Need some help with a mild overclock of a E8400

[HTC]

Right now, i have my E8400 @ 8*375=3000 MHz, using 1.1250 vCore, 2.02 vDIMM (8GB) and 1.38 vNB. It's OCCT stable (Linpack 1 hour).

If i change the CPU multi to 9, do i have to increase the vCore only or do i need to increase something else as well?

My goal is 9*385 or 390, but i'll happily start with 9*375, if i manage to get it stable ... which i haven't, yet

If this damned board had a better FSB capability, i would go for 8*460 or 470 but i can't even get a FSB of 415 ...

 

[Charper2013]

Only increase the Vcore if its not stable or doesnt boot into windows. Maybe even give some NB volts if unstable with a decent Vcore.

 

[HTC]

Only increase the Vcore if its not stable or doesnt boot into windows. Maybe even give some NB volts if unstable with a decent Vcore.

I'll have to increase the vCore, since i'm increasing the CPU's speed by 375 MHz. I was thinking about 1.2 vCore for this.

I've tried to have 9*385 with 1.25 vCore, 2.08 vDIMM and 1.42 vNB but it wasn't IntelBurn test stable

Since i'm maintaining the FSB, will the only change required be the vCore or will i have to change anything else as well? That's what i'm asking.

 

[Scheich]

The 8400 usually doesnt need much if any voltage increase upto 3200/3300 mhz.
Most likely you will need to increase NB to +0.3 v to get the ram stable. My board
barely supports 6gb@+0.3 NB, maybe 8gb is only meant for higher quality ones, dunno.
Would suggest to run the memory with standard speeds and no voltage increase first,
for stability tests.

 

[PlanetCyborg]

why do you use 8???why dont you use 9*333=3000 Any way my g31 board runs the E8400 @9*399 with stock volts no problem USE multi of 9 not 8!

 

[HTC]

why do you use 8???why dont you use 9*333=3000 Any way my g31 board runs the E8400 @9*399 with stock volts no problem USE multi of 9 not 8!

Because of the higher FSB that makes full use of my G-Skill PI Black 7200 RAM.

By stock volts, you mean "Auto" volts, right? Try using OCCT and check how much vCore you have while running the benchmark: you should have much more vCore then i do.


Anyways, i followed Scheich's advice and tested @ 9*375 with 1.2 vCore, same vDIMM and 1.42 vNB, and it passed OCCT's Linpack test.

 

[PlanetCyborg]

well mister smart pants it is on STOCK 1.285 volts!!i tried my E8400 on my friends p45 and got up to 3.8 ghz with stock volts(some have gotten do 3.9 with stock volts)

 

[HTC]

Something's wrong ...

Was encoding and it crashed the PC ...


It seems i have some more testing to do ...

 

[mlee49]

Does your board support half steps on the multi? Maybe try working your way up from 8,5 and increase your FSB in 10's.
From my experience I could do 375 FSB but at 1.4Vcore and my ram could not clock over 1200 even at 2.4V.

 

[HTC]

well mister smart pants it is on STOCK 1.285 volts!!i tried my E8400 on my friends p45 and got up to 3.8 ghz with stock volts(some have gotten do 3.9 with stock volts)

You obviously missed the point

When you have your vCore on auto, it's not a fixed value.

On my previous motherboard, vNB "stock volts" were @ 1.63 for a E6850 @ 3536 MHz with 2*2 Corsair CL4 800 RAM but, it ran just fine with 1.39 vNB as well.

What i'm trying to say is that, just because it's "Auto" doesn't mean it's 100% necessary to be @ that value. Have you even tried to apply less voltage and preformed a stress test to see if it would work flawlessly @ that voltage?

Does your board support half steps on the multi? Maybe try working your way up from 8,5 and increase your FSB in 10's.
From my experience I could do 375 FSB but at 1.4Vcore and my ram could not clock over 1200 even at 2.4V.

Yes, but the problem's in the board itself: it has a very low FSB wall

I did 9*400 but it means i would either have the RAM @ 800, which is too low, or @ 960, which is too high

 

[mlee49]

Too bad, most Asus' have good OC'ing potential.

 

[Kenshai]

You obviously missed the point

When you have your vCore on auto, it's not a fixed value.

On my previous motherboard, vNB "stock volts" were @ 1.63 for a E6850 @ 3536 MHz with 2*2 Corsair CL4 800 RAM but, it ran just fine with 1.39 vNB as well.

What i'm trying to say is that, just because it's "Auto" doesn't mean it's 100% necessary to be @ that value. Have you even tried to apply less voltage and preformed a stress test to see if it would work flawlessly @ that voltage?



Yes, but the problem's in the board itself: it has a very low FSB wall

I did 9*400 but it means i would either have the RAM @ 800, which is too low, or @ 960, which is too high

Adjust the divider in which the ram uses, I'm not familiar with asus boards, I'm sure someone can tell you how to do it.

 

[HTC]

Adjust the divider in which the ram uses, I'm not familiar with asus boards, I'm sure someone can tell you how to do it.

Those are the lowest possible choices: 1:1 and 5/6. The other choices are even higher.

 

[Kenshai]

Those are the lowest possible choices: 1:1 and 5/6. The other choices are even higher.

@400 fsb there isn't an option in between 800 and 1066? I'm pretty sure my gigabyte board didn't have such a big difference.

 

[HTC]

@400 fsb there isn't an option in between 800 and 1066? I'm pretty sure my gigabyte board didn't have such a big difference.

There are 2, actually: 960 and 1000.

Keep in mind that i'm using 4*2 GB sticks, and that's why i can't push the RAM as high as i would like to, and yes: i do need all that RAM, since i've seen my encoding prog use just over 6500 MB of RAM, according to Everest.

 

[Kenshai]

Will all 8 gigs not run at 960 if you bump the northbridge and the ram voltage up? I have 4 x1 a-data sticks @ 950 just fine with just a slight bump in voltages.

 

[HTC]

Will all 8 gigs not run at 960 if you bump the northbridge and the ram voltage up? I have 4 x1 a-data sticks @ 950 just fine with just a slight bump in voltages.

4*1 GB are much easier to OC then 4*2 GB.

I had 4*1 DHX Corsair CL4 @ 800 and i could OC them much more then 4*2 DHX Corsair CL4 also @ 800: the difference was from 940 (4*1) to 860 (4*2).

 

[PlanetCyborg]

I DONT HAVE MY SETTINGS ON AUTO!!! and what is wrong with 960mhz? my generic ram 800mhz is running at 960mhz to AND with 1.8 volts I HAVE A LOW VOLT RIG

 

[Kenshai]

4*1 GB are much easier to OC then 4*2 GB.

I had 4*1 DHX Corsair CL4 @ 800 and i could OC them much more then 4*2 DHX Corsair CL4 also @ 800: the difference was from 940 (4*1) to 860 (4*2).

Have you tried anything higher than 2.0 volts? I had no difference in OC, it's very possible one of your sticks isn't up to par with the others. Perhaps relax the timings a bit and bump them up.

 

[HTC]

Have you tried anything higher than 2.0 volts? I had no difference in OC, it's very possible one of your sticks isn't up to par with the others. Perhaps relax the timings a bit and bump them up.

Sorry for not replying sooner but i was away from the PC.

In order for these kits to run with CL4, 2.1 volts were required on vDIMM. My board @ the time was a Rampage Formula (i've since downgraded ...) and it overvolted by 0.06.

IRRC, max i put in these kits (2*2 GB) was 2.24 real (2.18 in BIOS), but that doesn't matter because i've sold those kits and am now using 4*2 GB PI Black 7200 @ 900 with 2.02 vDIMM: max i put on these PI black kits was 960, with 2.16 in BIOS, but not stable, @ all


I really made a bad choice for my board downgrade: it never occured to me this board had such a low FSB wall. My target would be 465-470 FSB with an 8 multi on the CPU but that's impossible with this board

I can still do 9*400 but that will either put the RAM @ 800 or put a BIG strain on the RAM, with 960, which is too high for my liking

 

[Kenshai]

Sorry for not replying sooner but i was away from the PC.

In order for these kits to run with CL4, 2.1 volts were required on vDIMM. My board @ the time was a Rampage Formula (i've since downgraded ...) and it overvolted by 0.06.

IRRC, max i put in these kits (2*2 GB) was 2.24 real (2.18 in BIOS), but that doesn't matter because i've sold those kits and am now using 4*2 GB PI Black 7200 @ 900 with 2.02 vDIMM: max i put on these PI black kits was 960, with 2.16 in BIOS, but not stable, @ all


I really made a bad choice for my board downgrade: it never occured to me this board had such a low FSB wall. My target would be 465-470 FSB with an 8 multi on the CPU but that's impossible with this board

I can still do 9*400 but that will either put the RAM @ 800 or put a BIG strain on the RAM, with 960, which is too high for my liking

Why is 960 too high for your liking? DDR2 runs just fine @ 2.3v and higher with most sticks. I believe P45's run different voltage northbridge wise, but stock is 1.25 right? Try it up another notch, those g-skills should do 960 no problem. Worst case scenario, relax the timings and increase the speed to 1066 or more. Really you won't notice a big difference between CL4 @ 800 and CL5 @ 1066. Only see a difference in benchmarks.

 

[HTC]

Last night, i encoded a movie using RAM @ 900 with 4-4-4-10 and @ 1000 with 5-5-5-15, just to prove you wrong ... but it didn't work simply because i have a low FSB.

Here are the pics:

And

You may be thinking that you were right but you're actually wrong: with an increase of 100 MHz on the RAM's speed (that's over 11% increase in speed), the gain was less then 2.5%. The reason it was faster with CAS 5 was because i used the same tRD for both tests (tRD was 6).

If i had the FSB @ over 450 and wanted to switch the CAS from 4 to 5, i would have to raise the tRD as well and that would negate the slim advantage, seen in the example above.

Right now i can't but, as soon as possible, i'll make another encode with this movie but using CAS 5 and a tRD of 7 and then you can see the difference.


You can do a similar test yourself: @ the same CPU speed, put your RAM with the lowest timings and tRD you can and encode any movie using any prog you have (if you have any such progs) and then put your RAM with the highest possible timings (and lowest tRD for those timings) and then check your results.

Which timings were faster?

 

[Kenshai]

Actually you proved my point exactly, you won't notice much of a difference except in benchmarks. I've run 800 @ 4-4-4-12 and 1066 @ 5-5-5-15 and it's no noticeable difference in any applications that I use or most people use. Honestly do you notice a difference?

 

[PlanetCyborg]

Last night, i encoded a movie using RAM @ 900 with 4-4-4-10 and @ 1000 with 5-5-5-15, just to prove you wrong ... but it didn't work simply because i have a low FSB.

Here are the pics:

http://img.techpowerup.org/090525/2009-05-24_194526.jpg

And

http://img.techpowerup.org/090525/2009-05-24_200551.jpg


You may be thinking that you were right but you're actually wrong: with an increase of 100 MHz on the RAM's speed (that's over 11% increase in speed), the gain was less then 2.5%. The reason it was faster with CAS 5 was because i used the same tRD for both tests (tRD was 6).

If i had the FSB @ over 450 and wanted to switch the CAS from 4 to 5, i would have to raise the tRD as well and that would negate the slim advantage, seen in the example above.

Right now i can't but, as soon as possible, i'll make another encode with this movie but using CAS 5 and a tRD of 7 and then you can see the difference.


You can do a similar test yourself: @ the same CPU speed, put your RAM with the lowest timings and tRD you can and encode any movie using any prog you have (if you have any such progs) and then put your RAM with the highest possible timings (and lowest tRD for those timings) and then check your results.

Which timings were faster?

dude now you are just being a stubborn!!run the memory at 960 and stop whining like a little girl