• 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.

Stability Question

Z

Zefram0911

[H]ard|Gawd
Joined
Apr 19, 2002
Messages
1,178
I'm running the computer in my sig... I can stress test and do normal activities on it 24/7... but when I start an smp folding client it will crash anywhere between ten minutes and ten hours. Any ideas on why folding causes my system to go nuts and what to do to fix that? I'd really like to start folding again... no sense in having a quad core at 3.8Ghz if you're not using the cycles.

TIA.
 
Try to lower the overclocking 100 or 200 MHz and see if it get stable. Folding stress the computer more than any other stability test ;)

 
Xilikon said:
Try to lower the overclocking 100 or 200 MHz and see if it get stable. Folding stress the computer more than any other stability test ;)

Click to expand...

This is very true.....

Very little will make a CPU beg for death like folding 24/7 on too high of an OC

 
I am no expert but from what I understand the calculations in the SMP process have a much lower tolerance for errors than the other stress tests out there. It's all math, SMP just requires it to be done to a higher rate of accuracy.


Correct me if I am wrong.
 
Wheresatom said:
I am no expert but from what I understand the calculations in the SMP process have a much lower tolerance for errors than the other stress tests out there. It's all math, SMP just requires it to be done to a higher rate of accuracy.


Correct me if I am wrong.
Click to expand...

Not 100% myself but I also think it's the shear amount of data running though the cache.

 
Zefram0911 said:
I'm running the computer in my sig... I can stress test and do normal activities on it 24/7... but when I start an smp folding client it will crash anywhere between ten minutes and ten hours. Any ideas on why folding causes my system to go nuts and what to do to fix that? I'd really like to start folding again... no sense in having a quad core at 3.8Ghz if you're not using the cycles.

TIA.
Click to expand...

Tried Mem test?
 
As science moves forward so does the difficulty. For quite a while prime-95 was the gold standard for stress testing because prime numbers were getting beyond huge.

Now we are asking machines to literally do life simulations on a speculative basis. Like it or not that gets complex.

Then comes the vicious circle, more computer power is going to equal science people asking bigger questions. It used to happen more slowly but now the growth has become exponential and frankly that is to be expected.

Look back 5 years and the hot rocks folder was a dual socket “A” box with something like 10% or so of the power now offered by an over clocked Quad.

The cost of a really massive Super Computer to do the work we do is frankly prohibitive even for and educational institute with some Government sponsorship. 10 years ago a Cray, the ultimate computer, cost something close to a billion dollars if you include all the accessories to make it run. I’d estimate just our little team would outperform that Cray on a grand scale.

Now comes the fun part, none of use here at the [H] are happy with what the industry would term the “out of the box experience” so we roll our own and we try to beat the manufactures at their own game. Face it, if we could over clock a capacitor we would.

There comes a point however where pushing a piece of hardware will cause it to error and frankly nobody builds a computer to make errors. The Quad as delivered today is capable of about an 800 MHz to 1 GHz over clock before things will go south. We quickly become way too used to things that get better and come to expect too much. Its human nature and we are (for the most part) humans.

When you over clock a Quad to 3.8 that is past wonderful, but if it is too unstable to do it’s job at 3.8 but will do it’s job at 3.2 or so it’s not wasted cycles, you only paid for 2.4 GHz to start with so you get a net gain.

Science wins, the team wins, and we all win.:)
 
BillR said:
When you over clock a Quad to 3.8 that is past wonderful, but if it is too unstable to do it’s job at 3.8 but will do it’s job at 3.2 or so it’s not wasted cycles, you only paid for 2.4 GHz to start with so you get a net gain.

Science wins, the team wins, and we all win.:)
Click to expand...
Yu're putting up arguments about things I wasn't even complaining about. I'm not angry about not having a 3.8Ghz system that's not stable. My point was... if I could get a system stable at 3.8Ghz... why not use it, because normal usage for most people won't use those cycles... it would be wasted.
 
Which CPU stress tests did you use and for how long ???

On my new Quads, I'm running two copies of the Orthos stress test in the gromacs mode, with affinity set so that each copy locked to its own side of the CPU.
My first quick test is only for an hour.
While I'm testing I've got CPuZ & Core Temp running so I can keep an eye on both core temps & volts.
Then its finish installing all the software, patches, etc, etc.
After that its running it for 24 hours as the final test before running the SMP client.
And even then I keep an eye on the new boxen for a month.
More than one EUE a week and I call that box unstable and go back and retest it at a slightly lower overclock or higher voltage as needed.

Luck ........................................ :D
 
Congrats on your overclock, my X3350 would not run more than a couple of minutes
at 3.8Ghz. I eventually had to back it down to around 3.4 to get it stable running F@H.
Obviously your results may very, I am using a much cheaper MoBo, GB P35 DS3L with
decent Crucial Ballistix. Good Luck and Keep on Folding!

 
BillR said:
As science moves forward so does the difficulty. For quite a while prime-95 was the gold standard for stress testing because prime numbers were getting beyond huge.

Now we are asking machines to literally do life simulations on a speculative basis. Like it or not that gets complex.

Then comes the vicious circle, more computer power is going to equal science people asking bigger questions. It used to happen more slowly but now the growth has become exponential and frankly that is to be expected.

Look back 5 years and the hot rocks folder was a dual socket “A” box with something like 10% or so of the power now offered by an over clocked Quad.

The cost of a really massive Super Computer to do the work we do is frankly prohibitive even for and educational institute with some Government sponsorship. 10 years ago a Cray, the ultimate computer, cost something close to a billion dollars if you include all the accessories to make it run. I’d estimate just our little team would outperform that Cray on a grand scale.

Now comes the fun part, none of use here at the [H] are happy with what the industry would term the “out of the box experience” so we roll our own and we try to beat the manufactures at their own game. Face it, if we could over clock a capacitor we would.

There comes a point however where pushing a piece of hardware will cause it to error and frankly nobody builds a computer to make errors. The Quad as delivered today is capable of about an 800 MHz to 1 GHz over clock before things will go south. We quickly become way too used to things that get better and come to expect too much. Its human nature and we are (for the most part) humans.

When you over clock a Quad to 3.8 that is past wonderful, but if it is too unstable to do it’s job at 3.8 but will do it’s job at 3.2 or so it’s not wasted cycles, you only paid for 2.4 GHz to start with so you get a net gain.

Science wins, the team wins, and we all win.:)
Click to expand...

Very well stated Bill....and looking back on all the technology that has evolved since I started getting into computers (and I've seen a lot since about 1980) It just amazes me how much things have progressed. Like you mentioned above, the Cray used to be the standard by which all others were judged. :eek:

That being said, it makes me wonder now just how many total gigabyte's worth of processing power that the [H]order brings to the table. :confused: Although it changes every day as we upgrade or add more boxen, does anyone have a baseline figure as of today?




 
Zefram0911 said:
Yu're putting up arguments about things I wasn't even complaining about. I'm not angry about not having a 3.8Ghz system that's not stable. My point was... if I could get a system stable at 3.8Ghz... why not use it, because normal usage for most people won't use those cycles... it would be wasted.
Click to expand...

No argument at all, have one of these
14063l.gif
as well as a doughnet and "attend more church oriented social activities". (Zappa)

Hell, how I wish I could get my quads to 6.0 stable but we all know better.

My only point was the work is getting harder and the hardware more critical.

Ok, we cool?;):)

 
You must log in or register to reply here.
Back
Top