I7 temp advice 3.8ghz on air

frozenreads

Distinguished
Jun 5, 2009
9
0
18,510
Hi,

Been steadily rising my i7 cpu up to 3.8ghz on air and was wondering if someone could confirm if my temps are ok. I am pretty new to overclocking and am not sure what 'good' temps are for the i7 920 when OC'ing. My sunbeam freezer cooler seems to be doing a good job.

So at 3.8ghz I am getting around 32-38c on idle and up to mid 40's for something like crysis. Running 4 x sp2004 produces mid 50's to 60C. Is this too hot, or is this ok? I have heard of people getting these temps at lower speeds, so I am assuming my cooling is doing a pretty good job. Internal system temps never get above 32c.

Any advice would be fantastic! many thx

 

D3LTA09

Distinguished
May 19, 2008
163
0
18,690
Sounds pretty good, but you really would want to download prime95 and do some blended stress tests and check temps with both real temp or core temp (make sure are the same). Dont need to leave it going for hours, 10-15mins is more than enough for thermal saturation to occur. As for good temps there is a guide written by computronix on these forums which will tell you all about what temps are good, should have a read. Essentially you would want under 70c but technically thermal thortling wont occur until 100c, personally I like mine at 60 or less which is why I backed mine down to 3.5 from 3.8.
 

Winly

Distinguished
those temps look ok. they would be better if they stay below 50 but its fine. i recommend you to run prime95 and some other programs that stress your pc so you can be sure you have a rock solid OC. and let them run at least 8 hours, that way you are really sure.
 

frozenreads

Distinguished
Jun 5, 2009
9
0
18,510
lol, thanks for the advice guys, its all running nice and stable, its just this is my first OC'ing in years. The last time I dabled with my own computer build was in the days of 486 overdrive chips and autoexec.bats! Thats how ancient i am! Great advice, tx. One thing I notice though is running at 3.6ghz I can drop vcore down to 1.35, but the leap to 3.8 requires an increase to 1.4 to get stable. Quite interesting!
 

CompuTronix

Intel Master
Moderator
frozenreads,

Overclocking and temperatures are about understanding Intel's voltage and thermal specifications, and not exceeding them. There are many variables involved, but let's go over just the basics.

Core i7s have a single CPU temperature sensor (Tcase), which is the temperature shown in BIOS, and four Core temperature sensors (Tjunction). For all Core i7 variants, Intel's Thermal Specification is 68c, which is shown in their Processor Spec Finder - http://processorfinder.intel.com/details.aspx?sSpec=SLBCH

68c is maximum CPU temperature (Tcase Max), NOT Core temperature, which is a common misconception among many users. Also, there's a 5c Gradient between CPU temperature (lower) and Core temperature (higher), which is shown in the following Intel document - http://arxiv.org/ftp/arxiv/papers/0709/0709.1861.pdf

Although maximum Core temperature (Tjunction Max) is 100c for Throttle and Shutdown protection, the corresponding CPU temperature would be 95c, which is far too hot for sane operation. As such, CPU temperatures above 68c and Core temperatures above 73c should be considered as an "overtemp" condition.

All Core i7 variants:

Vcore Max 1.375
Tcase Max (CPU temp) 68c
Tjunction (Core temp) 73c

SpeedFan shows CPU temperature AND Core temperature, while Real Temp shows Core temperatures ONLY. Both programs allow calibrations. Intel has stated that the Digital Thermal Sensors (DTS) used for monitoring Core temperatures are accurate at very high temperatures, become less accurate as temperatures decrease, and may be unreliable at idle temperatures.

On the other hand, the Analog Thermal Diode used for monitoring CPU temperature is linear from idle thru high temperatures, which is why, as per agreement with Intel, motherboard manufacturers do not include Core temperature in their monitoring utilities found on the installation disk, such as Asus Probe.

Never assume that default temperatures are accurate. The accuracy of CPU temperature is determined by BIOS calibrations, which can be re-calibrated in SpeedFan. The accuracy of Core temperatures are determined by Intel factory calibrations, which can also be re-calibrated in SpeedFan and Real Temp.

Prime95 Blend, or OCCT (Linpack), or CPU Burn Test (LinX) are cyclic workloads, which produce fluctuating temperatures. While these are useful for stability testing, they're inappropriate for thermal testing. Prime95 Small FFT's is the standard for thermal testing, because it's a steady-state 100% workload.

Since ambient temperature is a major variable, always try to test as close to 22c as possible, which is standard ambient. Also, test with case covers removed, and all fans at 100% RPM. This will eliminate cooling varables, and will provide more consistent results for comparison.

When testing, thermal saturation is reached within 7 to 8 minutes, so a 10 minute test is adequate. Keep in mind that even the most processor intensive games or applications will rarely exceed 70% to 85% sustained workload.

From the Core i7 and Core 2 Temperature Guide: http://www.tomshardware.com/forum/221745-29-sticky-core-core-temperature-guide


Scale 4: Quad
Ci7 9xx: Tcase Max 68c, Stepping C0, TDP 130W, Idle 16W<--Core i7
Ci7 9xx: Tcase Max 68c, Stepping D0, TDP 130W, Idle 16W<--Core i7
QX6x50: Tcase Max 65c, Stepping G0, TDP 130W, Idle 16W
QX6800: Tcase Max 65c, Stepping G0, TDP 130W, Idle 16W
QX6700: Tcase Max 65c, Stepping B3, TDP 130W, Idle 24W
QX9650: Tcase Max 64c, Stepping C1, TDP 130W, Idle 16W
QX9650: Tcase Max 64c, Stepping C0, TDP 130W, Idle 16W
QX9775: Tcase Max 63c, Stepping C0, TDP 150W, Idle 16W

-Tcase/Tjunction-
--65--/--70--70--70--70-- Hot
--60--/--65--65--65--65-- Warm
--55--/--60--60--60--60--Safe <--
--25--/--30--30--30--30-- Cool


For more detailed explanations, please refer to the Temperature Guide.

Comp :sol:
 

frozenreads

Distinguished
Jun 5, 2009
9
0
18,510
Thats great advice, wow, fantastic, thanks very much!

I have noticed a couple of strange things that happen when messing with settings and wondered if anyone can shed any light:

1. If I increase my memory bus multiplier by just one multiple i.e from x8 to x10 I always ALWAYS have a strange S3 issue where the computer fails to wake a second time (always second), with cycling debug codes on the bloodrage board. This happens as far as I can tell regardless of any other changes. I tested this by increasing one thing at a time i.e. uncore multiplier, voltages etc, and the issue only happens when memory mult is set to x10! I have even tested x10 multiplier at different frequencies by dropping bus speed, and the issue occurs at all freq! x8 everything runs absolutely fine, at every freq I can throw at it, why in the hell would S3 have an issue with x10 regardless of memory freq? I noticed that after rebooting the computer after the S3 failure, the memory freq reported in POST has dropped to the x8 multiplier even though BIOS settings have not changed. Its a real puzzler not critical, but the technician in me wants to get to the bottom. I am thinking possible BIOS bug?

2. I notice that if i drive my memory freq higher than 3.3ghz I have odd cold boot issues where if I hit the power button, machine starts, stops immediately and restarts absolutely fine. This only happens after power supply has been turned off and all power drained from mobo. I hear that there are cold boot issues with I7 architecture, I am assuming its nothing to be alarmed about? Possibly related, I notice if I drive uncore voltage to +140mv or higher, the same thing happens.

Again thanks for all the help
 
Why not to do it:

Assuming it's for gaming, if the game can't utilize the CPU at that speed I would clock it lower. Also, it may run stable for a month and then fry.

How to tell if your game uses the CPU:

Run the Task Manager (CTRL-ALT-DEL). Now, start a game leaving the TM running. Now, if you had a single core you could tell by looking at your CPU's history and if it didn't hit 100% then the game would run EXACTLY the same if you turned around and overclocked the CPU by 20%.

It's slightly more complicated for four cores (shows as 8 because of threading). If NONE of the 8 threads goes to 100% your okay. If one does it's hard to say. If two do it's probable you need that CPU speed.

So I'd run about 5 of your most demanding games and adjust your CPU accordingly. Personally, I wouldn't go above 3.4GHz.
 

CompuTronix

Intel Master
Moderator
The definition of an ideal overclock is simple:

Maximum performance within voltage and thermal specifications.

Vcore Max 1.375
Tcase Max (CPU temp) 68c
Tjunction (Core temp) 73c

See my signature.

Comp :sol:
 

frozenreads

Distinguished
Jun 5, 2009
9
0
18,510
lol, you know what having tried overclocking right up to 3.6ghz, I have come to the conclusion that from a personal experience point of view, I notice absolutely no difference in performance from stock speed. Using my Geforce 295GTX for gaming, pretty much everything runs maxed anyhow. I have come to the conclusion that a HT I7 920 at stock is more than enough for todays computing without having to OC! It might show higher numbers in these benchmark programs, but I really dont think its noticable to the eye. With all the S3 and cold power issues I seem to experience at higher speeds, I just dont think its worth it. The 920 is blindingly fast anyhow. I think these benchmark programs send us all down a false path. Yes its very nice that my new graphics card reports 120fps on COD4, but hey, my monitor has a 60hz refresh rate, and my eyes are happy to accept moving piccies at 25fps, so that 120fps is lost on me! lol.

 

D3LTA09

Distinguished
May 19, 2008
163
0
18,690
Ha yep your right about that, I havent noticed a difference between stock, 3.5, 3.8 and 4ghz except quite a lot more heat and if anything everything runs smoother at stock. But thats always the way with OCing when your not CPU bound when gaming.

I have also noticed a strange power issue the last 2 days that when I turn my rig on from cold it fails to post and I have to turn it off and on again and then everything is fine. Never happened at under 3.8 though so a bit different to the OPs situation. I assume its not stability related? as done quite a lot of stress testing with prime95 at all frequencies.

Also can someone explain why my workers in prime95 finish tests at slightly different times? doesnt seem to happen when im at stock bclk.

EDIT: should learn to read above^ lol
 

frozenreads

Distinguished
Jun 5, 2009
9
0
18,510
I have noticed power issues at 3.4ghz and beyond, all ways related to cold booting. From what I can understand its just a x58 chipset/i7 thing, nothing to worry about I believe.