Help with High Temps (stock performance)

Status
Not open for further replies.

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
Hi All,

I recently put together my first build (with help from my brother) and noticed high temps on both stock and OC. I don't know if it's due to poor airflow, location of system, or incorrect application of TIM and HSF. Or maybe it's all of these.

Room Temp: 26-27'c

Case: Raven 2 Evolution (RV02W-EW)

MoBo: Asus P8P67 Pro (REV 3.0) [BIOS v.1704]
CPU: i5 2500k (currently running stock)
HSF: Scythe Mugen 2 Rev.B (2 fans push/pull config)
--> used stock thermal paste (Thought it should be sufficient for low/moderate OC [40x-42x multiplier])
RAM: G.Skill RipjawsX DDR3-1600 (7-8-7-24-2T undervolted @ 1.55V [from 1.6V])
GPU: EVGA GTX570 Reference
PSU: Antec TP-750

4 HDDs (3x Seagate Barracudas [SATA1] and 1x Western Digital Caviar Blue [SATA3])
1 SDD (OCZ Agility 3 60GB boot drive)
1 BD ROM (Samsung)
1 DVD burner (Samsung)

System is sitting in an open cabinet under computer desk. It is open with no back panels or doors.

When booting to BIOS, temp shows 41'c (think it should be in the 30s?)

Windows using Aida64:

Idling shows:

MoBo: 29'c
CPU: 39'c
Core 1: 43'c
Core 2: 44'c
Core 3: 42'c
Core 4: 40'c
GPU: 38'c

But when idling (<3%) it's constantly bouncing between 1600 and 3700mHz.

Could be because of background programs (Steam, Online Armor, ESET Nod32, Webroot Spysweeper, Creative Audio, NVIDIA control panel).

When running Prime95 Blend test:
MoBo: 29 31 30.1
CPU: 39 58 47.5
Core1: 43 84 62.9
Core2: 44 88 66.0
Core3: 42 88 65.2
Core4: 40 84 61.8
GPU: 37 46 38.7

Fans:
CPU: 433 1387 907 (PWM Push Fan)
Chs1: 1610 2139 1852 (PWM Exhaust Fan)
Chs2: ~1800-1850 constant (3-pin Pull Fan)

I have the push/pull config as follows.

Low CFM PWM-Fan Push (900RPM)>HS>High CFM Non-PWM Fan Pull (1900RPM)>High CFM PWM-Fan Exhaust (1900RPM) -- equal RPM/CFM to pull fan

The AP fans on the bottom of case are set to low. Switching to high didn't make any difference to the temps.

When running blend test I noticed cores 1/4 and 2/3 match in temps with 1/4 running a bit cooler than 2/3. Indication I did something wrong?

Prime95 stable for a few hours (no errors). Except for the OCZ SSD crash/stall that occurred during the run.

Any ideas of what I did wrong? What can I do to improve temps if possible?
 
Solution
Assuming your temps stay the same after you re-install the cooler, I would recommend staying at that speed. You're reaching the temp threshold of what I recommend for a long-term everyday overclock.

To keep Sandy Bridge happy for a long-term overclock:
1. Keep peak Prime95 core temps below 75ºC
2. Keep CPU voltage below 1.4v

If you want to go higher safely, you will need a better cooler.
If I read the post correctly, the case is closed but the cabinet the case is in doesn't have a back panel or front door.

Just to be clear, the temps in AIDA when running Prime95 read as Current -- Peak -- Average ... is that right? It should never be getting that high at stock speed with an after-market cooler. It shouldn't even get that high with the stock cooler. Are you certain it's running at stock speed? ASUS boards have a tendency to auto-overclock if you aren't careful with the BIOS settings. Also, what is the CPU voltage reading under load?
 

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
The readings in Aida64 are Min - Max - Avg.

I just hit the F5 'Optimized Default' and changed DIMM settings. Timings @ 7-8-7-24-2T. DIMM speed @ 1600. DIMM voltage @ 1.55v.

I also enabled EPU. Disabled PLL overvoltage, CPU spread spectrum. Fans set to silent and warnings @ 200 rpms (Otherwise I get a CPU fan warning after cold boot each time). Everything else is at default.

Temps shot as high as 88'c for cores 2/3 and 83-84'c for cores 1/4 if I recall. I think CPU voltage spiked to ~1.3 and speed @ 3.7. Idling brings down to about 1.2 and jumping at 1.6-3.7. I'll post pics of the 7 hour blend test I performed overnight when I'm back home.
 

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
Okay, here's a screen of P95 blend test 7 hours running:

P95BlendTest-1.jpg


And P95 blend test post:

PostP95BlendTest-1.jpg


Temps are high and I only 'tweaked' the things I've listed and left everything else alone. I've been thinking that maybe the HSF isn't seated properly or the thermal paste is not fully covering the cpu/hsf (I used the pea method, but a very small pea size). Is that very possible?

Edit: reversed the pics and resized for better view
 
Yeah, it's just as I thought -- the 'optimized defaults' doesn't actually use default settings. It auto-overclocks the CPU. Your CPU is running at a non-stock speed of 3.7GHz using around 1.3v under full load. That's why your temps are high.

You're going to have to clear the BIOS using the jumper on the board. The manual will have its location. It may be called Clear CMOS or something like that. That will change the settings back to the defaults, giving you stock speeds back.

If you do ever overclock, you should be able to get 3.7GHz using stock voltage (which is 1.2v). 1.3v isn't usually needed until around 4.4GHz or so.
 

br4m

Distinguished
Dec 1, 2010
11
0
18,510
The push and pull fans are supposed to have the same RPM. When you're idling you're temps dont matter, unless they're above 50C. Try using the stock Mugen II fan in 20 minutes of prime, temps are supposed to be alot lower then.
 

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
@Leaps: Thanks for the info. I reset the CMOS and ran tests with br4m's suggestion.

Thanks br4m. 1 fan actually performed better than what I thought would have worked... Well, I knew I did something wrong, just couldn't figure it out. Newb mistake.

Here were the tests with a CMOS reset at ~20 mins:

P95 Blend Push/Pull mistake:
P95BlendClearCMOS20m.jpg


Post Blend Push/Pull mistake:
PostP95BlendClearCMOS20m.jpg


P95 Blend with the low RPM fan push
P95BlendClearCMOS20m1F.jpg


Post Blend with the low RPM fan push
PostP95BlendClearCMOS20m1F.jpg


P95 Blend with the high RPM fan push
P95BlendClearCMOS20m1FHRPM.jpg


Post Blend with the high RPM fan push
PostP95BlendClearCMOS20m1FHRPM.jpg


Edit: so stock temps are fine for this cooler, or is it a bit on the high side? Should I try to reseat the HSF? I was planning a modest OC of 40-42x multi. But looking at these temps currently, I think I still need to make some adjustments...
 

Kralnor

Distinguished
Jun 13, 2009
122
0
18,690
Your temps seem high for stock settings.

I have the same CPU and HSF, although I use an R3, which is not the best case in terms of airflow. It only has two case fans - one front for intake and one rear for exhaust. I only use the stock Scythe fan on the heatsink in a push configuration.

Under ideal conditions (front of the case open, all fans blowing at full) Prime95 maxes out at 55C on the hottest core during blend testing. This is with an ambient temp of about 24C and at full stock settings.
 

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
Thanks for your repsonse Kralnor.

In all honesty, I thought maybe I should be getting at least 30'-35'c idle and 60'-65'c on full load with the stability blend tests. At least with just stock default settings and no OC.

Well, I just played a little with OC and see how temps would be at 4.0GHz with manual voltage setting @ 1.195v and got the following results.

P95 Blend for ~1 hour
P95Blend4ghz1_195v1h.jpg


Post P95 Blend
PostP95Blend4ghz1_195v1h.jpg


I was actually expecting BSOD or an error or really high temps. But pleasantly surprised with the result. I know it's too soon to say it's rock solid stable at this setting, but I was just running more tests and seeing what I could do.

I'll try to reseat the HSF tonight and see if that was another newb mistake I made. Not looking forward to it as it was a pain to install in the first place. But I think I figured out some method which might make the install a little easier. Live and learn...
 
Assuming your temps stay the same after you re-install the cooler, I would recommend staying at that speed. You're reaching the temp threshold of what I recommend for a long-term everyday overclock.

To keep Sandy Bridge happy for a long-term overclock:
1. Keep peak Prime95 core temps below 75ºC
2. Keep CPU voltage below 1.4v

If you want to go higher safely, you will need a better cooler.
 
Solution

neograndizer

Distinguished
Sep 10, 2010
179
0
18,710
Yay! Success! It was my damn newb experience. When I took out the mobo, I noticed the back plate was on an angle and NOT flat. Apparently, my brother didn't screw it in 'tight' enough (I applied the paste; he screwed it on). The thermal paste covered the entire cpu, but it was still wet and oozy... A little difficult to remove (making sure it doesn't go all over the place when cleaning), but with a little patience, it came off. I applied some Noctua NT-H1 TIM with the instructions provided on the back.

I reseated the HSF and reconnected all the cables. Put it back in the cabinet, did a little prayer and pushed the power button. The fans amazingly enough, came on very silent (before it was audibly loud). I went directly into BIOS and rechecked settings (especially with SATA as I redid where the cables went into the SATA port). The BIOS showed CPU @ 30'c. Cool. It looks like it worked. Now to save settings and load into Win7.

Now I'm in Windows and doing a P95 blend test. Well, here are the results currently:

P95BlendHSFReseat.jpg


I think I can push my original OC target of 42 multi, provided the 40x multi lasts overnight test. That way I know I have a stable speed I can revert to if I can't reach target.

Edit: I'll make sure I post some pics of my rig. I'm damn happy and proud that I finally completed my build. I'll probably post my experience about it too...

Thanks again everyone!
 
Status
Not open for further replies.