of course if you are a Macbook owner.

Saying all of this, I'd love to get a Lucid Thunderbolt-to-pcie enclosure (x4 2.0) and benchmark it with NVidia and ATiAMD cards on a x1.2Opt, x2 2.0 and x4 2.0 Thunderbolt link on say a 13" Macbook Pro. That would be done by masking lanes on the video card using cellophane tape.

Link to comment
Share on other sites

Great to see the eGPU scene going on here, unfortunately the users are now split in different forums. I am following the eGPU scene for a while.

I am considering to drop my desktop and switch to a laptop+egpu, but dont know what to get atm. Expresscard is a nice solution with optimus atm i think, with also internal screen possibility.

But thunderbolt is available too now, did someone already test the Bplus th05? is it worth it with x2 2.0 (instead of max x4 2.0)?

Is the Lucid software for internal screen mode + compression a good alternative for optimus? Lucidlogix with the multi gpu solution

I think i have 3 options right now:

1. Buy cheap laptop with expresscard now (eventually buy new laptop when tb is more common)

2. Buy second hand macbook air 13" with tb-port (still relatively expensive, no 1080p screen, no quad core, no upgradeble ram)

3. Wait for hasswell (2013?, problably expensive)

What do you think guys?

Link to comment
Share on other sites

Great to see the eGPU scene going on here, unfortunately the users are now split in different forums. I am following the eGPU scene for a while.

I am considering to drop my desktop and switch to a laptop+egpu, but dont know what to get atm. Expresscard is a nice solution with optimus atm i think, with also internal screen possibility.

But thunderbolt is available too now, did someone already test the Bplus th05? is it worth it with x2 2.0 (instead of max x4 2.0)?

Is the Lucid software for internal screen mode + compression a good alternative for optimus? Lucidlogix with the multi gpu solution

I think i have 3 options right now:

1. Buy cheap laptop with expresscard now (eventually buy new laptop when tb is more common)

2. Buy second hand macbook air 13" with tb-port (still relatively expensive, no 1080p screen, no quad core, no upgradeble ram)

3. Wait for hasswell (2013?, problably expensive)

What do you think guys?

As much as it wasn't desirable to do so, moving content here was the only way I could communicate with users and update content. You can read the measures I went to in order to prevent fracturing of users in the NBR ban details in my sig. Anyway.. back to your query.

BPlus' TH05 Thunderbolt adapter with it's x2 2.0 performance is not worth it for the majority of users that run Windows, are OK with an iGPU/Optimus-equipped notebook and prefer a NVidia card. It's only benefit is then:

1. Macbook users who have no expresscard option. Note 13" Macbook Pro retina is rumored for release soon.

2. Users who want to run an ATiAMD card (no Optimus)

3. Users who have a notebook with dedicated graphics (no iGPU - no Optimus)

4. Users who predominantly run MacOS/Linux (no Optimus)

A expresscard-based x1.2Op NVidia using a $70 PE4L 2.1b is a more cost effective solution than a $180 TH05 Thunderbolt one. Performance will be pretty much the same since x1.2Opt's pci-e compression means x1 2.0 is approximates x2 2.0. That pci-e compression only engages upon detecting a x1 link. There are also far more expresscard equipped notebooks than Thunderbolt ones available at a lower cost too.

The cheapest x4 2.0 capable Thunderbolt enclosure is the $400 Sonnet's Echo Express SE Thunderbolt. There you'll need to add a $50 Thunderbolt cable and hack it to work with an external PSU. I believe it's designed for a x8 card, so need to cut the slot too to accept a x16 card.

So from your options I'd go 1. Buy cheap laptop with expresscard now (eventually buy new laptop when tb is more common and monitor the development of TB eGPUs, upgrading when it's performance-to-cost ratio improves. BPlus have said they'll be releasing a x4 2.0 Thunderbolt adapter next year.

REF: Internal Screen for ATI GPUs with Virtu Driver (timohour) shows the Virtua software giving good performance when rendering to the internal LCD. This is how one would get the equivalent of NVidia Optimus' internal LCD mode using an ATiAMD configuration.

  • Thumbs Up 3
Link to comment
Share on other sites

Hi,

does anyone know how to do a DSDT override on Windows 8? It's not working for me. It did work on Windows 7.

I extracted the .dsl file with iasl-20120913, edited it and then compiled and loaded it with ASL 4.0.0NT. The output was "Table overloading succeeded."

The registry entry does appear in System\CurrentControlSet\Services\ACPI\Parameters\DSDT\FUJ___\FJNBB13_\01060000.

But it is being ignored. No Large Memory entry appears in the device manager. I know that it's being ignored because normally when you install a DSDT override and then change the memory layout, for example by removing a memory stick, Windows will bluescreen on reboot. It does not.

Maybe the registry location for DSDT overrides changed?

Some observations I had on Windows 8:

Windows 8 reacts differently to the TOLUD issue. When a device gets connected for which Windows 8 can't allocate enough memory, it does not show Error Code 12. Instead there is an error message saying - I'm translating now, so it might not be correct - 'No drivers are installed for this device' in the Device Status field in Device Manager, even though there are drivers installed (I previously verified that the eGPU was working by removing memory to circumvent the TOLUD issue and the device manager even shows the correct driver version). Screenshot.

In addition to that, ntoskrnl.exe (or one of its threads) starts consuming 100% of one CPU core. As a consequence of that, all actions that will depend on it will not work anymore. For example searching for new devices in the Device Manger, disabling a device, connecting a different device, trying to search for drivers on Windows Update, etc. all result in the respective window freezing. You can't even shut down cleanly anymore, the shutdown will also hang indefinitely. This condition is permanent and can not be stopped by removing the offending device, only by doing a hard shutdown.

Windows 8 will also does this if you boot with the offending device connected. It will hang during the boot animation. It seems to be the same issue as I can hear the fan revving up, so I know it's hanging with high CPU load. The only thing left to do is to abort the boot with a hard shutdown.

As a bonus you'll have to deal with the "Automatic Repair" on the next boot because Windows detects it wasn't able to boot successfully.

Besides that, when removing memory to circumvent the TOLUD issue, my eGPU is fully operational on Windows 8. Optimus compression and rendering on the internal display both work with an unmodded desktop driver, version 306.23.

  • Thumbs Up 1
Link to comment
Share on other sites

Unfortunately your system is x1 1.0 capable only, so there is nothing more that can be done to extract better performance. Consider:

* PM55 is pci-e 2.0 compliant but only for power management. The link speed is still pci-e 1.0. You need a Series-6 chipsets or newer for pci-e 2.0.

* Last time I checked, the port listing of a DV7 wasn't capable to support x2. There would need [port1+2], [port3+4], [port5+6] or [port7+8] accessible as expresscard+mPCIe or mPCIe+mPCIe). To answer your question, x3 isn't possible.

* Putting an Arrandale CPU like a i5-580M into your system won't enable the iGPU to be the primary bootup video device. Sure, you can enable it using Setup 1.1x but the NVidia driver will only engage pci-e compression if the iGPU is the primary bootup video device. So no x1.Opt would be possible.

My suggestion is to offload the system for an inexpensive Sandy Bridge system like a 17" Dell Vostro 3750. Then you'll be able to enjoy x1.2Opt performance, approx 4 times more bandwidth than the best your DV7 can do (x1 1.0).

Hi Nando.

How about starting a new thread dedicated for eGPU?

I'm still very motivated to discuss this subject and try to help others.

Link to comment
Share on other sites

Hi i recently bought the PE4H-EC2C and im trying to hook the GTX 670 to my toshiba tecra r840

intel core i7-2620

8GB Ram

im using a cooler master 450w PSU

now heres the problem

after setting up everything my laptop was able to read the graphic card and isuccessfully installed the 670 driver from the disc (i am using a palit GTX 670 jetstream 2GB)

BUT despite all that my tecra r840 is unable detect the display on my external monitor which is meant for the gtx 670, i am using a DVI connector

can any expert please advice me, its my first time trying this :))

Link to comment
Share on other sites

does anyone know how to do a DSDT override on Windows 8? It's not working for me. It did work on Windows 7.

You are pioneering this work. I have as yet not loaded Win8 on my box to try it but if I can get a loaner Dell E4300, the test box I did the original DSDT override on, then I'll see if I can get a Large Memory Area defined in Win8. The different behavior of Win8 indicates the pci-e driver has been altered.

Hi i recently bought the PE4H-EC2C and im trying to hook the GTX 670 to my toshiba tecra r840

intel core i7-2620

8GB Ram

im using a cooler master 450w PSU

now heres the problem

after setting up everything my laptop was able to read the graphic card and isuccessfully installed the 670 driver from the disc (i am using a palit GTX 670 jetstream 2GB)

BUT despite all that my tecra r840 is unable detect the display on my external monitor which is meant for the gtx 670, i am using a DVI connector

can any expert please advice me, its my first time trying this :))

We've had a previous Toshiba R840 NVidia Optimus implementation here also with a PE4H. Having a PE4H 2.4 rather than a PE4L 2.1b means you'll require Setup 1.1x to downgrade the link to pci-e 1.x or else it will not work. That's because a PE4H 2.4 cannot transmit at the pci-e 2.x speed, the default your expresscard slot is set to.

My suggestion is rather than obtaining the Setup 1.1x software, offload the PE4H for a PE4L 2.1b so you get double the bandwidth (pcie 2.0 vs pci-e 1.x) and don't need donate for Setup 1.1x.

Hi guys!

Can anyone comment on this idea of a setup for my eGPU project:

GPU: GTX 660/GTX 660-ti

Power: 350w power brick

Interface: PE4L-PM060A

Do you think that the "power brick" be compatible as supplier for a GTX 660-ti setup?

A MSI GTX660Ti requires 164W of peak power as shown. An incredibly efficient card. That means a 12V/14A rail would be OK, though I'd go for 16A for a safety margin. The PSU you link is 12V/29A, so has plenty however, it doesn't have a 5V output. A PE4L 2.1b can then only use 3.3V/1.3A (4.3W) from the expresscard slot to drive your cards which might or might not be OK. Reason is a pci-e card is rated at up to 3.3V/3A (10W). You may need to run a single or double-ended USB cable from the notebook to the PE4L to supply an extra 5V/0.5A (2.5W - single or 5W - double), downregulated to 3.3V by the PE4L just to give slightly more power.

If this sounds overly complicated then may I suggest acquiring a 203W XBOX360 PSU. It's got a 12V/16.5A rail + 5V/1A rail. There's an example of how to wire it here and here.

Link to comment
Share on other sites

You are pioneering this work. I have as yet not loaded Win8 on my box to try it but if I can get a loaner Dell E4300, the test box I did the original DSDT override on, then I'll see if I can get a Large Memory Area defined in Win8. The different behavior of Win8 indicates the pci-e driver has been altered.

We've had a previous Toshiba R840 NVidia Optimus implementation here also with a PE4H. Having a PE4H 2.4 rather than a PE4L 2.1b means you'll require Setup 1.1x to downgrade the link to pci-e 1.x or else it will not work. That's because a PE4H 2.4 cannot transmit at the pci-e 2.x speed, the default your expresscard slot is set to.

My suggestion is rather than obtaining the Setup 1.1x software, offload the PE4H for a PE4L 2.1b so you get double the bandwidth (pcie 2.0 vs pci-e 1.x) and don't need donate for Setup 1.1x.

A MSI GTX660Ti requires 164W of peak power as shown. An incredibly efficient card. That means a 12V/14A rail would be OK, though I'd go for 16A for a safety margin. The PSU you link is 12V/29A, so has plenty however, it doesn't have a 5V output. A PE4L 2.1b can then only use 3.3V/1.3A (4.3W) from the expresscard slot to drive your cards which might or might not be OK. Reason is a pci-e card is rated at up to 3.3V/3A (10W). You may need to run a single or double-ended USB cable from the notebook to the PE4L to supply an extra 5V/0.5A (2.5W - single or 5W - double), downregulated to 3.3V by the PE4L just to give slightly more power.

If this sounds overly complicated then may I suggest acquiring a 203W XBOX360 PSU. It's got a 12V/16.5A rail + 5V/1A rail. There's an example of how to wire it here and here.

Ok I see, the reason I looked at this power brick was due to its dimensions. I want something that is shaped as a graphics card in order to make something slick. So maybe I can find something similar that have a 5v rail as well.

I'm amazed of the low consumption by the 660/660ti!

If I find something suitable I post it here for opinions, thanks a lot for keeping this AWESOME thread alive =)

What about this power brick?

  • Thumbs Up 1
Link to comment
Share on other sites

×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.