Quad Core heating issue

[krnmastersgt]

I'm assuming I have to get some grease or re-apply my compound on my quad core since I'm seeing a 14C difference between the highest and lowest cores with the other two cores relatively close to each other in temperature. Also the coretemp readings are lower than bios readings, so how am I supposed to adjust this?

 

[Idontcare]

I see you are running a TEC (thermo electric cooler) so you might actually be seeing some affects of the TEC not being uniform across the CPU.

Assuming this is not the case, then yes the first thing to do is clean the IHS and TEC surfaces and reapply some TIM (thermal interface material) of your choice.

AS5, TX-2, MX-2, ceramique, anything will work for helping to eliminate a 14C delta.

The other thing to check while you have the TEC off the CPU is look for warpage of the TEC surface or of the CPU's IHS. Just place a box cutter blade across the surface, check from a few different angles.

If it is badly warped (look to see if light can be seen between the blade and the TEC or IHS) then you may have a CPU or HSF candidate for lapping to reduce the delta.

In general you should see at most an 8C temp delta, 14C is pretty much unheard of. Was for me until I read your post.

 

[aigomorla]

Originally posted by: Idontcare
I see you are running a TEC (thermo electric cooler) so you might actually be seeing some affects of the TEC not being uniform across the CPU.

Assuming this is not the case, then yes the first thing to do is clean the IHS and TEC surfaces and reapply some TIM (thermal interface material) of your choice.

AS5, TX-2, MX-2, ceramique, anything will work for helping to eliminate a 14C delta.

The other thing to check while you have the TEC off the CPU is look for warpage of the TEC surface or of the CPU's IHS. Just place a box cutter blade across the surface, check from a few different angles.

If it is badly warped (look to see if light can be seen between the blade and the TEC or IHS) then you may have a CPU or HSF candidate for lapping to reduce the delta.

In general you should see at most an 8C temp delta, 14C is pretty much unheard of. Was for me until I read your post.

while i totally agree with idontcare...

b4 you do all this, you need to load up your processor so all cores are putting out 100% to see if you really have a large delta.

Idle temps dont mean jack squat with variance as long as its close to any other core.

 

[krnmastersgt]

So I ran Prime95 and got all 4 cores to 100%, still seeing a wavering 8-10C difference in temperature between core 0 and core 3, cores 1 and 2 are just 2C above core 3

Also the temperature sensor with TEC is pretty useless at this point since half the chip isn't heating up so the damn thing reads a constant 25C at load while coretemp reads 30-40C

So should I go about re-applying some compound/checking for smooth surfaces?

 

[Andvari]

14C is pretty large. I have 8-9C difference while idle, and a 5C difference under load. I've re-applied AS5, and re-seated the heatsink/processor for no improvement. It seems that temp deltas of several degrees are pretty common among the quad cores, so I'm fine settling with my delta.

14 and 10 seems a bit excessive though. Might try re-applying and re-seating a couple of times and see how your numbers change.

 

[krnmastersgt]

Is there a reason why Coretemp is reading from 30-40 while speedfan is reading a solid 20? I know speedfan reports about 15C under the actual but it registers all 4 cores at the exact same temp

 

[aigomorla]

Originally posted by: krnmastersgt
So I ran Prime95 and got all 4 cores to 100%, still seeing a wavering 8-10C difference in temperature between core 0 and core 3, cores 1 and 2 are just 2C above core 3

Also the temperature sensor with TEC is pretty useless at this point since half the chip isn't heating up so the damn thing reads a constant 25C at load while coretemp reads 30-40C

So should I go about re-applying some compound/checking for smooth surfaces?

can you take a picture of your mount?


core 0 being way off means you have the thing tighten too much up top and you need to loosen it a tad bit on the top right corner and tigthen it a bit on the bottom left corner.

 

[krnmastersgt]

Oh, that makes sense, when I was mounting the cooler it was a pain in the ass to get to the core 0 end so I guess I overtightened it without noticing.

 

[Markfw]

Something close to this, I was having a problem with my X3210. Lapped it, it went down several degrees, then I tightened the nuts on my Big Typhoon, and it went down several more !

 

[Idontcare]

Originally posted by: aigomorla

Originally posted by: krnmastersgt
So I ran Prime95 and got all 4 cores to 100%, still seeing a wavering 8-10C difference in temperature between core 0 and core 3, cores 1 and 2 are just 2C above core 3

Also the temperature sensor with TEC is pretty useless at this point since half the chip isn't heating up so the damn thing reads a constant 25C at load while coretemp reads 30-40C

So should I go about re-applying some compound/checking for smooth surfaces?

can you take a picture of your mount?


core 0 being way off means you have the thing tighten too much up top and you need to loosen it a tad bit on the top right corner and tigthen it a bit on the bottom left corner.

You can do that? Over-tighten I mean? I just tighten the holy mother living shizzy out of mine and call it a day. Granted I don't break out any extra hardware to do the tightening, still just "finger tight", but actually "backing off" after full-tight is not something I would have thought to try.

I have a quad here that reads nearly 8C temp delta (after lapping and multiple reseats)...I may have to give this "aigo theory #356" a try. Wouldn't be the first time you taught me something new.

 

[MTDEW]

I have 8-9C difference while idle, and a 5C difference under load. I've re-applied AS5, and re-seated the heatsink/processor for no improvement. It seems that temp deltas of several degrees are pretty common among the quad cores, so I'm fine settling with my delta.

THANK YOU!
Ive reseated my Tuniq Tower 4 times trying to get rid of the 6c difference of my core 2-3 vs my cores 0-1.
Now i know its just "normal".

I kept thinking it was me, since its next to IMPOSSIBLE to mount a heatsink anymore without need to "slide it around a bit" to get things all lined up.

I dont get why we dont have some extra holes for "guide pins" so i can drop the heatsink down onto the mobo and know its lined up right away so i dont have to move it to finish mounting and mess up my perfectly applied layer of thermal compound!

 

[Idontcare]

Originally posted by: MTDEW

I have 8-9C difference while idle, and a 5C difference under load. I've re-applied AS5, and re-seated the heatsink/processor for no improvement. It seems that temp deltas of several degrees are pretty common among the quad cores, so I'm fine settling with my delta.

THANK YOU!
Ive reseated my Tuniq Tower 4 times trying to get rid of the 6c difference of my core 2-3 vs my cores 0-1.
Now i know its just "normal".

I kept thinking it was me, since its next to IMPOSSIBLE to mount a heatsink anymore without need to "slide it around a bit" to get things all lined up.

I dont get why we dont have some extra holes for "guide pins" so i can drop the heatsink down onto the mobo and know its lined up right away so i dont have to move it to finish mounting and mess up my perfectly applied layer of thermal compound!

For what its worth, MTDEW, when I install my Tuniq I place the H-bar (the silver thing) and all four screws/springs thru the H-bar while holding it far above my mobo.

Then as slowly being my tuniq closer to the mobo I eyeball the lineup between my screws (dangling straight down from the h-bar) and the screwholes inthe mobo.

I even do so far as to get all four screws threaded a couple turns while still "hovering" the tuniq about 1/8-1/4 inch above the IHS. Yes it is possible.

That takes about 90% of the side-to-side play in where the Tuniq will end up when you finally set it down on the IHS.

Slipping and sliding your HSF around on top of your IHS is not a super good idea. For one it defeats the purpose of letting the HSF downforce decide where the thermal paste needs to go, and for another you can cause some rather nasty scratches in the IHS and HSF themselves from "fretting".

I get the most consistent results by doing the hover, align screws and thread a tad, then set-down approach. In 4 of my 5 tuniq quad systems this eliminated 60-80% of my temp delta's across the cores. (I have one outlier system which I mentioned above)

 

[MTDEW]

Idontcare,
Wow dude, you've got some real steady hands, when i tried that, everytime i tried to start threading a screw, my "H-Bar" would rock and fall off the "tiny peg" that aligns it on the heatsink.
Then i'd end up needing 3 hands to realign it while holding the heatsink and trying to thread the screw. (and i only have two hands BTW...LOL)

 

[Idontcare]

Originally posted by: MTDEW
Idontcare,
Wow dude, you've got some real steady hands, when i tried that, everytime i tried to start threading a screw, my "H-Bar" would rock and fall off the "tiny peg" that aligns it on the heatsink.
Then i'd end up needing 3 hands to realign it while holding the heatsink and trying to thread the screw. (and i only have two hands BTW...LOL)

Well coincidentally enough I drink a TON of MtDew so my hands can't be all that steady...but you are right that it took way too much practice for me to finally figure out just how to do it. Come to think of it, its kinda ridiculous of me to suggest its a walk in the park cause in recollection it really was a PITA process to do.

 

[MTDEW]

Yeah, its been driving me nuts ever since the first time i installed the heatsink.
Every time im done , i cant help but think that if i can get it installed without sliding the heatsink at all, then my temps would be better.

I just cant seem to pull it off.

And here i thought once i didnt have to screw with those god awful pushpins anymore, i'd be perfectly happy.

I may try it one more time this weekend.
Since all these cpu lapping threads that have been popping up on here recently, now i feel the need to give that a try too...LOL

Im sure all of us who have a quad on air cooling feel the same way.....I never have truely seen my cpu's max Overclock, because i have to backdown due to heat first.
I really dont care for my cpu hitting 60c+ under prime small ftt stress testing.

 

[krnmastersgt]

I seriously hope that the reading on core 0 is wrong for me, too busy with schoolwork to take out my mobo and re-do everything since it seems like it loads at 40-45C at 2.4ghz o_o

 

[Duvie]

I have seen 4-6c difference between high core and low core and this is having mounted different heatsinks and water coolers on this at least a dozen times....I get pretty constant this DELTA between the cores....

ONe thing I suggest is you go to Artic Silvers website and look at their application diagram for a quad core. Because of how the cores are under the IHS they suggest a thin line of paste but it needs to be orientated in a specific direction to assure coverage.

Check it out...

 

[BonzaiDuck]

Well, reading through a lot of these posts, I see that I gravitate toward a forum that draws some of the most anal-retentive, compulsive-obsessive, nit-picking detail-freaks in the world -- including ME -- voted most likely among Anandtech Geeks to write a book when three sentences would do fine. :laugh:

So -- OK -- it's (currently) a $250+ processor. You might worry about your warranty -- if you bought the retail-box.

But -- krnmastersgt -- why not lap that heatspreader? It cannot do much harm. You don't even need to take off the nickel-plating and grind down to copper, if the goal is simply to get it flat. You can check it with a straightedge, razor-blade -- anything reliably straight to the fraction of a millimeter.

My only other observation -- as I partake among friends and acquaintances I've found in these forums:

"Why haven't more people gravitated to synthetic-micronized-diamond thermal paste(s)?"

IC Diamond is $5 for maybe 3 applications-worth; it doesn't degrade over time like micronized-silver pastes; and if you're careful, you can actually re-use it by adding a touch of Jet-Art CK4800 (10% nano-diamond) TIM ($6 for a lot more) to wet the old particles a tad.

It's worth the 3C drop in temperatures, and we've proven this with some careful measurement, automated data-collection and statistical analysis under controlled conditions -- independently of Joe Citarella (OverClockers.com) and the Innovation Cooling company.

 

[krnmastersgt]

Oh I plan on lapping once I get the time, as it is I was in a hurry to put the system together since I knew I'd face a mountain of software and hardware issues, which I did, I think next weekend I'll have the time to really overclock the system and of course re-mount my cpu cooler, I find the IC Diamond paste intriguing, I wonder if I'd be able to find some at Fry's, in which case I'd grab some for my remounting I might even partially dis-assemble the cpu cooler since it has 2 copper plates with some random thermal compound inbetween and I could probably apply something better at least in terms of heat conductivity. Thanks for the info Duck

 

[BonzaiDuck]

Appreciating your congenial response Krnmastersgt.

Go here:

IC Diamond at Heatsink Factory

I mentioned the JetArt CK4800. By itself, it has performance neck-and-neck on par with AS5, per benchtest reviews. Also worth keeping in the PC Toolbox, for use in recycling the IC Diamond left on the heatsink and IHS if or when you ever need to remove and replace your cooler:

JetArt CK4800 at Sidewinder Computers

And I'll put in a plug for Gary Stoffer at Sidewinder. If you have a serious question or need technical guidance in choosing fans, coolers, water-cooling parts -- he's the man! He will e-mail customers. Good guy.

I don't think I use as much of the IC Diamond as the Heatsink Factory suggests. I spread it on with a razor-blade (credit-card is too soft for this stuff). I put a rice-grain-and-a-half in the center of the IHS, and then pull the razor-blade away from the center toward the four edges. Use a second razor-blade to remove the build-up on the first, and place the residue in the center -- or wherever it is still sparse. I then do the same thing with the heatsink base. By the time you're finished, the "ten minutes" will have expired anyway.

With more sparing usage, you should have enough for two CPU-heatsink applications, and one GPU-heatsink application before the tube is empty.

Meanwhile if anyone wants to buy 2 grams of 0 microns < [average particle size] < 2 microns of diamond powder -- let me know. You can mix it with JetArt, without going through the exercise posted at some forum using silicon grease. @ $20 per gram, I suppose I'd let it go for $10 per gram.

But you're just better off buying the IC Diamond product, and I won't feel over-invested if I have to keep all this diamond powder. . . .

As for lapping the IHS, I'm guessing that it takes between 30 minutes and an hour.

I fully appreciate what it's like to keep a lot of barbells in the air and find time to do this stuff. Ba-leeve me!! I know!

 

[Tempered81]

Originally posted by: Markfw900
Something close to this, I was having a problem with my X3210. Lapped it, it went down several degrees, then I tightened the nuts on my Big Typhoon, and it went down several more !

lol

 

[Thraxen]

Originally posted by: BonzaiDuck
I don't think I use as much of the IC Diamond as the Heatsink Factory suggests. I spread it on with a razor-blade (credit-card is too soft for this stuff). I put a rice-grain-and-a-half in the center of the IHS, and then pull the razor-blade away from the center toward the four edges. Use a second razor-blade to remove the build-up on the first, and place the residue in the center -- or wherever it is still sparse. I then do the same thing with the heatsink base. By the time you're finished, the "ten minutes" will have expired anyway.

I've seen people mention spreading IC Diamond "manually" like this several times. Is this basically just required when using IC Diamond since it's so thick? The generally accepted method of a dab in the middle and allowing the heatsink to spread the TIM just doesn't work with IC Diamond?

 

[BonzaiDuck]

Oh, I'm sure it works to just put a blob in the center, especially since the Heatsink-Factory web-site says to use an amount the size of a pea.

I've probably "concocted" that much of the stuff using the Penn-Scientific diamond powder I bought with JetArt CK4800. The result had the consistency of IC Diamond, and once I'd finished with it, there would have been a pea-sized pile of muck in between the IHS and heatsink-base, squishing out from the sides.

And be mindful -- "squishing out from the sides" is no risk whatever. The stuff is neither conductive nor capacitative -- unless you were to mix it with something like Arctic-Silver 3. I had thought of making a "diamond-silver slurry," but why use particles that have lower thermal conductivity? Why not just use a silicone TIM medium and diamond particles?

The reason I take pains spreading it arises only from certain knowledge of having lapped the base and IHS -- that the surfaces are flat, there are no pockets to fill, and -- why waste the stuff?

Arctic Silver's web-site cites their tests on their silver TIM and says the testing was done with a thickness of 0.001". So I try to make it just a bit thicker than a thousandth of an inch, but if I see any nickel plating through the haze, I want to paint it so the nickel is invisible. So maybe my own applications are 0.01" thick.

 

[krnmastersgt]

The shipping for the IC Diamond is more expensive than the actual solution anyways I rushed the installation of the heatsink so the AS5 I have in there is probably horribly applied but the system works, doesn't overheat and whatnot so it's enough for me till I get the spare time to properly put everything together, including using the 50 or so cable ties I have to make the internals more organized. I'll try and find any IC Diamond products that is locally sold and see if I can save myself some money, if not I guess I'll order it, I paid just as much for the AS5 at Fry's as the IC Diamond with shipping is going to cost me.

 

[aigomorla]

Originally posted by: Idontcare
You can do that? Over-tighten I mean? I just tighten the holy mother living shizzy out of mine and call it a day. Granted I don't break out any extra hardware to do the tightening, still just "finger tight", but actually "backing off" after full-tight is not something I would have thought to try.

I have a quad here that reads nearly 8C temp delta (after lapping and multiple reseats)...I may have to give this "aigo theory #356" a try. Wouldn't be the first time you taught me something new.

not my theory... someone on XS went though 25 remounts trying to perfect the correct mount on a quadcore.

http://www.xtremesystems.org/f...ghlight=quadcore+mount

And it works....

http://i125.photobucket.com/al...p73/aigomorla/Temp.jpg


How do you like those deltas

:]