Perpetual benchmark thread... - Page 39

starrynte · 2008-11-30, 22:06

Timings without overclocking and on 24.14 (Times for 512K and 640K removed for easier comparison):

Quote:

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3060.56 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 16 KB
L2 cache size: 1024 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 768K FFT length: 27.172 ms.
Best time for 896K FFT length: 33.198 ms.
Best time for 1024K FFT length: 38.122 ms.
Best time for 1280K FFT length: 46.649 ms.
Best time for 1536K FFT length: 56.462 ms.
Best time for 1792K FFT length: 68.155 ms.
Best time for 2048K FFT length: 76.344 ms.
Best time for 2560K FFT length: 100.902 ms.
Best time for 3072K FFT length: 121.601 ms.
Best time for 3584K FFT length: 147.856 ms.
Best time for 4096K FFT length: 163.470 ms.
Best time for 58 bit trial factors: 9.453 ms.
Best time for 59 bit trial factors: 9.536 ms.
Best time for 60 bit trial factors: 9.343 ms.
Best time for 61 bit trial factors: 9.450 ms.
Best time for 62 bit trial factors: 13.165 ms.
Best time for 63 bit trial factors: 13.158 ms.
Best time for 64 bit trial factors: 15.198 ms.
Best time for 65 bit trial factors: 15.172 ms.
Best time for 66 bit trial factors: 15.356 ms.
Best time for 67 bit trial factors: 15.302 ms.

Today's timings (slight improvements on 25.7 even without the multithreading) (overclocked)

Quote:

Compare your results to other computers at http://www.mersenne.org/bench.htm
Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3543.84 MHz, with hyperthreading
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: 21.778 ms.
Best time for 896K FFT length: 26.268 ms.
Best time for 1024K FFT length: 29.854 ms.
Best time for 1280K FFT length: 36.881 ms.
Best time for 1536K FFT length: 44.589 ms.
Best time for 1792K FFT length: 54.271 ms.
Best time for 2048K FFT length: 60.193 ms.
Best time for 2560K FFT length: 78.947 ms.
Best time for 3072K FFT length: 95.410 ms.
Best time for 3584K FFT length: 114.892 ms.
Best time for 4096K FFT length: 128.114 ms.
Best time for 5120K FFT length: 161.546 ms.
Best time for 6144K FFT length: 204.111 ms.
Best time for 7168K FFT length: 245.293 ms.
Best time for 8192K FFT length: 269.152 ms.
Timing FFTs using 2 threads on 1 physical CPUs.
Best time for 768K FFT length: 20.158 ms.
Best time for 896K FFT length: 24.427 ms.
Best time for 1024K FFT length: 26.779 ms.
Best time for 1280K FFT length: 34.835 ms.
Best time for 1536K FFT length: 41.751 ms.
Best time for 1792K FFT length: 50.565 ms.
Best time for 2048K FFT length: 56.097 ms.
Best time for 2560K FFT length: 73.249 ms.
Best time for 3072K FFT length: 89.654 ms.
Best time for 3584K FFT length: 108.813 ms.
Best time for 4096K FFT length: 120.581 ms.
Best time for 5120K FFT length: 156.782 ms.
Best time for 6144K FFT length: 191.514 ms.
Best time for 7168K FFT length: 231.898 ms.
Best time for 8192K FFT length: 255.446 ms.
Best time for 58 bit trial factors: 7.792 ms.
Best time for 59 bit trial factors: 7.820 ms.
Best time for 60 bit trial factors: 7.742 ms.
Best time for 61 bit trial factors: 7.803 ms.
Best time for 62 bit trial factors: 10.850 ms.
Best time for 63 bit trial factors: 10.895 ms.
Best time for 64 bit trial factors: 12.438 ms.
Best time for 65 bit trial factors: 12.623 ms.
Best time for 66 bit trial factors: 12.509 ms.
Best time for 67 bit trial factors: 12.327 ms.

MOBO: ABIT AS8

db597 · 2008-12-01, 10:57

Quote:

Originally Posted by stars10250

Regarding OC, the system just seems to like 3.2 GHz (8x), 400 MHz FSB (1600 MHz rated FSB), and 4-4-4-12 ram timing. I can run it a little faster (3.4 GHz), but it just doesn't "feel" right for the 24/7 performance I want out of it. I've spent several days playing around with the voltages and timings, and I've come to sense when it is going to go unstable or when it will run but it doesn't like it. This is hard to explain without trying it.

Welcome to the club! An overclocked Q6600 is incredible value! Mine has been running 24x7 for many months now with no problems. When you mention 4-4-4-12 RAM timings, is this for 1066MHz or 800MHz? I'm guessing the latter, since that's a very tight setting.

What voltages are you using? I'm using 1.35V for my Q6600 at 3.2GHz and 2.2V for my memory (1066MHz at 5-5-5-15). When left in "auto", my Gigabyte board uses more voltage than neccessary - it's stable but runs hotter than it needs to. The trick is to find the point where it's still stable, but with as low a voltage as possible.

You may want to run a day of torture test to check the stability of the system. Test both small 8K FFTs and large 1024K FFTs - when I use insufficent voltage on mine, it would pass the 8K FFT but fail on the 1024K. Use CoreTemp to monitor the temperature. If it's in the 60s that's fine. Start worrying about cooling if you see figures in the 70s.

fivemack · 2008-12-01, 11:42

I wonder if

http://math-atlas.sourceforge.net/errata.html#cpuid

might be relevant to the CPU-misdetection issue.

Apparently you're now supposed to use

int a = cpuid()
model = ((a&0xf0)>>4) + ((a&0xf0000)>>16)
family = ((a&0xf00)>>8) + ((a&0xff00000)>>20)

to get the model/family information, and Intel has started reusing old model+family pairs for new chips with the extra data in the extended fields.

henryzz · 2008-12-01, 14:27

Quote:

Originally Posted by starrynte

Timings without overclocking and on 24.14 (Times for 512K and 640K removed for easier comparison):
Today's timings (slight improvements on 25.7 even without the multithreading) (overclocked)
MOBO: ABIT AS8

would it be too much to ask to have the non-overclocked benchmark in 25.7 or the same overclocked in 24.14 so we can compare the differences easier

Quote:

Originally Posted by db597

Welcome to the club! An overclocked Q6600 is incredible value! Mine has been running 24x7 for many months now with no problems. When you mention 4-4-4-12 RAM timings, is this for 1066MHz or 800MHz? I'm guessing the latter, since that's a very tight setting.

What voltages are you using? I'm using 1.35V for my Q6600 at 3.2GHz and 2.2V for my memory (1066MHz at 5-5-5-15). When left in "auto", my Gigabyte board uses more voltage than neccessary - it's stable but runs hotter than it needs to. The trick is to find the point where it's still stable, but with as low a voltage as possible.

You may want to run a day of torture test to check the stability of the system. Test both small 8K FFTs and large 1024K FFTs - when I use insufficent voltage on mine, it would pass the 8K FFT but fail on the 1024K. Use CoreTemp to monitor the temperature. If it's in the 60s that's fine. Start worrying about cooling if you see figures in the 70s.

he did say it was PC2-8500 which is 1066mhz
edit: i just noticed that he actually wrote PC-8500 but the motherboard he stated doesnt support DDR memory so it must be a typo
also it would be a bit of a stupid combo IMHO

lavalamp · 2008-12-01, 17:10

Quote:

Originally Posted by henryzz

would it be too much to ask to have the non-overclocked benchmark in 25.7 or the same overclocked in 24.14 so we can compare the differences easier

he did say it was PC2-8500 which is 1066mhz
edit: i just noticed that he actually wrote PC-8500 but the motherboard he stated doesnt support DDR memory so it must be a typo
also it would be a bit of a stupid combo IMHO

There aren't many mainstream motherboards that don't support DDR in one version or another. The board that stars10250 bought supports DDR2.

4 ticks of a 1066 MHz clock though is only 3.75 ns, most RAM latency tends to float around 5 ns, so either he has SPANKINGLY good RAM, or the memory divider has clicked down a notch to run at 800 MHz.

Running at 1066 MHz and 5-5-5-15 should give a slight performance boost over 800 MHz 4-4-4-12.

5 ticks at 1066 MHz = 4.7 ns
4 ticks at 800 MHz = 5 ns

henryzz · 2008-12-01, 20:15

Quote:

Originally Posted by lavalamp

There aren't many mainstream motherboards that don't support DDR in one version or another. The board that stars10250 bought supports DDR2.

4 ticks of a 1066 MHz clock though is only 3.75 ns, most RAM latency tends to float around 5 ns, so either he has SPANKINGLY good RAM, or the memory divider has clicked down a notch to run at 800 MHz.

Running at 1066 MHz and 5-5-5-15 should give a slight performance boost over 800 MHz 4-4-4-12.

5 ticks at 1066 MHz = 4.7 ns
4 ticks at 800 MHz = 5 ns

what i was saying his motherboard supports DDR2 but not DDR and PC-8500 indicates DDR not DDR2 which would be called PC2-8500
i wasnt using DDR as a general term refering to all types of DDR

lavalamp · 2008-12-01, 21:36

1066 MHz would be awfully nippy for DDR, I think DDR2 can be assumed. A little on the low side for DDR3, but would perhaps be used in a low end i7 system.

Those PC numbers aren't the best way to label RAM in my opinion, the marketing departments like to use them because it's puts a bigger number on the packaging, but it's usually rounded off to the nearest hundred and therefore not as accurate as the frequency.

stars10250 · 2008-12-02, 01:19

Sorry about the typo...it's DDR2 as you guessed. I was running my ram at 4-4-4-12 at 400MHz 1:1, but just changed it to 5-5-5-15 at 533MHz, 3:4. I must admit I just figured out how to make this change today! The strapping settings on the motherboard were a bit confusing (I'm a novice recall). Right now it is stress testing but it looks good so far (few hrs in). For fun I started my four LL's and the iteration time dropped from 65 ms to 60 ms across all four cores, which is a nice improvement. Core temps are all 45 degrees C right now during the stress test.

For voltages, I had to crank things up from my earlier ram configuration. The VCore is happy at 1.3125V. Ram voltage is now at 2.1V which is the spec voltage (it didn't like anything less), and the northbridge is at 1.26V. The other voltages are on auto. Before the memory speed change it ran stable at 1.86 V, with the northbridge at 1.24V.

I've tried 3.4 GHz several more times but I can only get it to run stable for about 10 minutes. It didn't feel comfortable so I backed down to 3.2 GHz which it seems to like.

Anyway, I'm happy with my 8% performance improvement for the day :) Anything else I can tweak?

starrynte · 2008-12-02, 01:59

Quote:

Originally Posted by henryzz

would it be too much to ask to have the non-overclocked benchmark in 25.7 or the same overclocked in 24.14 so we can compare the differences easier

(non overclocked 25.7 1 thread has a slight improvement from non overclocked 24.14)
(best benchmark used, these are non-overclocked)

Quote:

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3060.56 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 16 KB
L2 cache size: 1024 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64

Quote:

Best time for 512K FFT length: 17.214 ms.
Best time for 640K FFT length: 22.350 ms.
Best time for 768K FFT length: 27.172 ms.
Best time for 896K FFT length: 33.198 ms.
Best time for 1024K FFT length: 38.122 ms.
Best time for 1280K FFT length: 46.649 ms.
Best time for 1536K FFT length: 56.462 ms.
Best time for 1792K FFT length: 68.155 ms.
Best time for 2048K FFT length: 76.344 ms.
Best time for 2560K FFT length: 100.902 ms.
Best time for 3072K FFT length: 121.601 ms.
Best time for 3584K FFT length: 147.856 ms.
Best time for 4096K FFT length: 163.470 ms.
Best time for 58 bit trial factors: 9.453 ms.
Best time for 59 bit trial factors: 9.536 ms.
Best time for 60 bit trial factors: 9.343 ms.
Best time for 61 bit trial factors: 9.450 ms.
Best time for 62 bit trial factors: 13.165 ms.
Best time for 63 bit trial factors: 13.158 ms.
Best time for 64 bit trial factors: 15.198 ms.
Best time for 65 bit trial factors: 15.172 ms.
Best time for 66 bit trial factors: 15.356 ms.
Best time for 67 bit trial factors: 15.302 ms.

Quote:

Best time for 768K FFT length: 25.056 ms.
Best time for 896K FFT length: 30.631 ms.
Best time for 1024K FFT length: 34.455 ms.
Best time for 1280K FFT length: 42.556 ms.
Best time for 1536K FFT length: 51.596 ms.
Best time for 1792K FFT length: 62.510 ms.
Best time for 2048K FFT length: 69.467 ms.
Best time for 2560K FFT length: 91.631 ms.
Best time for 3072K FFT length: 110.440 ms.
Best time for 3584K FFT length: 133.337 ms.
Best time for 4096K FFT length: 150.112 ms.
Best time for 5120K FFT length: 187.446 ms.
Best time for 6144K FFT length: 237.028 ms.
Best time for 7168K FFT length: 288.179 ms.
Best time for 8192K FFT length: 313.521 ms.
Timing FFTs using 2 threads on 1 physical CPUs.
Best time for 768K FFT length: 23.225 ms.
Best time for 896K FFT length: 28.191 ms.
Best time for 1024K FFT length: 30.888 ms.
Best time for 1280K FFT length: 40.017 ms.
Best time for 1536K FFT length: 48.296 ms.
Best time for 1792K FFT length: 58.522 ms.
Best time for 2048K FFT length: 64.314 ms.
Best time for 2560K FFT length: 84.704 ms.
Best time for 3072K FFT length: 103.787 ms.
Best time for 3584K FFT length: 126.496 ms.
Best time for 4096K FFT length: 140.476 ms.
Best time for 5120K FFT length: 183.097 ms.
Best time for 6144K FFT length: 222.241 ms.
Best time for 7168K FFT length: 271.558 ms.
Best time for 8192K FFT length: 297.570 ms.
Best time for 58 bit trial factors: 8.999 ms.
Best time for 59 bit trial factors: 9.073 ms.
Best time for 60 bit trial factors: 9.012 ms.
Best time for 61 bit trial factors: 9.056 ms.
Best time for 62 bit trial factors: 12.610 ms.
Best time for 63 bit trial factors: 12.590 ms.
Best time for 64 bit trial factors: 14.658 ms.
Best time for 65 bit trial factors: 14.912 ms.
Best time for 66 bit trial factors: 14.885 ms.
Best time for 67 bit trial factors: 14.638 ms.

P.S. I was going to put it all in a nice table, but the forum didn't like it lol

lavalamp · 2008-12-02, 02:23

Quote:

Originally Posted by stars10250

Sorry about the typo...it's DDR2 as you guessed. I was running my ram at 4-4-4-12 at 400MHz 1:1, but just changed it to 5-5-5-15 at 533MHz, 3:4. I must admit I just figured out how to make this change today! The strapping settings on the motherboard were a bit confusing (I'm a novice recall). Right now it is stress testing but it looks good so far (few hrs in). For fun I started my four LL's and the iteration time dropped from 65 ms to 60 ms across all four cores, which is a nice improvement. Core temps are all 45 degrees C right now during the stress test.

For voltages, I had to crank things up from my earlier ram configuration. The VCore is happy at 1.3125V. Ram voltage is now at 2.1V which is the spec voltage (it didn't like anything less), and the northbridge is at 1.26V. The other voltages are on auto. Before the memory speed change it ran stable at 1.86 V, with the northbridge at 1.24V.

I've tried 3.4 GHz several more times but I can only get it to run stable for about 10 minutes. It didn't feel comfortable so I backed down to 3.2 GHz which it seems to like.

Anyway, I'm happy with my 8% performance improvement for the day :) Anything else I can tweak?

It looks like there's quite a bit more head room in your system since your temps are low and so is your VCore. You should be fine all the way up to 1.5 V if this thread is correct. Having said that, it would obviously be a bad idea to jump straight there, one should always creep up to these things.

It would probably be a good idea to set yourself a hard limit that you won't pass, like no more than 65 C under load, or 1.45 max VCore.

Depending on how good your CPU cooler and case airflow are, you may want to be more restrictive or more lenient than that though.

stars10250 · 2008-12-02, 12:33

I'm interested in trying for more but here's my concern. Right now it is running 400FSB at 8x for 3.2 GHz. I can't use fractional multipliers, so I have to use 8x or 9x. Also, I'm not sure it would be worth lowering the FSB to say 378 MHz at 9x (to achieve 3.4 GHz), since I think I want to keep the FSB high for good memory speed (though maybe there is a good memory divider for a setting like this?). So I think the only way I can go faster is to go beyond 400 MHz FSB, like 425 MHz. The motherboard is rated for 1600 MHz, so is it ok if I push it to 1700 MHz? I haven't read as much about that. My core temperatures have always been low, partly because I have a big cooler (zerotherm-FZ120) and partly because it is winter time and the room temperatures are low. In the summer this room will be at least 10 degrees F hotter. I've run the Vcore in the 1.4V range before in my attempts at 3.4 GHz, but like I said I was only able to keep it stable for about 10 minutes. Do you have suggestions for 3.3 GHz, 3.4, or other FSB speeds?

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

2008-12-01, 11:42	#421
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 1100100011000₂ Posts	I wonder if http://math-atlas.sourceforge.net/errata.html#cpuid might be relevant to the CPU-misdetection issue. Apparently you're now supposed to use int a = cpuid() model = ((a&0xf0)>>4) + ((a&0xf0000)>>16) family = ((a&0xf00)>>8) + ((a&0xff00000)>>20) to get the model/family information, and Intel has started reusing old model+family pairs for new chips with the extra data in the extended fields.

2008-12-01, 21:36	#425
lavalamp Oct 2007 Manchester, UK 23·59 Posts	1066 MHz would be awfully nippy for DDR, I think DDR2 can be assumed. A little on the low side for DDR3, but would perhaps be used in a low end i7 system. Those PC numbers aren't the best way to label RAM in my opinion, the marketing departments like to use them because it's puts a bigger number on the packaging, but it's usually rounded off to the nearest hundred and therefore not as accurate as the frequency.

2008-12-02, 01:19	#426
stars10250 Jul 2008 San Francisco, CA C9₁₆ Posts	Sorry about the typo...it's DDR2 as you guessed. I was running my ram at 4-4-4-12 at 400MHz 1:1, but just changed it to 5-5-5-15 at 533MHz, 3:4. I must admit I just figured out how to make this change today! The strapping settings on the motherboard were a bit confusing (I'm a novice recall). Right now it is stress testing but it looks good so far (few hrs in). For fun I started my four LL's and the iteration time dropped from 65 ms to 60 ms across all four cores, which is a nice improvement. Core temps are all 45 degrees C right now during the stress test. For voltages, I had to crank things up from my earlier ram configuration. The VCore is happy at 1.3125V. Ram voltage is now at 2.1V which is the spec voltage (it didn't like anything less), and the northbridge is at 1.26V. The other voltages are on auto. Before the memory speed change it ran stable at 1.86 V, with the northbridge at 1.24V. I've tried 3.4 GHz several more times but I can only get it to run stable for about 10 minutes. It didn't feel comfortable so I backed down to 3.2 GHz which it seems to like. Anyway, I'm happy with my 8% performance improvement for the day :) Anything else I can tweak?

2008-12-02, 12:33	#429
stars10250 Jul 2008 San Francisco, CA 3×67 Posts	I'm interested in trying for more but here's my concern. Right now it is running 400FSB at 8x for 3.2 GHz. I can't use fractional multipliers, so I have to use 8x or 9x. Also, I'm not sure it would be worth lowering the FSB to say 378 MHz at 9x (to achieve 3.4 GHz), since I think I want to keep the FSB high for good memory speed (though maybe there is a good memory divider for a setting like this?). So I think the only way I can go faster is to go beyond 400 MHz FSB, like 425 MHz. The motherboard is rated for 1600 MHz, so is it ok if I push it to 1700 MHz? I haven't read as much about that. My core temperatures have always been low, partly because I have a big cooler (zerotherm-FZ120) and partly because it is winter time and the room temperatures are low. In the summer this room will be at least 10 degrees F hotter. I've run the Vcore in the 1.4V range before in my attempts at 3.4 GHz, but like I said I was only able to keep it stable for about 10 minutes. Do you have suggestions for 3.3 GHz, 3.4, or other FSB speeds?