The Dreaded 2684

C

cbarbour71

Weaksauce
Joined
May 26, 2011
Messages
88
So, for those with 2600k CPUs running -bigadv, if you happen to pull a 2684, the new points breakdown really hoses us.

32:00 TPF with a 4.6GHz 2600K nets an estimated 38.8K PPD. Makes you wonder if it's worth running -bigadv and risking getting nearly the same PPD as standard SMP.

Thoughts?
 
cbarbour71 said:
Makes you wonder if it's worth running -bigadv and risking getting nearly the same PPD as standard SMP.

Thoughts?
Click to expand...
I don't understand the quandary here, but you're certainly not the first to bring it up. How is it a risk when the worst unit you can get is still as high as the best points you could get with SMP. So long as there aren't consistent scenarios where SMP is faster on the same hardware then I don't think there's any real risk.

Am I just missing something?
 
Yes, the points are still above standard SMP, but it's the greater risk that worries me.

Should something cause the loss of a processed WU, the lost time and credit hurts. I realized this has always been a risk with the bigadv, but now that the credit for the worst bigadv is on par with regular SMP, the risk may not be worth the reward anymore.

I will caveat this with the fact that I have yet to have serious issues with system stability, so I personally haven't been affected by this, but I can see where others with stability or connection issues could be impacted pretty significantly.

Just my 2 cents.
 
you must still be in windows.....i'm still hitting atleast 40k PPD on 2684
 
W.Feather said:
you must still be in windows.....i'm still hitting atleast 40k PPD on 2684
Click to expand...

Nope, running Ubuntu. If your sig is correct, I'm betting the difference is CPU speed (you have me by 200MHz) and possibly RAM speed (I'll bet your timings are better than mine).
 
hmm, 200 mhz shouldnt be that much different on 2684....i know 100mhz on a 6900 improves me by 21sec
 
cbarbour71 said:
Nope, running Ubuntu. If your sig is correct, I'm betting the difference is CPU speed (you have me by 200MHz) and possibly RAM speed (I'll bet your timings are better than mine).
Click to expand...

You are most likely correct there ram speed and timings do make a pretty big difference when it comes to folding. At least they do on socket 1366 ;)
 
Zink said:
You can run linux in a VM if you need to stay on windows.
Click to expand...

Would this offer any advantage given its still windows underneath? Or would you run both on a hypervisor, I really don't know shit when it comes to virtual machines. :(
 
I tried a VM using VirtualBox and I was not impressed.

The 2684 I am working on using Windows is getting me 45.5k ppd with the 980x @ 4.4 GHz.
 
ugh, my SR-2 has a 2684 and its getting 100k ppd. why the hell does stanford insist on giving these units out when they clearly value my hardware at 200k+ ppd on other units? :rolleyes:
 
tjmagneto said:
I tried a VM using VirtualBox and I was not impressed.

The 2684 I am working on using Windows is getting me 45.5k ppd with the 980x @ 4.4 GHz.
Click to expand...
you were not impressed, as in, you didn't get any better PPD?
 
Vbox doesn't really support the "symmetric" part of SMP. It's mainly built for sequential or concurrent operations, not symmetric ones. For folding purposes, you can get away with 8 or fewer threads okay if you're in a pinch and have no other options. That is to say, it will work but it will still underperform VMware, which has an actual SMP implementation. Anything over 8 threads, and Vbox goes to dogdoo.
 
jebo_4jc said:
you were not impressed, as in, you didn't get any better PPD?
Click to expand...

It seemed not any better but I don't recall by how much. I'll do a little testing with a VM and report back with a comparison of Windows, Linux and Linux running in a VM.
 
Update regarding Virtual Machines: I know the data is not from a 2684 but here's what I am getting from the 980x rig in my signature. The WU is a 6901.

Windows 7 64 bit 52k ppd
VBox 54k
Kubuntu 64 bit 60k

How this will scale from one system to another as well across workunits I'm not sure.
My rig just started a 2685 in a vm and it looks like its netting a 5% gain in ppd versus windows which is about the same gain as I posted above with the 6901.
The vm image I am using is from the linuxforge website. I'd link to it but the site is currently unavailable.
 
You inspired me to try out a Linux VM on my SR2 to see if I can get any benefits running bigbeta compared to standard Windows wu.

I'll start a new thread
 
You must log in or register to reply here.
Back
Top