Core i7 920 stock speed running hot

Toggle sidebar Toggle sidebar
T

TaMaa

Junior Member
Sep 9, 2009
13
0
0
Hallo, I have a core i7 920 which is getting very hot, in idle it''s about 46 C on each core, and running prime95 on 8 threads it reaches 70 C in seconds, and keeps on rising till it hits about 85 C, which I think is extremely hot for a CPU.
If I set my ram on stock speeds (1600mhz on 1.65V) then the CPU gets a 5-10 C hotter.
Ambient temperatures are around 25 C.
In prime95 i sometimes get an error on a core after just a minute or so, and it stops the test on that core. But I have had it running over night without errors.
Ram is tested with memtest86+ over night without errors.

Is something wrong with the CPU, or are these temperatures normal for a core i7?
I have tried refitting the CPU cooler 10 times or so, with 3 different thermal paste, Noctua, AS5 and Zalman.
I checked the CPU socket, to make sure the CPU was higher than the socket brackets.
Any help, advice or good suggestions would be much appreciated =S

My specs are:
Antec p182 with 4 Noctua NF-P12 120mm fans running on lownoise ~1200rpm.
Asus Ramgape II GENE motherboard with everything set on auto in bios.
Intel Core i7 920 D0 stepping running stock speeds with turbo enabled, aircooled with Noctua NH-U12P push-pull on lownoise ~1200rpm.
6gb OCZ Reaper 1600MHz running 1066 MHZ 1.5V
Gainward GeForce GTX 275 with 2 fans
 
V

v8envy

Platinum Member
Sep 7, 2002
2,720
0
0
I get roughly those results with the stock heatsink at about 5400 feet of altittude (read: ~17% less air to cool with), although I've never received an error from prime95.

Try again with the stock heatsink, try rotating your heatsink 90 and 180 degrees, etc etc. If all that fails point a high flow (but more importantly high velocity fan) angled toward the heatsink. I used motherbord offsets to angle my fan and make sure my fan wasn't simply recirculating a column of hot air over the sink. That trick got me to the point where I'm satisfied with my CPU temps. They still reach mid 70s (50C over ambient) very quickly but at least they stop climbing to 100 and beyond.

Undervolting and disabling hyperthreading also worked to reduce heat, but seemed to turn off power management. I can run at 1.05vcore in perpetual turbo mode (3.3 ghz), or I can leave things on auto and have the board manage .92 to 1.22 volts as needed for 1.6 to 2.8 ghz.
 
A

alyarb

Platinum Member
Jan 25, 2009
2,444
0
76
if you don't plan on overclocking, then definitely undervolt a little, and you need to make sure that fresh air is constantly flowing past/over/through the heatsink. having a lot of fans doesn't mean you have a lot of effective airflow.
 
Y

YBS1

Golden Member
May 14, 2000
1,945
129
106
Whooa! That's crazy high for stock speeds. That's slightly above the temps I'm seeing when running at 4.2GHz and I'm even a little concerned that I've got a cooling/airflow/mounting issue causing mine to be a bit on the high side which I need to correct.
 
T

TaMaa

Junior Member
Sep 9, 2009
13
0
0
When I was going to put the stock heat sink on, I noticed that the zalman thermal paste wasn't all over the noctua heat sink, which it should be because I put the zalman thermal paste all over the CPU with the included "brush"

Picture of the CPU with zalman thermal paste after removing the noctua heat sink: Link
And of the noctua heatsink: Link

I am currently running with the stock heat sink, and it's idling at around 42 C.
Running prime95 it gets to ~67 C in seconds and after 5 minutes it gets to ~75 C, I still get the error in prime95 after a few seconds sometimes.
[Sep 10 15:48] Worker starting
[Sep 10 15:48] Setting affinity to run worker on logical CPU #5
[Sep 10 15:48] Beginning a continuous self-test to check your computer.
[Sep 10 15:48] Please read stress.txt. Choose Test/Stop to end this test.
[Sep 10 15:48] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Sep 10 15:50] FATAL ERROR: Rounding was 0.5, expected less than 0.4
[Sep 10 15:50] Hardware failure detected, consult stress.txt file.
[Sep 10 15:50] Torture Test ran 1 minutes - 1 errors, 0 warnings.
[Sep 10 15:50] Worker stopped.

Should I return the noctua heat sink?
And is my CPU damaged or something. I do also get the occasional crash, where everything just freezes and I then have to reboot.
 
A

ArchAngel777

Diamond Member
Dec 24, 2000
5,223
61
91
Mine hits 85C. I don't care though, the chip can handle it and NOTHING will tax my system like P95 does on all 8 threads.
 
E

EliteSentinel

Member
Sep 9, 2009
28
0
0
u should care, 85C isn't ideal for computing, get it down to at least 76c if u don't want to see your investments blow up ^
 
T

TaMaa

Junior Member
Sep 9, 2009
13
0
0
Has anyone ever gotten a Hardware failure in prime95, and if so, what was the reason?
 
GLeeM

GLeeM

Elite Member
Apr 2, 2004
7,199
128
106
Originally posted by: TaMaa
Picture of the CPU with zalman thermal paste after removing the noctua heat sink: Link
And of the noctua heatsink: Link
Click to expand...

The pictures show that you have a hump in the middle of either the heatsink base or the top of the CPU.
You can see a capital "I" shape (or laying down "H" shape) and that the heatsink and CPU are not touching on either side in the middle!

Was the heatsink tightened down really good? If it was you might want to search the forums for "lapping" :(
 
A

ArchAngel777

Diamond Member
Dec 24, 2000
5,223
61
91
Originally posted by: EliteSentinel
u should care, 85C isn't ideal for computing, get it down to at least 76c if u don't want to see your investments blow up ^
Click to expand...

I hope you are joking. 85C isn't the temperature that will be 24/7. In fact, there is no application that I run that would heat my system up to that temperature. It is only during a torture test. Your also kidding yourself if you think 85C is going to 'kill' my investment. I hate to break it to you, but my chip will be worth pennies on the dollar when it fails and will have lived a long life past its usefullness.
 
V

v8envy

Platinum Member
Sep 7, 2002
2,720
0
0
85C sounds suspiciously even, and may be due to the CPU thermal throttling. If you use RealTemp it will report whether or not thermal (or max power draw?) throttling is in progress. It reads this info right from the CPU (much like the i7z program for linux), and is independent of whatever right or wrong numbers are being displayed for the core temperature.
 
A

AstroGuardian

Senior member
May 8, 2006
844
0
0
Originally posted by: TaMaa
When I was going to put the stock heat sink on, I noticed that the zalman thermal paste wasn't all over the noctua heat sink, which it should be because I put the zalman thermal paste all over the CPU with the included "brush"

Picture of the CPU with zalman thermal paste after removing the noctua heat sink: Link
And of the noctua heatsink: Link


Should I return the noctua heat sink?
And is my CPU damaged or something. I do also get the occasional crash, where everything just freezes and I then have to reboot.
Click to expand...

Dude are you blind? That's why you get high temps. Your heat sink does not take away the heat from the cpu. You definitely have a surfaces contact problem. Try lapping your cpu/heatsink. Here is a guide: http://forum.overclock3d.net/showthread.php?t=4735
When i did that successfully my E7300 @ 4GHz dropped temps at least 15 degrees. I wanted to write lapping guide myself but didn't have the time. Anyways it functions great. Note that you are losing CPU warranty if you lap it. So try lapping your heatsink first and check the results. If there is not good contact still, try lapping your CPU.

But make sure to READ the guide THOROUGHLY and ask questions here if unsure about some pf the procedures. You can also write me mail personally if you like
 
L

Loki427

Junior Member
Jul 1, 2009
4
0
0
Glad folks were able to point out the need for lapping, You should not need to lap CPU as they should be flat already (but it will help slightly) . Here is a good cheap kit /w video:

http://www.frozencpu.com/produ...tsink_Lapping_Kit.html

My 920 @3.8ghz runs ~27c idle, ~58c load, @~22c ambient. Xigmatek HDT-S1283 cooler /w direct heatpipe contact, AS5 paste.
I had it running over 4ghz, but the temp increases were too dramatic for such a small gain in performance.

If you are unable to resolve temperatures and continue to crash even at stock, I would try to get a replacement.

(I am at sea level so have dense air, good point v8envy!)
 
GLeeM

GLeeM

Elite Member
Apr 2, 2004
7,199
128
106
Originally posted by: Loki427
My 920 @3.8ghz runs ~27c idle, ~58c load, @~22c ambient. Xigmatek HDT-S1283 cooler /w direct heatpipe contact, AS5 paste.
I had it running over 4ghz, but the temp increases were too dramatic for such a small gain in performance.
Click to expand...

1. At 3.8 what vcore do you have set?

2. What vcore did you need at 4.0? And what temps did you have?

3. What program are you using to measure temp?

Thanks
 
L

Loki427

Junior Member
Jul 1, 2009
4
0
0
1. 1.247 vcore @ 3.8 .

2. I don't remember what I had it on over 4.0, sorry. Foggily remembering the temps I want to say it was in the 6-10C range.

3. I've discovered a discrepancy with between software and system so I'll provide both examples. I had been reading from Speedfan's results, as I thought them to be accurate since it and HWmon agreed on System 1. System 2 I had never bothered to test this way.

System 1: XP 32-bit, e4600 @ 3.2, SpeedFan v4.38= 58C load, Hardware Monitor v1.14= 58C load, RealTemp v3.0 = 64C load

System 2: Vista 64-bit, i7 920 @ 3.8, SpeedFan= 60C load, HWmon= 71C, RealTemp= 73C(also 43C idle)

Since on System 2 it is SpeedFan which is not consistent, I am inclined to believe the other 2. Interestingly, the "CPU" temp listed in HWmon was the same temp as Speedfan.

So assuming the 2C higher ambient, I would say my previous temp should be 70C, not 58C! Which totally solves the temp mystery for me. Thanks for your interest!
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom