Processor GFlops Compilation

[newtekie1]

Table Updated.

[Athlonite]

average GFlops = all runs added together \ divided by number of runs

so your twilyth your avg gflops would workout to be 138.8102 \ 5 = 27.76204

[AphexDreamer]

I'm sad I can run folding@Home for days and play games and what not with out a single crash but I can't do more than one run of this burn in test.

1.48 votls @ 4095mhz. Any tips to make it stable? Temps 43C Idle.

It's not really stable unless you can run at full load for at the very least, several hours, w/o any problems. You'll probably need to tweak a few of the parameters that are normally involved in getting a good overclock beyond where they are now - if it's possible to do safely. If not, you'll have to settle for a slower speed.

I think that question is worthy of it's own thread though and I don't think the OP would want this thread to go off on that kind of tangent. I'm just guessing but it is pretty narrowly focused, and seriously, I think you would get a good response.

[Mussels]

Originally Posted by AphexDreamer

I'm sad I can run folding@Home for days and play games and what not with out a single crash but I can't do more than one run of this burn in test.

1.48 votls @ 4095mhz. Any tips to make it stable? Temps 43C Idle.

best make your own thread.

[Hardi]

just some undervolting stabilty testing, but might as well add this to the list too.

Hardi | AMD | Athlon II X3 435 @ 2.9GHz | 025.1375 | 08.668 | 64bit |

[exodusprime1337]

Exodusprime1337 | AMD | Amd Phenom II 1090t @ 4.265Ghz | 080.1186 | 64bit | 6 threads on water 1.50v.

[AphexDreamer]

Originally Posted by twilyth

It's not really stable unless you can run at full load for at the very least, several hours, w/o any problems. You'll probably need to tweak a few of the parameters that are normally involved in getting a good overclock beyond where they are now - if it's possible to do safely. If not, you'll have to settle for a slower speed.

I think that question is worthy of it's own thread though and I don't think the OP would want this thread to go off on that kind of tangent. I'm just guessing but it is pretty narrowly focused, and seriously, I think you would get a good response.

Originally Posted by Mussels

best make your own thread.

Ok Understood guys, however I think its temp issue for me. No matter what voltage I give it as soon as it hits 63C it BSODs. So bottom line is I need a better cooling solution.

[trickson]

Did mine on MAX settings .

trickson Intel Q9650 @ 4.0GHz 47.3793 64-bit

I see mine there but you have the wrong OS listed and I updated some as well .

[newtekie1]

Table updated.

[AphexDreamer]

YAY Finally got it stable! And its faster than before!

AphexDreamer | AMD | Amd Phenom II 965 @ 4047Mhz | 130.71 | 64bit | 4 threads Air 1.496 Volts.





I'm so happy

AVG Came out to be 52.444 GFlops

[newtekie1]

Table updated.

[Mussels]

there was a typo in mine (NB clocks) so i went and abused my mod powers and edited it in the first post.

(typo was my fault, btw)

[Hayder_Master]

nice thread, i will upload my results soon

[Mussels]

could someone with an i7 run this test, but do it with and without HT enabled? i'm curious about how the Gflops per thread will be affected, and how much performance HT actually adds in this kind of situation.


for example, sorting by average Gflops and dividing by threads:

10.77 on my cores, averaged

fitseries xeon at 4.5Ghz (900Mhz faster!)
7.59 Gflops per thread

what makes me curious here, is that if the HT cores are doing really badly at this test, then disabling HT would let us know how much of the score is from the real cores, how much from the HT, and let us get a performance per core comparison between i7 and thuban

[joytime360]

i will try it

[newtekie1]

I'll try it.

[newtekie1]

Wow some interesting results for you Mussels:

newtekie1 | Intel | Core i7 875K@3.6GHz | 043.9826 | 12.2174 | 64-bit| Hyperthreading ON and 8 Threads
newtekie1 | Intel | Core i7 875K@3.6GHz | 048.9807 | 13.6057 | 64-bit| Hyperthreading ON and 4 Threads
newtekie1 | Intel | Core i7 875K@3.6GHz | 049.3956 | 13.7210 | 64-bit| Hyperthreading OFF and 8 Threads
newtekie1 | Intel | Core i7 875K@3.6GHz | 050.6025 | 14.0562 | 64-bit| Hyperthreading OFF and 4 Threads

[Mussels]

k so the ones i'm interested in:

HT off/4 threads


newtekie1 | Intel | Core i7 875K@3.6GHz | 050.6025 | 14.0562 | Hyperthreading OFF and 4 Threads

60.60 / 8 = 15.15 per thread

thats about what i'd expect - per core, you're faster than my thuban... but you got 2 less cores, so my multithreaded performance leaps ahead.

HT on/8 threads


newtekie1 | Intel | Core i7 875K@3.6GHz | 043.9826 | 12.2174 | Hyperthreading ON and 8 Threads

43.98 / 8 = 5.4975 per thread


so.... not only are the HT threads slower than the real cores, they slow down the whole thing overall?

If i was on i7 right now, i'd be disabling HT for sure :P

[newtekie1]

You mean 50.60 / 4 = 12.65

Still faster per core, but your extra 2 physical cores pushes you ahead of me.

And hyperthreading really seems to throw this application for a loop. Even with HT off and 8 threads, the Windows scheduler seems to do a better job switching between the threads then Hyperthreading does.

[Mussels]

Originally Posted by newtekie1

You mean 50.60 / 4 = 12.65

Still faster per core, but your extra 2 physical cores pushes you ahead of me.

And hyperthreading really seems to throw this application for a loop. Even with HT off and 8 threads, the Windows scheduler seems to do a better job switching between the threads then Hyperthreading does.

i'm dividing by threads, not by cores. if i still screwed something up, lemme know.. its late here

[somebody]

Originally Posted by Mussels

so.... not only are the HT threads slower than the real cores, they slow down the whole thing overall?

If i was on i7 right now, i'd be disabling HT for sure :P

AFAIK that's not how it works. One core has 2 threads when HT is on, it's still one core but by having 2 threads while one thread is not working the execution units 100% it leaves some room for the other thread to execute. If you were to run linpack with only one thread on core 1 say, the result would probably be very close if you did the same on just using the second thread on core 1. Problem is with linpack running on one thread it will already saturates the core leaving not much in the way of spare execution cycles so using 2 threads on the same core will likely result in ~half the execution speed for each thread wrt using just one thread on one core.

Something else you might find surprising is the core temperature differential when running linpack on just one core.

[trickson]

I redid mine this time normal setting and with windows 7 64 bit .

trickson Intel Q9650 @ 4.0GHz 53.3787 13.34 64-bit

[newtekie1]

Originally Posted by Mussels

i'm dividing by threads, not by cores. if i still screwed something up, lemme know.. its late here

The one you are using that scored 50.60 is 4 cores and 4 threads. With 4 cores and 8 threads I got 49.39.

[newtekie1]

Table Updated.

Very nice score on that Q9650 Trickson! Amazing what 775 is still capable of