Beta version 24.6 - Athlon users wanted - Page 11

Prime95 · 2004-12-29, 22:57

Quote:

Originally Posted by biwema

I also played around with Prime95 vers 24.6 on Sierpinski factoring:

vers 24.0.1
B1=60000, 660000: 117 minutes

vers 24.6
B1=55000, 632500: 185 minutes (varying quite a bit)

The new version seems to be quite a bit slower. Are there any changes in the implementation?

No. But I thought the 1024K FFT was broken in 24.0.1. Try a smaller N that uses the 768K FFT.

Uncwilly · 2005-01-02, 05:14

Just turned in 12898861. Done with 24.6 Listed in cleared.txt only the soft errors as mentioned above.

Peter Nelson · 2005-01-07, 13:03

I have a Pentium 3 laptop which has been running 23.8.1 successfully.
It has previously been listed in my primenet stats page under a heading Pentium 3 cpus.

I tried upgrading to 24.6 using the ftp binary.

This was successful. Timings we marginally faster on the 512k fft, marginally slower for bigger ones. Perhaps not significant difference, but that's not what I'm posting about.

Looking at results.txt I see that my system has changed from being a

Intel(R) Pentium(R) III processor
CPU speed: 497.56 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE

to a

Intel Celeron 2 (Mobile/XBOX) processor
CPU speed: 497.55 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE

Although I am very happy with a Celeron D I have in another machine (great overclocker), this reported downgrade of my laptop is quite worrying because most Celerons are lame.

Primenet stats now count the machine under the "celeron" heading too.

I ran CPU-Z to make sure I wasn't going crazy and was relieved to see it report:
Intel Pentium III E (Coppermine), brand id 2, 0.18u technology.
Family 6 model 8 stepping 1 revision cA2.

Memtest 86 agrees with me, it's still a Pentium 3 fortunately.

What I have discovered may be a bug in the 24.6 processor detection algorithm.
I understand why you would be changing processor detection to make the Athlons run optimised code, but perhaps in doing so, the existing detection has been broken.

Please could this be fixed so my machine is reclassified back to a Celeron, because I need to look after my credibility you know!

While you're working on that you might take the opportunity to include some recent cpu types eg why not determine and report Northwood or Prescott; and how about Celeron D as opposed to just classifying D as Celeron. I know the code does not fully query the processor just its capabilities and cache size, but I think you could differentiate these simply by establishing whether the processor could run SSE3 instructions. One reason to do this would be for benchmark comparisons - a Northwood and Prescott P4 machine speed result is only comparable if you know which chip you have inside, regardless of MHz/GHz. By the way, the improvement to AMD benchmarks is impressive.

Kind regards, Peter

2005-01-11, 08:44

George says these are slow so any experience on tweaking my PC greatly appreciated! I did originally via the bios have TURBO enabled and it was only after using Prime that I found out this was causing errors.

Phil

Here is the before and after. It is now looking very fast!?

Before:

AMD Athlon(tm) XP 3200+
CPU speed: 2191.24 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 256
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 53.340 ms.
Best time for 448K FFT length: 60.985 ms.
Best time for 512K FFT length: 65.846 ms.
Best time for 640K FFT length: 88.676 ms.
Best time for 768K FFT length: 105.773 ms.
Best time for 896K FFT length: 127.277 ms.
Best time for 1024K FFT length: 141.019 ms.
Best time for 1280K FFT length: 184.985 ms.
Best time for 1536K FFT length: 220.334 ms.
Best time for 1792K FFT length: 268.004 ms.
Best time for 2048K FFT length: 295.786 ms.
[Mon Jan 10 21:57:43 2005]
Compare your results to other computers at http://www.mersenne.org/bench.htm
That web page also contains instructions on how your results can be
included.

After:
AMD Athlon(tm) XP 3200+
CPU speed: 2191.11 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 256
Prime95 version 24.6, RdtscTiming=1
Best time for 512K FFT length: 28.953 ms.
Best time for 640K FFT length: 39.764 ms.
Best time for 768K FFT length: 97.905 ms.
Best time for 896K FFT length: 119.919 ms.
Best time for 1024K FFT length: 133.308 ms.
Best time for 1280K FFT length: 171.875 ms.
Best time for 1536K FFT length: 206.033 ms.
Best time for 1792K FFT length: 123.722 ms.
Best time for 2048K FFT length: 279.577 ms.

Wolf · 2005-01-12, 18:35

Quote:

Originally Posted by geoff

I came across this problem when I tried ECM on 10^311-1:

ECM found a factor in curve #1, stage #1
Sigma=7393313418913594, B1=110000000, B2=4290000000.
ERROR: Factor doesn't divide N!
10^311-1 completed 1 ECM curves, B1=110000000, B2=4290000000

I tried some smaller bounds and got mixed results, including one where the 'ERROR: Factor doesn't divide N!' message repeated over and over until I hit stop.

Quote:

Originally Posted by Prime95

I'll look into the bug. BTW GMP-ECM should clean prime95's clock on this number.

Any progress on a fix for this

Prime95 · 2005-01-12, 22:39

I think that bug was fixed a while back. Let me know if the latest 24.6 still fails. If so, I can email a 24.9 to you.

Wolf · 2005-01-13, 18:53

Quote:

Originally Posted by Prime95

I think that bug was fixed a while back. Let me know if the latest 24.6 still fails. If so, I can email a 24.9 to you.

Using the 14th Dec version of 24.6 I get this on 20^146 + 1 :-

ECM found a factor in curve #1, stage #1
Sigma=8212270044474278, B1=3000000, B2=300000000.
ERROR: Factor doesn't divide N!

On 80^101 - 1 Prime95 crashed with an unhandled exception. 0xC0000005 Access Violation.

phantom · 2005-01-15, 14:42

Is it okay to run regular LL test now or what? =)

Prime95 · 2005-01-15, 14:54

Quote:

Originally Posted by phantom

Is it okay to run regular LL test now or what? =)

Yes. Dozens of successful double-checks have come in.

phantom · 2005-01-15, 15:08

ahh good :), will start using it immediantly. Its way faster :)

renuac · 2005-01-19, 02:11

Prime95 Plz where I can make download version 24.9 of Prime 95?

2005-01-07, 13:03	#113
Peter Nelson Oct 2004 23² Posts	version 23.8.1 to 24.6 test on Pentium 3 / 500 Mhz I have a Pentium 3 laptop which has been running 23.8.1 successfully. It has previously been listed in my primenet stats page under a heading Pentium 3 cpus. I tried upgrading to 24.6 using the ftp binary. This was successful. Timings we marginally faster on the 512k fft, marginally slower for bigger ones. Perhaps not significant difference, but that's not what I'm posting about. Looking at results.txt I see that my system has changed from being a Intel(R) Pentium(R) III processor CPU speed: 497.56 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE to a Intel Celeron 2 (Mobile/XBOX) processor CPU speed: 497.55 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE Although I am very happy with a Celeron D I have in another machine (great overclocker), this reported downgrade of my laptop is quite worrying because most Celerons are lame. Primenet stats now count the machine under the "celeron" heading too. I ran CPU-Z to make sure I wasn't going crazy and was relieved to see it report: Intel Pentium III E (Coppermine), brand id 2, 0.18u technology. Family 6 model 8 stepping 1 revision cA2. Memtest 86 agrees with me, it's still a Pentium 3 fortunately. What I have discovered may be a bug in the 24.6 processor detection algorithm. I understand why you would be changing processor detection to make the Athlons run optimised code, but perhaps in doing so, the existing detection has been broken. Please could this be fixed so my machine is reclassified back to a Celeron, because I need to look after my credibility you know! While you're working on that you might take the opportunity to include some recent cpu types eg why not determine and report Northwood or Prescott; and how about Celeron D as opposed to just classifying D as Celeron. I know the code does not fully query the processor just its capabilities and cache size, but I think you could differentiate these simply by establishing whether the processor could run SSE3 instructions. One reason to do this would be for benchmark comparisons - a Northwood and Prescott P4 machine speed result is only comparable if you know which chip you have inside, regardless of MHz/GHz. By the way, the improvement to AMD benchmarks is impressive. Kind regards, Peter

2005-01-11, 08:44	#114
philipwinfield 13·17·29 Posts	XP 3200 results George says these are slow so any experience on tweaking my PC greatly appreciated! I did originally via the bios have TURBO enabled and it was only after using Prime that I found out this was causing errors. Phil Here is the before and after. It is now looking very fast!? Before: AMD Athlon(tm) XP 3200+ CPU speed: 2191.24 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE L1 cache size: 64 KB L2 cache size: 512 KB L1 cache line size: 64 bytes L2 cache line size: 64 bytes L1 TLBS: 32 L2 TLBS: 256 Prime95 version 23.8, RdtscTiming=1 Best time for 384K FFT length: 53.340 ms. Best time for 448K FFT length: 60.985 ms. Best time for 512K FFT length: 65.846 ms. Best time for 640K FFT length: 88.676 ms. Best time for 768K FFT length: 105.773 ms. Best time for 896K FFT length: 127.277 ms. Best time for 1024K FFT length: 141.019 ms. Best time for 1280K FFT length: 184.985 ms. Best time for 1536K FFT length: 220.334 ms. Best time for 1792K FFT length: 268.004 ms. Best time for 2048K FFT length: 295.786 ms. [Mon Jan 10 21:57:43 2005] Compare your results to other computers at http://www.mersenne.org/bench.htm That web page also contains instructions on how your results can be included. After: AMD Athlon(tm) XP 3200+ CPU speed: 2191.11 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE L1 cache size: 64 KB L2 cache size: 512 KB L1 cache line size: 64 bytes L2 cache line size: 64 bytes L1 TLBS: 32 L2 TLBS: 256 Prime95 version 24.6, RdtscTiming=1 Best time for 512K FFT length: 28.953 ms. Best time for 640K FFT length: 39.764 ms. Best time for 768K FFT length: 97.905 ms. Best time for 896K FFT length: 119.919 ms. Best time for 1024K FFT length: 133.308 ms. Best time for 1280K FFT length: 171.875 ms. Best time for 1536K FFT length: 206.033 ms. Best time for 1792K FFT length: 123.722 ms. Best time for 2048K FFT length: 279.577 ms. Last fiddled with by philipwinfield on 2005-01-11 at 08:45

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LLR beta Version 3.8.13 (deprecated)	Jean Penné	Software	111	2015-01-26 21:41
Prime95 beta version 28.3	Prime95	Software	68	2014-02-23 05:42
Beta version 24.12 available	Prime95	Software	33	2005-06-14 13:19
Early Beta of version 24.11	Prime95	Software	113	2005-05-24 17:05
Beta version of PRP	Prime95	PSearch	15	2004-09-17 19:21

2005-01-02, 05:14	#112
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 2672₁₆ Posts	Just turned in 12898861. Done with 24.6 Listed in cleared.txt only the soft errors as mentioned above.

2005-01-12, 22:39	#116
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 7543₁₀ Posts	I think that bug was fixed a while back. Let me know if the latest 24.6 still fails. If so, I can email a 24.9 to you.

2005-01-15, 14:42	#118
phantom Jan 2005 3 Posts	Is it okay to run regular LL test now or what? =)

2005-01-15, 15:08	#120
phantom Jan 2005 3 Posts	ahh good :), will start using it immediantly. Its way faster :)

2005-01-19, 02:11	#121
renuac Jan 2005 3 Posts	Prime95 Plz where I can make download version 24.9 of Prime 95?