SPCR Folds Team Blog
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
I didn't actually try overclocking on stock voltage. But from reading a lot of the threads here and elsewhere it looks like a good rule of thumb is that you can either overclock about 20% on stock voltage, or undervolt about 20% at stock speed. Now, since power consumption is proportional to the square of voltage, that 20% undervolt will get you a 36% power saving for the cpu (not quite so much for the whole system). And since power consumption under load is roughly proportional to clock speed, the 20% overclock will increase power consumption by about 20%. But in practice I think it will be somewhat less.
So looking just at cpu power consumption, overclocking on stock voltage looks like it should slightly increase points per watt. While undervolting at stock speed will significantly increase points per watt. So points per watt is better when undervolting at stock speed.
But when you bring other system power consumption into the picture then it's not so clear. For my folding machines which have low power integrated graphics and no hard drive or optical drive, I think undervolting is still significantly better in terms of points per watt. But I haven't measured.
Undervolting also makes the machines easy to run quietly while folding 24/7, which is the reason I originally started fooling around with undervolting. The energy savings are just a bonus.
So looking just at cpu power consumption, overclocking on stock voltage looks like it should slightly increase points per watt. While undervolting at stock speed will significantly increase points per watt. So points per watt is better when undervolting at stock speed.
But when you bring other system power consumption into the picture then it's not so clear. For my folding machines which have low power integrated graphics and no hard drive or optical drive, I think undervolting is still significantly better in terms of points per watt. But I haven't measured.
Undervolting also makes the machines easy to run quietly while folding 24/7, which is the reason I originally started fooling around with undervolting. The energy savings are just a bonus.
Even without the voltage bump, the OC will draw more watts. If the #1 goal is low power consumption, an OCed CPU is probably not a good choice.aristide1 wrote:It seems this would also give you more points per kilowatt, the amount you can OC without increasing the voltage.
But what do I know? A couple days ago I said all I get are the 1760 pointers, and now I'm getting a whole bunch of weird ones
[edit: Iganu beat me with a much better, thorough explanation, thanks!]
Yes but it's not optimal in his case. The answer is in iganu's post. Change in speed vs. voltage has a linear vs. squared relationship to power (approximately).aristide1 wrote:I agree, but wouldn't slowing the processor down also save power? Not that I want to do that.
To illustrate (not real #s):
You undervolt 0%, underclock 2% and see power savings of 2%.
You undervolt 2%, underclock 0% and see power savings of 4%
You undervolt 2% and manage to overclock by 1%. Power savings 3%.
Etc.
I'm sure every proc. has a "sweet spot" but it first depends on your goals. In iganu's case, it was low heat (easy to silence) 24/7 folding speed. I can see reasons to underclock, but not on a serious folding machine. Go change the lightbulbs and leave the CPU alone! Well, ok, you can undervolt it.
I said his priority is folding 24/7, and his machines are serious folders, so he should (obviously) not underclock.VanWaGuy wrote:Not on a serious folding machine? He is top daily folding rate on our team is all.
Although I can see reasons to overclockPlissken wrote:it first depends on your goals. In iganu's case, it was low heat (easy to silence) 24/7 folding speed. I can see reasons to underclock, but not on a serious folding machine.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
I guess that I need to reinstall F@H on my brother's iMac -- it has "refused" to credit his user; despite repeated efforts to get it to do so.
My home SMP machine has been finishing a work unit, and then balking on starting the next one. It has "stray files" that need to be deleted -- has anybody else seen this, and if so, is there a fix?
I guess that I need to reinstall F@H on my brother's iMac -- it has "refused" to credit his user; despite repeated efforts to get it to do so.
My home SMP machine has been finishing a work unit, and then balking on starting the next one. It has "stray files" that need to be deleted -- has anybody else seen this, and if so, is there a fix?
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Neil I think you now have 2 that require reinstalling.
I got a Proj 3064 WU, 25 minute checkpoints.
My E6300 used to finish a wu in 1410 minutes, gaining half an hour a day, but not lately. Its lucky it keeps up now. Stuff is changing.
If there's a way to get your pc to stop the FAH development team seems gifted at finding it.
I got a Proj 3064 WU, 25 minute checkpoints.
My E6300 used to finish a wu in 1410 minutes, gaining half an hour a day, but not lately. Its lucky it keeps up now. Stuff is changing.
If there's a way to get your pc to stop the FAH development team seems gifted at finding it.
Neil,
Often you can just delete the working folder and restart folding. I used to need to clean things up, but have not needed to in months. (and I know I will now as soon as I type this) If you have the latest version, you can keep just the config file and the one you download and re-install and run quickly. If you download again, and they might have changed anything, then don't even save the config. Last update, they changed some of the info in there like work unit size that would result in less than optimal folding.
Often you can just delete the working folder and restart folding. I used to need to clean things up, but have not needed to in months. (and I know I will now as soon as I type this) If you have the latest version, you can keep just the config file and the one you download and re-install and run quickly. If you download again, and they might have changed anything, then don't even save the config. Last update, they changed some of the info in there like work unit size that would result in less than optimal folding.
-
- Friend of SPCR
- Posts: 381
- Joined: Fri Aug 01, 2003 12:13 pm
- Location: Rochester, MN
-
- Friend of SPCR
- Posts: 381
- Joined: Fri Aug 01, 2003 12:13 pm
- Location: Rochester, MN
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Iganu,iganu wrote: Motherboard: GA-G33M-DS2R (uATX with BIOS support for undervolting)
I dont see this mobo on NewEgg anymore, they have the good board
http://www.newegg.com/Product/Product.a ... 6813128085
and the el cheapo board
http://www.newegg.com/Product/ProductRe ... 6813128078
Wow, the cheap board may have only 2 memory sockets, but it still has solid state caps. Impressive.
I just don't know of either one of them undervolt.
And what exactly are my undervolt options these days?
Aris
I think the GA-G33DS2R has been discontinued. It's a shame, because it's a really nice uATX form factor board. I was hoping to get another to put a 45nm quad in at some point.
I suspect that both of the boards you linked to will have undervolting in the BIOS. But it's less of a gamble with more expensive boards so the G33 based board would be the safer bet if you want to undervolt.
I suspect that both of the boards you linked to will have undervolting in the BIOS. But it's less of a gamble with more expensive boards so the G33 based board would be the safer bet if you want to undervolt.
Download the user manual of the prospective mobo. Some, like the older P5B-D, will list every CPU voltage setting available - from 1.2750 to 1.7000V Not sure about Giga boards though.aristide1 wrote:I just don't know of either one of them undervolt.
And what exactly are my undervolt options these days?
GA-P35-DS4 (Rev 2.0) has these options
DDR2 Over Voltage: +0.05v to +1.55v in 0.05v increments
PCI-E Over Voltage: +0.05 to +0.35v in 0.05v steps
FSB - as PCI-E
(G)MCH - northbirdge - Overvoltage: +0.05v to +0.375v in 0.025v steps
CPU Voltage: The adjustable range depends on the CPU - plus the blablah warnings about you might b*gger up your CPU.
That's the whole set of voltage adjustments.
PCI-E Over Voltage: +0.05 to +0.35v in 0.05v steps
FSB - as PCI-E
(G)MCH - northbirdge - Overvoltage: +0.05v to +0.375v in 0.025v steps
CPU Voltage: The adjustable range depends on the CPU - plus the blablah warnings about you might b*gger up your CPU.
That's the whole set of voltage adjustments.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
I knew it about 5 days ago. The pc that I have folding far away had stopped. I watched my average score nose dive and I couldn't do a thing about it. Of course when I started it up after a week off no matter which SMP module it was working on it said "this one expired", but at least it didn't tell me how many checkpoints it had completed before being tossed. I have no idea why it stopped but it could have been a small power outage. I guess I will need at least one UPS soon as.
The wacky WU's continue.....
The wacky WU's continue.....
-
- Posts: 78
- Joined: Fri Jul 28, 2006 1:07 am
- Location: UK
Check out the stats! http://folding.extremeoverclocking.com/ ... s=&t=31574
-
- Posts: 78
- Joined: Fri Jul 28, 2006 1:07 am
- Location: UK
looks like iganu might have something to do with that...
http://folding.extremeoverclocking.com/ ... =&u=267697
http://folding.extremeoverclocking.com/ ... =&u=267697
Lots of 2605 and 2619 ...
Iganu's sudden spurt looks like the result of a catch up by the stats team. Many many folders had been beefing on the support site about their completed WUs not getting credit. All the three WUs credited to me in the three hour period to midnight CST last night were catch-ups which I had copmleted between last wednesday and sunday morning. I imagine that is true for most, or all, of Iganu's 12.
-
- Posts: 78
- Joined: Fri Jul 28, 2006 1:07 am
- Location: UK