theory behind cooling

[c.ruel]

The easiest way to explain my question is through a metaphor. This is a little weird but bear with me. When you play COD or some other shooter the guns each do different damage in the same way that different coolers have different levels of efficacy. 25 dmg is an important cut of point along with 33 and 50. If a gun does 24 dmg it will take 5 bullets to kill and if a gun does 25 dmg it will only take 4. Therefore this 1dmg difference makes a massive difference in the efficacy of these guns. Conversely a gun that does 32 dmg is really that much more effective than one which does 26.

So on to my question. I was recently reading a of a review of the swiftech watercooling kit (which performs better than an AIO) compared to the top air coolers. All of the coolers where basically much of a muchness. Water did better a bit better than air temp wise but was a bit noiser. So for arguments sake let's say that most of the best coolers are within 5 degrees of one another (a reasonable assumption). Could 1 degree on an overclock be the difference between stability and instability or is stability more on a sliding scale.

[TotesTurtle]

Unless you were hitting TJ Max those 1-5 degrees are going to make 0 difference.
Stability is only indirectly related to temperature, temperature usually only dictates how far you can OC a chip. Voltage is what affects your stability, and with voltage comes more heat.
Usually you'll be pushing a chip really hard if you manage to overvolt it enough to reach nearly 100 degrees on any cooler anyway.

I have a Hyper212 EVO with an NF-P12 PWM fan on it, and it does me just fine, 4.4 Ghz at 1.22 volts, 75c full load (prime 95 for 20 mins).

[kalniel]

Roughly speaking, the heat generated for a given load is the same, regardless of cooler. A more effective cooler just transfers this heat to the environment more quickly. A less effective cooler transfers more slowly at a given CPU temp, so the CPU temp increases, at which point the temperature differential to the environment works in the cooler's favour and the heat is dissipated and the temperature should stabilise. The only problem with CPU stability comes when this temperature stabilisation occurs above the CPU rated temperature, or TJ Max as noted above. At this point the CPU will throttle to ensure stability.

There are some minor downsides to having the CPU stable temperate higher - slightly increased power draw, and slightly increased substrate wear rate. The latter is trivial for desktop usage.

[c.ruel]

Ok. But for example could you have a chip that was stabled overclocked to 5ghz at 60 degrees but not stable at 70 degrees with the same overclock where 70 degrees was within the TJ max of the chip?

[TotesTurtle]

Ok. But for example could you have a chip that was stabled overclocked to 5ghz at 60 degrees but not stable at 70 degrees with the same overclock where 70 degrees was within the TJ max of the chip?

If the TJ max was 70 then the CPU would throttle back it's speed until it wasn't so hot, so yes I guess you could call that unstable.
Most chips have a TJ max of around 100c nowadays.

[herulach]

Ok. But for example could you have a chip that was stabled overclocked to 5ghz at 60 degrees but not stable at 70 degrees with the same overclock where 70 degrees was within the TJ max of the chip?

If you can hit the max stable clock for a chip on air you won't get it any faster on water. However, when you get up the range you won't be able to do that. My WC set up (2 360 rads 2 970s and a 4790k) can just about cope with the CPU at 4.6 and the GPUs at about 1500. (89C on the cpu with prime & furmark running).

[watercooled]

It's not strictly true that temperature makes no difference to stability, but whether a few C are going to make or break an overclock is another matter, and other things are likely to impact borderline-stable clocks, possibly overshadowing the temperature difference.

Increased temperature leads to undesirable effects on processors such as increased resistance which can directly lead to instability for pretty much the same reason as decreased voltage as described in some detail here: http://www.overclock.net/t/794098/th...cience-warning

An example demonstrating this effect would be liquid nitrogen overclocking. Allow such an overclock to warm up slightly (still far, far below TJmax) and you get errors quite quickly. As I said one or two C may not have a clear impact on regular overclocking, though such an experiment would no doubt be interesting to see.

[Pob255]

yes it is possible to have an unstable overclock while below the TJ max of a chip and yes it is possible to have an overclock that is stable at one temperature but not at another.

You have to remember that temperatures will effect the conductivity and resistance within the chip, remember that the world record overclocks are achieved with liquid nitrogen, because you have to get the silicon down below 0c to get the oc stable.

I'm not saying your oc will be stable at the right temperature, it just MIGHT be.
That's one of the thing you have to remember about overclocking, you are pushing your components above their rated ability, there are many factors that come into play.

[c.ruel]

Really good answer pob255! You mention below zero cooling, that is quite a big variance from say 60 to 70 degrees. So in the "real world" if you use normal coolers that can't go below ambient, can a small difference in temperature like 5 degrees make a difference? For example 45 instead of 50? I remember someone saying that with phenom IIs low temps were more important than high volts when trying to achieve a particularly overclock.

[watercooled]

You'd have to try it, or find someone that had, and it could be different for each processor; as I said though it would be an interesting test to see.

[Pob255]

As I said, a 5 degree difference MIGHT effect your overclock stability, there's too many unknown factors in play, the power stability both from psu and motherboard VRM, the quality of the silicon in the cpu (every cpu is slightly different).
Finally there's one of the biggest factors, how much you plan on pushing the overclock.
The more you push the overclock the more likely it is you'll hit limiting factors.

[watercooled]

Am I on ignore lists or something, since I posted roughly the same answer 40 minutes in advance but got overlooked?

Edit: Sorry I really got up on the wrong side of bed. It's easy enough to miss posts if you get two new ones at once...

[kalniel]

Did someone say something?

[c.ruel]

So to summarise: many factors affect an overclock one of which is temperature. Crucially temperature can affect an overclock way below the TJ max of a cpu. Yes?

As an experiment to explore the relationship with temperature and overclock when using above ambient cooling, could you get a ridiculously overpowered custom loop say a quad radiator with 3000rpm fans, then find your max overclock on your chip and once you have done that you could raise the ambient temperature in increments of 5 while running prime until you hit instability. Then lower overclock and until you hit stability again and repeat. This would tell you the key temperatures for your cpu at what clock. I know the results would only be useful for that particular cpu but it would be interesting to see what kind of pattern formed e.g. linear or whatever.

[watercooled]

Pretty much, but I really wouldn't know what to expect at normal OC temperatures. Rather than going for completely max OC, you could try leaving the voltage alone and take up the clock; this way it would be easier to keep the CPU cooler and therefore be able to test over a wider temperature range, just turning down the fans to let things heat up more.

[TotesTurtle]

Am I on ignore lists or something, since I posted roughly the same answer 40 minutes in advance but got overlooked?

Edit: Sorry I really got up on the wrong side of bed. It's easy enough to miss posts if you get two new ones at once...

I can see you mate