precision01

Hi, @Dschijn. Thanks for sharing your project. Your power tests shows that apparently Akitio standard PSU is probably enough for Geforce GTX 750 ti. And your GPU benchmarks confirms that we have to expect an average loss of 20% of performance for the eGPU, if compared with the card directly connected to the PCIe slot. I will do the tests with a external PSU + powered riser to check if I have the same results and I hope to share them here ASAP. - - - Updated - - - Hi, @Dschijn. I don't really know. I will repeat the tests and pay attention to that.

Hi, @johndshea, thanks for replying. I pretty much agree with you about the 4-lace PCIe connection of Akitio Thunder 2 pie box. Even if that's the case, I believe that we can be satisfied with the fact that Akitio setup cost much less than other manufacturers solutions. Since I live in Brazil, very specific computer parts, like this eGPU setup had to be imported from US, but finding common parts like a powered riser is easy. I can try to use this powered raiser thing, but I must admit that I still didn't understand very well how does this powered riser connect to the PSU and o the GPU. Since most Geforce GTX 750 ti doesn't have any 6 pin power connector (mine is from Zotak), I believe that a powered riser is the solution, right? I would like to know how much Watt power Akitio is really feeding the GPU. Is it 25w (I doubt), 60w or maybe more? Hope someone has equipment to do a test to confirm that. About stability, when quitting iRacing (in Windows 8.1) I had a few black screens due to nvidia driver fail (windows said that!), but after 2 or 3 seconds the screen was back again without any apparent issue. In OSX 10.10.1, I don't remember to have had issues, but I didn't test too much. When I first read your post, I was already almost sure that this setup is the "Golden standard" for low budget eGPU. Both on notebook check and toms hardware you can find benchmarks to confirm that this Akitio + GTX 750 ti setup is 3x to 4x faster than discreet 750m. By the way, integrated intel Iris PRO 5200 is usually 5-8% faster than Geforce 750M, which makes me wonder what's the point in Apple putting 2 GPU's with the same performance.

External screen. I didn't had time to try to make the egpu setup work on internal screen.

Hello. First I would like to thank johndshea, the author of this thread, and everyone else for the help building this setup! I've just done a setup almost identical to johndshea: - Macbook PRO Retina 13" (late 2013) - OSX Yosemite 10.10.1 - Windows 8.1 via bootcamp - Akitio Thunder 2 PCI-e Box - Zotac Geforce GTX 750 ti 2gb - Installed latest nvidia drivers (download from nvidia on jan/23/2015 Important! I´m not using any external power, just the standard Akitio Thunder 2 package. Results --------- I could successfully do this eGPU setup both on OSX Yosemite 10.10.1 and Windows 8.1. Just a few benchmark numbers: 3DMark (2013) - Fire Strike Standard Graphics 1920x1080 (Windows 8.1) I've ran 3dMark Firestrike Standard and a got a 3.554 score. This is a little disappointing to me, since on a similar spec PC this GTX 750 ti can do a 4.500 score. So it is a loss of 20% in performance, where I was expecting a 10% loss. Other tests ------------- I´ve build this setup mainly for GPU rendering purposes on 3D Softwares using Octane Renderer, but also for gaming (most of the time iRacing). iRacing What I experienced in general is a satisfactory result. As in 3dmark, the game is running, like, 20% slower than when was connected to a PICe slot on a similar specs PC. In general the GPU is temperature is low most of the time. But sometimes (after 20-25 minutes racing), I´m getting significant frame rate drops, where it drops from 60fps (I´ve mannualy set this frame rate limit on iRacing) to 35 - 40 fps without any reason. I don´t know if this is related to drivers issues, power issues, temperature issues (althought temperature seems normal), or something else. So, I must ask: Am I missing something important here? Would I have a more stable experience with external power? Can I still improve that 3dmark score from 3500 to at least 4000? Can someone ( johndshea?) run 3dmark to compare with my 3500 score?