Duvie
08-14-2003, 10:46 PM
Here are some synthetic benchmarks first.....
Test Systems:
#1
Gateway 1.5ghz Centrino
1024mb pc2100 DDR
40gb 4200rpm HDD
#2
P4 2.4b@3.15ghz
512mb pc3000 ddr @438mhz (pc3500)
40gb 7200rpm HDD (NTFS)
#3
P4 2.4b@1.8ghz (400fsb)
512mb pc3000 ddr @ 266mhz (pc2100)
40gb 7200rpm HDD (NTFS)
#4
P4 2.4b@2.01ghz (448fsb)
512mhz pc3000 ddr @ 280mhz (pc2200)
40gb 7200rpm HDD (NTFS)
Benchmarks used:
Prime95 version 22.12 (I tried vertsion 23 and all scores increasesd equally but had same separation)
Sissoft Sandra (non max version, tried it no difference)
PCMark2002 (all test)
____________________________________________
Prime95 benchmark 1792k
p4 3.15ghz (#2) -------- 76.22ms ---- (+223.5%)
p4 2.01ghz (#4) -------- 106.86ms --- (+130.7%)
p4 1.8ghz (#3) --------- 131.49ms --- (+87.5%)
p4m 1.5ghz (#1) ------- 246.49ms --- **basis**
_____________________________________________
Sissoft CPU (ALU (mips))
p4 3.15ghz (#2) ------- 8770 --- (+78.0%)
p4 2.01ghz (#4) ------- 5112 --- (+3.8%)
p4m 1.5ghz (#1) ------ 4927 ---- **basis**
p4 1.8ghz (#3) -------- 4596 ---- (-7.2%)
Sissoft CPU (FPU (mflops))
p4 3.15ghz (#2) ------- 4143 --- (+39.6%)
p4m 1.5ghz (#1) ------ 2966 ---- **basis**
p4 2.01ghz (#4) ------- 2643 --- (-12.2%)
p4 1.8ghz (#3) -------- 2351 ---- (-26.2%)
Sissoft MM (integer)
p4 3.15ghz (#2) ------- 12484 --- (+46.7%)
p4m 1.5ghz (#1) ------ 8514 ---- **basis**
p4 2.01ghz (#4) ------- 7973 --- (-6.8%)
p4 1.8ghz (#3) -------- 7111 ---- (-19.7%)
Sissoft MM (float)
p4 3.15ghz (#2) ------- 15951 --- (+61.2%)
p4 2.01ghz (#4) ------- 10184 --- (+2.9%)
p4m 1.5ghz (#1) ------ 9896 ---- **basis**
p4 1.8ghz (#3) -------- 9086 ----- (-8.9%)
______________________________________________
PCMark2002 (CPU)
p4 3.15ghz (#2) ------- 7721 --- (+57.8%)
p4 2.01ghz (#4) ------- 4902 --- (+0.2%)
p4m 1.5ghz (#1) ------ 4894 ---- **basis**
p4 1.8ghz (#3) -------- 4407 ----- (-11.0%)
____________________________________________
Conclusions so far is that the centrino at 1500mhz falls around the speed of a 2.0ghz 512kb chip or slightly better. My concern is though it scores well in floating points in the sandra test it really struggles in the benchmark in prime95. I tried newest version that says it supports the p4m but does not mention centrino exactly. difference was that all p4's across the board gained slightly from optimized code of the newer version.
The question for me is that several of the programs like prime95 and sandra saw the centrino 1500 as 600mhz in speed. Is this how the chip is preceived in an idle state and hence its limited power use or throttling like effect to conserve power?
I am going to run sandra for 10pass cpu and look at the scores to see if they increase with heavier continues use. I briefly ran the cpu/bios immediately after a test and it told me the cpu was now 800mhz, so there appears to be some sort of changing going on.
Test Systems:
#1
Gateway 1.5ghz Centrino
1024mb pc2100 DDR
40gb 4200rpm HDD
#2
P4 2.4b@3.15ghz
512mb pc3000 ddr @438mhz (pc3500)
40gb 7200rpm HDD (NTFS)
#3
P4 2.4b@1.8ghz (400fsb)
512mb pc3000 ddr @ 266mhz (pc2100)
40gb 7200rpm HDD (NTFS)
#4
P4 2.4b@2.01ghz (448fsb)
512mhz pc3000 ddr @ 280mhz (pc2200)
40gb 7200rpm HDD (NTFS)
Benchmarks used:
Prime95 version 22.12 (I tried vertsion 23 and all scores increasesd equally but had same separation)
Sissoft Sandra (non max version, tried it no difference)
PCMark2002 (all test)
____________________________________________
Prime95 benchmark 1792k
p4 3.15ghz (#2) -------- 76.22ms ---- (+223.5%)
p4 2.01ghz (#4) -------- 106.86ms --- (+130.7%)
p4 1.8ghz (#3) --------- 131.49ms --- (+87.5%)
p4m 1.5ghz (#1) ------- 246.49ms --- **basis**
_____________________________________________
Sissoft CPU (ALU (mips))
p4 3.15ghz (#2) ------- 8770 --- (+78.0%)
p4 2.01ghz (#4) ------- 5112 --- (+3.8%)
p4m 1.5ghz (#1) ------ 4927 ---- **basis**
p4 1.8ghz (#3) -------- 4596 ---- (-7.2%)
Sissoft CPU (FPU (mflops))
p4 3.15ghz (#2) ------- 4143 --- (+39.6%)
p4m 1.5ghz (#1) ------ 2966 ---- **basis**
p4 2.01ghz (#4) ------- 2643 --- (-12.2%)
p4 1.8ghz (#3) -------- 2351 ---- (-26.2%)
Sissoft MM (integer)
p4 3.15ghz (#2) ------- 12484 --- (+46.7%)
p4m 1.5ghz (#1) ------ 8514 ---- **basis**
p4 2.01ghz (#4) ------- 7973 --- (-6.8%)
p4 1.8ghz (#3) -------- 7111 ---- (-19.7%)
Sissoft MM (float)
p4 3.15ghz (#2) ------- 15951 --- (+61.2%)
p4 2.01ghz (#4) ------- 10184 --- (+2.9%)
p4m 1.5ghz (#1) ------ 9896 ---- **basis**
p4 1.8ghz (#3) -------- 9086 ----- (-8.9%)
______________________________________________
PCMark2002 (CPU)
p4 3.15ghz (#2) ------- 7721 --- (+57.8%)
p4 2.01ghz (#4) ------- 4902 --- (+0.2%)
p4m 1.5ghz (#1) ------ 4894 ---- **basis**
p4 1.8ghz (#3) -------- 4407 ----- (-11.0%)
____________________________________________
Conclusions so far is that the centrino at 1500mhz falls around the speed of a 2.0ghz 512kb chip or slightly better. My concern is though it scores well in floating points in the sandra test it really struggles in the benchmark in prime95. I tried newest version that says it supports the p4m but does not mention centrino exactly. difference was that all p4's across the board gained slightly from optimized code of the newer version.
The question for me is that several of the programs like prime95 and sandra saw the centrino 1500 as 600mhz in speed. Is this how the chip is preceived in an idle state and hence its limited power use or throttling like effect to conserve power?
I am going to run sandra for 10pass cpu and look at the scores to see if they increase with heavier continues use. I briefly ran the cpu/bios immediately after a test and it told me the cpu was now 800mhz, so there appears to be some sort of changing going on.