New core i7 hot enough to boil water!

S

schizo

[H]ard|Gawd
Joined
Nov 6, 2004
Messages
1,689
:confused:So I just put together my new rig; P6T, 3GB corsair DDR3-1600, i7 920, and a noctua nh-u12p se1366. Everything goes together fine, I use the grain of rice technique for the noctua paste and get the HSF on tight with both fans blowing towards the back of my P182 case at full speeds. I get vista x64 installed, etc, start overclocking, and *BAM* I hit a brick wall. At default vCPU (1.2v) and a smidgen over default for vDIMM, at only 3100Mhz CPU (155Mhz BCLK) and around 1400Mhz RAM, my 920 hits 100C (212F!) after 2 minutes of 8 threads of prime95. The CPU almost immediately hits its thermal threshold and starts downclocking and dropping voltage, eventually hitting 2.0Ghz or so. Prime95 never throws a calculation error.

I've never seen temps like this before on any computer, ever, and it's got me a bit spooked. I've never had a CPU hit its thermal threshold before. In the past, when I overclocked too aggressively Prime95 threw errors and I knew to back off. I watched Kyle's video and expected to see something like 65-75C at full load on air at default voltages and my relatively conservative overclock. Maybe 85-90C at 3.6+Ghz and 1.35vCPU, but not at 3.1Ghz and 1.2vCPU. The CPU idles at maybe 45-55C, which I thought looked about right for an i7.

What's going on here? Is it something silly like a badly mounted HSF or do I just have unrealistic expectations?
 
A little bit of both I reckon.
Try to remount your hsf again.
Remember 80-85C on air cooling is quite common. Remember Kyle tested his air cooling in an open bench setup so your temps would be higher in a case. In my water setup, 1 of the core hits 78C in prime (I think the one farthest from the inlet :()
This is at 3.9GHz and 1.3 Vcore.
 
That's pretty much what I thought. I'll give it another shot when I get home from work. Thanks guys.
 
Yup definitely your heatsink, no way you should be hitting those temps at only 3.1.
 
I dont want to start a flame war but I was never happy about that grain of rice deal. I start off with a blob that size and with my finger inside a baggie (to keep paste clean and my finger ungreased) spread it out over the IHS in a thin layer and almost always end up adding a bit more. Thats a big IHS, without going overboard try a bit more paste.

Hmm LOL I cant figure out how those sping loaded screws work other than it appears you have to take the fans off to adjust them. An extra turn for more cllamping force might be in order. No clue what the installation instructions say but I would carefully and evenly (go back and forth between the two screws 1/4 turn at a time, and tighten unil a good wiggle of the heatsink gave absolutely no movement at all. Be careful better too little than too much and break something.
 
I used to do the baggie finger method too. The noctua TIM is particularly "stiff" and prevailing wisdom is grain of rice, so that's what I did.

You do have to take the fans off to adjust/remount the heatsink. The noctua mounting mechanism is the easiest I've ever seen (although I apparently managed to mess it up, so take with salt) but getting those little fan clips back on is a real pain in the butt. Not looking forward to it as my hands are already shredded hamburger from the build last night.
 
BillParrish said:
I dont want to start a flame war but I was never happy about that grain of rice deal. I start off with a blob that size and with my finger inside a baggie (to keep paste clean and my finger ungreased) spread it out over the IHS in a thin layer and almost always end up adding a bit more. Thats a big IHS, without going overboard try a bit more paste.

Hmm LOL I cant figure out how those sping loaded screws work other than it appears you have to take the fans off to adjust them. An extra turn for more cllamping force might be in order. No clue what the installation instructions say but I would carefully and evenly (go back and forth between the two screws 1/4 turn at a time, and tighten unil a good wiggle of the heatsink gave absolutely no movement at all. Be careful better too little than too much and break something.
Click to expand...

For a quad core Intel, you put a line of paste across both chip modules. Then clamp your heatsink on and twist back and forth a degree or two. You don't have to spread it out, since putting paste the very edges of the IHS is pretty much a waste since nearly all of the heat is conductive directly above the cores.
 
Yeah I used to do the "grain of rice/line of paste" thing and I never noticed a difference either way when compared to spreading it over the entire cpu. I now use the Zalman STG thermal paste (with the applicator brush). I just use the brush to spread out a thin layer over the entire cpu...never have had a problem with high temps using this method. I'm installing a q9550/Sunbeam Core Contact Freezer this weekend, and I plan to do the same thing.
 
Yeah I kinda agree with the too much is better than too little argument. I mean, if I put too much chicken of the sea brand dolphin safe tuna in my sandwich and press down on the bread to make it a manageable size to eat, the excess tuna will just ooze out the sides. If I put too little though I am risking air bubbles devoid of flavor.

Don't go overboard obviously.
 
Just an update; incorrect HSF install was indeed the problem. Actually the paste didn't cover the entire top of the heatsink face. Adding another little half grain of rice fixed it. I'm now hitting around 75C at 100% load at 3.1Ghz. While that still feels really high compared to C2Ds, it's not insane for air cooling with a closed case. I'm going to see how high I can go before toping off at 85C at full load, then I'll probably back it off one step and stop there.
 
schizo said:
Just an update; incorrect HSF install was indeed the problem. Actually the paste didn't cover the entire top of the heatsink face. Adding another little half grain of rice fixed it. I'm now hitting around 75C at 100% load at 3.1Ghz. While that still feels really high compared to C2Ds, it's not insane for air cooling with a closed case. I'm going to see how high I can go before toping off at 85C at full load, then I'll probably back it off one step and stop there.
Click to expand...


lol, just take a look at all the other i7 overclocking threads in here.. and you will relize how low 78C in an i7 really is.. :p
 
Oh I read all of em. 75C still feels fairly high. The i7 is the first consumer-level CPU that really screams for watercooling, IMO.

Try as I might, I couldn't quiiiite make a solid 3.6Ghz overclock; 1.250 vCPU bluescreened and 1.275 vCPU hit 90C (my self-imposed temp limit). I'm currently burning in 3.5Ghz at 1.250 vCPU sitting at 84C with 8 prime95 threads and OCCT running. Looks like that'll be where I stop. Not too I'm complaining, with turbo mode at 3675Mhz that's a 38% OC and that ain't bad.

This noctua paste isn't supposed to cure like AS, but I still have hopes I'll drop another degree or two.
 
give some time for the grease to cure.. probably by next week or so you should be able to start working on higher temps.. even though im sure some one will say im full of crap blah blah blah.. a week does make a difference.. it has with all of my systems ive overclocked.. only other thing id suggest is changing the fans with a higher CFM.. ive seen a few reviews with this heatsink where people have swapped the fans out and gained an extra 4-5C cooling.. though personally i find the noctua 120mm 1300rpm vortex fan to do a better job on that heatsink then the stock fan that comes with it..
 
schizo said:
Oh I read all of em. 75C still feels fairly high. The i7 is the first consumer-level CPU that really screams for watercooling, IMO
Click to expand...
You're still not mounting it right.

Draw 1 line of paste across the chip, covering approximately 3/4 of the total width. You can also draw 2 smaller lines(not shorter) that run parallel to each other.
 
I put on as much as I could without it spilling off the side of the IHS.

I have the same setup as you, btw.

And I've heard that Intel has no idea what the real tjmax is on these chips, and 100 was just a number. If you look at the processor spec sheet for Core i7, you'll see the maximum temp is missing from the Core i7 specs.

Considering this is a tock, it's still on the same 45nm process as the core 2s, and the highest max temp I saw for a 45nm c2q was 71.4.

All looks kind of sketchy to me. I'm trying to keep my temps sub 70 until someone knows for sure. That means no fully loading the CPU.
 
I'm pretty sure I have it correctly mounted now. I have an even distribution of paste across the top of the IHS, and my temps dropped dramatically after reseating it.

Latest update, I seem to be able to make 3500Mhz (3675 w/ turbo) at stock voltage of 1.20 vCPU! At full load for >1 hour at 3500Mhz my CPU is topping out at 78F, according to both coretemp and realtemp. I guess my CPU just needs a ton more voltage to stably run at 3600Mhz, and sadly running at 1.30 vCPU causes me to hit a whopping 90C. I'm pretty happy with this overclock on air, and the i7 920 overall. It's clear that this CPU has a ton more headroom if you have the cooling to match. Air doesn't cut it.

100C is definitely the tjmax for my chip. When I had the HSF mounted incorrectly, the CPU would immediately throttle when it hit 100C. Based upon nothing but reading a crapload of forums over the past week or two, you basically just want to keep the CPU below 85-90C at 100% sustained load.
 
Yea, heat is the limiting factor in most of the oc. Intel specs the max vcore to 1.55v but that means insane temps :(
 
Intel knows exactly what the real TJmax is on these chips, since they programmed it in.
It's not going to hurt the CPU to have it hot, I doubt it'll even shorten the life a reasonable amount.
These don't actually produce that much more heat than say a QX6850 but they have a far smaller area than the Qxxx0 series processors. Mine on stock cooling hits ~ 75-79ºC under full prime load @ 3.2GHz, 1.150 Vcore.
 
Pretty impressive that you managed to overclock while undervolting. But I guess it's comparable to my much larger overclock at stock voltage.

I've been rolling my own computers and overclocking since the pentium1 days, and the core i7 is the first CPU I've ever owned that was actually temperature limited. Sure, temps always mattered, and watercooling would have eked out a bit more speed, but I always managed to hit rapidly diminishing returns with just air cooling, and I always hit a point where the CPU started throwing errors in stress tests. I can't make the core i7 fail prime95 no matter what I do. I get the feeling that it has a HUGE speed margin that's unreachable with air cooling. And that's a first, for me.

The next nehalem release on a smaller process should be very interesting.
 
What i do is put a very small dot in the center of the chip, then spread it out for a very thin layer. My Q6600 at 3.2 under full load does not get above 50C
 
Thats the old way of doing it. Groupthink is that doing it that way has a chance of creating tiny insulating air bubbles.

Really I doubt it matters as long as you have coverage everywhere and it isn't too thick.
 
Gotta remember when applying the 1 grain method that you "pre-coat" the heatsink surface and the IHS on the chip with some thermal paste, ie just enough to change the color of the metal to know its in the cracks and fine imperfections. Then you use the 1 grain of rice method across the cores (when I think grain, I think a bit bigger than typical grain of rice).

q6600 still humming along at 3.6 closed case antec P160 with all 1500rpm fans. ie a grand total of 3. I suspect that if I were to go i7 then I could probably go with the TRUE still but hit the thermal wall of the i7 quick due to the damn heat its outputting.
 
schizo said:
Latest update, I seem to be able to make 3500Mhz (3675 w/ turbo) at stock voltage of 1.20 vCPU! At full load for >1 hour at 3500Mhz my CPU is topping out at 78F, according to both coretemp and realtemp.
Click to expand...



Now that you've got your cooling problems solved, try core damage and see if you still make 100°


I'm running at 3.2ghz on the stock intel heatsink & fan. I've got the voltage set to 1.1 and I'll hit TJ max in about 7 minutes.
 
I've got mine running at 3.9GHz and max temp on it during burntest was 75c. This morning it's been running around 37c-43c just messing around on the internet.
14.jpg
 
just read this and had posted another for benchmarking. my intel stcok heat sink and prime95 are producing 95 - 100 degrees. just built it this weekend (first timer).

I used artic silver5 (line across the processor method), sat the heat sink, twisted back and forth a very little bit and then hooked the latches I checked below the Motherboard, and it seemed all four latches took.

idle = low 40s
normal load (internet) = 40 - 50 degrees
Prime 95 = immediate jump to 95 - 100 degrees no crash after 15-20 minutes, but running hot

Played Oblivion as well for 30 minutes without issue.

Can I reseat the heatsink without pulling it off completely and having to clean the chip and heatsink again?

Does artic silver 5 just take time to cure?
 
I am using TRUE and got mine to 3.6Ghz last night with stock voltages... Although it crashed this morning, it was running decent with Prime 95 and Far Cry 2 while temperature was around 67-72C per cores.

I put about three grains of rice worth of AS5. I'm gonna go home tonight and try increasing voltage and hopefully it should run more stable.
 
koss said:
I am using TRUE and got mine to 3.6Ghz last night with stock voltages... Although it crashed this morning, it was running decent with Prime 95 and Far Cry 2 while temperature was around 67-72C per cores.

I put about three grains of rice worth of AS5. I'm gonna go home tonight and try increasing voltage and hopefully it should run more stable.
Click to expand...

That might be a bit too much AS5. I just put a drop or so, smear it around with my heatsink and call it a day xD
 
Good question about the intel stock heatsink. It uses that wax instead of normal thermal interface paste. Not sure if it's OK to reseat it or not. Your temps are probably about right for the stock HSF-- remember the core i7 runs hot. Even if you hit tjmax at 100C you won't crash. The corei7 simply doesn't crash or give errors due to temps. It very smoothly throttles back speed and voltages instead.

AS5 does take a week or so to cure. And yes probably core damage and intelburntest will hit tjmax, but those aren't representative of "normal" 100% CPU loads. IMO 8 threads of prime95 burned in for >8 hours is a much better stability measurement.

Now that I've got my cooling fully worked out, I'm sure I could hit 3.8+ by pumping up the voltage but frankly I'm pretty happy with 3.6Ghz and my very reasonable load temps. I don't do a lot of video conversion and it would take 2 GTX280s to bottleneck a 3.6Ghz corei7 in games. And I don't even have one!
 
below is the post from the other board. I may try to reseat the heat sink anyway. I stripped the stock wax in favor of artic silver 5. Thanks for the heat advice. any thing else you can assist with. None of the below has so far caused my PC to crash.

QUOTE=Meat1000;1033456567]Built a new i7 machine this weekend (first timer). I am running a few benchmarks and am having issues with two. I read a forum in the CPU board but it mentioned a hotfix which I have. So here is the issue. Any help is appreciated:

PC Mark05 - runs 7 of 11 tests. Completes, but no score, not sure why some tests are failing.

3DMark Vantage - runs 3 of 4 tests, stalls on the 4 test which tests the CPU for the second time. This is the Physix test.

Note: 3D Mark 06 ran fine and scored 14500 however, may not be related

Prime95 - runs along with RTHDribL, hasn't crashed, but I haven't gone more then twenty. But the real issue is the CoreTemp test which shows the CPU maxed at 100% and the temperature of the cores at 95-100 degrees, which is worrisome.

Side Note: 120mm Fans: Any thoughts on the extra two fan placements for making the case cooler?

Last Side note: I played Oblivion for 30 minutes no problem and averaged 60 FPS even outside while battling

System Info:

Intel i7 920 - stock heat sink
asus p6t deluxe
WD 6400 aaks 640mb
evga gtx260 216 core superclocked
bfg 800 watt power
3 x 1 gb corsair xmsr3 ddr3 1333
cooler master rc 590 case
acer 22"
samsung 22x dvd burner
2 extra 120mm fans (top vents blowing out of the machine)
Saitek Eclipse Keyboard
Vista 64 bit[/QUOTE]
 
will long exposure to high core temperatures cause long term damage to the chip?

i'll run memtest 86 tonight and then retest prime95 after reseating the heatsink. Since I am using AS5, do I have to clean the chip and all again if I move it? or can you reuse since it is a grease?
 
Oh that other post was you too. You shouldn't leave it that hot for extended periods of time. Basically you want it <90C at all times; 85C ideally. You can reuse the AS5.
 
I was reading these posts again. I've got my i7 920 running at 3.9GHz. All four cores are idling around 33c-37c. DRAM freq. 585.0MHz. It has passed the intel burn test set on max stress at ten tests several times. I won't be doing this again don't want to kill the damn thing. I've also ran the 3D Mark bench several times also with no problems. IMOP this chip is rock solid at 3.9GHz. I can't get 4.0GHz on air though. Won't pass a burn test. it crashes after 5 runs. Any questions please ask!
 
Forgot to add the max temp I saw durning all testing was 81c. I brought down the DRAM volts. I was seeing around 84c-87c before I lowered the freq. but still passed all tests at the higher freq. just ran hotter.
 
Reading your guys post's I think I really lucked out because under max load with a 1.32 vcore I hit 58-60oC max. My idle temps are at around 27-29. This is with a room temperature of around 21 to 22.
 
You must log in or register to reply here.
Back
Top