Perpetual benchmark thread... - Page 16

S485122 · 2008-11-05, 06:48

Quote:

Originally Posted by petrw1

I'm pretty much sold on the idea of 64 Bit Vista (I only shudder slightly as I type this).

If I do so is there any noticeable GIMPS benefit to having MORE THAN 4GB of RAM (DDR2-1066)?

I know of some gamers using Vista 64 bits : they say there are no problems.

About the memory, at the moment 2GB is enough for Pime95 v25.7 (32 and 67 bits) : because of a bug it will not accept allocation of more than 1703 MB. When that bug is fixed 8GB or more will be usefull during the memory intensive phases (P-1 2 and ECM...) For that reason I still run one CPU on a quadcore with Prime95 v24.14.

Jacob

starrynte · 2008-11-06, 02:42

how do you run a benchmark WITHOUT prime95 submitting results automatically to primenet? (prime95 v25.7)

petrw1 · 2008-11-06, 05:17

Quote:

Originally Posted by S485122

I know of some gamers using Vista 64 bits : they say there are no problems.

About the memory, at the moment 2GB is enough for Pime95 v25.7 (32 and 67 bits) : because of a bug it will not accept allocation of more than 1703 MB. When that bug is fixed 8GB or more will be usefull during the memory intensive phases (P-1 2 and ECM...) For that reason I still run one CPU on a quadcore with Prime95 v24.14.

Jacob

I has surmised that most ECM tests have stopped at 4,294,000,000 because that is the 32 bit limit but with a 64 bit OS and more memory available that could be MUCH higher.

gdf · 2008-11-06, 10:02

Just installed V25.7 (upgraded from V24.x)

Code:

Intel(R) Pentium(R) D CPU 2.80GHz
CPU speed: 300.00 MHz, 2 cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 16 KB
L2 cache size: 1 MB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 32-bit version 25.7, RdtscTiming=1
Best time for 768K FFT length: 26.076 ms.
Best time for 896K FFT length: 31.651 ms.
Best time for 1024K FFT length: 36.140 ms.
Best time for 1280K FFT length: 44.623 ms.
Best time for 1536K FFT length: 54.022 ms.
Best time for 1792K FFT length: 64.489 ms.
Best time for 2048K FFT length: 72.079 ms.
Best time for 2560K FFT length: 96.052 ms.
Best time for 3072K FFT length: 115.435 ms.
Best time for 3584K FFT length: 138.142 ms.
Best time for 4096K FFT length: 154.372 ms.
Best time for 5120K FFT length: 196.196 ms.
Best time for 6144K FFT length: 249.801 ms.
Best time for 7168K FFT length: 299.634 ms.
Best time for 8192K FFT length: 331.630 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 15.539 ms.
Best time for 896K FFT length: 18.181 ms.
Best time for 1024K FFT length: 21.830 ms.
Best time for 1280K FFT length: 24.538 ms.
Best time for 1536K FFT length: 29.307 ms.
Best time for 1792K FFT length: 34.487 ms.
Best time for 2048K FFT length: 39.034 ms.
Best time for 2560K FFT length: 50.853 ms.
Best time for 3072K FFT length: 62.487 ms.
Best time for 3584K FFT length: 76.587 ms.
Best time for 4096K FFT length: 89.447 ms.
Best time for 5120K FFT length: 121.796 ms.
Best time for 6144K FFT length: 136.877 ms.
Best time for 7168K FFT length: 173.027 ms.
Best time for 8192K FFT length: 200.139 ms.
Best time for 58 bit trial factors: 9.880 ms.
Best time for 59 bit trial factors: 9.914 ms.
Best time for 60 bit trial factors: 9.792 ms.
Best time for 61 bit trial factors: 9.863 ms.
Best time for 62 bit trial factors: 13.488 ms.
Best time for 63 bit trial factors: 13.509 ms.
Best time for 64 bit trial factors: 16.144 ms.
Best time for 65 bit trial factors: 16.241 ms.
Best time for 66 bit trial factors: 16.005 ms.
Best time for 67 bit trial factors: 16.258 ms.

Not bad for a 300MHz???? machine ?

Jeff Gilchrist · 2008-11-06, 17:15

Quote:

Originally Posted by starrynte

how do you run a benchmark WITHOUT prime95 submitting results automatically to primenet? (prime95 v25.7)

I didn't think the client did submit it automatically. You need to copy and paste the benchmark info from your results.txt file into a form on the v5 server for it to appear on the PrimeNet server.

sdbardwick · 2008-11-06, 17:55

Left to its own devices, v25.7 automatically submits benchmark data to PrimeNet; just did one on my laptop.

James Heinrich · 2008-11-06, 23:48

Quote:

Originally Posted by sdbardwick

Left to its own devices, v25.7 automatically submits benchmark data to PrimeNet

Somewhat related is a bug I have noticed in the v5 server:
http://www.mersenneforum.org/showpos...&postcount=161
This only applies if a full benchmark is enabled (via override setting in prime.txt), so really shouldn't affect anybody but us weirdos who like to break things

Meikel · 2008-11-12, 22:48

I can't believe it, but it seems like I am the first one to post a test with a brand-new Core i7 here...

The machine is brand new, no overclocking whatsoever. 6 GB of RAM installed (3x2GB DDR3-1066), Win Vista 64Bit.

I wonder, why the benchmark does not try to start 4 threads on 4 physical CPUs - it stubbornly uses 4 threads on two physical CPUs (and two hyperthreaded ones) - and all 4 real cores only get used with 8 threads.

This way, the "sweet spot" seems to be 6 cores on 3 cpus...

George, should you search for someone to try out some special tweaks maybe for SSE4.2 or the way the Core i7 has its three levels of cache organized ... don't hesitate to contact me!

Code:

Intel(R) Core(TM) i7 CPU         940  @ 2.93GHz
CPU speed: 2940.03 MHz, 4 hyperthreaded cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2, SSE4
L1 cache size: 32 KB
L2 cache size: 256 KB, L3 cache size: 8064 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 64-bit version 25.7, RdtscTiming=1
Best time for 768K FFT length: 13.368 ms.
Best time for 896K FFT length: 16.293 ms.
Best time for 1024K FFT length: 18.486 ms.
Best time for 1280K FFT length: 23.668 ms.
Best time for 1536K FFT length: 28.368 ms.
Best time for 1792K FFT length: 34.041 ms.
Best time for 2048K FFT length: 38.716 ms.
Best time for 2560K FFT length: 50.189 ms.
Best time for 3072K FFT length: 60.043 ms.
Best time for 3584K FFT length: 72.395 ms.
Best time for 4096K FFT length: 81.989 ms.
Best time for 5120K FFT length: 107.426 ms.
Best time for 6144K FFT length: 128.328 ms.
Best time for 7168K FFT length: 148.590 ms.
Best time for 8192K FFT length: 167.032 ms.
Timing FFTs using 2 threads on 1 physical CPUs.
Best time for 768K FFT length: 6.689 ms.
Best time for 896K FFT length: 8.195 ms.
Best time for 1024K FFT length: 9.917 ms.
Best time for 1280K FFT length: 11.875 ms.
Best time for 1536K FFT length: 14.864 ms.
Best time for 1792K FFT length: 18.063 ms.
Best time for 2048K FFT length: 19.861 ms.
Best time for 2560K FFT length: 25.788 ms.
Best time for 3072K FFT length: 31.210 ms.
Best time for 3584K FFT length: 38.233 ms.
Best time for 4096K FFT length: 41.931 ms.
Best time for 5120K FFT length: 54.641 ms.
Best time for 6144K FFT length: 65.475 ms.
Best time for 7168K FFT length: 79.814 ms.
Best time for 8192K FFT length: 90.862 ms.
Timing FFTs using 4 threads on 2 physical CPUs.
Best time for 768K FFT length: 7.477 ms.
Best time for 896K FFT length: 8.506 ms.
Best time for 1024K FFT length: 11.261 ms.
Best time for 1280K FFT length: 11.120 ms.
Best time for 1536K FFT length: 13.407 ms.
Best time for 1792K FFT length: 16.011 ms.
Best time for 2048K FFT length: 18.202 ms.
Best time for 2560K FFT length: 24.070 ms.
Best time for 3072K FFT length: 30.140 ms.
Best time for 3584K FFT length: 35.505 ms.
Best time for 4096K FFT length: 39.085 ms.
Best time for 5120K FFT length: 50.311 ms.
Best time for 6144K FFT length: 61.976 ms.
Best time for 7168K FFT length: 75.375 ms.
Best time for 8192K FFT length: 84.356 ms.
Timing FFTs using 6 threads on 3 physical CPUs.
Best time for 768K FFT length: 5.492 ms.
Best time for 896K FFT length: 5.965 ms.
Best time for 1024K FFT length: 8.925 ms.
Best time for 1280K FFT length: 7.579 ms.
Best time for 1536K FFT length: 9.484 ms.
Best time for 1792K FFT length: 11.080 ms.
Best time for 2048K FFT length: 12.399 ms.
Best time for 2560K FFT length: 15.691 ms.
Best time for 3072K FFT length: 18.883 ms.
Best time for 3584K FFT length: 22.388 ms.
Best time for 4096K FFT length: 26.680 ms.
Best time for 5120K FFT length: 32.753 ms.
Best time for 6144K FFT length: 43.160 ms.
Best time for 7168K FFT length: 53.001 ms.
Best time for 8192K FFT length: 63.023 ms.
Timing FFTs using 8 threads on 4 physical CPUs.
Best time for 768K FFT length: 5.892 ms.
Best time for 896K FFT length: 6.615 ms.
Best time for 1024K FFT length: 8.571 ms.
Best time for 1280K FFT length: 7.954 ms.
Best time for 1536K FFT length: 9.800 ms.
Best time for 1792K FFT length: 12.616 ms.
Best time for 2048K FFT length: 13.749 ms.
Best time for 2560K FFT length: 16.434 ms.
Best time for 3072K FFT length: 18.823 ms.
Best time for 3584K FFT length: 23.083 ms.
Best time for 4096K FFT length: 28.770 ms.
Best time for 5120K FFT length: 33.957 ms.
Best time for 6144K FFT length: 44.094 ms.
Best time for 7168K FFT length: 54.124 ms.
Best time for 8192K FFT length: 63.830 ms.
Best time for 58 bit trial factors: 2.534 ms.
Best time for 59 bit trial factors: 2.538 ms.
Best time for 60 bit trial factors: 2.854 ms.
Best time for 61 bit trial factors: 2.989 ms.
Best time for 62 bit trial factors: 3.492 ms.
Best time for 63 bit trial factors: 4.391 ms.
Best time for 64 bit trial factors: 4.673 ms.
Best time for 65 bit trial factors: 4.860 ms.
Best time for 66 bit trial factors: 4.872 ms.
Best time for 67 bit trial factors: 4.746 ms.

Meikel · 2008-11-12, 23:04

Hi, it's me again. I've now switched off hyperthreading in the BIOS - the benchmarking now thinks, that i have two hyperthreaded cores. This is wrong, of course: I have four "full" cores now.

Still, this gives me the fastest results yet:

Code:

Compare your results to other computers at http://www.mersenne.org/bench.htm
Intel(R) Core(TM) i7 CPU         940  @ 2.93GHz
CPU speed: 2940.07 MHz, 2 hyperthreaded cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2, SSE4
L1 cache size: 32 KB
L2 cache size: 256 KB, L3 cache size: 8064 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 64-bit version 25.7, RdtscTiming=1
Best time for 768K FFT length: 12.747 ms.
Best time for 896K FFT length: 15.277 ms.
Best time for 1024K FFT length: 17.934 ms.
Best time for 1280K FFT length: 23.013 ms.
Best time for 1536K FFT length: 28.270 ms.
Best time for 1792K FFT length: 33.592 ms.
Best time for 2048K FFT length: 37.665 ms.
Best time for 2560K FFT length: 47.850 ms.
Best time for 3072K FFT length: 58.709 ms.
Best time for 3584K FFT length: 70.034 ms.
Best time for 4096K FFT length: 79.401 ms.
Best time for 5120K FFT length: 100.429 ms.
Best time for 6144K FFT length: 121.148 ms.
Best time for 7168K FFT length: 145.074 ms.
Best time for 8192K FFT length: 163.836 ms.
Timing FFTs using 2 threads on 1 physical CPUs.
Best time for 768K FFT length: 6.676 ms.
Best time for 896K FFT length: 7.966 ms.
Best time for 1024K FFT length: 9.456 ms.
Best time for 1280K FFT length: 11.804 ms.
Best time for 1536K FFT length: 14.461 ms.
Best time for 1792K FFT length: 17.122 ms.
Best time for 2048K FFT length: 19.235 ms.
Best time for 2560K FFT length: 24.964 ms.
Best time for 3072K FFT length: 30.479 ms.
Best time for 3584K FFT length: 36.404 ms.
Best time for 4096K FFT length: 41.065 ms.
Best time for 5120K FFT length: 51.963 ms.
Best time for 6144K FFT length: 62.756 ms.
Best time for 7168K FFT length: 75.391 ms.
Best time for 8192K FFT length: 85.677 ms.
Timing FFTs using 4 threads on 2 physical CPUs.
Best time for 768K FFT length: 4.953 ms.
Best time for 896K FFT length: 5.564 ms.
Best time for 1024K FFT length: 7.762 ms.
Best time for 1280K FFT length: 6.416 ms.
Best time for 1536K FFT length: 7.598 ms.
Best time for 1792K FFT length: 8.918 ms.
Best time for 2048K FFT length: 10.136 ms.
Best time for 2560K FFT length: 13.041 ms.
Best time for 3072K FFT length: 15.907 ms.
Best time for 3584K FFT length: 19.533 ms.
Best time for 4096K FFT length: 21.483 ms.
Best time for 5120K FFT length: 27.949 ms.
Best time for 6144K FFT length: 34.438 ms.
Best time for 7168K FFT length: 42.808 ms.
Best time for 8192K FFT length: 51.298 ms.
Best time for 58 bit trial factors: 2.533 ms.
Best time for 59 bit trial factors: 2.536 ms.
Best time for 60 bit trial factors: 2.837 ms.
Best time for 61 bit trial factors: 2.981 ms.
Best time for 62 bit trial factors: 3.467 ms.
Best time for 63 bit trial factors: 4.313 ms.
Best time for 64 bit trial factors: 4.600 ms.
Best time for 65 bit trial factors: 4.794 ms.
Best time for 66 bit trial factors: 4.752 ms.
Best time for 67 bit trial factors: 4.729 ms.

stars10250 · 2008-11-13, 15:55

Thanks much for the numbers, and please keep them coming! I have some questions...

While the timing numbers for employing various threads is interesting, isn't one of the biggest questions the ability of 4 cores to accomplish work simultaneously on 4 separate exponents? I don't understand how to interpret the benchmark results of 4 threads all working on the same FFT compared to 4 threads working on 4 unique FFTs. Don't we all want to use a quad-core chip to calculate 4 exponents (or maybe 8 with hyperthreading) simultaneously rather than just 1 exponent as fast as possible? I guess what I'm getting at is I'd like to know if the i7 deals better with the memory bottle-neck that people reported with existing quad-core chips running 4 exponents (users reported about 3 cores-worth of performance when running all 4 cores).

Your result of ~50 ms for a 2560K FFT is consistent with another post I saw, but not as good as I was hoping for. My E8500 already gives me this performance level (with its admittedly much larger L2 cache but with only 2 cores). However, if i7 can maintain this across 4 cores simultaneously then it would beat the "3 cores worth of performance" seen on current quads.

Lastly, do the listed performance specs make you guys want to run out and buy an i7, or would better results (and price) be had by overclocking the existing line of hardware. I'm leaning toward the latter.

petrw1 · 2008-11-13, 16:03

The Core i7 or at least, any current motherboard for the Core i7 as of November 13, 2008 (two ASUS boards, the intel DX58SO, gigabyte board -- data available from manufacturer website) does not support ECC memory.[citation needed] Some experts, such as Daniel Bernstein[10], recommend that systems without ECC support not be used for scientific computing, and not in general unless the user does not mind errors in critical data.

mdettweiler · 2008-11-13, 17:03

Quote:

Originally Posted by petrw1

The Core i7 or at least, any current motherboard for the Core i7 as of November 13, 2008 (two ASUS boards, the intel DX58SO, gigabyte board -- data available from manufacturer website) does not support ECC memory.[citation needed] Some experts, such as Daniel Bernstein[10], recommend that systems without ECC support not be used for scientific computing, and not in general unless the user does not mind errors in critical data.

Pardon my stupidity but...what the heck is ECC?

(Yeah, I know, I should probably Google this, but I'll probably get a better and clearer answer here.

)

Phantomas · 2008-11-13, 18:46

Quote:

Pardon my stupidity but...what the heck is ECC?

best answer I found so far http://en.wikipedia.org/wiki/Dynamic...ror_correction

Meikel · 2008-11-13, 18:54

More numbers!? OK... scalability of the Core i7.

Of course, you are right, stars10250, normally I also would run multiple different LL tests and not one multi-threaded one. So, how does that scale?

I don't use the benchmark for that, but use my regular work-units; double-checking in the 23000000-range currently (FFT-Length 1280K). I let it run for a minute or so and quote the best per-iteration time here. Then I'll start further tests on the other cores and watch, if and how the times for the first test increase.

This setting in local.txt guarantees, that I only use the real CPUs and not the hyperthreaded ones (HT was activated in BIOS):
[Worker #1]
Affinity=0
ThreadsPerTest=1

[Worker #2]
Affinity=2
ThreadsPerTest=1

[Worker #3]
Affinity=4
ThreadsPerTest=1

[Worker #4]
Affinity=6
ThreadsPerTest=1

And these are the results:

1 Test : Per iteration time 23.057ms
2 Tests: Per iteration time 23.714ms
3 Tests: Per iteration time 23.988ms
4 Tests: Per iteration time 24.240ms

I think, thats not too shabby. But now I am interested in Hyperthreading. What if I start 8 different tests? Of course, the times are expected to go up significantly. But I would hope for something a bit LESS than factor 2, as HT is supposed to improve utilization of every core. So, I add this to local.txt:
[Worker #5]
Affinity=1
ThreadsPerTest=1

[Worker #6]
Affinity=3
ThreadsPerTest=1

[Worker #7]
Affinity=5
ThreadsPerTest=1

[Worker #8]
Affinity=7
ThreadsPerTest=1

and let it go.

8 Tests: Per iteration time 48.920ms

So, thats 200% workload on each core and the result is about 200% time on each core. No gain from HT. But on the other hand: Not much of a loss, which is interesting, because with 2 threads per core, each one only has 256K of L2-cache!

Maybe it would be optimal to do 4 different tests and have each test run on two threads on the same core? That should probably avoid the L2-cache problem, because the two threads on the same FFT would also use the same L2-data, right? Unfortunately, the second worker threads can't be assigned to CPU numbers, and I can observe, that they tend to jump from core to core.

Can anyone tell me, how to assign the affinity of the helper thread?

mdettweiler · 2008-11-13, 20:38

Quote:

Originally Posted by Phantomas

best answer I found so far http://en.wikipedia.org/wiki/Dynamic...ror_correction

Ah, thanks for the link! It was quite enlightening.

Based on what this article said, though--namely, that most modern computers are *not* equipped with ECC memory--it would seem that Core i7 motherboards' not supporting ECC memory is not at all uncommon, and thus this would not render them any less suitable for GIMPS use.

stars10250 · 2008-11-14, 03:42

Thanks for the new i7 numbers Meikel. Your results seem to indicate true quad performance, in which all 4 cores have nearly identical throughput.

I'd like to build a new number cruncher (~$1k) so I've been sketching out the performance tradeoffs of oc i7 vs oc penryn vs a non-oc q6600 and I came to a funny conclusion. The first two systems are close in price and performance, depending on the build and oc, and can achieve nearly twice the performance of a non-oc q6600 (with its memory bottleneck)...quite a feat. But the q6600 system parts are becoming so cheap (even before i7 release) that one could buy three systems for the price of either the oc i7 or oc penryn. And three q6600s will outperform either of the other two systems readily. Yes, three will consume more power than one (which is a non-negligible cost), and you have to deal with everything times three, but it surprised me. The oc'd i7 and penryn require expensive cpu's & motherboards, expensive DDR3 ram, a video card, quality power supplies, sata HD, etc. I'm leaving out a bunch of details here, but I hope you see my point. The only thing is, my girlfriend is going to kill me if I order three more computers!

petrw1 · 2008-11-14, 17:22

Quote:

Originally Posted by stars10250

Yes, three will consume more power than one (which is a non-negligible cost), and you have to deal with everything times three, but it surprised me. The oc'd i7 and penryn require expensive cpu's & motherboards, expensive DDR3 ram, a video card, quality power supplies, sata HD, etc. I'm leaving out a bunch of details here, but I hope you see my point. The only thing is, my girlfriend is going to kill me if I order three more computers!

I have a similar situation here. My employer (a large Government corp.) has a PC refresh policy that gives everyone a BRAND NEW, CURRENT PC every 3 or 4 years ... as for the old PC's (not that old): some are given to schools and the rest are offered for sale to the public (employees included) cheap. For example they are now removing and replacing 3 Ghz single core PCs and selling them (monitor and mouse included) for about $200.

I just ordered a new home PC (Q9550) but first toungue-in-cheek suggested to my wife that I could get 10 slightly used PCs from work for the same price. She would just have to clear out the storage room and install a second air conditioner and separate power box there first. I too was unsuccessful in convincing her. Realistically, though I don't really want a whole storage room full either.

fivemack · 2008-11-14, 18:40

Quote:

Originally Posted by petrw1

The Core i7 or at least, any current motherboard for the Core i7 as of November 13, 2008 (two ASUS boards, the intel DX58SO, gigabyte board -- data available from manufacturer website) does not support ECC memory.[citation needed] Some experts, such as Daniel Bernstein[10], recommend that systems without ECC support not be used for scientific computing, and not in general unless the user does not mind errors in critical data.

I am not convinced by this; for example, I have just run a very error-sensitive calculation for 850 hours on a quad-core Phenom without ECC, and it completed without trouble.

fivemack · 2008-11-14, 18:43

Quote:

Originally Posted by stars10250

Thanks for the new i7 numbers Meikel. Your results seem to indicate true quad performance, in which all 4 cores have nearly identical throughput.

I'd like to build a new number cruncher (~$1k) so I've been sketching out the performance tradeoffs of oc i7 vs oc penryn vs a non-oc q6600 and I came to a funny conclusion. The first two systems are close in price and performance, depending on the build and oc, and can achieve nearly twice the performance of a non-oc q6600 (with its memory bottleneck)...quite a feat. But the q6600 system parts are becoming so cheap (even before i7 release) that one could buy three systems for the price of either the oc i7 or oc penryn.

Have you also factored in that at least two of the three Q6600 don't need a hard disc or graphics card? I think it'll be a while before the best price-performance doesn't come from the cheapest available quad-core - though I admit that I have three at the moment, and getting another three would make the study Very Full and Very Loud.

db597 · 2008-11-15, 11:50

Quote:

Originally Posted by stars10250

... oc i7 vs oc penryn vs a non-oc q6600

Why wouldn't you oc the q6600 as well? In fact, it's acknowledged to be an even better overclocker than the penryns. Most are Prime stable at 3-3.2Ghz with only a minimal bump in voltage. If you consider there's easily another 30% performance for free in there, the value proposition of the q6600 is even better.

stars10250 · 2008-11-15, 14:19

fivemack, I've never booted xp from a network so I'm not sure how to run them diskless. I looked on google, and it didn't seem straightforward. I have lots of old IDE drives I could use, though I do still like the diskless idea as it would be tidy and draw less power. If you have any good instructions, please post a link. I'm reasonably computer savvy, I've just never done it. I'd be using xp (my xp pro disk broke!). I ran ubuntu for a bit on one of my machines, but I disliked the feedback I got with prime95 so I reverted to xp. The motherboard that I was considering has built in video, so that's not an issue.

db597, I'm reluctant to oc because I've never done it and generally one needs a more expensive motherboard to do so. With the more expensive motherboards, they tend to use better ram (I was considering cheap DDR2 PC6400), and then often a video card. I suppose there are tons of instructions for the oc'd q6600 by now, so I guess I should try it. In my price estimate, I was considering an intel motherboard that costs around $85 which people have said doesn't oc. If you know an inexpensive one to use instead, let me know. I'll start researching this, as I've also heard good things about its oc ability.

On convincing my girlfriend, she said it was ok as long as I got rid of an equivalent number of older computers (which I was going to do anyway!). Power/cooling is an issue, but not right now in the cold months :)

db597 · 2008-11-15, 14:54

Quote:

Originally Posted by stars10250

db597, I'm reluctant to oc because I've never done it and generally one needs a more expensive motherboard to do so. With the more expensive motherboards, they tend to use better ram (I was considering cheap DDR2 PC6400), and then often a video card. I suppose there are tons of instructions for the oc'd q6600 by now, so I guess I should try it. In my price estimate, I was considering an intel motherboard that costs around $85 which people have said doesn't oc. If you know an inexpensive one to use instead, let me know. I'll start researching this, as I've also heard good things about its oc ability.

Overclocking these days is pretty simple - the options are all in the BIOS. Also, the P45 motherboards are all rated for 1333MHz FSB, so with the 1066MHz FSB Q6600 you're not even overclocking the motherboard at all (only the CPU).

The Penryns and the Q6600 are both going to be similarly bottlenecked by the RAM (both don't have low latency integrated memory controllers) - so if you're going the cheap DDR2 route, I'd still advise going for DDR1066. The price difference isn't huge these days, and it makes a significant impact on performance. Either way, they're still a lot cheaper than DDR3.

In terms of motherboards, I've been very happy with the Gigabyte EP45 series. You might want to have a look.

stars10250 · 2008-11-21, 12:38

I ordered the EP45 last night with a Q6600, DDR21066, good heat sink, cheap video, etc. The build price (after a ton of rebates from newegg and w/o case) is around $360. I've also spent the time reading up on overclocking, but I'm still a bit nervous. I guess worse-case I'll just run it stock if I can't figure out oc. If things work out well I'll by a second one, as my goal is to best the i7's performance for less $ (including a margin for the added electrical costs). I'll post detailed specs when my experiment is complete.

Cruelty · 2008-11-21, 20:43

Quote:

Originally Posted by stars10250

I ordered the EP45 last night with a Q6600, DDR21066, good heat sink, cheap video, etc. The build price (after a ton of rebates from newegg and w/o case) is around $360. I've also spent the time reading up on overclocking, but I'm still a bit nervous. I guess worse-case I'll just run it stock if I can't figure out oc. If things work out well I'll by a second one, as my goal is to best the i7's performance for less $ (including a margin for the added electrical costs). I'll post detailed specs when my experiment is complete.

Overclocking is very simple, at least on my EP45-DS4 motherboard - just set FSB to 400MHz and CPU multiplier to 8 and you'll have a 3.2GHz machine.

petrw1 · 2008-11-24, 04:20

Q9550 2.83 (NOT OC'd)
4GB Ram DDR2-1066
Vista-64
L1 Cache 32KB
L2 Cache 6 MB

Code:

Compare your results to other computers at http://www.mersenne.org/bench.htm
Intel Pentium III Xeon processor
CPU speed: 2833.01 MHz, 4 cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2, SSE4
L1 cache size: 32 KB
L2 cache size: 6 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 256
Prime95 64-bit version 25.7, RdtscTiming=1
Best time for 768K FFT length: 14.315 ms.
Best time for 896K FFT length: 17.312 ms.
Best time for 1024K FFT length: 19.695 ms.
Best time for 1280K FFT length: 24.660 ms.
Best time for 1536K FFT length: 30.352 ms.
Best time for 1792K FFT length: 36.097 ms.
Best time for 2048K FFT length: 40.171 ms.
Best time for 2560K FFT length: 52.687 ms.
Best time for 3072K FFT length: 64.813 ms.
Best time for 3584K FFT length: 77.192 ms.
Best time for 4096K FFT length: 86.378 ms.
Best time for 5120K FFT length: 110.774 ms.
Best time for 6144K FFT length: 134.046 ms.
Best time for 7168K FFT length: 162.474 ms.
Best time for 8192K FFT length: 177.730 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 7.583 ms.
Best time for 896K FFT length: 9.068 ms.
Best time for 1024K FFT length: 10.481 ms.
Best time for 1280K FFT length: 12.761 ms.
Best time for 1536K FFT length: 15.739 ms.
Best time for 1792K FFT length: 18.723 ms.
Best time for 2048K FFT length: 21.028 ms.
Best time for 2560K FFT length: 27.472 ms.
Best time for 3072K FFT length: 33.636 ms.
Best time for 3584K FFT length: 39.994 ms.
Best time for 4096K FFT length: 45.127 ms.
Best time for 5120K FFT length: 57.362 ms.
Best time for 6144K FFT length: 69.912 ms.
Best time for 7168K FFT length: 84.464 ms.
Best time for 8192K FFT length: 93.099 ms.
Timing FFTs using 3 threads.
Best time for 768K FFT length: 10.788 ms.
Best time for 896K FFT length: 11.471 ms.
Best time for 1024K FFT length: 21.142 ms.
Best time for 1280K FFT length: 11.142 ms.
Best time for 1536K FFT length: 13.561 ms.
Best time for 1792K FFT length: 16.093 ms.
Best time for 2048K FFT length: 18.102 ms.
Best time for 2560K FFT length: 23.742 ms.
Best time for 3072K FFT length: 28.923 ms.
Best time for 3584K FFT length: 33.718 ms.
Best time for 4096K FFT length: 37.996 ms.
Best time for 5120K FFT length: 46.522 ms.
Best time for 6144K FFT length: 56.103 ms.
Best time for 7168K FFT length: 67.684 ms.
Best time for 8192K FFT length: 74.683 ms.
Timing FFTs using 4 threads.
Best time for 768K FFT length: 10.147 ms.
Best time for 896K FFT length: 11.036 ms.
Best time for 1024K FFT length: 17.728 ms.
Best time for 1280K FFT length: 10.069 ms.
Best time for 1536K FFT length: 10.855 ms.
Best time for 1792K FFT length: 12.564 ms.
Best time for 2048K FFT length: 14.444 ms.
Best time for 2560K FFT length: 18.652 ms.
Best time for 3072K FFT length: 22.751 ms.
Best time for 3584K FFT length: 26.769 ms.
Best time for 4096K FFT length: 30.662 ms.
Best time for 5120K FFT length: 37.507 ms.
Best time for 6144K FFT length: 45.039 ms.
Best time for 7168K FFT length: 53.756 ms.
Best time for 8192K FFT length: 60.621 ms.
Best time for 58 bit trial factors: 2.995 ms.
Best time for 59 bit trial factors: 3.010 ms.
Best time for 60 bit trial factors: 3.414 ms.
Best time for 61 bit trial factors: 3.634 ms.
Best time for 62 bit trial factors: 4.097 ms.
Best time for 63 bit trial factors: 5.227 ms.
Best time for 64 bit trial factors: 5.676 ms.
Best time for 65 bit trial factors: 6.295 ms.
Best time for 66 bit trial factors: 6.244 ms.
Best time for 67 bit trial factors: 6.207 ms.

Can someone tell me why it thinks I have a PIII Xeon?

And why it thinks I only have 6MB L2 Cache? Jeff #371 also shows 6Mb. My vendor tells me I have 12 MB.

2008-11-06, 17:55	#381
sdbardwick Aug 2002 North San Diego County 1010101101₂ Posts	Left to its own devices, v25.7 automatically submits benchmark data to PrimeNet; just did one on my laptop. Last fiddled with by sdbardwick on 2008-11-06 at 17:55

2008-11-13, 15:55	#385
stars10250 Jul 2008 San Francisco, CA 3·67 Posts	Thanks much for the numbers, and please keep them coming! I have some questions... While the timing numbers for employing various threads is interesting, isn't one of the biggest questions the ability of 4 cores to accomplish work simultaneously on 4 separate exponents? I don't understand how to interpret the benchmark results of 4 threads all working on the same FFT compared to 4 threads working on 4 unique FFTs. Don't we all want to use a quad-core chip to calculate 4 exponents (or maybe 8 with hyperthreading) simultaneously rather than just 1 exponent as fast as possible? I guess what I'm getting at is I'd like to know if the i7 deals better with the memory bottle-neck that people reported with existing quad-core chips running 4 exponents (users reported about 3 cores-worth of performance when running all 4 cores). Your result of ~50 ms for a 2560K FFT is consistent with another post I saw, but not as good as I was hoping for. My E8500 already gives me this performance level (with its admittedly much larger L2 cache but with only 2 cores). However, if i7 can maintain this across 4 cores simultaneously then it would beat the "3 cores worth of performance" seen on current quads. Lastly, do the listed performance specs make you guys want to run out and buy an i7, or would better results (and price) be had by overclocking the existing line of hardware. I'm leaning toward the latter. Last fiddled with by stars10250 on 2008-11-13 at 15:57

2008-11-13, 16:03	#386
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 3×5×313 Posts	From Wiki The Core i7 or at least, any current motherboard for the Core i7 as of November 13, 2008 (two ASUS boards, the intel DX58SO, gigabyte board -- data available from manufacturer website) does not support ECC memory.[citation needed] Some experts, such as Daniel Bernstein[10], recommend that systems without ECC support not be used for scientific computing, and not in general unless the user does not mind errors in critical data.

2008-11-15, 14:19	#396
stars10250 Jul 2008 San Francisco, CA 311₈ Posts	fivemack, I've never booted xp from a network so I'm not sure how to run them diskless. I looked on google, and it didn't seem straightforward. I have lots of old IDE drives I could use, though I do still like the diskless idea as it would be tidy and draw less power. If you have any good instructions, please post a link. I'm reasonably computer savvy, I've just never done it. I'd be using xp (my xp pro disk broke!). I ran ubuntu for a bit on one of my machines, but I disliked the feedback I got with prime95 so I reverted to xp. The motherboard that I was considering has built in video, so that's not an issue. db597, I'm reluctant to oc because I've never done it and generally one needs a more expensive motherboard to do so. With the more expensive motherboards, they tend to use better ram (I was considering cheap DDR2 PC6400), and then often a video card. I suppose there are tons of instructions for the oc'd q6600 by now, so I guess I should try it. In my price estimate, I was considering an intel motherboard that costs around $85 which people have said doesn't oc. If you know an inexpensive one to use instead, let me know. I'll start researching this, as I've also heard good things about its oc ability. On convincing my girlfriend, she said it was ok as long as I got rid of an equivalent number of older computers (which I was going to do anyway!). Power/cooling is an issue, but not right now in the cold months :) Last fiddled with by stars10250 on 2008-11-15 at 14:31

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Perpetual "interesting video" thread...	Xyzzy	Lounge	43	2021-07-17 00:00
LLR benchmark thread	Oddball	Riesel Prime Search	5	2010-08-02 00:11
Perpetual I'm pi**ed off thread	rogue	Soap Box	19	2009-10-28 19:17
Perpetual autostereogram thread...	Xyzzy	Lounge	10	2006-09-28 00:36
Perpetual ECM factoring challenge thread...	Xyzzy	Factoring	65	2005-09-05 08:16

2008-11-06, 02:42	#377
starrynte Oct 2008 California 2²·59 Posts	how do you run a benchmark WITHOUT prime95 submitting results automatically to primenet? (prime95 v25.7)

2008-11-13, 18:54	#389
Meikel Nov 2008 3² Posts	More numbers!? OK... scalability of the Core i7. Of course, you are right, stars10250, normally I also would run multiple different LL tests and not one multi-threaded one. So, how does that scale? I don't use the benchmark for that, but use my regular work-units; double-checking in the 23000000-range currently (FFT-Length 1280K). I let it run for a minute or so and quote the best per-iteration time here. Then I'll start further tests on the other cores and watch, if and how the times for the first test increase. This setting in local.txt guarantees, that I only use the real CPUs and not the hyperthreaded ones (HT was activated in BIOS): [Worker #1] Affinity=0 ThreadsPerTest=1 [Worker #2] Affinity=2 ThreadsPerTest=1 [Worker #3] Affinity=4 ThreadsPerTest=1 [Worker #4] Affinity=6 ThreadsPerTest=1 And these are the results: 1 Test : Per iteration time 23.057ms 2 Tests: Per iteration time 23.714ms 3 Tests: Per iteration time 23.988ms 4 Tests: Per iteration time 24.240ms I think, thats not too shabby. But now I am interested in Hyperthreading. What if I start 8 different tests? Of course, the times are expected to go up significantly. But I would hope for something a bit LESS than factor 2, as HT is supposed to improve utilization of every core. So, I add this to local.txt: [Worker #5] Affinity=1 ThreadsPerTest=1 [Worker #6] Affinity=3 ThreadsPerTest=1 [Worker #7] Affinity=5 ThreadsPerTest=1 [Worker #8] Affinity=7 ThreadsPerTest=1 and let it go. 8 Tests: Per iteration time 48.920ms So, thats 200% workload on each core and the result is about 200% time on each core. No gain from HT. But on the other hand: Not much of a loss, which is interesting, because with 2 threads per core, each one only has 256K of L2-cache! Maybe it would be optimal to do 4 different tests and have each test run on two threads on the same core? That should probably avoid the L2-cache problem, because the two threads on the same FFT would also use the same L2-data, right? Unfortunately, the second worker threads can't be assigned to CPU numbers, and I can observe, that they tend to jump from core to core. Can anyone tell me, how to assign the affinity of the helper thread?

2008-11-14, 03:42	#391
stars10250 Jul 2008 San Francisco, CA 3·67 Posts	Thanks for the new i7 numbers Meikel. Your results seem to indicate true quad performance, in which all 4 cores have nearly identical throughput. I'd like to build a new number cruncher (~$1k) so I've been sketching out the performance tradeoffs of oc i7 vs oc penryn vs a non-oc q6600 and I came to a funny conclusion. The first two systems are close in price and performance, depending on the build and oc, and can achieve nearly twice the performance of a non-oc q6600 (with its memory bottleneck)...quite a feat. But the q6600 system parts are becoming so cheap (even before i7 release) that one could buy three systems for the price of either the oc i7 or oc penryn. And three q6600s will outperform either of the other two systems readily. Yes, three will consume more power than one (which is a non-negligible cost), and you have to deal with everything times three, but it surprised me. The oc'd i7 and penryn require expensive cpu's & motherboards, expensive DDR3 ram, a video card, quality power supplies, sata HD, etc. I'm leaving out a bunch of details here, but I hope you see my point. The only thing is, my girlfriend is going to kill me if I order three more computers!

2008-11-21, 12:38	#398
stars10250 Jul 2008 San Francisco, CA 201₁₀ Posts	I ordered the EP45 last night with a Q6600, DDR21066, good heat sink, cheap video, etc. The build price (after a ton of rebates from newegg and w/o case) is around $360. I've also spent the time reading up on overclocking, but I'm still a bit nervous. I guess worse-case I'll just run it stock if I can't figure out oc. If things work out well I'll by a second one, as my goal is to best the i7's performance for less $ (including a margin for the added electrical costs). I'll post detailed specs when my experiment is complete.