Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Neat little website! I was hoping for something similar to the iGPU and didn't realize this card was much weaker. I hope I don't have any issues. [EDIT]It actually looks like that website doesn't have the iGPU i'm using. My Intel HD Graphics 3000 is only 512MB.[/EDIT]

I didn't want to take a chance on another card and have to use an external power supply as not much info is available on the max watts in the Helios. I really like the Quadro K1200 but it's like $300 bucks and I didn't feel like it made sense to spend that much. I scored the Helios off craigslist for $80, so all in, this setup was super inexpensive to accomplish my goal of a second display at 2560x1440.

I also considered buying a new Mac Mini with the two TB ports but am really hard pressed to buy a machine that has no user upgradeable RAM / HDD and only supports 16GB max. I currently have that much RAM, an SSD + 2.0GHz i7. A new one configured w/ i7 comes in at $1600 and all I would gain is 1GHz extra on the processor and the extra TB port. All the base models have an i5. Kind of ridiculous. My next machine may end up being a hackintosh.

Will the eGPU be viewed as the main GPU or can it be set as secondary?

You can set the eGPU as main or as secondary as you want. Same as with an external monitor connected directly to a Mac.

As you already have a Mac Mini 2011 with Thunderbolt 1 and the Helios I would try the NVS 310 for the time being and for first eGPU experiences, and later buy a better card (Nvidia Kepler and Maxwell cards support for example 4K Displays, Maxwell cards need the Nvidia web driver, Kepler cards work with native OS X driver). The newest Pascal cards are not supported in OS X at all.

Short: Avoid old Fermi and newest Pascal cards. Look for Kepler or Maxwell cards.

By the way: The NVS 310 is an old Fermi generation card, that might not work with newest OS X versions.

GF means Fermi generation, GK = Kepler, GM = Maxwell, GP = Pascal


https://en.wikipedia.org/wiki/GeForce_600_series
https://en.wikipedia.org/wiki/GeForce_700_series
https://en.wikipedia.org/wiki/GeForce_900_series

This thread is helpful: https://forums.macrumors.com/thread...out-nvidia-pc-non-efi-graphics-cards.1440150/
[doublepost=1479412086][/doublepost]
Hey guys, I have a question about eGPU's. I have a MacBook Pro (late 2016) 15" with 460 on order. Right now I have a late 2013 rMBP 15" that I use in clamshell mode with 3 monitors. That is 1 27" 1440p display and 2 24" 1200p displays. At the moment I connect all my displays with each display on another port. Which means I use the HDMI and the two mini DP/TB2 ports.
For my new MBP I am looking for a solution with which I can connect one or two cables which would connect all my displays and possibly ethernet.

If I'm not mistaken, this should be possible by using an eGPU like the Razer Core or PowerColor Devil Box. They both supply 60w of power to the laptop and have an ethernet port. If these eGPU enclosures work on the late 2016 MBP, do you guys think this is a good setup? Are eGPU's stable enough so I can plugin my laptop and boot/wake from sleep? And if the GPU (for example AMD RX 480) support DisplayPort 1.3 or higher, would it be possible to use DisplayPort 1.3 or higher monitors through this GPU?

Thanks in advance!

You have to edit IOThunderboltFamily kext with Razer Core or Devil Box, newest findings here: http://forum.netkas.org/index.php/topic,11654.msg34140.html#msg34140

If the card and the monitor have DisplayPort 1.3 or 1.4 you can surely exploit all benefits. I'm using my Mac Mini 2012 via DisplayPort 1.2 eGPU with a 4K monitor @60 Hz.
 
Last edited:
This thread is helpful: https://forums.macrumors.com/thread...out-nvidia-pc-non-efi-graphics-cards.1440150/
[doublepost=1479412086][/doublepost]

You have to edit IOThunderboltFamily kext with Razer Core or Devil Box, newest findings here: http://forum.netkas.org/index.php/topic,11654.msg34140.html#msg34140

If the card and the monitor have DisplayPort 1.3 or 1.4 you can surely exploit all benefits. I'm using my Mac Mini 2012 via DisplayPort 1.2 eGPU with a 4K monitor @60 Hz.

I already saw that thread, it's really interesting what's happening there. If that all works, how stable is an eGPU in combination with a MacBook Pro? I saw in your signature you are running one with a 13" MBP. Especially in OSX I want stability.
 
I already saw that thread, it's really interesting what's happening there. If that all works, how stable is an eGPU in combination with a MacBook Pro? I saw in your signature you are running one with a 13" MBP. Especially in OSX I want stability.

As long as you don't use sleep mode, do not remove the Thunderbolt cable while the Mac is on (no hot plugging support), and have a widely used Nvidia Kepler or Maxwell card (Exception: I made bad experience with Gygabite GTX 750 Ti - iTunes did not play movies), the whole thing is as stable as with a card installed in the PCI-E-Slot of a classic Mac Pro.

No problems with these cards I tested: GTX 680, GTX 770, GT 630 (GK107 version), GTX Titan X. And soon GTX 960 to be tested.
 
Last edited:
As long as you don't use sleep mode, do not remove the Thunderbolt cable while the Mac is on (no hot plugging support), and have a widely used Nvidia card (I made bad experience with GTX 750 Ti), the whole thing is as stable as with an internal card in a Mac Pro.
If sleep mode isn't available, then what do you do if you want your mac to sleep? And wasn't AMD supported through the Apple drivers?
 
If sleep mode isn't available, then what do you do if you want your mac to sleep? And wasn't AMD supported through the Apple drivers?

Instead of sleep mode just shut down the Mac. :D

I have no eGPU experience with AMD cards. Almost all the guides in the net are for Nvidia cards.
[doublepost=1479416075][/doublepost]
There's no profit on making AMD eGPU because AMD graphics cards don't use PCIe packing that Nvidia does. The performance loss is too high.

I am indeed not very sure about that: https://translate.google.com/transl...base.de/2011-08/test-grafikkarten-mit-pcie/4/
 
Last edited:
By the way: The NVS 310 is an old Fermi generation card, that might not work with newest OS X versions.

What should my plan of attack be? I am running 10.12.

1) Try without any drivers but edit the kexts to include IOPCITunnelCompatible?

2) If that doesn't work try with the Web Drivers?

Or should I just try automate-eGPU.sh right off the bat?

Is there any benefit to me installing CUDA with this GPU?

Lastly what is the process to set the eGPU as main or as secondary and the monitors? I think I'd like the eGPU to be secondary as I want to be able to see the boot screens over my main monitor.
 
Last edited:
What should my plan of attack be? I am running 10.12.

1) Try without any drivers but edit the kexts to include IOPCITunnelCompatible?

2) If that doesn't work try with the Web Drivers?

Or should I just try automate-eGPU.sh right off the bat?

Is there any benefit to me installing CUDA with this GPU?

Lastly what is the process to set the eGPU as main or as secondary and the monitors? I think I'd like the eGPU to be secondary as I want to be able to see the boot screens over my main monitor.

I don't know if the card will work or not. So 1) Yes 2) Yes 3) If 1 & 2 do not work why not.

CUDA benefit: I doubt. I guess it's best the eGPU to be secondary unless you have a better graphics card.
 
I don't know if the card will work or not. So 1) Yes 2) Yes 3) If 1 & 2 do not work why not.

CUDA benefit: I doubt. I guess it's best the eGPU to be secondary unless you have a better graphics card.

Sorry for so many questions... I truly appreciate the help!

Do you have information on how to set the eGPU as main or as secondary? Any info on this? Will the main monitor always be connected to the main GPU?
 
Sorry for so many questions... I truly appreciate the help!

Do you have information on how to set the eGPU as main or as secondary? Any info on this? Will the main monitor always be connected to the main GPU?

Control panel -> monitors -> define main monitor (it's really not difficult).
 
Control panel -> monitors -> define main monitor (it's really not difficult).

Got it.

Well I received the card and realized I made a stupid mistake. This is for an x16 slot and the Helios TB1 only supports x1, x4 or x8 PCIe cards and PCIe 2.0! The TB2 version supports x16. Damn. Any recommendation on what I should get now?
 
Last edited:
Got it.

Well I received the card and realized I made a stupid mistake. This is for an x16 slot and the Helios TB1 only supports x1, x4 or x8 PCIe cards and PCIe 2.0! The TB2 version supports x16. Damn. Any recommendation on what I should get now?

The following card is all I have been able to find which meets my required specifications. It's a Kepler card. Am I crazy for trying this?

https://www.cdw.com/shop/products/NVIDIA-GeForce-GT730-graphics-card-GF-GT-730-2-GB/3868655.aspx

Update 1:
Based on the discussion on some hackintosh boards it looks like the 730 is hit or miss depending on OS X version. Argggg!

Update 2:
Reading that you can cut the x8 lane to fit an x16 card and the card will operate at x8. Going to try it... seems like a dumb idea but you never know.

Update 3:
Serious progress but not working... Here are some pictures:

Modified PCIe slot:
6Nl7nwF.jpg


Helios w/ card installed:
ntoMvJI.jpg


Card recognized!
vpZw5Sm.png


Hanging on first boot:
63mNLLr.jpg


Booting in verbose mode got it further but no dice:
T2xowxZ.jpg



The systems keeps hanging when the monitor is plugged in, otherwise it boots. I have another Mac Mini here that I am going to do a clean install on and see if that changes anything. Any other suggestions?
 
Last edited:
The following card is all I have been able to find which meets my required specifications. It's a Kepler card. Am I crazy for trying this?

https://www.cdw.com/shop/products/NVIDIA-GeForce-GT730-graphics-card-GF-GT-730-2-GB/3868655.aspx

Update 1:
Based on the discussion on some hackintosh boards it looks like the 730 is hit or miss depending on OS X version. Argggg!

Update 2:
Reading that you can cut the x8 lane to fit an x16 card and the card will operate at x8. Going to try it... seems like a dumb idea but you never know.

Update 3:
Serious progress but not working... Here are some pictures:

Modified PCIe slot:

Helios w/ card installed:

Card recognized!

Hanging on first boot:

Booting in verbose mode got it further but no dice:

The systems keeps hanging when the monitor is plugged in, otherwise it boots. I have another Mac Mini here that I am going to do a clean install on and see if that changes anything. Any other suggestions?

The card is well recognized by the system, but the driver hangs with the NVS 310, as I presumed. Have you tried both drivers, OS X native and Nvidia web drivers? Otherwise buy a GTX card.

The mentioned GT 730 card is a Fermi card (GF). Make sure the GTX 730 is a Kepler card (GK208). There are three different versions:
GT 730 B.png

Two of them draw less than 25 Watt (TDP), the Fermi version draws 49 Watt: https://en.wikipedia.org/wiki/GeForce_700_series
[doublepost=1479517434][/doublepost]

-------------------------------------------------------------------------------------------------------------------
Building a straightforward eGPU

If you have a Thunderbolt 1 Mac and don't want to hassle with power requirements: PE4C v 3.0, reaches 400 Mbit via the Sonnet Thunderbolt-Adapter http://www.mfactors.com/pe4c-v3-0-ec060a-pm060a-hp060a/

Supports up to 220 Watt with a Dell DA-2 AC adapter http://www.ebay.com/sch/i.html?_fro...apter.TRS0&_nkw=Dell+DA-2+AC+adapter&_sacat=0 ore more with an ATX PSU. When I receive the Dell AC adapter I will do some tests, especially if the 400 Mbit have enough bandwidth for a 4K monitor.

Some pics:
IMG_1030.JPG IMG_1029.JPG IMG_1026.JPG IMG_1027.JPGIMG_1025.JPG
 
Last edited:
The card is well recognized by the system, but the driver hangs with the NVS 310, as I presumed. Have you tried both drivers, OS X native and Nvidia web drivers? Otherwise buy a GTX card.

The mentioned GT 730 card is a Fermi card (GF). Make sure the GTX 730 is a Kepler card (GK208). There are three different versions:
View attachment 673346

Two of them draw less than 25 Watt (TDP), the Fermi version draws 49 Watt: https://en.wikipedia.org/wiki/GeForce_700_series

It works on 10.8.5! Going to try to do a fresh install of 10.12 and see what happens. Check it out:

Screen Shot 2016-11-18 at 11.03.01 PM.png
Screen Shot 2016-11-18 at 11.33.13 PM.png
IMG_5901.JPG
 
  • Like
Reactions: Synchro3
Afaik Apple/Nvidia dropped support for (some?) Fermi cards with El Capitan or Sierra. But Mountain Lion should work.

If you have only problems with newest OS X, get a Kepler or Maxwell card.

Be aware that a lot of the low cost cards, i.e. GT 630, GT 730 and lower, are available with different chips, older Fermi and newer Kepler, Nvidia messed it up. Avoid Fermi (GF), look for Kepler (GK). Read wikipedia GTX sites. No low cost Maxwell cards at present.

By the way: Support for VGA also dropped with El Capitan.
 
Last edited:
Afaik Apple/Nvidia dropped support for (some?) Fermi cards with El Capitan or Sierra. But Mountain Lion should work.

If you have only problems with newest OS X, get a Kepler or Maxwell card.

Be aware that a lot of the low cost cards, i.e. GT 630, GT 730 and lower, are available with different chips, older Fermi and newer Kepler, Nvidia messed it up. Avoid Fermi (GF), look for Kepler (GK). Read wikipedia GTX sites. No low cost Maxwell cards at present.

By the way: Support for VGA also dropped with El Capitan.

The real problem is that I can't find a Kepler 730 with Displayport... They all have DVI, HDMI, VGA. All other cards are higher wattage and double width. I wonder how many watts can actually be pulled from that card slot as it reads 'minimum 25W' but the double width is still a problem if I use the current enclosure.

There are a bunch more NVDA and GeForce related kexts in the extensions folder... Is it worth a shot to start adding the <key>IOPCITunnelCompatible</key> to them?
 
Last edited:
Hi guys,

Forgive my noob questions, but I've been following the eGPU concept off/on for a while now, and I might be ready to take the plunge. Yesterday, I got the LG 27UD88 monitor (brilliant) that works at 4K 60Hz with 2 of 3 of my Macs. Late 2013 Mac Pro and Early 2016 MacBook do fine via DP and USB-C respectively.

That leaves me with my Late 2014 Mac mini, only displaying 4K @ 30Hz, as expected. It doesn't look terrible, but I think it's giving me a bit of headache/eye strain at 30Hz.

So, here is my ask - can I build a Thunderbolt 2-fed eGPU outputting HDMI 2.0 4K @ 60Hz with the following requirements?
  1. Low power - my mini stays on 24/7 (server), but I only use the monitor a few hours of the day. Looking for a low power solution if it needs to run 24/7, or something I can easily switch off and on.
  2. Low cost - prefer under $200 for the total solution
  3. No kext/macOS mods - I don't care if I don't have a boot screen, but I don't want to be constantly fiddling with drivers, etc. I always keep my machines on the newest public OS revision. So if the eGPUs are prone to breaking with every 10.12.x update, it's probably not a solution for me.

Thanks for any insight.
 
The real problem is that I can't find a Kepler 730 with Displayport... They all have DVI, HDMI, VGA. All other cards are higher wattage and double width. I wonder how many watts can actually be pulled from that card slot as it reads 'minimum 25W' but the double width is still a problem if I use the current enclosure.

There are a bunch more NVDA and GeForce related kexts in the extensions folder... Is it worth a shot to start adding the <key>IOPCITunnelCompatible</key> to them?

Additional instructions for OS X 10.10 - 10.12 in post -> https://forums.macrumors.com/threads/building-external-gpus-on-mac-egpu.1893792/page-2#post-22013272

Seems like GTX low end cards with DisplayPort are really rare, especially up to 25 Watt.

Do not draw more than 25 Watt over the Helios PCI-E (without modifications)! You will fry the Helios and the graphics card. - > https://forums.macrumors.com/threads/building-external-gpus-on-mac-egpu.1893792/page-5#post-22940819
 
Last edited:
Additional instructions for OS X 10.10 - 10.12 in post -> https://forums.macrumors.com/threads/building-external-gpus-on-mac-egpu.1893792/page-2#post-22013272

Seems like GTX low end cards with DisplayPort are really rare, especially up to 25 Watt.

Do not draw more than 25 Watt over the Helios PCI-E (without modifications)! You will fry the Helios and the graphics card. - > https://forums.macrumors.com/threads/building-external-gpus-on-mac-egpu.1893792/page-5#post-22940819

This looks like the animal right here: http://www8.hp.com/uk/en/products/oas/product-detail.html?oid=5285071

It has to be the Kepler version as it is the only 630 with PCIe 3.0.

Screen Shot 2016-11-19 at 10.55.01 AM.png


Problem is its TPD is 50W... Are you absolutely sure I need a powered riser for this? Even if I am only using 1 port out of the 3? Why would OWC use the word 'Minimum 25W'? The power supply they included is 12V 5A which produces 60W total.

If I get the powered riser can I use something like this which would supply an additional 60W rather then a PC power supply? https://www.amazon.com/Power-supply-4PIN-molex-connector/dp/B00OVC92QQ/
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Yes, this is exactly the HP GT 630 with Kepler GK107 I have. Works without problems in all OS X. Faster than the internal Intel HD 3000 or 4000 iGPU. Supports two 4K and even one 5K monitor (via Dual DisplayPort cables).

You could try without powered riser cable , directly connecting the 60 Watt AC adapter to the Helios. ATX PSU is not obligatory. Maybe the AC adapter should have a little more Watt than the card.

But don't complain if you fry the Helios and the card. I fried a Thundertek/PX, but I must admit that I don't know the Helios. Workaround with powered riser cable is always the safest way.
 
Problem is its TPD is 50W... Are you absolutely sure I need a powered riser for this? Even if I am only using 1 port out of the 3? Why would OWC use the word 'Minimum 25W'? The power supply they included is 12V 5A which produces 60W total.

30W overhead has to be provided for powered Thunderbolt pass through. Some enclosures (ex Akitio Thunder2) make at least part of that power unofficially available to x16 cards when thunderbolt chaining isn't in use. I have no clue if that is the case with your enclosure. Maybe the Helios uses the same circuitry. Maybe the 50W rumor if not chaining thunderbolt devices is correct. Maybe the Synchro's fried card was a fluke.

We aren't absolutely sure that a powered riser is required, its just the safe bet given the number of maybes.
 
30W overhead has to be provided for powered Thunderbolt pass through. Some enclosures (ex Akitio Thunder2) make at least part of that power unofficially available to x16 cards when thunderbolt chaining isn't in use. I have no clue if that is the case with your enclosure. Maybe the Helios uses the same circuitry. Maybe the 50W rumor if not chaining thunderbolt devices is correct. Maybe the Synchro's fried card was a fluke.

We aren't absolutely sure that a powered riser is required, its just the safe bet given the number of maybes.

Two questions:

1) Im going to give it a shot with the powered riser... Can you answer the question of whether or not a power supply like this with the molex connector connected to the riser will work: https://www.amazon.com/Power-supply-4PIN-molex-connector/dp/B00OVC92QQ/ I can't see why not but maybe I'm missing something. I don't want to use an ATX PSU.

2) You are talking about something like this for the riser to connect the GPU (I need to convert 8x to x16 even though I already melted down the connector on the main board)? The link to your eBay listing earlier in the thread wasn't working.

pcie8xto16x.jpg





I am going to attempt to fit this all into the same enclosure as well. The PCI x8 board connects to the main board by a x4 connector. I am going to buy this riser and try to flip the board that connects to the card upside down on top of standoffs so I have space to install the GPU correctly with enough space to connect its riser.

41upWYb.jpg
From this:

IMG_5905.JPG

To This:

IMG_5906.JPG


I'll keep you posted on the progress!
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
This is the right powered riser cable (with capacitor). It gives power to both sides, the Helios board and the graphics card.
 
This is the right powered riser cable (with capacitor). It gives power to both sides, the Helios board and the graphics card.

Looks like the Sonnet Echo Express SE is exactly the same as the Helios TB1... The boards and enclosure are identical (besides enclosure color)!

You didn't answer if the following P/S can be used: https://www.amazon.com/110v-Power-Supply-Molex-5000mA/dp/B002TJNDU4. It will only provide 60W (12V 5A = 60W) but if I use this additional power supply with the riser and the original P/S (12V 5A = 60W) will it combine to 120W?
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
I would buy a bigger PSU than the graphics Card (120 Watt), to have extra reserve for power peaks, but yes, this can be used.

Connect it to the riser cable, the riser should power both, the card and the Helios board. No Need for second AC adapter, if it works like my Thundertek/PX.

Google if the Sonnet Echo Express can provide 75 Watt via PCI-E port, maybe you don't need the riser cable.
 
Hi, has anyone had any joy with any of the following on any OSX platform: GTX 1080, GTX 1070, or GTX 1060's ??
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.