Q9550 oc'd: I don't think I'm happy with Load Temps

[francisA]

So I've been searching for my 24/7 setup and found a good one, 471*8.5 4GHz @ 1.296v. Am now running with the Xigmatek SD964 with the 92mm fan set to PWM mode.

Testing for stability, I ran Prime95 (Large FFTs) and it went on to 16hrs and still going to 17hrs but I stopped it noticing the temps were 67-69C on the cores. This is with 71-72F ambient. {Edit, can't live in 71C ambient - I'll toast}

Cooling setup:
Front: 1 Xigmatek 120mm intake @ 100%
Side: 1 Antec 120mm intake @ 50-55% blowing directly over HD4850
Rear: 1 Xigmatek 120mm exhaust @ 100%
Corsair PS fan: exhaust

With my current setup for cooling and voltage, I would think my core temps should be lower, no?

I do understand that 67-69C may still be within Intel spec for the Q9550 but do you think it should be lower? I'm thinking of getting the big brother of the SD964, the 3-heatpipe S1283, to hopefully lower the temps.

Any thoughts? Am I overreacting to these temps?

Thanks in advance
-francis

 

[dguy6789]

You LinX to test your real load temps. 67-69 for full load across all four cores is just fine.

 

[dbcooper1]

I'm running the same setup at the same speed/voltage on a UD3Rwith the larger S1283 and getting about 5-10 degrees lower, so I'd guess you're just about right. The 5 degree variance is in the cores on my CPU with core0 consistently running warmer. 71C ambient- yikes. #

 

[francisA]

Lol. Sorry it was 71F ambient

 

[francisA]

Thanks. I'll do some LinX and IBT later today.

 

[Phunk0ne]

if your current setup is giving you 67-69C with prime95, you defenitely do not want to do the intel burn test!

Prime95 is between 5 and 10C less stressful than the linpack/intel burntest.
You're better off getting the S1283 first before you are going to do the intel burn test!

 

[Idontcare]

Originally posted by: francisA
Testing for stability, I ran Prime95 (Large FFTs) and it went on to 16hrs and still going to 17hrs but I stopped it noticing the temps were 67-69C on the cores. This is with 71-71F ambient. {Edit, can't live in 71C ambient - I'll toast}

Large FFT is really more for testing the stability of the NB, FSB, and Ram.

If you want to test the CPU stability then you need to run small FFT.

 

[Phew]

Intel Burn Test makes Prime95 look like Minesweeper when in comes to stressing a CPU. I personally won't run it on my CPU, since Small FFTs already stress the CPU more than any game does.

I've seen upwards of 65C on Core0 of my Q9650 during Small FFTs, but never more than 55C while running Warhammer Online (a very CPU-intensive game).

If you are encoding, folding, etc, then I can see the value of running IBT. However, for just about anything else, Prime95 is a more realistic test of 'maximum' CPU load.

 

[francisA]

Thanks for the reminder. It's almost sounding like my 69C core temp is high for my current OC. As if I do IBT, it might literally Burn my Q9550
Look like I'm going to stop my MicroCenter after work today

 

[aigomorla]

guys look at his overclock.

70C for 4ghz is pretty darn good for air.

 

[error8]

Originally posted by: aigomorla
guys look at his overclock.

70C for 4ghz is pretty darn good for air.

That is amazing indeed, especially with such a low voltage.

 

[soccerballtux]

That's a little warm; not a problem though. Heat doesn't age chips nearly as much as high voltage-> electrons tunneling through the dielectric.

Rarely will your cpu actually run that hot, only if you were folding or encoding videos will it actually get this warm. So, I wouldn't worry about it.

 

[SunnyD]

Originally posted by: aigomorla
guys look at his overclock.

70C for 4ghz is pretty darn good for air.

But is it viable for long-term 24/7 use? Hardly.

It does raise the question though - OP, what are your IDLE temps?

 

[francisA]

Thanks all for the tips/advice as well as props for my oc. Gotta agree myself, hitting 4GHz was great. I even had it running also up to 490*8.5, and I can at 1.328v. 495*8.5 and up is where the drama starts. I can hit 500*8 and be stable for about 1.75hrs and 500*8.5 for only 15mins LOL The UD3P and Q9550 has served me well so far, not mentioning my poor G.Skills that have been pushed to DDR2 1000.

Anyways back on the subject. Here are my temps

Idle:
CoreTemp: 38-38-43-40
RealTemp: 33-33-38-35

Loaded (Prime95 Large FFT)
CoreTemp: 67-69C
RealTemp: 62-64C

-- Cores 2 and 3 are somehow stuck and doesn't go any lower than that. But when they're all loaded, all the temps are within 2C but Core 0 seems to be a little warmer almost all the time.

I'm going to try Prime95 Small FFTs then Blend. I remember seeing lower temps when doing Blend. The Large FFTs yield higher temps - not sure if this really should be the case. Thought I'd do some more testing before I invest in a new cooler.

What's interesting, or maybe not, is when I had it at 490*8.5 and did a run of OCCT Mix, max temps were 62-64C as reported by CoreTemp. RealTemp was 57-59C.

 

[OCGuy]

Are you ever going to stress the cores as much as you are in the simulated tests?


You are waisting too many clock cycles. 8 hrs is enough.....

~71 C is the intel max for 3+ year lifespan.

 

[Idontcare]

Originally posted by: francisA
I'm going to try Prime95 Small FFTs then Blend. I remember seeing lower temps when doing Blend. The Large FFTs yield higher temps - not sure if this really should be the case. Thought I'd do some more testing before I invest in a new cooler.

You should see highest temps with small FFT, specifically test 3 IIRC.

Large FFT is hitting the cache and ram a lot, the matrices are intentionally large to force the hardware to stress-test these parts of the system. As such the core's themselves spend more idle cycles waiting for cache fetches with Large FFT than they do with small FFT, hence the core temps with large FFT will be less than with small FFT.

Blend is just that, the program oscillates cycles between basically doing a small FFT test followed by a large FFT test. It is in the middle, but depending on whether you happen to look at CPU temps while the blend run is doing a small FFT test or a large FFT test will determine whether it appears to you that the blend test is producing more or less heat/temps.

In general there is really no point to blend because an error during blend tells you nothing for debug purposes and it also stresses neither the ram/NB/FSB or the CPU for a continuous duration as the small vs. large FFT stability tests themselves will do.

If you like, for fun, run small/large/blend test with Coretemp set to log temps in 5s intervals then plot the temps after a 1-2hr run. Small FFT test #3 is a real barn-burner.

 

[TidusZ]

Not to brag or anything but I get similar temps on 1.392 volts and 4.25. I have lapped the processor and TRUE 120 to mirror finish and used ocz freeze, with 2 scythe s-flex @1600rpm push/pull. I haven't played with the voltage much so I think after reading your awesome results with under 1.3 I might go lower and do some stability testing.

 

[n7]

Nothing wrong with your temps based on your cooling, voltage, CPU, etc.

As has been mentioned though, Large is not really a CPU test.

Run LinX or IBT @ max RAM (will set larger problem size & consequently higher temps) or P95 small FFTs to stress the CPU.
Use Large for FSB/NB testing, or Blend for everything.
HCI Memtest & Memtest86+ 2.11 for RAM stability.

 

[francisA]

Well ..... I re-ran P95 last night, both Small and Large FFTs, but this time just 2-hr durations as IDK suggested and captured the temps.

Small FFTs: 65 (Min); 68 (Max)
Interesting is that the max of 68C was reached within the 1st 17mins then after 20mins, it leveled off to between 65-66C

Large FFTs: 57 (Min); 65 (Max)
A broader range of temps. Didn't hit 60C until after 50mins. Stayed in the 62-63C range up until the 1hr 50min mark. After that, the temps stayed at 64-65C.

That's for a 2-hr run. In my op, it ran for a little over 16hrs so maybe it's just getting progressively hotter with Large FFTs as more components are being stressed and I'm sure if I run Small FFTs for the same duration {which I won't now }, it will probably reach same temps or 2-3C higher.

Also just for fun, tried running IBT 5-passes but watching it closely and temps went up to the 71-73C range. Had my finger on the Abort button just in case

 

[Idontcare]

Originally posted by: francisA
Also just for fun, tried running IBT 5-passes but watching it closely and temps went up to the 71-73C range. Had my finger on the Abort button just in case

I can appreciate the feeling of angst that can develop regarding thoughts of your hard earned money going up in smoke should your CPU start to melt and burn...but really you have absolutely nothing to worry about unless the thermal throttling hardware on your CPU is malfunctioning somehow (which has never been reported by anyone, ever).

If your CPU gets to hot but Intel's standards (and who best knows what is acceptable temps for your CPU than its manufacturer?) then it will automatically throttle down in clockspeed as well as use process halt states to immediately cool off the CPU.

Your concerns would be well-placed a decade ago before thermal throttling became an everyday feature of modern CPU's...but nowadays it is a misplaced concern. (but still common, don't feel like you are the only one thinking like this)

 

[SunnyD]

After seeing your idle temps - for an overclocked system where you're at, I'm going to change my stance and say you're probably okay unless you plan on loading your cores with Prime-like loads 100% of the time.

 

[TidusZ]

Originally posted by: Idontcare

Originally posted by: francisA
Also just for fun, tried running IBT 5-passes but watching it closely and temps went up to the 71-73C range. Had my finger on the Abort button just in case

I can appreciate the feeling of angst that can develop regarding thoughts of your hard earned money going up in smoke should your CPU start to melt and burn...but really you have absolutely nothing to worry about unless the thermal throttling hardware on your CPU is malfunctioning somehow (which has never been reported by anyone, ever).

If your CPU gets to hot but Intel's standards (and who best knows what is acceptable temps for your CPU than its manufacturer?) then it will automatically throttle down in clockspeed as well as use process halt states to immediately cool off the CPU.

Your concerns would be well-placed a decade ago before thermal throttling became an everyday feature of modern CPU's...but nowadays it is a misplaced concern. (but still common, don't feel like you are the only one thinking like this)

anyone remember the video where the guy was playing quake on the athlon thunderbird and every time he'd take the heatsink off the game would freeze, then he left it off and within 3-4 seconds it started to smoke and catch fire? That was awesome.

 

[francisA]

TidusZ: you got awesome oc there on your Q9550, with good temps to boot. Would you say the lapped TRUE played a huge role in attaining that oc?

Would you be so kind to share your bios settings? Or maybe you already have a thread about it that you can direct me to? I can maybe at least try and see how close I can get.

 

[francisA]

Thanks all for the great responses/advice.

I now feel better about the oc and temps.

I might not even get close to the kind of stress P95 or any stress testing apps can do to the system. Once I get to doing more video-encoding, I could probably revisit it but so far, it has been good. Gaming doesn't come anywhere close to stressing the Q9550, at least the ones I play.

 

[Idontcare]

Originally posted by: TidusZ
anyone remember the video where the guy was playing quake on the athlon thunderbird and every time he'd take the heatsink off the game would freeze, then he left it off and within 3-4 seconds it started to smoke and catch fire? That was awesome.

http://www.youtube.com/watch?v=cCmai6Q9DoM