It depends on cards used. Different cards have different memory onboard (DDR5 or DDR3 or even DDR2). So the bandwidth of the memory and chip makes this difference. That is why even modern cards sometimes can not show us increase in calculation speed, and that is why one card (even with older chip) may cost twice as a new one. So, don't hestitate to post not only then name of your cards chip, but the model number and manufacturer, type of VRAM and width of a memory bus.
That`s strange, but if you want I can offer a couple ideas how to solve the problem:
Little check:
1. Left running on the box only enshtein
2. In app make 0.5 CPU + 1 GPU and 2 threads
3. Change process priority to *realtime*
4. The most important things are not to load CPU with the other projects and process priority
It takes 154 minutes when I have 3 WU's running at the same time.
It takes 104 minutes when I have 2 WU's running. With or without the cpu's doing Malaria.
It takes 53 minutes when 1 have WU running.
So there is hardly any gain in time on my system with a EVGA GTX460 1GB FTW.
And here only .20 cpu is used no matter what I do.
Have assigned different cpu's (real time) to every WU.
And I want my systems to produce as much as possible so besides the GPU running E@H my quads run Malarianet with their spare cycles.
@ Vikk : Afterburner gave a max. load of 48% running two or three WU's.
It takes 154 minutes when I have 3 WU's running at the same time.
It takes 104 minutes when I have 2 WU's running. With or without the cpu's doing Malaria.
It takes 53 minutes when 1 have WU running.
So there is hardly any gain in time on my system with a EVGA GTX460 1GB FTW.
And here only .20 cpu is used no matter what I do.
Have assigned different cpu's (real time) to every WU.
And I want my systems to produce as much as possible so besides the GPU running E@H my quads run Malarianet with their spare cycles.
------------------------------
Can you check GPU load on 1/2/3 threads? (With afterburner or gpu-z)
It depends on cards used.
)
It depends on cards used. Different cards have different memory onboard (DDR5 or DDR3 or even DDR2). So the bandwidth of the memory and chip makes this difference. That is why even modern cards sometimes can not show us increase in calculation speed, and that is why one card (even with older chip) may cost twice as a new one. So, don't hestitate to post not only then name of your cards chip, but the model number and manufacturer, type of VRAM and width of a memory bus.
That`s strange, but if you
)
That`s strange, but if you want I can offer a couple ideas how to solve the problem:
Little check:
1. Left running on the box only enshtein
2. In app make 0.5 CPU + 1 GPU and 2 threads
3. Change process priority to *realtime*
4. The most important things are not to load CPU with the other projects and process priority
It`s work pretty well for me
RE: ???? gtx 470 on my
)
That's very nice. Would like mine to do that.
It takes 154 minutes when I have 3 WU's running at the same time.
It takes 104 minutes when I have 2 WU's running. With or without the cpu's doing Malaria.
It takes 53 minutes when 1 have WU running.
So there is hardly any gain in time on my system with a EVGA GTX460 1GB FTW.
And here only .20 cpu is used no matter what I do.
Have assigned different cpu's (real time) to every WU.
And I want my systems to produce as much as possible so besides the GPU running E@H my quads run Malarianet with their spare cycles.
@ Vikk : Afterburner gave a max. load of 48% running two or three WU's.
RE: RE: ???? gtx 470 on
)
------------------------------
Can you check GPU load on 1/2/3 threads? (With afterburner or gpu-z)
------------------------------
windows_intelx86
0.200000
0.500000
BRP3cuda32
change to 0.5
windows_intelx860.5000001.000
)
windows_intelx86
0.500000
1.000000
BRP3cuda32
The 1.00000 is standard here. I've interrupted malaria for a few hours to see if it changes anything.
The gpu load must be around
)
The gpu load must be around 80%. If not - that means something wrong
RE: The gpu load must be
)
Nice to know but what ?
At the moment it is 45% running 3 WU's.
It`s look like you doesn`t
)
It`s look like you doesn`t change process priority. Use process tamer and make a rule for brp cuda
RE: It`s look like you
)
I use process lasso with which I assign a core to every wu in realtime.
Should be sufficient.
You may be able to avoid
)
You may be able to avoid using all these third-party tools when the new v1.05 application is running on your hosts.