[font='courier,courier new']Just an informal poll, but I've been noticing that SETI WUs (and Einstein WUs) are taking much longer to crunch than they used to. I've had two ten-hour-long SETI WUs and a 16-hour-long Einstein WU is still in my queue.

I'm only running BOINC Menubar, so there's no graphics, and I occasionally fire up BOINC Manager (just to look at what's in the to-do list), but that's for maybe a total of 10 sec. tops.

Is anyone else experiencing this?[/font]

---

My times seem to be about normal...6 hours for SETI, and 12 hours for Einstein.

Don

---

[font='courier,courier new']Just an informal poll, but I've been noticing that SETI WUs (and Einstein WUs) are taking much longer to crunch than they used to. Is anyone else experiencing this?[/font]

Wow, I just came by to check if anyone else has noticed this, and here you are!

I've seen my SETI WU times go from 2.5 to 3.5 hours and just can't figure it out. There were a few others who posted in the last few days that their computers mysteriously slowed down.

I know this: It's not the P4 3.2E downclocking from heat (I can watch it with Cpuz). It's not a virus (McAfee hasn't screamed). It's not another process (I get close to 2x50% of the CPU for two WU at a time in HT mode). And I am running the SSE3 optimized SETI.

So now I'm benchmarking the memory to see if I overheated it... But we've been having a cold spell here in the SF Bay area. Did I miss anything?

---

[font='courier,courier new']In my case, I suspected that maybe there was a change in the client apps, or that I was running on PPC and not x86, or that the presence of the BOINC Manager somehow affected the times. I checked my logs just to make sure that my installation of Tiger wasn't the source of the problem (it wasn't).

I can say for certain that it isn't an overheating problem - I've had my laptop running as many as 10 days straight with SETI at 100%, and if it didn't act up then, it's got no reason to do so now. I'm now draining out my cache so that I can try to reinstall stuff, but you know how long those CPDNs go.

At least I know I'm not crazy :-P[/font]

---

If you are talking about new builds, yes they take longer. E@H is currently chasing the problem. It doesn't appear to happen on all machines though.

Edit: I haven't yet run the new S@H 4.18, but I'd assume the results are the same as with the new E@H app.

First unit with 4.18 (non-optimized) looks to be done in 7h 23m 05s (26,585 seconds), whereas my units with 4.09 usually took 25,000 - 29,000 seconds. So for the moment I don't see a change yet.

In the end it took exactly 26,298.92 seconds. I'll check it again with 2 more units. ;)

---

[font='courier,courier new']
At least I know I'm not crazy :-P[/font]

That's the great thing about these boards - we all look sane compared to everyone else, no? :-)

Don't assume that just because your laptop didn't crash for 10 days in a row, that it wouldn't decide to slow down now. If temps run hot, materials get mushy, and things start to fail without rhyme or reason.

I'm now trying to figure out how to set (or even read) the memory timing settings on my Toshiba Satellite A75 laptop. The Phoenix BIOS is useless. SiSoft says my memory bandwidth is shot, maybe due to "Safe" down-clocking of memory timing. That could possibly happen on a smart motherboard when things get hot. I can reset these easily on a desktop with a good BIOS, but laptops are opaque.

BTW, the completion times are now all over the board, great to bad.

---

This thread caused me to have a look at my results. I have to say I cannot see any differences beyond +/- 2% in crunch times. Thats on P4 and P3 and Linux and Win XP/Win 2003. Perhaps you just had a challenging batch of WUs?

---

[font='courier,courier new']The PowerBook3,5 has a ppc7450, which doesn't have a thermal sensor available to the OS to be polled - Even OpenFirmware can't get to it. (it's hard-wired to the PMU, and it in turn hits the fans or can unilaterally sleep the machine until it cools off).

I've let it cool for six hours, but no change - Everything is just slugging along. The project app is getting all the CPU possible (top -o cpu -w -s 5), but is taking much longer than it used to (est. 18hrs., which is about 25% longer)[/font]

---

[font='courier,courier new']Just an informal poll, but I've been noticing that SETI WUs (and Einstein WUs) are taking much longer to crunch than they used to. I've had two ten-hour-long SETI WUs and a 16-hour-long Einstein WU is still in my queue.

I'm only running BOINC Menubar, so there's no graphics, and I occasionally fire up BOINC Manager (just to look at what's in the to-do list), but that's for maybe a total of 10 sec. tops.

Is anyone else experiencing this?[/font]

are you looking at the estimated WU time or the actual time? Actual time will be shorter than estimated and it helps to turn the computer off and let it rest once in a while. I had my laptop running at max for about a month before it decided to take a break on its own.

---

are you looking at the estimated WU time or the actual time? Actual time will be shorter than estimated and it helps to turn the computer off and let it rest once in a while. I had my laptop running at max for about a month before it decided to take a break on its own.

Whether the estimate is shorter or longer than your actual crunch time depends on your computer. Computers with small L2 caches seem to take much longer than the original estimate (up to a factor of 2).

---

BOINC WIKI

are you looking at the estimated WU time or the actual time?

[font='courier,courier new']Both.

SETI's estimate has always been way too low; Einstein's, just a bit. I can hazard a guess why SETI's is off the mark (I think it has to do with SETI's approach to combing through the data), but the estimate isn't the problem.

It's just.. well, all-around slow.[/font]

---

are you looking at the estimated WU time or the actual time?

[font='courier,courier new']Both.

SETI's estimate has always been way too low; Einstein's, just a bit. I can hazard a guess why SETI's is off the mark (I think it has to do with SETI's approach to combing through the data), but the estimate isn't the problem.

It's just.. well, all-around slow.[/font]

Just looked at my laptop which I know is now running 4.18 of the client. All the WU's seem to completing in 7-9K CPU secs.

Going back it appears there is not much change..

[edit] You are still crazy!! [endedit]

---

I'd rather speak my mind because it hurts too much to bite my tongue.

American Spirit BBQ Proudly Serving those that courageously defend freedom.

You are still crazy!!

[font='courier,courier new']Says a fellow TFFEer. :-P

I'll reset everything tonight and see if what happens.[/font]

---

You are still crazy!!

[font='courier,courier new']Says a fellow TFFEer. :-P

I'll reset everything tonight and see if what happens.[/font]

Did I say I wasn't crazy? - No Did you say you weren't crazy? - yes!!!

---

I'd rather speak my mind because it hurts too much to bite my tongue.

American Spirit BBQ Proudly Serving those that courageously defend freedom.

/me warms up the straight jackets and calls Nurse Rachette out of her hiding place.

---

[font='courier,courier new']The problem is that my results page is either empty or full of aborted WUs, but here's the two cases I can point to and say "Something's wrong":

E@H #1338735 host 283438.
It took longer and claimed less than anyone else who's returned a result. One normally expects that the longer it takes, the higher the claim.

S@H #18107243 host 1070936.
It took longer than usual (by about 10Ksec.), took more time than anyone else, and is a faster machine than one of its co-crunchers (host 819275)! The claim is higher than usual because it took more time.

I've re-run the benchmarks and they're in line with its previous results, so that can be ruled out. No new hardware has been added, there's nothing running in the background, the caches are all enabled and the processor is running at maximum.

Something's wrong.

I'll start resetting everything tonight.[/font]

---

OK, I have a theory:

Looking at my WU processing times, it's hard not to notice the beautiful cyclicality. Up and down in a nice sine wave on a 24-hour cycle. Hmmmm. What is it that I do at 00 hours UTC (5:00 PM Pacific Daylight Time here in the beautiful but cold San Francisco Bay Area)?? I WORK ON MY COMPUTER!! DUUUUH!

Or it could be Space Aliens. It won't be the first time the little buggers have driven me insane.

---

[font='courier,courier new']The problem is that my results page is either empty or full of aborted WUs, but here's the two cases I can point to and say "Something's wrong":

E@H #1338735 host 283438.
It took longer and claimed less than anyone else who's returned a result. One normally expects that the longer it takes, the higher the claim.

S@H #18107243 host 1070936.
It took longer than usual (by about 10Ksec.), took more time than anyone else, and is a faster machine than one of its co-crunchers (host 819275)! The claim is higher than usual because it took more time.

I've re-run the benchmarks and they're in line with its previous results, so that can be ruled out. No new hardware has been added, there's nothing running in the background, the caches are all enabled and the processor is running at maximum.

Something's wrong.

I'll start resetting everything tonight.[/font]

It claimed less but was granted the same your computer is probably active with other items more often

---

If you are talking about new builds, yes they take longer. E@H is currently chasing the problem. It doesn't appear to happen on all machines though.

Edit: I haven't yet run the new S@H 4.18, but I'd assume the results are the same as with the new E@H app.

I also noticed, but only on my laptop the recent slowdown from 3-4 hours to over 8 hours per wu of BOINC. I thought it was my computer, but even after cleaning it up I only can get the speed down to 6 hours per.

Nick

---