David_Schroth

Administrator
Staff member
Joined
Apr 23, 2019
Messages
780
Points
93




Introduction



One of the areas that we’ve been working on behind the scenes to refresh and get moving has been for CPU cooler testing. We’ve had a few AIOs (All-In-One) coolers in the backlog for a bit and as I stared at the boxes I saw one that claimed that it could dissipate “500+ W” of heat. Thinking about this a bit it dawned on me that an overclocked Ryzen 1700 has exactly zero chance of testing that claim.



I didn’t know exactly what that change would end up being, but the puzzle pieces started to come together when I saw the powerful (pun intended) potential that lay in the Intel 10980XE chip, especially when overclocked. This was put into the back of my mind until a few months ago when I was looking for cases available on Fry’s Electronic’s website and it said that my local store had a...
Continue reading...
 

Burticus

FPS Regular
Joined
Aug 15, 2019
Messages
1,159
Points
113
Slow golf clap. I look forward to your AIO torture sessions.
 

Terminus

n00b
Joined
Jun 20, 2019
Messages
39
Points
8
I like it! Well done, looking forward to reading some AIO reviews!
 

Grimlakin

FPS Junkie
Joined
Jun 24, 2019
Messages
2,290
Points
113
Looking forward to these articles. Now I'm wondering if my 280 rad is enough for my 3900x!
 

magoo

Sort-of-Regular
Joined
Jul 9, 2019
Messages
241
Points
43
How are you planning to measure heat output?
I am curious as to what metric one uses?
Most reviews simply graph a line-up of coolers versus each other based on a standard system, ie a single CPU overclocked to a certain GHz level.
It would be interesting to change that up by looking at AIO versus custom loops as well.
I think "The Real RedRaider used to do that as well as measure the heat created by the CPU.....it was pretty scientific, if I recall.
 

David_Schroth

Administrator
Staff member
Joined
Apr 23, 2019
Messages
780
Points
93
How are you planning to measure heat output?
I am curious as to what metric one uses?
Most reviews simply graph a line-up of coolers versus each other based on a standard system, ie a single CPU overclocked to a certain GHz level.
It would be interesting to change that up by looking at AIO versus custom loops as well.
I think "The Real RedRaider used to do that as well as measure the heat created by the CPU.....it was pretty scientific, if I recall.
Quite frankly, heat output will not be 100% scientific - I'm going by total system power as provided by a Kill-a-watt. It stands to reason that when idle is ~100w and max OC under load is ~600w, that we're somewhere in the ballpark of 500w of power likely going into the CPU (sure, memory/SSD/board will be drawing some of it, but not a material amount).

After working with two different AIOs and seeing what they could do, I settled on 5 tests to be performed at 3 different frequencies. Specifically:

Frequencies (approx total system draw):
  • Stock/3.8Ghz All Core/Default Voltage (300w)
  • Low OC/4.3GHz All Core/1.15v (450w)
  • Max OC/4.7GHz All Core/1.25v (600w)
For each of the frequencies, I've run the following tests:
  • Max Fans, Max Pump Speed
  • 1500 RPM Fans, Max Pump Speed
  • 1000 RPM Fans, Max Pump Speed
  • 600 RPM Fans, Max Pump Speed
  • 600 RPM Fans, 50% Pump speed
If a test in the above order fails (thermal throttling kicks in - as indicated by tracking core clocks and the bright red light that turns on from the board), then the quieter settings will get skipped (i.e. if it throttles at 1500RPM, I'm not testing 1000RPM). The initial approach was to only do stock at max OC, but let's just say not much data got collected as a result 🙃.

The differentiation that we will have here is that our "standard" system is capable of putting far more wattage into a cooler than just about any other that I've seen. When I've looked around, I've seen the other cooler reviews using 8700K, 9700K, Ryzen 7 chips and so on. While those represent a real world workload for a majority of users, it doesn't test the claims made by some of the marketing materials.

I'm not familiar with the "Real RedRaider", but I did consider building a custom rig that uses the hot side of a pelt to allow for dialing in a specific amount of wattage. It really wouldn't be that expensive to build (maybe a couple hundred bucks), but then it comes back to how to measure the relative performance - how could I consistently measure how the unit performed when attached? About the only possibility would be sandwiching a probe between the pelt and block, but that would cause mating issues... End of the day, I need to get some form of content moving, but I'm not opposed to revisiting this concept and adding it into the mix...
 

magoo

Sort-of-Regular
Joined
Jul 9, 2019
Messages
241
Points
43
thanks for that info.
my friend at RealRedRaider shut his site down about 4 or 5 years ago.
He had a gang of real scientists that contributed on a number of topics from cool to pumps to flow in various system assemblies to plating in blocks.
He also used Martin's Lab alot.
The site was a little abrasive so it did turn some folks off. But he was a little like Kyle Bennett.....no BS and he called people out, but less tact......so.
Their claim to fame was the EK nickle plating problem several years back.
Id still like somebody to run AIO against custom loops, straight up.
There are times when Ive been frustrated with my custom stuff.....read care and maintanence........and have been tempted to let someome else do the work.
 
Become a Patron!
Top