• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Can't pass prime95 1440k

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

ElDavo

New Member
Joined
Mar 5, 2018
EDIT: Solved: Tertiary timings not set to auto were the problem

Hi!
I'm trying to overclock my PC:
Mobo : P8P67 Deluxe and i7-2600k @ 3.40ghz.
Overclocked BCLK (106.4) and multiplier (42) which gives 4467mhz. LLC is set to high as with medium I have a 0.05V spike when I stop IBT and very high goes beyond values.
I'm using offset voltage, that is +0.015. Under load, VCORE is about 1.31V.
Now, memtest and even an hour of IBT passes without problems, but prime95 with in-place 1440K FFTs almost immediately fails with rounding error (no BSOD). With other FFTs in that range, it may fail in 30 minutes or ine hour.
The same prime95 with small FFTs is fine.
At stock settings, everything is fine. Temperatures are fine too (below 80C).
I've tried setting an offset voltage up to +0.060 (so vcore becomes 1.35V under load), but it didn't change anything (apart from more heat of course).
Now the question (obviously is): Why is it failing?
I overclocked my RAM too, but p95 fails with fft in-place, which don't stress the memory much. Larger FFTs, memtest and IBT don't fail, so I don't think this is related to RAM or IMC.
But this isn't related to the CPU either because upping the voltage doesn't change anything.
Any thoughts? Should I just ignore it and live with that?
 
Last edited:
Why are you messing with the bclk instead of just overclocking via the multiplier. Changing the bclk is effecting other system components like the cache that can be causing problems.
 
Looks like you are indeed messing with the BCLK. A no, no with Intel chips for several generations now.
 
Wrong. BCLK is fine at 106.4 and doesn't cause problems (if I go higher than that it does). I can't set a higher multiplier because of temps, so I don't have the problem of "higher BCLK so less multiplier".
However, I've set the tertiary RAM timings to auto and now it's stable! So today we learned that even in-place FFTs stress RAM :)
So this thread should be moved to the memory section instead :)
Thanks to trents and lochekey! :)
 
SB had a bit of BCLK headroom to play with... not much though. Anything over 105 is really good.

That said, 4.5Ghz via 100x45 or 106x42 should yield the same temps...its the same clock.

I also have no idea what in plcae ffts are and that particular length. Typically users use small fft for CPU and blend for Memory and CPU. Large is just MEH to use.

EDIT: Also, BCLK can be adjusted, and quite freely on some boards. Either SB or IB was the last to lock that down. Many boards have BCLK generators(chips) on them as well. I mean, The ASUS Z370 apex I think has a PROFILE for 370 BCLK... ;)
 
Last edited:
Yes, there is some wiggle room with the bclk but I agree with ED. 6 mhz over stock is more than typical. I usually find that I can vary it 2-3 mhz without any problem but very unstable on many or most boards I have used beyond that.

I don't know if it works the same way with Intel but it was common knowledge on the AMD side (pre FX) that you often could get by with a little lower voltage at a given overclock level by using a combo of the FSB and the multiplier rather than the multiplier alone. That might help with temps.
 
If only 2-3 mhz over stock is acceptable, is it really necessary to mess with it at all?(considering multiplayer is unlocked)

Eit: NVM, trents second paragraph I read it xplains
 
Last edited:
I don't know if it works the same way with Intel but it was common knowledge on the AMD side (pre FX) that you often could get by with a little lower voltage at a given overclock level by using a combo of the FSB and the multiplier rather than the multiplier alone. That might help with temps.
I dont see why that would be. X.xGHz is X.xGHz. if anything raising up HT or FSB/BCLK would use more power (negligible) considering it is going faster than spec..

If only 2-3 mhz over stock is acceptable, is it really necessary to mess with it at all?(considering multiplayer is unlocked)

Eit: NVM, trents second paragraph I read it xplains
Nope. Its a 6 of one, half dozen of the other situation. I guess trents is saying it happened on pre FX amd cpus though ( we talking like s754/s939 days???? Still fx there chips there too though - what is a pre FX chip?)?
 
Last edited:
My original point of mentioning the Bclk change was more to do with what else is being effected than just the the core speed.

If I remember correctly the cache, imc, pcie, etc. Is all tied to the bclk so having the bclk pushed this far is probably either requiring unnecessary voltage increases or is adding instability that can be easily avoided.

As ED mentioned the same core speed can be obtained with a lower bclk so I see no need to add the instability of bclk overclocking into this equation.
 
My Thanks! is emptied for today, so I'm doing dedicated thanks.
It's interesting how it got inverted. Back in generationX58, bclk over clocking was the only way to overclock unless you had ExpensiveEdition cpu(965)
 
Pre Fx chip like before Bulldozer and Vishera. As in Athlon II and Thuban and Phenom.

Don't know why using combo of FSB and multiplier was easier on voltage but it was widely reported. I don't know that I ever heard an explanation why.
 
I dont see why that would be. X.xGHz is X.xGHz. if anything raising up HT or FSB/BCLK would use more power (negligible) considering it is going faster than spec..

Nope. Its a 6 of one, half dozen of the other situation. I guess trents is saying it happened on pre FX amd cpus though ( we talking like s754/s939 days???? Still fx there chips there too though - what is a pre FX chip?)?


Only way I could OC my locked Phenom X4. Had to bump (big bump, 243 from stock 200) my HT and lower my RAM speed because Hyundai Electronics RAM overclocks like Ethel Merman looked good in a bikini. The joys of a non BE K8. LOL

Don't know why using combo of FSB and multiplier was easier on voltage but it was widely reported. I don't know that I ever heard an explanation why.

Funny you should mention that. Voltage bumps never helped my OC on that chip. Did it all (3038 MHz) on stock voltage, including my daily driver OC @2750 MHz.
 
Back