Dont want to reinvent the wheel testing, if the answer is already out there somewhere .... definitive not anecdotal.

Is PPS Sieve GPU app sensitive to GPU memory bandwidth?

For example, with Collatz (needs memory high) the datasets going back and forth CPU-GPU are large and need all the GPU memory bandwidth you can give it, there is a very noticeable difference in output if GPU memory is set at (say) 500 compared to set at (say) 1250 or more.

With MilkyWay(ideally memory low), its the opposite. The datasets are small, and GPU memory bandwidth setting is irrelevant - it can go as low as 180 with no discernable difference in output. Running MW at (say) 1150 or more, is a waste of power, and only produces a more efficient space heater.

So for PPS GPU Sieve - memory high or low?

Regards
Zy

IIRC the current CUDA GPU app for PPS Sieve is not sensitive to memory bandwidth.

I run my GTX275 OC'd at 703 with memory DC'd to 799.
nvidia-smi -q returns ---
Utilization
Gpu : 99%
Memory : 1%

I do not know anything about the version of the for ATI cards.
____________
There's someone in our head but it's not us.

Okie Doke - Low memory it is. I've already had a go for a couple of days, but want to properly settle down to an extended session for a while as I've not really crunched PrimeGrid in the past as such, and its time I gave it a whirl - accepting the "hit" being AMD card.

Will not be that much difference as such to the dataset size shuttling around between AMD & NVidia for this purpose in terms of low or high memory

First few are through, seems ok for an AMD card here.

5970 @ 840GPU & 500 memory on a 1090T @3.7Ghz, through in either side of 2000secs per GPU. Temperatures way down, not an issue - nice to see. Might crank it up a bit more later on if it stays that way over the next 24 hrs.

Any AMD GPU Players out there to whom that rings incorrect/on the low side for the card - grateful you shout, hard to get comparitors with Primegrid being primarily a NVidia house, to know if I've got it set up in the right ball park for the card. Not going to push it to the edge though .... I like to sleep nights :)

Regards
Zy