9 air-coolers for Haswell tested

[JimmiG 244]

Over at Toms Hardware:

http://www.tomshardware.com/reviews/best-heat-sink-haswell,3554.html

 

They tested with a 4770K at 4.5 GHz / 1.25V (a fairly typical clock speed and voltage). The best coolers managed about 65C over ambient at full load. This means with a room temperature of 25C, the CPU will run at about 90C. Some coolers that used to be considered high-end only a few years ago now fail to keep the CPU under 100C at this overclocked speed.

 

This mirrors my experience with the Phanteks PH-TC14PE. At 4.3 GHz / 1.217V, it runs about 85C during stress tests, with a room temp around 25-27C.

 

 

 

The most noticeable feature of our temperature charts is the narrow range of approximately 12° between highest and lowest CPU core temperatures. Dismantled Core i7-4770s suggest that this is due to the interface material Intel uses. At best, this shows that huge improvements in cooling technology are needed to create relatively small drops in temperature.

 

 

So how are people claiming 4.6, 4.8 and even 5 GHz overclocks at over 1.3V on air? Simple - they don't run those demanding stress tests. In something like FSX, with HT disabled, and AffinityMask tweak applied to lower CPU usage of Core #0, temps will be at least 20 - 25C lower than Linpack, Prime95 etc. This means you can still achieve much higher overclocks since you now have a huge amount of thermal headroom. The system will not pass a 24-hour torture session in Linpack without throttling, but it will happily run FSX all day. If the CPU is repeatedly hitting TjMax during a run, it will permanently throttle back, giving the impression that temps are good and the overclock is stable. Whether or not this counts as a "valid" overclock is debatable, but if you use your system only for flight sim, it doesn't really matter.

 

If you want to achieve 4.6+ GHz with 100% stability, you have a few choices:

-Custom-built water cooling setup (though with the diminishing returns of high-end air coolers Tom's noted, even that may not be enough for more than a ~5C reduction)

 

-De-lid

 

-Clock speeds don't have nearly the same effect on temperatures as the core voltage. Keep shopping until you find a "golden" CPU that can reach high clock speeds without going much above 1.25V.

[martin-w 3,235]

If the CPU is repeatedly hitting TjMax during a run, it will permanently throttle back, giving the impression that temps are good and the overclock is stable. Whether or not this counts as a "valid" overclock is debatable, but if you use your system only for flight sim, it doesn't really matter.

Sorry but I don't agree with that hypothesis.

 

Anyone that runs a CPU at a temp that's inducing throttling would know it. Performance drop would be noticeable, locks up's common etc.

 

As for the article at Tom's... beware of unilateral analysis. The opinions at Tom's Hardware have made me raise an eyebrow more than once.

 

And why on Earth didn't they test the NH-D14, the Silver Arrow, the Corsair H110??? They seem to have avoided the best cooling solutions.