Sign in with
Sign up | Sign in
Your question

Uneven Core Temps?

Last response: in CPUs
Share
December 26, 2007 11:24:41 PM

I just installed a Q6600. Within the operating system, on the internet, etc, I'm getting around 50C for all four cores, within a two or so degrees either way. However, when I torture test in Prime95, two cores are running at between 70 and 73, and two are running at between 62 and 65. Why such a large gap in temps between the individual cores, under load?

More about : uneven core temps

December 27, 2007 12:52:10 AM

Could be uneven spread of thermal paste. You could re-seat it and reapply the paste.
December 27, 2007 1:06:40 AM

I followed the arctic silver directions precisely. If it was an uneven spread, wouldn't the cores show more of a disparity at idle?
Related resources
December 27, 2007 1:27:55 AM

Okay...here's some temps:

Within the operating system, with the processor only running windows and whatever minor background apps that might be going, these are my readings:

Core 0: 50C
Core 1: 48C
Core 2: 50C
Core 3: 47C

And now under prime 95 torture test:

Core 0: 73C
Core 1: 62C
Core 2: 73C
Core 3: 65C

If it is an uneven spread of the thermal material, which I don't think is the case, is there any possibility that it would distribute itself evenly over time?
a c 126 à CPUs
December 27, 2007 1:28:01 AM

I doubt it is a uneven thermal paste issue as I used what came with my Zalman CPNS 9700(its in a little bottle with a small application brush) and made a thin even spread. I am thinking of going to Arctic Silver 5 but maybe later.

My Q6600 normally runas a 32c-40c idle(if its cold out its 32c) with core0/1 at 35c and core2/3 at 30c. I think it is an effect of the 2 dual cores as a quad.

Also it would make sense if each die ran at different temps this way but maybe not. But yours run really hot compared to mine and I have mine OC'ed to 2.7 GHz and prime barely pushes it to 51c average between cores when Prime95'ing it. Are you OC'ing yours to like 3GHZ+ or is it the stock cooler?
December 27, 2007 1:32:27 AM

Maybe 2 of the cores isn't exactly running up to speed? When you run prime95, do see that the cooler cpus are running slower than the hotter ones?
December 27, 2007 1:37:06 AM

jimmysmitty: What CPU temp software are you using? If you're using anything other than Coretemp (including BiOS, Speedfan, etc), add 15C to your numbers to get your actual core temps.
a c 126 à CPUs
December 27, 2007 1:38:22 AM

When I Prime95 mine the cooler CPUs run at 2.7GHz along with the other ones. I think its just those cores run cooler. As long as he doesn't notice the performance dropping drastically he should be fine.
December 27, 2007 1:40:04 AM

Are you sure Prim95 is running on all four cores? Try using Orthos instead.
December 27, 2007 1:41:37 AM

How do you see clock frequencies with Prime95?
a c 126 à CPUs
December 27, 2007 1:42:06 AM

That is with Speedfan and 15c added. It shows cores0/1 at 19c and cores2/3 at 15c. I use the BIOS, speed fan and others to match up the CPU total temp. Also I have a temp monitor on my case attatched to the base of my heatsink and it shows normally 35c while my case total temp is usually 30-32c.

I think your CPU is fine and that it is just some of the cores run cooler than the others.
December 27, 2007 1:53:34 AM

.. to be honest i think this is a thing with the quad cores..

im using artic silver 5, with zalman9700nt

my temperatures at idle are
37, 35, 31, 37


and im running this at 9x333.. i tried reseating the paste several times.. same results but sometimes different cores like

31, 37, 34, 37
a c 126 à CPUs
December 27, 2007 1:55:39 AM

Oh yea Prime 95 is using all 4 cores. On both task manager all cores show 100% activity and Pime95 show work on all 4 cores.

Also I don't think only one core would work at the top speed and the others wouldn't as this is the early version of SpeedStep where all cores downclock and upclock. The next step(Nehalem I believe) will have a SpeedStep where each core can underclock and if you are using a single threaded app, one core will overclock and the others will down clock in order to speed up that thread.

So I doubt its a faulty core. I looked for a program that would allow us to see each cores speed individually but couldn't find one. CPUz only shows one speed for all cores.
December 27, 2007 2:06:08 AM

I ran Everest just to verify Prime95 was doing its thing, and these are the results I got:

Core 0: 73
Core 1: 63
Core 2: 71
Core 3: 65

These are almost identical to what I got in prime95. Given that other people have had similar experiences with the Quad, I'm just going to assume it's the nature of the beast...unless someone steps in to inform me otherwise.
a c 125 à CPUs
December 27, 2007 2:30:13 AM

i have a gap of 7c from my coolest to hottest core. it will not hurt anything....

If you want to try to redo the thermal paste apply a thin layer with a clean razer blade or credit card over the full heat spreader.....this may give more even results....other then that, i do not think you have too much to worry about....

Q6600 3.0(gigabyte increased the fan speeds[in smart fan]...used to run 57-58 at full load). This is folding all day..may flash back for slower fan speeds :) 
December 27, 2007 3:51:25 AM

If you're desperate, you can always try lapping both the heatsink and Cpu.
December 27, 2007 4:33:18 AM

I'd keep an eye on the core with the highest temp when you're overclocking.

If the temps are still bothering you, as a lot of people suggested try reapplying the thermal paste. This time spread it evenly all over the heatspreader instead of using the AS5 method.
December 27, 2007 6:19:06 AM

As long as I can keep the hottest core under 80 during torture testing, I'm fine. We'll see what happens. Given how tedious it was working with a Micro ATX case, there would have to be a serious problem for me to want to disassemble the thing to get at the CPU. In fact, I'd rather keep it stock and not OC as opposed to taking it apart again. I'll try to eek as much speed out of it as possible, but once I'm brushing up against 80 I'll stop.
December 27, 2007 6:52:15 AM

just for ur refrence. mine at idle is
37 33 35 31

at load with primer.

60 57 59 58

n i've put smart fan enable. since i wanted to reduce the noise levels.
December 27, 2007 7:28:28 AM

Hey, maybe you guys might know the answer to this...

How would increasing the clock alone, without touching the voltage, impact performance? I've heard of many instances where people OCed to around 3.0 GHz without increasing stock voltage, and in some cases lowered it. How much would temps rise going from 2.4GHz to around 3.0 GHz, without increasing voltage?
a c 125 à CPUs
December 27, 2007 12:36:00 PM

maybe 3-6c in most cases without extra voltage. Its over 3.2 when things start to get REAL hot(due to the need for more voltage in most cases, some people have golden chips that dont need extra) without good cooling....

You just increase the FSB from 266 to 333(for 3.0), make sure you do lots of prime95 to test it for stability. 24 hours is good...keep an eye out that this tweaking does not push your ram over specs(you don't want it to hold you back). If possible run the ram at 1:1 this gives lots of headroom(up to 3.6 for DDR2 800 3.0 for 667)

On the idea of voltage dropping. It depends on the chip, it can save you close to 10c in the right circumstances. My E6000 @ stock is dropped to 1.2(stock is 1.325) and my A64 3200 to 1.4 (The board was feeding it 1.55) so the temp drops there are close to 10c.

At stock you may get down to 1.2 or even lower. According to speedfan my Q6600 @ 3 is getting 1.17 - 1.18 at load(V droop)...
a b à CPUs
December 27, 2007 1:31:22 PM

I have my Q6600 @ 3.0, 9x333 and it runs great without an issue. I have a 1/2" ID water loop on my CPU, northbridge and GPU, so temps never get above ambient at idle, and rarely get above 35C under load on all 4 cores. From what I have read from almost everyone else is that once you start clocking up the quads, they become small room heaters.

Make sure you have a decent aftermarket cooler to keep up...I never have liked the stock coolers sent with processors...ever.
December 27, 2007 6:31:07 PM

I really don't have much of an option as far as aftermarket heat sinks go, as I'm working with a micro ATX case and have very little room. This actually isn't even my computer, it's a build for my brother. It's running perfectly fine with everything stock. His opinion is, if we can get it OCed without messing with any of the hardware, great. If not, oh well. Given how hard and tedious it was to get everything in that case, there is no way I'm going to go fooling around in there if I don't have to.

I'm hoping to get up to 3.0 after messing with the voltage, etc. However, if I can only get it to 2.5, then so be it.
a c 125 à CPUs
December 27, 2007 7:08:44 PM

you can try stock with low voltage....cool and still fast enough.....
December 27, 2007 8:35:31 PM

I'm going to lower the voltage as much as possible. In the process, I'm going to see how far I can get it to OC while still remaining within a reasonable temp level. My thinking is, I can get it to at least 2.6 as it stands.
a c 125 à CPUs
December 27, 2007 11:34:18 PM

let us know how it goes...
December 28, 2007 12:34:03 AM

Thats perfectly normal, even my lapped X2 3800+ with lapped ninja have uneven core temps.
January 5, 2008 12:45:43 PM

So, just an update for all interested parties...

When I ran Prime95, I was doing the standard test, not small FFTs. When I did that, I ended up around 77*C with my hottest core, stock (yikes!). This case does indeed have poor circulation on the CPU side. Damn you micro ATX! ::shakes fist at the sky::

Anyway, I figured I'd tool around in BIOS before I undid my heat sink and refastened. I'm not sure what the voltage was set to stock, because the choices are either AUTO or choose your own, but it must have been pretty darned high. I was able to over-clock to 2.7 GHz (roughly 11% increase), AND lower my temps a full 6*C from stock. Voltage is at 1.175. The hottest core is 72 after 12 hours of stable Prime 95.

I stopped trying to get 3.0 GHz when it would not run stable at 1.25, at which point it was already beyond 80*C. I had it going stable at 2.9 (can't remember the voltage), but the temps were right at 81*C, which was just too much. At 2.8 GHz, the machine required around 1.23, which was running too hot for me, at 78*C.

Regarding my current settings, I may or may not be able to go lower on the voltage. I had it set one notch lower than 1.175, and went into the other room. I came back in, and my brother said "the OC programs crashed." I asked him if it rebooted or anything like that, and he said no. My guess is he just Xed out of them, but whatever. 72*C for the hottest core seems more than comfortable for me, considering I was shooting for under 75. After 12 hours of Prime95, I really don't think it's worth it to keep the computer out of service any longer. Based upon the trends I've seen at other frequencies, I'm probably not going to get the voltage much lower than it is right now, if at all.
January 5, 2008 1:24:26 PM

well for one this is a torture test and since that case is restricting airflow its gonna get hot and also i dont think prime 95 supports more than 1 or 2 cores(correct me if im wrong) which may explain the uneven temps because only a couple are actually working hard
a c 125 à CPUs
January 5, 2008 3:07:15 PM

As long as its a newer Prime95 It will run on as many cores as it can(or as many threads as you tell it to use) :) 




@ JJBlanche - What case do you have?
January 13, 2008 7:06:35 PM

This build, with the heat, was for my brother, whose running a Falcon Northwest Fragbox. It has got to be the most tedious case to work with.

Me, on the other hand...I'm running an Antec 900 with fans out the wazoo. 66*C hottest core, full load Q6600 3.6 GHz.
!