I examined the code of Prime95 some months ago because one of my friends wanted to more about it, so i know that Prime95 isn't a best choice for testing OC.
Yep, thats true. I once overclocked my XP2800, ran Prime95, all seemed well.
But with climateprediction I experienced a machine reboot from time to time. CPU-temperature wasn't an issue, it was well under max. specs.
Based on the experience that S41.07 is actually significantly slower on my computer than S41.06 I did some calculations.
Based on a total of 92 results (51 with S41.06 and 41 with S41.07) I came up with these results.
The 51 results with S41.06 took a total of 164667 seconds which gives an average of 3228.764706 seconds per WU.
The 41 results with S41.07 took a total of 145041 seconds which gives an average of 3537.585366 seconds per WU.
It gives a difference on average of 308.82066 seconds per WU.
The computer I use is an AthlonXP 2500+ (Barton) running at stock speed.
Based on these numbers and the fact that I'm not affected by the bug it fixes, I don't think I'll upgrade to S41.08 if it's based on the code of S41.07.
Which version is better for Athlon64 3500+? I saw Athlon64 with S41.07 and U41.04 so I don´t know which version is better or faster for my CPU... Thanks for Answer
I'm sure that S41.07 is faster on K7 cores than S41.06 with 1-2%.
I don't remember where in this thread, but we have actually seen previous results that suggests that for some K7 cores there is little or no difference in speed between S41.06 or S41.07.
Quote:
Probably you measured more short wus with S41.06.
Actually the only short WUs in the calculations (only 1 or 2) were processed with S41.07, but since they didn't make a difference to the result I didn't mention them.
I've been running S41.07(HT) - based on little more than I thought it was probably the most appropriate version, though I don't know whether there was any real benefit of using it instead of the standard S41.07 client.
The question now is - should I change to S41.08 instead?
Thanks for any advice. Note - the computers running the clients require HT to be enabled for other purpose, so disiabling HT isn't a viable option in this case.
I've been running S41.07(HT) - based on little more than I thought it was probably the most appropriate version, though I don't know whether there was any real benefit of using it instead of the standard S41.07 client.
The question now is - should I change to S41.08 instead?
Thanks for any advice. Note - the computers running the clients require HT to be enabled for other purpose, so disiabling HT isn't a viable option in this case.
I've found that on my P4 3.0 HT(on) that the fastest version was the S40.04. Here is a previous post in this thread I'd wrote about comparing the S40 & S41HT. After trying the newest S40, S41, & S41HT versions since that post I still find the S40 being the fastest for the 3.0 HT (I leave the HT on too because my computers get used for everything and are not 'dedicated' for crunching.) On the other hand, the new S41.08 (and S41.07 non-HT) are the fastest for the Cel. 2.4. You can take it as advice, but these are my observations.
RE: OC sometimes causes
)
Yep, thats true. I once overclocked my XP2800, ran Prime95, all seemed well.
But with climateprediction I experienced a machine reboot from time to time. CPU-temperature wasn't an issue, it was well under max. specs.
Greetings groundhog
Based on the experience that
)
Based on the experience that S41.07 is actually significantly slower on my computer than S41.06 I did some calculations.
Based on a total of 92 results (51 with S41.06 and 41 with S41.07) I came up with these results.
The 51 results with S41.06 took a total of 164667 seconds which gives an average of 3228.764706 seconds per WU.
The 41 results with S41.07 took a total of 145041 seconds which gives an average of 3537.585366 seconds per WU.
It gives a difference on average of 308.82066 seconds per WU.
The computer I use is an AthlonXP 2500+ (Barton) running at stock speed.
Based on these numbers and the fact that I'm not affected by the bug it fixes, I don't think I'll upgrade to S41.08 if it's based on the code of S41.07.
RE: Based on the experience
)
I'm sure that S41.07 is faster on K7 cores than S41.06 with 1-2%.
Probably you measured more short wus with S41.06.
Which version is better for
)
Which version is better for Athlon64 3500+? I saw Athlon64 with S41.07 and U41.04 so I don´t know which version is better or faster for my CPU... Thanks for Answer
RE: I'm sure that S41.07 is
)
I don't remember where in this thread, but we have actually seen previous results that suggests that for some K7 cores there is little or no difference in speed between S41.06 or S41.07.
Actually the only short WUs in the calculations (only 1 or 2) were processed with S41.07, but since they didn't make a difference to the result I didn't mention them.
U41.xx is faster than S41.xx
)
U41.xx is faster than S41.xx on any machine that can support SSE3
RE: U41.xx is faster than
)
Thanks for the answer... i tested both of them and i think U41.xx isn´t much faster than S41.xx ... 1-2 minutes on a WU.
It's a small edge, but as
)
It's a small edge, but as fast as the app is only a massive change could shave more than a few minutes off.
I've been running S41.07(HT)
)
I've been running S41.07(HT) - based on little more than I thought it was probably the most appropriate version, though I don't know whether there was any real benefit of using it instead of the standard S41.07 client.
The question now is - should I change to S41.08 instead?
Thanks for any advice. Note - the computers running the clients require HT to be enabled for other purpose, so disiabling HT isn't a viable option in this case.
.
RE: I've been running
)
I've found that on my P4 3.0 HT(on) that the fastest version was the S40.04. Here is a previous post in this thread I'd wrote about comparing the S40 & S41HT. After trying the newest S40, S41, & S41HT versions since that post I still find the S40 being the fastest for the 3.0 HT (I leave the HT on too because my computers get used for everything and are not 'dedicated' for crunching.) On the other hand, the new S41.08 (and S41.07 non-HT) are the fastest for the Cel. 2.4. You can take it as advice, but these are my observations.