AMD Launches Project to Ease HPC Development on GPUs

[Mumps]

http://www.eweek.com/developer/amd-launches-project-to-ease-hpc-development-on-gpus.html

So, I wonder if this will translate to moa-betta GPU apps for BOINC. :)


AMD is building on its work around heterogeneous computing to develop tools that officials say will increase the performance and efficiency of high-performance computing (HPC) systems that can leverage both CPUs and discrete GPUs.

[Shandia]

http://www.eweek.com/developer/amd-launches-project-to-ease-hpc-development-on-gpus.html

So, I wonder if this will translate to moa-betta GPU apps for BOINC. :)

I hoped it goes into good amd gpu apps. :)) Einstein is a good reminder that the playing field is not level towards amd, I know, if it's good on amd, it's not that great for nvidia and vice versa. I just wish Einstein was a little more fair......or I had a few nvidia cards sitting on the sideline to be swapped out.

[FourOh]

I hoped it goes into good amd gpu apps. :)) Einstein is a good reminder that the playing field is not level towards amd, I know, if it's good on amd, it's not that great for nvidia and vice versa. I just wish Einstein was a little more fair......or I had a few nvidia cards sitting on the sideline to be swapped out.

What GPUs are you running on Einstein? Nvidia has an advantage right now because Beta (CUDA only) work units finish faster than regular WUs, but I see only a marginal benefit in credits/watt on current generation Nvidia cards versus Tahiti AMD cards. The only real outlier in my fleet is the 750 Ti, which is about 25% more efficient than a 280X. This table is based on non-beta work units:


GPU
WU Avg Completion Time
WU Value
Credits/Day
TDP
Credit/Day per Watt


GTX 750Ti x2
16780
4400
45311
60
755


GTX 960 x3
16900
4400
67484
120
562


GTX 970 x2
8260
4400
92048
145
635


7970 x4
10700
4400
142116
250
568


R9 280X x4
9960
4400
152675
250
611

[Shandia]

What GPUs are you running on Einstein? Nvidia has an advantage right now because Beta (CUDA only) work units finish faster than regular WUs, but I see only a marginal benefit in credits/watt on current generation Nvidia cards versus Tahiti AMD cards. The only real outlier in my fleet is the 750 Ti, which is about 25% more efficient than a 280X. This table is based on non-beta work units:


GPU
WU Avg Completion Time
WU Value
Credits/Day
TDP
Credit/Day per Watt


GTX 750Ti x2
16780
4400
45311
60
755


GTX 960 x3
16900
4400
67484
120
562


GTX 970 x2
8260
4400
92048
145
635


7970 x4
10700
4400
142116
250
568


R9 280X x4
9960
4400
152675
250
611



I'm running my hd 5850. I checked the hd5770 today and it's fine. The hd5850 is at 30 hours and 83% complete. It seems I have a big problem going on here. Temps are at 60*c and output is at 0%. WTF??? I'd be thinking its fried butr I'm using it with no problems..l....except it's barely crunching. What are the best drivers for it? I really don't want to start messing around with a bunch of different ones and get it more confused than it is now.

I did have the 5850 in bipolar working something else. When I get the cache cleared, I'll start bunkering there too. It's always good to get a second opinion on drivers from an identical card. I did do the round about on my main due to a beta I was in a few weeks ago. (game) It's possible the newest drivers did something or just didn't take.

EDIT -- WAHAHAHAHAHAHA......WAHAHAHAHAHA. Got it figured out. That little comment I made not too long ago about a computer relying on a gpu far more than anything......yeah. My VM was sucking the life right out of the 5850 despite me not doing anything with the gpu on that front. Ok. The VM work is getting finished up today and going back into the nether for a 3while until I need it for a cpu task again.

[Mumps]

With VirtualBox, it's important that you install the Extension Pack on both the host and guest O/S's. That helps a lot on the GPU utilization for a VM. (And even things like USB benefit from that more efficient code that adds.)