Jump to content
EwinRacing Flash Series Gaming Chairs
bsohn

2012 15" MBP GT650M + GTX970@10Gbps-TB1 (AKiTiO Thunder2) + Win10/7/OSX10.10 [bsohn]

Recommended Posts

My Hardware at the time of build:

Computer Hardware (option #1 - Win7pro or Win10pro)

15" Mid 2012 Macbook Pro (non-retina)

i7-3615QM 2.3GHz quad-core CPU

Intel HD4000 iGPU + NVidia GT650M dGPU

10Gbps Thunderbolt 1 port

Hybrid HD bootcamp with OSX 10.10.3 + Windows 7 Professional (64bit)

16GB Memory

2012 MacMini (option #2 - Win10pro)

2012 MacMini + GTX970@10Gbps-TB1 (AKiTiO Thunder2) + Win10 [bsohn]

Intention for the build - External GPU to host multiscreen set-up for gaming in Windows but would also like it to work in OS X

eGPU Hardware

AKiTiO Thunder2 PCI expansion box ($210 Amazon or $189 from AKiTiO Sales)

Galax GTX970OC ($337 Galax)

220W Dell DA2 Power Supply ($18 eBay)

Misc wiring, plugs and switches ($20 local electronics stores)

Custom Fan Grill ($35 customcargrills.com and paint)

Software @Tech Inferno Fan Setup 1.2/1.3 ($25)

Custom CNC parts, wiring, and case adjustments (done by me at Penguin r/c)

Total cost: $645 (Sans the custom CNC work as I have no price for that since I was able to do it)

The Case

Goal of the build was to create an eGPU that would be as small as possible to take up as little extra space as possible. In looking through all of the builds and doing research I found that the Thunder 2 case By Akitio was the best option to create a reliable, minimally connected, solid unit. It is small and well built. Only issue with the Akitio is actually it's small size as I found that very, very few cards of the latest style were able to fit fully inside the case.

The Card

First thoughts on the cards were the Gigabyte and Asus ITX X970s. I was leaning toward the Gigabyte due to the majority display port outputs. These two card I found however did not fit dimensionally within the Akitio (they fit length wise but not height wise. I then saw a build that used an EVGA 960 that fit perfectly inside the Akitio's case. For a short while I contemplated getting that EVGA 960 (seen in @elfxmilhouse 's build http://forum.techinferno.com/provisional-guides/9754-2013-13-macbook-pro-gtx960%4016gbps-tb2-akitio-thunder2-win8-1-%5Belfxmilhouse%5D.html) but in the 4GB rather than 2GB card (see note1). I still wanted a 970 so I pursued all of the builds I could find and studied the cards, their layout and their length and found one tell tale sign as to whether or not a card will fit height wise in the Akitio. That tell tale sign was what height the actual circuit board extended above the PCI screw plate flat. If the flat was at the same level as the bottom of the SLI pins or above then generally the card would fit, if the bottom of those pins was any amount above the plate height there was no way. By this criteria and a little information about dimensions I decided on going for the Galax GTX970OC.

note1: I contemplated the 4GB card for the 960 only due to the fact I wanted to run multiple monitors. When running multiple monitors more memory is a good thing because more pixels have to be drawn for each frame. i.e. a normal 1980x1020 would need to only draw and store one 1920x1080 frame but if you have 3 monitors it needs to draw 3 frames at 1920x1080 and retain all of them in memory.

The Power Supply

Do to my goal of compactness I was dreading the power supply. I kept reading you need 500watts of power and would need to use an ATX style power supply. Problem was I just didn't like that idea, after doing a ton of research I found out that that level of power was needed by the card AND the rest of the system and that an ATX for just powering the card would be overkill. I researched out a few different power supplies but always came back to the Recommended Dell DA2 that @Tech Inferno Fan and @Dschijn have used in their builds in the past.. Most of the other power supplies I found output 14-15V DC rather than the needed 12V that the card and the Akitio required. The DA2 is 220watts, fairly small, and more than enough to power the Nvidia recommended power for the GTX970 of 145watts, granted the Galax would probably draw a bit more than that due to being overclocked slightly. The issue with this power supply is the odd Molex plug.

The Build

OK so what to do now. I have all the parts but how to fit them together. I wanted it all self contained so I wanted to keep everything in the Akitio (except of course the power supply). I really didn't want to modify the DA2 so that if I ever needed to replace it I could easily. I wanted a power switch to turn the DA2 on and off so I didn't have to disconnect everything all the time (found out the remote on the DA2 allowed for this via @Tech Inferno Fan 's DA2 writeup http://forum.techinferno.com/enclosures-adapters/9426-220w-dell-da-2-ac-adapter-discussion.html), I didn't really want to modify the Akitio unless it was required, and I really didn't want to screw anything up. So head to the Electronic Store with the DA2 and one of the Galax supplied PCIe Power adaptors in hand to try to find a Molex plug for the DA2 and some unbuilt PCIe connectors.. Well as it turns out nothing was to be found.. The DA2 uses an odd Molex shaping that appears to be exclusive to it and the PCIe 6pin plugs are again specialized for that purpose so standard Molex Mini plugs are a no go. Upon finding this out I was a little discouraged but decided that I would de-pin pre-made PCIe power extenders and I happened to run across a 12pin Molex Mini that just so happens that 8 pins line up with the 8 pins on the DA2. I figured that it was small enough that I could get it to work. At this time I also came across a small SPST switch that I had originally planned on cutting a small hole in the front panel to install it.

So now I have all of my wires and am ready to built. (this thing is way tighter inside than I thought it would ever be). I put the card in the Akitio and just stared at it for a while with ll of the parts hanging around, placed the plugs and the switches all over the place seeing how they might fit in my mind. I knew I would have to make custom short height PCIe power connectors but arghhh 12 wires. Turns out for the PCIe 6Pin not all of those wires are needed (thank you). I was looking at the Galax adaptors that came with the card and realized that there were only 4 -12volt lines and 3 grounds to each plug so I did some research, turns out the PCIe 6 Pin Spec is only 2x 12volt, 2x Ground on the outer 4 pins and then the middle ground is only used as a Sense line for the card to determine that there is actually a 6pin plug connected. So cool Only 5 wires are needed (down to 10) so far so good we are making it easier at the 12pin molex as all of these wires have to go into the 6 x12v and 6 Ground spots on the DA2. In further research I learned a lot about Parallel wiring and the goods and bad about it. As it turns out the card could actually be power by 1 x 12v and 1 x Ground run across all of the 12v positions. Only problem with that is that is the wire would have to be fat to carry the amperage and the Molex pins themselves cannot handle enough current to do so. Parallel wiring allows the Amperage load and pin load to be spread across many lines evenly IF the wires are the same length and generally the impedance is the same. Understanding this and the Way that the PCIe standard was the key to my next decision. As it turns out I only need 4 wires running from each PCIe 6pin and a Ground Bridge to the Sense to turn on the 970's external power acceptance. I calculated that each wire would at most, by Nvidia Spec of 145watts for the card, need to carry 36.5 watts of power or 3-4 Amps of power. Molex pins are rated to carry 9AMPS of power per pin and 18GA wire is more than sufficient for that level of draw so with this set-up granted at the DA2 side now with 4 wires going into 2 we are looking at 8A per pin which is getting close to the max. That being said this calculation is based on the Nvidia 970 total power draw excluding the fact that the card can as well draw from the PCIe bus. If any wattage of significance is drawn through the bus then it would reduce the load on the PCIe Supplemental power plugs. This realization helped out significantly in wiring the DA2 as I could now use 4 of the pins for the supplemental PCI connections with 2 power and 2 pounds going to each plug. This left the final 12V Position and Ground for the DA2 to power the Akitio. I would also use this last ground to round the sense for the remote through the power switch.

Wiring all figured out it was off to work on putting everything into effect. In the meantime while visualizing what could be done I found that the rear venting on the Galax looked interesting.. The 4 cutouts then made in that area right above the single DVI-D Port looked like a good place to put the switch (cool its on the back rather than the front).. then I thought wait there is more area there, could I fit a plug. Turns out I could, not initially but with a lot of looking I found that if I cut out the Molex 4 pin positions that I was not using I could make this area fall behind the plate and it would fit. To do all this though I had to trim the crossbars out of the Galax card PCI plate to make one large opening. By this point I already knew I would have to cut a hole in the side of the Akitio for cooling so I wasn't too worried about covering that small area. I also had to make a custom mounting plate and found that I could fit it behind the plate and it would allow the Molex for the DA2 and the switch to be attached to it solidly. My worry The number of wires and the space provided. Since I had already been able to modify the PCIe plugs to be short enough I decided to modify the Molex the same way. this however was more difficult being that there was less space between the plug and the fan. Needless to say I got it all wired well. At this point I realized that the power functionality was all contained on the Video Card itself except for the connection to the Akitio. I had already planned to attach the power lined for the Akitio to the backside of the Barrel plug so that I didn't have to have an external connection. I decided at this point to put a plug inline with this (chose a 2 pin molex) so that I can now detach the Graphics Card and all power connections from the Akitio by that plug.

Build process photos in

All in all working well so what about cooling?

Cooling

To get the cooling I decided to cut a hole the shape of the Galax fans in the side of the Akitio. I placed a custom mesh grill over the hole to make the case look professional. The Grill is made from Aluminum grill plate that I got from customcargrills.com and then created a hole pattern and hammered it to shape in that pattern, this made the grill face almost flush with the outer surface of the case even though it is mounted inside the case. The Grill mesh is fastened to the case internally with epoxy. I did need to trim the metal out of the center of rear of the internal carriage to be able to get it to pass the grill when closing the Akitio. (you can see the trim in the last picture of the spoiler) Once done I tested the fan direction and found that the Galax Fans pull air into the case and as well found out that the Akitio Fan does the same. This would not allow for proper airflow through the case so I disassembled everything and removed the Akitio Fan, Turned it around and reinstalled it. This now creates an airflow pattern where the Galax pulls air into the case and the Akitio's fan pulls air out of the case. I believe that this will provide adequate cooling but if it should prove not to I will probably add vent holes to the top of the Akitio's case.

Next step is getting it working.

Windows Setup - (Macbook Pro 15" with Windows 7)

For Windows Pro 7 (64bit) (which I run through bootcamp) I actually found the eGPU really easy to install but I did figure out a few things in the process. It didn't seem to take as much to get done as some others have had in their builds. The first thing I did was download the latest drivers from Nvidia for the GTX970 and Windows 7 (don't remember the version number). Installed them as normal with the normal installer. This updated the internal GPU with the latest drivers. I then turned on the eGPU and restarted Windows, it of course froze at the boot loader immediately after the bootcamp system select and I had to hard restart. Trying to load with the eGPU again did the same thing. So, I turned the eGPU off and hard restarted agin and allowed Windows to load normally. Once loaded fully I turned the eGPU back on and restarted the computer, went through the bootcamp loader, and windows appeared to freeze however a few seconds later it loaded into Windows just fine no hangups. It passed the boot loader it just took it a few seconds (probably 15) longer this time (the first two times I let it sit and it was sitting for a minute or so). Once loaded Windows installed a bunch of plug and play software for the Akitio and the PCI bridge (note: I had never hooked the Akitio up to Windows prior to this with or without the card and this may have made a difference in the manner in which Windows recognized the eGPU).

At this point I looked at the Windows device manager and found that the eGPU was being recognized as a normal VGA Adapter and the devices control panel had the Yellow Triangle and the normal (as mentioned in the forums) code 12 resources not allocated error. I restarted again without turning off the eGPU and the computer froze again at the boot loader. So I repeated what I did earlier and loaded without the eGPU turned on, turned it back on after windows fully loaded and restarted. Once I restarted it again took a few additional seconds to get past the boot loader but it did load into Windows again with the VGA Adaptor and the code 12. To test I restarted the computer a couple of times with everything connected and it went straight into windows (after selection in the Bootcamp loader) with no issues.

At this point I decided to manually install the Nvida driver to the eGPU, it took successfully and now the eGPU was being recognized as a GTX970 in device manager. Again restart, Again Freeze, Again load without GPU on, again turn GPU on and restart, again a few seconds and again successful load. Starting to see a pattern here. Any Subsequent loads of Windows 7 were successful and recognized as a GTX970 but with code 12

Turns out through all of this loading and installing there is something with Windows 7 and its clean-up procedure after installing new components that causes it to freeze immediately at the boot loader if new software has been installed and the eGPU is connected. Once Windows is started normally once it allows the clean-up procedures to take place uninterrupted and it will then recognize the eGPU without issue the next time around. I am guessing the extra few seconds at the boot loader with the eGPU connected is windows trying to allocate resources but being unsuccessful. So for me from this point on I just made sure that if anything was installed in windows I did a clean restart prior to turning on the eGPU and it seemed to work fine (had to do a System update in the process as well as updated the bootcamp drivers, both coming after I figured out that lock-up procedure)

At this point I knew I had to get rid of the code 12 error to get this to work and that it was probably the only thing standing in my way (it was). Researching i was thinking that I could just do a DTST override. Turns out it was not possible. I received information that @Tech Inferno Fan ' s Setup 1.3 was the only thing that would work but I really didn't want to use more software so I still researched a bit. In the meantime I contacted @Tech Inferno Fan about the DTST override and basically was informed that Setup 1.3 was the only thing that could do it through PCI Compaction. So I ordered Setup 1.3 from @Tech Inferno Fan http://forum.techinferno.com/diy-e-gpu-projects/2123-diy-egpu-setup-1-30-a.html and told him about my set-up technical specs. He recommended using Setup 1.2 which he also provided a link to and pointed me toward @errin 's implementation http://forum.techinferno.com/implementation-guides-apple/9312-2012-15-macbook-pro-gtx960%4010gbps-tb1-akitio-thunder2-win8-1-bios-%5Berrin%5D.html as he was running the same computer that I was running.

Installation of Setup 1.2 was a breeze, followed the instructions provided by @Tech Inferno Fan and proceed to do compaction to get the eGPU code 12 eliminated. It took one compaction to get Setup to recognize the eGPU and I decided that I would skip actually trying to see if that alone worked and followed the steps and modifications in @errin 's implementation. Everything happened just as described in his implementation, even the request for "Remove disk a, Insert disk b" so I did the modifications to the pci.bat file. The only thing that I didn't get right was after the editing of the pic.bat I ran the test run on the compaction and couldn't get the right side Setup 1.2 menu to show “PCI Write” and “all_alloc (256MB)” as “YES”. Turns out I was running the wrong test and needed to run the test on the startup.bat after I did that all came up good. I restarted and everything worked perfectly no Code 12 and full external monitor support.

One thing to note is that with Setup (confirmed by @Tech Inferno Fan) you must always load the PCI Compaction by going through Setup's startup procedure, One more step but not a big deal Just load Setup 1.2 and select 1 (auto load) as all of the files should be correct then select start Windows and it just works. So, basically, the procedure is (boot loader/setup 12./boot loader/Windows) each time you load and wish to use the eGPU. In Talking to @Nando there is a way to eliminate this procedure using Setup 1.3 or you could eliminate the boot loader steps by opening Setup 1.2 directly from a USB stick. I haven't tried either of these as I am not sure the stick changes that much and @Tech Inferno Fan informed me that the other method if it didn't take properly could cause having to reload BOTH systems.

Windows Setup - (2012 Macbook Pro 15" with Windows 10 64bit on External USB3 Drive)

I installed windows 10 on an External USB 3 SSD Drive and was able to get it working with the same basic method that was used for Windows 7. I did run into a few Macintosh boot screen issues which I finally figured out were related to Windows starting up in Selective boot mode rather than in Normal boot mode. Once these were corrected for both Windows 7 and Windows 10 through msconfig and System Configuration respectively (and made permanent) everything worked as it should without any major issues.

Steps in order for Windows 7 Professional (64bit) or Windows 10 Professional (64bit USB external) on 15" Macbook Pro in spoiler.

I eliminated some of the things that I did above to streamline process as it should work.

eGPU Install

1. Install newest Nvidia drivers to internal GPU (GT650M)

2. Restart Windows

3. Turn eGPU on after Windows has started fully.

4. Restart Windows (Windows should code 12 with eGPU as VGA device after automatic installation of Thunderbolt and PCI bridge drivers)

5. Manually install Nvidia Driver to eGPU VGA device.

6. Restart Windows with eGPU tuned off

7. Turn eGPU on after Windows has started fully.

8. Restart Windows. (Windows should now load with code 12 and eGPU fully recognized with the correct driver)

9. Install Setup 1.2 from @Tech Inferno Fan

10. Follow PCI Compaction process in Setup 1.2 to get eGPU recognized. (all devices except SATA, Sound, ect...)

11. Follow PCI Compaction process in Setup 1.2. (all devices except SATA, Sound, ect...)

12. Edit and Save pci.bat as indicated by @errin (change all a:x.x and b:x.x to 0a:x.x and 0b:x.x)

13. Use Test to test the startup.bat and check for “PCI Write” and “all_alloc (256MB)” being “YES”

14. Once everything comes up correct start windows and everything should be good to go.

Startup Procedure

1. Windows boot loader select Setup 1.2

2. Select option 1 in Setup 1.2

3. Windows boot loader select Windows

4. Enjoy the use of the eGPU

Windows Issues if anyone knows fixes for these please let me know.

#1 - Problem:

Windows Crash w/Shutdown upon Hibernate Recovery as noted by the bootloader upon restart after it happened. - cause: Tried to close computer to use only the external display.

#2 - Problem:

Nvidia Surround (3 monitor) causes slow performance, crashy programs, and messed up bootup.

Fix:

Turns out that this issue is within the Nvidia driver being used (current as of 6.01.15) in that it doesn't co-exist well with Internet Explorer in the Windows 7 environment. The Boot issues were due to forgetting to restart Windows with the eGPU disconnected after a crash as Windows has to do some clean-up to its system as noted when installing. I now use Google Chrome when I am using the eGPU with the 3 monitor set-up in Surround and to be safe as there was a little weirdness going from the GTX7970 to the GT650M Internal Monitor I have also made the Surround display my primary and turned off Extended Desktop.

Update: The latest Nvidia Driver (as of 7.01.2015) still does not like Explorer when running Surround (3 Monitor) in Windows 7, The New Win 10 Web Browser has no issues

Benchmarks

3dmark13-FS.gpu=10508

3dmark11.gpu=13387

GT650M dGPU results in spoiler:

3DMark11 v1.0.132 - dGPU Nvidia GT650M - Intel i7 2.3Ghz[/b ]

Final - P2310 (GS2086 / PS7829 / CS1851)

3DMark13 v1.5.893- dGPU Nvidia GT650M - Intel i7 2.3Ghz

Firestrike - 1065 (1241G / 8348P / 316C)

Cloud Gate - 9702 (11672G / 6100P)

Ice Storm - 78455 (106250G / 40956P)

OSX 10.10.3

I have decided to wait to try to get this to work with OSX until the next major System release in October. It is not a priority to get OSX implementation for me as I am using it for gaming within Windows.

Thank you,

Brion Sohn

post-35125-14495000127177_thumb.jpg

post-35125-14495000129192_thumb.jpg

post-35125-14495000131277_thumb.jpg

post-35125-14495000132647_thumb.jpg

post-35125-14495000133717_thumb.jpg

post-35125-14495000135628_thumb.jpg

post-35125-14495000137545_thumb.jpg

post-35125-14495000126106_thumb.jpg

post-35125-14495000128178_thumb.jpg

post-35125-14495000130167_thumb.jpg

post-35125-14495000132_thumb.jpg

post-35125-14495000134707_thumb.jpg

post-35125-14495000136565_thumb.jpg

  • Thumbs Up 5

Share this post


Link to post
Share on other sites

Wow that is a lot to read ^^

I really love the build and the idea to route the plug through the Galax IO. By that you don't have to make any modifications to the case.

Some things I would like to mention:

- You can use (with a bit force) any 6 or 8-Pin PCIe plug on the DA-2

- 20 or 24 Pin ATX plugs can connect to the DA-2 in two ways without any force

- 145W for the GTX 970 is for the (not existing) reference card. All the cards on the market will use much more power (up to 200W at stock settings)

- I doubt that one slim 60mm fan will be capable to get all the heat out of the case

Regarding the cooling. It could work the way you inintended it to work, if you seal up the intake side so good, that the fans can only pull in fresh air. So ou should remove any gap between the fan (shroud) and the case, make it like a intake tunnel.

By that you would create a higher pressure inside the case and would force all the air out of it. With the 60mm fan on the front beeing the only "exit" it could work (maybe even better without the fan because it is limiting the amount of air).

Keep up the testing and the good work! Love all the creative ideas people come up with :D

Share this post


Link to post
Share on other sites

Ingenious mod of the dual-slot PCIe card slot plate. One I hadn't considered.

I'll add that the on-off switch is nice but not entirely necessary. The AKiTiO manages it's on/off when getting a TB signal when the notebook is on. Sure, the Dell DA-2 remains on (green) but with no load on it will not be using much power. When you "off" the DA-2 the light goes orange meaning standby.

Your idea raises the possibility of someone creating a ready-to-buy faceplate solution or at least a kit. One which:

- provides 75W slot power via the 4-pin connector: remove the AKiTiO yellow/black cable and have a cable that directly attaches to the 4-pin plug on the TB board. Only loss would be some filtering done on the DC input which seems not necessary given some ppl were "back powering" using a powered PCIe riser.

- provide 6P+8P PCie connectors of suitable length, preferrably as low-profile connectors that could allow the top cover to slide over.

Great stuff :)

Share this post


Link to post
Share on other sites
Wow that is a lot to read ^^

I really love the build and the idea to route the plug through the Galax IO. By that you don't have to make any modifications to the case.

Some things I would like to mention:

- You can use (with a bit force) any 6 or 8-Pin PCIe plug on the DA-2

- 20 or 24 Pin ATX plugs can connect to the DA-2 in two ways without any force

- 145W for the GTX 970 is for the (not existing) reference card. All the cards on the market will use much more power (up to 200W at stock settings)

- I doubt that one slim 60mm fan will be capable to get all the heat out of the case

Regarding the cooling. It could work the way you inintended it to work, if you seal up the intake side so good, that the fans can only pull in fresh air. So ou should remove any gap between the fan (shroud) and the case, make it like a intake tunnel.

By that you would create a higher pressure inside the case and would force all the air out of it. With the 60mm fan on the front beeing the only "exit" it could work (maybe even better without the fan because it is limiting the amount of air).

Keep up the testing and the good work! Love all the creative ideas people come up with :D

On the power I figure the Galax is going to pull more as it is Overclocked a bit, standards say that it could pull up to 225watt based on the power config (75 PCIe + 75 PCIeSUP + 75 PCIeSUP) which still puts me at the 4A per PCIe pin (37.5 watt/3.125A per pin) and 7-8A at the DA2 Connection (still within spec). If it actually does pull 225 watt then I will be a little out of luck as the DA2 is only 220 watt.

Yea, I am a bit concerned with the cooling but haven't had issue yet running one monitor (not sure when I get the three running) normal cooling methodology would say to have similar to electrical wiring in that the input and output being equal so there is a cooling bottleneck with what I currently have.. It actually probably wouldn't be to hard to seal fan gap with some window foam to create a higher pressure system. At least I realized that both fans were putting air into the box, not realizing that would have been bad for heat.

On the Molex the 12 pin I found required no force for the 8Pin match on the one end so that is probably the smallest plug that can be used direct without tearing up an Optiplex or forcing the fit. I just cut down the four unused and actually ended up using one of the empty pin location to use a screw to more securely lock the custom board to the card (the screw right above the DA2 port).

This was actually a little more fun to do than I thought it would be.

Brion

- - - Updated - - -

Ingenious mod of the dual-slot PCIe card slot plate. One I hadn't considered.

I'll add that the on-off switch is nice but not entirely necessary. The AKiTiO manages it's on/off when getting a TB signal when the notebook is on. Sure, the Dell DA-2 remains on (green) but with no load on it will not be using much power. When you "off" the DA-2 the light goes orange meaning standby.

Your idea raises the possibility of someone creating a ready-to-buy faceplate solution or at least a kit. One which:

- provides 75W slot power via the 4-pin connector: remove the AKiTiO yellow/black cable and have a cable that directly attaches to the 4-pin plug on the TB board. Only loss would be some filtering done on the DC input which seems not necessary given some ppl were "back powering" using a powered PCIe riser.

- provide 6P+8P PCie connectors of suitable length, preferrably as low-profile connectors that could allow the top cover to slide over.

Great stuff :)

I chose to keep the on/off switch because I knew from reading that there can be start-up issues at times with the system modifications. By having the on/off switch I don't have to always be connecting or disconnecting the power or thunderbolt cables to cycle the eGPU. For general on off though it isn't really needed as @Tech Inferno Fan is correct in that the Akitio turns everything off with Thunderbolt signal loss, the DA2 stays in active mode.

Brion

Share this post


Link to post
Share on other sites

How did you cut the side panel? Looks extremly clean! :D

Did you paint the cutting edges? I simply used a black permanent marker.

Share this post


Link to post
Share on other sites
How did you cut the side panel? Looks extremly clean! :D

Did you paint the cutting edges? I simply used a black permanent marker.

I have access to a CNC Machine with my business so I was able to do the side opening and the custom power plate on the machine. The cut edge was sprayed with a satin black paint this is easy as all you need to do for masking is to put tape all around the edge and then cut it back with an exacto to the shape. Cover the rest of the box and if you want stick a piece of paper inside and spray. I actually thought of the Marker for a quicker way of doing it but I don't like the red haze Sharpie ink leaves and I was already having to paint the grill (straight aluminum) so the paint was there.

Brion

  • Thumbs Up 1

Share this post


Link to post
Share on other sites

@bsohn: Did you monitor your temperatures of that build? I'm interessted if that exhaust is enough to get rid of all the hot air. By intaking fresh air the temperatures should be fine. :)

Also did you consider building a "shroud" to make sure the GPU fans are only getting fresh air and are not reusing warm air from the inside of the case, as there should be a gap betwenn the fans and the intake mesh.

Share this post


Link to post
Share on other sites
@bsohn: Did you monitor your temperatures of that build? I'm interessted if that exhaust is enough to get rid of all the hot air. By intaking fresh air the temperatures should be fine. :)

Also did you consider building a "shroud" to make sure the GPU fans are only getting fresh air and are not reusing warm air from the inside of the case, as there should be a gap betwenn the fans and the intake mesh.

I didn't really monitor the temps (not sure how to software wise but I think there is a method) but during the benchmark tests the temperature of the air coming from the case through the front was not insane (warm yes, hot no) and the actual Akitio case never got too warm.. My Macbook n the other hand when running the internal GPU tests with the 650M got quite hot, couldn't really touch above the keyboard for more than a second (typical macbook heat buildup location) under the load.

I did note that the X970 did put out some whine when running but I have read that is common with the X970. I'm still considering the shroud and what would be the best method of accomplishment for that as I have to be able to of course slide the internals in and out without catching it.

Brion

Share this post


Link to post
Share on other sites

You can use MSI Afterburner or EVGA Precision X to monitor the temperatures. These tools also enable you to OC the card or to change the fan profile.

Most GTX 970 have a coil whine/buzzing problem because all manufactures are running them beyond the "reference" specifications. You might get a less noisy card by downclocking and lowering the power target. Of course that will also lower your performance!

For the sliding mechanism I tried to attach everything to the inner cage. Only the outer fan grills are screwed to the outer case. With them beeing countersunk screws there is luckily no collision.

Share this post


Link to post
Share on other sites
You can use MSI Afterburner or EVGA Precision X to monitor the temperatures. These tools also enable you to OC the card or to change the fan profile.

Most GTX 970 have a coil whine/buzzing problem because all manufactures are running them beyond the "reference" specifications. You might get a less noisy card by downclocking and lowering the power target. Of course that will also lower your performance!

For the sliding mechanism I tried to attach everything to the inner cage. Only the outer fan grills are screwed to the outer case. With them beeing countersunk screws there is luckily no collision.

I am guessing I should just log from one of these programs while running the Benchmark tests for a temp profile. The coil whine actually wasn't all that bad I just noticed it at times under the heavier loading segments of the Benchmarking. if it wasn't sitting right next to me while doing it I probably wouldn't have even noticed it.

On mine with the shroud might take a bit of experimentation because the grill is epoxied to the inside of the case, I had to trim the internal carriage flap next to the outputs so that it could slide past the grill. So the issue for me would be that the shroud would catch on the grill as I am sliding it in. I'm not totally sure the primary gaming I am going to be doing will use the amount of power this is capable of so I might not need the shroud but it could be a good test.

Brion

Share this post


Link to post
Share on other sites

Please log the temps one time :D

I would be really interessted.

Share this post


Link to post
Share on other sites
Please log the temps one time :D

I would be really interessted.

OK I downloaded the EVGA version.. Not sure how to get it to log but I watched the Graphs and took some notes during the 3DMark13 test.

With stock settings with 1379mhz boost these were the max numbers and some other noted numbers:

Base Idle temp (low usage just displaying normal stuff) is approx. 30-35c

Base Running Temperature is approx. 50c

Base Fan Speed is 1100rpm (30% speed) estimated MAX speed 4000rpm

Ice Storm 1.2 running numbers -

Max temp noted 62c

Max fan noted 2075rpm

Cloud Gate 1.1

Max temp noted 70c

Max fan noted 2500rpm

Fire Strike 1.1

Max temp noted 80c

Max fan noted 2825rpm

Other notes: I do think that the push pull of the fans works to a good advantage as cool down time is about 2c every second so temps do not appear to build when running continuously. Between tests the eGPU cooled to what I would consider the base running temp of 50c. The 80c peak on Firestrike was only for a second and it cooled to and more or less ran in the 75-77c range.

I did note a little bit of MHz and voltage fluctuation during the Fire Strike test dropping intermittently to 1366mhz and about 11.5V - 11.75V consumption. Generally with the other tests it was pegged at 1379mHz at 12V.

It actually ran 4 points better on the benchmark than it did in my first test on this run through (8734, 10518GS)

anyway there are some numbers, no shroud to tunnel air with the card fans pulling in air and the Akitio's fan set up to push air out the front.

Brion

  • Thumbs Up 1

Share this post


Link to post
Share on other sites
Thanks Brion! That looks good. Seems like your card is hitting differnt limits:

- power

- termperature (target at 80°C)

If any of that targets is reached, the card will clock down to get below the target.

The limits that were set for it were the EVGA Programs Default which I believe was 100% power and 80c Temperature with maintaining power as the primary.. Maybe it was flirting on that 80c limit and dropping the mHz slightly.. Wondering why it didn't decide to just press the fans harder though to maintain temp as from what it looked like it was only running at about 75% fan speed at 80c.. I probably won't push it for Maximum performance as I am not sure I would need it but it looks like from those numbers that I might have some room there to be able to if I chose as the fan speed wasn't maxed and it fell pretty much with in default targets.

Brion

Share this post


Link to post
Share on other sites

The EVGA tool just used the stock settings of the card. You could consider adjusting the fan profile in the EVGA tool… but of course that will result in more fan noise to get lower temperatures.

Share this post


Link to post
Share on other sites

Hi, bsohn. Im working with the same laptop as you have and I was wondering if it is necessary to have a fresh install of windows? I've had a windows 8.1 partition for quite awhile now. Though, all the projects I have read about start with a new installation.

Share this post


Link to post
Share on other sites

I didn't need to do any Fresh install I just had to use @Tech Inferno Fan's Setup program.. I don't know if Windows 8.1 will need a fresh install as I am on Win7Pro. The only Windows Quirk that I noted that is in Win 7 is that if Windows does an automated update or you install something (anything) You must load into windows without the eGPU turned on to allow windows to do whatever clean-up is needed or the Macbook will freeze trying to load the boot loader initially with the eGPU on. Once this clean-up is done you should be able to have the eGPU on when you turn on the computer and load as normal without having to do any weird (plug in at this moment types of things).. I still havent attempted OSX but now that 10.4 is coming out I may just wait though there have been some major advancements on External use for OSX recently.

Share this post


Link to post
Share on other sites

Hi bsohn,

awesome job! could you let me know how certain you are Gigabyte GTX970 mini doesn't fit in Akitio? i've just googled for specs and Akitio height is 149, gigabyte - 129 - will that just not fit at all or does it leave not enough space for something i don't know yet?

thanks

Share this post


Link to post
Share on other sites

Sorry for the late reply. I am 95% sure that due to the height of the Board above the PCI Bracket that though the Gigabyte GTX970 would fit in the case you would not be able to plug the Auxiliary power plugs into the ports on he top of the card and be able to close the case.

Brion Sohn

  • Thumbs Up 1

Share this post


Link to post
Share on other sites
Hi bsohn,

awesome job! could you let me know how certain you are Gigabyte GTX970 mini doesn't fit in Akitio? i've just googled for specs and Akitio height is 149, gigabyte - 129 - will that just not fit at all or does it leave not enough space for something i don't know yet?

thanks

I'm 100% certain it does not fit fully enclosed. I had the Gigabyte card when I first bought the Akitio and confirm it's too tall to have the outer case closed. However, if you don't plan on modifying the Akitio to allow proper cooling, it doesn't really matter (presuming you are not looking to use the outer case as protection for traveling only). If its just going to sit on your desk as a chassis, then you are good to go.

Share this post


Link to post
Share on other sites
I'm 100% certain it does not fit fully enclosed. I had the Gigabyte card when I first bought the Akitio and confirm it's too tall to have the outer case closed. However, if you don't plan on modifying the Akitio to allow proper cooling, it doesn't really matter (presuming you are not looking to use the outer case as protection for traveling only). If its just going to sit on your desk as a chassis, then you are good to go.

thanks martinm0, i've figured that out by browsing more threads :) got myself kfa/galax card that should fit entirely. waiting for all that stuff to be delivered.

Share this post


Link to post
Share on other sites
thanks martinm0, i've figured that out by browsing more threads :) got myself kfa/galax card that should fit entirely. waiting for all that stuff to be delivered.

The Galax/KFA card will fit with no issues except you will need to create some sort of low profile plug for the auxiliary power. Take a look at @seefew 's build for an interesting circuit based method of doing this (kind of like his method better than mine wire management).. Personally as you can see in my build I used the shortened plug with bent pin method but I also limited the number of wires. I reduced them to the PCIe6 standard 4 wire setup with a bridge on the ground for the trigger (to remove the 5th wire). Most of what you see with the 3x3 wiring is because with that wire setup it allows for the PCIe8 plug expansion which is not needed for this card. (realized by the PCIe Supplemental adaptors provided by Galax with my card.

Brion

Share this post


Link to post
Share on other sites

Did you have issues getting the eGPU recognized in setup 1.20? I've read through most of the MBP implementations, and I can't find any that cover getting the EGPU recognized in the setup environment.

Share this post


Link to post
Share on other sites
Did you have issues getting the eGPU recognized in setup 1.20? I've read through most of the MBP implementations, and I can't find any that cover getting the EGPU recognized in the setup environment.

I have found that the eGPU will not be recognized by Set-up 1.2 IF it is not fully turned on (Akitio and Card) and attached prior to either Windows load (Direct Load) or the Mac EFI Loader screen. It seems as if the Mac checks the Thunderbolt port immediately on start and plugging it in after that point will not allow recognition in the early boot processes.

If you follow the directions in the spoiler you should be able to get recognition in Windows 7 and Setup 1.2 - If not you could have other issues beyond software (Power, Installation, etc..)

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • 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.