johndshea

@goalque, thanks! I can confirm that both my setups support wake from sleep without kernel panic using the Nvidia cards. I can also work with the lid closed, although OSX doesn't disable the internal display properly - rather, it simply turns it off but still displays it in System preferences and you can still drag items onto the display as if it was turned on. Also, I forgot to note: my eGPU is not hot-pluggable on either setup on OSX or in Windows. I have to reboot to plug it in whenever I want to connect or disconnect it. One issue I've been running into is that my laptops always end up suffering from intermittent crashes starting 3-5 days into my eGPU usage. The crashes don't seem to happen under heavy usage, and appear to be pretty random. Also they don't occur during the first couple of days, only once I've been running the eGPU for a week or so (periodically). Very strange. Anyone have any ideas as to why this might be happening?

It's a Mid-2014 13" Macbook Pro with Retina Display with an Intel Iris GPU. I did the exact same setup that I did with my Early 2014 Macbook Air (Nvidia Web Drivers, AKiTiO unit w. 500W ATX PSU, MSI Mini-ITX GTX760), and it worked equally well. Oddly enough, the performance difference between the TB1 MBA and the TB2 MBP was negligible. But they both were quite stable in OSX Yosemite with the eGPU.

@MVC yes, my Early 2014 MBA is TB1. I also have a Mid-2014 MBP which is TB2. Both are working with my GTX760/AKiTiO eGPU.

@MVC you are very kind. I'm not sure how to rename the thread but I'll give it a try. Thanks! @Tech Inferno Fan I believe you created this thread, are you the only person who can rename it?

Thanks, @goalque, I wasn't aware of the different types of risers. That explains it! I appreciate the electrical help, I'm not very experienced with that. Update: I received my MSI mini-ITX GTX 760 today and am running it now on the AKiTiO box with my 500w ATX PSU providing power through the PCIe 8-pin connector on the 760 as we speak. It appears to be completely stable in OSX Yosemite (so far). It runs an external display perfectly both on my 11'' Early 2014 Macbook Air and on my 13'' Mid 2014 Macbook Pro (Retina). Interestingly, benchmarks on each are pretty close to being the same (see below). It appears that there is little-to-no performance difference between a Thunderbolt 2 connection (on the Macbook Pro) and a Thunderbolt 1 connection (on the Macbook Air) when using an eGPU with the AKiTiO. My speculation is that this is because the 4 PCIe lanes the AKiTiO provides already max out at 5Gbps themselves, well within the throughput range of both TB1 and TB2. In summary: 1) Early 2014 11'' Macbook Air (TB1) with GTX750Ti running purely on the AKiTiO's own power (no riser): works under both OSX and Windows, but unstable under load, scores roughly 440 in Unigine Heaven benchmark "Extreme" @ 1080p. 2) Mid 2014 13'' Macbook Pro (TB2) with GTX750Ti running purely on the AKiTiO's own power (no riser): works under both OSX and Windows, but unstable under load, scores roughly 440 in Unigine Heaven benchmark "Extreme" @ 1080p. 3) Early 2014 11'' Macbook Air (TB1) with GTX750Ti running on the AKiTiO with a 500W ATX PSU via a powered riser: works under both OSX and Windows, stable under load, scores roughly 440 in Unigine Heaven benchmark "Extreme" @ 1080p. 4) Mid 2014 13'' Macbook Pro (TB2) with GTX750Ti running on the AKiTiO with a 500W ATX PSU via a powered riser: works under both OSX and Windows, stable under load, scores roughly 440 in Unigine Heaven benchmark "Extreme" @ 1080p. 5) Early 2014 11'' Macbook Air (TB1) with Mini-ITX GTX760 running on the AKiTiO with a 500W ATX PSU via 8-pin PCIe connectors (no riser): works under both OSX and Windows, stable under load, scores roughly 660 in Unigine Heaven benchmark "Extreme" @ 1080p. 6) Mid 2014 13'' Macbook Pro (TB2) with Mini-ITX GTX760 running on the AKiTiO with a 500W ATX PSU via 8-pin PCIe connectors (no riser): works under both OSX and Windows, stable under load, scores roughly 680 in Unigine Heaven benchmark "Extreme" @ 1080p. Update: The Mini-ITX MSI/Nvidia GTX760 appears to run properly as an eGPU even when using the stock Mac OS X graphics driver in Yosemite! I'm not entirely sure why this is - perhaps because Yosemite supports the GTX760 natively? I'm surprised it will allow it to run in an eGPU though. Can anyone else with a GTX760 and an AKiTiO confirm this?

Hi @MVC, I didn't actually do anything particularly special other than the process outlined by @nesone here. Basically, you have to do some modifications to the NVIDIA web drivers to get them to install, program your NVRAM so it accepts unsigned drivers, modify some kext files, and you're all set (see @nesone 's instructions, he lays it all out very well). It took a little bit of trial and error and creativity, but I managed to get an AKiTiO/GTX750Ti combo working both on my TB1 MBA and my TB2 MBP. For some reason, the MBA is able to run the AKiTiO/GTX750Ti without any additional power such as a powered riser, though it is not stable. The MBP absolutely requires the AKiTiO/GTX750Ti to have additional power to even boot properly. I don't know whether this is related to any difference between TB1 and TB2.

Another question - I was just looking at the box for the AKiTiO and it appears to state that it has a "4 lane PCI Express 2.0 compliant interface at 5.0Gbps" (see picture below). Does this mean that a GPU attached to the AKiTiO board is only going to get 5Gbps of throughput regardless of the 16Gbps capacity of Thunderbolt 2? In other words, does this mean that there's no functional difference between using the AKiTiO with a TB1 port and a TB2 port? Could it also mean that you can daisy-chain up to three AKiTiO units together without maxing out TB2's bandwidth? For me, for example, benchmarking 13'' MBP (with TB2) against my base model 11'' MBA (with TB1) produces identical results.

Hi All! Just wanted to share a piece of information - I recently (last week) purchased an AKiTiO Thunder2 box from Amazon. Before Amazon, I went to B&H (in-person) and was informed that AKiTiO had discontinued the Thunder2 PCIe box and that B&H would be receiving a new model in the next one to two weeks. Compare: Akitio Thunder2 PCIe Box Expansion Chassis AK-T2PC-TIA B&H Photo (listed as "Discontinued") To: Akitio Thunder2 PCIe Expansion Chassis AK-T2PC-TIA-AKTU B&H (listed as "Usually ships in 7-10 business days") The old version seems to have the manufacturer model number AK-T2PC-TIA and the new version carries the number AK-T2PC-TIA-AKTU. The AKiTiO website also lists AK-T2PC-TIA-AKTU as the current model number: AKiTiO Thunder2 PCIe Box | PCI Express Expansion Box with Thunderbolt 2 Interface | AKiTiO. Does anyone know what the change was that AKiTiO made? Does anyone have the new version yet? I'm pretty sure the one I got from Amazon is the old version, as it carries the model number "T2PC-TIA."

Very odd. Thanks, [MENTION=31783]Tim Weri[/MENTION]

Gotcha. Thanks!

I can confirm that my AKiTiO unit also does not receive any power from a powered riser, it only turns on when plugged in via its own power source. I've tried multiple risers.

Thanks, @Dschijn! Do you know of any mac software to measure power draw? I don't. I'm less concerned with damaging the AKiTiO's power adapter (as those are easily replaceable) and more for the board itself. If the board itself is probably fine, I feel very relieved. I have a Mini-ITX GTX 760 coming today, so I should be able to test that with an ATX PSU and report as well. So I'll be able to compare all six combinations of a MBP (TB2) or MBA (TB1) with a GTX750Ti (both with and without powered riser) or a GTX 760 with external ATX power.

Here's a question for @Tech Inferno Fan or @goalque or one of the other forum members with way more electrical engineering expertise - is it possible to damage the AKiTiO unit by running the 750Ti off of the AKiTiO's power alone (i.e. without a powered riser or any other power source other than that provided by the AKiTiO board alone). I would think that the 750Ti would simply be power-starved and unstable and wouldn't be capable of drawing excessive voltage through the AKiTiO, but something tells me it might not be that simple. Many thanks, guys! I appreciate the help.

I just tried with the new riser and had the same result - the AKiTiO still needs its own power supply. I also tested the AKiTiO/gtx750Ti combo with a brand-new 13'' Macbook Pro and it got the exact same benchmarks as the MBA (which is odd - you'd expect that the jump from Thunderbolt 1 to Thunderbolt 2 would result in a performance jump, not to mention the spec jump from the base MBA to a MBP with twice the RAM and a faster CPU). Lately, I've tried booting the AKiTiO with the MBP and its own internal power supply and it won't do it at all. I am hoping that I haven't permanently damaged the AKiTiO, which was my primary worry. I have a GTX760 coming tomorrow and will test with that. Very confused.

I did, and the AKiTiO will not run with only the powered riser. So that could mean my riser is defective. I should have another one tomorrow to test.

Hi @salut, Heat doesn't appear to be an issue. The 750ti runs very cool, usually even cool to the touch even when running, for example, Dragon Age Inquisition at Ultra settings, 1080p (which runs off the Frostbite III engine, very demanding). However, my machine is unstable lately - I've been getting periodic crashes both in Windows and in OS X. I'm 90% sure the eGPU is to blame, but the crashes don't appear to follow any discernible pattern. For example, they seem to be no more likely to occur when I'm stressing the GPU (gaming, benchmarking, etc) than any other time, and they don't appear to be more likely to occur if the computer has been running for some time. As far as I can tell, they are random. The obvious suspect is the power supply issue on the AKiTiO. I'm currently testing a hunch that the cause may have been the fact that I daisy-chained a Thunderbolt ethernet adapter with the AKiTiO and that the ethernet adapter might have been the culprit creating unpredictable power requirements on the AKiTiO's PSU unrelated to the power draw of the 750Ti. It sounds pretty out-there, but I'm testing that theory and we'll see how it goes. Update: removing the daisy-chained thunderbolt ethernet adapter didn't seem to improve things. Neither did running the 750Ti with a 500w power supply and a powered riser, so either my riser is busted or the power draw isn't the culprit. Something else is awry here, and I'm not sure what it is.

Hello all! I've been a lurker on this thread for several months now and decided to finally bite the bullet and do my first eGPU build with the AKiTiO. It's up and running beautifully in Yosemite after following @entzoe's instructions here, after a few hiccups (e.g., apparently the driver installation and kext editing must be done in a very specific order). Right now, I'm running the 750Ti off the AKiTiO's power alone, with the enclosure on too. And it seems to be running fine! This is confusing to me because my understanding is that the AKiTiO only provides 25W to the card and even the 750Ti requires 60W. So I'm concerned that a) I'm damaging the AKiTiO by exceeding its power envelope or my 750Ti won't run properly because it's power-starved, or both. Update: running the 750Ti on the AKiTiO's 60W PSU alone proved to be moderately unstable. I eventually shifted to using a 500W ATX PSU + powered riser with the 750Ti, which was more stable, then to running a Mini-ITX GTX760 and later a Mini-ITX GTX 970 with the AKiTiO, both of which were even more stable. I'm well behind the learning curve on electrical engineering compared to most people on this thread. However, if my setup really is working well, it's possible that the AKiTiO is capable of running a midrange Maxwell GPU without any other hardware needed. OpenGL benchmark with the GTX750Ti running only on the AiKiTO's own power supply (not bad for a base model 11'' MBA!): OpenGL benchmark with the GTX750Ti running on an EVGA 500w ATX PSU through a powered riser (as you can see, barely any difference): For comparison - OpenGL benchmark for the MBA's original integrated Intel HD5000 GPU: For good measure: here's the same benchmark replacing my 750Ti with a GTX 970: