Hi all,i've recently put together an SR-2 BUILD using water cooling , and i'm getting weird temperature readings from one of the CPU's, even though i've checked and double checked the cooling and even added an additional radiator to the loop, and the difference is minimal...Here's a picture of the beast in question:
 
 

 
 
One of the CPU's is always running hotter than the other, and the spread between CPU 1 and CPU 2 can easily reach a good 20*C, so it just simply can't be the cooling given what i'm using, so is there a known glitch at the BIOS level for temperature reporting, or is it a software issue perhaps with the temperature monitoring software itself?...

Thanks for the compliment and the offer, but i think i'll pass...;)
 
 
I think i heard at one point that it was a bug, but other than that the machine performs great and is stable....It's just that one thing that annoys me, and wair until you see the water cooling loop completed for the video cards...
 
 

I'm not talking about the socket temperature readings and potential differences between both, but that of the individual cores inside the processors themselves, and there's a huge gap between CPU 0 and CPU 1 using Eleet or any other temperature mesuring program....
 
For instance, i ran a quick run of cinebench 11.5 wich stresses all the cores on both processors to 100%, and all cores on processor 0 have a hard time breaking 55*C at max load....Processor 1 however is a whole different ballgame, easily hitting 85*C average across all 6 cores at 100% load on each core for the duration of the benchmark run...???

Tried twice already....Thanks for the suggestion anyway.
 
The crazy thing is that immediately after i end a workload that hammers all the CPU cores to 100% for sustained periods, the temperature readouts on the hottest CPU drop for 85*C to 45*C in less 1/2 a second, so good contact is being made to dissipate the heat coming out of the CPU.....
 
However unlikely it may be, is it possible that the thermistor built within the CPU that reports core temperatures may be defective somehow?....Oh and the LED display on the motherboard reports the same high temperatures for the hottest core, as the E-leet software i'm using(it's the latest version too), so it doesn't seem to be a software issue at least... 
 
Leaves a bad CPU(thermistor really) or a bios bug regarding temperature reporting being inacurate...I'm still somewhat in shock that the additional radiator i added at the back of the case in the picture, hardly made no difference at all relative to the offending CPU...It's basically hitting the same temperatures with or without it in the loop....

True, i can try that to see if it moves from one socket to the other, and if it does it has to be the CPU.....If it doesn't, then it's something related to the motherboard itself or a Bios issue.
 
It's a great idea, i'll switch them when i get the remaining parts for the video card water cooling loop and see what happens.... Thanks

I would swap CPU's like previously mentioned. At least then you'll know if it's the CPU or otherwise.
 
Some of these have been mentioned, but this is what I would check.
 
- unbalanced tension on mounting hardware (unlikely as you've un-mounted it twice as you said)
- over-tightening. simple mistake that is often made.
- over application of tim. One rice grain sliver is adequate.
- unbalanced motherboard. Sometimes mounting hardware on the backplate can interfere with the tray in some cases. Thus lifting the SR-2 up making so it cannot lay flat.
- finger print or debris on gold contact pads of CPU.
- bent pins in socket.
 
 
From the sounds of this though, I wouldn't be surprised if you had to RMA something.
 
Good Luck

Hi Shadow,

i had the same problems, but only about 10C. In my case the cpu cooler wasn´t screwed correctly. That means that the four srews of each coller were tied differenly strong.
Hope that works for u.

greets

 
Well, with the Swiftech CPU water blocks, it's pretty much impossible to screw them with different tensions since all 4 screws only screw in up to a certain point and then just stop, so the only way for that to happen is if one of the 4 spings is weaker than the others, then the tension Could be different perhaps. 
 
http://img.ncix.com/images/59621_l.jpg

There is an unusual aspect of your CoreTemp readings which deserves investigation. The temperatures in the "Low" and "High" columns appear "reversed". To be sure that this was not an error or incompatibility of the CoreTemp software, I ran the program on my computer and as shown below, the values in "Low" column were, as expected, less (rather than greater) than those in the "High" column.


http://img638.imageshack....638/5145/coretempi.jpg

The correct enumeration of the "Low" and "High" temperatures might seem trivial. But, it deserves attention because it could be a sign that a specific portion of your BIOS is corrupt. If this is indeed the case, I would try resetting the BIOS and select OPTIMAL CONFIGURATION in the SETUP (or something like that). Then verify that the lows and highs read by CoreTemp (hwMonitor, etc) are appropriate. If your BIOS version is old (e.g., prior to A46), you might try updating it.

Wish you the best --- RP

The temperature disparity between CPUs shown in your last post is NORMAL. Some of us believe that CPU-1 doesn't require as high voltage as CPU-0. For example,  the CPU voltages I used in overclocking @ 184 x 19 is:
 
CPU 0 VCORE = 1.2000 V
CPU 1 VCORE = 1.1565 V
 
My computer runs stable and cool with those settings. And, with respect to your interest, the temperature of my two CPUs are very similar at full load.

Hey from my experience with my sr-2 the one socket always runs atad hotter then the other usually at stock for me they idle at 23c while the other cpu at 26c im not using water but i am using 2 corsair h70 coolers. Its always been that way for me anywaise hope this helps. Oh and even when ocing and running 100 percent full load 1 core is always about 3 degrees more hotter then the other.
 
 

shadow001
I guess that's an option to try should all else fail, but why does it happen in the first place when the voltage settings for both CPU's are the same...I mean either they both overheat or they both run cool as the workload is the same across all the cores, they're both cooled by the same method as well...

I don't know the answer to your question. I suppose that only those involved in the design of the motherboard know. But I could speculate. Here's the possibility came to mind . When you set each CPU voltage to 1.20 V, the two CPUs might not have identical electrical power input. That is because the voltage is not DC but rather is modulated by a power regulator circuitry. Because of the topographical location of the two CPUs, the voltage regulation for one CPU might be more efficient than another. So setting their voltages identically might actually supply more juice to one, i.e., CPU-1 (according to my speculation).

Take a look at the motherboard (below) and notice how there is relatively less space  available for CPU-0 PWM components in comparison to the CPU-1. For the sake of clarity, I circled the two power regulation components in red. The area around the CPU-1 PWM is slightly larger or, at least, its heatsink is larger. Such differences might suggest that, perhaps, the power regulation is not identical or something like that.

shadow001

That's the main problem though....It's a 25~30*C difference under load between each CPU, and about 10*C at idle.....I wouldn't even care at all if it was something as small as 3*C difference, as that's pretty much considered a margin of error variation.

I'd love it if it was that small a difference in fact...

 
I would try reseating your cpu and respasting both of them to be honest sounds like a small problem try using some better thermal paste also i dont personally know what your using but i myself use the artic silver stuff works great. I believe that if u do that then you should be seeing some better temps.

Lol after seeing your pics there lower your vtt and ioh voltage thats too much i can get  stable oc with my 5650s at 1.21vcore and 1.25vtt and ioh and thats at 3.8ghz. vtt causes more heat i would defently lower that abit lower your vcore too u dont need 1.3vcore for 3.2ghz

shadow001

I would try reseating your cpu and respasting both of them to be honest sounds like a small problem try using some better thermal paste also i dont personally know what your using but i myself use the artic silver stuff works great. I believe that if u do that then you should be seeing some better temps.

Will do, but only when i receive the last batch of fittings for the water cooling loop and the full cover blocks for the GTX580's over the next week or so, then do all the work at the same time....The full fury of this setup is yet to be unleashed, but it's completely stable when all CPU cores aren't being pushed to the limit like that, wich the only things that can are professional applications anyhow.

These are the blocks in question:

 
  unless your doing 4.0ghz on that chip man u really need to lower your voltages like i said above i can do 3.8ghz with turbo on at 1.22vcore 1.25vtt and ioh 1.25v  your pushing too much voltage for a 3.6ghz that would defently increase your heat id back down if your not pushing higher overclock.

i'm not doing 3.2 Ghz...It's at 3.8 Ghz and with turbo on, it goes up to 4.18 Ghz if i'm doing my math right( 2 multiplier jump).
 
In any case, since both CPU's are using the same voltage, it still doesn't explain why one of them is running so much cooler than the other by 30*C under load....