Back in the saddle.(Success! post 26!) (2Million points for me!)

[N64link]

I finally had time to put in my GTX465 and @ stock it's doing one frame every three minutes or so.
But as I thought about how annoyed I am about the other rig I was trying to set up for the GT430, I bought at the same time. I remembered something awesome.
I don't need SLI/CrossFire to run two GPUs on the same board. (I have a cross fire board.) So I looked around EOCF, foldingforum, and evga forums and learned how to just set up a dual card system. Set my MoBo to 8x/8x mode for the slots and just barely got the GTX and the GT next to each other. Figured out how to switch between the two cards in MSI Afterburner. Set up my clients and am now folding on both cards, with 99% percent usage each, and my CPU with SMP.
Based off of what other people report as average PPD, I should get ~18K with this.

I've seen some opinions on the net but I trust the long time folders and big points earners of EOCF over all others. Is OCing a GPU still a no-no, for science results? I remember being told, I think by MtM, that OCing my HD3870 could result in unreliable results and that going stock was best for science.
If OCing is ok now, how do we test to make sure it's F@H stable and accurate?

[PatonB]

UnfortunaTELY WITH FERMI CARDS, THEY DO NOT OVERCLOCK LIKE THE OLD G92, gtx200 caqrds did, i.e. independant shaders/core

I found my 560ti really didnt get alot of overclock so barely a ppd boost, for the amount of extra heat.

But you may have better results so all you can do is try yourself.

[jimh425]

Welcome back. I can't help with the GPU folding.

[N64link]

Originally Posted by PatonB

UnfortunaTELY WITH FERMI CARDS, THEY DO NOT OVERCLOCK LIKE THE OLD G92, gtx200 caqrds did, i.e. independant shaders/core

I found my 560ti really didnt get alot of overclock so barely a ppd boost, for the amount of extra heat.

But you may have better results so all you can do is try yourself.

The GT 430 I can actually unlink the shader and core clock. But the GTX is locked. I did a quick test for only a couple of minutes in Afterburners kombuster and the GTX went to 750 core with no artifacts, But I only ran it for a few minutes. I have no idea if Kombuster will give an accurate gauge of F@H stable clocks or not. And I don't know if memtest g80 stresses the whole card or just the vram.

Additional Comment:

Experimenting with OCing. Upped the GTX465 100MHz core/ 200MHz shader & 100MHz VRAM shaved a bit over 30 seconds per frame off ( currently @ 2:32 per frame). The GT430 I upped by 50MHz all around it is now @ 8:25 per frame, from 9:22
And my SMP is still slow as molasses @ 24:46 Per frame.

Only been at it for twenty-ish minutes with no EUEs so here is hoping.

[johnerz]

Originally Posted by N64link

The GT 430 I can actually unlink the shader and core clock. But the GTX is locked. I did a quick test for only a couple of minutes in Afterburners kombuster and the GTX went to 750 core with no artifacts, But I only ran it for a few minutes. I have no idea if Kombuster will give an accurate gauge of F@H stable clocks or not. And I don't know if memtest g80 stresses the whole card or just the vram.

Additional Comment:

Experimenting with OCing. Upped the GTX465 100MHz core/ 200MHz shader & 100MHz VRAM shaved a bit over 30 seconds per frame off ( currently @ 2:32 per frame). The GT430 I upped by 50MHz all around it is now @ 8:25 per frame, from 9:22
And my SMP is still slow as molasses @ 24:46 Per frame.

Only been at it for twenty-ish minutes with no EUEs so here is hoping.

download evga precision and use that to oveclock your card
it is debatabe about the extra point versus the extra power used and heat produced is worth the effort

[PatonB]

Ruhn -smp 3 for your best results. a3 hates sharing and slows dramatically when it does.

[N64link]

Originally Posted by johnerz

download evga precision and use that to oveclock your card
it is debatabe about the extra point versus the extra power used and heat produced is worth the effort

But I'm too lazy to make an account to DL it from EVGA.
Honestly, the heat doesn't concern me, I set my fans to 80-100% anyway. And my apartment is rent controlled. I don't think it will use more than my older card,(3870x2) even with the GT430.

Originally Posted by PatonB

Ruhn -smp 3 for your best results. a3 hates sharing and slows dramatically when it does.

Is that a flag? What does it do?

[PatonB]

You can get evga precision from majorgeeks btw.

Yes to -smp 3. It tells the program to only run on 3 cores. This frees up 1 core for the gpus so they are not inpeeding the smp setup.

I'm unsure on how v7 sets core flags though.

[N64link]

I'm bumping the cards a bit again, since they ran all night with no EUEs. The GT430 is now at 850MHz core, 1700MHz shader, & 925MHz vram. GTX465 is siting at 750MHz core, 1500MHz shader, & 1775MHz vram. I'll see how stable this is after I get back from my final. And hopefully they both have a bit better frame times.

Additional Comment:

Oh, I'm still on a dual core 7750BE Kuma.

[PatonB]

Oh crap, I though you had a c2q

Ditch the cpu folding then

[N64link]

But then I just have 85% of my CPU just sitting idle. Could the SMP really slow down the GPUs that much?

EDIT: The GTX is at 2:22 per frame and the GT is at 7:31 per frame. They are both working on Project 8009s ATM.

[Maniac]

Originally Posted by N64link

But then I just have 85% of my CPU just sitting idle. Could the SMP really slow down the GPUs that much?

EDIT: The GTX is at 2:22 per frame and the GT is at 7:31 per frame. They are both working on Project 8009s ATM.

You could always try it and find out...

[N64link]

I will. I'll run with no SMP for a week and run with SMP the next week. Then see what it does. I need to borrow another skt775 rig so I can start testing the other components of that side rig. I would like to put that C2D on an SMP at least. Or see if my 3870x2 will be ok in that for a folder. For some reason, it stopped receiving WUs and I couldn't justify not putting the GTX in my rig, anymore.

Thanks for answering my questions, guys. Hope I can get a little more PPD in a month or so. (I still have a loan to pay off for a rebuilt transmission.)

[PatonB]

Your 3xxxati cards are done. Theyve stopped putting out units for them.

If you don't pay for power, i'd run a unicore a4 on the 1core but not smp.

[N64link]

Originally Posted by PatonB

Your 3xxxati cards are done. Theyve stopped putting out units for them.

If you don't pay for power, i'd run a unicore a4 on the 1core but not smp.

Well that makes me sad. I guess I'll put my 3870x2 in my nephew's PC so he can play Diablo 3 a little better. And see if anybody wants the two 3870s I have lying around. Still good cards for low level use and older games. The Bad thing about these new cards is the lack of a S-video/component port. My TV's HDMI slot doesn't work and I already miss watching Hulu on my TV.

So I lock both units to one core and run a single core F@H on the other?

[PatonB]

You really don't need to do anything other than just start them up.

I beleve other DC projects still can make use of older ati cards.

The reason behind the death of older ati, is ati switch from Brooke to openmm. I.E. the switched up their programing language

[N64link]

Ok, so since it's just A3 that hates to share, would two instances of the single core client hurt the GPU clients?

[johnerz]

! would not the only way to prove with 2 is run them

Don't forget to set affinity, to one core for the single core client and the other core to the GPU client - if you only use one core

[PatonB]

You don't need to set affinity anymore as far as i'm aware.

[N64link]

I seem to be getting very large WUs on my GPUs when I set the parameter for small in both set ups. I keep getting WUs that are worth 7 kpoints but they take so long that they will barely get 9k ppd, plus they are so intense that I've had to back down the OC which folded the smaller 2.8k point units @ 2:20 per frame and was getting me close to 15k ppd. My net connection isn't very fast and it take a while to DL/UL these bigger WUs.

So is the parameter for the size of WU just worthless or is there something I have to do to force it to only give me smaller units?