SuperPI 32M for testing CPU and system speed and stability

[Robert]

This is Venice 3200+ @ 2890Mhz on air. It didn't stable at 2.9Ghz
I need more vcore.

ADA3200DAA4BP LBBLE 0515APAW
idle=43C load=52C 1.58v

CAS2.5 6-4-3 T2 1:1 HTT=304 1.7v

 

[Gautam]

So far so good. This should make the performance/MHz calcs interesting...

Intel Pentium M Dothan Revision C0, 2672MHz
Asus P4C800-E Deluxe, OCZ DDR Booster, VCore 1.55v, VDimm 3.3v
Kingston HyperX 3200 BH5 based 243MHz 5-2-2-2
Stock Asus cooler with Torin blower, temperature readings too inaccurate
Windows Server 2003 Enterprise Edition
Completed 32M pass in 25' 29.636"

Again, the temperature readings are just whacked, as is the +12v rail reading. BIOS reads it to be around 35C.

 

[Eldonko]

hitechjb1 I have some new results with different RAM. The BIOSs that were released today (0618-X) allow for higher CPU and RAM OCs. I also will be freezing my CPU next week with a custom single stager built by PC_ICE so will have more results to post then.

CPU: San Diego 3700+ CABGE 0515
memory: OCZ pc3200 Gold (BH) 2 x 512 MB
motherboard: DFI LanParty UT NForce4 Ultra-D BIOS 0618-3
cooling: H20
OS: Windows XP Professional SP2

- CPU: 2839 MHz = 258 MHz x 11, 1.7 V
- memory: 258 MHz, 2-2-2-5 1T, 3.6 V
- SuperPI 32M completed in 25 min 51.172 sec

EDIT: typo fixed

 

[Gautam]

You mean 2.89GHz... scared me for a second there.

 

[Clevor]

Too bad this forum doesn't allow larger attachments. 200 kb like Extreme would be nice. I didn't know you are supposed to open all the boxes on one screen (I assumed they were composite shots), so Bacchus didn't log one of my 32M runs in his UTT database. However I now open all the boxes on one screen, except it's hard to get them below 180 kb or so with decent quality.

Otherwise I would have three more BH-5/UTT runs to show: 270 with VX3200 , 267 with XMS3500, and 261 with VX4000.

But Hitechi, thanks for trusting me earlier .

 

[freeagent]

i have to wait till it gets a litle cooler before i can make a bombing run with my barton, looking at my tbred b in the results is makeing me feel nostalgic. i did the 32m in 35m 41.359s at a touch over 2600mhz, but i screwed up my screenie with my cpuz shots. hopefully itll rain soon and cool off the house, i dont expect this to be official, but since i spoke up i figured id at least show you the best i could lol

 

[deathstar13]

754 sempron +2800 2.6ghz @1.6v for one last trip on this bench.thats a 1ghz oc as its 1.6ghz stock.
dfi-nf3
ocz 3200 rev2 plat tcc5 on a 140 mhz devider which it seems nothing reads correctly.
below is the calculations showing its at 216 mhz on the memory.
cooling is air still with TT typhoon.
8 x 325 = 2600 7:10(141) x 325 = 216.66 3 x 325 = 975

this will be the last bench with this cpu.
i need more power and will replace it with a 1m DTR if not fully switch to 939.

 

[adriana]

ahh my score suxors

[/IMG]

 

[deathstar13]

ahh my score suxors

[/IMG]

with some help and a new psu id bet that thing would flat out fly!
just stick around and ask as many questions as you can and read as much as you can and eventually it will all come together.
most the people in this thread started where you are we just didnt pop a cpu in and say "done"

 

[hitechjb1]

Finally, it goes below 27 min SuperPI 32M, by disabling an unnecessary background ultility (a TV scheduler to be exact). The system is not just for benchmarking, but rather a system for daily usage with lots of programs and utilities installed, ....

CPU: Winchester 3000+ CBBHD 0447
memory: G. Skill 4400 LE 2 x 256 MB (Samsung TCCD)
motherboard: DFI LanParty UT NForce4 Ultra-D (rev. A02, bios 414-3)
cooling: XP-90, 90 mm Enermax fan (< 3000 rpm)
OS: Windows XP Professional SP2

- CPU: 2.80 GHz = 311 MHz x 9, 1.54 V (1.52 V works also)
- memory: 311 MHz, 2.5-4-4-8 1T, 2.8 V
- SuperPI 32M completed in 26 min 55.312 sec

With CnQ enabled (idle x5, under load x9)
- CPU: 2.77 GHz = 308 MHz x 9, 1.17/1.49 V
- memory: 308 MHz, 2.5-4-3-7 1T, 2.8 V
- SuperPI 32M completed in 26 min 56.456 min

 

[Gautam]

Windows 2000 is significantly faster than windows xp here, somewhere around 20 seconds. Windows 2003 should fall in between them. Also, you want to minimize your hard drive usage as much as possible, which is why 1 GB works better than 512 MB in this test, comparing to Clevor's Claw results by close to half a minute...so if you still have your UTT based memory, you might want to try it out.

 

[hitechjb1]

Thanks, points well taken.

I don't have a system just for benchmarking, otherwise I would explore the "100 system variables" and find out the performance sensitivity of each of them to minimize the benchmarking objective for individual benchmarking programs.

If I really want to push benchmarking results, I would build a system just for benchmarking, I would rebuild the OS every so and then, start with fresh Windows registry, use faster HD and more memory. disable as much background and networking utilities as possible, pick the most friend OS, create a performance sensitivity matrix for each benchmarking program vs the many system variables, ....

The system I am running the tests now is with XP SP2 which has been upgraded from Windows ME to XP to SP1 to SP2 over the years. There are lots of junks in the Windows registry and the system is not running with minimal background tasks. Probably all these add 30 sec to the SuperPI 32M run.

 

[Gautam]

Your point well taken too...I do most of that and it's a headache.

But I'd still like to see a straight UTT vs TCCD comparison if you've still got the former.

 

[hitechjb1]

I have done some testings and preliminary analysis on TCCD and UTT, including SuperPI runs.

Testing UTT and TCCD memory modules in Winchester and DFI NF4 Ultra-D setup


I don't have the UTT any more, because I don't like using 3.3+ V on the memory modules with the DFI 5V Vdimm option, too much stress is put on the DFI motherboard voltage regulator. Since the DFI board and TCCD memory I am using can run 300 - 320 MHz 2.5-3/4-3/4-x 1T which is better or equal to UTT/BH-5 at around 250-260 MHz with tight timing. I prefer the flexibility of TCCD over a wider memory frequency range and timing.

I am also not trying to squeeze the last 0.01% of benchmarking results (don't have the time and the video card and cooling hardware to get to the very top).

 

[Clevor]

Your point well taken too...I do most of that and it's a headache.

But I'd still like to see a straight UTT vs TCCD comparison if you've still got the former.

This is anecdotal info as I got the results checking for stability and did not take screenies: 264x11, 2-2-2-5 got a 27.963 1M run. With TCCD at 290x10, 2.5-3-3-6 got a 28.131 time. This is about same 2900 mhz clock. TCCD can still be plenty fast.

At 300x9, 2.5-3-3-6 I've broken 29 sec easily (upper 29s). Pretty good for 2700 mhz. I got slightly quicker times at 270x10, 2-2-2-6. I know I did not break 28 sec.

Board is DFI-UltraD with a 3700+ San Diego. I still say the 9NDA3+ is even faster at PI, 'cept I cannot run TCCD anywhere near as high (certainly not 300, 2.5-3-3-6 at 1T).

As far as SuperPI 32M with UTT/BH-5 at 260-270, who cares about the damn times - just completing a run is an achievement. Due to the 10x multi, only break 26' here. Beats doing it with 9x though. Will have to wait for phase to run 11x.

 

[Gautam]

Was the TCCD 2x512 or 2x256? That's the real issue in 32M. In 1M and other tests, yes, we've established the difference between UTT/BH5 and TCCD to be minimal. In 32M, though, the amount of memory plays a larger role.

 

[hitechjb1]

SuperPI 32M itself requires 268 MB. If SuperPI 32M and the rest of the programs/utility plus OS can fit under the total memory size, than paging won't occur and should not impact performance.

32M was tested with 2x256 MB. In my runs, I kept everything under 512 MB.

As SuperPI 32M uses quite a bit of memory, that is why it is good for testing the system performance and stability as a whole. But due to more system resources and variables are involved, the actual time would vary more and have wider spread.

Prime95 can be set to run even with more memory, up to very large virtual memory space to test system paging and stability.

 

[Clevor]

Was the TCCD 2x512 or 2x256? That's the real issue in 32M. In 1M and other tests, yes, we've established the difference between UTT/BH5 and TCCD to be minimal. In 32M, though, the amount of memory plays a larger role.

Oh yes, I only run 2x512 (let me make that perfectly clear . . .).

I think Eva2000 can also vouch that the 1MB L2 helps a bit too. T_M over at Eva's forums ran 264x11, 2-2-2-6 with a CAA2C Winnie and cannot break 27 sec in 1M. Eva says his 4000+ SD at 2750 beats his Winnies and Venices at 2850-2880 in pretty much everything.

Other tidbits: at 290x10, 2.5-3-3-6 I got a 40.53 (maybe 40.053?) PiFAST time.

But let's not ignore Five-Mark. How does 264x11 (2904 mhz), 2-2-2-5 compare to 290x10, 2.5-3-3-6? The BH-5 beats the TCCD by 400 points in Aquamark and around 250 each in 3DMark2000/2001 (31-34,000 territory though). I did not do PiFAST with the BH-5 but it will beat it for sure.

 

[Clevor]

I just finished running SuperPI 32M at 270x10, 2-2-2-6 @ 3.5 VDIMM with OCZ Gold Edition VX3200. You may have seen my post at Bacchus's thread at Extreme. Dang, I forgot to do Five-Mark, since didn't expect that would be a problem ('cept mebbe 3DMark2005).

Next time I get the sticks in I will get data. One day I will post a comparison between 300x9, 2.5-3-3-6 (got those benchmarks already) and 270x10, 2-2-2-6. Five-Mark, SuperPI 1M, PiFAST. I can't do 32M with the TCCD there yet, though. It will be closer than people think. 270x10 didn't feel that fast. And of course, it depends on how fast I can run the Trrd-Trwt timings on the TCCD; with UTT stuff, everybody and his sister is running 2-2-1-2.

Give credit where credit is due: the TCCD that I can run 300, 2.5-3-3-6 are PQI 2-2-2-5 DC kit and Vitesta PC600.

 

[Gautam]

Those some are some golden sticks you got there...very very impressive, and just 3200 at that.