Hardware Some questions about our new CPU cooler test systems

I'd like to see delidded results, but I think for the general populus, lid-on is where it's at. In terms of the graphs, grouping by cooler type works for me, but I'd echo the requests for links to other graphs for those 'keen' readers

PS Delidding CPUs is what the cool kids do. And me. But I'm the exception that proves the rule?

 

Hello.

It's great that you let us participate this way.

My main concern is with the graph, because when I'm reading reviews and reach it I find myself looking (inside that huge graph) for the few models I know and I have under control to compare them alone. This does not make the graph very user friendly to me, but I still want the rest of the data to discover new heatsinks.

Ideas about it:

• Smaller graph
• Interactive graph so I can select and compare the heatsinks I care about
• With the heatsinks I selected fixed on the graph, I still can hover the mouse over the rest of the heatsink list and the value will appear.
• What if I identify a heatsink I don't know about and it performs great? link to the review, from the graph heatsink list.

Hope it helps.

 

If the temperature sensor is on the chip, then the motive of testing the efficiency of a cooler is lost if you do not delid. Personally, I wouldnt delid a chip, but purely in terms of a test, the measurement point is separated from the object of interest by a "dodgy" interface which leads to inconsistency.

 

While interactive content within an article is generally a massive pain (not one website I've seen implement it has been enjoyable to use), a separate interactive chart would be very nice to have. If a unified test rig is being set up, then it should be possible to compile all results from that rig into a central database and produce on-demand charts from that (e.g. graph of all coolers compatible with socket X, graph of degrees above ambient vs. current sale price, etc).

 

Yeah, okay - I'm an idiot.

That 1 response for "Show graphs for either air or liquid coolers" is me, because I misread it. I forgot to untick it before submitting - apologies and feel free to remove/ignore/ridicule.

Didn't have time to post but have now ended up posting because....well, I've already highlighted that.

 

Disagree. Why would you buy a >£50 cooler, and then swap out the fans? You test the product as is. You may as well compare it to some other form of cooling if you're going to change the actual components it comes with.

 

Reviewing cooler performance should be about comparing the review sample to other, previous review samples. In this way we can determine how well it performs amongst its peers. So it should not be relevant whether the test cpu is delidded or not purely from a performance comparison perspective.

The benefit of keeping the original package intact is so that most readers have an idea of the expected level of performance for the cooler in question. Naturally the readers experience will vary due to manufacturing tolerances, how the user mounts the cooler, thermal paste, ambient conditions etc. but at least they will have a reasonable ball park of how the cooler performs from the review. Reviewing with a delidded sample skews that ball park figure in the wrong direction for the reader. So the reader sees the review and the product performs at temperature x under load. They get the product but for them it performs at temperature x+10 under load because they haven't delidded their CPU. Essentially their experience would not reflect your review unless they de-lid their CPU.If most users delidded their CPUs then the tests should of course reflect this, but most users don't from what I can see.

The graphs should be kept together in my opinion. All in one coolers perform more similarly to high end air coolers than they do custom loops. Buying, installing and maintaining an all in one cooler is also a more similar experience to an air cooler than it is a customer loop. So people looking to buy a new CPU cooler should be able to compare the two kinds easily. Unless they were set on one kind or the other before doing their research (how unscientific!) In which case keeping them on the same graph but with a different colour as was suggested seems like a reasonable thing to do.

If the graphs are getting massive, I would suggest pick the top 10 or 15 most popular or best performing for the site and host a seperate public doc with all the results like an public google document or something, for those that are interested. I don't think that extra overhead would be too much. If you are reviewing something like a low profile cooler, then pick suitable comparisons, rather than the best 10 or 15. This would be in a similar vein to forgoing 1080p benches on top tier graphics cards review or forgoing 4k tests when reviewing iGPU

 

Frickin stoked that you're bringing back AMD. I lost a lot of interest in cooler reviews (and a lot of respect for the site itself, frankly) when you dropped AMD. And it certainly doesn't hurt that I also run a FX-6300

On the Intel side, though: don't delid. It's no good all around - too many variables get introduced. And it's hardly representative of the experience for the vast majority of users.

As to the graphs, I'd like to see performance for price, like the cost divided by delta temperature over the stock cooler. As far as the graph getting too large, maybe within the reviews keep them close in price. If I'm looking to spend, say, $30 on a cooler, seeing how well a $100 AIO performs doesn't do me much good. That said, keep a full graph around. There are people that want performance above all, and a full graph keeps that easy.

 

difficult answer.
to test a cooler you should test it regardless of CPU and fan model: this will give the best comparison.
so ditch the cpu altogether and insert a reliable heat generator with maximum heat transfer to the cooler mating surface.
test it with different power output and different fan speed/flow.
Instead of simple steady state also some measuring of thermal inertia will be useful as could provide further details about real world behaviour.

 

Me

I'm using the noctua fans that came with my d14.

Ugly colours though, could do with a paint job.

 

Don't De-Lid

With the graphs - make the bars 50% thinner = twice as much space for results.
Or make them "Smart" so people can tick boxes to show results - by default it only shows the last two coolers reviewed (i.e. the graph is small) and has additional tick boxes for users to tick to compare if they wish.
You could even include a "show all" box for those who want to see everything.

 

As we are testing the cooler, why not drop the whole problem and just leave a skt 1150 out of the tests?

You already have a 2011 & AM3+ which gives a broad enough scope as they are much hotter running than 1150.

 

Testing coolers has become far too complicated.

 

If you're starting the list afresh, please consider remeasuring some "popular coolers" for reference.

The last performance king, the last silent king, best sold/most popular ones (air and water) things like that.


(I'd say simply take the ones from the last few hardware-buyers-guides but...nevermind)

 

No de-liding of CPU
Different colors for types of cooling.
different colors for SLI/crossfire/ 2gpu card.
Hardware buyers guide! budget - mainstream - 4k.

-EVRE

 

Although the techie in me would like to see the de-lid results it would only show the reader: 'wow the lided results are really bad', so it sort of defeats the point. Intel already knows this but perhaps the extra publicity would influence them to improve things. So I'm of two minds here. Now if AMD produced a killer CPU, that might spur a change.