i5 2500K OC'd to 4.6Ghz, Hyper 212+ Push Pull, High Temperatures

M

MrFoo

Limp Gawd
Joined
Dec 14, 2002
Messages
129
So I've built a new rig:

Asrock Z86 Extreme4 Gen3
8GB Corsair XMS 3 1600 at 1.65V
2x Galaxy Geforce 460 768MB in SLI
i5 2500K at 4.6Ghz using +.005V offset, level 3 LLC
Biostar Nano Diamond TIM
CM Hyper 212+ in Push pull using the included fan as push and the included back case fan from the case mounted to the back of the HSF as pull.
Corsair Carbide 400R Case
The front 120's included in the case, no other fans.
Zalman 850W supply

The case has excellent cable manageability and I have this system the cleanest I've ever built. I can provide pictures upon request.

MB temp as measured in the AXTU software, 35-38C. Idle temps pre and post remount were in parity with the MB temp. AXTU measures Vcore about the same as CPU-Z, within 3mV. At 4.6Ghz Turbo, Vcore reads 1.314V in CPU-Z.

I used IBT, Prime95, and LinX for load testing, then played BF3 at 1080P on High.

IBT gave me outrageous temperatures, one core reached 92C in a couple minutes (the rest 89 or so), so I stopped it.

I respread and checked the mating surfaces and remounted the Hyper 212+ and got the same temperature behavior. No indication of problematic mating surfaces.

LinX gave me temps in the 82C maximum range during a 15min run as measured by the Realtemp Maximum

Prime was lower, around 75C max after 20min.

In BF3 over about 45min of play, I got CPU max at 80C, the top 460 at 80C, the bottom 460 at 76C. The 460's have AS5 on them.

Running another test this morning while I write this, LinX gets the cores up to 75C. with an ambient of about 20-25C in the room and 31C on the Mobo.

I think these temps are crazy. I don't dare run IBT with the 90C+ temps. Do I have something weird going on here or is the 212+ just not cutting it? I considered an H100, but I already spent $500 on the chip, mobo, ram, case, HSF (I already had the rest from my old QX6850 rig). If I have to go RASA, then I'll have to look into it or find a deal on an RS240 or something.

I've read a lot of threads around here and elsewhere on this chip and 212+ performance. I think I am running 20C high at load vs other data and similar Vcore's.

Please let me know if you need more data. Thanks for any assistance.
 
Actually, no. I've only run it overclocked. I can look into that. The surface looked pretty good when I pulled the sink, no high or low spots, I filled the grooves in the HS and then used the spreader tool to basically scrape away the layer on the HS above the level of the grooves. The CPU TIM is translucent after spreading. I applied the HS and used light twisting motion and then screwed it down in the criss-cross pattern.
 
I find those temps a little suspect. That cooler should be able to handle better than that imho. However, if you are going to do 4.6 or more I would highly suggest getting a high end air cooler or a water cooler. For OC'ing in general I recommend watercooling as it is much more stable temp wise.

My first impressions here are you don't have a tight enough seal between the heatsink and the chip. Perhaps either too much compound or not a flat enough surface. You can try lapping the heatsink and trying to get more contact with the chip.
 
MrFoo said:
I think I am running 20C high at load vs other data and similar Vcore's.
Click to expand...

Not sure about the 212+ per se but yeah, you're running high there's no doubt about that.
Hell my 2600K barely goes mid 60s at LinX load @ 4.6 w/1.35 Vcc under a TRUE push/pull using S-Flex SFF21E's (and that's with HT enabled).

It's either the TIM or the cooler. Easy answer - I would try replacing the TIM with something else, since you've already re-applied it once. Try a different kind. If your temps are still high, get another cooler. Grab a TRUE Spirit, those are arguably the best bang-for-buck out there.
 
That's high for sure.

I have a 212+ on my 2500K. 4.4 GHz @ 1.312-1.32 vCore (it fluctuates between the two under load) on an AsRock board. My max temp for a core under prime over 12 hours was 64C and usually that core was 61-62C.

EDIT: I see you are running AxTu. Uninstall it and see if that changes anything. When I had AxTu loaded up it was changing the voltages by itself and running way more voltage through the chip than what I had set in BIOS (I had it set for 1.28 and it was running over 1.4V through as soon as AxTu would launch) . Uninstalling AxTu fuxed the issue.
 
Thanks for the advice so far. I will recheck the flatness of the cooler, clean it out and apply a different TIM. I don't see fluctuation in the Vcore more than what would be realistic if each tool (CPU-Z and AXTU) are polling at slightly different times. I can uninstall it and check. It was handy for me to use to drop my offset from .020 to .005 and raise the multi while running the testing apps in real time while monitoring voltage and temperature.
 
I have heard of chips (not recently) where the heat spreader didn't make good contact with the cores. If you can't get it to xfer well and all else looks good send the chip back. As lapping the CPU will void warrenty (iirc). You can lap the hs though.
 
OK, I left everything else alone except dropping the voltage offset to 0mv and the multi to 33x. I ran 5 cycles of IBT with a max temp of 60C on the hottest core. Idle temp 28C on that same core. Some cores have higher idle (32C) but lower max (54C). I set it back to 46 and 0.005 and back to 91C (stopped after less than 1 run). I have a line on TRUE 120. I considered a RASA RS240, but I'm not really in position to spend $140 on cooling, though the more HSF's I buy, I might have been smarter to do it right once. Maybe an H80 is enough to handle this. Gotta pay to play. :D
 
Last edited:
Just get a TRUE Spirit, there's really no reason to spend more than that IMHO. You might be a few C higher on a different air cooler that costs 2-3+x the TS but the performance is completely respectable. [H] review
I used a TS on my X58's as well as when I had an SR-2 and was completely happy with it. The only reason I'm using a TRUE now is because the TS's were sold with my SR-2 bundle and I had to get something compatible with 1155, and it was faster to run to MC to get an 1155 TR bracket than wait 30+ days for Noctua to send me the free kit for my U12P.

I really don't see any reason to get an H Corsair, a Kuhler, or a custom loop of any sorts.
If a $30 TS doesn't lower your temps, then your cpu has the IHS issue that Stupid_Newbie described and you should try to exchange it (if you still can).
 
Thanks, if I ever hear from the guy I paid for the TRUE 120, then I'll get it in there and test. Otherwise, hope MC takes the chip back.
 
For reference here's my temps when running BF3 using 2600k @ 4.9ghz using hyper 212+ push/pull with low rpm fans...

5201settings.png
 
That's fantastic. I need to pull the sink and take another look. I'll try a razor blade on the IHS and see what it does. The idle in here right now is really low too, so these temps are nuts. I can see 7C spread at idle between core 0 and 1 as well, but that could be single threadedness on my running apps right now focusing on the first core.
 
Ok, so big update. I checked things out and reseated twice this afternoon. IF anything I thought the IHS was slightly convex. Though when I pulled the chip and checked with the razor blade it seemed pretty good to me. Both times I used the spread technique and both times in seconds IBT brought it back to 92-94. So I pulled again, went through the full clean and this time did the 'pea' technique. This time magic happened, IBT got it no higher that 75 after a 5 cycle test. This is the first 5 cycle it's done at 4600, all others were killed by me in 10-15s because of heat. Of course, now that I've got it right I don't want to disassemble it, but someday I have that TRUE Rev3 1366 coming. I also bought a thing of ICY Diamond 7. Currently the sink is mounted with the Biostar Diamond.

Interestingly, LinX on max mem for a 5 cycle test got to 85. So, it is not perfect, but better-ish?

I did a second run of IBT about an hour later and the temps started sucking again...hmm.

The lesson learned is that even with direct heat pipe, the pea technique is the best, apparently.
 
Last edited:
I had really bad results with IC Diamond. Just pick up a tube or two of MX-2 from Mwave for $2.99 a tube with free shipping. It performs well, its cheap and its reliable.

What core voltage does cpu-z report while the chip is under full load in Intel Burn Test?
 
Darn, I just burned $7 for the ICD7 at MC. I remounted again with the ICD7 and am writing this while in the 23rd of 25 cycles of IBT are going. It's currently maxed out at 81C on 2 cores (73,81,81,78), which is not bad. This mount felt pretty good too, no hopping up of the HS when screwing it down or anything. Core voltage never tops 1.344, currently at 1.336 and 80C on the hottest core. I'll likely grab the MX-2 to add to my sack. Somewhere I have older Noctua NT-H1 too.

Update: Played BF3 for a while and it maxed out at 52C, heh.
 
Last edited:
Your temps seems rather high on a 212+. I have a similar experience with the 2500K. First installation, 2500K @ 4.3Ghz, idle was 39c and prime95 was around 92c in 30mins. Took the h/s off and made sure every inch of the cpu was covered with Artic Silver 5. On the second run, idle @ 33c and prime95 max 72c for 24 hours.
 
MrFoo said:
Darn, I just burned $7 for the ICD7 at MC. I remounted again with the ICD7 and am writing this while in the 23rd of 25 cycles of IBT are going. It's currently maxed out at 81C on 2 cores (73,81,81,78), which is not bad. This mount felt pretty good too, no hopping up of the HS when screwing it down or anything. Core voltage never tops 1.344, currently at 1.336 and 80C on the hottest core. I'll likely grab the MX-2 to add to my sack. Somewhere I have older Noctua NT-H1 too.

Update: Played BF3 for a while and it maxed out at 52C, heh.
Click to expand...

If you already have IC diamond you might as well use it. Some people have good results with it, I wasn't one of them though but I was using the super thick 24 caret. I like MX-2 and I think that $3 shipped is a good price.

After five runs of IBT here is what Realtemp said my max core temps were with a 2500k at 4.6ghz with a Venomous X and two S-flex fans. I would imagine that after 25 runs temps would be a bit higher too. IDK, take what people on the forums say with a grain of salt. I've seen some pretty outrageous claims.

Venomous X
63 70 72 68
Click to expand...

alanon said:
Your temps seems rather high on a 212+. I have a similar experience with the 2500K. First installation, 2500K @ 4.3Ghz, idle was 39c and prime95 was around 92c in 30mins. Took the h/s off and made sure every inch of the cpu was covered with Artic Silver 5. On the second run, idle @ 33c and prime95 max 72c for 24 hours.
Click to expand...

There is a big difference between Prime 95 and Linpack with AVX support. Try Intel burn test and watch your temps skyrocket.

Prime blend is the better test for sandybridge anyways.
 
What's weird is that I got the high temp behavior on IBT without rebooting, the 5cycle to 75 was great, then a couple hours or so later, suddenly it was much higher again. I ran another 5 now after playing some BF3. Idle was about 31-34 across the cores. I am getting 91C max now. Really weird. Either my 212+ is really uneven (though it seemed pretty flat to me with the blade test) or something just isn't right in there. There is some noticeable discrepancy between HWmonitor and Real Temp's measurements. My GPU's are 50C and 39C respectively, at idle when I ran the IBT. So Prime blend is better? I think I get much lower temps there than LinX or IBT.

Edit: Hmm, maybe it was just too soon after BF3. Now the GPU's are at 45,35.
 
Last edited:
Either a bad chip or the mount is still not right. I have a 212+ in P/P with the HSF fan pulling and an Antec Tricool at medium pushing. 45 minutes of P95 Max Heat running 4.7Ghz and the cores are sitting at 65/71/69/68. Last idle was 37/38/35/37.

Unrelated, but I think I got a gem of a 2500k. Currently running [email protected].
 
Interesting that you note the voltage. Since I am running .1V greater, that could help to explain the temp difference, though 100Mhz slower. Though I still think the mount is odd. Also, the push pull might be an issue, I think my fans have a several hundred RPM difference. I put the faster fan on the pull side, perhaps that will help. I am running the P95 max heat test now and will edit with results.

Edit: Ran 59m of this test, max 71,76,76,73. Avg at max load was around 73. Idle after stopping the tests 36,30,32,34

For fun I am trying at 4700 to see what happens in the maxc heat test. Maybe it will even be stable. :D

Is it reasonable that some of my trouble could be that the older config had the push fan 800RPM higher than the pull? I swapped them now.
 
Last edited:
Possibly. I'm no aerodynamics guru or anything, but maybe the slower speed fan wasn't pulling the hot air out of the HSF fast enough and letting it linger?
 
adurn03 said:
Either a bad chip or the mount is still not right. I have a 212+ in P/P with the HSF fan pulling and an Antec Tricool at medium pushing. 45 minutes of P95 Max Heat running 4.7Ghz and the cores are sitting at 65/71/69/68. Last idle was 37/38/35/37.

Unrelated, but I think I got a gem of a 2500k. Currently running [email protected].
Click to expand...

Try Intel Burn Test and watch your temps skyrocket.
 
Sorry if this is thread necromancy.

Update:

I replaced my hyper 212+ with a TRUE 120 Rev C with the CU base, the Thermalright branded 1600RPM pushing and a 1600RPM S-FLEX pulling. Ran 5 runs of LinX on max and max temps were 72C on one core (20min run). Very nice. If I read my earlier results right, this is a drop of ~15C over the 212+. This run was at 4700Mhz with no offset boost. For full stability, I may have to bump to +0.005V. I had to purchase the 1156 Bolt-Thru Rev 2 kit ($13 at my Microcenter) to mount it.
 
You must log in or register to reply here.
Back
Top