Benchmarks wierdness

delta_t · 2003-12-11, 20:44

This is very strange. Below are two benchmarks from the same machine. This is a dual Xenon 2.4 on an Asus PC-DL Deluxe board with 1G of dual channel RAM, running XP Pro. These are HT capable processor.

The first benchmark is only one client running with XP running as usual. The second benchmark is still one client running, but with XP in Safe Mode. I could repeat this. Is there something that XP is doing (or not doing) in Safe Mode?

Typically P95 runs faster in Safe Mode, but not in this case. Wondering if anyone else could try this on an HT capable P4. My regular (non-HT) P4 has slightly better results in Safe Mode as expected.

** Regular **

Intel(R) Xeon(TM) CPU 2.40GHz
CPU speed: 2525.71 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 23.7, RdtscTiming=1
Best time for 384K FFT length: 14.163 ms.
Best time for 448K FFT length: 16.795 ms.
Best time for 512K FFT length: 19.072 ms.
Best time for 640K FFT length: 22.952 ms.
Best time for 768K FFT length: 27.943 ms.
Best time for 896K FFT length: 33.117 ms.
Best time for 1024K FFT length: 37.283 ms.
Best time for 1280K FFT length: 48.679 ms.
Best time for 1536K FFT length: 60.194 ms.
Best time for 1792K FFT length: 71.050 ms.
Best time for 2048K FFT length: 80.715 ms.

** Safe Mode **

Intel(R) Xeon(TM) CPU 2.40GHz
CPU speed: 2525.71 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 23.7, RdtscTiming=1
Best time for 384K FFT length: 19.738 ms.
Best time for 448K FFT length: 23.529 ms.
Best time for 512K FFT length: 26.581 ms.
Best time for 640K FFT length: 31.619 ms.
Best time for 768K FFT length: 38.587 ms.
Best time for 896K FFT length: 46.073 ms.
Best time for 1024K FFT length: 51.387 ms.
Best time for 1280K FFT length: 67.707 ms.
Best time for 1536K FFT length: 82.960 ms.
Best time for 1792K FFT length: 99.243 ms.
Best time for 2048K FFT length: 111.275 ms.

xtreme2k · 2003-12-12, 06:46

Most likely the HT kernel is not loaded in Safe mode. Performance is like that if you run a HT capable in Win2k also(as W2k also doesnt support HT).

Dresdenboy · 2003-12-12, 12:16

Maybe Windows has a (slightly) different memory management in safe mode and provides memory in a disadvantegous form (for example memory pages not being contiguous in physical memory).

Jeff Gilchrist · 2004-06-24, 13:35

Does anyone know if this would be the same with a true SMP machine? In SafeMode would it disable the support for the second CPU?

2003-12-11, 20:44	#1
delta_t Nov 2002 Anchorage, AK 545₈ Posts	Benchmarks wierdness This is very strange. Below are two benchmarks from the same machine. This is a dual Xenon 2.4 on an Asus PC-DL Deluxe board with 1G of dual channel RAM, running XP Pro. These are HT capable processor. The first benchmark is only one client running with XP running as usual. The second benchmark is still one client running, but with XP in Safe Mode. I could repeat this. Is there something that XP is doing (or not doing) in Safe Mode? Typically P95 runs faster in Safe Mode, but not in this case. Wondering if anyone else could try this on an HT capable P4. My regular (non-HT) P4 has slightly better results in Safe Mode as expected. Regular Intel(R) Xeon(TM) CPU 2.40GHz CPU speed: 2525.71 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 23.7, RdtscTiming=1 Best time for 384K FFT length: 14.163 ms. Best time for 448K FFT length: 16.795 ms. Best time for 512K FFT length: 19.072 ms. Best time for 640K FFT length: 22.952 ms. Best time for 768K FFT length: 27.943 ms. Best time for 896K FFT length: 33.117 ms. Best time for 1024K FFT length: 37.283 ms. Best time for 1280K FFT length: 48.679 ms. Best time for 1536K FFT length: 60.194 ms. Best time for 1792K FFT length: 71.050 ms. Best time for 2048K FFT length: 80.715 ms. Safe Mode Intel(R) Xeon(TM) CPU 2.40GHz CPU speed: 2525.71 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 23.7, RdtscTiming=1 Best time for 384K FFT length: 19.738 ms. Best time for 448K FFT length: 23.529 ms. Best time for 512K FFT length: 26.581 ms. Best time for 640K FFT length: 31.619 ms. Best time for 768K FFT length: 38.587 ms. Best time for 896K FFT length: 46.073 ms. Best time for 1024K FFT length: 51.387 ms. Best time for 1280K FFT length: 67.707 ms. Best time for 1536K FFT length: 82.960 ms. Best time for 1792K FFT length: 99.243 ms. Best time for 2048K FFT length: 111.275 ms.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Benchmarks	Pjetrode	Information & Answers	3	2018-01-07 23:23
Benchmarks	ET_	Operazione Doppi Mersennes	18	2013-04-24 06:38
Where are the Benchmarks	Sandman192	Homework Help	17	2012-04-05 19:03
ggnfs sieving yield wierdness	jrk	Factoring	10	2009-05-07 17:41
Benchmarks for 24.12	Prime95	Software	60	2005-06-11 07:35

2003-12-12, 06:46	#2
xtreme2k Aug 2002 2·3·29 Posts	Most likely the HT kernel is not loaded in Safe mode. Performance is like that if you run a HT capable in Win2k also(as W2k also doesnt support HT).

2003-12-12, 12:16	#3
Dresdenboy Apr 2003 Berlin, Germany 551₈ Posts	Maybe Windows has a (slightly) different memory management in safe mode and provides memory in a disadvantegous form (for example memory pages not being contiguous in physical memory).

2004-06-24, 13:35	#4
Jeff Gilchrist Jun 2003 Ottawa, Canada 3·17·23 Posts	Does anyone know if this would be the same with a true SMP machine? In SafeMode would it disable the support for the second CPU?