msi p45 neo3 and q8400 overclock

[eliron82]

hey guys

specs:

msi neo3 p45

q8400 3.5ghz 1.355v (437x8)

8gb ddr2 800 kingston value ram [email protected] dram

hd 6950 unlocked (950/1420) 1.18v

seasonic x-660

bios setting: mch - 1.304v

vtt - auto, ich- auto

doesn't have all other refrence gtl voltage, and when i raised mch and vtt to the max safe of 1.4v and vcore to 1.43v

my pc stuck after 1-2 minute prime95.. i can feel that i can get it much higher..

also want to mentioned that try to play with the clock skew by putting cpu on 200ps and mch 100ps but it make the pc more unstable!

what am i missing? thanks!

[wevsspot]

Eliron - messing with GTL reference voltages and clock skews is an exercise in diminishing returns unless you're shooting for bragging rights or world records. If you can get your q8400 overclocked to 3.3Ghz or a little higher at reasonable cpu voltages then you are setting pretty much where most other overclockers are with that particular chip.

 

I think since you're at this point the best thing to do is clear your cmos, reset BIOS defaults and start with a clean slate. After the BIOS clear begin with these settings;

 

cpu voltage - 1.35v

cpu multiplier - 7

fsb - 400

set memory divider to yield memory frequency equal to or less than your memory's rated speed

begin pushing the FSB until she won't boot or you fail stress testing

at that point bump NB(MCH) voltage up .05v and retest

rinse / repeat until you've got your MCH voltage anywhere between 1.3 and 1.4v (the northbridge gets hot - so I never went above 1.30 personally without active NB cooling)

if needed add a bump to VTT voltage - .05v should be fine

 

Repeat this test until you've hit a maximum FSB that is achievable with reasonable voltages - now you know the max FSB your board should be able to hit

 

Now go back and set your FSB to 400 and change the CPU ratio to 8 and start pushing the FSB again. You won't get as high on the FSB this time because the cpu will be operating at a higher frequency than before. But you will know when you hit the true cpu wall because you've already vetted your maximum FSB in the first set of testing. One of the most important parts of overclocking on this platform is watching your strap and memory dividers. They must be set to keep your memory within a reasonable range of it's stock operating frequency if you are going for max cpu clocks.

 

If you've got good cooling I would be comfortable with pushing the cpu vcore as high as 1.40v, but I'd go easy on the other voltages and settle for whatever you can get stable. In my book a 3.2 to 3.4Ghz overclock on the q8400 for 24/7 use would be dandy as long as voltages and cpu temps were reasonable.

[battery]

another thing to consider is playing with TRD or PL (they're the same thing.) this setting greatly effects the load on your northbridge and can drastically change the effects a high fsb will have on your system. with q8400 you're limited to lower multiplier so in order to get the high clocks you need the fsb to overclock. raising your fsb will cause your TRD to raise itself in order to maintain stability. this usually offsets the memory performance gained by the increase in fsb and can even produce lower system speeds than was seen at the lower fsb with the lower TRD settings. sometimes its worth the sacrifice to your memory bandwidth to have the extra clock speed, sometimes not. i personally run my x38 chipset at 400mhz with a TRD setting of 6. this yielded the best overall system performance and i still landed at 3.8ghz with an e7500.

Edited August 9, 2012 by battery

[eliron82]

Eliron - messing with GTL reference voltages and clock skews is an exercise in diminishing returns unless you're shooting for bragging rights or world records. If you can get your q8400 overclocked to 3.3Ghz or a little higher at reasonable cpu voltages then you are setting pretty much where most other overclockers are with that particular chip.

 

I think since you're at this point the best thing to do is clear your cmos, reset BIOS defaults and start with a clean slate. After the BIOS clear begin with these settings;

 

cpu voltage - 1.35v

cpu multiplier - 7

fsb - 400

set memory strap to yield memory frequency equal to or less than your memory's rated speed

begin pushing the FSB until she won't boot or you fail stress testing

at that point bump NB(MCH) voltage up .05v and retest

rinse / repeat until you've got your MCH voltage anywhere between 1.3 and 1.4v (the northbridge gets hot - so I never went above 1.30 personally without active NB cooling)

if needed add a bump to VTT voltage - .05v should be fine

 

Repeat this test until you've hit a maximum FSB that is achievable with reasonable voltages - now you know the max FSB your board should be able to hit

 

Now go back and set your FSB to 400 and change the CPU ratio to 8 and start pushing the FSB again. You won't get as high on the FSB this time because the cpu will be operating at a higher frequency than before. But you will know when you hit the true cpu wall because you've already vetted your maximum FSB in the first set of testing. One of the most important parts of overclocking on this platform is watching your strap and memory dividers. They must be set to keep your memory within a reasonable range of it's stock operating frequency if you are going for max cpu clocks.

 

If you've got good cooling I would be comfortable with pushing the cpu vcore as high as 1.40v, but I'd go easy on the other voltages and settle for whatever you can get stable. In my book a 3.2 to 3.4Ghz overclock on the q8400 for 24/7 use would be dandy as long as voltages and cpu temps were reasonable.

thanks, i try this on this weekend.. i'm using air cooling corsair A70 with 1.41v max temp 70c which is pretty high i know.. a few days ago i test my memory with memtest86+ on these setting:

fsbx460 multiplier x7

cpu vcore - 1.36

vtt - 1.35v - max safe for c2q i'v heard.

mch - 1.3v

ich - auto

dram - 1.896v

the test run complete with no errors..i think that 465 is the max that my mobo can get post..

now when i used prime95+ on this 460 fsb with above setting it failed..so what does that tells me? it's not the memory of course and not cpu ( decrease the multiplier to 7 and raise voltage to 1.36v), so according to what you said it's must be the mch and vtt voltage setting, or could be the clock skew..

but hear this.. last night when i change the fsb to 455 (multiplier x8) and only my clockgen setting: via

cpu clock drive - auto

pci express - auto

cpu clock skew - 100ps - was on auto besfore.

mch clock skew - auto

it seems to be stable (didn't ran test - was very late) played gta iv for half our and did some bench of metro 2033 and crysis , but of course i need to do at least 1 hour of occt or prime95+ 1h+ and see..i'l post when i do the tests..thanks a lot..

[eliron82]

another thing to consider is playing with TRD or PL (they're the same thing.) this setting greatly effects the load on your northbridge and can drastically change the effects a high fsb will have on your system. with q8400 you're limited to lower multiplier so in order to get the high clocks you need the fsb to overclock. raising your fsb will cause your TRD to raise itself in order to maintain stability. this usually offsets the memory performance gained by the increase in fsb and can even produce lower system speeds than was seen at the lower fsb with the lower TRD settings. sometimes its worth the sacrifice to your memory bandwidth to have the extra clock speed, sometimes not. i personally run my x38 chipset at 400mhz with a TRD setting of 6. this yielded the best overall system performance and i still landed at 3.8ghz with an e7500.

thanks for reply

I tried to play with it, but don't forget i have very budget memory..4 sticks of kingston value ram 800 MHz cl6 which very limiting..

i think that i tried setting it to 7 but it won't post so i left that on auto..maybe 8 or 9 will post..

[wevsspot]

Yep, 7 is very aggressive. 8 or 9 is the sweet spot if you are really pushing everything else pretty hard. Battery makes a great point on that setting's impact on memory performance, but sometimes in order to achieve it you have to underclock your ram and shoot for tighter primary timings in conjunction with tightening the Trd (PL).

[battery]

TRD (PL) overclocks your northbridge and actually has no effect on your memory. what happens is when you run to low of a TRD you have to run more chipset voltage to compensate.

 

i would read This and then make your own educated decision on what works for you.

[wevsspot]

While I'm not a huge fan of Anandtech, I'll have to give credit where credit is due. That is a very good write up on the subject.

[eliron82]

ok guys, finally got sometime to test, and here is the process and the result so far:

I started to raise fsb to 450 but had some problem to pass post, so I raised vtt fsb to 1.4v (safe according to bitech or tom's hardware review if I remember right), but it didn't resolve the post problem. then I noticed that there is a ratio of 1:2 between mch voltage to fsb voltage, so I kept this ratio by setting fsb voltage to 1.4v (default 1.2v) and mch 1.208v (default 1.1v) and it did resolve the post problem. next I lowered the multiplier to 7 and leave cpu voltage on auto and ran occt linpack but failed after ~30 min. here I started to get frustrated but I remember that my ram is a value ram (meaning cheap and very limited ram especially because it's high default latency) and I started to raise it's voltage. first, to 1.9v - didn't help, 2.0v - again, eventually on 2.1 and 455 fsb it pass 2 hours test of occt linpack with only 1.29v on cpu core (impressive), btw ich voltage left on auto. anything above 455 (with multiplier on 7) give me bsod randomly.

after that I raised the multiplier back to 8 so the cpu frequency is 3640 MHz which is nice and again it pass occt linpack for 2 hours so it stable for now. about overclock ram: lower timing is out of the question, maybe lower trd timing to 11 (on 10 I got bsod) and tfrc timing (doesn't know if I gain anything from doing that). back to cpu and mch skew - still don't understand that and don't know how to benefit from playing with that (like to learn and try suggestion). another important thing, anybody that know this motherboard knows that it does not give gtl ref voltage tweak options, so don't bother to give me instructions how to change that..

last, I would like to know if based on the result above I reached the max oc on the ram or there is a way I could push it a little more to 3.7~3.8ghz because from what I saw it the test that motherboard temp stayed below 40c(is that represent ram temp to?) and cpu below 70c on ambient temp of 20c, so I think I get achieve these desired frequencies...

anyway, thanks a lot for the help guys, couldn't done it without your help and hope that it will help others who got stuck like I did.. replies are welcome...

[battery]

set your MCH voltage to 1.55 - 1.65v

[eliron82]

set your MCH voltage to 1.55 - 1.65v

ok..but is that safe? so far the highest I set was 1.4v but it didn't had any effect..what do I need to monitor? mobo temp?

[battery]

does your motherboard have active cooling on the chipset? if so then you should be safe. i run 1.65v 24/7