CPU Cooler Reviews using non-soldered IHS CPU's

[El_Capitan]

I was reading a review on the Noctua NH-D15, and they used an i7-4770K and Handbrake for the load tests. Temps were great even overclocked to 4.4GHz at 1.225V's, but having used an i5-3570K with a custom 120.2 watercooling solution, it got up to 90C within 15 minutes of using Prime95 with similar overclocks.

 

Once I delidded the 3570K and put the CPU back in the loop, load temps never got past 55C during load on Prime95.

 

I think reviews for CPU Coolers should revert back to using 2600K's, or use delidded 4770K's. What are your thoughts?

[ir_cow]

give me a delid 4770k

[l1il]

Your concern is good Cap...

 

I was pondering on that, but in the end, I think testing coolers on newest CPU's is what people are looking for. The i7 980x is outputting a lot of heat, but nobody is testing coolers on this CPU, even if it can hold it's spot on a SMP benchmark.

 

Whether we like it or not, Intel is going to use the cheapest assembly technique to splurt out as many CPUs as they can. If they are using only medium-grade thermal paste between IHS and the die on new CPUs, testing on a 2500K is going to be pretty irrelevant. It is even worse thinking that Intel is looking into BGA only CPUs for future versions (or at least this was rumored in a bunch of review sites). So buying a cooler that was great at cooling a 3930K might not improve cooling on a 4770K compared to a 30$ cheaper cooler, due to higher Tdie-Tcase resistance.

 

The Socket 2011 is one that has it's IHS soldered to the die. Testing the coolers on this platform will have different results from testing on socket 1150. People are mostly buying socket 1150 processors, so this is the one that reviews are going to focus on, with a reason.

[cchalogamer]

give me a delid 4770k

Ship yours to me and I can fix that for you

 

Back on topic, there is a clear reason to test as the CPU was shipped from Intel.  For one most people are far too scared to risk their CPU doing the work so testing the coolers on the unmodified platform gives the more "standardized" review.  With that said it would be very easy to argue that you want to see a cooler tested on every current socket (as of now 1150/2011/AM3+/FM2+) to see how it performs in different tests, after all a 2011 CPU's IHS is MUCH larger than 115X CPUs so coolers with different sized or different levels of flatness on the base will perform VERY differently across the varying IHS designs.  Then you get into in a case vs not and using a standardized fan vs not and 100 other things that can change the way a cooler works.

 

In short, there's never going to be a site with enough time and money to test it all, but if someone wants to send me a nice AM3+ board/CPU and the same for FM2+ a long with a huge supply of coolers and ~$10/hr for my time I'll be glad to test them all out for you in any test you want.

Edited April 26, 2014 by cchalogamer

[panjang110]

I was reading a review on the Noctua NH-D15, and they used an i7-4770K and Handbrake for the load tests. Temps were great even overclocked to 4.4GHz at 1.225V's, but having used an i5-3570K with a custom 120.2 watercooling solution, it got up to 90C within 15 minutes of using Prime95 with similar overclocks.

 

Once I delidded the 3570K and put the CPU back in the loop, load temps never got past 55C during load on Prime95.

 

I think reviews for CPU Coolers should revert back to using 2600K's, or use delidded 4770K's. What are your thoughts?

 

that is a very hot chip maybe intel cheap you out on the thermal paste. mine never get past 70C with my cooler master v8 and now mid 50's with a custom loop ([email protected])

 

what i think reviewer should do is do multiple cpu testing on all platform so all of us with older cpu can get fresh data, the trend i'm seeing right is now almost all review is done with 1150 cpu. i can barely see testing done on am3 let alone fm2

Edited April 26, 2014 by panjang110

[ir_cow]

honestly it wouldn't be more challenging to have multiple platforms for a case review. just add 4-5 hours per platform on top of the 10-20 it takes (case dependent) to write a reivew and there you have it.

 

The extra costs include a motherboard , CPU , CPU cooler and ssd for each platform so you can just drop it in without having to deal with partitions and swaping stuff other than physical.

 

In the end your looking at the reviewer forking over a lot of money for a few numbers. now if companies wanted to donate their products I would be interested in working with that But it's not going to happen and my 2011, Am3 and Fm2 are all tired up in other working computers I can't be taking apart for ever review .

[ccokeman]

We used to use a 2600K since it was one of the most popular out at the time but have migrated to the 4770K.  The smaller die and IVR help generate a lot of heat that has to be discharged through a smaller surface area than on previous chips. You can get comparable results but air cooling on its own is not going to give you a lot of head room on a socket 1150 CPU. Hopefully the rumored Haswell refresh will fix that problem.  Even so testing with the socket 1150 platform in a case gives realistic comparable results. If you do the math you can figure out where it might perform with one CPU or the other.   

It would be nice to perform testing across multiple platforms but you introduce added time and complexity where you do not need to. Using one platform will give you the data that can translate over across platforms. Introducing a higher thermal load than the design limit of the cooler does not show you anything but the fact that it is overwhelmed much like the stock Intel heat sink was in my last CPU Heatsink review. 

 

By using one platform tested in a case you get a good idea of how it will perform against a comparison field. Testing is more than just slapping a cooler on and running it for 15 minutes on a variable synthetic load. The applied TIM does not even have the time to burn in as it may be.

By using the same fan on all coolers you automatically throw out the window how the cooler performs as shipped from the manufacturer. The best way to compare coolers is to test them as they  are delivered so you can see how it performs instead of as a modified assembly. If you want a better fan on the cooler thats up to the end user.

[doyll]

Honestly, to test coolers a CPU isn't even needed.  All that is needed is an adjustable heat source same size as chip under an IHS.  There is some who test this way. 

 

To me the biggest problem with cooler testing is the use of room ambient as baseline for CPU temperature increase and not the temp of air that is going into cooler.  In open bench testing the cooler intake and room temps will be similar.  But when testing in a case the temps change dramatically; from cooler to cooler, even with changing fan speed on cooler because of heated cooler exhaust air circling back into intake air.  We also have GPU's heated exhaust causing changes in cooler intake temps.

 

This temperature difference shows up most when comparing CLC / AIO and air coolers..  and air coolers like SST HR01,  Silver Arrow SB-E Extreme and Silver Arrow IB-E Extreme because of their high cfm airflow.

 

I'm not proposing using only the cooler intake temp.  Having room ambient is useful in knowing how the testers' systems cool..  But it really has little relationship in comparing different coolers.. unless reader has same system.. and very few of us have same system as testers' review we are readlng.

 

I'll get off my soapbox now.

Sorry for the hi-jack. 

[ccokeman]

Honestly, to test coolers a CPU isn't even needed.  All that is needed is an adjustable heat source same size as chip under an IHS.  There is some who test this way. 

 

To me the biggest problem with cooler testing is the use of room ambient as baseline for CPU temperature increase and not the temp of air that is going into cooler.  In open bench testing the cooler intake and room temps will be similar.  But when testing in a case the temps change dramatically; from cooler to cooler, even with changing fan speed on cooler because of heated cooler exhaust air circling back into intake air.  We also have GPU's heated exhaust causing changes in cooler intake temps.

 

This temperature difference shows up most when comparing CLC / AIO and air coolers..  and air coolers like SST HR01,  Silver Arrow SB-E Extreme and Silver Arrow IB-E Extreme because of their high cfm airflow.

 

I'm not proposing using only the cooler intake temp.  Having room ambient is useful in knowing how the testers' systems cool..  But it really has little relationship in comparing different coolers.. unless reader has same system.. and very few of us have same system as testers' review we are readlng.

 

I'll get off my soapbox now.

Sorry for the hi-jack. 

 

Correct, a variable heat load can simulate a CPU to some extent and can be tuned to a specific heat load. That would be great if the IHS on all CPU's were flat. Most current CPU's have a concave IHS and CPU coolers are now designed to compensate for this so the base of the cooler best fits the IHS . A variable that you will have using just a heat load. In the past it was a great idea when all heat sinks were flat, but not so right now.  

 

99% of people do not use an open bench for their systems and testing with a variable heat load does not bring all variables into play.The key is to have your airflow through the test chassis high enough that the ambient temperature inside the chassis is a close to the external ambient as possible. The variables you list are the main reasons for testing in a chassis. Sure a cooler might be the best  cooling, quietest thing on an external bench, but that changes when slapped in the chassis because of the variables you noted. All of a sudden it is a noisy beast that cools fairly well and does not garner the same praise. Like I said earlier based on how it performs in our test setup will vary from the results you see on an open bench. One perspective vs. another. Open bench is nice but in a chassis is as real world as it gets regardless of the system used. If you look at the test system we use for testing coolers many of the concerns you listed are taken into account with a reference cooled card and an SSD.

 

Thanks for your opinion!      

[doyll]

Correct, a variable heat load can simulate a CPU to some extent and can be tuned to a specific heat load. That would be great if the IHS on all CPU's were flat. Most current CPU's have a concave IHS and CPU coolers are now designed to compensate for this so the base of the cooler best fits the IHS . A variable that you will have using just a heat load. In the past it was a great idea when all heat sinks were flat, but not so right now.  

 

99% of people do not use an open bench for their systems and testing with a variable heat load does not bring all variables into play.The key is to have your airflow through the test chassis high enough that the ambient temperature inside the chassis is a close to the external ambient as possible. The variables you list are the main reasons for testing in a chassis. Sure a cooler might be the best  cooling, quietest thing on an external bench, but that changes when slapped in the chassis because of the variables you noted. All of a sudden it is a noisy beast that cools fairly well and does not garner the same praise. Like I said earlier based on how it performs in our test setup will vary from the results you see on an open bench. One perspective vs. another. Open bench is nice but in a chassis is as real world as it gets regardless of the system used. If you look at the test system we use for testing coolers many of the concerns you listed are taken into account with a reference cooled card and an SSD.

 

Thanks for your opinion!      

 

What is happening is testing is of a system as a complete unit.  That is testing the cooler in a way very few of the readers use it..  because more than 99% of their systems are different from test system.  .

 

Using variable heat load instead of CPU is really a non-issue.  Contour can easily be the same on both. Heatsinks have not been flat for many years. My Ultra 120A (2007) is almost flat but Cogage Arrow (2009) in convex.

 

The variables I listed are not the issue..  except for the cooler intake air temperature that varies so much from room temperature.

 

A cooler that is the "quietest thing" does not become "a noisy beast"  when put in a case.

 

Having airflow through chassis is all fine and good, as is ambient chassis temperature readings.  But neither of them will consistently be the same as cooler intake air temperature. So  why not monitor and use the cooler intake temperature as the base? 

 

"in a chassis is as real world"

The "real world" is only the system being tested... at most it could include systems with identical components and cable management.  Change any components or cable management in airflow and results will most likely change.  

 

Being in a case or not.. Using a CPU or other heat source..  These are not the problem I'm trying to address.  Using room ambient as base temperature instead of cooling intake air temperature is the real problem..  and one that is very easy solved.

Room ambient temperature is topically 20-22c. 

Cooler intake air temperature can easily be 20-40c..  resulting in radical differences is CPU temps and therefore test results are very inaccurate.  To me CPU delta temps that are up to 18c different than delta is based on temperature of air going into cooler is a variable we can easily resolved by installing a temperature sensor a few cm in front of cooler air intake and using that temperature for cooling baseline instead of room ambient.

By all mean, keep a record of room ambient too.  To show how much the a cooler intake air temperature varies from room temperature.  

 

Using the cooler intake air temperature removes most variables from resutls.  

 

Please understand, I'm trying to discuss this problem, not argue.

Edited May 17, 2014 by doyll

[ccokeman]

 

Correct, a variable heat load can simulate a CPU to some extent and can be tuned to a specific heat load. That would be great if the IHS on all CPU's were flat. Most current CPU's have a concave IHS and CPU coolers are now designed to compensate for this so the base of the cooler best fits the IHS . A variable that you will have using just a heat load. In the past it was a great idea when all heat sinks were flat, but not so right now.  

 

99% of people do not use an open bench for their systems and testing with a variable heat load does not bring all variables into play.The key is to have your airflow through the test chassis high enough that the ambient temperature inside the chassis is a close to the external ambient as possible. The variables you list are the main reasons for testing in a chassis. Sure a cooler might be the best  cooling, quietest thing on an external bench, but that changes when slapped in the chassis because of the variables you noted. All of a sudden it is a noisy beast that cools fairly well and does not garner the same praise. Like I said earlier based on how it performs in our test setup will vary from the results you see on an open bench. One perspective vs. another. Open bench is nice but in a chassis is as real world as it gets regardless of the system used. If you look at the test system we use for testing coolers many of the concerns you listed are taken into account with a reference cooled card and an SSD.

 

Thanks for your opinion!      

 

What is happening is testing is of a system as a complete unit.  That is testing the cooler in a way very few of the readers use it..  because more than 99% of their systems are different from test system.  .

 

Using variable heat load instead of CPU is really a non-issue.  Contour can easily be the same on both. Heatsinks have not been flat for many years. My Ultra 120A (2007) is almost flat but Cogage Arrow (2009) in convex.

 

The variables I listed are not the issue..  except for the cooler intake air temperature that varies so much from room temperature.

 

A cooler that is the "quietest thing" does not become "a noisy beast"  when put in a case.

 

Having airflow through chassis is all fine and good, as is ambient chassis temperature readings.  But neither of them will consistently be the same as cooler intake air temperature. So  why not monitor and use the cooler intake temperature as the base? 

 

"in a chassis is as real world"

The "real world" is only the system being tested... at most it could include systems with identical components and cable management.  Change any components or cable management in airflow and results will most likely change.  

 

Being in a case or not.. Using a CPU or other heat source..  These are not the problem I'm trying to address.  Using room ambient as base temperature instead of cooling intake air temperature is the real problem..  and one that is very easy solved.

Room ambient temperature is topically 20-22c. 

Cooler intake air temperature can easily be 20-40c..  resulting in radical differences is CPU temps and therefore test results are very inaccurate.  To me CPU delta temps that are up to 18c different than delta is based on temperature of air going into cooler is a variable we can easily resolved by installing a temperature sensor a few cm in front of cooler air intake and using that temperature for cooling baseline instead of room ambient.

By all mean, keep a record of room ambient too.  To show how much the a cooler intake air temperature varies from room temperature.  

 

Using the cooler intake air temperature removes most variables from resutls.  

 

Please understand, I'm trying to discuss this problem, not argue.

 

 Not taking it as an argument and I understand the point you are trying to make.