• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Client Configuration Comparison

C

Choder

[H]ard|Gawd
Joined
Mar 14, 2000
Messages
1,276
After reading Evil's post in the "So you want a low cost High PPD.." thread, I thought it would help some of us to see a broad comparison of configurations and how well they are preforming. A critique of our current configs wouldn't hurt either. So please post what you have and let us know if you see a way we can improve. Every little bit helps :D

If posting your config, you don't have to follow my convention, but you need to tell us what CPU you are using, how much RAM that machine has and how you are running the F@H client!

Please quote the configuration if you are posting a critique.

-------------------------------------------
My current F@H setup:
All CPUs are at stock speeds unless otherwise noted. PPD are according to FAHMon.

  • Xeon 3060 @ 3150GHz (450x7) | 4GB RAM | Linux Folding CD Flash Drive | SMP = 2500 PPD
  • Q6600 | 4GB RAM | 2x Linux Folding CD VM on Vista 32-bit | SMP = ETA 2/20/2008
  • T7100 | 1GB RAM | Linux Folding CD | SMP = 1300 PPD
  • E4300 | 2GB RAM | Vista 32-bit | SMP | Affinity Changer = 1050 PPD*
  • PentiumD 3.2 | 1GB RAM | Linux Folding CD | SMP = 1200 PPD
  • Pentium4 3.4 | 1GB RAM | Win2k3 | console client = 600 PPD
  • Dual Xeon 2.8 (Nacona) | 2GB RAM | Win2k3 R2 | SMP | Affinity Changer = 800 PPD
  • T5450 | 1GB RAM | Linux Folding CD | 2x console clients = 90 PPD each

* This machine is used for 8-10 hours a day and may have less than typical results.

Notes:
Affinity Changer has made a difference for me, raising PPD by around 20% on the machines tested.


 
Q6600 @ 3.52 (440 x 8), 2 x 1 GB RAM, Windows SMP, 3103.3 PPD
E6600 @ 3.375 (375 x 9), 2 x 1 GB RAM, Windows SMP, 2189.5 PPD
E4500 @ 2.926 (266 x 11), 1 x 1 GB RAM, Windows SMP, 1823.3 PPD


 
2x X5355's @2.66Ghz (stock), 4x 2Gb Ram, Win XP64-> 4x VM-> CentOS-5, 6,400 PpD.
2x X5355's @2.66Ghz (stock), 6x 1Gb Ram, Win XP64-> 4x VM-> CentOS-5, 6,000 PpD.
2x 285's @2.60Ghz (stock), 4x 1Gb Ram, Win XP-> 2x VM's -> CentOS-5, 2,450 PpD.
Q6600 @2.40Ghz (stock), 2x 2Gb Ram, Win XP-> 2x VM's -> CentOS-5, 3,260PpD.
E6600 @ 3.00Ghz (333 x 9), 2x 1Gb Ran, WinXP, 1,800 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,706 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,698 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,723 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,723 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,690 PpD.

10 Boxen = ~28,500 PpD.

Dont know why my second X5355 system running 400 PpD less than my first.
But its a consistant 100 PpD less per VM.
Same VM image so I may need to re-install Windoze on it.

Luck ............... :D
 
Dual Xeon E5345's @ 2.33GHz [Stock] 2x 2GB DDR2 FBDIMM [only addressing 3.2GBs until I switch from Vista Business x86 to x64 after this semester is over and I don't have school work to be afraid of losing] running 2x Windows SMP Clients

When I leave it running all day, it usually rakes in 3,500-ish PPD [3520 to be exact, I think]...and thats on a daily usage machine, so it'd probably be a little higher if it were dedicated to folding, but I can't afford to dedicate almost $3000 to folding, sadly :(


In the works to raise this, however, is plans to lock each SMP client to 3 cores each, and then have 2 cores reserved for my own usage, such as Adobe Flash, Photoshop, or BF2142 :cool:
 
Tigerbiten said:
2x X5355's @2.66Ghz (stock), 4x 2Gb Ram, Win XP64-> 4x VM-> CentOS-5, 6,400 PpD.
2x X5355's @2.66Ghz (stock), 6x 1Gb Ram, Win XP64-> 4x VM-> CentOS-5, 6,000 PpD.
2x 285's @2.60Ghz (stock), 4x 1Gb Ram, Win XP-> 2x VM's -> CentOS-5, 2,450 PpD.
Q6600 @2.40Ghz (stock), 2x 2Gb Ram, Win XP-> 2x VM's -> CentOS-5, 3,260PpD.
E6600 @ 3.00Ghz (333 x 9), 2x 1Gb Ran, WinXP, 1,800 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,706 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,698 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,723 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,723 PpD.
E6600 @ 2.65Ghz (297 x 9), 2x 1Gb Ran, CentOS-5, 1,690 PpD.

10 Boxen = ~28,500 PpD.

Dont know why my second X5355 system running 400 PpD less than my first.
But its a consistant 100 PpD less per VM.
Same VM image so I may need to re-install Windoze on it.

Luck ............... :D
Click to expand...

It might be because you've got a different ram config on the 2nd that might be holding it back some? Just a thought...

I'll get back to you in a couple days on how my lowly pentium dual core folds. I'm midway through replumbing the watercooling. But before I took it down it was just under 20mins a frame while running at 2.61ghz in vista 32bit SMP. I'm shooting for 3.2ghz after I get the WC installed.
 
One PS3 - approx. 900 PPD
Two E6400's (stock) 1x2gb ram 1x1gb ram XP Pro SMP client - approx. 1450 PPD each
One T7500 laptop (stock) 4gb ram Vista Home Premium 32bit SMP client - approx. 1100 PPD
 
>80% Workers:
5 x X5355's @2.66GHz (stock), 4x 1GB, WinXP32-> 2x Windows SMP-> 4,400 PpD
1 x E6700 @2.66GHz (stock), 4x 1GB, WinXP32 -> Windows SMP -> 1500 PpD
1 x E2160's @1.8GHz (stock), 4x 1GB, Windows Home Server-> ??? PpD.

<50% Workers:
6 x P4 @2.54GHz (stock), 2x 512MB, WinXP32 -> 100 PpD :D

Not sure about the E2160, it's new and I haven't even looked at it except to see if it was running, will update it later.


 
E6400@3.4, 2x1 GB, openSUSE 64 10.3: 2300PPD
X2 4000+@3.0, 2x512MB, openSUSE 64 10.3:1350PPD
Q6600@3.6, 2x2GB, openSUSE 64 10.3->2VMs openSUSE 64 10.3: 4600PPD

That's just off the top of my head but I believe the numbers are pretty close. If anything they are a little low in each instance.

 
SMP Clients
Q6600 @ 3.1 (345*9) | 2x2 Gig RAM | 2x VMs using Xubuntu 64-bit 7.4 | ~4200 PPD
X2 4800 @ 2.5 | 4x512 RAM | 1x VM using Xubuntu 64-bit 7.4 | ~1000 PPD

Standard and Part Time Clients
T7400 Laptop @ ~75% Time | 2x1 Gig RAM | 1x Standard Client in Vista | ~250 PPD
A64 3500 | 2x512 RAM | 1x Standard Client in Xubuntu 64-bit 7.4 | ~175PPD
P4 1.8 | 1x256 RAM | 1x Standard Client in XP Pro | ~70 PPD

 
Killer[MoB];1032084698 said:
Q6600 @ 3.52 (440 x 8), 2 x 1 GB RAM, Windows SMP, 3103.3 PPD
E6600 @ 3.375 (375 x 9), 2 x 1 GB RAM, Windows SMP, 2189.5 PPD
E4500 @ 2.926 (266 x 11), 1 x 1 GB RAM, Windows SMP, 1823.3 PPD


Click to expand...

Compare your E6600 to my Xeon 3060 at 3.15GHz (450x7). Memory amount and OS differences aside, I wonder if the increased bus speed also gives an advantage.

 
Choder said:
Compare your E6600 to my Xeon 3060 at 3.15GHz (450x7). Memory amount and OS differences aside, I wonder if the increased bus speed also gives an advantage.
Click to expand...

Bus won't make much difference. I would say that you have the most efficient machine ever built or your numbers might be off. Using all the info I have available at this time, I think you should be more in the 2129.4 to 2160.9 PPD range.

Linux smp straight up will be a max of 10% more efficient than windows smp straight up.

 
It takes about 17 hours and 30 minutes to process a WU, so I agree with you in that FAHMon's estimations are optimistic. Real world performance seems to be more around 2350PPD, which seems to be prefectly in line with your estimation of Linux SMP being 10% better than the Windows client. I'm in the process of moving that box from a Folding CD to VMWare. Maybe that will prove if the 10% figure is correct. I do still think that the higher bus / lower multiplier will be faster though. It certainly can't hurt.

Killer[MoB];1032086645 said:
Bus won't make much difference. I would say that you have the most efficient machine ever built or your numbers might be off. Using all the info I have available at this time, I think you should be more in the 2129.4 to 2160.9 PPD range.

Linux smp straight up will be a max of 10% more efficient than windows smp straight up.

Click to expand...
 
Choder said:
It takes about 17 hours and 30 minutes to process a WU, so I agree with you in that FAHMon's estimations are optimistic. Real world performance seems to be more around 2350PPD, which seems to be prefectly in line with your estimation of Linux SMP being 10% better than the Windows client. I'm in the process of moving that box from a Folding CD to VMWare. Maybe that will prove if the 10% figure is correct. I do still think that the higher bus / lower multiplier will be faster though. It certainly can't hurt.
Click to expand...

2350 sounds about right for that as I'm doing just a little bit under that with my E6400 clocked at 3.4 running Linux. I would probably get more out of it if I had the extra cache.

 
Choder said:
It takes about 17 hours and 30 minutes to process a WU, so I agree with you in that FAHMon's estimations are optimistic. Real world performance seems to be more around 2350PPD, which seems to be prefectly in line with your estimation of Linux SMP being 10% better than the Windows client. I'm in the process of moving that box from a Folding CD to VMWare. Maybe that will prove if the 10% figure is correct. I do still think that the higher bus / lower multiplier will be faster though. It certainly can't hurt.
Click to expand...

One other thing we don't ususally take into account around here is the chipset on the mobo. My E6600 is on a 680i. Testing my 680i vs my P35 I found the P35 to fold faster than the 680i with the same chip, ram, psu, clock speed and ram settings.
 
Killer[MoB];1032087939 said:
One other thing we don't ususally take into account around here is the chipset on the mobo. My E6600 is on a 680i. Testing my 680i vs my P35 I found the P35 to fold faster than the 680i with the same chip, ram, psu, clock speed and ram settings.
Click to expand...

Yeah, I hadn't really considered the chipset as a determining factor. Good to know.
 
Choder said:
I do still think that the higher bus / lower multiplier will be faster though. It certainly can't hurt.
Click to expand...

I don't know if there's any differences but I do know the fact using a higher bus and a lower multi will give out more heat than lower bus and a higher multi would.

Regardless...

Q6600 @ 3.6 (400*9) | 2x1 Gig RAM | Windows Vista Business 64 | ~3200 PPD
 
You must log in or register to reply here.
Back
Top