Tech Inferno Fan

According to this guide on NBR, your iGpu should work by removing a file in windows.

So maybe you can get optimus working

User who wrote that had a MBA with an iGPU but no dGPU. We've discovered above that the iGPU isn't even present on the pci-e bus after boot on a dGPU equipped 15" MBP - the system boots with the dGPU as the primary active video device. In which case there's no chance of getting the x1 pci-e compression and transparent internal LCD mode that NVidia Optimus offers only when it detects the iGPU as the primary bootup device (4500MHD, HD, HD3000, HD4000). This requires Apple to modify how their 'bios' or whatever they use to boot the system with.

Wish it wasn't so. 15" MBP/MBPr is otherwise a impressive package: TB, dGPU-HD6750M/GT650M, i7-quad, thin-and-light. Just need a way to make the iGPU the primary bootup device in Windows.

Would you be so kind to respond to my post, please?

Your Radeon Expresss 1100 is considered a Legacy device by AMD: ATI Catalyst . Your only chance then of getting it to work with your HD6790 would be by doing INF mods. However, even TWL's experimental driver set works on HD2000-6000. I'd suggest start reading at AMD Catalyst (Modified Build) - Guru3D.com Forums and work from there. INF modification is a whole topic in it's own right better discussed and investigated elsewhere like on that linked guru3d post.

I did ask earlier (you probably missed it) if it was possible to use an ethernet port via an hdmi ethernet extender adapter along with hdmi to mhdmi cable, assuming a 1GB or higher ethernet port. Then (and only then) I'd be able to set up a x2 1.0 link.

EDIT: How about 4 photos?

Thank you for the photos.

Now to answer your query. The eGPU needs access to the pci-e bus. Your ethernet port would presumably provide the second port for a x2 link. Now, the ethernet port is provided via an ethernet chip that acts as an interposer between pci-e and ethernet. So you'd need to hijaak the pci-e lines coming into the ethernet chip to be able to do a x2 link. You'd be working with miniscule tracks so we can say it's not feasible.

The faster and painfree approach to getting more performance on a budget would be to offload your system for a budget Sandy Bridge box with an expresscard slot that's x1.2Opt capable such as a Lenovo E420/E520, Dell Vostro 3350/3450/3550/3750 or HP Probook 4330s/4430s/4530s/4730s/6360b/6460b/6560b. Any of those would get you a lot more performance with GTX660 if you used a pci-e 2.0 capable PE4L-EC060A 2.1b. A 17" Vostro 3750 with a GT525M dGPU close in spec to what you have now.

Thanks. I'll try it as i saw some pcie options for mm.

In that case, there is an easier way to solve your error 43. Boot Setup 1.1x and run the Video cards->Initialize on your GTX660Ti. It will give you a PT commandline where you'll see the memory address of the 0x114 pci-e register of your GTX660Ti. Just translate that to your mm command, apply it, then boot Win7 and you should be golden.

Great news,

Did you get your hands on a retina 13"??

I recently got windows 8 so I might try this out when I get time.

Thanks Shelltoe for the info, really great work.

Have my eyes on a ~AU$900 2011 13" Macbook Pro with Thunderbolt port just to have a devel box to play with. A 13" MBPr is like > AU$1800 or a third-gen 13" MBP is > $1200. I can't justify that sort of $$ when it's highly likely my 12.5" Dell E6230 with i5-3320M that costs 1/3 or 1/2 of that is more satisfying to use (internal HSPA+ 3G, traditional keyboard, matte LCD, 1.6kgs, 66Whr 6-cell).

Only other benefit of the latest MBPr is it does have two Thunderbolt ports so might be possible to SLI/XF two eGPUs to get more performance. The earlier MPBs only have 1 Thunderbolt port. We discovered a while back that need at least a x4 link to do SLI so would need a pricier Sonnet/OWC Thunderbolt enclosure to try it.

All right, my eGPU is up, running, and very smooth.

...

Future Changes:

Ditch the hdmi to expresscard adapter to go for the ultramon internal lcd, since the adapter seems to need ultramon anyways.

Install better processor

P.S. I'd love to hear others' opinions of my benchmark results.

Your benchmark results is what can be expected from a x1 1.0 link. Unfortunately your system lacks a 4500MHD iGPU so the NVidia x1 Optimus pci-e compression doesn't engage. Can see my i5-3320M +GTX660 results here where x1 1.0 and x1.Opt are compared. You can see 3dmark06 sees over 3 times improved performance going from x1 1.0 to x1.Opt. If there was a 4500MHD version of your systemboard then maybe you could scout around in the forums to see if somebody would be willing to do a swap for your 9600GT one?

Question I have for you: how did you manage to get the HDMI expresscard and eGPU to co-exist at the same time? Are you running a mPCIe-based eGPU? AFAIK there is only one expresscard slot on a HDX16T.

There was certainly interest in using that expresscard HDMI adapter to gain internal LCD functionality without impacting eGPU pci-e bandwidth (ultramon, NVidia Optimus method) so a photo or three would help understand how you've got your system wired up.

I have got it worked, finally!! But at cost.

I have a gtx 460 se that I am using it on my thinkpad edge, so I switch places with amd hd6790.

So my new question is: Will amd radeon hd 6790 work on my thinkpad edge 14?

The HD6790 will work with the Edge14 far easier than getting it to work with your Inspiron 1501. The trouble your Inspiron 1501 is having with the HD6790 stems from getting the Radeon Express 1100 iGPU and HD6790 drivers to co-exist. The error 31 you see is requires a unified driver for both as a fix, hence why you've been directed to the DNA drivers or start hacking the INF files of your ATiAMD drivers to add your HD6790.

returns in 'Warning: No device selected for "4.b=0"'

That means the HD4000 iGPU isn't on the pci-e bus. Explains why Optimus isn't possible in Win7.

compaction hangs after a success while trying to reenter the menu

calling pci.bat at startup results in hang too

Please remove the \config\pci.bat or the pci.bat line in startup.bat. Doesn't appear to like the result of pci compaction. The Macbooks have a low TOLUD so Win7 may be able to sort things out.

results in systemfailure and reboot (I saw error message followed by some lines of HexCode flashing up)

This is the #1 issue. If can't get chainloading out of Setup 1.1x to work then can't use it to do pci fixups on a Macbook. Good news is I am in the process of organising a 13" Macbook Pro now for delivery in approx 2 weeks. Will help tremendously to work out these issues. That's all except the 15" Macbook Pro iGPU activation which would require Apple's involvement.

The vidinit eradicates error43 by setting register 0x114 to 0x1 on the NVidia video card. Maybe you can use mm to do that? Do note that a lot of utilities won't allow pci-e registers beyond 0xFF to be modified. If mm is affected, but can work on a memory address then it's possible to calculate the mem offset where 0x114 of the NVidia video card is. MCFG + bus + slot + reg (google).

Tested an GTX 660 Ti and received error 43 in Device Manager.

Use Setup 1.1x's Video Cards->Initialize function to eradicate error 43. So your startup.bat would look like:

:: Wait up to 60 secs for user to hotplug eGPU
call vidwait 60 10de:1183

:: Initialize NVidia eGPU to eradicate error43
call vidinit -d 10de:1183

:: IGPU Intel HD 4000
:: I noticed some improvements in boottimings while deactivating the Intel HD
:: mm 00020004 1 ;PCI :0
setpci -s 0:2.0 4.b=0

:: eGPU PCI Bridge 
:: this line does the magic by enabling VGA Output
:: mm 0001013E 1 ;PCI :8
setpci -s 0:1.1 3E.b=8

:: Run result of PCI compaction (if any)
if exist pci.bat call pci.bat

:: Chainload to your OS. Depending on your partitioning, 
:: it would be one of the following:
call grub4dos uefi
call grub4dos mbr

Good work in figuring this out . . .

echo -off
echo "Setting Registars"
# IGPU Intel HD 4000
# I noticed some improvements in boottimings while deactivating the Intel HD
mm 00020004 1 ;PCI :0

# eGPU PCI Bridge 
# this line does the magic by enabling VGA Output
mm 0001013E 1 ;PCI :8

echo "Booting Windows"
fs0:\EFI\Boot\bootx64.efi

If can chainload to Win7 from Setup 1.1x using 'grub4dos mbr' or 'grub4dos uefi' then you could achieve the same result with setpci. Ie:

:: Disable the iGPU
:: mm 00020004 1 ;PCI :0
setpci -s 0:2.0 4.b=0

:: Enable (VGAEN) This bit controls the routing of processor initiated transactions
:: targeting VGA compatible I/O and memory address ranges.
:: mm 0001013E 1 ;PCI :8
setpci -s 0:1.1 3E.b=8

Then chainload to Win7.

The setpci command also allows bitmasking, so you could type 'setpci -s 0:1.1 3E.b=8:8' to just flip the 0x8 bit on (VGA enable).

<span style="font-size:large">Win8: Perform a DSDT substitution within Setup 1.x to overcome TOLUD issues<span>

Moved to http://forum.techinferno.com/diy-e-gpu-projects/7476-%5Bguide%5D-dsdt-override-fix-error-12-a.html#post102526

Hey Nando4 Thanks For Replying Bro , I Will Try What You Said But I need to know something when i do the video card --> initialize i get a message down on 1.x Setup saying that wait for 10dev:1180 (which is my egpu card id ) to be hot plugged and load initialize to it , so i need to know what hot plug exactly means and after that shall i do the PCI compression and what version shall i use (32-bit , 32-bitA - 36 bit ) . because i have Win7 x64 , and on the chainload shall it be on mbr or shall i change it to win7 , please give me clear steps to do , and please clarify the hotplug thing for me and when exactly do i do the hotplug mentioned in the setup 1.x after loading initalize option .

Thanks

A 'call vidwait 60 10dev:1180' will wait for up to 60 seconds for you to hotplug that video card. Hotplugging means attaching your eGPU once your system is already running until it's detected. That may be as simple as powering on the eGPU with the EC2C already plugged in (usually doesn't work). More successful is having the eGPU powered on and then plugging in the mHDMI cable into the EC2C end that is already inserted into the system, or plugging the EC2C into your system. Check that your eGPU is detected by hitting F5 to update the status.

You should try all combinations of 32bitA+32bit compaction. 36bit compaction shouldn't be used as you don't have a DSDT override to extend the root bridge into 36-bit space. Ie:

- 32bitA compaction on the eGPU

- 32bitA compaction on the dGPU+eGPU

- 32bit compaction on the eGPU

- 32bit compaction on the dGPU+eGPU

- 32bitA compaction on ALL devices

- 32bit compaction on ALL devices

Chainload should be set to MBR. Only time I suggest changing it to Win7 is when you have another bootloader such as Linux' grub2 and want to skip it's selection step.

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?

DSDT override testing with Win8

I did a brief test using Win8 without secure boot confirming your results with the DSDT override failing to register at all. I used a standard bios/MBR boot. DIY eGPU Setup 1.1x works, chainloading works. Got rid of the painfully slow graphical bootloader UI using 'bcdedit /set {current} bootmenupolicy Legacy'

The DSDT override is completely ignored in Windows 8. 'asl /loadtable dsdt.aml' changes the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\ACPI\Parameters\DSDT\DELL__\CBX3___\00000021 key on my machine. I manually also modified HKEY_LOCAL_MACHINE\HARDWARE\ACPI\DSDT\DELL__\CBX3___\00000021 key to match the one loaded by 'asl /loadtable dsdt.aml' finding that a Win8 had reset it back to it's original form on a reboot. It appears Win8 enumerates the ACPI on every reboot based on the in-memory/BIOS copy.

Potential workarounds??

- seek some override switch/registry setting somewhere. Google hasn't shown anything as yet

- get acpiexec compiled for FreeDOS. Then load the DSDT override prior to chainloading Win8 in Setup 1.1x

- revert back to Win7 where the DSDT override works well

First My Laptop Iam Using Is Dell XPS 1645 (i7 , 2GB RAM , 500 GB HDD , ATI 5730 Raedon , INTEL CHIPSET PM55 ) And My eGPU Is ( Nvidia GTX 680 Classified 4 GB GDDR5 , PSU 550 Watt)

Setup I GOT : ( 1.x Setup ) , (Nvidia - Verde Driver V.306.97 WHQL , With Modded nvam.inf) , (PE4H v 2.4 With EC2C , Delay Switch Putted On SW1 , Default (0) Seconds)

Screenshot Of My Setup 1.x :

3. Boot eGPU Setup 1.x ( eGPU Was Connected Already!)

- select Video cards->Load to load the PCI space dump created when it was working with no error 43. <-------- ( When Dump Created Using Windows Error 43 Was Still There , and As Instructions It Says It's Okay)

Per Video cards->PCI space.!Save dump use the Video cards->Initialize function instead of the save/load dump method. In addition, use the standard desktop driver (not notebook) and load it without modifying any files.

@Tech Inferno Fan, I assume you missed my recent post. From the posts I've seen, I get the impression that not many other people have used the DSDT override, so I'm relying on your help!

Once you have successfully compiled your .dsl file you load it into the registry with a command like 'iasl /loadtable dsdt.aml'. The DSDT override requires skills in navigating the commandline and dealing with any errors during compile time. If lacking in those areas then suggest ask a friend for help. REF too: DSDT override

Well, looks like I'm stuck unitil either DNA open their website or they upload their drivers to online-source.net.

When I first installed the driver for my eGPU, there was no error. But there was no signal on external monitor. Is there something I can try to make it work while DNA upload the drivers? or should I wait?

You could try modifying the INF file of your standard mobile drivers to include your desktop driver. There you'd need to find where the PCI IDs are stored and add your HD5770 (1002:68b8) to it.

IMHO I dont think Optimus will happen.

Optimus is dependent on the HD4000 being the active bootup video device. So if you can't get the HD4000+GT650M going then it's going to be a tough ask to try to get say a HD4000+GTX660Ti (eGPU) working. What would you lose if can't get Optimus working? You'd have no x1 pci-e compression and the ability to render using the eGPU but display to the internal LCD. The former isn't going to work if using a x2/x4 Thunderbolt link (TH05/Sonnet Echo Express) so not a loss to those ppl. The latter can be provided by LucidLogix Virtu software.

Question for nando/everyone: do you expect that the zenbook ux32vd will have problems with running egpu? Tolud/dgpu/bios/whitelisting/whatever?

Hmm i really cant decide now what to choose, since thunderbolt is already working under x4.2 link, but its very expensive.

UX32VX = HD4000 + GT620M. Now under Win7, the DIY eGPU Setup 1.1x software can disable the GT620M if your bios can't thereby liberating 32-bit PCI space for the eGPU. So no TOLUD/dGPU issue then. Whitelisting? Can't say if Asus does that but Setup 1.1x again has undo-whitelisting components to help. If using Win8 then it complicates things if it's configured for UEFI+Secure boot. The Linux folks are devising a way around that atm. If BIOS/MBR boot is used in Win8 then Setup 1.1x can chainload without any issue.

Agreed. Thunderbolt eGPUs are certainly more costly than mPCIe/expresscard ones. Cheapest I found being a $380/$400 OWC_Helios/Sonnet_Echo_Express from which the board could be extracted and treated like a x4 capable TH05.

Can I please ask a simple question?I have searched this forum and others regarding a working egpu on a Macbook Pro and i'm really really confused.Would a 13" Retina Macbook Pro running through bootcamp (Windows 7) using either a ViDock or Echo Express Pro with whatever interface to work with thunderbolt (TH05?) on a 660 / 680GTX, etc work using the Macbook Retina's screen? That's all i want to know. Cheers.

The 13" rMBP has a HD4000 iGPU only. So then as long as a NVidia Fermi/Keplar eGPU can be detected and made to work on the Thunderbolt bus then the NVidia Optimus drivers can render the image using the eGPU but display it on the internal LCD. Shelltoe's advised a few posts back about getting the eGPU detected:

There seems to be some success installing Windows in EFI mode (without sound as usual). Another blog about TH05 and MAC.

Sandy/Ivy bridge systems without iGPUs and/or using AMD card owners could try LucidLogix Virtu software. It too can render using an eGPU but display to the internal LCD.

Yeah I saw that, the only thing is that if i can't use the Intel driver then i can't use optimus which i really would like to.

So I guess i am stuck then?

Stuck are all MBP owners with iGPUs + dGPUs. See http://forums.macrumors.com/showthread.php?t=1263534 for a progress status. First thing is to get Optimus running with your GT650M dGPU.

The only problem is that if I try to use Tech Inferno Fan's directions in the DSDT override, I have nothing to input the code into. I've tried iasl.exe, but it only flashes onscreen for a second and disappears. Trying to run both the DSDT Editor and iasl.exe as an administrator doesn't help. What am I missing?

Run those commands in a DOS prompt screen (WIN+R, type 'cmd').

Did anyone tried a egpu to a asus zenbook ux32vd? Im considering to buy this laptop, because it gorgeous IPS screen.

Will the ux32vd be able to run a egpu?

I found a picture with bottomplate removed, it seems like theres only one (used) wlan slot(usable mpcie slot?).

[ATTACH=CONFIG]5313[/ATTACH]

Another question will it possible to make a diy connector for the pe4l cable? (I mean just cutting of the cable, and use anything like this?)

If theres a way to make a custom connector for the cable, then its possible to make only a small hole in the bottomplate, to let the cable come out.

Noone has done a DIY eGPU implementation on a Zenbook ultrabook. You may be the first.

See http://forum.techinferno.com/diy-e-gpu-projects/2158-diy-egpu-guide-sony-vaio-vpc-z2-svz13.html for a very nice mPCIe implementation using a custom underside access panel to the mPCIe slot. A PE4L 1.5, or PE4L-L060A has a socketted PM3N option making it easier to just connect/disconnect a mHDMI cable. Problem there is it can only run at pcie 1.x Gen1 speed, as opposed to the pcie 2.0 Gen2 that your system is capable of. In which case you'd want the PE4L-PM060A 2.1 wired as shown in the link above for best performance.

Hey nando,

I have sent you a PM, care to reply it?

Hi nando,

Looks like I'm stuck. I have purchased HD 6790 eGPU, the drivers is detected and installed the desktop driver version, and I have connected it to external screen using all kinds of outputs, from vga to hdmi, but there was no signal available. Is there something wrong I did there? Is there some kind of driver that will make it work?

Thanks.

You said earlier you were using a DVI-to-VGA adapter which are sometimes problematic. Try connecting to a LCD with HDMI or DVI input and work from there.

I've got everything up and running and made some tests. Only on internal screen, sorry.

Specs:

Lenovo x220 (max config) i7-2640M\8GB\HDD320GB

ZOTAC GeForce GTX 670 AMP! Edition 2 GB (without overclocking)

3D Mark 11 (basic version, default preferences) 6034 NVIDIA GeForce GTX 670 video card benchmark result - Intel Core i7-2620M Processor,LENOVO 4286CTO score: P6034 3DMarks

RE 5 (all preferences are on max, AA x4, V-Sync is off, 1366x768) 74,6 fps

[ATTACH=CONFIG]5300[/ATTACH]

..

Keep in mind my notebook specs and that all test were ran on internal screen, are all this results OK?

The internal screen tests of my i5-3320M 2.6 [email protected] DIY eGPU implementation got RE5-var-dx9-1280x800 of 79.7 and 3dmark11 of P5050. Your results indicate x1.2Opt is engaged and your GTX670 a superior performer.

And if somebody could answer mine would be so great! I really would like to get this running!

Setup 1.1x is confirmed to have issues when used on a Macbook platform. Namely, need to do a USB installation and the pci compaction doesn't work right. Those problems have been passed on to the developer of the compact utility.

I just realized this is completely false. After finding AIDA64 (Everest's successor) online to find my PCI ports, I discovered my PCI-E ports are

1: Intel Wi-Fi Link

2: Realtek Ethernet

4: Empty

5: JMicron Firewire

6: Empty

so a 2x config is out of the question... unless I take out the Firewire, which I never use anyway. Is that possible (from a software perspective, I know I'll probably have to modify the outside of my laptop)? Based on PanzerHauptmann's experience, I know my 4 port is the ExpressCard, so if I can repurpose the Firewire, I could run a 2x config after all!

Also, out of curiosity, does anyone know why do I not have a 3 port?

The systemboard designers chose those ports to which to attach devices. They would not have been thinking "let's make port1+port2 available so a user can do a x2 DIY eGPU implementation". Repurposing the Firewire would need to be done at the chip level. Ie: identify the pci-e lines to it and rewire it to be able to connect the mHDMI cable to it. So far nobody has done this as it's quite a technical challenge to complete.

My suggestion is to grab a budget Sandy Bridge notebook like a Lenovo E520, Dell Vostro 3550/3750 or HP Probook 4530s/6560b/4730s and enjoy the faster and more convenient x1.2Opt performance.

If anyone has any ideas, let me know?

Otherwise, I hope this is enough to encourage other X230 owners that it's possible.

Lastly, huge thanks to to everyone involved for posting all your info and results, but most especially Tech Inferno Fan and kizwan.

I recall reading another X230 owner had problems running with 16GB of RAM. So first thing would be to run with 8GB or 4GB instead to see if that helps. If no go, then I'd suggest use Setup 1.1x to do a 32-bitA PCI compaction on your system. One reason for the slowdown may be WIN7's relocation of PCI devices on startup. Using Setup 1.1x to do it beforehand may overcome some problem Win7 is introducing.

What I meant in my question was, Should I download notebook, or the desktop graphics card driver?

First try the desktop driver. If there is any problem running the notebook (iGPU-ATi) and desktop (eGPU-AMD) drivers together then try the DNA-drivers which unify the two together.

Hey guys, finally to see some familiar people here:

I have already asked my questions to Tech Inferno Fan, and now it's your turn to answer it:

I have another laptop, which is Dell inspiron 1501, and I have recently upgraded its amd processor from sempron +3000 1.8GHz to turion 64 x2 tl-68 2.4 GHz, which is significantly good.

After the upgrade, its specs have become:

OS: Windows 7 Ultimate

RAM: 2.5 GB

Processor: AMD turion 64 X2 TL-68 2.4GHz

iGPU: ATI Radeon Xpress 1100, which is the weakest iGPU I have ever seen in my life.

And my eGPU:

PE4H-EC2C 54mm

PSU:Silverstone ST-ST50F-ES 500W with +12V/34V rail output

eGPU: AMD Radeon HD 7750

I have many questions:

1- It's my first time doing it on AMD processor and ATI iGPU, Will it work without disabling iGPU?

2-I have LG Flatron M1717S monitor, which has a VGA output. The screen is pretty good, and I don't want to spend money for another monitor. Unfortunately, the graphics card only supports DVI, HDMI, and Display port. Will DVI-to-VGA adapter work for my eGPU setup?

3- What type of driver shall I install for eGPU in order to make it work?

No reason why it shouldn't work. Possible issues being (1) error 12 - you'd need a strategy to deal with it if encounter it and (2) ensure a Win7 driver exists for your iGPU as the HD7750 Win7 driver will only work if the primary video card's Win7 driver is functional. The DVI wiki tells us that your passive DVI-to-VGA adapter will work:

In addition to digital, the DVI connector includes pins that carry an analog signal, compatible with the RGBHV signals in a VGA connector. This allows a VGA monitor to be connected with a passive plug adapter and offers a universal solution for the DVI interface.

Moved to http://forum.techinferno.com/diy-e-gpu-projects/2747-12-5-dell-e6230-gtx660%40x1-2opt-hd7870%40x1-2-pe4l-ec060a-2-1b.html#post37197 .

how about rEFIt? macs disk utility and rEFIt should allow me to boot egpu without usb disk easily.

fedit works great while setedit hangs (as in all other situations with blinking cursor / no freeze)

scanmode.com results in hang too. i also tried another tool called SCODE.COM but with the same result.

maybe so problem with the shell?

i'll proceed to change the screenshotkey anyways and test some other things.

So fedit gives full edit capabilities without keys getting mangled? If so, I'll change pedit with fedit for the next version. Yes, please do try some other hotkeys in \bin\capture (snarfit screencapture utility) and advise if can get another good one to work. I avoid ones like ALT+S since they are editors to search or save, preferring ones like ALT+ or ALT+DEL.

I am unaware of any other bootloader that allows direct image writes like grub4dos does. Other utils like grub2 use memdisk so read the image into memory and writes occur to memory too. The disk image being a way to avoid needing to repartition the harddisk to boot. A couple of workarounds to the failing disk image writes come to mind, listed in order of complexity below:

1. Just stick with a USB stick install. It allows writes without issue.

2. Create a USB Setup 1.1x stick and do all write activity. That is, generate the requires pci.bat via pci compaction and your required startup.bat file. Then create a disk image install and copy across the \config directory from the USB stick to the disk image. The disk image's benefit over a USB stick being much faster response time and more convenience.

3. [experimental] Move the Setup 1.1x disk image to the first 2GB of space on the boot HDD using one of the better defrag utilities. Seems grub4dos may use CHS access with writes working if the Macbook bios emulation lacks EBIOS write support (per tinybit comment, a grub4dos developer).

4. [experimental] Create a FAT32 DOS partition and move the Setup 1.1x disk image contents across. Boot the FAT32 partition using your favorite bootloader. Then can use the disk image to do all write activities. This is by far the most complex workaround.

I did all the setup and everything is running well till now, plug and play. Is there anyway i can check that my eGPU is using 1.2Opt?

I'm doing some benchmark with 3dmark11 and will update soon.

Thansk!

Checking x1.2Opt is working

Check with GPU-Z. Does it report x1 2.0 as the link speed during the first few secs when it starts up? When you open the NVidia control panel, then 3D settings, is there a Program settings tab where you can choose whether to run a program using "Integrated graphics" or "High-performance NVidia processor". If yes to both, then x1.2Opt link speed is working. These are shown below:

Hi nando,

Just one more silly question: do I need to do anything special to enable Optimus compression for my Nvidia card?

Recent NVidia desktop drivers automatically enable pci-e compression on a x1 link if using a NVidia Fermi or Keplar card and the bootup video device is an Intel 4500MHD or HD iGPU. Meaning you don't need to do anything special.

yeah sure i'll do so but at first i'll need some hours of sleep as its 4:30 in germany

i'll do the tests when im back home from work.

Unfortunately I don't have a Macbook to test with so your help there is appreciated

I got a reply back about the Setup 1.1x disk image writes failing on a Macbook Pro from a grub4dos developer. tinybit wrote here

If I remember rightly, Mac should not be the first one with this problem.

Sevral years ago I encountered a DELL which had the similar issue.

The DELL has no EBIOS write(ie., a failed int13/AH=43h). Int13/AH=42h worked fine. Normal CHS mode BIOS calls(int13/ah=2 read, int13/ah=3 write) also work OK.

Workaround 1: You may place your IMG at (near) the very beginning of your disk and expect it could be accessed in CHS mode.

Workaround 2: You may "map --mem" your IMG so it could be written in memory.

So if you have an advanced defrag utility, you could see if placing the DIYeGPU-Setup-1105.img file at the front of the HDD allows disk writes.