Hardware Some questions about our new CPU cooler test systems

One BIG enhancement to cooler testing would be switching from Core Temp readout to Intel's recommended Tcase measurement method (thermocouple milled into centre of IHS). The internal Digital Temperature Probe that can be polled from the OS/BIOS becomes WILDLY inaccurate outside of a few degrees away from TCC. At idle, deviations of 10°C from the true temperature are well within the acceptable error bars! Plus, Intel specs thermal performance in terms of Tcase, not Tj, so measuring Tcase could be more useful for puchasing considerations.
I am not familiar with AMDs thermal specifications and recommended measurement methodology..

 

I'm in favour of standard test results. I run a 4770k which I delidded out of necessity as I got a terrible chip. De-lidding made a huge difference but I didn't want to do it. Would rather invest in a more capable cooler than delid.

 

I can see what you're saying, but surely what's under review is not the cooler's ability to cool, but the cooler's ability to cool a CPU. It's all well and good one cooler being better than another but if it can't keep an 'average' temperature CPU cool then it's a bit useless.

Plus I'd wager a lot of people skim read articles and may miss the part about any delidding, making people think BT is rather useless if their results vary so much from everyone else.

^+1 for this!

 

I agree with de lidding since that was a manufacturing error, either de lid or don't use CPU's that require de lidding, a CPU that needs de lidding isn't fit for sale or if it is then it should have the price slashed dramatically IMHO.

Or have one de lid & one stock.

 

Don't de-lid, it doesn't represent an average home user.

 

There is not a single CPU that requires de-lidding, it is an after market method of gaining a combination of extra performance and extra cooling. So I really don't understand what it is you are trying to say.

 

Wouldn't any inaccuracy in the sensor remain constant as long as the CPU isn't changed, I.e if one CPU is out by 10°C as long as that CPU is used in all test it would be an equal playing field.

IMHO the important thing with a test system is consistency.

 

Depends if you are happy with temps being higher than they should be, sounds like you'd be happy with that, I wouldn't but I also wouldn't want to void warranty due to poor manufacturing so I'd simply avoid buying known CPU with this thermal problem.

But if I bought one before knowing & I couldn't get my money back too right I'd risk fixing it.

 

rather not see chips de-lidded, adds new variables to a test that is simply about comparisons between coolers on the same chip. the coolers are the things being tested so the chip should be out-of-the-box and as ordinary as possible.

would like to see water and air cooling sharing a graph but with different colours so i can see at a glance at what point water cooling becomes worthwhile considering the extra cost/complexity. if this makes for extra long charts perhaps just have a snapshot of the chart with the 5 nearest results above/below the tested product in the review and a full long chart of everything tested and at what speeds at the back of the magazine or on the website ?

 

Can I ask a question here?

With regards to the TIM used inside the CPU, does 'long term baking' become an issue? I am wondering if a test with a cooler when the CPU is new might show a lower temp than one performed after the CPU is a couple of years old as this may skew results towards any cooler that was tested when said CPU was newer and the TIM performed better. Has a test been done with a new CPU and then again a couple of years later with the same actual CPU and cooler to see if results vary much?

 

When I did cooler reviews, I de-lidded my 4770K. The temperatures dropped by 23c across all cores.

The reason I did it was to 1) lap the IHS to get a perfectly smooth base (mirror shine) and 2) to get more accurate temperatures which aren't limited to the **** TIM in the middle of the CPU.

Sure, it may not be a representation of the "real" CPUs out there, but you also need to remember that one non de-lidded CPU can be far better, or far worse, than another chip. No two are the same. My very first 4770K would hit 100c (throttle temps) at stock under a Silver Arrow SB-E, so you tell me...

 

So you didn't just de-lid, you also lapped the internals and you wonder why those results shouldn't be used? Your comment is exactly why de-lidded and/or lapped CPUs should not be used when REVIEWING coolers because the data returned is very chip specific as well as having to factor in the TIM used. Generic is best when it comes to reviewing.
Yes a generic processor may vary slightly from one chip to another however, they are ALL certified to run at their stated clock speeds within stated thermal limits, only with OCing does the chip lottery start to play a part.

Besides, isn't a review on coolers supposed to be a method to differentiate between different coolers? If they all end up showing very similar min/max temps isn't that making it pointless? I want to be able to decide between coolers, not just say 'Well done everybody, you're all the same'.

 

They don't show similar temperatures, at all. Doing such things to the CPU eliminates the CPU from being a variable.

De-lidding your CPU won't allow you to run at 5GHz on the stock cooler, and give you the same temperatures as a custom water loop...

 

Can we have separate charts for regular and delidded (great word) results? Regular better reflects the real-world, but delidded would emphasise performance differences.

Would be good to split off air and water coolers, but a reference would need to be included.
Eg. in the water-cooling chart include a single bar for the best and worst air-cooler and maybe the average (not including stock)

 

hah, bet you never even tried it

Anyway, I wouldn't mind seeing delidded results as well as stock, if nothing else, it would show what differences delidding would make. (if you happened to have a delidded version of the same CPU lying around).

 

i'd be interested in a feature/article about benefits to be had from de-lidding various sorts of CPU and/or testing different thermal goos & application/lapping techniques but i maintain that for the purposes of this thread the cooler is the thing on test not the CPU, therefore CPU must be as average as any home users out-of-the-box performance.

 

Nope, it goes under LN2 instead.

Like I said, guys...

It went back for RMA and the retailer confirmed it had an issue with temperatures.

 

Q1. Keep the chips in factory spec / no de-lidding

Q2. As long as it is easy enough to understand i'm not bothered, if it's AIO's only, maybe slip in the highest rated air cooler just for quick comparison.

 

At least they asked the community's opinion, you don't see many (if any sites) doing that.

I've voted above, Q1. Don't de-lid and Q2. All in one graph (basically so its easier to compare, different colours might be a good idea though.)

 

Don't delid the processor.

Mounting and remounting processors is going to mess with the results.