SPCR Folds Team Blog
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
I'm not sure if dual channel RAM does much for folding. When I was testing the Q6600, I was using 512MB DDR2 RAM in single channel mode and it was getting roughly the same numbers as other people were posting on the net.
From what I've read, cache does make a difference - Ryboto can probably attest to this as he's using an Opteron which puts out more ppd than a X2 w/ 512k cache. I'd go for the oc'd 6320 or wait for the new cheaper 1333 fsb Conroe's due out July 22nd.
From what I've read, cache does make a difference - Ryboto can probably attest to this as he's using an Opteron which puts out more ppd than a X2 w/ 512k cache. I'd go for the oc'd 6320 or wait for the new cheaper 1333 fsb Conroe's due out July 22nd.
-
- Posts: 333
- Joined: Wed Apr 05, 2006 9:54 pm
- Location: USA
I think it would make sense to try to use the RAM you already have. I wouldn't worry about the price of a video card, one can be had for very little new and even less used. Also, I would try to get a board that allows you to overclock, as Core2Duos seem highly overclockable, and I think this can greatly improve performance. What about a socket 939 Athlon 64 X2? I believe socket 939 still used DDR 400, and new or used Athlon 64 X2s are available at exceptional prices and are overclockable as well.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
You make an excellent point, I'd like a table of processors, how many PPD they are capable of, and how many watts they use.vg30et wrote:I'm not sure if dual channel RAM does much for folding. When I was testing the Q6600, I was using 512MB DDR2 RAM in single channel mode and it was getting roughly the same numbers as other people were posting on the net.
From what I've read, cache does make a difference - Ryboto can probably attest to this as he's using an Opteron which puts out more ppd than a X2 w/ 512k cache. I'd go for the oc'd 6320 or wait for the new cheaper 1333 fsb Conroe's due out July 22nd.
Initial cost is secondary to operating costs to a large extent.
I did a quick comparison between the Q6600 + ECS motherboard against a Brisbane x2 3600+ & Biostar TF7025 motherboard. The mildly overclocked Q6600 numbers are mine while the X2 numbers are taken from another post on the board. I estimated the x2 3600+ PPD numbers based on what I got from my X2 3800+ @ 2.5ghz. Here's what I came up with:
If we ignore initial cost outlay then Intel comes out ahead. However, it's hard to ignore that the $174 price difference equates to approximately 2 years of operating costs for the AMD machine...
I am very tempted to run the numbers again with a G0 stepping Q6600 overclocked to 3ghz+. I have a feeling we'll see Intel's PPW numbers climb even higher.
Code: Select all
AMD Intel
Mobo + CPU Combo Price $125 $299
Power usage (Idle) 34 66
Power usage 100% @2.71 ghz 86 140
Idle vs 100% Delta 52 74
Points per Day 1100 2300
Point per Watt 21.15 31.08
I am very tempted to run the numbers again with a G0 stepping Q6600 overclocked to 3ghz+. I have a feeling we'll see Intel's PPW numbers climb even higher.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
So would the watts used, no free ride. It will become more interesting if as suggested the Q6600 is priced at $250.I am very tempted to run the numbers again with a G0 stepping Q6600 overclocked to 3ghz+. I have a feeling we'll see Intel's PPW numbers climb even higher.
That 7025 board sure isn't runnng Ubuntu, 32 or 64 bit. Can we pick up some PPD that way
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
good comparison. I calculated similar numbers for my opteron overclocked, and then for undervolted. I find that overclocked the chip uses more kilowatt-hours per WU than it does when undervolted and slightly overclocked. I did this by measuring the maximum cpu load while folding, the time it takes per framex100, convert that to hours, and multiply by the power load. I would do this for higher overclocks, but running fanless, I can't really push this chip to far on stock volts, undervolting is a necessity.vg30et wrote:I did a quick comparison between the Q6600 + ECS motherboard against a Brisbane x2 3600+ & Biostar TF7025 motherboard. The mildly overclocked Q6600 numbers are mine while the X2 numbers are taken from another post on the board. I estimated the x2 3600+ PPD numbers based on what I got from my X2 3800+ @ 2.5ghz. Here's what I came up with:If we ignore initial cost outlay then Intel comes out ahead. However, it's hard to ignore that the $174 price difference equates to approximately 2 years of operating costs for the AMD machine...Code: Select all
AMD Intel Mobo + CPU Combo Price $125 $299 Power usage (Idle) 34 66 Power usage 100% @2.71 ghz 86 140 Idle vs 100% Delta 52 74 Points per Day 1100 2300 Point per Watt 21.15 31.08
I am very tempted to run the numbers again with a G0 stepping Q6600 overclocked to 3ghz+. I have a feeling we'll see Intel's PPW numbers climb even higher.
-
- *Lifetime Patron*
- Posts: 1740
- Joined: Sun Dec 21, 2003 11:24 am
- Location: 'Sunny' Cornwall U.K.
No wattage's (I think...) but http://fahinfo.org does a Damn Good job with user inputed PPD's for lots of chips.aristide1 wrote:I'd like a table of processors, how many PPD they are capable of, and how many watts they use.
KansaKilla, I'm not sure ASRock boards are the way to go for you...
I had one and it died.. merely due to 'some' *cough* Vodka and a little session of unprotected vacuuming.....
Pete
-
- Posts: 333
- Joined: Wed Apr 05, 2006 9:54 pm
- Location: USA
That's because the switching power is proportional to f*V^2. Additionally, frequency changes do not affect leakage currents, but I believe voltage does. Leakage currents are more difficult to express. Also, the Q6600 really trounces the X2 3600+ in PPD, but how much of that is due to the 2 extra cores on the Q6600 and how much is due to architecture/cache/clock speed differences? It's hard to believe the Core2 architecture is that much better.ryboto wrote: good comparison. I calculated similar numbers for my opteron overclocked, and then for undervolted. I find that overclocked the chip uses more kilowatt-hours per WU than it does when undervolted and slightly overclocked. I did this by measuring the maximum cpu load while folding, the time it takes per framex100, convert that to hours, and multiply by the power load. I would do this for higher overclocks, but running fanless, I can't really push this chip to far on stock volts, undervolting is a necessity.
As I recall, didn't KansaKilla once post about a problem caused by a session of drunken overclocking?peteamer wrote: KansaKilla, I'm not sure ASRock boards are the way to go for you...
I had one and it died.. merely due to 'some' *cough* Vodka and a little session of unprotected vacuuming.....
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
I know vcore has a much greater effect. Though, frequency does come in to play when overclocking, underclocking has yet to make any impact on power draw when I run undervolted. It's easy to see that heat output and power consumption are directly related to the voltage, which is the obvious relationship. The Core2 arch is very fast at folding. At 2.8ghz, I was seeing 22min/frame with my opteron, the typical frame time for Core 2 users was 16min or under, from what I've read in various forums.disphenoidal wrote: That's because the switching power is proportional to f*V^2. Additionally, frequency changes do not affect leakage currents, but I believe voltage does. Leakage currents are more difficult to express. Also, the Q6600 really trounces the X2 3600+ in PPD, but how much of that is due to the 2 extra cores on the Q6600 and how much is due to architecture/cache/clock speed differences? It's hard to believe the Core2 architecture is that much better.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Yeah, fahinfo ispeteamer wrote: No wattage's (I think...) but http://fahinfo.org does a Damn Good job with user inputed PPD's for lots of chips.
Pete
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
This is what I was looking for:
From:
http://www.tomshardware.com/2007/07/11/ ... age11.html
That's embarrassing, the Core 2 Duo is wiping the floor with the Pentium D 840.
From:
http://www.tomshardware.com/2007/07/11/ ... age11.html
That's embarrassing, the Core 2 Duo is wiping the floor with the Pentium D 840.
Last edited by aristide1 on Wed Jul 18, 2007 5:39 pm, edited 1 time in total.
-
- Friend of SPCR
- Posts: 381
- Joined: Fri Aug 01, 2003 12:13 pm
- Location: Rochester, MN
Thanks for the advice, peteamer. I'll try not to vacuum when I overindulge. Karkov Vodka was the poison last time (emphasis on poison), and I blew out a hard drive. I was thinking that since those asrock boards are relatively cheap, it wouldn't hurt too much if I blew one up with the next round of ethanol. Might be time for some Evan Williams. (Why Karkov and not Stoli? Why Evan and not Jack? Because I'm a cheap bastard, that's why!)
I'm thinking the 6320 might just be the way to go to efficiently add points to the team. We're probably gonna get lapped one way or another in less than two weeks, but I gotta do something.
I'm thinking the 6320 might just be the way to go to efficiently add points to the team. We're probably gonna get lapped one way or another in less than two weeks, but I gotta do something.
I have some White Lightning I think you will like...(Why Karkov and not Stoli? Why Evan and not Jack? Because I'm a cheap bastard, that's why!)
http://www.walsallwonderland.co.uk/trampdrink.htm
-
- *Lifetime Patron*
- Posts: 1740
- Joined: Sun Dec 21, 2003 11:24 am
- Location: 'Sunny' Cornwall U.K.
Funny you should mention Stoli, I was attacked by a bottle of the very same last night... (yes attacked, I have no other explanation for todays hurting head. ) Tonight it's Smirnoff Norsk.
My last HD died from a Bacardi overdose... I nudged it out of it's elastic suspension whilst adjusting a fan, as it dropped it found an escaped molex 12V line and died whilst providing an excellent spark show that impressed the girlfriend at the other end of the room. Gotta admire the amount of amps a Seasonic 330 can provide... unfortunately the sparks eroded the side of some surface mounted thingamee and the magic smoke leaked out.
When the MOBO died the CPU died at the same time, I did a real good job there. Oddly when looking for replacements, an AM2 4400+ and 2GB Crucial RAM were the cheapest items I could find Had to have a new graphics card also, I couldn't find any AGP compatible MOBO's... in the PCI-E section... odd that.
My last HD died from a Bacardi overdose... I nudged it out of it's elastic suspension whilst adjusting a fan, as it dropped it found an escaped molex 12V line and died whilst providing an excellent spark show that impressed the girlfriend at the other end of the room. Gotta admire the amount of amps a Seasonic 330 can provide... unfortunately the sparks eroded the side of some surface mounted thingamee and the magic smoke leaked out.
When the MOBO died the CPU died at the same time, I did a real good job there. Oddly when looking for replacements, an AM2 4400+ and 2GB Crucial RAM were the cheapest items I could find Had to have a new graphics card also, I couldn't find any AGP compatible MOBO's... in the PCI-E section... odd that.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Get ready to drop another position. Team Engadget, is about to go from #24 to #22, and their trend is totally downhill.
Now we're hovering around 120 users.
Check out this page
http://folding.extremeoverclocking.com/ ... s=&t=31574
Im sorry we recently lost fgump, he was running really good.
Hey Jaganath, how did you go 2 days without completing any WUs?
Hey Neil, why are you stuck at 600 PPD?
Now we're hovering around 120 users.
Check out this page
http://folding.extremeoverclocking.com/ ... s=&t=31574
Im sorry we recently lost fgump, he was running really good.
Hey Jaganath, how did you go 2 days without completing any WUs?
Hey Neil, why are you stuck at 600 PPD?
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
I just recently was able to start folding using a lab computer. It's a dual Opteron 285 based machine. Trouble is, I'm not seeing any performance gain over a single Opteron 170 at the same clockspeed(2.6ghz). I'm thinking it's the RAM configuration, since the owner of the machine installed 4gb for one cpu, and only 1gb for another. Also, isn't there a way to get windows to see 4gb of ram? I was under the impression that there was, some BIOS setting, I read about it somewhere. I may just uninstall the 1gb set, and just keep the entire system running on 4gb.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
I do need to get over to my brother's house to find out why his production stopped -- this could be possible on Sunday...
I'm not "stuck" at 600PPD -- two days earlier, I was about 1,250PPD, and now I'm back up over 800PPD...Several of my machines are in other places, so they might not be running all the time. My home machines go through cycles, so they end up with some 0 point days.aristide1 wrote:Hey Neil, why are you stuck at 600 PPD?
I do need to get over to my brother's house to find out why his production stopped -- this could be possible on Sunday...
Well thanks for pointing that out. I must have misread that somewhere. I tried the Windows SMP Client and it's working fine now including resuming.ryboto wrote: it doesn't resume? strange...I've used the smp client in linux, in linux through Vmware, and in windows. The vmware isn't much more efficient than using the windows smp client. native linux is surely the best, but I can't undervolt in linux. I'm not sure what you mean though. I start and stop the client all the time in windows, and it picks right up where it left off. It'd probably fold faster using one Win smp client than two linux SMP clients in vmware. You're basically running 4 folding cores on each pair of cpu cores. Maybe it's about the same as just one Windows client though, I can't be sure. Still, the windows client resumes fine for me.
I think that's the better option than two virtual machines. It needs a lot less RAM and it's easier to maintain.
On Project 2610 my PPD is 2585 now. Not a bad number =)
Starting the SMP client as a scheduled task keeps it off the taskbar too
I periodically check the task manager to make sure it's still folding. Or I just fire up speedfan to verify my system is still blazing hot ("60C... yep, it's working!")
I hate those 1500 point WUs (2610), they take longer than the 1700 pointers.
I periodically check the task manager to make sure it's still folding. Or I just fire up speedfan to verify my system is still blazing hot ("60C... yep, it's working!")
I hate those 1500 point WUs (2610), they take longer than the 1700 pointers.
The 2652 WU is even worse. It is slow as hell. I only get around 1860 PPD with the QX6700.Plissken wrote: I hate those 1500 point WUs (2610), they take longer than the 1700 pointers.
I'm so tempted to buy a new cpu. I still have a P5B-VM Mainboard laying around which could replace my old Athlon XP setup. Well I wait for the price cut and see if there are any nice deals.
these 1500pt SMP WUs are giving me fits too.
the timing for my old dual xeon box is WAY off so i get WUs that it cant handle, and thus endup missing deadlines. I changed that one back to running 2-3 single process clients.
but with my "current" dual opteron box its hit & miss. aparently your given very little "buffer" for the SMP enabled WUs so when i have fah shut off over the weekend when i use my pc a lot.. it misses deadlines.
if it misses 1 more then im abandoning the SMP and going back to single process fah.
i figure low points is better than zero points due to missed deadlines.
the timing for my old dual xeon box is WAY off so i get WUs that it cant handle, and thus endup missing deadlines. I changed that one back to running 2-3 single process clients.
but with my "current" dual opteron box its hit & miss. aparently your given very little "buffer" for the SMP enabled WUs so when i have fah shut off over the weekend when i use my pc a lot.. it misses deadlines.
if it misses 1 more then im abandoning the SMP and going back to single process fah.
i figure low points is better than zero points due to missed deadlines.
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
the SMP WU's have a 4 day deadline. My Single opteron takes ~2.1 days to complete a WU. The Dual 2.6ghz Opteron PC in my lab complets WU's in ~1.2 days. You shouldn't have much of an issue as long as you let it finish a WU uninterupted. If you stop for a few days, you should delete the work folder and queue file before starting again.kittle wrote:these 1500pt SMP WUs are giving me fits too.
the timing for my old dual xeon box is WAY off so i get WUs that it cant handle, and thus endup missing deadlines. I changed that one back to running 2-3 single process clients.
but with my "current" dual opteron box its hit & miss. aparently your given very little "buffer" for the SMP enabled WUs so when i have fah shut off over the weekend when i use my pc a lot.. it misses deadlines.
if it misses 1 more then im abandoning the SMP and going back to single process fah.
i figure low points is better than zero points due to missed deadlines.
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
2.0ghz is fast enough, you just have to give it time. Any pauses longer than a few hours will definitely make it late.kittle wrote:4 days? that agrees with what i see for the listed deadlines.
I stop fah on my pc at nite while im working on it, so all that time is lost.
until they give longer deadlines, or i get a faster PC I can see it causing problems. Aparently 2ghz x2 isnt fast enough anymore..
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Hey Neil what do you think of this?NeilBlanchard wrote:Hello,
I'm not "stuck" at 600PPD -- two days earlier, I was about 1,250PPD, and now I'm back up over 800PPD...Several of my machines are in other places, so they might not be running all the time. My home machines go through cycles, so they end up with some 0 point days.aristide1 wrote:Hey Neil, why are you stuck at 600 PPD?
I do need to get over to my brother's house to find out why his production stopped -- this could be possible on Sunday...
http://reilly.homeip.net/folding/cd64.html
-
- Posts: 215
- Joined: Sun Jun 05, 2005 6:21 pm
- Location: St-Hubert, Qc