Perpetual benchmark thread... - Page 19

henryzz · 2008-12-30, 11:06

how do you get it to do that huge benchmark

patrik · 2008-12-30, 11:42

I didn't try it myself with the v25 client, but I guess you do it like this:

Quote:

Originally Posted by undoc.txt

The benchmarking code by default only times a subset of the large FFT sizes.
In prime.txt, set
FullBench=1
to have the benchmarking code time all of the large FFT sizes.

starrynte · 2008-12-30, 16:56

correct

jon_jtech · 2009-01-01, 17:54

Code:

[Thu Jan 01 11:23:19 2009]
Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Pentium(R) Dual  CPU  E2180  @ 2.00GHz
CPU speed: 3000.03 MHz, 2 cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: 1 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 256
Prime95 64-bit version 25.8, RdtscTiming=1
Best time for 4K FFT length: 0.044 ms.
Best time for 5K FFT length: 0.065 ms.
Best time for 6K FFT length: 0.076 ms.
Best time for 7K FFT length: 0.093 ms.
Best time for 8K FFT length: 0.096 ms.
Best time for 10K FFT length: 0.135 ms.
Best time for 12K FFT length: 0.164 ms.
Best time for 14K FFT length: 0.197 ms.
Best time for 16K FFT length: 0.206 ms.
Best time for 20K FFT length: 0.279 ms.
Best time for 24K FFT length: 0.340 ms.
Best time for 28K FFT length: 0.419 ms.
Best time for 32K FFT length: 0.437 ms.
Best time for 40K FFT length: 0.571 ms.
Best time for 48K FFT length: 0.695 ms.
Best time for 56K FFT length: 0.851 ms.
Best time for 64K FFT length: 0.910 ms.
Best time for 80K FFT length: 1.286 ms.
Best time for 96K FFT length: 1.577 ms.
Best time for 112K FFT length: 1.904 ms.
Best time for 128K FFT length: 2.061 ms.
Best time for 160K FFT length: 2.546 ms.
Best time for 192K FFT length: 3.201 ms.
Best time for 224K FFT length: 3.857 ms.
Best time for 256K FFT length: 4.261 ms.
Best time for 320K FFT length: 5.476 ms.
Best time for 384K FFT length: 6.726 ms.
Best time for 448K FFT length: 8.026 ms.
Best time for 512K FFT length: 8.867 ms.
Best time for 640K FFT length: 11.819 ms.
Best time for 768K FFT length: 14.555 ms.
Best time for 896K FFT length: 17.293 ms.
Best time for 1024K FFT length: 20.157 ms.
Best time for 1280K FFT length: 24.219 ms.
Best time for 1536K FFT length: 29.683 ms.
Best time for 1792K FFT length: 35.379 ms.
Best time for 2048K FFT length: 39.361 ms.
Best time for 2560K FFT length: 52.097 ms.
Best time for 3072K FFT length: 64.104 ms.
Best time for 3584K FFT length: 76.723 ms.
Best time for 4096K FFT length: 86.342 ms.
Best time for 5120K FFT length: 111.757 ms.
Best time for 6144K FFT length: 137.566 ms.
Best time for 7168K FFT length: 167.985 ms.
Best time for 8192K FFT length: 185.773 ms.
Best time for 10240K FFT length: 240.579 ms.
Best time for 12288K FFT length: 298.871 ms.
Best time for 14336K FFT length: 363.477 ms.
Best time for 16384K FFT length: 410.165 ms.
Best time for 20480K FFT length: 579.829 ms.
Best time for 24576K FFT length: 718.615 ms.
Best time for 28672K FFT length: 884.232 ms.
Best time for 32768K FFT length: 1004.593 ms.
Timing FFTs using 2 threads.
Best time for 4K FFT length: 0.044 ms.
Best time for 5K FFT length: 0.065 ms.
Best time for 6K FFT length: 0.077 ms.
Best time for 7K FFT length: 0.093 ms.
Best time for 8K FFT length: 0.096 ms.
Best time for 10K FFT length: 0.134 ms.
Best time for 12K FFT length: 0.139 ms.
Best time for 14K FFT length: 0.129 ms.
Best time for 16K FFT length: 0.170 ms.
Best time for 20K FFT length: 0.171 ms.
Best time for 24K FFT length: 0.206 ms.
Best time for 28K FFT length: 0.246 ms.
Best time for 32K FFT length: 0.259 ms.
Best time for 40K FFT length: 1.030 ms.
Best time for 48K FFT length: 1.027 ms.
Best time for 56K FFT length: 1.089 ms.
Best time for 64K FFT length: 1.133 ms.
Best time for 80K FFT length: 1.111 ms.
Best time for 96K FFT length: 0.983 ms.
Best time for 112K FFT length: 1.137 ms.
Best time for 128K FFT length: 1.246 ms.
Best time for 160K FFT length: 1.504 ms.
Best time for 192K FFT length: 1.915 ms.
Best time for 224K FFT length: 2.291 ms.
Best time for 256K FFT length: 2.644 ms.
Best time for 320K FFT length: 3.281 ms.
Best time for 384K FFT length: 3.977 ms.
Best time for 448K FFT length: 4.723 ms.
Best time for 512K FFT length: 5.461 ms.
Best time for 640K FFT length: 7.055 ms.
Best time for 768K FFT length: 8.610 ms.
Best time for 896K FFT length: 10.028 ms.
Best time for 1024K FFT length: 11.862 ms.
Best time for 1280K FFT length: 14.170 ms.
Best time for 1536K FFT length: 16.934 ms.
Best time for 1792K FFT length: 20.513 ms.
Best time for 2048K FFT length: 22.899 ms.
Best time for 2560K FFT length: 29.250 ms.
Best time for 3072K FFT length: 37.551 ms.
Best time for 3584K FFT length: 49.926 ms.
Best time for 4096K FFT length: 52.799 ms.
Best time for 5120K FFT length: 82.134 ms.
Best time for 6144K FFT length: 83.787 ms.
Best time for 7168K FFT length: 107.766 ms.
Best time for 8192K FFT length: 124.614 ms.
Best time for 10240K FFT length: 184.584 ms.
Best time for 12288K FFT length: 184.116 ms.
Best time for 14336K FFT length: 238.359 ms.
Best time for 16384K FFT length: 283.345 ms.
Best time for 20480K FFT length: 459.663 ms.
Best time for 24576K FFT length: 465.853 ms.
Best time for 28672K FFT length: 672.147 ms.
Best time for 32768K FFT length: 809.754 ms.
Best time for 58 bit trial factors: 2.976 ms.
Best time for 59 bit trial factors: 2.999 ms.
Best time for 60 bit trial factors: 2.957 ms.
Best time for 61 bit trial factors: 3.366 ms.
Best time for 62 bit trial factors: 3.477 ms.
Best time for 63 bit trial factors: 4.100 ms.
Best time for 64 bit trial factors: 5.006 ms.
Best time for 65 bit trial factors: 5.282 ms.
Best time for 66 bit trial factors: 5.894 ms.
Best time for 67 bit trial factors: 5.816 ms.

Note: This was run on an MSI P7N SLI motherboard with an Intel stock HSF from a Pentium D 915 and the CPU temp never went about 68C during 8 hours of stress testing. I thought that the results were very good considering the CPU and wanted to share them so I registered here just to post them.

henryzz · 2009-01-03, 18:45

pretty rubbish with two threads
i though all dual cores scaled better than than

James Heinrich · 2009-01-03, 20:30

Quote:

Originally Posted by henryzz

pretty rubbish with two threads
i though all dual cores scaled better than than

If you look at all the numbers, across most of the normally-used FFT sizes you'll see that 2-thread version runs at about 170% of the speed of the 1-thread version. The very-small FFT sizes (4K-10K) have zero performance change in 2-thread version. 40K-80K FFTs actually perform significantly worse in 2-threaded than single-threaded. But this isn't much to worry about... from 96K onwards it's pretty consistent, and 96K is tiny (used for M1648000 to M1966000, by default). See attachment for graph.

henryzz · 2009-01-03, 20:55

Quote:

Originally Posted by James Heinrich

If you look at all the numbers, across most of the normally-used FFT sizes you'll see that 2-thread version runs at about 170% of the speed of the 1-thread version. The very-small FFT sizes (4K-10K) have zero performance change in 2-thread version. 40K-80K FFTs actually perform significantly worse in 2-threaded than single-threaded. But this isn't much to worry about... from 96K onwards it's pretty consistent, and 96K is tiny (used for M1648000 to M1966000, by default). See attachment for graph.

sorry i looked first at the lower end and then the upper end and thought that the upper end should be better than 120-130% of one core
in the middle it gets better

the core 2 architecture does it better though
One core: Best time for 32768K FFT length: 812.106 ms.
Two cores: Best time for 32768K FFT length: 442.065 ms.
Three cores: Best time for 32768K FFT length: 381.942 ms.
Four cores: Best time for 32768K FFT length: 326.699 ms.

Code:

Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Core(TM)2 Quad CPU    Q6600  @ 2.40GHz
CPU speed: 3005.39 MHz, 4 cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: 4 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 256
Prime95 32-bit version 25.8, RdtscTiming=1
Best time for 4K FFT length: 0.043 ms.
Best time for 5K FFT length: 0.064 ms.
Best time for 6K FFT length: 0.076 ms.
Best time for 7K FFT length: 0.092 ms.
Best time for 8K FFT length: 0.095 ms.
Best time for 10K FFT length: 0.133 ms.
Best time for 12K FFT length: 0.163 ms.
Best time for 14K FFT length: 0.197 ms.
Best time for 16K FFT length: 0.207 ms.
Best time for 20K FFT length: 0.279 ms.
Best time for 24K FFT length: 0.340 ms.
Best time for 28K FFT length: 0.421 ms.
Best time for 32K FFT length: 0.441 ms.
Best time for 40K FFT length: 0.575 ms.
Best time for 48K FFT length: 0.697 ms.
Best time for 56K FFT length: 0.853 ms.
Best time for 64K FFT length: 0.910 ms.
Best time for 80K FFT length: 1.282 ms.
Best time for 96K FFT length: 1.564 ms.
Best time for 112K FFT length: 1.885 ms.
Best time for 128K FFT length: 2.001 ms.
Best time for 160K FFT length: 2.460 ms.
Best time for 192K FFT length: 3.022 ms.
Best time for 224K FFT length: 3.614 ms.
Best time for 256K FFT length: 4.001 ms.
Best time for 320K FFT length: 5.123 ms.
Best time for 384K FFT length: 6.303 ms.
Best time for 448K FFT length: 7.595 ms.
Best time for 512K FFT length: 8.456 ms.
Best time for 640K FFT length: 11.293 ms.
Best time for 768K FFT length: 13.987 ms.
Best time for 896K FFT length: 16.747 ms.
Best time for 1024K FFT length: 19.003 ms.
Best time for 1280K FFT length: 23.776 ms.
Best time for 1536K FFT length: 28.999 ms.
Best time for 1792K FFT length: 34.539 ms.
Best time for 2048K FFT length: 38.345 ms.
Best time for 2560K FFT length: 50.701 ms.
Best time for 3072K FFT length: 61.676 ms.
Best time for 3584K FFT length: 73.614 ms.
Best time for 4096K FFT length: 82.409 ms.
Best time for 5120K FFT length: 105.837 ms.
Best time for 6144K FFT length: 128.228 ms.
Best time for 7168K FFT length: 155.753 ms.
Best time for 8192K FFT length: 170.559 ms.
Best time for 10240K FFT length: 226.718 ms.
Best time for 12288K FFT length: 274.480 ms.
Best time for 14336K FFT length: 332.630 ms.
Best time for 16384K FFT length: 365.440 ms.
Best time for 20480K FFT length: 505.670 ms.
Best time for 24576K FFT length: 617.289 ms.
Best time for 28672K FFT length: 739.000 ms.
Best time for 32768K FFT length: 812.106 ms.
Timing FFTs using 2 threads.
Best time for 4K FFT length: 0.043 ms.
Best time for 5K FFT length: 0.064 ms.
Best time for 6K FFT length: 0.076 ms.
Best time for 7K FFT length: 0.092 ms.
Best time for 8K FFT length: 0.095 ms.
Best time for 10K FFT length: 0.118 ms.
Best time for 12K FFT length: 0.114 ms.
Best time for 14K FFT length: 0.127 ms.
Best time for 16K FFT length: 0.138 ms.
Best time for 20K FFT length: 0.170 ms.
Best time for 24K FFT length: 0.200 ms.
Best time for 28K FFT length: 0.242 ms.
Best time for 32K FFT length: 0.254 ms.
Best time for 40K FFT length: 0.734 ms.
Best time for 48K FFT length: 0.760 ms.
Best time for 56K FFT length: 0.838 ms.
Best time for 64K FFT length: 0.875 ms.
Best time for 80K FFT length: 0.926 ms.
Best time for 96K FFT length: 0.905 ms.
Best time for 112K FFT length: 1.060 ms.
Best time for 128K FFT length: 1.134 ms.
Best time for 160K FFT length: 1.348 ms.
Best time for 192K FFT length: 1.595 ms.
Best time for 224K FFT length: 1.895 ms.
Best time for 256K FFT length: 2.104 ms.
Best time for 320K FFT length: 2.681 ms.
Best time for 384K FFT length: 3.299 ms.
Best time for 448K FFT length: 3.968 ms.
Best time for 512K FFT length: 4.442 ms.
Best time for 640K FFT length: 6.102 ms.
Best time for 768K FFT length: 7.536 ms.
Best time for 896K FFT length: 8.909 ms.
Best time for 1024K FFT length: 10.413 ms.
Best time for 1280K FFT length: 12.668 ms.
Best time for 1536K FFT length: 15.483 ms.
Best time for 1792K FFT length: 18.498 ms.
Best time for 2048K FFT length: 20.757 ms.
Best time for 2560K FFT length: 27.268 ms.
Best time for 3072K FFT length: 33.267 ms.
Best time for 3584K FFT length: 40.072 ms.
Best time for 4096K FFT length: 44.816 ms.
Best time for 5120K FFT length: 57.447 ms.
Best time for 6144K FFT length: 69.971 ms.
Best time for 7168K FFT length: 85.222 ms.
Best time for 8192K FFT length: 95.668 ms.
Best time for 10240K FFT length: 121.431 ms.
[Wed Dec 31 14:21:43 2008]
Best time for 12288K FFT length: 149.783 ms.
Best time for 14336K FFT length: 181.500 ms.
Best time for 16384K FFT length: 200.955 ms.
Best time for 20480K FFT length: 269.814 ms.
Best time for 24576K FFT length: 333.306 ms.
Best time for 28672K FFT length: 402.629 ms.
Best time for 32768K FFT length: 442.065 ms.
Timing FFTs using 3 threads.
Best time for 4K FFT length: 0.043 ms.
Best time for 5K FFT length: 0.064 ms.
Best time for 6K FFT length: 0.076 ms.
Best time for 7K FFT length: 0.092 ms.
Best time for 8K FFT length: 0.095 ms.
Best time for 10K FFT length: 0.153 ms.
Best time for 12K FFT length: 0.178 ms.
Best time for 14K FFT length: 0.184 ms.
Best time for 16K FFT length: 0.204 ms.
Best time for 20K FFT length: 0.241 ms.
Best time for 24K FFT length: 0.270 ms.
Best time for 28K FFT length: 0.304 ms.
Best time for 32K FFT length: 0.340 ms.
Best time for 40K FFT length: 0.917 ms.
Best time for 48K FFT length: 0.947 ms.
Best time for 56K FFT length: 1.055 ms.
Best time for 64K FFT length: 1.080 ms.
Best time for 80K FFT length: 1.248 ms.
Best time for 96K FFT length: 1.376 ms.
Best time for 112K FFT length: 1.538 ms.
Best time for 128K FFT length: 1.700 ms.
Best time for 160K FFT length: 1.746 ms.
Best time for 192K FFT length: 1.526 ms.
Best time for 224K FFT length: 1.732 ms.
Best time for 256K FFT length: 1.967 ms.
Best time for 320K FFT length: 2.371 ms.
Best time for 384K FFT length: 2.880 ms.
Best time for 448K FFT length: 3.501 ms.
Best time for 512K FFT length: 3.924 ms.
Best time for 640K FFT length: 7.960 ms.
Best time for 768K FFT length: 8.996 ms.
Best time for 896K FFT length: 10.254 ms.
Best time for 1024K FFT length: 15.080 ms.
Best time for 1280K FFT length: 11.106 ms.
Best time for 1536K FFT length: 13.356 ms.
Best time for 1792K FFT length: 15.397 ms.
Best time for 2048K FFT length: 17.938 ms.
Best time for 2560K FFT length: 23.359 ms.
Best time for 3072K FFT length: 28.905 ms.
Best time for 3584K FFT length: 33.949 ms.
Best time for 4096K FFT length: 38.708 ms.
Best time for 5120K FFT length: 48.092 ms.
Best time for 6144K FFT length: 58.344 ms.
Best time for 7168K FFT length: 72.284 ms.
Best time for 8192K FFT length: 79.083 ms.
Best time for 10240K FFT length: 102.562 ms.
Best time for 12288K FFT length: 122.527 ms.
Best time for 14336K FFT length: 151.461 ms.
Best time for 16384K FFT length: 167.008 ms.
Best time for 20480K FFT length: 229.974 ms.
Best time for 24576K FFT length: 283.547 ms.
Best time for 28672K FFT length: 338.915 ms.
Best time for 32768K FFT length: 381.942 ms.
Timing FFTs using 4 threads.
Best time for 4K FFT length: 0.043 ms.
Best time for 5K FFT length: 0.064 ms.
Best time for 6K FFT length: 0.076 ms.
Best time for 7K FFT length: 0.092 ms.
Best time for 8K FFT length: 0.095 ms.
Best time for 10K FFT length: 0.146 ms.
Best time for 12K FFT length: 0.161 ms.
Best time for 14K FFT length: 0.170 ms.
Best time for 16K FFT length: 0.188 ms.
Best time for 20K FFT length: 0.215 ms.
Best time for 24K FFT length: 0.246 ms.
Best time for 28K FFT length: 0.270 ms.
Best time for 32K FFT length: 0.294 ms.
Best time for 40K FFT length: 0.865 ms.
Best time for 48K FFT length: 0.935 ms.
Best time for 56K FFT length: 0.943 ms.
Best time for 64K FFT length: 0.990 ms.
Best time for 80K FFT length: 1.122 ms.
Best time for 96K FFT length: 1.237 ms.
Best time for 112K FFT length: 1.395 ms.
Best time for 128K FFT length: 1.476 ms.
Best time for 160K FFT length: 1.567 ms.
Best time for 192K FFT length: 1.405 ms.
Best time for 224K FFT length: 1.596 ms.
Best time for 256K FFT length: 1.792 ms.
Best time for 320K FFT length: 2.157 ms.
Best time for 384K FFT length: 2.524 ms.
Best time for 448K FFT length: 2.955 ms.
Best time for 512K FFT length: 3.412 ms.
Best time for 640K FFT length: 7.289 ms.
Best time for 768K FFT length: 8.117 ms.
Best time for 896K FFT length: 9.332 ms.
Best time for 1024K FFT length: 14.240 ms.
Best time for 1280K FFT length: 9.966 ms.
Best time for 1536K FFT length: 11.693 ms.
Best time for 1792K FFT length: 13.406 ms.
[Wed Dec 31 14:26:44 2008]
Best time for 2048K FFT length: 15.890 ms.
Best time for 2560K FFT length: 19.637 ms.
Best time for 3072K FFT length: 25.388 ms.
Best time for 3584K FFT length: 28.973 ms.
Best time for 4096K FFT length: 33.721 ms.
Best time for 5120K FFT length: 41.201 ms.
Best time for 6144K FFT length: 50.045 ms.
Best time for 7168K FFT length: 61.223 ms.
Best time for 8192K FFT length: 68.953 ms.
Best time for 10240K FFT length: 85.743 ms.
Best time for 12288K FFT length: 103.263 ms.
Best time for 14336K FFT length: 124.100 ms.
Best time for 16384K FFT length: 143.934 ms.
Best time for 20480K FFT length: 192.120 ms.
Best time for 24576K FFT length: 235.059 ms.
Best time for 28672K FFT length: 282.375 ms.
Best time for 32768K FFT length: 326.699 ms.
Best time for 58 bit trial factors: 3.597 ms.
Best time for 59 bit trial factors: 3.580 ms.
Best time for 60 bit trial factors: 3.583 ms.
Best time for 61 bit trial factors: 3.535 ms.
Best time for 62 bit trial factors: 3.558 ms.
Best time for 63 bit trial factors: 5.959 ms.
Best time for 64 bit trial factors: 5.949 ms.
Best time for 65 bit trial factors: 5.533 ms.
Best time for 66 bit trial factors: 5.490 ms.
Best time for 67 bit trial factors: 5.526 ms.

thinking about it memory bottleneck might be his problem as with that age cpu i would be surprised if he has fast DDR2 memory
i am used to looking at the normal benchmark which is actually a much better comparison of cpus

James Heinrich · 2009-01-03, 21:25

Quote:

Originally Posted by henryzz

the core 2 architecture does it better though
thinking about it memory bottleneck might be his problem as with that age cpu i would be surprised if he has fast DDR2 memory

The E2180 is Core2 architecture too, same generation as the Q6600, just less cache.

henryzz · 2009-01-03, 21:32

Quote:

Originally Posted by James Heinrich

The E2180 is Core2 architecture too, same generation as the Q6600, just less cache.

again i missed something
Intel(R) Pentium(R) Dual CPU misled me
why doesnt it say core 2
is it a misrecognized cpu

James Heinrich · 2009-01-04, 01:01

Quote:

Originally Posted by henryzz

Intel(R) Pentium(R) Dual CPU misled me
why doesnt it say core 2

Because it's a Pentium Dual-Core.
Intel has several brand names for different target budget/performance ranges: Celeron < Pentium < Core.

henryzz · 2009-01-04, 19:56

Quote:

Originally Posted by James Heinrich

Because it's a Pentium Dual-Core.
Intel has several brand names for different target budget/performance ranges: Celeron < Pentium < Core.

ok that link makes me understand

jon_jtech · 2009-01-07, 02:21

Quote:

Originally Posted by henryzz

again i missed something
Intel(R) Pentium(R) Dual CPU misled me
why doesnt it say core 2
is it a misrecognized cpu

Your's says that your CPU only has 4MB L2 Cache and not 2x 4MB or 8MB. But we know the truth. :)
This CPU is about the same age as your Q6600 and I benchmarked it with DDR2-800 RAM running at 800MHz. I will try pushing the RAM more later but that was not what I was benchmarking this time.

I just wanted to share with everyone the results of this CPU overclocked so everyone could see that a lower-end CPU with OK cooling still can get good results. When compared to other similarly clocked CPUs it did better than I expected but still the lower L2 hurts it. This is CPU has the potential to be a decent economy gamers CPU when overclocked.

T.Rex · 2009-01-07, 18:52

Quote:

Originally Posted by Xyzzy

The boxes run significantly cooler compared to our older 65nm Q6600 C2Q. We think the 9550 is a 65nm CPU with a 95W TDP. We verified that the CPU has the fixed "B3" stepping (?) to eliminate the "TLB" problem.

Hi,
I plan to buy a Phenom Quad 9950 (2.6 GHz - TDP 140W) or a 9850 (2.5 GHz - TDP 95W), with 2 GB G-Skill memory ((2x1024) DDR2 PC8500 (1066Mhz) - 5-5-5-15) and with a GigaByte mother-board.
The reseller says that I need a thermical dissiper (Noctua - NH-C12P) for the 9950. Do you think it is necessary ? Or maybe I should stay with the 9850 which is only 10% slower than the 9950 for Prime95 I think.
There is a new Phenom Quad 9950 2.6 GHz with only 125W. But it is not yet widely available....

And what is this TLB problem ?

Thanks,
Tony

ET_ · 2009-01-07, 19:53

Quote:

Originally Posted by T.Rex

Hi,
I plan to buy a Phenom Quad 9950 (2.6 GHz - TDP 140W) or a 9850 (2.5 GHz - TDP 95W), with 2 GB G-Skill memory ((2x1024) DDR2 PC8500 (1066Mhz) - 5-5-5-15) and with a GigaByte mother-board.
The reseller says that I need a thermical dissiper (Noctua - NH-C12P) for the 9950. Do you think it is necessary ? Or maybe I should stay with the 9850 which is only 10% slower than the 9950 for Prime95 I think.
There is a new Phenom Quad 9950 2.6 GHz with only 125W. But it is not yet widely available....

And what is this TLB problem ?

Thanks,
Tony

IIRC, the TLB problem has been solved by AMD with the stepping B3 release of Phenom. Processors numbered xx5x are not affected by that bug.

I also heard that on next January 9th Phenom II will be released, up to 3 GHz with 45nm technology. Voices tell that Phenom II will be perfectly suited for overclocking, AMD folks pushed it up to 6 GHz with a liquid nitrogen cooler (-195 degrees Celsius).

Luigi, an AMD fan.

metallurge · 2009-01-07, 21:31

Quote:

Originally Posted by T.Rex

Hi,
I plan to buy a Phenom Quad 9950 (2.6 GHz - TDP 140W) or a 9850 (2.5 GHz - TDP 95W), with 2 GB G-Skill memory ((2x1024) DDR2 PC8500 (1066Mhz) - 5-5-5-15) and with a GigaByte mother-board.
The reseller says that I need a thermical dissiper (Noctua - NH-C12P) for the 9950. Do you think it is necessary ? Or maybe I should stay with the 9850 which is only 10% slower than the 9950 for Prime95 I think.
There is a new Phenom Quad 9950 2.6 GHz with only 125W. But it is not yet widely available....

And what is this TLB problem ?

Thanks,
Tony

Very similar to a couple systems I am currently building. The 125W 9950 is very available here, in fact I even just picked one up as an "open box" item at NewEgg for $132 delivered. Regular price for the 9950 125W is about $169. Nothing else makes sense, IMHO, in this price range. Not the 9850, for sure.

Be aware also that Phenom II 940 will be available literally any day now. 3 Ghz, larger cache, improved core, lower power (45 nm vs 60 nm). Approximately $300.

In either case, you will need a good (125-140W capable; heatpipe) CPU cooler (I like the Scythe Ninja+ rev B myself), a well-ventilated case (I like the Antec 300), and probably heatsinks for the motherboard mosfets in the power supply section, if they are not included on your motherboard already. Also, look for solid capacitors in the power supply section. Gigabyte has many choices here, and I have used them. My latest builds are with Biostar 790GX boards.

The TLB problem only applies to 9x00 Phenoms, not 9x50 Phenoms or 9x0 Phenom IIs. TLB Phenoms are to be avoided. Good luck.

James Heinrich · 2009-01-08, 01:27

Quote:

Originally Posted by metallurge

Be aware also that Phenom II 940 will be available literally any day now.

They're actually around, your local computer store very likely has them in stock (where I work has had them since 24-Dec), but AMD has strictly told retailers they can't advertise the Phenom II until release date. But I'd suggest waiting a month and skipping the AM2+ PhenomII's and waiting for the AM3 version -- you then get the choice of using AM2+ or AM3 motherboard, and DDR2 or DDR3 RAM. With the first-release AM2+ versions you're stuck with AM2+ boards and DDR2 RAM (not really a bad thing right now, but things will look different in a year or two and you may regret having the "crippled" version).

cheesehead · 2009-01-08, 08:00

Quote:

Originally Posted by T.Rex

And what is this TLB problem ?

AMD's formal designation is "Erratum 298".

From http://techreport.com/discussions.x/13742

Quote:

Erratum 298 will be described as follows: "The processor operation to change the accessed or dirty bits of a page translation table entry in the L2 from 0b to 1b may not be atomic. A small window of time exists where other cached operations may cause the stale page translation table entry to be installed in the L3 before the modified copy is returned to the L2. In addition, if a probe for this cache line occurs during this window of time, the processor may not set the accessed or dirty bit and may corrupt data for an unrelated cached operation. The system may experience a machine check event reporting an L3 protocol error has occurred. In this case, the MC4 status register (MSR 0000_0410) will be equal to B2000000_000B0C0F or BA000000_000B0C0F. The MC4 address register (MSR 0000_0412) will be equal to 26h."

metallurge · 2009-01-08, 17:59

(Regarding Phenom II's)

Quote:

Originally Posted by James Heinrich

They're actually around, your local computer store very likely has them in stock (where I work has had them since 24-Dec), but AMD has strictly told retailers they can't advertise the Phenom II until release date. But I'd suggest waiting a month and skipping the AM2+ PhenomII's and waiting for the AM3 version -- you then get the choice of using AM2+ or AM3 motherboard, and DDR2 or DDR3 RAM. With the first-release AM2+ versions you're stuck with AM2+ boards and DDR2 RAM (not really a bad thing right now, but things will look different in a year or two and you may regret having the "crippled" version).

First, to let people know, Phenom IIs are now available. I picked a 940 up from newegg this morning for $275.

James, you are right to let people know about the AM3 Phenom II DDR3-capable versions that are coming in a few months (April for the 3 Ghz 945). For me, I intend to populate with not less than 8 Gb of RAM, which would be unaffordably expensive in DDR3 anytime soon. 1066-DDR2 is where the sweet spot currently is price-performance-wise. With the Phenom II's larger cache size, it matters less anyway.

T.Rex · 2009-01-08, 20:31

Quote:

Originally Posted by metallurge

(Regarding Phenom II's)

First, to let people know, Phenom IIs are now available. I picked a 940 up from newegg this morning for $275.

NUTS ! In France, I can see it for 275 Euros !!!!! Instead of the expected 200 Euros ! That's NOT fair !
Tony

metallurge · 2009-01-08, 23:35

Quote:

Originally Posted by T.Rex

NUTS ! In France, I can see it for 275 Euros !!!!! Instead of the expected 200 Euros ! That's NOT fair !
Tony

You know, you might be able to arrange to import one. Shipping would be reasonable. For that much price difference, I might think about it. Know anyone in the States who could help?

James Heinrich · 2009-01-09, 01:54

Quote:

Originally Posted by T.Rex

NUTS ! In France, I can see it for 275 Euros !!!!!

In Canada I see it for C$370 (~225€, US$312), so halfway between US and France.

Aapje · 2009-01-10, 23:06

Quote:

Originally Posted by T.Rex

NUTS ! In France, I can see it for 275 Euros !!!!! Instead of the expected 200 Euros ! That's NOT fair !

The Newegg price is without sales tax (which USians would technically have to pay). So adding 20% sales tax (the french rate) would bring you up to 240 Euros. That is not that far off and might be explained by higher shipping costs.

James Heinrich · 2009-01-11, 01:53

Quote:

Originally Posted by Aapje

The Newegg price is without sales tax

The Canadian price I quoted is also pre-tax, once you add in 13% sales tax (variable by province) would be ~262€, so pretty close.

T.Rex · 2009-01-11, 09:11

OK, thanks for the information about taxes. And, yes, the 275 € include taxes.
So the difference is now only 10-20 %. Shipping, probably.
Thanks again.
Tony

jon_jtech · 2009-01-13, 01:44

Code:

[Mon Jan 12 19:27:07 2009]
Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Core(TM)2 Duo CPU     E8400  @ 3.00GHz
CPU speed: 4050.35 MHz, 2 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 25.8, RdtscTiming=1
Best time for 4K FFT length: 0.032 ms.
Best time for 5K FFT length: 0.047 ms.
Best time for 6K FFT length: 0.056 ms.
Best time for 7K FFT length: 0.068 ms.
Best time for 8K FFT length: 0.071 ms.
Best time for 10K FFT length: 0.099 ms.
Best time for 12K FFT length: 0.120 ms.
Best time for 14K FFT length: 0.145 ms.
Best time for 16K FFT length: 0.152 ms.
Best time for 20K FFT length: 0.206 ms.
Best time for 24K FFT length: 0.252 ms.
Best time for 28K FFT length: 0.310 ms.
Best time for 32K FFT length: 0.323 ms.
Best time for 40K FFT length: 0.422 ms.
Best time for 48K FFT length: 0.514 ms.
Best time for 56K FFT length: 0.629 ms.
Best time for 64K FFT length: 0.671 ms.
Best time for 80K FFT length: 0.948 ms.
Best time for 96K FFT length: 1.153 ms.
Best time for 112K FFT length: 1.389 ms.
Best time for 128K FFT length: 1.481 ms.
Best time for 160K FFT length: 1.806 ms.
Best time for 192K FFT length: 2.236 ms.
Best time for 224K FFT length: 2.667 ms.
Best time for 256K FFT length: 2.958 ms.
Best time for 320K FFT length: 3.749 ms.
Best time for 384K FFT length: 4.625 ms.
Best time for 448K FFT length: 5.512 ms.
Best time for 512K FFT length: 6.113 ms.
Best time for 640K FFT length: 8.017 ms.
Best time for 768K FFT length: 10.150 ms.
Best time for 896K FFT length: 12.343 ms.
Best time for 1024K FFT length: 14.226 ms.
Best time for 1280K FFT length: 17.697 ms.
Best time for 1536K FFT length: 21.774 ms.
Best time for 1792K FFT length: 25.950 ms.
Best time for 2048K FFT length: 28.938 ms.
Best time for 2560K FFT length: 37.955 ms.
Best time for 3072K FFT length: 46.581 ms.
Best time for 3584K FFT length: 55.470 ms.
Best time for 4096K FFT length: 62.073 ms.
Best time for 5120K FFT length: 78.901 ms.
Best time for 6144K FFT length: 95.362 ms.
Best time for 7168K FFT length: 115.543 ms.
Best time for 8192K FFT length: 126.468 ms.
Best time for 10240K FFT length: 168.583 ms.
Best time for 12288K FFT length: 203.554 ms.
Best time for 14336K FFT length: 246.152 ms.
Best time for 16384K FFT length: 270.126 ms.
Best time for 20480K FFT length: 376.375 ms.
Best time for 24576K FFT length: 458.414 ms.
Best time for 28672K FFT length: 546.438 ms.
Best time for 32768K FFT length: 596.896 ms.
Timing FFTs using 2 threads.
Best time for 4K FFT length: 0.032 ms.
Best time for 5K FFT length: 0.047 ms.
Best time for 6K FFT length: 0.056 ms.
Best time for 7K FFT length: 0.068 ms.
Best time for 8K FFT length: 0.070 ms.
Best time for 10K FFT length: 0.068 ms.
Best time for 12K FFT length: 0.080 ms.
Best time for 14K FFT length: 0.091 ms.
Best time for 16K FFT length: 0.097 ms.
Best time for 20K FFT length: 0.122 ms.
Best time for 24K FFT length: 0.146 ms.
Best time for 28K FFT length: 0.176 ms.
Best time for 32K FFT length: 0.185 ms.
Best time for 40K FFT length: 0.443 ms.
Best time for 48K FFT length: 0.452 ms.
Best time for 56K FFT length: 0.524 ms.
Best time for 64K FFT length: 0.542 ms.
Best time for 80K FFT length: 0.548 ms.
Best time for 96K FFT length: 0.655 ms.
Best time for 112K FFT length: 0.774 ms.
Best time for 128K FFT length: 0.825 ms.
Best time for 160K FFT length: 0.981 ms.
Best time for 192K FFT length: 1.175 ms.
Best time for 224K FFT length: 1.394 ms.
Best time for 256K FFT length: 1.549 ms.
Best time for 320K FFT length: 1.964 ms.
Best time for 384K FFT length: 2.404 ms.
Best time for 448K FFT length: 2.863 ms.
Best time for 512K FFT length: 3.185 ms.
Best time for 640K FFT length: 4.308 ms.
Best time for 768K FFT length: 5.407 ms.
Best time for 896K FFT length: 6.566 ms.
Best time for 1024K FFT length: 7.736 ms.
Best time for 1280K FFT length: 9.581 ms.
Best time for 1536K FFT length: 11.922 ms.
Best time for 1792K FFT length: 14.217 ms.
Best time for 2048K FFT length: 16.192 ms.
Best time for 2560K FFT length: 20.969 ms.
Best time for 3072K FFT length: 25.782 ms.
Best time for 3584K FFT length: 30.428 ms.
Best time for 4096K FFT length: 34.600 ms.
Best time for 5120K FFT length: 43.313 ms.
Best time for 6144K FFT length: 52.656 ms.
Best time for 7168K FFT length: 63.182 ms.
Best time for 8192K FFT length: 70.347 ms.
Best time for 10240K FFT length: 89.865 ms.
Best time for 12288K FFT length: 108.971 ms.
Best time for 14336K FFT length: 131.091 ms.
Best time for 16384K FFT length: 145.132 ms.
Best time for 20480K FFT length: 199.812 ms.
Best time for 24576K FFT length: 243.035 ms.
Best time for 28672K FFT length: 288.927 ms.
Best time for 32768K FFT length: 317.888 ms.
Best time for 58 bit trial factors: 2.158 ms.
Best time for 59 bit trial factors: 2.174 ms.
Best time for 60 bit trial factors: 2.200 ms.
Best time for 61 bit trial factors: 2.386 ms.
Best time for 62 bit trial factors: 2.504 ms.
Best time for 63 bit trial factors: 2.950 ms.
Best time for 64 bit trial factors: 3.700 ms.
Best time for 65 bit trial factors: 3.903 ms.
Best time for 66 bit trial factors: 4.345 ms.
Best time for 67 bit trial factors: 4.310 ms.

Note: This test was run using a Thermalright Ultra 120 eXtreme with a S-Flex 12cm 1600rpm FDB fan. System is air cooled only. Motherboard is an ASUS P5K Deluxe.

2009-01-11, 09:11	#474
T.Rex Feb 2004 France 2²×229 Posts	Don't forget taxes ! OK, thanks for the information about taxes. And, yes, the 275 € include taxes. So the difference is now only 10-20 %. Shipping, probably. Thanks again. Tony

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

2008-12-30, 11:06	#451
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 7×29² Posts	how do you get it to do that huge benchmark

2008-12-30, 16:56	#453
starrynte Oct 2008 California EC₁₆ Posts	correct

2009-01-03, 18:45	#455
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 7×29² Posts	pretty rubbish with two threads i though all dual cores scaled better than than