I'm currently running 4 SoB WUs on my Core2Quad Q6600. One of those WUs started running two days ago. It appeared to be running at a pace that would set it to complete in about 10 days.

Yesterday, I started running the other three SoB WUs. Those seem to be looking at a 12 day duration. In addition, the first WU has also slowed down.

During that first day, the other cores were not idle; they were running tasks from another project (SIMAP to be specific).

It appears that running multiple instances of this application causes some type of interference amongst the WUs.


____________
My lucky number is 75898⁵²⁴²⁸⁸+1

It appears that running multiple instances of this application causes some type of interference amongst the WUs.

Unfortunately, that's a known effect, but the reason is still unknown. See this thread.

Edit: OK, I was too slow this time. :)
____________

It appears that running multiple instances of this application causes some type of interference amongst the WUs.

There is no solution for this yet. See this post: LLR slowdown across multiple cores

My four SoB WUs have working sets of 52 MB each. I'm also running 32 bit Vista SP2, on a Q6600 (same as you), so it seems kind of odd that yours is so different.

Then again, I seem to recall the working sets being a different size, so perhaps it varies over time.

My WUs are checkpointing every 15 CPU minutes. I have my minimum checkpoint interval set to 300 seconds (5 minutes), but I don't think that's affecting this application's checkpointing, although I could be wrong. Anyway, it does checkpoint, but not every minute like some other projects do.

IIRC, when running PPS (LLR), it also did checkpoints every 15 minutes, which meant it never checkpointed on the Q6600 since it took less than 15 minutes to run. It did checkpoint at 15 minutes on slower computers.

I see you're running 6.10.18 on that machine. There's a relatively new feature available in the BOINC manager. If you select a task and hit the "Properties" button, a dialog box is displayed that shows you information about the selected task. One of the things it tells you is when the last checkpoint was.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

This slowdown also occurs with Prime95. Exactly what the cause is, I am not sure. I find that running only two or three cores increases the speed by as much as 10%.

These applications cause the CPU temps to rise significantly and it may be the built-in Intel protection mechanism kicking-in. I suggest you download one of the core temp monitoring apps and see what the temps are doing.

Perhaps this might help in figuring out what the problem is:

So I have these four SoB WUs running, and they slowed down.

You know what also slowed down by about 15%? The GPUGRID task running on my GPU. The slowdown we see in the LLR processes appears to also be affecting the CPU thread of the GPUGRID application.

The GPUGRID application hasn't changed in a while, and consistently runs at around 77% GPU utilization (according to GPU-Z). It's now running at about 60%, and fluctuating more than it normally does.

I suspended the four SoB tasks, one by one. As each was suspended, the GPU load went up. Even a single SoB task caused the GPU to slow down. That's really odd since it's just a single thread running on a quad core system with three other cores almost idle.

It should be noted that other applications don't cause this slowdown.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

I've got 4 running on my PII 940 OCed @ 3.5Ghz.

I think he is talking of a Phenom II 940