Thinking about lapping my q6600, temps too high

A

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
4,086
Hi everyone,

I've had my system up an running for about a month now and I'm quite pleased with everything but my cpu temps. I was going to let it go until I realized tjmax is not 90c as I thought, it's probably more like 100c. I am using ocz freeze thermal paste for gpu cores, chipsets and cpu, ceramique on the gpu ram. My system is in my sig, and these are the max observed temps taken from everest, note this was with a 90c tjmax:

CPU: cores spread between 59c-62c during prime95 for 34 hours, idle 36-39c
GPUs: folding@home seems to put the biggest stress on them, the cores have never passed 48c, idle around 38c
MCP and SPP: throwing some volts at these and yet neither have gone past 41c, idle 35-36c

So this means my cpu load temps were more like 69-72c! I'm seeing about 20 degrees difference from idle to load too, which shouldn't happen with my cooling loop. I've tried both the drop in the middle as per ocz's site and have switched to the line technique as per arctic silver's site. Both the line, and quad insert washer, were oriented correctly. Even though my system is stable I can't help but think that elusive 4ghz would be possible with some lower temps. Thoughts? Thanks!
 
All that on one loop?

No wonder your temps are high.

Might want to break up that loop. Maybe keep the dual 280's on their own loop, NB/SB and CPU on their own loop.
 
Not possible to seperate loops, I don't have one of those enourmous double wide cases. I'll post pics later. This system was custom built taking me over 9 months of weekend time, no changes to the physical layout are happening. How is it possible that the other components are 20c cooler than the cpu if it's just all one loop? Certainly the 260's put out as much heat as the cpu. I just don't understand.
 
I'd still venture to say it's too much on your loop. I don't watercool much, so my knowledge on WC'ing a quad is nil, but WC'ing an OC'd E6600 I've done. Even then I had to do 2 loops, keeping the vid cards out of the CPU loop. Temps would just get out of control.

One thing you could try is adding another rad. Or better yet, could you provide some pics? I really don't know where to start with your rig. How is the loop layed out? Are the temps of your other cooled parts rising as well with the CPU? If not, then that is a good place to start. Have you thought about an Apogee GTZ or FuzionV2?
 
loop is in my sig. I'm not taking this thing apart, the difference between a v1 and a gtz/v2 is like 2 degrees. The temps all increase at load, please see the op. Generally the cpu only goes up 2 degrees sitting idle with gpu at full. I just moved and cant find my memory card with the pics, I may have lost it. :( It's a cooler master stacker case with the triple rad up top and the single on the back. Again, no changes to the loop or setup are allowed, the only thing I can do at this point is change thermal pastes or lap the processor. Again, how is it possible that my gpus, which are known to run much hotter than cpus, running much cooler? This tells me the cooling loop is doing a hell of a job, but something between the cpu block and the cpu isn't meshing.
 
So, as an update, I attempted to see if I've just got too much going on in my loop. So I popped the second gpu from the motherboard and removed the power connectors from it, effectively removing it from the cooling loop. the results are as follows:

cpu: cores between 55c and 59c during full prime95 stress, 33-35c idle
gpu: stress at 44c, idle at 38c
mcp/spp stressing at 39/40c.

So by removing the second gpu, I lost anywhere between 2 and 4 degrees in all temperatures. This is far from the drastic reduction in temperature I would of expected, if the loop was overloaded, as was previously thought. Remember my gpu downlocks when idle so it reduces its heat output greatly when not in use.

My conclusion is that there is something seriously wrong with the cpu portion of my setup. I have drawn this conclusion from the items listed below:
1. ~25c difference between idle and load on a medium overclock is too much for a liquid cooled setup
2. The cooling loop is adequate for this setup, as the heat output of two gpus at full load and an idle quad outweighs the heat output of a loaded quad and one idle and downclocked gpu.
3. Even with the cpu roasting away, the loop maintains the gpu's at under 49c and the chipsets at under 43c at all times - if the cpu heat was being transfered to the loop correctly and it really ran that hot, these components could not maintain such low temperatues (at least not the gpus, which are actually designed to run 90c just fine)
4. All my testing is based off an inccorect tjunction max of 90c - the commonly accepted setting is 100c. That means I'm actually idling at high 40's and loading at low 70's!

So I have three courses of action to take, listed easiest first:
a) experiment with different TIM applications, as well as trying a different product
b) my q6600 is just stupid concave, and needs a serious lap job STAT
c) Something is wrong with my cpu block. Maybe the quad insert rotated, maybe somehow something got into the loop and has clogged it, I don't know. Not looking forward to checking this, as draining my loop is nearly impossibly without making a terrible mess.

Will keep this updated. Look forward to anyone else chiming in, thanks!
 
I'd have to agree with the other poster, it seems that the problem lies with the loop. I understand your hesitation to take everything apart though; I have the same issues with my system, having so many components in a tiny antec 900, and i'm not even watercooling. However, I'm pretty sure the REASON for your high temps is what the other guy said. For reference, my q6600 at 3.6 idles at 32, 29, 27, 27 and at full load gets up in the 60's. However, this is with intel burntest, so these are like the absolute max temps I could ever get with any stress test. And I'm running aircooled - a TRUE 120 og edition with a scythe s-flex on it. HOWEVER, my CPU and TRUE base are both lapped - i think i went up to 2000 grit on both. This is the reason for my low temps.

So in conclusion, to at least help your problem, i highly suggest lapping both the cpu and maybe the water block as well. Absolutely no reason not to (besides voiding the warranty, pfft) and your temps will go down! Plus, my temps got even lower after I scrapped the AS5 line-across-the-chip method and went to MX-2 with a thin layer spread evenly. Good luck man! Lapping FTW!
 
Yeah, I was thinking of trying the thin layer technique next. That's what I used on the gpus, and they run very cool. Will definetly lap the proc. You guys really think it's still a loop issue even after removing a gpu completely from it reduced temps only 2-4 degrees? O well, loop is NOT changing. I'd have to throw my $5,000 computer away and start over basically, and I'm not doing that. Remember I have two pumps in series pushing serious flow, and a pa120.3 with 6x fans on it. Anyway I might take both gpus out of the loop tonight and use a pci vid card, just to put this debate to rest. I predict 2-6 degrees lower than 1 gpu, meaning idle of 36-38c and load of 59-62c in a cpu and chipset only high flow loop. We'll see.
 
Yeah, at the very least give lapping the CPU and CPU waterblock a shot. I'd venture to say you won't see much more than maybe 1-2c drop though.

IMO simply dropping the clock and voltage on one of the GPU's doesn't remove it from the loop. It's still running, still putting out heat. And you seeing a 4c drop by doing that all the more solidifies my previous statement of too much on one loop. If you removed those 2 cards from the loop you could expect to see roughly a 10c drop in temps. Slap on a FuzionV2 or GTZ (heard good and bad things about V1's with quad insert) and that should put you in the temp range you're looking for, which is what most people had to do for those temps on a 65nm quad. And, 3.825Ghz is far from a modest overclock. That's pretty much the upper end of that chips capabilities. It's throwing out some mad heat.

Regardless, 69-72c at load for that much on one loop isn't something I'd be concerned about. You're well within the chips limits for temperature.

I know tearing apart the WC system is the last thing you want to do. Most of the time it's neccessary to troubleshoot. Becomes a bottomless pit of money, time and aggrevation until you get it right. Those three reasons are why my wife made me sell all my crap and give up water cooling. I, personally, will never go back to it.

Welcome to water cooling! :)
 
Yeah. 10c isn't adequate though. I should be seeing MUCH lower temps on such a beefy loop if its just cpu and chipset, though. Something like 50c load I would expect, and I'm nowhere near that. I'm also only using 1.45v for this oc, people push 1.6v so I've just doen soemthing terrible wrong. That's my real problem. And I had a slap on swiftech apex kit previously, so it's not my first venture into watercooling, but thanks. I just made mistakes during my first case mod, which makes loop maintence damn near impossible. Thanks for all the help, but I quit. Just gonna leave it as is, and stop monitoring temperatures.
 
Firstly im new to the forums, but in no way new to Hardware tweaking, overclocking or watercooling for that matter. Now I firstly ask which rad your running, because when i looked up your radiator it comes up as a 120x1 Single core rad, and your attempting to run? on it 2 gtx260 a Quadcore e6600, north and south bridge? with just a 120x1 rad.
IMO radiator is far to small to sufficently cool that setup.

As of Current, my Cooling loop consists of a Swiftech Mcp650, A Danger Den 120x3 radiator, cooling a 3ghz Amd opteron 170 dual core @ 3.0Ghz, (which also has the IHS removed) and a 8800gts 640mhz on the core 1050mhz on the memory, with the Shader Bios modded to 1564mhz @ 640 core.

Though its no where relevent to the temps dual gtx260' a north south bridge and a Q6600 throws out. i run at about 29'c cpu and 28'c core temp idle and under 40'c loaded 100% both cores. while my Vga stays under 45'c Loaded playing crysis.

So IMo Two DDC 3.2 pumps is un nessisarry, reason being, is you can have all the Water pressure and Flow rate in the world, but if you dont have sufficient Surface area on the radiator end of the loop, your temperatures are going to suck.
Further more a single 120 rad isnt exactly a beefy rad, and in my personal experience i wouldnt put more than a dualcore Cpu and a single video card on a single 120 rad.
maybe 2 cards if you dont plan on over clocking anything, but 2 cards a q6600 and a northbridge south bridge is just farr too much for that loop to handle.

Id accualy like to see some pics of the machine posted up so we could try to help you out a bit more. Anything is fixable, and it may not cost as much as you think so dont get discouraged.
 
let me correct my self, before i wind up getting flamed. I just noticed the TC 120.3 in which case I would deffintly like to see some pictures of your case and setup, because Theres deffinitly somthing going wrong in there if your hitting temps that high with 4x120 worth of radiator surface.
 
I have a similar radiator for my loop. The BI120.3. Your radiator should barely be adequate for the heat your putting out on those components, assuming you have the fans maxed. I know on my radiator, if I have the fans down to minimum, I have pretty significant heat build up fairly quickly. Those radiators are so thick, they need some powerful fans to move air through them properly. Since you have 6 in a Push/Pull configuration, I assume, your airflow should be adequate.

That said, if it were me, I'd be looking at the contact between the CPU and the waterblock. It sounds to me like you have a concave heat spreader on your CPU that's now allowing a good contact across all cores. Lapping the CPU would help a great deal I'd imagine, if that is the case. That's where I'd personally be looking next, since lapping the CPU can only help and you've nothing to lose but time. :)
 
You must log in or register to reply here.
Back
Top