Milky Way@H ATI vs Nvidia

[mindzx]

gpu-z reads incorrect load on gpu, using PC Meter + GPU gadget ,I see easily 99% usage with 2 WU's running at once.

 

[Sunny129]

gpu-z reads incorrect load on gpu, using PC Meter + GPU gadget ,I see easily 99% usage with 2 WU's running at once.

that explains alot. strange that you were getting a bad GPU load reading w/ GPU-Z - even though i use MSI Afterburner to monitor my GPU vitals, i can double check it w/ GPU-Z, which seems to be working just fine on my machine. at any rate, i'm glad you found that the error was in the GPU-Z GPU load reading.

 

[StitchExperimen]

I've found that the video playback was hampered by the lack of cpu power not GPU when loading multiple instances and AMD GPUs needing cores/threads.

 

[Sunny129]

I've found that the video playback was hampered by the lack of cpu power not GPU when loading multiple instances and AMD GPUs needing cores/threads.

well its going to depend on a number of factors, which is why i said choppy video playback is a distinct possibility, but not necessarily a certainty, for the OP. for example, we know that Milkyway@Home separation tasks tend to put a full load on the GPU, and yet those who crunch MW@H and run their displays on the same GPU experience little to no GUI lag provided they aren't multitasking to the point of putting the CPU and/or GPU under substantial additional load. Einstein@Home on the other hand isn't nearly as cooperative in this respect. a single Einstein@Home BRP4 task might only load my GPU to 50-60%, which is a far cry from the full load induced by a single MW@H separation task...and yet the GUI becomes noticeably laggy if i'm also running the display on that same GPU. i'm certainly not ruling out the possibility that CPU load contributes to this - after all, a single MW@H GPU task consumes substantially less CPU resources than an Einstein@Home GPU task. but all my machines are 6-core machines...and knowing that BOINC will take advantage of as many cores as one can throw at it, i have to assume that CPU load isn't the main contributing factor to the GUI lag i experience when crunching E@H on a display GPU. its also going to depend on an individual machine's hardware combination.

 

[Assimilator1]

Old thread I know , had totally forgotten about it!.

For the record the HD 4830 would of owned the 260 c216 in MW, from ATs article http://www.anandtech.com/show/2549/5 , the double precision/FP 64 multiplier of the GT 200 series is 1/8th.

So their DP powers are :-
HD 4830, 512 MB = 147.2 GFLOPs
GTX 260 c216, 900 MB = 100.6 GFLOPs

My son's got my old HD 4830 now, & although performance wise we'd both be better off swapping cards (me for MW, him for games) it would increase his power usage ( http://www.anandtech.com/show/2647/10 ) which wouldn't go down well with the bill payer, not him!