Anandtech CPU cooler reviews

[imported_wonderwrench]

Does anyone have any idea why all reviews that use the new test setup show max overclock for nearly every cooler ever tested but only show temp charts for the reviews since the test setup was changed? I'm guessing because the max overclock tests were not re run on on all coolers using the new test setup. If this is true the max overclock tests are invalid. They always were deceiving because at the same frequency the coolers are in no order what so ever. AKA without looking close you would think the higher a cooler is on the chart the better it is which is not always true. If this is not the case why would the idle and load temps not be reported for all coolers tested.

If you compare old review temps of the Thermalright ultra 120 extreme to the new reviews there is 10C difference @ 3.83 ghz. New reviews are higher. This tells me you need to add a minimum of 10C to any cooler tested in a old review to be able to compare results from newer reviews. This 10C increase could be much more on coolers that are lacking in heat exchanger surface area.

The reason for this post is many people seem to think all Anandtech's CPU cooler reviews can be directly compared and the can not. Several coolers that have dropped from the temp charts people still think are in the top 5. Tuniq Tower 120, The Zalman 9700, Synthe Ninja plus B.

Tuniq Tower 120 which looked so good in the old reviews would show 60C @ 3.83 GHZ if tested using Anandtech's new system. The Zalman 9700 would be 68C, Ninja plus B with stock fan would be 70C. This makes sense because the new test system could have higher case temps. Its also possible that temps reported by the motherboard change make up some of the temp increase. From what I can see the Cooler Master Hyper 212 = 57C and even the Arctic Cooling Freezer 7 Pro = 58C beat all three coolers listed above. The Tuniq Tower 120 should come back at higher heat loads do to larger heat exchanger. I would hope anyway.

To prove this one way or another. Anandtech needs to retest all coolers from the previous tests (if they have them) on the new test rig if they are still for sale new. Report the idle and load temps for all and either ditch the overclock charts or fix them so they are useful. To be useful the charts need to list the coolers in order of performance and should list the resulting temp at the given max overclock. The data can't be from two different test rigs.

 

[imported_Scoop]

I think the 'max OC' chart is completely useless, I'm only interested in the scaling of cooling performance. But what you said about the temp differences between the new system and the old one isn't as simple as you played it to be. Take a look at the retail Intel cooler results. Old system: 71C @ 3.73Ghz, New system: 65C @ 3.73Ghz. So you can't just add 10 degrees to the temps of all the old coolers based on the TRUE results.

The result difference in the test systems with Corsair Nautilus 500 and the Swiftech Compact is strange as well. When using watercooling, the case and such shouldn't really make any difference in the performance when the radiators aren't inside the enclosure. Yet both end up with 2C higher results @ 3.83Ghz in the new system.

There is something strange going on with the cooler tests at Anandtech for sure. It would be nice if they did a round up and spice it up with a Q6600.

 

[imported_wonderwrench]

Originally posted by: Scoop
I think the 'max OC' chart is completely useless, I'm only interested in the scaling of cooling performance. But what you said about the temp differences between the new system and the old one isn't as simple as you played it to be. Take a look at the retail Intel cooler results. Old system: 71C @ 3.73Ghz, New system: 65C @ 3.73Ghz. So you can't just add 10 degrees to the temps of all the old coolers based on the TRUE results.

The result difference in the test systems with Corsair Nautilus 500 and the Swiftech Compact is strange as well. When using watercooling, the case and such shouldn't really make any difference in the performance when the radiators aren't inside the enclosure. Yet both end up with 2C higher results @ 3.83Ghz in the new system.

There is something strange going on with the cooler tests at Anandtech for sure. It would be nice if they did a round up and spice it up with a Q6600.

I see what you are saying. I guess in the end I do not believe any of the results as the more I
look at them the more unbelievable they are.

 

[toadeater]

Originally posted by: wonderwrench
I see what you are saying. I guess in the end I do not believe any of the results as the more I
look at them the more unbelievable they are.

You're right not to believe them. Whenever you see comparison reviews, they're old reviews compared to the latest review. Quality control isn't always perfect for review sites doing this, and most of the time they will not even use the same fans (as seen in Frostytech's anomalous results for the TRU and TRUE) so that some coolers will do better or worse with a new fan than previous coolers tested with some other fan. Noise levels don't stay the same for the same reason. The best you can do is to average the results of reviews for a cooler to get an estimate of how good it is. Then see what results people are getting in the real world.

Since most 120mm coolers today have nearly identical performance except at extreme temps (are you really planning for a 4GHz+ OC?), it really comes down to noise, ease of installation, cost, longevity, and whether you need a top-down cooler, or a tower cooler.

The other deciding factor is case cooling. DX10 cards putting out ~200W of heat can raise your CPU temps by 10c in a badly-ventilated case, not to mention raise NB and SB temps. So in a review a cooler might seem great, but in your case in the real world it won't necessarily get the same results.