Beta version 24.6 - Athlon users wanted - Page 5

koekie · 2004-12-11, 10:18

Looks like P95 itself also decided that it isn't THAT much faster. It just unreserved most of the work by itself

Matthias C. Noc · 2004-12-11, 12:40

It seems with P4s the smaller FFTs run a bit faster, but the higher FFTs are slower.

Intel(R) Pentium(R) 4 CPU 2.66GHz
CPU speed: 2799.78 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 8 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
<DELETED>
Best time for 512K FFT length: 17.927 ms.
Best time for 640K FFT length: 21.514 ms.
Best time for 768K FFT length: 26.153 ms.
Best time for 896K FFT length: 30.723 ms.
Best time for 1024K FFT length: 34.493 ms.
Best time for 1280K FFT length: 45.197 ms.
Best time for 1536K FFT length: 55.248 ms.
Best time for 1792K FFT length: 65.832 ms.
Best time for 2048K FFT length: 74.999 ms.

With 24.6

Intel(R) Pentium(R) 4 CPU 2.66GHz
CPU speed: 2799.84 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 8 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 24.6, RdtscTiming=1
Best time for 512K FFT length: 17.742 ms.
Best time for 640K FFT length: 21.416 ms.
Best time for 768K FFT length: 26.266 ms.
Best time for 896K FFT length: 30.597 ms.
Best time for 1024K FFT length: 34.499 ms.
Best time for 1280K FFT length: 45.238 ms.
Best time for 1536K FFT length: 57.118 ms.
Best time for 1792K FFT length: 66.192 ms.
Best time for 2048K FFT length: 75.344 ms.

Prime95 · 2004-12-11, 14:58

Quote:

Originally Posted by gdf

Unfortunately, this 24.6 version crashes on my K6.

I've uploaded a new 24.6 to fix this.

gdf · 2004-12-11, 17:06

where,
the p95v246.zip in the ftp directory has a newer date, but is not different from the one I downloaded a day ago.

Danath · 2004-12-11, 17:48

Don't forget to run several benchmarks, one after other, and post the average result (or which is met more often). It happens that the difference between one benchmark run and another (just after the first) is considerable.

And the difference within 1 millisecond per iteration is not a difference at all, for the iteration time may vary from one benchmark run to another by even more than one ms/iter. So it looks like the new version brought no speed change for P4s (looking at the benchmarks posted by louis_net and Matthias C. Noc).

Prime95 · 2004-12-11, 18:30

Quote:

Originally Posted by gdf

where,
the p95v246.zip in the ftp directory has a newer date, but is not different from the one I downloaded a day ago.

My bad, I rebuilt the debug version and then zipped the release version. Please download again. Sorry.

alpertron · 2004-12-11, 20:20

After several successful runs, using the following worktodo.ini:

ECM2=1,2,2976221,-3,2000,200000,22

produced page fault errors using both the old and the new versions named 24.6.

This is the output in the new version 24.6 (sorry, my Windows 98 Second Edition is in Spanish, but I think it is easy to understand).

Code:

PRIME95 provocó un error de página no válida en el 
módulo PRIME95.EXE de 01af:00406e66.
Registros:
EAX=00016b4f CS=01af EIP=00406e66 EFLGS=00210212
EBX=02014aac SS=01b7 ESP=014bf81c EBP=0206f7f4
ECX=00016b4f DS=01b7 ESI=02014ab4 FS=0fb7
EDX=00000000 ES=01b7 EDI=00000000 GS=0000
Bytes en CS:EIP:
,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x 
Volcado de pila:
,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x

The microprocessor is a Duron 2000+ (1300 MHz).

gdf · 2004-12-11, 20:31

Quote:

Originally Posted by Prime95

My bad, I rebuilt the debug version and then zipped the release version. Please download again. Sorry.

This version stil crashes, details :

PRIME95_24_6_2 executed an invalid instruction in
module PRIME95_24_6_2.EXE at 015f:004c9d3d.
Registers:
EAX=00000000 CS=015f EIP=004c9d3d EFLGS=00210247
EBX=00000000 SS=0167 ESP=012afb98 EBP=00000040
ECX=01848800 DS=0167 ESI=012c2040 FS=2ba7
EDX=00044000 ES=0167 EDI=0085f85c GS=0000
Bytes at CS:EIP:
0f 18 54 0d 00 8d 49 20 04 80 0f 83 2e ff ff ff
Stack dump:
010a12f0 0085f85c 00000018 012afbd4 0041612b 4050c001 00000000 4050c000 00000000 0043ad71 012afbd8 0041c34a 016c1000 0007ffff 00000000 012afea0

Not that it is important, I just was curious what it would do on my K6-2 which I will use to upgrade my pentium linux server box.

gdf

alpertron · 2004-12-11, 20:44

This is the disassembly of the subroutine that didn't finish execution:

Code:

00406E40   push        ebx
00406E41   push        esi
00406E42   mov         ebx,dword ptr [esp+0Ch]
00406E46   push        edi
00406E47   mov         esi,dword ptr [esp+14h]
00406E4B   mov         eax,dword ptr [ebx]
00406E4D   mov         edi,dword ptr [esi+4]
00406E50   mov         dword ptr [esi],eax
00406E52   mov         eax,dword ptr [ebx]
00406E54   mov         esi,dword ptr [ebx+4]
00406E57   cdq
00406E58   xor         eax,edx
00406E5A   sub         eax,edx
00406E5C   lea         ecx,[eax*4]
00406E63   shr         ecx,2
00406E66   rep movs    dword ptr [edi],dword ptr [esi]
00406E68   pop         edi
00406E69   pop         esi
00406E6A   pop         ebx
00406E6B   ret

In the offending instruction, located at 00406E66, EDI = 0, that means a null pointer.

Tihy · 2004-12-13, 15:17

The new version uses a lot of memory for the P-1 test. P95 v23.8 doesn't pass the 512M physical limit of the memory but the new version uses about 1024M (physical+swap). The OS is Win2000 Pro, Barton 2600,nForce2 Ultra,512M RAM.

arjanscholl · 2004-12-13, 17:31

New version:

Code:

[Mon Dec 13 18:21:24 2004]
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.

AMD Athlon(tm) 
CPU speed: 2205.09 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 64 KB
L2 cache size: 256 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: 28.607 ms.
Best time for 640K FFT length: 39.223 ms.
Best time for 768K FFT length: 47.854 ms.
Best time for 896K FFT length: 58.246 ms.
Best time for 1024K FFT length: 64.782 ms.
Best time for 1280K FFT length: 83.595 ms.
Best time for 1536K FFT length: 100.587 ms.
Best time for 1792K FFT length: 122.159 ms.
Best time for 2048K FFT length: 135.351 ms.

Old version:

Code:

[Mon Dec 13 18:22:29 2004]
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.

AMD Athlon(tm) 
CPU speed: 2205.24 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 64 KB
L2 cache size: 256 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: 26.150 ms.
Best time for 448K FFT length: 29.895 ms.
Best time for 512K FFT length: 32.292 ms.
Best time for 640K FFT length: 43.288 ms.
Best time for 768K FFT length: 51.372 ms.
Best time for 896K FFT length: 61.977 ms.
Best time for 1024K FFT length: 68.556 ms.
Best time for 1280K FFT length: 90.077 ms.
Best time for 1536K FFT length: 107.867 ms.
Best time for 1792K FFT length: 132.331 ms.
Best time for 2048K FFT length: 152.358 ms.

On a Athlon XP 2700+ @ 2200MHz

2004-12-11, 12:40	#46
Matthias C. Noc Dec 2003 30₈ Posts	It seems with P4s the smaller FFTs run a bit faster, but the higher FFTs are slower. Intel(R) Pentium(R) 4 CPU 2.66GHz CPU speed: 2799.78 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2 L1 cache size: 8 KB L2 cache size: 512 KB L1 cache line size: 64 bytes L2 cache line size: 128 bytes TLBS: 64 Prime95 version 23.8, RdtscTiming=1 <DELETED> Best time for 512K FFT length: 17.927 ms. Best time for 640K FFT length: 21.514 ms. Best time for 768K FFT length: 26.153 ms. Best time for 896K FFT length: 30.723 ms. Best time for 1024K FFT length: 34.493 ms. Best time for 1280K FFT length: 45.197 ms. Best time for 1536K FFT length: 55.248 ms. Best time for 1792K FFT length: 65.832 ms. Best time for 2048K FFT length: 74.999 ms. With 24.6 Intel(R) Pentium(R) 4 CPU 2.66GHz CPU speed: 2799.84 MHz CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2 L1 cache size: 8 KB L2 cache size: 512 KB L1 cache line size: 64 bytes L2 cache line size: 128 bytes TLBS: 64 Prime95 version 24.6, RdtscTiming=1 Best time for 512K FFT length: 17.742 ms. Best time for 640K FFT length: 21.416 ms. Best time for 768K FFT length: 26.266 ms. Best time for 896K FFT length: 30.597 ms. Best time for 1024K FFT length: 34.499 ms. Best time for 1280K FFT length: 45.238 ms. Best time for 1536K FFT length: 57.118 ms. Best time for 1792K FFT length: 66.192 ms. Best time for 2048K FFT length: 75.344 ms. Last fiddled with by Matthias C. Noc on 2004-12-11 at 12:42

2004-12-11, 20:20	#51
alpertron Aug 2002 Buenos Aires, Argentina 5F3₁₆ Posts	After several successful runs, using the following worktodo.ini: ECM2=1,2,2976221,-3,2000,200000,22 produced page fault errors using both the old and the new versions named 24.6. This is the output in the new version 24.6 (sorry, my Windows 98 Second Edition is in Spanish, but I think it is easy to understand). Code: PRIME95 provocó un error de página no válida en el módulo PRIME95.EXE de 01af:00406e66. Registros: EAX=00016b4f CS=01af EIP=00406e66 EFLGS=00210212 EBX=02014aac SS=01b7 ESP=014bf81c EBP=0206f7f4 ECX=00016b4f DS=01b7 ESI=02014ab4 FS=0fb7 EDX=00000000 ES=01b7 EDI=00000000 GS=0000 Bytes en CS:EIP: ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x ,02x Volcado de pila: ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x ,08x The microprocessor is a Duron 2000+ (1300 MHz). Last fiddled with by alpertron on 2004-12-11 at 20:22

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

2004-12-11, 10:18	#45
koekie Dec 2002 Amsterdam, Netherlands 2²×19 Posts	Looks like P95 itself also decided that it isn't THAT much faster. It just unreserved most of the work by itself

2004-12-11, 17:06	#48
gdf Aug 2003 12₈ Posts	where, the p95v246.zip in the ftp directory has a newer date, but is not different from the one I downloaded a day ago.

2004-12-11, 17:48	#49
Danath May 2003 Republic of Moldova 2³·5 Posts	Don't forget to run several benchmarks, one after other, and post the average result (or which is met more often). It happens that the difference between one benchmark run and another (just after the first) is considerable. And the difference within 1 millisecond per iteration is not a difference at all, for the iteration time may vary from one benchmark run to another by even more than one ms/iter. So it looks like the new version brought no speed change for P4s (looking at the benchmarks posted by louis_net and Matthias C. Noc).

2004-12-13, 15:17	#54
Tihy Oct 2004 Romania 9₁₆ Posts	The new version uses a lot of memory for the P-1 test. P95 v23.8 doesn't pass the 512M physical limit of the memory but the new version uses about 1024M (physical+swap). The OS is Win2000 Pro, Barton 2600,nForce2 Ultra,512M RAM.