When is ATI going to get into F@H?

[p2501]

you'll have to try and see what happens.... you have 16 cores, so that would be 8 smp clients to maximize the thread... (going from it takes 2 smps on a quad to get alll the cycles).
and I think you are allowed 8 machine id's... so it's probably doable. no idea how it scales software wise on 16 cores though... that would be 32 threads of fahcore_a1

I think when I last set up my SMP clients they showed a machine ID up to 16, not that anyone would run 16 SMP's on a dual neha...

[Marvin_The_Martian]

Not good enough, esp. that P3065 total just 9xxx ppd totally drain off the performance ...


...

That cpu usage graph isn't at 100% for any core... no from me

Edit: you know what, I think most would understand this is a joke, but for the one or two who would like to jump in 'ahahah it's 100% just taskman shows diffrent then you kentsfield, learn something new' let me just say it is up front

[road-runner]

Hi, I'm curious ...
Are you run all this F@H on one rig ?
Is your QX9650 oc'ed ? B'cause I used Gainestown with 4SMP just gained +/-7xxx ppd .... points every SMP are +/-18xx ppd ...
Can I run CPU's 8SMP to gains the exact same ppd per SMP ? Or it'll just reduce the performance ?? I do intend to try it ... Just need some input ...

PS : VISTA 64Bit ...

THX

No, this is several different rigs. My QX9650 is overclocked to 4ghz and running an 8800 also. You should be able to run 2 cores per SMP client, I have not had any luck doing in Windows but works rather well in Ubuntu using the taskset to assign cores...

[JCornell]

No, this is several different rigs. My QX9650 is overclocked to 4ghz and running an 8800 also. You should be able to run 2 cores per SMP client, I have not had any luck doing in Windows but works rather well in Ubuntu using the taskset to assign cores...

Thanks for your enlightened ...

[MikeB12]

yeah, with windows,
Affinity changer is what maxes the 8 threads' of fahcore_a1 efficiency http://distributed.org.ua/forum/inde...showtopic=1149
I'm using it to run dual smps in vista64 with one gpu2 client...

last time I tried it in xp, gpu wouldn't play nice with it... because I had to isolate fahcore_a1 to 3 cores, and affnity changer was written to use all 4 cores.

[JCornell]

yeah, with windows,
Affinity changer is what maxes the 8 threads' of fahcore_a1 efficiency http://distributed.org.ua/forum/inde...showtopic=1149
I'm using it to run dual smps in vista64 with one gpu2 client...

last time I tried it in xp, gpu wouldn't play nice with it... because I had to isolate fahcore_a1 to 3 cores, and affnity changer was written to use all 4 cores.

So you're running SMP with 3 rigs, right ? And each rig setup with 2 SMP's set to 2 cores each SMP, right ?

But with Gainestown case, actual can run 8SMP's, but it seems doesn't boost enough performance ... too bad ...
My thought is maybe 2*SMT's can't perfectly handle 2*SMP's(8*Fahcore_a1) at the same time compare to physical cores ...

Maybe I'll try like yours on my yorkie ... 2SMP's for each cores ...

[JCornell]

That cpu usage graph isn't at 100% for any core... no from me

Edit: you know what, I think most would understand this is a joke, but for the one or two who would like to jump in 'ahahah it's 100% just taskman shows diffrent then you kentsfield, learn something new' let me just say it is up front

This is rough testing ... I doesn't manually set the affinity. But my experience to set affinity or not doesn't boost the performance ...

[MikeB12]

yeah J, that's 3 separate rigs with 1 Q6600 in each...