• 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.

SMP 2610 and Memory Bandwidth

A

aldamon

Supreme [H]ardness
Joined
May 24, 2000
Messages
6,716
My new E4300 file server was getting dreadful folding times in SMP 2610 (36 minutes per stage!) at 2.7 GHz, so I ordered another 1GB stick of RAM this morning to enable dual channel. Did I waste my money or do you guys think this will help the times? Everything I've read has indicated that 2610 is extremely bandwidth limited. My DS3/E6300 was pulling 19 minute stages at 2.94 GHz, but it was in dual channel with a much higher FSB. I'm assuming this upgrade in bandwidth will help me get into the 20s with the E4300 at 2.7. Does anybody have any experience with this? I suppose I could pull a stick out of my E6600 rig to test it, but then I'd have to stop folding for 30 minutes. :p I was hoping somebody already experimented with this.

 
aldamon said:
My new E4300 file server was getting dreadful folding times in SMP 2610 (36 minutes per stage!) at 2.7 GHz, so I ordered another 1GB stick of RAM this morning to enable dual channel. Did I waste my money or do you guys think this will help the times? Everything I've read has indicated that 2610 is extremely bandwidth limited. My DS3/E6300 was pulling 19 minute stages at 2.94 GHz, but it was in dual channel with a much higher FSB. I'm assuming this upgrade in bandwidth will help me get into the 20s with the E4300 at 2.7. Does anybody have any experience with this? I suppose I could pull a stick out of my E6600 rig to test it, but then I'd have to stop folding for 30 minutes. :p I was hoping somebody already experimented with this.

Click to expand...

I can do 15 mins with my E4300 but it's at 3.3 GHz and 2 Gb. Those WU are sensitive to ram timings and bandwidth indeed, but also to L2 cache.

I bet your E6600 rig will be able to crunch the 2610 much faster.

 
Xilikon said:
I can do 15 mins with my E4300 but it's at 3.3 GHz and 2 Gb. Those WU are sensitive to ram timings and bandwidth indeed, but also to L2 cache.
Click to expand...

Thanks. My CPU is only 18% slower than yours so hopefully the 20s will be possible with dual channel. I'm hoping I can push for 2.9 GHz as well, but my board doesn't have a working PCI lock and starts crapping out at 315. 2.7 - 2.8 might be my max.

Xilikon said:
I bet your E6600 rig will be able to crunch the 2610 much faster.
Click to expand...

Yeah, my E6600 is pulling 10:50 frames but I didn't want to put a $200 CPU in a file server. I'm being a cheapskate, hence the E4300, open-box motherboard ($50 P5L-VM 1394), and temporary single-channel experiment. :p I'll slide over the E6600 when I go quadcore in my main rig.




 
Anyone know if enabling or disabling NUMA on Opteron systems affects SMP times?
 
Hmmm...my 2.8Ghz X2 takes 32-33 minutes per frame. It has 512KB+512KB L2 cache. My FSB is 267 (DDR is at 700Mhz). Is that low performance for that processor?
 
Um, holy crap???!!!

I just popped in the second stick of RAM to enable dual channel and my time per stage dropped from 36 minutes to 22! :eek: :eek: :eek: A 40% improvement for $40? I'll take it.

[18:05:17] Completed 55000 out of 500000 steps (11 percent)
[18:41:34] Writing local files
[18:41:35] Completed 60000 out of 500000 steps (12 percent)
[19:17:51] Writing local files
[19:17:52] Completed 65000 out of 500000 steps (13 percent)
[19:54:08] Writing local files
[19:54:08] Completed 70000 out of 500000 steps (14 percent)
[20:30:26] Writing local files
[20:30:26] Completed 75000 out of 500000 steps (15 percent)
[21:02:04] Project: 2610 (Run 1, Clone 50, Gen 21)
[21:02:04]
[21:02:05] Entering M.D.
[21:02:11] Calling FAH init
[21:02:12] Writing local files
[21:02:12] Completed 77069 out of 500000 steps (15 percent)
[21:02:12] ein
[21:02:12] Writing local files
[21:02:12] Completed 77069 out of 500000 steps (15 percent)
[21:02:14] Extra SSE boost OK.
[21:15:21] Writing local files
[21:15:21] Completed 80000 out of 500000 steps (16 percent)
[21:37:51] Writing local files
[21:37:51] Completed 85000 out of 500000 steps (17 percent)
[22:00:12] Writing local files
[22:00:12] Completed 90000 out of 500000 steps (18 percent)
[22:22:33] Writing local files
[22:22:33] Completed 95000 out of 500000 steps (19 percent)

 
Caffeinated said:
Hmmm...my 2.8Ghz X2 takes 32-33 minutes per frame. It has 512KB+512KB L2 cache. My FSB is 267 (DDR is at 700Mhz). Is that low performance for that processor?
Click to expand...
Considering that my 2.3GHz Opterons accomplish a frame in around 50 minutes, I'd say that is a good time and about right. Be glad you can run the SMP client at all. ;) :)

aldamon: awesome!
 
You must log in or register to reply here.
Back
Top