Need some folding Advice
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- *Lifetime Patron*
- Posts: 288
- Joined: Sat Oct 22, 2005 10:39 am
Need some folding Advice
Hi there, i've seen a large drop in my folding preformance when i started out with my new comptuer (Q9450) I was seeing about 1500 PPD from it, but lately its only doing about 600-700 PPD and i dont know why it has changed.
Does anyone have any ideas that come to mind off the top of your head?
Does anyone have any ideas that come to mind off the top of your head?
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
-
- *Lifetime Patron*
- Posts: 288
- Joined: Sat Oct 22, 2005 10:39 am
Im actually running 4 instances of the client and I don’t have any antivirus software installed. I just remembered some more detail when I first started a 250 point project would take about 8 hours so 1 core could do almost 1000 PPD or at least that’s what FahMon was telling me, but now it takes more than a day to do a 250 point project so that’s dropped each core to 175 PPD or there about.
I’m pretty new to folding really so I don’t know what kind of performance I should be getting, but my Q9450 is clocked at 3.2Ghz if that gives you any idea. Thanks for your help I would love to be able to help out the SPCR group more than I already am.
I’m pretty new to folding really so I don’t know what kind of performance I should be getting, but my Q9450 is clocked at 3.2Ghz if that gives you any idea. Thanks for your help I would love to be able to help out the SPCR group more than I already am.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello Dave,
You should definitely be running the SMP client, and not four instances of the regular client. That should yield much higher PPD. And on a quad core, there are steps you can take to run two instances of the SMP client that yield much more PPD than a single instance of the SMP client...with one SMP, you should be in the 2,000+ PPD range, and with two SMP clients, maybe 3,000+ PPD.
You should definitely be running the SMP client, and not four instances of the regular client. That should yield much higher PPD. And on a quad core, there are steps you can take to run two instances of the SMP client that yield much more PPD than a single instance of the SMP client...with one SMP, you should be in the 2,000+ PPD range, and with two SMP clients, maybe 3,000+ PPD.
I might get a Q6600 or X3220 just to inch above aristide1 in the rankings..NeilBlanchard wrote:Hello Dave,
You should definitely be running the SMP client, and not four instances of the regular client. That should yield much higher PPD. And on a quad core, there are steps you can take to run two instances of the SMP client that yield much more PPD than a single instance of the SMP client...with one SMP, you should be in the 2,000+ PPD range, and with two SMP clients, maybe 3,000+ PPD.
(especially now when they're getting alot cheaper)
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Wibla, you were way ahead of me when I had one, and sometimes two, machines not running. Even know if I stop to sneeze you will pass me.
This is the first time in months I stopped checking prices and yes they are down, but not that much. The Q6600 at NewEgg is still $219, and the selection of 45nm quads in stock is very small. This is not what I call technological progress. Even the E8400 is still $185.
You would think as energy costs rise more companies with many servers would be looking at cpus that use less electricity. The guy over at extremeoverclocking.com has cut his electrical use by a large margin, using less power now at full load than his prior systems used at idle!!!!
How are socket 775 Xeons different that regular 775 processors besides perhaps a larger cache????
I'd like to get one before Intel makes them all FSB1600 (real 400MHz FSB). Those will be very difficult to OC and they will have very low multipiers.
Oh, they have a baby 45nm dual core. Its 2.53MHz and FSB is 266, multipier 9.5, for $139, and they cut the cache to 3MB, but that's a lot more than those 2MB and 1MB super low cost dual cores out there. But I think Iganu's 80 watt Q6600 undervolted is the cheapest way to fold right now.
This is the first time in months I stopped checking prices and yes they are down, but not that much. The Q6600 at NewEgg is still $219, and the selection of 45nm quads in stock is very small. This is not what I call technological progress. Even the E8400 is still $185.
You would think as energy costs rise more companies with many servers would be looking at cpus that use less electricity. The guy over at extremeoverclocking.com has cut his electrical use by a large margin, using less power now at full load than his prior systems used at idle!!!!
How are socket 775 Xeons different that regular 775 processors besides perhaps a larger cache????
I'd like to get one before Intel makes them all FSB1600 (real 400MHz FSB). Those will be very difficult to OC and they will have very low multipiers.
Oh, they have a baby 45nm dual core. Its 2.53MHz and FSB is 266, multipier 9.5, for $139, and they cut the cache to 3MB, but that's a lot more than those 2MB and 1MB super low cost dual cores out there. But I think Iganu's 80 watt Q6600 undervolted is the cheapest way to fold right now.
Haha, yeah, and I lost a folder this week, dangit.. my E2140 @ 2.3GHz is offline and i dont know why...
a Q6600 is cheap as hell in norway now, same for the X3220's, which is basically the same cpu, down to cpu speed, cache and architecture..
Remember that you dont necesarrily need ALOT of cache to fold fast, atleast not in linux where you have the offset speed of not having to drag windows' dead weight around... my E2140 was almost as fast at 2.4GHz in linux as my E6600 was at 3.2GHz in windows, that says something...
a Q6600 is cheap as hell in norway now, same for the X3220's, which is basically the same cpu, down to cpu speed, cache and architecture..
Remember that you dont necesarrily need ALOT of cache to fold fast, atleast not in linux where you have the offset speed of not having to drag windows' dead weight around... my E2140 was almost as fast at 2.4GHz in linux as my E6600 was at 3.2GHz in windows, that says something...
-
- *Lifetime Patron*
- Posts: 288
- Joined: Sat Oct 22, 2005 10:39 am
Ok well at the moment im running the SMP client i didn't know about it before i just have 4 folders with a normal client in running them all at the same time. So im Running the SMP client now and it seems to be running all 4 cores, what did you meen by running 2 instances of the SMP client?Wibla wrote:Hello Dave,
You should definitely be running the SMP client, and not four instances of the regular client. That should yield much higher PPD. And on a quad core, there are steps you can take to run two instances of the SMP client that yield much more PPD than a single instance of the SMP client...with one SMP, you should be in the 2,000+ PPD range, and with two SMP clients, maybe 3,000+ PPD.
-
- *Lifetime Patron*
- Posts: 288
- Joined: Sat Oct 22, 2005 10:39 am
Well I ran 1 work unit with a single instance of SMP and it was running at 3000PPD. Now i have started running 2 instances of SMP using something called "FAH SMP Affinity Changer" which I found after looking through some forum threads and FahMon tells me its running at 4000PPD, so im alot happier now.Fat_bloater_dave wrote:Ok well at the moment im running the SMP client i didn't know about it before i just have 4 folders with a normal client in running them all at the same time. So im Running the SMP client now and it seems to be running all 4 cores, what did you meen by running 2 instances of the SMP client?Wibla wrote:Hello Dave,
You should definitely be running the SMP client, and not four instances of the regular client. That should yield much higher PPD. And on a quad core, there are steps you can take to run two instances of the SMP client that yield much more PPD than a single instance of the SMP client...with one SMP, you should be in the 2,000+ PPD range, and with two SMP clients, maybe 3,000+ PPD.
Im also running a Athlon 64 3800 which is giving me 800PPD so thats 4800PPD so hopefully ill be up in the top 20 producers soon.
Thanks for all your help guys.
Dave