Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
There is a script that makes all the kext mods and it is very very comfortable!

For the price of the Sonnet Echo Expresscard you can also get the AKiTiO Thunder2 and rely on the builds of MANY other people with the AKiTiO.
 
  • Like
Reactions: jblagden
synchro3 says in the OP...

- When I began my first eGPU the only affordable adapter was the PE4L (PCIE-expresscard):http://www.hwtools.net/Adapter/PE4L V2.1.html together with Sonnet Echo ExpressCard Prohttp://www.sonnettech.com/product/echoexpresscard34thunderbolt.html

Does anyone know if these exact same kext edits and procedure would work with the PE4L (PCIE-expresscard)? Or are they thunderbolt specific? How might I get the computer to recognize the expresscard eGPU without the Thunderbolt Sonnet adapter? The OP never mentions this guide in particular works just that when he first began he used the PE4L. Did he ever use the PE4L with El Capitan? Thanks is advance!

If you use the PE4L directly connected to an old MacBook Pro via ExpressCard-Slot you don't have to edit the KEXT's. Expresscard is PCI-Express. eGPU will work OOTB.

The same with Expresscard-Adapter for Mac Pro, no KEXT editing needed: https://forums.macrumors.com/threads/mac-pro-with-express-card-34-adapter.1695547/

Expresscard.png
 
Last edited:
  • Like
Reactions: laimis17
If you use the PE4L directly connected to an old MacBook Pro via ExpressCard-Slot you don't have to edit the KEXT's. Expresscard is PCI-Express. eGPU will work OOTB.

The same with Expresscard-Adapter for Mac Pro, no KEXT editing needed: https://forums.macrumors.com/threads/mac-pro-with-express-card-34-adapter.1695547/

View attachment 610770

wow thanks for reply! this helps give me a bit more confidence, but I'm not quite using an old macbook pro, I'm using an hp 4530s with clover UEFI to boot El Capitan. I'll give it a try but I thought I'd at least have to add device ID or something. I guess I'll just turn the card on when clover boots.
 
I am working towards building my first eGPU. The purpose is not for gaming but instead, support multiple displays. I am merely looking to have a MacBook Pro output to 5 displays via two native Thunberbolt-2 ports.

I have some existing hardware listed below.
  • Laptop: MacBook Pro (Retina, 13-inch, Late 2013)
    • Processor: 2.8 GHz Intel Core i7
    • Memory: 8 GB 1600 MHz DDR3
    • Graphics: Intel Iris 1536 MB
    • OS X EL Capitan v. 10.11.2
  • Video Card #1: EVGA GeForce 8800 Ultra
  • Video Card #2: EVGA GeForce 8800 Ultra
  • Power: ZALMAN ZM850 HP
  • Display #1: Apple Thunderbolt Display 27” (2560 x 1440)
  • Display #2: Apple Thunderbolt Display 27” (2560 x 1440)
  • Display #3: Dell 3007WFP 30” LCD - DVI-D (2560 x 1600)
  • Display #4: Dell 3007WFP 30” LCD - DVI-D (2560 x 1600)
  • Display #5: Dell 3007WFP 30” LCD - DVI-D (2560 x 1600)
I am reaching out to see if anyone can recommend a configuration which will support what I’m working to accomplish.

5rsMZ_cZp_IFQrYv3r1MpWr47lxjnDga3m8pgxD2i1-ULnbDe1Z9fhe-3P4EbUZ6t3wIEbwcVrXJINFbBJdez2zlZjdRH5NhZ2rWQgeNtlLfrmSUojZL0sQwgmNJ1c2skEfcExgwugRZYPBoex1bYxAfeQYrfR64RAk8YQq7F23WhFC6E-XhHITOkA47Pa-3ZKSlZG2BuIdHhlgwUW1ONdfLGtg7Y5semwQG9Bjha6lJ-f8Aybdmq4jSgG_Bc6W7RWEBfLiAbhDA7F4fFQnZD4uChILQHSmGgxEl_IZxEjCVsAJVB_m_BQuIPhQXAU1UPl9lqdRW6G82Vk0Ee9Gy2S5ZxuBX-Q7WMiqQeJTlCr2d4NEXSRRz7HrutaTJdC00iVEWEsm-DUrgfY30-Cmi9Fl60bD-RdLsFZXTZRbhYPhlQTh_plbdHubaSVEJP-r5dzGnfJdkFtd7MOGPhKRepNO7WPE8rB07w7JHjq3lPY-T8MVqkBrv8JNRorsdce7PDY5qbTNB7ad_8_al7k2zBliJ9bqCxA_EDy3JiDf0A0veZkOa6CWnd_aFR-4mJNSyn0PKuA=w1280-h720-no
 
I had 4 displays running on my eGPU and even 8 displays with two eGPU daisychained.
Not sure if that 8800 cards are good enough for that purpose...
 
64 watts should be fine for a GT 740, but yeah, a 76 watt PSU might not be sufficient.

Well, ordered a 12V 10A power adapter and hooked it up but still getting the same results. GT 740 is detected and shows up in system profile but no monitor is shown attached and no video out on that cinema display. Old GT210 still works fine though.
 
I had 4 displays running on my eGPU and even 8 displays with two eGPU daisychained.
Not sure if that 8800 cards are good enough for that purpose...

I appreciate the feedback. I'm open to picking up a couple newer cards. I just need a configuration which will support the displays and resolution. Any recommendations on the hardware I should purchase would be a big win.
 
Any current Nvidia based graphics card should work, if you don't mind using adapters. Can you displays only handle DVI?
 
If you use the PE4L directly connected to an old MacBook Pro via ExpressCard-Slot you don't have to edit the KEXT's. Expresscard is PCI-Express. eGPU will work OOTB.

The same with Expresscard-Adapter for Mac Pro, no KEXT editing needed: https://forums.macrumors.com/threads/mac-pro-with-express-card-34-adapter.1695547/

View attachment 610770


So I followed your guide and didn't touch the kext stuff because I'm using the expresscard slot and the PE4L (your original setup).

1)I downloaded my 10.11.1 nvidia web drivers package.
2)Turned off SIP. confirmed disabled.
3)Rebooted and at login screen plugged everything in and turned on (I did this once more but turning everything on after login).
4)I got my GT240 card to turn on (or at least the fan was spinning) but no video output or external monitor recognized by OSX. Expresscard slot does show menu bar as having a expresscard in the slot, but thats all I see that OSX recognizes.

I can try a GT430 or GTX460 but then I need to send them more power, so I assumed the GT240 would be easiest to test.

Am I missing something or is this maybe a complication from Clover bootloader?... I think it's the only difference I can tell between my setup and a macbook pro that might make the hardware or software behave differently.

update:GT240 uses native osx drivers not the nvidia web drivers. should I uninstall the web driver utility? Anyway now I see card listed in about this mac under graphics as...

NVIDIA CHIP MODEL
PCIe LANE WIDTH x1
DeviceID: 0x0ca3
ROM rev. preset 1.0.0

I've got it hooked up to a monitor but still no signal being transferred. I get a black screen on laptop if booting with eGPU already plugged in and turned on. I have to force power off.
 
Last edited:
What I struggle with are the other things needed to make an eGPU. Assuming I have a power supply and graphic cards.

You will need:
- AKiTiO
- PSU
- barrel mod to power the AKiTiO with the PSU
- GPU (e.g. a EVGA GTX750Ti which has DVI, HDMI 1.4a, DisplayPort 1.2, each 1x)
 
  • Like
Reactions: kwickstrom
There are many instruction blogs/posts on how to build an eGPU, and taken all together (including the OP of this thread) there's a lot of good information there. Also, still, some vague or misleading information, like is the CUDA driver needed?
Mmm, not likely, unless you wish to do parallel computing on your eGPU, and in that case there's also OpenCL (part of XCode) on the Mac.

I finally built an eGPU for my late 2014 base model Mac Mini, and people may be interested in my complete hardware/software report, https://dl.dropboxusercontent.com/u/5115339/egpu.pdf.

Enjoy!

Thanks for write-up... I'm using an expresscard and wondering if I need to do the step below (from your mac mini 2014 guide) or if it is only for thunderbolt or nvidia web driver? I have the card recognized in system profile but it is not available to use with my external monitor.

Make eGPU selectable The late-2014 model Mac Mini will not enable the eGPU without one more kext modification, for which you’ll have to know the board-ID of your Mac Mini. Do this with the Terminal command, ioreg -l |grep ’board-id’ which showed it to be Mac-35C5E08120C7EEAF in my case. Find 7 • AppleGraphicsDevicePolicy.kext in the AppleGraphicsControl.kext/C*/P*/ folder and look in its Info.plist for the appropriate entry and change its value to none, ConfigMap ... Mac-35C5E08120C7EEAF/key> none ... Adjust as necessary for a different board-ID.
 
Well, ordered a 12V 10A power adapter and hooked it up but still getting the same results. GT 740 is detected and shows up in system profile but no monitor is shown attached and no video out on that cinema display. Old GT210 still works fine though.
If you are using the original Akitios thunder2 board for powering the card then I think it maxes out sending 60W to the card via the PCI slot. I think you'd have to split the power in your 12v 10A (120W) supply to send power to board and card (or just power to card? via a 8pin (6+2) connector). You should read the DA-2 power supply mods at tech inferno as I think you have a power issue. The PCI lane power that comes with the Akitios t2 isn't designed to power eGPU cards so you need some mods. This explains why the GT210 has no issue... its very low power.
 
So I followed your guide and didn't touch the kext stuff because I'm using the expresscard slot and the PE4L (your original setup).

1)I downloaded my 10.11.1 nvidia web drivers package.
2)Turned off SIP. confirmed disabled.
3)Rebooted and at login screen plugged everything in and turned on (I did this once more but turning everything on after login).
4)I got my GT240 card to turn on (or at least the fan was spinning) but no video output or external monitor recognized by OSX. Expresscard slot does show menu bar as having a expresscard in the slot, but thats all I see that OSX recognizes.

I can try a GT430 or GTX460 but then I need to send them more power, so I assumed the GT240 would be easiest to test.

Am I missing something or is this maybe a complication from Clover bootloader?... I think it's the only difference I can tell between my setup and a macbook pro that might make the hardware or software behave differently.

update:GT240 uses native osx drivers not the nvidia web drivers. should I uninstall the web driver utility? Anyway now I see card listed in about this mac under graphics as...

NVIDIA CHIP MODEL
PCIe LANE WIDTH x1
DeviceID: 0x0ca3
ROM rev. preset 1.0.0

I've got it hooked up to a monitor but still no signal being transferred. I get a black screen on laptop if booting with eGPU already plugged in and turned on. I have to force power off.

'NVIDIA CHIP MODEL' -> This is a symptom when the card has not enough power

Did you power the PE4L separately? It has to deliver 75 Watt. Do not use the USB-Port to power the PE4L. Or are you using a powered riser (with capacitor)? I can run a GTX 570 with the PE4L via Expresscard.
PE4L-.jpg
 
Last edited:
'NVIDIA CHIP MODEL' -> This is a symptom when the card has not enough power

Did you power the PE4L separately? It has to deliver 75 Watt. Or are you using a powered riser (with capacitor) and PSU? I can run a GTX 570 with the PE4L via Expresscard.

Thanks for reply! I am using ATX 400W PSU. I've got the 4pin (molex to 4pin fan) on just like you, but I dont have the dupont 2pin connected to the SWEX (I am using the SWEX to start my PSU)... I dont have that cord. I have a usb m to usb m in the usb plug and connect that to my laptop which supposedly adds a more stable 3.3v. Is the dupont 2pin required? Do I need to mess with modifying the board-id like the thunderbolt tutorial to enable eGPU screen output? Or is this native with my setup?
 
No, afaik the dupont 2pin is not required if the button (3 possible positions) is on the right position. Also check the On/Off-Button etc on the PE4L.
PE4L B.jpg


Or the GT240 is not supported by actual drivers… It's indeed a very old card. Fermi, Kepler and Maxwell cards should work. Maxwell only with Web drivers.
 
Last edited:
yes that is how I have my switches set SW1=1, SW2= 2-3. Maybe I need to set SW2=1 for x1(default) instead of 2-3 x2 which I have it set to now? I think my PCIe lane maxes out at PCIe 2.0 x1.

not sure what the jumper is... the swex? there is only one switch for on and off.

update: tried to put SW2=1 for x1(default). Card fans did not come on as before. I assume SW2=2-3 is the better setting since at least that powered the fan for the card and got something in system profile.

next: ok i'll try my gtx460 - fermi and I'll keep use native OSX drivers enabled.

update: got gtx460 to power on, but it has the same response. shows up in system profiler as Nvidia Chip Model. Plenty of power from my 380W bronze PSU (different than one I tried with GT240). Tried with native and then with web drivers.

I'm out of ideas... not sure where my process is failing. anyone got some advice for an expresscard non-thunderbolt process?
 
Last edited:
No, afaik the dupont 2pin is not required if the button (3 possible positions) is on the right position. Also check the On/Off-Button etc on the PE4L.

Since I'm not using the dupont 2pin are you suggesting I try the other two positions on SW1? This switch changes the perst delay. I will try positions 2 and then 3. I'm supposed to be turning my eGPU power on and plug in expresscard in at my login screen, right? or do I need to turn everything on at my clover bootloader screen? monitor connnected to card and powered on (waiting for signal) at all times?

edit: tried settings 2 and 3 on SW1 and nothing changed.

edit: will try a Radeon HD4850 AMD card.
 
Last edited:
If you are using the original Akitios thunder2 board for powering the card then I think it maxes out sending 60W to the card via the PCI slot. I think you'd have to split the power in your 12v 10A (120W) supply to send power to board and card (or just power to card? via a 8pin (6+2) connector). You should read the DA-2 power supply mods at tech inferno as I think you have a power issue. The PCI lane power that comes with the Akitios t2 isn't designed to power eGPU cards so you need some mods. This explains why the GT210 has no issue... its very low power.
I thought the Akitio sent out up to 75 watts to the card.
 
'Some pics' are in my first post.


Costs:
PSU 500W $70
Elite 130 $50
SWEX $28 (incl. shipping from China)
Powered riser $7
Thunderbolt cable 2m $50

Thunderbolt Adapter
Akitio $220 or ThunderTek/PX $200

Graphics card depending on your needs: From $0 if you have already one lying around to $1000 for a GTX Titan X.

Let's say a with a new midrange card like GTX 970 for $350 the whole eGPU would cost you $750.
That's pretty expensive
 
I thought the Akitio sent out up to 75 watts to the card.

The stock PSU of the AKiTiO has only 60W, which have to be split up between the AKiTiO itself and the PCIe slot. The official support limits the PCIe slot to 25W because of the weak PSU.
But AKiTiO themself confirmed that you use a more powerful PSU for the input and power 75W to the slot (without modifications on the board or circuit). Most eGPU users with the AKiTiO are using a more powerful PSU and a barrel plug connected to it, to achieve that.
 
The stock PSU of the AKiTiO has only 60W, which have to be split up between the AKiTiO itself and the PCIe slot. The official support limits the PCIe slot to 25W because of the weak PSU.
But AKiTiO themself confirmed that you use a more powerful PSU for the input and power 75W to the slot (without modifications on the board or circuit). Most eGPU users with the AKiTiO are using a more powerful PSU and a barrel plug connected to it, to achieve that.

This is what I also found via research. Went with the GT 740 specifically because it was powered by the slot. Want to keep my setup quiet and organized without cables hanging out everywhere. Since I've tried a 96W and now a 120W power supply, it shouldn't be a power supply issue. Also since the card is recognized by the OS, I would think the card is getting enough power.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.