Discussion in 'Hardware' started by Bloody_Pete, 14 Dec 2011.
The loop is run off a 360 Radiator sat infront of the window, only three fans on it (need to upgrade to push/pull high RPM fans when I've got some money spare) Loop order is; Res, Pump (Swiftech MCP 335 W/EK top.) VRM block, CPU block (Heatkiller 3.0) NB block, Rad, res again. The entire loop is bodged on the wiring end, but the loop itself is sound (Bloody well should be after all the work I had to do to get it to full potential)
I just wonder if the Heatkiller's sheer restriction is slowing down the water through the NB block. (EK GA AMD kit, btw. SB block doesn't fit on any of the 9xx series chipset boards, though, so I'm not using that.)
Nice work! I doubt I'll be catching that. Not with this chip anyway. I may have got your last one, but I think this is a little outside what this chip can do. I've never yet managed to run anything at 5.4. I can get to the desktop and that's about it.
What were your volts?
What clock speed was that last one run at? Not 4.7 surely.
Must have been at 5Ghz surely?
5Ghz with a H80 and be stable? Please, show me how!!!!
Download Intel Burn Test. Set it to Maximum stress level, and run it for at least 20 loops. If it passes, it's stable.
Any stress level less than Max will not use all your RAM. I can get this to say it's stable at 4.7 with around 1.28V when set to standard stress level, but when set to max, it needs 1.36. It makes a difference.
Unless you test using all your RAM it's not stable.
What was your Vcore at 5GHz?
Depends how much voltage your using to get 5Ghz stable ;D
What where you running at then 4.7Ghz? If so that's an strangely amazing score.
Yes... he never did clarify.. was that last Geekbench score at 5GHz? I think that's what you meant.
UltraNEO*... Yours is not on yet until you confirm your clock speed for the 20107 run... it can't be 4.7.. so I'm assuming 5GHz. Can you please confirm?
1.6v set in the bios. 1.67v was the max when running the 64bit test, using Extreme LLC.
5Ghz! He would not had got Over 20K with 4.7Ghz. All 3960X and K's run about the same ~200 points.
It's quite easy to run Geekbench @ 5Ghz on a H80 if you have a golden sample. Intelburn test will be a completely different thing getting it stable with a H80.
Lets not blow this out of per-portion. We need images of CPU-Z next to the Geekbench run, like my uploaded images. Otherwise, it didn't happen.
There is noway he can get over 20K at 4.7Ghz, no matter how good a golden sample you have.
20K Scores require at least 4.9Ghz.
Rules should be; Image of CPU-z next to your Geekbench score, and a upload link. The same as my images. Then this will take away the guessing game, and someone trying to bluff a high score with lower clocks...
I don't know what you folks do when you run your benches... i attempted to clock as high as possible (4.8) but I notice with every run, my scores were slightly different, then So with this in mind I started terminating unwanted background tasks, ending window system processes and generally killed system services until such time the OS was running at minimum spec. Then I re-ran geekbench several times, making minor adjustments to voltages with each boot... saving the results to disk and eventually it hit 20k.
TBH, I didn't expect it'll happen..
Would running a benchmark task over and over and over again, then only to submitting your highest score constitute to being a cheat??
Next time, I'll grab a bucket of ice
This is 5 runs at 4.7Ghz Lowest score was the first run, and the highest score was the third run.
I will now do another 5 tests at 4.8Ghz.
Quit your AI suite dude...
4.8Ghz again, first run was the lowest score, and forth run was the highest.
4.9Ghz. Again first run was the lowest and the highest was the third run.
UltraNEO*.. was this....
I don't want to start changing the rules so that the only valid ones include a CPU-Z report in teh same screen as the geekbench results, as then we'll have to wipe the boards and start again... one rule for all etc.. but given the data on the leaderboards, it's exceptional to get over 20K with a 4.8 overclock.
If this is indeed true, then just to satisfy the scepticism, can you run it again with a CPU-X screenshot? Experience has taught me that it should be very repeatable, as Simon has demonstrated. I can also concur that I can run Geekbench several times and receive numbers within 100 points or so each time.
My screenshots would look suspicious as hell. For one simple reason I shall show in a second.
Edit; This is the score i just produced. http://browse.geekbench.ca/geekbench2/562454
This is the screenshot to accompany. (Demonstration only, Pook, don't worry about recording it.)
I get a max of ~250 points when running multiple tests.
That's at 4.7
@Simon. Why do I get CPU overvolt error at the BIOS screen when I set over 1.56v? Do I have to mess with the jumper switches or something?
Separate names with a comma.