DDR1 SuperPI 32M 2.5/3.0GHz results

[Sparksnl]

Hi all, please excuse me for my english. Be free to talk German in here. I can read most of it.

I was thinking about Super PI 32M results for DDR1. The one thing I still miss is a list of best results for a given speed.
So here it is, rules for posting:

- There is a 2.5GHz, 3GHz and 3.5GHz cpu speed list.
- CPU speed can be 10MHz lower or 5 MHz faster max.
- CPUz validation is a must.
- A screenshot is a must: including CPU main tab, mainboard, memory (latest normal or vintage CPUz version) and Super PI.
- Super PI Mod 1.5XS is to be used.
- PI 32M has to be run.
- DDR1 is to be used.

Please post including a string like this:
CPU Brand and type | CPU frequency | mem frequency | mem timings | dimm voltage | Memory brand and type | Socket | Name | OS | Link | time (min / sec)

Good idea’s? Please post them.

2.5GHz
…

3.0GHz
…

3.5GHz
…

 

[Biolante]

there is one hardware-config to smash all other DDR1 S-pi results. if you use Asrocks 775 Agp/DDR1 hybrids + Yorkfield...good luck to beat this lol
this is merely a cpu benchmark and OS efficiency test at this point.

 

[Sparksnl]

No problem if that is faster. You will see results per platform in the list.

 

[digitalbath]

No problem if tgat is faster. You will see results per platform in the list.

okay, that would be my question. A seperation per socket should be nice and more comparable. Are there any other limitation in plattforms?
I would like to make some results in S462 and S478. With these test, I can improve the effiency in sips and dimm timings.

e. is it possible to add 3,5Ghz for S478?

 

[Sparksnl]

I can still change or make the rules. 3.5GHz can be added. A per platform list seems wise.
Edit: I changed the string for info about the result.

 

[The Sandman]

First attempt

2.5GHz
Athlon 64X2 4800+ Rev. E6 | 2497-MHz | 208-MHz | 2-3-2-5 | 2.5V | 2x 512MB MDT PC400 CL2.5 | Abit AV8 | 939 | The Sandman | Win XP | Result, Valid | 30m 01.594s

Okay?

 

[ItsFun]

3.0GHz
Opteron 170 Rev. E6 (LCBQE 0712UPMW) | 3000 MHz | 300 MHz | 2,5-4-4-8 | 2.8V | 2x 512MB Corsair 4400C25 | Shuttle SN25P V2.0 | 939 | @ItsFun | Win XP | Result, Valid | 25m 10.688s

 

[digitalbath]

2.5GHz
XP-M 2400+ (AQYHA0711XPMW) | 2500 MHz | 250 MHz | 2-2-2-5 | 3.2V | 2x 256MB Infineon AT-6 | ASRock K7NF2-RAID | 462 | @digitalbath | Win XP | Result, Valid | 34m 00.875s

first try. I will try to make some tweaks with timings and sips to get a better result.

 

[Sparksnl]

I will edit the list soon. I see I missed some scores.
Edit: Seems Ok to me so far. I shall make a run in a couple of weeks as well.

 

[emissary42]

@Sparksnl There already is a 32M list like that in the memory sub, with a very similar format and ruleset. If you are interested, I'll add any DDR1/DDR2 platforms with lower target CPU frequencies.

[Sammelthread] - Der Hardwareluxx SuperPi 32M Effizienz Wettbewerb @ 6.0 / 5.0 / 4.5 / 4.0 GHz & Retro (AMD+Intel) [V3]

Der HardwareLUXX SuperPi 32M Effizienz Wettbewerb @ 6.0 / 5.0 / 4.5 / 4.0 GHz & Retro (AMD + Intel) 0.) Vorwort Hier geht es nicht um absolute Stabilität oder den maximalen Takt, sondern um die Effizienz des Speichers und Systems. Als Maßstab dient der SuperPi Benchmark. Das Ziel ist die...

www.hardwareluxx.de

 

[Sparksnl]

Interesting. That would be better. 2 lists is not effective.
If others are not against it?

 

[emissary42]

Agreed, a single thread would probably end up being more active in the long run. Let me know if there any rules that have to be adjusted for very old platforms. Also my allowed deviation from target frequency are stricter than yours and I am not sure how well that works out, with lets say only up to +2MHz at 3GHz and +1.5MHz at 2.5GHz (+1MHz at 2GHz). In theory any amount of being under-frequency should be fine, if the handicap is worth it for the overall score/time.

 

[Tzk]

I am not sure how well that works out,

At least on nforce 2 this isn't an issue as we got a new tweaker which got a very fine range for fsb adjustments. It's possible to adjust in ~0.1Mhz steps. So +1.5Mhz for the cpu clock should be achievable. I just looked through my old results and by default my nf2 board applies 250.6Mhz FSB when set to 250 in bios. This results in 2506Mhz cpu clock at multi x10 which is by far out of your submission window.

 

[digitalbath]

At least on nforce 2 this isn't an issue as we got a new tweaker which got a very fine range for fsb adjustments. It's possible to adjust in ~0.1Mhz steps. So +1.5Mhz for the cpu clock should be achievable. I just looked through my old results and by default my nf2 board applies 250.6Mhz FSB when set to 250 in bios. This results in 2506Mhz cpu clock at multi x10 which is by far out of your submission window.

As far as I know, the FSB steps in the Tweaker depend on the FSB speed. In higher FSB ranges, the steps are higher than 0.1 MHz. For example: I could only use 260.00MHz (makes 259,xx Mhz) FSB or 261,xx MHz.
Then, there are differences between boards or probably BIOS manfacturers (AWARD vs. AMI). Setting 250,6MHz in the tweaker on my ASRock board (AMI BIOS) results i a perfect 250,00MHz. The board sets always a ~0,6MHz lower FSB clock then the tweaker (all FSB tweaker!) setting. AWARD related boards don't do this though.

The tighter deviation target is not a problem i.m.o. Applying a lower FSB clock is always doable.

 

[Sparksnl]

So then I see no issue to close this thread and integrate it in the other one.

Edit: please give me an update when the thread is updated.

 

[emissary42]

Machen wir so. Hier geht es also weiter:

[Sammelthread] - Der Hardwareluxx SuperPi 32M Effizienz Wettbewerb @ 6.0 / 5.0 / 4.5 / 4.0 GHz & Retro (AMD+Intel) [V3]

Der HardwareLUXX SuperPi 32M Effizienz Wettbewerb @ 6.0 / 5.0 / 4.5 / 4.0 GHz & Retro (AMD + Intel) 0.) Vorwort Hier geht es nicht um absolute Stabilität oder den maximalen Takt, sondern um die Effizienz des Speichers und Systems. Als Maßstab dient der SuperPi Benchmark. Das Ziel ist die...

www.hardwareluxx.de

Bitte die Regeln beachten und das Timing Tool (CPU-Tweaker o.ä.) im Screenshot nicht vergessen. Im Beispiel-Screenshot ist zu sehen, was erforderlich und optional ist.

Kategorien für neue Plattformen/Sockel werden mit den ersten Ergebnissen hinzugefügt.