Super Pi 1M Performance to MHz Ratio: Post Yours!

[T_M]

Most of the tricks are in the screeny

nice one funk, where'd you get the 300x8 idea - heheheh
PS - you coulda got fractionally lower if you had let CPU-Z idle for a while and capture the screeny when the CPU MHz floated to 2400.5

[dinos22]

Wanna share the tricks?

yeap

make sure you run at 2.4GHz with BH5 RAM maxed out at 267 with tightest possible timings....surprised he didn't use MAL=7...guess the Opteron must have been upset

few of the registy tweaks, fast hard drives, SuperPI in root directory, run 16k a few times before going for 1M.........and run a lot of 1M runs.....

for better results than T_M's >>>>>> Win2003 Server

[T_M]

yep, still trying to get my hands on a working 2003 (been trying for weeks now...)

MAL 7 at 300HTT is just unstable
maybe if it was 290.... but not 300

[Aphex_Tom_9]

so i guess it's all about the ram frequency/timings

[misteroadster]

hummm, after some thinkings , the formula is good, but the rules aren't, to compete it must be a range of frequency :
Logically at low cpu frequency you need less bandwith than at high.

As i said previously, superpi times are propotionnals with the same ratio CPU/RAM, so to compete with 2000/200 you need 3000/300 or 4000/400.

[MaxxxRacer]

sigh.. well I guess everyone is gonna be running at exactly 2.4ghz now..

[misteroadster]

hummmm me think a range of 2.4-2.6Ghz would be better.
There is more choice in term of settings as well

[dinos22]

sigh.. well I guess everyone is gonna be running at exactly 2.4ghz now..

pretty much.....because the lower you go the better your results will be

run below 2.4 and see what you'll get

you shouldn't complain....you've got 300MHz 2.5-3-3-6 stable TCCD.......mine are memtest stable up to 295/296 at those timings but not in windows

I hate my expert mobo............it will not run half multies for the life of me

[T_M]

SO why not update the list when you were in the thread maxxx?

[MaxxxRacer]

It will take about 40 minutes of copy paste, copy pate, copy paste to get it done. I will do it tonight.


Dinos, its actaully only 2.5-4-3-6 stable.. Ive been trying to mess with certain settings to get it stable at 2.5-3-3-6 but its really hard as it takes 20 minutes for it to error out even at 2.5-3-3-5-7-10 at 300mhz. And If I cant boot one mhz higher than 300 or else it will reboot. Because of this I cant stest to see which settings give the least errors. I will try again later tonight. Will be messing around with drive strength and a few others.

[dinos22]

Dinos, its actaully only 2.5-4-3-6 stable.. Ive been trying to mess with certain settings to get it stable at 2.5-3-3-6 but its really hard as it takes 20 minutes for it to error out even at 2.5-3-3-5-7-10 at 300mhz. And If I cant boot one mhz higher than 300 or else it will reboot. Because of this I cant stest to see which settings give the least errors. I will try again later tonight. Will be messing around with drive strength and a few others.

ok maybe i mixed up your and Bachus's posts..........

btw what is the difference between memtest stable and windows stable speeds for you on TCCD RAM...........I think mine is at least 10MHz

on BH5 i tested in the past is a different story....basically whatever i got memtest stable was also windows SPI stable........actually even when it would error in memtest I could get it 32M stable in windows....that was a surprise i remember but hell it didn't stop me trying

p.s. sup with your sig

[chew*]

So dual cores can compete to if we use this formula right? >>>>
core 0 = 27.703 * 2951.7 / 2000 = 40.88547255
core 1 = 27.699 * 2951.7 / 2000 = 40.87956915
40.88547255 + 40.87956915 / 2 = 40.88252085

Will that be acceptable?

[MaxxxRacer]

the difference betwen memtest and windows stable is the same.. 307mhz in memtest and windows.

Chew, no need to aveage the scores. Just run one of the cores.. the better one if u can.

[chew*]

the difference betwen memtest and windows stable is the same.. 307mhz in memtest and windows.

Chew, no need to aveage the scores. Just run one of the cores.. the better one if u can.

cool core 1 it is then BTW i posted for some of your opinions on cooling an x1800xt in the watercooling thread. Need help can't make up my own mind.

[MaxxxRacer]

OK, its all updated now.

removed the numbering as it is just a pain in the ass to mess with the numbers when organizing... Im no freeking computer so sorting takes more time than my T1i-89 sorting Ebay's daily listings. (takes 24 hours wherehouse full of powerful servers)

[chew*]

33.859 x 1mhz / 2000 = 0.0169295



After you get a laugh out of that ill explain what happened later and post a real score as im still dialing in this ati board

[dinos22]

33.859 x 1mhz / 2000 = 0.0169295



After you get a laugh out of that ill explain what happened later and post a real score as im still dialing in this ati board

beat that Dothan bytches

[killingspreez]

33,641s*2403,5/2000 = 40,428

40,428---Killingspreez----------3700 SD

here you go:



mfg

[T_M]

Update

39.869 --- T_M ---------- Opteron 146

[FireDragon]

2.7x29/2000= 39.15

245x11

opty 175

Dragon

[MaxxxRacer]

LOL Chew*


dinos, what about my sig?

[funkflix]

The link of the second placed does not work properly.

Thx for the list and work MaxxxRacer!

[dinos22]

LOL Chew*


dinos, what about my sig?

he's a fanboy he wouldn't know any better

also are those figures right ONLY 110W burn

[T_M]

2.7x29/2000= 39.15

245x11

opty 175

Dragon

you have read the rules right?
SUperPI mod 1.4 (i.e. 3 decimal points behind the seconds)
MHz on CPU as accurate as you can show (CPU-Z shows .1 accuracy)
Screenshot of the run including CPU-Z and A64 Tweaker.

[MaxxxRacer]

T_M, dont worry, Fire's score didnt go up for those reasons.


Pretty soon I will also stop posting scores that dont follow the posting rules. IE. the correct format. For now I am fixing them, since its only a few who dont read the rules, but its very time consuming to have to do alot of them.

Dinos, yes thats right. 110w read BEFORE the mosfets. that means the 110w is what the cpu and mosfets are drawing. So it doesnt take into account the innefeciency of the MOSFET. all of that means that the actual power draw is a few watts lower, but its not possible to take that reading.

funk, I will fix the link.