Clarkdale, SS and PI32M

[dinos22]

yes so what is your point?

[youngpro]

haha bro you cant compare efficiency with that formula for 32m

[dinos22]

youngpro i am disappointed with your clarkdale efficiency, even mAssman can kick your ass

[[XC] Oj101]

So how would you do it? It's the way I have for years

[dinos22]

there is no way to do it. you can get a rough estimate from it if you know the irregularities with the formula when it comes to different CPU clocks. We've been discussing formula back in AMD days but yeah no one has really come up with anything that can take all the variables into equation.

[massman]

You guys do realize that this thread was nothing about clock-per-clock, but rather score-per-cooling efficiency right? I mean ... I know in general people get all worked up when efficiency is compared (the irony - PP was supposed to be something factual) as it became, for some reason, a tool to measure super-dooper-tweaking-skills.

That no one was able to come up with anything is because everyone's lazy. You have all the variables that are measureable (score - cpu - NB - mem), do some scaling research on those, put weight factors on the variables and throw all that in a nice formula. What you're left with is some number that says something about the memory (sub)-timing settings and the OS optimisation. Of course, it's not as handy as the one we have today. Then it will be possible to say "look at me having the best efficiency with tCL=9". Woohoo! E-penis growth!

In the end, we'll just have the same discussion over and over again. Fanboys shouting their favorite hardware is more efficient (= "the best out there"), ego-tripping maniacs claiming their OS tweaking is far beyond anything any human has ever done and some arrogant brainiacs realizing it's always a balance between hardware limitations, hardware tweaking abilities and OS tweaking abilities. What you relate the end-result to is subjective and, quite frankly, irrelevant. The tragedy of all this is that an efficiency-rating takes away the first concern of all overclockers: pushing the CPU frequency. That's what matters the most in benchmarking and what will determine the end result the most. In other words, 'crippling' the system so much you lose 100MHz on the CPU, but gain on memory subsystem, is retarded. Oh, yeah, it will be very 'efficient', but in a very gay way. In fact, there should be a formula saying you have bad efficiency because you didn't manage to squeeze the best result out of your system.

In the end, it's all about making the most with the tools you have and the limitations you're facing. In my case, this was limited cooling (temps turning postive ... awesome!), memory sticks that can not be pushed over 1.7V (sorry, Leeghoofd) or put on extreme cooling. It's also about setting up challanges for yourself. I, for one, was into 'no compromise' overclocking where I forced myself to use the 2:10 divider because memory overclocking with Clarkdale is not very easy.

Then, just look up all the scores matching your hardware (cpu, cooling) and try to beat as much as possible by optimising your system on hardware and software level.

So, to get back at my original story.

Since this phase-change cooling sucks so damn hard, I'm very limited in terms of operating temperature. I did, however, manage to take down Ross' score with a lower (or equal) CPU clock, which was my original intend. It's good to assume I will need more CPU power if I ever want to take down 0Ro! So far, I've been able to run the highest CPU clock on the GD80 (5.32G consistently), but the fastest runs were on the M3F (5.28G + DDR3-2110). Too bad of the OCP kicking in on the H55M board, because it can run higher memory/qpi clocks than the others I tested so far, but stable CPU clocks are sub-par, resulting in worse results.

Onto P55-UD6.

[dinos22]

yes very true...sorry about the derailment

[chuchnit]

In other words, 'crippling' the system so much you lose 100MHz on the CPU, but gain on memory subsystem, is retarded. Oh, yeah, it will be very 'efficient', but in a very gay way. In fact, there should be a formula saying you have bad efficiency because you didn't manage to squeeze the best result out of your system.

hehe I wanna sig that. Good points you bring up though. You did exactly what I find myself doing when benching single stage though.

[massman]

Expanding the search field for more challenges:

- SuperPI 32M
- Clarkdale
- 5200 < Frequency < 5700
- No cooling limitations

- 419.219 by Ross with Core i5 655k @ 5628.40 and GA-H55N-USB3
- 424.594 by uncle fester with Core i5 655k @ 5600.00 and GA-H55N-USB3
- 427.625 by jabski with Core i3 530 @ 5696.00 and GA-P55-UD5
- 427.937 by Moonman with Core i5 655k @ 5608.00 and GA-P55A-UD7
- 430.531 by GraduS with Core i5 670 @ 5670.00 and GA-P55A-UD4
- 431.672 by pro with Core i5 655k @ 5388.00 and GA-H55N-USB3
- 432.047 by splmann with Core i5 670 @ 5664.00 and GA-H55N-USB3
- 436.859 by Hicookie with Core i3 530 @ 5500.00 and GA-H55M-USB3
- 437.406 by gbglassen with Core i5 670 @ 5659.00 and GA-P55-UD6-C
- 437.453 by trans am with Core i5 655k @ 5670.40 and GA-H55N-USB3
- 439.047 by 0Ro! with Core i5 661 @ 5589.60 and GA-P55-UD3
- 441.453 by 1Day with Core i3 530 @ 5588.80 and GA-H55N-USB3
- 442.547 by Massman with Core i5 655k @ 5276.00 and Maximus III Formula
- 442.719 by KkWong with Core i3 530 @ 5478.70 and GA-H55-UD3H
- 443.000 by Ross with Core i5 670 @ 5315.00 and GA-P55-UD6
- 443.516 by Moonman with Pentium G6950 @ 5481.30 and GA-H55M-UD2H
- 443.656 by chew* with Core i5 655k @ 5357.00 and Sabertooth 55i
- 443.891 by Massman with Core i5 655k @ 5317.00 and P55-GD80
- 444.343 by momoto with Core i5 670 @ 5586.50 and P55 FTW
- 445.766 by Toolius with Core i5 650 @ 5525.00 and GA-P55A-UD4P
- 446.078 by Massman with Core i5 655k @ 5221.80 and GA-H55M-UD2H
- 446.800 by 1Day with Core i3 530 @ 5632.60 and GA-H55N-USB3
- 446.800 by 1Day with Core i3 530 @ 5632.60 and GA-H55N-USB3
- 447.484 by 69_Goat with Core i5 670 @ 5587.30 and P55 Classified 200
- 447.781 by over@locker886 with Pentium G6950 @ 5646.00 and GA-P55A-UD7
- 448.640 by Massman with Core i5 655k @ 5217.00 and GA-H55M-UD2H
- 450.390 by 69_Goat with Core i5 670 @ 5405.00 and P55 Classified 200
- 450.422 by eagle58 with Core i3 560 @ 5648.00 and GA-P55-UD3R
- 450.547 by RULE with Core i5 670 @ 5670.00 and GA-H55M-UD2H
- 450.640 by tiborrr with Core i3 540 @ 5683.00 and Sabertooth 55i

[massman]

DDR3-2280 baby!

(now need to verify findings on other boards )

[dinos22]

WOW

[massman]

Cool! Also managed to beat 0Ro!'s score now. Also, here's the CPU-Tweaker shot that allowed me to go from a hardwall @ DDR3-2100 to whatever I reached.



//edit: what am I doing! CPU-Tweaker on HWBOT forums of course LOL

[Blue Storm]

P55-UD6 rocksvery nice efficiency PJ

[massman]

A close up of my mainboard can be found here: http://www.xtremesystems.org/forums/...d.php?t=234723

[M.Beier]

A close up of my mainboard can be found here: http://www.xtremesystems.org/forums/...d.php?t=234723

You should work for PR

Now stop teasing me about that f***ing ram frequency, I reached my goal, and will not try to compare with you, using this chip, its barely good for 950mhz mem, grrr

[0Ro!]

now I must take out the clarkbar and rebench it ... will retest it on P55A-UD6...maybe efficientier results can be made

[Perica_Barii]

Great efficiency massman....