I noticed in your signature you are running a 680 eGPU off your mini. I am thinking of doing the same. Was that hard/expensive to set up and are you happy with it?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
GTX 980 SuperClocked or GTX 680, Any guide on how to flash it?
- Thread starter troy14
- Start date
- Sort by reaction score
Last edited:
Wow - Thanks for the great info! Looks like some hard work ahead getting things together. Read somewhere else that MVC and Netkas might be putting an eGPU together for sale as well, I wonder how thats going!
I use my cards principally for 3d work (Mac Rhino for modelling and Blender for its cycles engine to render).
I've started with a real GTX 285, and while it was ok for open gl in rhino it had no current cuda support so blender renders were CPU only. Horribly slow.
So I got a Mac flashed 4gb 680, and while it supported gpu compute for blender it wasn't all that much faster (maybe 2x). It also had issues, artifact info and other problems. So I returned that and just went the stock card route with nvidia drivers. In looking closely at the various benchmarks people reported with the 970's and 980's for cuda rendering there really wasn't much of a difference (maybe 15% faster with the 980), which didn't justify the almost double price in the cards. Once I figured out I could run 2 970's with out mods or external power supplies just buy getting a pair of $6 power splitter cables I went that route, and couldn't be happier. I pcked up a pair of EVGA 4gb 970's (an acx 2.0 Sc and an ACX 2.0 FTW), and for just a tad more than a 980 I have a render solution that's almost twice as fast as a ingle 980, and is considerably faster than a titan X for well less than half the price of that.
I have a gt120 in slot 4 that I use to drive a pair of 1920x1200 monitors and as it's a real efi card I still get boot screens and all that. I have occasioanlly used the video outs off the 970's and it works fine as well although I tend to stick to the 120 for rendering performance reasons.
As others have mentioned, OS updates will break the drivers for a few days, but the simple solution is to just turn off auto updates, wait for the update to come out, then wait a few days until nvidia updates the drivers and THEN do the updates.
Since then I've updated the procs as well to 6 core 3.46's and am extremely happy with the bang for the buck. I've also built a second cMP with another pair of 970's in it as a render node.
Between the two boxes I have a render solution that's well closing on 3x as fast as a TitanX for less than the cost of the TitanX .
Something to consider...
I've started with a real GTX 285, and while it was ok for open gl in rhino it had no current cuda support so blender renders were CPU only. Horribly slow.
So I got a Mac flashed 4gb 680, and while it supported gpu compute for blender it wasn't all that much faster (maybe 2x). It also had issues, artifact info and other problems. So I returned that and just went the stock card route with nvidia drivers. In looking closely at the various benchmarks people reported with the 970's and 980's for cuda rendering there really wasn't much of a difference (maybe 15% faster with the 980), which didn't justify the almost double price in the cards. Once I figured out I could run 2 970's with out mods or external power supplies just buy getting a pair of $6 power splitter cables I went that route, and couldn't be happier. I pcked up a pair of EVGA 4gb 970's (an acx 2.0 Sc and an ACX 2.0 FTW), and for just a tad more than a 980 I have a render solution that's almost twice as fast as a ingle 980, and is considerably faster than a titan X for well less than half the price of that.
I have a gt120 in slot 4 that I use to drive a pair of 1920x1200 monitors and as it's a real efi card I still get boot screens and all that. I have occasioanlly used the video outs off the 970's and it works fine as well although I tend to stick to the 120 for rendering performance reasons.
As others have mentioned, OS updates will break the drivers for a few days, but the simple solution is to just turn off auto updates, wait for the update to come out, then wait a few days until nvidia updates the drivers and THEN do the updates.
Since then I've updated the procs as well to 6 core 3.46's and am extremely happy with the bang for the buck. I've also built a second cMP with another pair of 970's in it as a render node.
Between the two boxes I have a render solution that's well closing on 3x as fast as a TitanX for less than the cost of the TitanX .
Something to consider...
FWIW I have an Asus Strix OC GTX970 which only has 1x8 pin PCIe connector. The fan doesnt even come on until the gpu hits 69C. Something to consider if you have a Mac Pro as well, just get 2x 6-8 pin adapters and you are good to go with 2 GTX970s overclocked. I havent tried this yet but Im sure it would work OK if you test the wattage output first..
Last edited: