• 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.

what am I missing

J

joeb0b

Limp Gawd
Joined
May 25, 2005
Messages
141
So I added a new 3ghz box and i started folding on it before I noticed it was HT
So it went over night folding at about 50% CPU
Logs showed:
[20:53:49] Completed 0 out of 125000 steps (0)
[21:12:15] Writing local files
[21:12:15] Completed 1250 out of 125000 steps (1)
[21:30:22] Writing local files
[21:30:22] Completed 2500 out of 125000 steps (2)
[21:48:29] Writing local files
[21:48:29] Completed 3750 out of 125000 steps (3)
As you can see about ~18 minutes for 1250 steps

Then I realized it so I turned off Hyper threading
[14:35:11] Completed 72500 out of 125000 steps (58)
[14:53:12] Writing local files
[14:53:13] Completed 73750 out of 125000 steps (59)
[15:11:14] Writing local files
[15:11:14] Completed 75000 out of 125000 steps (60)

Now its like 95+% CPU and Still 18 minutes for 1250 steps. Is it going faster and I can't tell or are there other factors about hyperthreading or the way this log works that I'm missing?

-joeb0b
 
Hyperthreading is, to make a long story short, fake multi-processing.

With HT on, one processor was maxed, and the other was not, hence 50%.

With HT off, you only have one, and it's maxed.

Either way, F@H is still doing the same calculations at the same rate.

I would recommend turning HT on (wouldn't you want it on anyways?), and running two instances of F@H. I'm not sure of the specifics for doing that, as I don't have a box with HT myself, but someone here should be able to help you out.
 
HT is really just two instruction paths with one processor behind them. Windows sees it as two processors, but there's really only one. When running multiple processes, the two paths help smooth the transition while slicing up the processor time between them. Your "50%" usage is really just one of the instruction paths being used at 100% and the other with minimal usage (OS overhead). The core of the processor that does the work is working at 100% in both instances. That's why you don't see any time difference.

Enabling HT is generally better. FAH can use one instruction queue while the OS and other tasks use the other, reducing the time taken to clear out the queue for a new process. Or, you can run two instances of FAH and it will see each instruction queue as a separate processor. Each would only fold at about 50-55% the speed of a single one, but this slight total speed gain over one is because each would have its own queue. Overall, HT will give you about 5-15% speed gain, depending on how the computer is used and what speed it is.
 
I have a Hyperthreading P4 at work and what I do is run normal units on one instance and let it run full out. On another instance I run timeless WUs and have it set to use 15-20% CPU. 10-20% is around the extra work that gets done when you run two copies on a Hyperthreaded P4. This way, the time per frame on the main instance isn't affected much and I still get some extra production out of the machine.

This falls in line with what Stanford prefers, getting the WUs back as soon as possible, while I still get a PPD increase by taking my time running the timeless WUs slowly.

 
hmm thats interesting. Basically I turned off HT thinking that when enabled my one instance was using only half the cpu and I was trying to use most the cpu with the least amount of effort. To get that extra 10% or so I guess I'll have to re-enable it and setup another instance of FAH.

Thanks for the responses,
-joeb0b
 
You must log in or register to reply here.
Back
Top