First i7 920 OC

[speedy523]

Hey everyone, i just started OC'ing my 920, and i thought itd be a good idea to post here and get some feedback from everyone whos already been at for a while

Anyways here are some screenshots of the OCs i did:

4 GHz:

4 GHz ( Lower Voltages ):

4.2 Ghz:

Can someone please explain to me the speed column on the burn test and explain to me which is the best, because at 4 ghz the speed is shown faster than the 4.2 ghz and the lower voltage 4 ghz.

 

[R1P5AW]

I never look at the time column lol. Just if I were you I'd run OCCT 1hr, Prime 95 6~8hrs and LinX 20 runs

 

[R1P5AW]

BTW what cooler are you using?

 

[Sinsear]

LinX stresses your CPU much more than Prime95. To be truly "stable" (at least in my books), you have to be able to complete 50 passes of LinX.

 

[BlackDragon24]

Everyone's definition of stable is different.

Since you have 12GB of memory you should technically be running at least 20 passes of Linx with around 10,500-11,000MB of memory selected.

 

[BillParrish]

Flops are FLoating point Operations Per Second. Computer do two kinds of "math", integer which just uses whole numbers like 1 2 3 4 .... and any remainder/fraction is discarded and floating point where remainder/decimal places are maintained. Floating point is much more difficult/hard for the processor to do and the number of floating point operations a CPU can perform in a second is used as a performance indicator. More on the net, this is a very basic explanation.

http://en.wikipedia.org/wiki/FLOPS

What your are seeing is that when one execution run of Linpack calculates whatever it calculates to give the result in the third column as the speed of the run decreases the GFLops increase (faster run time means more math operations per unit time)

the slight variations in run time is likely due to cache, windows disk access, realtemp running at same time, /shrug.

The only thing I can think of to account for the lower speeds in the 4 GHz lower voltage case is that you where getting data errors that required retrys. Almost every "handoff" of data, say from the core to the memory controller, memory controller to memory, is parity/error checked and retransmitted if there is an error. It is just a guess, I have not looked into it deeply and someone certainly might have a more correct explanation.

 

[BlackDragon24]

1.54 VTT is also probably waaaaaaaaaaaaay more than you need at 21 x 200. Too much VTT can cause just as many problems as too little.

i7's are a tricky beast. I find that even when I am "stable," I can still tweak voltages to get a little more "stable." Finding the right VTT/DRAM/Vcore combo can often lead to higher benchmarking scores, especially in memory related applications like superpi. It can also lead to more consistent gflop readings in IBT and Linx.

And its all related to what Bill just posted....and how the cpu handles error checks.

 

[speedy523]

thanx for the replies everyone

and im using a TRUE Rev C in push/pull with 2 Scythe Ultra Kazes at a little over 50% power, but i might swap them out for Slipstreams and see the difference.

also, last night i tried running LinX to make sure everythings stable on the lower voltage 4 ghz clock, but LinX just destroyed the processor and proved it wasnt stable

so im back to readjusting the voltages, but one thing i wasnt sure about is the vdroop. Like i read a little bit about its purpose to make sure the processor never exceeds its maximum specified voltage, but whenever the processor goes under load its drops the voltage a little too much, and results in my bsods, so i was wondering is it ok to disable it? ive read a lot of places arguing one side or the other, but i dont want to have a higher idle voltage in order for the cpu to get enough power under load

EDIT: i running LinX at 4.2 ghz right now with new settings, and its not having any problems right now except that the temps for Cores 2,3,4 are low 80s ( like 80-83 ), while core 1 is at 87. im also running with HT on, so are these temps normal for these settings, or too high?