Ever get the feeling that overclocking is going too smoothly?

w1llyg

w1llyg

w1llyg

Associate
Joined
23 Dec 2009
Posts
549
*EDIT* New problem
Hi,

Well its my first time overclocking properly, i did it once b4 and my computer restarted and it worried me so i decided to stay with the stock speeds.

Now i have read more guides and researched more, i am in the process of overclocking and currently have my phenom ii x4 965 BE C2 running at 3.9Ghz @1.45v and its around 48degrees on prime95. Im going to aim for 4GHz but i dont fancy having a vcore of 1.5 :/

When i do prime95 stability test (Small FFT) It runs smoothly without crashing. But if i use the Blend test (which includes RAM in the test) one or two threads stop working. The RAM is not overclocked and its running at 1066MHz with timings of 7-7-7-16-27. Iv upped the DDR3 voltage to 1.65v and prime95 still stops.

I dont know what to do :/
 
Last edited:
Blend test will point at instability at the RAM. Maybe loosen up the timings? How long do you run the Prime test for?
 
Well on the Small FFT 1-2 hours then if no fail ill bump up the multiplier etc.
But on the Blend one of the threads will stop working within 5 minutes.
I've lowered the multiplier so im running at 3.6ghz and ill try loosening the timings now, but they are on stock timings and @ 1066mhz which the Bio has lowered it to :/
i also dont want my lower my timings too much or else ill bottleneck i guess?
 
Blend also tests your northbridge, but generally RAM is the first thing to test. What I'd suggest is this:
- drop your multiplier right down so you're at stock speeds and timings, and then run blend.
- if problems persist you've ruled out your cpu for sure, and most likely your northbridge (notice how your NB is running at 2GHz, this is due to how the multipliers interact. Big Wayne explained it in one of my threads, so perhaps email him and and he could explain it better than me).
- try removing one stick of ram and try again, if problems persist repeat with the other dimm/dimms. This is to rule out a bad stick of RAM. You could also try running memtest to see if any errors crop up.

Let us know how you get on. Good luck bro

EDIT: found his post, not sure if it only applies to LGA775, but if your RAM testing doesn't pan out, perhaps it's another avenue you could investigate

Big.Wayne said:
The 4.5GHz Fritz results are 9.8% faster which doesn't help corroborate your initial claim of performance decrease? . . at least it seems your CPU crunches faster at the higher frequency which is your main concern . . . I'm not that familiar with Vantage CPU benchmark and therefore don't know if 500 points is a lot *or* a little! :D


A quick brute force explanation would be that lowering the LGA775 CPU multi from its default overclocks what you think is the FSB, although its actually called the NBCC (Northbridge Core Clock) which is in fact a processor called the Northbridge and acts in a similar way to how you understand a regular processor i.e higher frequency = faster . . . and the faster it gets the more volts (vNB) it needs and the hotter it runs . . .

p45sm.jpg

"A Veteran P45 Express Northbridge Chip with heatsink removed . . . . PIII Coppermine anyone?"

Most LGA775 clockers have a good enough understanding of what FSB does but few of them have the knowledge that the FSB you see in CPU-z really only relates to working out the Processor MHz and memory speed through the use of memory multipliers and doesn't really relate to the true system "Northbridge" frequency and the way that effects system bandwidth! :)

Using your E8400 as an example, lets say we set it at 9x400=3600MHz and I asked what is the FSB? . . . most people would say 400MHz which would be correct:
9400.gif



now lets say we lowered the CPU multi from [x9] to [x6] and again I asked what is the FSB? . . . most people would say 400MHz which would be incorrect? . . . . the correct answer would be 600MHz :eek:

6400.gif


In the two examples given above we have *increased* the Bus Bandwidth from 12.5GB/s to 18.75GB/s which is a mighty 50% gain just by lowering the CPU multi! :eek:

To work the NBCC speed out you take the native CPU multi and divide it by the set CPU multi, you then take the resulting figure and multiply it by the set FSB to work out what is the actual NBCC frequency . . . using the two E8400 examples above:

  • 9/9=1
  • 1x400=400 MHz-NBCC

  • 9/6=1.5
  • 1.5x400=600 MHz-NBCC
This system ignores the half CPU multipliers and always rounds down, so [x8.5] is calculated as [x8] and the only exception to this rule is the eXtreme-Edition Processors that count as native multi no matter what you set.

NBCC MHz determines the speed of the data transfers between the CPU and the System Memory so its worth understanding if you want maximum performance, most advanced LGA775 tweakers will also *not* run their memory [1:1] sync and instead use a Memory-Multiplier along with a more conservative set FSB and a knowledge of NBCC and tRD (aka Performance Level) to speed up the data transfers!

To sum up, on the LGA775 platform factor in a lower CPU multi combined with an upward memory divider for best performance! :cool:
Click to expand...
 
Last edited:
Thanks so much for a helpful reply :D

I put everything back to stock settings etc, and still failed the Blend test within minutes.

Took out one RAM, (noticed a massive performance drop :O ) and it failed again in minutes. Trying other RAM now and its been running for 25mins now, which isnt along time but b4 it was failing within 5mins so its an improvement.

Also in the stress test now im using 1x 2GB RAM sometimes one or two cores go off 100% to about 7% and then go back up to 100% load is this normal? I have prime95 on high priority on windows task manager too.

One last thing, if it is dodge RAM whats the next step? Buy a new stick? Also does this mean i cant return it as faulty as ive overclocked?
 
The RAM is not overclocked and its running at 1066MHz with timings of 7-7-7-16-27. Iv upped the DDR3 voltage to 1.65v and prime95 still stops.
Click to expand...

Is 1.65v the correct voltage for the ram?

Also you could remove your overclock and test the memory with memtest86 if you believe the problem lies with one of the sticks.
 
I tested the RAM over night, and passed all the tests over the 10hours.
I tested the other RAM this morning and failed after 4minutes :/
So do i loosen the timings or do i have a 'duff' RAM? :'(

Btw bobby, 1.65volts is the correct voltage.
 
I think i can confirm i have a duff RAM module -.- Only brought jan :(
They dont even sell my RAM any more here :/
Looks like im going to hav to put my speakers on hold and buy some more RAM
Does anyone know about the return policy @ OC for RAM?
 
w1llyg said:
I think i can confirm i have a duff RAM module -.- Only brought jan :(
They dont even sell my RAM any more here :/
Looks like im going to hav to put my speakers on hold and buy some more RAM
Does anyone know about the return policy @ OC for RAM?
Click to expand...

Is this at stock speeds? If it fails at stock speeds then it's faulty, so you shouldn't have problems returning it.
 
You must log in or register to reply here.
Back
Top Bottom