SPCR Folds Team Blog

aristide1 · Post by **aristide1** » Wed Nov 26, 2008 2:10 pm

I just realized, if the dual core frees up substantially I can start another SMP project.

Post by **NeilBlanchard** » Sun Nov 30, 2008 12:33 pm

Hi,

I need help figuring out why my new 24" iMac won't start Folding. I installed the verion with the installer, and the fah6 client is running in the Activity Monitor. But there are no cores running (there should be 4 of 'em?), and I can't get registered at the FoldingForum.org -- I think my ISP is blocking the confirmation email...

I've got the F@H clientset to run when ever the machine is on, but nothing is happening.

BTW, I added RAM to my brother's machine (up to 2GB) and so his iMac should be running much faster!

aristide1 · Post by **aristide1** » Sun Nov 30, 2008 3:03 pm

I never get email notification from the folding forums.

Post by **NeilBlanchard** » Tue Dec 02, 2008 5:36 pm

Hi,

I think I found the reason that the iMac wasn't Folding: I had to download the a2 core manually and install it. Now fah6 and mpiexec are active, but I've not seen the FahCore_a2.exe running yet. I'm checking the log file and I've gotten a widget to monitor it, called F@H WUdget.

So, it's better than it was, but I can't yet confirm that it is actually Folding...

Post by **NeilBlanchard** » Sat Dec 06, 2008 6:35 pm

Okay, my iMac is definitely Folding now -- it has gotten 55% in the last 22 1/2 hours. The last step that I had to do was to run chmod +x on the FahCore_a2.exe to make it into an executable (because I had downloaded it manually).

[Correction on the time: it took my 2.8gHz Core2 Duo iMac just ~21 hours to complete it's first work unit! The points have not shown up yet. The log file and the F@H WUdget I am monitoring it with both appear to be using GMT so I got it confused with local time, and did the math wrong. The log starts at 14:54:35 and ends at 11:35:16, so under 21 hours! Wow.]

Post by **NeilBlanchard** » Wed Dec 10, 2008 7:07 pm

Well, I could kick myself -- I didn't wait a few more minutes for a WU to complete, and I shut down to plug in my Kill-A-Watt; and I lost the whole thing...grrrr. My daily output would have been 1920 points higher.

My iMac consumes 124-126watts while running SMP Folding@Home. Very nice. I'll just have to wait another ~21 hours...

aristide1 · Post by **aristide1** » Wed Jan 07, 2009 7:42 am

Neil,

When you shut down FAH do you do it the proper way? CTRL + C?

Anyway, it's been a little less than 2 years since this thread was started. The threat back then, TLC Russia, is now the 7th team. They passed us when we were about team #20. Now we're at #40, and I looked at the team list this morning. A lot of new teams are going to pass us, which is OK, active folders are a good thing. These new teams are so strong we could double the number of people folding here and we wouldn't be able to keep up with them. As long as it pays off that's a good thing. I think GPU folding points really got folding moving. Lots of points on the cheap.

I can't say for others but my main rig running SMP and GPU folding has memory leaks under Vista. About once a month I start getting low memory messages and find I'm using over 2.3GB of memory (I have 4GB but under Vista 32 I see 3GB available). After a restart and I'm using 1.2 - 1.3GB. My XP rig only doing GPU folding doesn't seem to have his issue. Funny how the system slows down after using more than 60-70% of memory even though it has 600MB of memory free. Same old Windows. I can even close Internet Explorer and still see memory allocated for it.

If some of you haven't noticed lately - oh my God! - AMD dual cores can be had for under $40. Even a BE unit or 2, the energy efficient ones rated 45 watts. Simply amazing. The quad core equivalent of a BE would be a 95 watt quad priced at $170. I don't think I could be bothered with the latter. Anyone on a single AMD core can run SMP on the cheap. 300 watt APFC power supplies can be had cheap as well. I may pick up a BE just to compare with my current one.

The sorry state of affairs with PCs in general is the lack of progress with quad video motherboards. The AMD CPU lineup frankly has more to offer and for less money as well. Both NVidia and AMD have chipsets with 4 PCI-E slots for a bargain compared to Intel. FoxConn has a quad motherboard for barely over $100 after rebate on NewEgg (a solid board that has fallen out of favor because of one misplaced power connector.) Asus best quad motherboard (an OC special) is under $200. Go over and look at quad motherboards for Intel processors, it's a joke. Oddly enough Biostar, which many give high marks for OC capability and new technology leader, doesn't have a single offering.

The quad boards will become more viable for folding as Nvidia switches from 65nm to 55nm manufacturing. They have already done that with 2 current video cards, but I don't see any way to identify them. NVidia is planning to rename everything (again) so at that time the 9600GTs at least will be all 55nm. I think ATI is looking to go to 40nm manufacturing, but I've had bad luck with them, late drivers, and by the time my X1950GT got decent drivers it was no longer a good candidate to fold. $170 down the drain as far as I'm concerned. I gave the card away, upgraded copper GPU cooler and all.

aristide1 · Post by **aristide1** » Sat Jan 10, 2009 5:05 pm

SKeptical Thinker wrote:Or XP with the 180.60 drivers. My CPU load on the GPU core went from 100% to barely measurable. There was a slight hit on PPD (~100).

Wibla wrote:the new drivers does this? then I can start folding on my 8600GT again, as the current driver smashes rosetta throughput (yes, I defected, helping my local rosetta@home team)

I just installed these drivers now, and I must agree 100%, my cpu usage is now less then 1% for 1GPU where before it was 50% of a dual or core 100% of 1 core.

Another SMP on the way. And a free one at that.

aristide1 · Post by **aristide1** » Mon Jan 12, 2009 1:45 pm

Hey Neil, shake a leg man, I'm not waiting for you all day. Let's go!

Post by **NeilBlanchard** » Mon Jan 12, 2009 6:33 pm

Hi Aris,

I'm waiting for the Windows SMP version to get the kinks worked out, and then I get 1 (or maybe 2) machines up to speed (currently only running the single threaded version). And I need to get back to a Mac Pro to install the latest Mac version, and it'll get back on track, too.

My son doesn't want me to run it on his MacBook (and it does get a little bit loud), so I'm only running my iMac and my Linux machines with the SMP version, and one (or two) single threaded clients.

aristide1 · Post by **aristide1** » Thu Jan 15, 2009 5:08 pm

XP GPU folding is a swine. I rebooted with no monitor attached to GPU #2 (it did have a dongle) and it folded with errors. I finally figured it out after 2 days.

Goodness - The X2 on the same pc is running SMP. The 1066DDR2 memory I originally left at default setting 5-5-5-18 2T it was checkpointing every 39 minutes. I fixed the memory settings in the BIOS to 4-4-4-12 1T and now it checkpoints every 34 minutes and 40 seconds, a reduction of 4 minutes and 20 seconds per checkpoint. Not bad for a freebie.

I lowered the cpu voltage .05 volts and lowered consumption by 15 watts.

Post by **NeilBlanchard** » Thu Jan 22, 2009 3:22 am

Hiya,

My brother's iMac (17" Core Duo) is back Folding again (finally)!

http://folding.extremeoverclocking.com/ ... =&u=189150

angelkiller · Post by **angelkiller** » Fri Jan 23, 2009 3:28 pm

It's good to be back!

Yeah, I'm number 20, but still....

Right now I have my 4.05GHz E8400 and my 2.8GHz E2180 folding 24/7. I also have 2 4830's that are getting RMA'ed atm, so I'll add those in later. Any estimates on how many ppd the 4830's can get?

Interesting though, the E2180 is running 64bit Ubuntu and sometimes scores more ppd than my E8400.... I've been meaning to install Ubuntu on the E8400 system for a while.... We'll see. I was just really surprised how much of a difference the OS made.

Post by **NeilBlanchard** » Fri Jan 23, 2009 7:10 pm

Hi,

I think that Linux handles the MPIEXEC natively -- Windows has to be kludged a bit to work. BTW, has the Windows SMP client been updated recently? I would really like to get my work computer (WinXP) running that again...

ryboto · Post by **ryboto** » Sat Jan 24, 2009 7:35 am

im having trouble after 2weeks of zero issues. i've just recently overclocked my q9400 to 3ghz with 1.15v . i used it for 2weeks with no problems. then i tried a more aggressive undervolt which failed p95 testing. l set it back to the original overclock and ran p95 for 15 hours, (small fats) and had no errors. now whenever i run the f@h client i get the "critical error" message after a few % completion. my network isn't disconnecting, so i guess my oc is bad?

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 7:11 am

ryboto wrote:....so i guess my oc is bad?

Have you checked the stanford forums if this is a known error?
What board are you using?

I personally don't OC to off frequencies anymore, I stay with even multiples of 66MHz, which kicks in the northbridge divider and keeps the board running at stock frequencies. You may have less trouble at 8 * 400 or 3.2 GHz.

Also, no matter what speed your memory is, if it's CAS 4-4-4-12 it won't OC far and going immediately to 5-5-5-15 is a good safety precaution.

ryboto · Post by **ryboto** » Sun Jan 25, 2009 8:41 am

aristide1 wrote:
ryboto wrote:....so i guess my oc is bad?
Have you checked the stanford forums if this is a known error?
What board are you using?

I personally don't OC to off frequencies anymore, I stay with even multiples of 66MHz, which kicks in the northbridge divider and keeps the board running at stock frequencies. You may have less trouble at 8 * 400 or 3.2 GHz.

Also, no matter what speed your memory is, if it's CAS 4-4-4-12 it won't OC far and going immediately to 5-5-5-15 is a good safety precaution.

Well, I've been using 375x8, just because I don't want to go to far with it. I don't have a beefy power supply, and I didn't want to worry about cooling in my little mATX box. As for the RAM, it's at 5-5-5, at DDR2-900+, can't remember exactly.

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 9:06 am

400MHz stresses the board less, it changes multipliers on the Northbridge and returns everything on the board to stock frequencies. That's why many people have what they call an OC hole - a range they can't OC and then OCing starts to work again. The hole is always at the upper range just before the divider switch. Sometimes its only 10MHz wide, sometimes it 30+MHz wide.

You probably should bump your NB voltage up 1 notch if it's still at the stock voltage.

ryboto · Post by **ryboto** » Sun Jan 25, 2009 1:19 pm

aristide1 wrote:400MHz stresses the board less, it changes multipliers on the Northbridge and returns everything on the board to stock frequencies. That's why many people have what they call an OC hole - a range they can't OC and then OCing starts to work again. The hole is always at the upper range just before the divider switch. Sometimes its only 10MHz wide, sometimes it 30+MHz wide.

You probably should bump your NB voltage up 1 notch if it's still at the stock voltage.

If I move the NB voltage from the "auto" setting, the BIOS reports that the "OC has failed" and it makes me reset settings if I try an FSB from 333+ to 400. I've tried small voltage boosts, to large ones, and it's never stable in windows. At my current settings, I'm Prime95 stable. I'm going to try running StressCPU2 right now.

So, you're suggesting 400x7.5?

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 1:50 pm

Try 400 * 7 then 400 * 8. I'm not fond of 1/2 multipliers.

Can you hold the NB cooler in your hand without burning yourself?

ryboto · Post by **ryboto** » Sun Jan 25, 2009 1:56 pm

aristide1 wrote:Try 400 * 7 then 400 * 8. I'm not fond of 1/2 multipliers.

Can you hold the NB cooler in your hand without burning yourself?

yes, NB temps have never been bad, max I've ever seen was around 45C.

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 2:01 pm

Are you OCing your memory as well?

ryboto · Post by **ryboto** » Sun Jan 25, 2009 2:49 pm

aristide1 wrote:Are you OCing your memory as well?

It's DDR2-1000 RAM, running at DDR2-900. Rated timings at rated speed are 5-5-5-15, and that's what they're at now, with 100mhz less speed. Rated voltage is 2.0v, It's set to 2.0v, though on Mushkins forum, users have used less voltage to overclock the sticks over 1100mhz. Also, I've ran HCI's memtest for a day with the current ram settings, and it found zero errors.

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 3:21 pm

You got your bases covered. PCI-E slot should be locked at 100MHz, and PCI slots locked at 33MHz. Not much left after that.

What board do you use?

Have you checked how much your voltage on the CPU droops under load? Some boards will droop more than others. On mt C2D when I took it from 266MHz FSB to 400FSB I had to set the voltage to 1.47 for it to be 1.4 under load. Now I don't OC it so hard anymore. I run a 333MHz FSB, again to leave the board frequencies stock. I have less droop and can also run on lower voltage.

Typical reports for 45nm processors are running on air well into the high 3GHz range, some running 4GHz on air. YMMV.

ryboto · Post by **ryboto** » Sun Jan 25, 2009 5:28 pm

aristide1 wrote:You got your bases covered. PCI-E slot should be locked at 100MHz, and PCI slots locked at 33MHz. Not much left after that.

What board do you use?

Have you checked how much your voltage on the CPU droops under load? Some boards will droop more than others. On mt C2D when I took it from 266MHz FSB to 400FSB I had to set the voltage to 1.47 for it to be 1.4 under load. Now I don't OC it so hard anymore. I run a 333MHz FSB, again to leave the board frequencies stock. I have less droop and can also run on lower voltage.

Typical reports for 45nm processors are running on air well into the high 3GHz range, some running 4GHz on air. YMMV.

I only wanted 3.0ghz because I want to keep power consumption low, and my airflow is kind of restricted. There doesn't seem to be a droop, but an increase under load. I set it to 1.15v, and at idle, CPUZ/Everest reports 1.136v, and under full CPU load, that goes up to 1.144v. The board is an Asus P5Q-EM.

aristide1 · Post by **aristide1** » Sun Jan 25, 2009 6:37 pm

Have you considered taking the voltage up to 1.2v? That's still a lot lower than what my E6400 runs with.

ryboto · Post by **ryboto** » Mon Jan 26, 2009 5:55 am

aristide1 wrote:Have you considered taking the voltage up to 1.2v? That's still a lot lower than what my E6400 runs with.

When I get home I'll be able to check what happened with running StressCPU2, and I'll consider it after I see the results. I don't know why my CPU would need 1.2v for 3.0ghz, when that's what most require to reach 3.6ghz...and this is an R0 revision(E0 with half cache).

aristide1 · Post by **aristide1** » Mon Jan 26, 2009 6:36 am

I have an X2 Black Edition, people have run them as high as 3.7GHz, mine refused to be 100% stable at 3.2GHz even with excessive voltage, so I run it now at 2.8 and say the hell with it. Mine clearly is a dog. An unlocked multiplier dog, but still a dog.

In general you are correct, with less cache you should have more opportunity to OC, less circuitry stressed and lower power requirements.

I'm considering the Q9400 on my next build as well.

Iganu had the right idea back then. He undervolted and underclocked a Q6600, 85 watts under full load. Most quads need more than that at idle.

ryboto · Post by **ryboto** » Mon Jan 26, 2009 6:45 pm

So, I just logged back into my PC, and sure enough, StressCPU V2 is still running, though I don't know what an error would look like, it hasn't reported anything. I started it at 11:30pm last night, it's almost been 24hrs. So, probably something is wrong with F@H? StressCPUV2 uses the gromacs core, and stresses all 4 cores in case you hadn't used it.

aristide1 · Post by **aristide1** » Tue Jan 27, 2009 6:48 am

You may want to consider just testing 3.2GHz, because it will leave the motherboard running at stock/normal frequencies.

Biostar T boards do not undervolt, I need to find one that does.