Quantified: How high temperatures, cooling affect CPU performance

[Jos]

If you’re looking for authoritative information on how much cooling is enough for your CPU, you’ll be hard-pressed to find an answer. While we all know that modern processors need some degree of active cooling, very little official information exists to say how different temperatures affect a CPU's performance.

Older CPUs would simply fail if they started to overheat, but modern CPUs adjust their frequency based on temperature (among other things) to prevent a dramatic failure. Because of this, it stands to reason that once you reach certain temps, you will no longer be getting the maximum performance from your CPU because it will be busy protecting itself.

But what is that temperature? And do you really need a high-end liquid-cooled system to get peak performance, or is the little stock cooler that comes with most CPUs enough? In this article we will answer these questions and more.

Read the complete article.

 

[Burty117]

Am I blind (which is entirely plausible) or did this article only mention Intel CPU's and not compare them to anything made by AMD?

 

[Guest]

Apparently AMD does not matter and no one needs AMD CPUs (at least this is what I can understand from this article). Even so, I think an AMD CPU would behave differently on a stock cooler.
As for the temperatures on Intel CPUs, I can tell you that on my i5 760 I get maximum 56 degrees C with a CM 212 cooler, while on a haswell i5 you get like 75 degrees C in stress applications with the same cooler (I've tested this). And also take into consideration that the Lynnfield CPU consumes more power than Haswell (without integrated GPU).
So why do we have this kind of difference between the 2 CPUs ? Well, it's because Intel decided that soldering of the IHS is too good for the average consumer, and thermal paste is good enough. What this means is that a Lynnfield CPU can survive with a stock Intel cooler in a stress tool with around 80 degrees C, while on Haswell you get like 100 degrees (on stock frequency).
What can I say, Intel is the best ...

 

[Guest]

Well, I am curious about how would an Athlon x4 860K's over-100°C chart would look like compared to Intel's.

 

[Guest]

Am I blind (which is entirely plausible) or did this article only mention Intel CPU's and not compare them to anything made by AMD?

Well, they kind of need to compare apples with apples don't they, and AMD in this day and age may as well be an olive, or a sour grape with the kinds of performance and kit they put out. Intel is well known for their thermal capabilities, something AMD is not known for among another long, long list of things they cant compare on. AMD at best can be known as a budget company producing mediocre performance on higher energy use but for cheaper prices (You get what you pay for), something that as far as this article is concerned doesn't warrant a mention as you wouldn't need to put their mid range to low end merchandise through anything close to these types of temperatures.

 

[Guest]

Unbelievable the kind of bigotry some people exhibit behind the guise of a guest account.

 

[GhostRyder]

Well that is an interesting article, guess we all could assume that higher temps can degrade performance over time however its cool to see what it actually does at a higher temp. Its actually shocking the CPU will allow itself to stay that high before really cutting back and even so its only cutting back very slowly. Nice to see it in a definitive article specifically on the subject and to see how even 20 extra bucks for a CPU cooler can be all the difference in the world.

On the subject of AMD not being here, I would be curious to see what they do specifically at higher temps but I would assume they probably downclock the same but at a much higher pace and if I recall they have a lower temp threshold as well. I think it would be nice to see how an APU performs since those are very common in laptops and such if they were to compare them for a taste but that's all.

 

[LNCPapa]

Pretty sure the title of the article is Impact of Temperature on Intel CPU Performance. Perhaps you can wait till there is one called Impact of Temperature on AMD CPU Performance?

 

[loading]

Pretty sure the title of the article is Impact of Temperature on Intel CPU Performance. Perhaps you can wait till there is one called Impact of Temperature on AMD CPU Performance?

Pretty much this.

On point, I'm also surprised that a CPU can stay working that long at full load. I figured the fail safe would have kicked in and shut down the machine. Something interested to test would be the different generations of CPU's such as a soldered Sandy vs a thermal Ivy for instance or the differences between dies.

More interestingly yet, would these CPU's follow a linear path when overclocked compared to stock. I normally recommend to my friends building a rig to set aside a little money for a better CPU cooler especially if they ever plan on overclocking. The Hyper 212 EVO has been a great little cooler and for $35 I'd say it's worth it.

 

[Kibaruk]

While stock cooling only causes around a 2.5% drop in performance, even a budget after market cooler will dramatically improve CPU temperatures

I bought one of these CoolerMaster Hyper TX3 -and to add a bit if anyone cares- and back then when I read all the benchmarks and reviews it was the absolute best budget after market cooler out there. You could've bought a Noctua or what not for a lot more for little improvement over the TX3.

Interesting thread.

 

[Capaill]

I'm confused by what the article means by the CPU losing performance after 20% of its time. I'd prefer if they used actual time. I can't see why the overall length of time the CPU is working affects the point at which it starts reducing in performance. For example, if it's 10 minutes at 100 degrees, it's going to do it at that point regardless of whether it runs for a total of 20 minutes or 20 hours.

But I like articles like this that inject a little reality into our crazy expectations. With everyone determined to turn their CPU into a block of ice through cooling, it's nice to see an expert recommend letting it get a little warmer.

 

[Cryio]

Awesome article. Seriously.

Now make a similar one using the FX 9590, 8370E, 8350 and/or latest Kaveri APU or Athlon II x4 based on Kaveri.

 

[SuperVeloce]

You do not need a CM Hyper TX3 cooler for i3. Tcase would need to be stupidly high for i3 to hit 100*c in anything lighter than linpack-like stress-testing. If you live in warm climate, then you could take aftermarket cooler for more silent system I guess...

I have never heard of anyone destroy a stock "core I" series cpu with those throttling-inducing temperatures.

 

[Adhmuz]

The last time I had experience with AMD CPU's and high temperatures it resulted in system shut downs, right around the 105 degree C mark, there was extreme throttling up to that point as well. Nothing like how these new Intel chips perform, its really remarkable, considering how bad the issue use to be, going back to last generations of P4, those had issues, unless you used sub zero cooling to achieve ludicrous overclocks. Now if only there was a way to test the long term effects of running your brand new I series chip at 95 degrees all day for over a year.

 

[hahahanoobs]

What is the difference between Tcase and TJ max?

 

[Guest]

Intel specified the Tcase for this CPU at 72.72°C. So if you intend to run your i7 4790 at full load for prolonged time it would be wise to make sure it stays below this temperature. So running it at 85°C is not a good recommendation if you want to keep your CPU for years. Besides, this is a very high end CPU - a 50$ cooler or an additional case fan should help greatly.

However if you do not use your CPU at full load all the time even the stock cooler will keep the temps in check most of the time. Then again people who need such CPU power (and don't just own one for bragging rights) will probably run intensive tasks.

 

[dividebyzero]

Am I blind (which is entirely plausible) or did this article only mention Intel CPU's and not compare them to anything made by AMD?

The article is written for Puget Systems blog, TS reprint the Puget articles. Puget are Intel-centric. AMD options are limited and a downgrade (system cost), so the company likely just tailor their articles to their system fit-out, sales and marketing base.

What is the difference between Tcase and TJ max?

Tcase is the recorded temp of the (whole) CPU package. TJmax are the individual core temps. Depending upon CPU loading you could have a high TJmax (some cores running at ~100%) but a dissimilar Tcase temp (idle cores, effective heat removal from cores via thermal compound/solder > heatspreader > heatsink)

 

[Jad Chaar]

Am I blind (which is entirely plausible) or did this article only mention Intel CPU's and not compare them to anything made by AMD?

Yeah I am really curious about AMD CPUs especially since they usually run hotter. Is there throttling point the same?!

 

[treetops]

Damn so your down a ghz at 60c.

 

[Guest]

This sounds like a ad for intel. every other sentence is a compliment.

 

[cliffordcooley]

This sounds like a ad for intel. every other sentence is a compliment.

That is because it was, now lets see if AMD can make the same advertisement. Should I start laughing now, or wait for results?

 

[nickc]

Clifford, go ahead and start giggling. I have run computers at 100%, and overclocked to the max. with water cooling for years and although in some instances I have had to increase the voltage to the cpu the motherboard is generally what has given up, of course that was probably because of temp. to and yes I am sure every thing got slower as it got older.

 

[dividebyzero]

That is because it was, now lets see if AMD can make the same advertisement. Should I start laughing now, or wait for results?

Slightly different paradigm for AMD processors as a general rule. AMD CPUs don't typically have the same overclocking headroom as Intel processors (Intel have always been conservative in stock clocks) for mainstream air/water cooled systems, consequently to gain somewhere near the same percentage overclock requires a considerable voltage bump in comparison.
The different processes being used (HKMG vs SOI) also make comparisons difficult. SOI tends to love esoteric (sub-zero) operation, but overclocking tends to be more limited with less aggressive forms of cooling.

 

[Guest]

My rule of thumb is never go over 70° celcius stay below 60°c is the best . system ( northbridge, southbridge, cpu, video card, ) running constantly over 70° is prone to experience problem . some video card run at higher temperature but is is best to keep it lower.

 

[Cryio]

Running my FX 6300 at 4.5 GHz (not 100% and not all the time) and in games I never reach 80 celcius. It stays in 67-77.