qaywsx
asked on
slow timings: 203MHz slower than 179MHz?
I finished testing my new memory now!
I have a better score w/ 179x13 = 2327MHz than w/ 203x11.5 = 2335MHz. I think, the timings are the reason for this: at 179MHz I have CAS-TRP-TRCD-TRAS=2-2-2-7 as the fastest timings that work (Tras 7 is faster than Tras 5 and 6). At 203MHz the best is 2.5-3-3-8. The biggest difference makes the TRCD-timing.
Benchmark results (3DMark01SE, RAM:FSB=1:1 in all test):
179x13 = 2327MHz (2-2-2-7): 14238 (Prime stable 24 hours)
179x13 = 2327MHz (2-2-2-5): 14201 (not tested for stability)
181x13 = 2353MHz (2,5-2-3-7): 14102 (not tested for stability)
181x13 = 2353MHz (2,5-2-3-6): 14042 (not tested for stability)
181x13 = 2353MHz (3-3-3-8): 13947 (Prime stable 6 hours)
188x12,5 = 2350MHz (2,5-2-3-7): 14053 (not tested for stability)
188x12,5 = 2350MHz (3-3-3-8): 14057 (Prime stable 6 hours)
203x11,5 = 2335MHz (2,5-3-3-8): 14043 (Prime stable 30 min.)
203x11,5 = 2335MHz (3-3-3-8): 14081 (Prime stable 6 hours)
My question: Can I also expect in games, that 203x11.5 is slower than 179x13, just because the timings are lower althought FSB and overall clock are faster? What setting do you think is the fastest in games?
About memory-voltage: I use 2.77V and this will not change, if it isnt stable at a lower voltage: I have posted in 32 forums what voltage I should use and 32 people gave replies, which answered the question: 16 (50%) said, that I should use 2.9V or more, 13 (40.6%) said I should use 2.77V and 3 (9.4%) said I should use 2.63V. I will test, if 2.63V is also stable at the best setting I can get w/ 2.77V.
[URL=http://www.nethands.de/pys/show.php4?user=MrBurns]My System:[/URL]
AMD Athlon XP 2700+
Epox 8RDA+
2x512 MB DDR-400 Infineon
Thermaltake Volcano 7+
Creative GeForce 4 Ti 4600 w/ 8xAGP, 128MB
80GB/8MB Western Digital HDD
16x48 Liton DVD-ROM
Acer 20x10x40 Burner (doesnt work anymore)
I have a better score w/ 179x13 = 2327MHz than w/ 203x11.5 = 2335MHz. I think, the timings are the reason for this: at 179MHz I have CAS-TRP-TRCD-TRAS=2-2-2-7 as the fastest timings that work (Tras 7 is faster than Tras 5 and 6). At 203MHz the best is 2.5-3-3-8. The biggest difference makes the TRCD-timing.
Benchmark results (3DMark01SE, RAM:FSB=1:1 in all test):
179x13 = 2327MHz (2-2-2-7): 14238 (Prime stable 24 hours)
179x13 = 2327MHz (2-2-2-5): 14201 (not tested for stability)
181x13 = 2353MHz (2,5-2-3-7): 14102 (not tested for stability)
181x13 = 2353MHz (2,5-2-3-6): 14042 (not tested for stability)
181x13 = 2353MHz (3-3-3-8): 13947 (Prime stable 6 hours)
188x12,5 = 2350MHz (2,5-2-3-7): 14053 (not tested for stability)
188x12,5 = 2350MHz (3-3-3-8): 14057 (Prime stable 6 hours)
203x11,5 = 2335MHz (2,5-3-3-8): 14043 (Prime stable 30 min.)
203x11,5 = 2335MHz (3-3-3-8): 14081 (Prime stable 6 hours)
My question: Can I also expect in games, that 203x11.5 is slower than 179x13, just because the timings are lower althought FSB and overall clock are faster? What setting do you think is the fastest in games?
About memory-voltage: I use 2.77V and this will not change, if it isnt stable at a lower voltage: I have posted in 32 forums what voltage I should use and 32 people gave replies, which answered the question: 16 (50%) said, that I should use 2.9V or more, 13 (40.6%) said I should use 2.77V and 3 (9.4%) said I should use 2.63V. I will test, if 2.63V is also stable at the best setting I can get w/ 2.77V.
[URL=http://www.nethands.de/pys/show.php4?user=MrBurns]My System:[/URL]
AMD Athlon XP 2700+
Epox 8RDA+
2x512 MB DDR-400 Infineon
Thermaltake Volcano 7+
Creative GeForce 4 Ti 4600 w/ 8xAGP, 128MB
80GB/8MB Western Digital HDD
16x48 Liton DVD-ROM
Acer 20x10x40 Burner (doesnt work anymore)
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
I would think that without changing the mutliplier that you should be able to overclock 10-15%
ASKER
> I think that this has a lot to do with your CAS latency setting. Obvisously you know the
> difference, but there is a big difference between 2 and 2.5 or 2 and 3.
From my benchmark resukts it looks, like that the CAS-timing doesnt make any difference at all: I made 3 tests at CAS2.5 w/ 179MHz, 3 tests at CAS 2 w/ 179MHz, 1 test at CAS 2.5 w/ 203MHz and 3 tests w/ CAS 3 at 203MHz. The result was, that the CAS-timing never made a big difference, all differences were much smaller than the random error of 60. Tests, that only had a difference in the CAS-setting:
179x13 (2.5-2-2-7): 14241 179x13 (2-2-2-7): 15238
179x13 (2.5-2-2-5): 14188 179x13 (2-2-2-5): 14201
203x11.5 (3-3-3-8): 14081 203x11.5 (2.5-3-3-8): 14043
So you see, once I get -3 from the lower CAS, once +13 and once -38, so this shows, that a lower CAS-setting doesnt make any w/ 3DMark measureable difference.
The only timing, that made a difference of at least 200 points in my banchmark was TRCD.
> 1. in order to get your clock to 203 you cranked up the V-core voltage creating more heat then you would have at 173.
I always used a vcore of 1.75V (default is 1.65V) for testing my max. oc, because I know, that 1.75V is the max. I want to have, because my CPU should last at least 3 years and I dont want to risk a sudden death syndrom. I dont have any heat problems, the max. CPU-temp I had is 47°C in prime95. I use CPUCool to measure my temps in Windows. The program reports the same temps as the BIOS.
When I finished my stabilty testing w/ a memory voltage of 2.63V, I will try, if my system is also stable at a lower Vcore w/ the clock I will decide to use.
> Have you tried increasing your multiplier? This my be impossible now depending on how you unlocked the lower multipliers.
My multiplier is unlocked from the factory, because I have a Thoroughbred B. I know, that on some mainboards you have to connect bridges if you want to unlock the multis even on Tbred B CPUs, butthis is because these mainboards lock the multiplier by default, but my mobo is unlocks the multi by default when used w/ a Tbred B, so I can use all multis available in the BIOS.
I tried higher multis, the highest stable clock I got at 13.5x175=2363MHz, but I only get 138xx 3DMarks w/ this clock. w/ 14x166 I even get a lower overall clock, so the best multiplier seem sto be 13 or 11.5, but according to 3DMark 13 is better because of my RAM, but I want to know, if this is also w/ games.
@buckeyes33: The highest clock I got w/o chjangingf the multi was 181x13=2353MHz, my default clock is 166MHz, which is an overclock of 8,6%. I know from other forums, that this isnt bad for an 2700. The high models are always not as ocable as the low models, and the fatsest available Tbred is only 66MHz fasterat stock than mine. Also I think, that my CPU has an old revision.
> difference, but there is a big difference between 2 and 2.5 or 2 and 3.
From my benchmark resukts it looks, like that the CAS-timing doesnt make any difference at all: I made 3 tests at CAS2.5 w/ 179MHz, 3 tests at CAS 2 w/ 179MHz, 1 test at CAS 2.5 w/ 203MHz and 3 tests w/ CAS 3 at 203MHz. The result was, that the CAS-timing never made a big difference, all differences were much smaller than the random error of 60. Tests, that only had a difference in the CAS-setting:
179x13 (2.5-2-2-7): 14241 179x13 (2-2-2-7): 15238
179x13 (2.5-2-2-5): 14188 179x13 (2-2-2-5): 14201
203x11.5 (3-3-3-8): 14081 203x11.5 (2.5-3-3-8): 14043
So you see, once I get -3 from the lower CAS, once +13 and once -38, so this shows, that a lower CAS-setting doesnt make any w/ 3DMark measureable difference.
The only timing, that made a difference of at least 200 points in my banchmark was TRCD.
> 1. in order to get your clock to 203 you cranked up the V-core voltage creating more heat then you would have at 173.
I always used a vcore of 1.75V (default is 1.65V) for testing my max. oc, because I know, that 1.75V is the max. I want to have, because my CPU should last at least 3 years and I dont want to risk a sudden death syndrom. I dont have any heat problems, the max. CPU-temp I had is 47°C in prime95. I use CPUCool to measure my temps in Windows. The program reports the same temps as the BIOS.
When I finished my stabilty testing w/ a memory voltage of 2.63V, I will try, if my system is also stable at a lower Vcore w/ the clock I will decide to use.
> Have you tried increasing your multiplier? This my be impossible now depending on how you unlocked the lower multipliers.
My multiplier is unlocked from the factory, because I have a Thoroughbred B. I know, that on some mainboards you have to connect bridges if you want to unlock the multis even on Tbred B CPUs, butthis is because these mainboards lock the multiplier by default, but my mobo is unlocks the multi by default when used w/ a Tbred B, so I can use all multis available in the BIOS.
I tried higher multis, the highest stable clock I got at 13.5x175=2363MHz, but I only get 138xx 3DMarks w/ this clock. w/ 14x166 I even get a lower overall clock, so the best multiplier seem sto be 13 or 11.5, but according to 3DMark 13 is better because of my RAM, but I want to know, if this is also w/ games.
@buckeyes33: The highest clock I got w/o chjangingf the multi was 181x13=2353MHz, my default clock is 166MHz, which is an overclock of 8,6%. I know from other forums, that this isnt bad for an 2700. The high models are always not as ocable as the low models, and the fatsest available Tbred is only 66MHz fasterat stock than mine. Also I think, that my CPU has an old revision.
Have you tried anything different?
ASKER
If you mean, if I tried changing the multiplier: I tried all multis from 11 to 14.
w/ 13.5 I got the highest clock: 175x13=2363MHz, but I had lower 3DMarks (14119@2.5-2-2-6) than w/ 13x179.
w/ 11 I got the highest FSB (11x206MHz=2266MHz), but it wasnt much higher than w/ 11.5, so I didnt even benchmark.
w/ 13.5 I got the highest clock: 175x13=2363MHz, but I had lower 3DMarks (14119@2.5-2-2-6) than w/ 13x179.
w/ 11 I got the highest FSB (11x206MHz=2266MHz), but it wasnt much higher than w/ 11.5, so I didnt even benchmark.
If I was you I would just go with what has the fastest marks. Have you used any other benchmarking software? Sandrasoft
ASKER
For all other multis see original Thread (I didnt post multi 12 and, buecause I didnt got a usefull clock w/ these multis).
ASKER
@buckeyes: I havent used any other benchmarking software yet, but I will use Unreal Tournament 2003.
I dont know, when I will do the benchmark, because I have a big exam at Friday on the uni, and I had to reinstall my system because PartitionMagic (boot disk version!) crashed when I wanted to make additional partitions for WinXP and Linux (I will nver use that program again). I had all data backed up, but it still takes a lot of time to reinstall all the programs.
I dont know, when I will do the benchmark, because I have a big exam at Friday on the uni, and I had to reinstall my system because PartitionMagic (boot disk version!) crashed when I wanted to make additional partitions for WinXP and Linux (I will nver use that program again). I had all data backed up, but it still takes a lot of time to reinstall all the programs.
Mhz Is not the be all and end all.
slower CPU speed can often be faster than Higher CPU speed. It really depends on CPU architecture, but at the end of the day memory is a damn site slower than most CPU's and slowing the CPU to the benefit of memory bandwidth will usually yield a benefit.... ie the chip is getting the data it needs at a faster rate and isn't idly sitting back twiddling its thumbs.... or whatever CPU's have that are opposable
slower CPU speed can often be faster than Higher CPU speed. It really depends on CPU architecture, but at the end of the day memory is a damn site slower than most CPU's and slowing the CPU to the benefit of memory bandwidth will usually yield a benefit.... ie the chip is getting the data it needs at a faster rate and isn't idly sitting back twiddling its thumbs.... or whatever CPU's have that are opposable
any more questions or comments. If not you should probably close out this thread.
ASKER
I finally had the time to make a few UT2k3 benchmarks, and this benchmark were also faster w/ 179x13, althought the difference was max. 1 fps in most of these tests, but ins some it was even 3-5fps.
Because of this results I decided to use the 179x13 setting.
First I used the Benchmark.exe in the \UT2003\Benchmark directory, than I used the [H]ard|Ocp UT2K3 Benchmark Utility v2.1, which can be downloaded here: http://www.hardocp.com/article.html?art=NDQy
My results (I only posted the overall/avarage score, click on the links for
details):
Benchmark.exe (all test were run @ 1024x768):
179x13:
Flyby: 171.84
Botmatch: 78.61
203x11.5:
Flyby: 171.23
Botmatch: 77.39
I have lost the file w/ the details of this tests and I haven written them down.
[H]ard|Ocp UT2K3 Benchmark Utility v2.1 (all test are Flyby):
CPU-Test:
179x13:
640x480: 111.94
1024x768: 110.43
Details: http://members.chello.at/edith.matuskovics/results_179x13_cpu.htm
203x11.5:
640x480: 110.44
1024x768: 109.58
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_cpu.htm
low details:
179x13:
640x480: 271.38
1024x768: 218.13
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
203x11.5:
640x480: 266.12
1024x768: 215.34
Details: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
high details:
179x13:
640x480: 199.31
1024x768: 141.06
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
203x11.5:
640x480: 196.17
1024x768: 140.72
Details: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
Because of this results I decided to use the 179x13 setting.
First I used the Benchmark.exe in the \UT2003\Benchmark directory, than I used the [H]ard|Ocp UT2K3 Benchmark Utility v2.1, which can be downloaded here: http://www.hardocp.com/article.html?art=NDQy
My results (I only posted the overall/avarage score, click on the links for
details):
Benchmark.exe (all test were run @ 1024x768):
179x13:
Flyby: 171.84
Botmatch: 78.61
203x11.5:
Flyby: 171.23
Botmatch: 77.39
I have lost the file w/ the details of this tests and I haven written them down.
[H]ard|Ocp UT2K3 Benchmark Utility v2.1 (all test are Flyby):
CPU-Test:
179x13:
640x480: 111.94
1024x768: 110.43
Details: http://members.chello.at/edith.matuskovics/results_179x13_cpu.htm
203x11.5:
640x480: 110.44
1024x768: 109.58
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_cpu.htm
low details:
179x13:
640x480: 271.38
1024x768: 218.13
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
203x11.5:
640x480: 266.12
1024x768: 215.34
Details: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
high details:
179x13:
640x480: 199.31
1024x768: 141.06
Details: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
203x11.5:
640x480: 196.17
1024x768: 140.72
Details: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
ASKER
Oops, I mixed up the links for the high and low details (cpu-test is right) and only posted the links for the low details. It should be:
low:
179x13: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
203x11,5: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
high:
179x13: http://members.chello.at/edith.matuskovics/results_179x13_high.htm
203x11,5:http://members.chello.at/edith.matuskovics/results_203x11,5_high.htm
low:
179x13: http://members.chello.at/edith.matuskovics/results_179x13_low.htm
203x11,5: http://members.chello.at/edith.matuskovics/results_203x11,5_low.htm
high:
179x13: http://members.chello.at/edith.matuskovics/results_179x13_high.htm
203x11,5:http://members.chello.at/edith.matuskovics/results_203x11,5_high.htm
those our still some good scores.