Compare OCs.

[drdallon]

I just wanted to see what other people using the same processor, and similar setups are getting.

Im using a G9550 @ 3.4GHz, been running Core Temp/Prime95 for a bit now and im peaking at 48c on core 0 and core 2, core 1 and 3 are both staying under 45c.

Also a quick question, what is a reasonable temp to be running at? I want to see if i can take it further but i also dont want to melt my chip

Thanks for the imput//help ive gotten from you guys!(Special thanks to zombine210 who helped me with the OC).

PS: How long did you all leave prime95(or w.e program) on to test?

*Edit* Forgot to say this but my rig is in my SIG! *Edit*

 

[Jet]

those are really good temperatures! I would try to keep it under 65C. Voltage is what really kills a processor, though--what are you running it at?

 

[drdallon]

So funny thing happened, i ran prime95 for about 2 hours no problems(i know not long enough but i got ancey), and then played some CSS i got weird graphical problems and it gave me a blue screan. Thing is my temps are fine and voltages stayed the same(from what i can tell), i ran prime95 again and it instantly failed the torture test. I took the OC off and same thing CSS gets weird and insta-fail the torture test failed. Any ideas of whats up?

 

[drdallon]

Good news, got it to not fail on standard settings. Now just to OC it with out failing again ill keep you all updated.

PS:Any one have any ideas why it only happened after i launched CSS?

 

[drdallon]

i have managed to get it stable at 3.45, im running an over night stab. test. Ill post tomorrow with my results.

 

[linkin]

Graphical corruption generally points towards your GPU - have you overclocked that? overclocking your CPU may also be raising the NB clock, which will raise the PCI-E clock. too high on either will cause instability/corruption. If possible, it is recommended to overclocked in "unlinked" mode if your board supports it. basically it makes it so that when you overclock, it won't overclock all your other hardware, only the CPU.

 

[drdallon]

Graphical corruption generally points towards your GPU - have you overclocked that? overclocking your CPU may also be raising the NB clock, which will raise the PCI-E clock. too high on either will cause instability/corruption. If possible, it is recommended to overclocked in "unlinked" mode if your board supports it. basically it makes it so that when you overclock, it won't overclock all your other hardware, only the CPU.

I have not OCed my GPU in a long time, and it is unlinked. Thank you for the information though.

Also i ran it in prime95 @4.45GHz last night and it failed on torture test 4, better then what it was, but i dont want any fails. Any one else have any ideas? My temps are all reasonable, never went above 49c. I UoCed it too 4.40GHz so well see how well it does now.

 

[bigrich0086]

Keep in mind that prime is testing memory as well. you may not have enough Voltage to the dram. Set your timings and voltages manually on ram. the q9550 usually runs well 3.4gh o na 1.30 vcore. and a 1.31-1.32 nb

 

[drdallon]

1.31-1.32 nb

Is that the voltages for the RAM?

PS:Sorry for my ignorance, first OC. But hey yah gotta start somewhere right?

 

[bigrich0086]

No that 1.31-1.32 is for the North Bridge voltage.

The timings and voltage the ram is rated at is on the RAM its self printed on sticker.

Most q9450 and 9550 are able to achieve 3.4-3.6ghz OC with a 1.31 Vcore wit ha 1.31-1.32 NB.

BUt keep in Mind not ALL cpus are made the same where some will need more voltage and some will use less.

 

[drdallon]

Okay here is what my clocks are set too.

Q9550 @ 3.40GHz (400x8.5)
Memclock 800MHz(It is suppose to be 1066, but i wanted to get my OCdown first)
FSBclock 1600MHz

My voltages screen in the bios looks like this


Parameters-Setting-Current Vaulue

Core - Auto - 1.35000V
CPU FSB - Auto - 1.30V
Memory - Auto - 1.900V
nForce SPP - Auto - 1.40V
NF200 voltage level - Auto - 1.20V
-------------------------------------------
GTLVREF Lane 0 - Auto - +00mv
GTLVREF Lane 1 - Auto - +65mv
GTLVREF Lane 2 - Auto - +00mv
GTLVREF Lane 3 - Auto - +00mv

What should i chage(hope this helped).

 

[Jet]

Try bumping the memory voltage up to 2.1V--from what I see those RAM modules are rated for 2.1V-2.2V--my OCZ DDR2 runs best at high voltages.

 

[drdallon]

Also my friend said that i had vDrop and to disable it, i do not see any option in the Bios that lets me disable this.

he said i had vDrop because my clockspeeds change when there under/not under load.

 

[drdallon]

Try bumping the memory voltage up to 2.1V--from what I see those RAM modules are rated for 2.1V-2.2V--my OCZ DDR2 runs best at high voltages.

Thanks ill try that now, ill post back in a few mins what happens. (Im using CSS as my crash test, along with memtest).

 

[Jet]

Also my friend said that i had vDrop and to disable it, i do not see any option in the Bios that lets me disable this.

he said i had vDrop because my clockspeeds change when there under/not under load.

First off, take all voltages off auto. You don't want the motherboard messing with them, because then if you make changes there are multiple variables changing and you never know what helps.

vDrop/vDroop is when your CPU voltage is fine at idle, but then drops when you put load on the CPU. To test this:

1. Set the CPU voltage in the BIOS manually
2. Boot the computer, and run CPU-z. Record the idle CPU voltage
3. Run something like Prime95 (make sure all cores are at 100%), and record the voltage that CPU-z reports.

 

[drdallon]

This is what my voltages look like now

Core - 1.35000V - 1.35000V
CPU FSB - 1.30V - 1.30V
Memory - 2.100V - 2.100V
nForce SPP - 1.40V - 1.40V
NF200 voltage level - 1.20V - 1.20V
-------------------------------------------
GTLVREF Lane 0 - Auto - +00mv
GTLVREF Lane 1 - Auto - +65mv
GTLVREF Lane 2 - Auto - +00mv
GTLVREF Lane 3 - Auto - +00mv

Should i have changed the bottom half too?

 

[drdallon]

So left is under stress right is idle

Also when you use prime95 do you run the blend test or one test at a time?

Thanks for the help so far!

 

[drdallon]

after changing the voltages CSS hasnt crashed after 20mins.

It was crashing between 5-10mins. Ill run Prime95 and see if that fixed it completely.

*edit* core 0 failed test 1 after voltage change *edit*

Also just curious if i wanted to take it higher then 3.6 i would have to change the voltages? and what too?

 

[Jet]

So left is under stress right is idle

Also when you use prime95 do you run the blend test or one test at a time?

Thanks for the help so far!

Those aren't bad voltages--it looks like from what you set in the BIOS, it's .3 less at idle and right around .5 less at load--try raising the voltages some more. Intel rates the Q9550 at a max of 1.3625V, which will be a very very safe max Vcore. So, with the .3 drop at idle and .5 drop at load, you're looking at say 1.4V safe max in BIOS--though you still could go higher.

Your temps aren't an issue, so I don't see any problem with you going up to 1.4V, though it's at your own risk .

With Prime95, run the one that stresses the processor most (I think it's small FFTs or something of that nature).

--------------

As far as other voltages, you'll eventually have to raise the CPU FSB voltage (which I believe would be the Northbridge voltage), as you're getting pretty high FSB speeds. However, never raise voltages until you need to, and write down what works, etc on paper so you don't have to do tests over again.

 

[bigrich0086]

you need to disable Speed step in the bios. Find it and diasble. its like amds cool n quiet. When cpu is not in use it will down clock the cpu to save power. Not a good thing to have on when OC.

i run blend last. and run small fft first so it tests cpu for max temp and pushes cpu harder.