Which is faster, 9x333 or 8x375 on a q6600? Results inside

[Clunk]

Superpi stores your fastest time in the list that you have posted, but I was meaning something more like this...



If I posted my list, it would show a 13.x second 1M run, which isnt correct for this run, if you see what I mean?

As you can see from the pic, it isnt even close.

[graysky]

Wonder if it's because you're running 4:5 DRAM:CPU? I keep it 1:1. Also, I think we're using different superpi versions. If I do an about I get version 1.1... I see from your screenshot that you have some modified version. Here is the screenshot you requested:

[Supershanks]

Good work graysky! Always nice when the numbers prove your common sense and intuition correct.

Like Clunk i'm a bit uncertain about that 16s result.

it's an old version of superPi i think the best is the modded as this uses verification, but more importantly goes to millisecs. Super PI Mod 1.5 Using v1.1 for short runs is a bit like timing a 100m sprint without a second hand

I think in general you'll find a slight improvement in running at a lower multi/higher fsb:-

9x333 SuperrPI 1M

8x375 SuperrPI 1M

9x333 mem 1067 SuperPi 32m + metabench

8x375 mem 1067 SuperPi 32m + metabench


I've used metabench for a long while the score is an average of 50+ tests which give a reasonable emulation of usage see 7Bytes Metabench comparatively may give a better indication than superpi, although there is a reasonable correlation in my results.

At the end of the day there are several ways to get to the same point. It depends on your equipment & preferences which road you go. if using cheap ram you user the higher multiplier lower and/or strap.

Have seen similar discussions around the amd64, results here are the same - slight improvement from lower multi higher fsb.

[Clunk]

Wonder if it's because you're running 4:5 DRAM:CPU? I keep it 1:1. Also, I think we're using different superpi versions. If I do an about I get version 1.1... I see from your screenshot that you have some modified version. Here is the screenshot you requested:

This is the version that everyone uses.

It is actually slower using 1:1.

This is why ours has a validation checksum and yours doesnt.

I will try the standard version and see what the results are.

[graysky]

Cool... I'll d/l and re-run the 9x333 for you guys.

[graysky]

Here you go: 17.2 sec, so it's consistent.



I'll bet if you d/l version 1.1 and test your q6600 on it, you'll hit 16 sec as well.

[graysky]

@supershanks: in the lower part of your screen shot, you have a display of temps and voltages in a 6x2 grid. What software is that?

[Clunk]

Here you go: 17.2 sec, so it's consistent.

I'll bet if you d/l version 1.1 and test your q6600 on it, you'll hit 16 sec as well.

Its much slower.

Very possibly, but that version only rounds up to the nearest second, which isnt too good for trying to illustrate something that takes less than 18 seconds to complete, thats why nobody uses it any more.

I think the analogy is " Like measuring the 100 metres without a second hand on the stopwatch"

[Supershanks]

motherboard monitor seeOfficial Motherboard Monitor 5 (MBM5) Update 3 with Core(TM) Temp plugin

Here you go: 17.2 sec, so it's consistent

So your 8x375 comnes to what sorry?? be nice to sdee the straight comparison

That's why the mod version of superpi came from as the earlier versions were flaweed, with results being manipulated. 1.2S is a fair old difference, on a short 1m run.

[Clunk]

edit: pics broke..brb

[graysky]

So your 8x375 comnes to what sorry?? be nice to sdee the straight comparison

Ah, I think I'm gonna leave sleeping dogs lie on this one. I don't wanna repeat those experiments. The more I think about it, including super pi in the "real world" category was a mistake... unless you calculate pi out to 32 million places on a day-to-day basis

[Supershanks]

including super pi in the "real world" category was a mistake... unless you calculate pi out to 32 million places on a day-to-day basis

Sorry so your now saying what?? There is a difference or there isn't a difference??

What's wrong with looking at metabench that i linked to b4, aren't 50+ different tests a sufficient simulatiomn of real world??

[graysky]

I mean I'm too burnt out on changing settings and running benchmarks for a while Given the data I collected, there isn't a difference. I might have a look at metabench later in the week... are you up for the testing w/ metabench on your 6600?

[Phil_P]

Ah, I think I'm gonna leave sleeping dogs lie on this one. I don't wanna repeat those experiments. The more I think about it, including super pi in the "real world" category was a mistake... unless you calculate pi out to 32 million places on a day-to-day basis

Depends what you do in the real world. I run floating point/FFT transformations (SETI) all day long, 24/7, so to me it's pretty important... but I take your point

[Supershanks]

Depends what you do in the real world.

Absolutely there will be no benchmark that provides a perfect match but try running Metabench & looking at the ramge of tests covered, looks fairly comprehensive to me.

As i said earlier

At the end of the day there are several ways to get to the same point. It depends on your equipment & preferences which road you go. if using cheap ram you user the higher multiplier lower and/or strap.

What most people aim at is maximising cpu clock. Having done that i often see if i can squeeze the last drop out. Occasionally by trying playing with multi, more often it will be maximising memory through changing strap.

[Supershanks]

I mean I'm too burnt out on changing settings and running benchmarks for a while

Think it's more the effort of keeping upto date with all the threads on this u have:-
9x333, 8x375, or 7x428 on a Q6600 - Which is faster? Toms hardware
PCPerspective
9x333, 8x375, or 7x428 on a Q6600 - Which is faster? Xtreme
[h]ardforum
Extreme Forums

I might have a look at metabench later in the week... are you up for the testing w/ metabench on your 6600?

Results are already there, at least for your 375x8 & 333x9 options see earlier post. But yep will run a few if u want, just remember where you heard of it