I7 Temps

[Greg_77]

For point of comparison, I thought it would be good to create a thread for people to list the clock speed, cpu temp (idle and load), software to measure CPU temp, and the cpu cooler being used for their i7. This reference could be used by people to get a rough estimate of what temperature their i7 should be. I will start:

Model: i7 920

Clock: 2.66 GHz

Temp Idle: 35 C

Temp load: 55 C

Temp Software: Real Temp (Tj max 100 C)

CPU Cooler: Coolermaster V8

 

[xthekidx]

Model: i7 920

Clock: 3.0 GHz (150x20)

Temp Idle: 42 C

Temp load: 81 C

Temp Software: Real Temp (Tj max 100 C)

Stock Cooler for the moment, going to install V8 this weekend.

 

[CompuTronix]

Greg_77,

When comparing temperatures, the formula just isn't quite that simple. It's all about variables and specifications, so in order to accurately compare apples to apples, you also need to include other basic test settings, which directly affect idle and load temperatures such as:


Ambient Temperature: ?

Computer Case Covers: Removed ?

Computer Case Fans: Manual 100% RPM ?

CPU Fan: Manual 100% RPM ?

Internet: Disconnected ?

Load: Prime95 Small FFT's ?

Programs : Closed ?

Tray Trash: Closed ?

Vcore: ?


Comp

 

[Greg_77]

CompuTronix, thanks for correcting me. I wrote this late at night and now see how many variables I missed. To correct myself, my load testing software was prime 95 on large FFT, my internet was connected, all programs (except temperature software) were closed, the ambient temperature was 22 C,my case was completely closed, my CPU fan was at 100%, other fans were at 50%, and my idle was done at all programs closed except, of course, realtemp. My V-core is 1.25 V at idle (have not looked at load yet). If you want to see my configuration, I added it under member configuration.

 

[cjl]

Prime95 smallFFT
i7 965
Antec 1200, low fans
Ambient around 21C

Stock clocks:
Full load: ~58C
Idle: 30C

3.73GHz @ 1.22V:
Full load: ~62C
Idle: 32C

4.00GHz @1.32V:
Full load: 70C
Idle: 35C

4.26GHz @ 1.38V:
Full load: 78C
Idle: 39C

Cooler: TRUE black with dual fans, Arctic Silver 5. All listed overclocks are fully stable as tested with linpack and prime95 for at least 4 hours.

 

[CompuTronix]

... prime 95 on large FFT ...

Greg_77,

The accepted standard for processor thermal testing and processor stability testing is Small FFT's.

Large FFT's is described in Prime95 as "maximum heat, power consumption, some RAM tested". This statement is somewhat misleading. Large FFT's will create overall maximum heat & power among cores and memory.

Small FFT's is described in Prime95 as "maximum FPU stress, data fits in L2 cache, RAM not tested much". This means that although there is less overall heat, there is maximum heat concentrated among the cores only.

These differences between Large FFT's and Small FFT's can be seen in SpeedFan's "Charts" which graphs the thermal signatures of the cores over time. Large FFT's will run a few degrees below maximum temperatures for approximately 20 to 25 minutes, but will then ramp up to the same maximum temperatures of Small FFT's. This fluctuating workload pattern is repeated when certain passes are completed. The workload fluctuations seen in "Blend" are much more extreme.

Small FFT's is a steady-state 100% workload which does not vary. This is obvious in SpeedFan's Charts, where it can be seen that thermal saturation is reached within 7 to 8 minutes, so a 10 minute test is adequate for thermal benchmarking. It is also widely accepted that Small FFT's should be run for a minimum of 8 hours to validate processor stability.

As such, Small FFT's is the correct test to use for processor temperatures, and not OCCT, Burn Test, or any other shortcut "easy button" thermal test solutions which are currently being tossed around the Forums. If we don't adhere to a standard test setup so that everyone uses the same variables, specifications and methods, then temperatures are just apples and oranges thermal fruit salad in a blender.

Comp

 

[cjl]

If you really want maximum temps though, the correct test to use is some form of Linpack. It does get hotter than Prime95, and is an excellent stress and stability tester.

 

[cjl]

Here's an example:

My stock clocked full load Prime95 temps (smallFFT):

Linpack temps:

(all other settings were identical)

 

[CompuTronix]

cjl,

I don't agree with using Linpack for thermal testing, and neither does Kevin Glynn, the author of Real Temp.

Just as TAT was popular at one time, Linpack also loads all registers with all one's. The result is unrealisticly high temperatures due to a 115% worload, which also fluctuates, and can quickly cause temperatures on some overclocked rigs to spike well beyond safe values, which your post at stock settings does not illustrate.

Even when video editing or running games which are heavily CPU bound, such as M$ Flight Simulator X, the processor workloads rarely exceed 80% of Prime95 Small FFT's. Again, because Prime95 Small FFT's maintains a steady-state 100% workload, there are no surprises or radical changes in temperatures.

Comp

 

[benningt]

Model: i7 920

Clock: 3.4 GHz (170x20)

Temp Software/Stress Test: OCCT

CPU Cooler: Coolermaster V8

Case: Nzxt Zero (mod)

Ambient Temperature: 24C

Computer Case Covers: On

Computer Case Fans: Manual 100% RPM

CPU Fan: Auto, Normal (Asus P6T Deluxe)

Internet: Idle

Load: OCCT (mix for 1 hour)

Programs : Closed

Vcore: 1.275

 

[Greg_77]

Benningt, am I missing something, or did you forget to post your temperatures? :lol: Whatever the case, all the comments here are right. In order to get useful temperature comparisons, we need to set some sort of standard. Maybe someone more experienced than myself could create it, and we could all agree on whether to use it or not. I believe that temperature data consolidated on one thread could be very useful for the overclocking community. Finding good temperature data with a google search for the i7 processor turns up little.

 

[cjl]

cjl,

I don't agree with using Linpack for thermal testing, and neither does Kevin Glynn, the author of Real Temp.

Just as TAT was popular at one time, Linpack also loads all registers with all one's. The result is unrealisticly high temperatures due to a 115% worload, which also fluctuates, and can quickly cause temperatures on some overclocked rigs to spike well beyond safe values, which your post at stock settings does not illustrate.

Even when video editing or running games which are heavily CPU bound, such as M$ Flight Simulator X, the processor workloads rarely exceed 80% of Prime95 Small FFT's. Again, because Prime95 Small FFT's maintains a steady-state 100% workload, there are no surprises or radical changes in temperatures.

Comp

I'm not saying that any normal workload will even remotely approach the level of loading (or temperatures) that linpack will. However, a truly stable CPU should be able to run linpack without difficulty, and it will often demonstrate an instability in a much shorter time than prime95 will. Note that Linpack is a real load that could possibly be run on the computer (specifically a linear algebra package, which solves large systems of linear equations). Also, as far as temperatures "well beyond safe levels" are concerned, the modern CPUs are quite good at thermal protection, and if they reach levels that are truly detrimental to the processor's health, they are capable of downclocking themselves. No, it would not be good to run 24/7 right at TJmax, but it isn't as harmful as you might think to run it quite hot. This is also shown by notebook CPUs, which are the same dies as the desktop parts, and will often run all day at 80C in heavy load situations (depending on the notebook's cooling).

In addition, Linpack gives an excellent floating point benchmark. Mine, for example, scores around 44 GFlops stock, and around 55GFlops at full overclock (and yes, it is within temperature spec the whole time).

 

[randomizer]

Here is an example of how temperatures vary with 3 different benchmarks; IBT (Linpack) and Prime95 Small and Large FFTs. If Linpack didn't bottom out the load every couple of minutes it might be more useful. But with the CPU load constantly dropping for a good 10-15 seconds at a time, I don't know how much good it would be for stability testing, let alone monitoring high-load temperatures.

 

[cjl]

I've used it for stability testing, followed by Prime95, and every time I can pass 4-6 hours of Linpack, I can subsequently pass 24 hours of Prime95 with ease. I still use Prime95 to make sure, but Linpack does seem to work well.

 

[randomizer]

I would be interested in seeing which one fails first on a slightly unstable OC (not one that fails in 5 minutes, but after an hour or so).

 

[benningt]

Sorry for missing the temps on my first post...

Model: i7 920

Clock: 3.4 GHz (170x20)

Temp Software/Stress Test: OCCT

CPU Cooler: Coolermaster V8

Case: Nzxt Zero (mod)

Ambient Temperature: 24C

Computer Case Covers: On

Computer Case Fans: Manual 100% RPM

CPU Fan: Auto, Normal (Asus P6T Deluxe)

Internet: Idle

Load: OCCT (mix for 1 hour)

Programs : Closed

Vcore: 1.275

Core1: Idle: 37 Load: 68
Core2: Idle: 34 Load: 65
Core3: Idle: 40 Load: 67
Core4: Idle: 34 Load: 64

 

[OmegaStalker]

Out of curosity a few questions? Firstly Whats the highest Stable overclock people have got? second whats considered a safe load temp on prime95 after a few hours? and the TRUE black 120 does it come with the fans? for an i7920 what have people got as the safestish, highest OC (not too high Vs, not too high Temps?), and if i've missed anything go easy on me and explain plz

 

[OmegaStalker]

come on?

 

[randomizer]

i7 OCs look to be tied fairly close to batch #. See if your batch number is on here and it will give you an indication of what to expect. It's not a guarantee though.

 

[cjl]

My highest stable overclock was 4.26GHz so far, but I'm somewhat conservative with voltages - I won't push them higher than the 1.38V that it took to achieve 4.26. As for the TRUE, it doesn't come with a fan, but any standard 120x25mm fan will attach easily enough.

As for safe load temps, I'd feel comfortable with anything below about 75C or so. 80C should still be OK, but is starting to push it a bit.

Oh, and batch number for me is 3837A871

 

[kylen1487]

Anyone got any advice? I cant get my CPU any hotter! Its running at -195C right now and I think its crazy!!

You overclockers are gonna love this!!!

Hahahaha! enjoy!!!

 

[kylen1487]

This is a video on how to get the CPU temp down to -195C

http://ca.youtube.com/watch?v=evtGvukJ3KI

enjoy!!!

 

[cjl]

Not tremendously useful though - what matters for this thread is useful, day to day overclocks and temperatures.

 

[Greg_77]

I looked at that batch list Randomizer, and don't see a correlation between batch number and clock speed. It could just be because of the different cooling methods used and the intentions of the overclocker. Most seem to get around the same temperature at 4 GHz using aircooling. The good news is that they all can get to around 3.8 and 4 GHz.

 

[randomizer]

I looked at that batch list Randomizer, and don't see a correlation between batch number and clock speed. It could just be because of the different cooling methods used and the intentions of the overclocker. Most seem to get around the same temperature at 4 GHz using aircooling. The good news is that they all can get to around 3.8 and 4 GHz.

I thought the same thing more or less when I first saw it, but after seeing a few other people's overclocks and batch numbers (who aren't in the spreadsheet), there appears to be some correlation. Although to make a proper assessment, a far larger sample size is required, because all of these OCs are on different hardware with different people and different cooling. You'd need several people with the same hardware and cooling to get a proper idea of the connection between batch # and OCing capabilities. Of course, having the same person OCing them all on the exact same components is the best option. The spreadsheet in it's current form is inherently flawed and obviously can't be used to make any definitive judgements.