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

[johnerz]

Originally Posted by PatonB

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

Not on Intel, still recommended for ATI

[N64link]

I just noticed that I got an SMP A4 Unit, I'm on a PHII X4 955 now. With what PatonB mentioned earlier, anybody know if the SMP A4 would play nice with 4 cores and my GPUs? i'm running -SMP 3 ATM and to be safe have process lasso keeping my SMP on 0,1,2 and my GPUs on 3 but am curious about the advent of A4 SMPs.

[PatonB]

Yha, i is not uncommon for a4s to drop instead of a3.

a4 are a hybrid uni/smp unit, anbd the server will toss them out if its low on a3. Biggest issue are they are crappy units.

You're all good still.

[N64link]

So I shouldn't try to run the a4s on all four then? (With my GPUs going)

Even if they aren't high reward units, it's an extra 6-7K PPD with them, So I'm okay with that.

[PatonB]

I think you'll run into the same issues as a3s have. Had you 1 gpu you might be fine, but 2 meh...

But hell, prove it out... fire up 4 cores and see what happens.

[N64link]

Okay, I forgot, If I redo my config file will I lose my current WU? I've switched from 4 to 3 when I last had an A3 with processor lasso but I redid my config for -SMP 3 since then.

Additional Comment:

Well after that unit finished I redid my config file to get rid of -smp 3 and started fiddling around.
When I first started SMP, it was only using 25% of the CPU, I closed SMP and opend it back up used 100%. Started PL and set the F@H core_a4 process to use all 4 cores and set it to have above normal priority, (I had to do this with my Dual core and the 3870 and when I had the 3870x2).
SMP cranked away and after it finished 3 frames, HFM.net had confirmed what I thought. It's under 2:15 per frame (2:19 while surfing and typing this out).
Next I fired up my GPU client for my GTX465, It's running normally and SMP is still using 99% of CPU while maintaining frame times on the SMP and my standard GPU frame times, according to HFM.net.
Next i turned on my GT430's client and SMP is still using 99% of CPU with same frames. I'm waiting on the GT430 to finish some more frames to make sure it's times are right. But it take 8:45 on these units per frame.


So SMP, so far, has gone from 2:58 per fame to 2:17ish per frame on P8042 units. HFM was reporting around 7.3KPPD for these and is now calculated at 9.5KPPD.
Both GPUs are running P8009 units ATM and that's all I've gotten for the last couple of days.
GT430 is calculated on average, I'm still waiting for one more frame, 4.1k PPD.
GTX465 is calculated at 15.8K PPD.

i should get 29.4K PPD, up from previous calculated 26.6K PPD. So far, success on running a4 SMP with GPUs.

EDIT: Oh and CPU temps are no more than 53C. I'm getting some better fans to go on my S1284 in a week or so. So that should go to around 50C then.

[Maniac]

I have my SMP and GPU running low priority, with the GPU thread priority bumped up a little. Why do you have your priorities set so oddly?

[PatonB]

yha, you should have cpu set for idle, and gpu set slightly higher.

You really don't need process lasso, windows does a good job with managing things.

If youre using hfm, make sure youre letting it run atleast 5 frames.

I think you will see a varying in tpf as the units switch and the gpu gets more cpu intense wu.

[N64link]

Originally Posted by Maniac

I have my SMP and GPU running low priority, with the GPU thread priority bumped up a little. Why do you have your priorities set so oddly?

My thinking was that since my GPUs hadn't taken more than a 10 second hit in TPF or much PPD no matter how much my of the CPU I was using, they would be fine at idle. But I wanted to make sure the CPU wouldn't give too much to the GPU clients, when it uses the CPU for it's folding. There is probably a flaw in my logic. I'll switch them up and see if it makes a change on these small units or not.

Originally Posted by PatonB

yha, you should have cpu set for idle, and gpu set slightly higher.

You really don't need process lasso, windows does a good job with managing things.

If youre using hfm, make sure youre letting it run atleast 5 frames.

I think you will see a varying in tpf as the units switch and the gpu gets more cpu intense wu.

I'm using Process lasso, because I was just trying the a4 units and A4 and a3 have different processes. So if a3 didn't like to play nice with the GPUs and a4 did, I would only have to change the default affinity for those processes. instead of switching them manually. Every time I got a different unit.

I've only gotten the P8009s and similar sized units for the last week, on the GPUs. I hope it stays that way, they give me the best PPD, of the few I've seen. Since those huge ones take so long.

I've got an A3 now (P11070), which I had default to 3 cores.It was at frame 40 when I walked in and was around 8:40 TPF. I'm gonna try it on all 4 and see if it plays nice.

[Maniac]

Yeah, for process priority, you could set a process that's not doing anything to realtime and it still wouldn't do anything.

Since your SMP is *always* going to be using CPU, you want it on a very low setting. However, the GPU client only needs a bit of CPU once in a while, so in order to keep it running at top performance, it needs to be able to access the CPU when it needs to.

This will allow your SMP to run at full tilt when nothing else is going on, but when the GPU, or say your web browser, needs some cycles - they'll get them.

[N64link]

Yep. I'm as bright as a box of broken light bulbs.

On the bright side, this A3 is playing nice and is sitting at 4:32 TPF. I did put the wrong TPF before it had 6:23 when it was using 3 cores. (I was looking at my GT430s TPF)

Additional Comment:

At the 6PM update, I'll finally hit 2 million! only took my 4-5 years.

[jimh425]

Congratulations!