CPU Cooler Test System Update 2021

[bug]

How so?
A candle is 1400C, my oven is 200C - yet the candle is much slower to cook a roast, and definitely slower at heating a room


The AMD's have a lesser amount of heat, condensed into a smaller area

Well, you said nothing about density in your first post, you simply said "intel puts out more heat, AMD is harder to cool". That's the part I explicitly quoted.

Even considering density, Intel was constantly bashed for using allegedly sub-standard IHSes, AMD is supposed to have the upper hand here. I mean, yes, the chiplet is small. But with the IHS over it and the base of the cooler on top, you'd expect plenty of room foe the heat to dissipate.
So far, cooler have been rating based on the wattage they can handle. Should we start them to be rated based on the wattage and positioning(?) from now on?

Intels 10nm should be similarly dense so I guess when Alderlake drops we will see who is better at it.

It'll be interesting when amd transitions to 6nm or 5nm how hard their chips will be to cool then.

I don't think the problem is (just) the process, the problem is when there are two chiplets, they are next to each other. If AMD managed somehow to put the IO die in between the chiplets, the heat would have been much more evenly distributed. But I'm sure they know that and they have good reasons for not going down that route.

 

[Mussels]

TDP ratings for coolers have always been made up

They can do 200W on a heatplate in 100% contact with their heatsink base with the fan at 100% in a windtunnel at 15C ambients - we never find out, they dont specify
according to google:
5800x is 80.7 mm²
10900k is 206.1mm²

AMD needs significantly better soldering and IHS to keep up when intel has far more than double the space to work with - they'll have to friggin flood under the IHS with some kind of non conductive paste at this rate

 

[oxrufiioxo]

Even considering density, Intel was constantly bashed for using allegedly sub-standard IHSes, AMD is supposed to have the upper hand here. I mean, yes, the chiplet is small. But with the IHS over it and the base of the cooler on top, you'd expect plenty of room foe the heat to dissipate.

8th/9th gen were pretty terrible... 10th/11th are pretty decent given the massive amount of wattage. Amd seems to be above average since at least zen 1.

 

[Mussels]

AMD increased heat density without an TIM/IHS solution
intel incresed wattage without a new TIM/IHS solution

End result: Both companies have higher and higher temperatures, but intels is easier to counteract, as an end user (by throwing ridiculous cooling at it)

It'd be interesting to see with a basic cooler like the cryorig C7 how each chip compares on auto stock settings, where the stock boost and throttle mechanisms are unhindered.
Does Ryzen or X lake throttle harder when cooling is insufficient?

 

[80-watt Hamster]

Despite the fact that it's at the bottom of every performance ranking, I can't help but love the Cryorig C7. I don't know why.

It's also interesting to see how well the Hyper 212 Black performs, given how often it and its siblings are slagged off by folks these days. Perf/$ is danged impressive.

 

[bug]

TDP ratings for coolers have always been made up

They can do 200W on a heatplate in 100% contact with their heatsink base with the fan at 100% in a windtunnel at 15C ambients - we never find out, they dont specify
according to google:
5800x is 80.7 mm²
10900k is 206.1mm²

AMD needs significantly better soldering and IHS to keep up when intel has far more than double the space to work with - they'll have to friggin flood under the IHS with some kind of non conductive paste at this rate

I know TDP is a rather vague metric with virtually no chance of being fixed.

I was just saying, what we're looking at here, possibly for the first time, is a rather popular cooler (NH-U12S) that can handle overclocking a 125W part, but unable to handle overclocking a 105W part. That's got to end with unpleasant surprises for the less informed users.

 

[ca_steve]

What's the noise floor of the testing room?

 

[Night]

You need to add Arctic cooling Liquid freezer 2 AIOs in this mix, they will probably blow the other AIOs out of the water (pun intended), since its the only AIO with a thicker 38mm rad, whereas all the other ones use a slimmer rad... For the prize and simplicity, I think its the best AIO out there. I have the 360mm and couldn't be happier. The only gripe is their AM4 mounting system, you need 3 arms to mount it lol.

Yeah, I was really impressed with the performance of their 240/280mm AIOs, I'm definitely considering it for the new build, but there's an issue with case compatibility, as it's 38 mm think (probably the thickest radiator out there), it's hard to find cases that can fit it on top, personally I don't feel like mounting it in front because of intake airflow obstruction. I like cases like Fractal Design Meshify series (2, 2 Compact), but what I read is nothing but trouble for Arctic's AIO in those cases. Usually it's RAM being too tall and colliding with the fans, Fractal's cases report maximum RAM height of 40 mm, and basically there are only low profile sticks that can fit to that measurement, also radiator thickness of 40 mm max. Anyone know of cases similar to Meshify series that can support this AIO? My CM Haf XM can definitely support it, because the radiator can be mounted on the top side of the chassis, while only fans are near the motherboard area excluding the radiator, so no clearance issues are possible here. While it does support it, I'll be buying a new case with updated onboard connections, dust filters, etc.

 

[bobbybluz]

You need to add Arctic cooling Liquid freezer 2 AIOs in this mix, they will probably blow the other AIOs out of the water (pun intended), since its the only AIO with a thicker 38mm rad, whereas all the other ones use a slimmer rad... For the prize and simplicity, I think its the best AIO out there. I have the 360mm and couldn't be happier. The only gripe is their AM4 mounting system, you need 3 arms to mount it lol.

I just replaced a Corsair H115 RGB Platinum with a Freezer II 360. It knocked almost 14 degrees C off the temp of my i9 9900K under load. The Corsair H105's also had 38mm thick radiators. IMO the H105 with aftermarket high speed fans was the best 240 AIO until the Liquid Freezer II's came out. As my H105's end their useful service lives I'm replacing them with the larger Freezer II 280's.

 

[heky]

"We also had to increase the SOC voltage to 1.4 V to maintain stability, which had Ryzen Master report an increase from 105 W on the CPU and 39 W on the SOC to 150 W on the CPU and 44 W on the SOC."
Surely that has to be a typo? Pushing 1.4v on the SOC is mental and surely not needed...

 

[tabascosauz]

"We also had to increase the SOC voltage to 1.4 V to maintain stability, which had Ryzen Master report an increase from 105 W on the CPU and 39 W on the SOC to 150 W on the CPU and 44 W on the SOC."
Surely that has to be a typo? Pushing 1.4v on the SOC is mental and surely not needed...

I shudder to think of the effect on that 3900X after testing 20 coolers 8 times each at 1.4 VSOC......JEDEC timings 3200 should require no more than sub-1.0V VSOC. Someone please tell @crazyeyesreaper that chiplet Zen 2 and Zen 3 should never exceed 1.2V ever, and shouldn't ever need more than 1.1V below 3800MT/s.

That I/O die runs at like 15-20W at Auto 1.1V, and is made for a maximum of something like 30W tops at the max safe of 1.2V. Running VSOC like this I imagine is akin to running a 10900K on ambient with 1.8V VCCSA.

 

[claes]

What's the noise floor of the testing room?

I’m curious about this as well.

Also, @crazyeyesreaper, any chance you could add standardized fans, like the old SPCR reviews? Thanks for the great work!

 

[crazyeyesreaper]

I shudder to think of the effect on that 3900X after testing 20 coolers 8 times each at 1.4 VSOC......JEDEC timings 3200 should require no more than sub-1.0V VSOC. Someone please tell @crazyeyesreaper that chiplet Zen 2 and Zen 3 should never exceed 1.2V ever, and shouldn't ever need more than 1.1V below 3800MT/s.

That I/O die runs at like 15-20W at Auto 1.1V, and is made for a maximum of something like 30W tops at the max safe of 1.2V. Running VSOC like this I imagine is akin to running a 10900K on ambient with 1.8V VCCSA.

The default voltage applied by the motherboard was 1.35v on the SOC (software), that said ill look into it, keep in mind those these are software reading and in the various apps they did fluctuate in regards to reporting.

BIOS reports 1.1 volts. So its likely software reading the voltage wrong. I had 1.125v set via BIOS, I will update the article to reflect the BIOS voltage settings ( went via software as the voltage set in BIOS is never what you really see. CPU voltage is accurate, but SOC is not.

Further Edit, appears to be software conflict, SOC volts are 1.1 in BIOS / 1.125 in BIOS for OC Windows software reporting tends to be a bit wacky with the combination of software I have but issue has been resolved. Article updated with BIOS set voltages.

I’m curious about this as well.

Also, @crazyeyesreaper, any chance you could add standardized fans, like the old SPCR reviews? Thanks for the great work!

Straight from the article below: That said, current testing time, photos, video / editing, troubleshooting, thermal paste spread check, multiple mounts to verify integrity of the data, and finally writing, means each cooler done as specified takes about 22-28 hours. Standardized fans would add another 18 hours to that.

All the above tests are repeated three times in total. The entire process is then restarted for noise normalized testing. For those wondering, this works out to 18 hours of testing excluding the time taken by reboots and swapping from one system to another. That figure also doesn't include all the other work necessary to produce a review, so if you want to ask why we do not test with standardized fans, etc., this is it. There is also the fact that the vast majority of users tend to use a cooler as sold.

 

[claes]

Don’t know how I missed that

Noise floor though?

 

[crazyeyesreaper]

Don’t know how I missed that

Noise floor though?

Noise floor is 32 to 33 dBA (fluctuates between the two. I have seen it drop as low as 31 but the typical reading in the room is 32 to 33.

 

[Mussels]

I know TDP is a rather vague metric with virtually no chance of being fixed.

I was just saying, what we're looking at here, possibly for the first time, is a rather popular cooler (NH-U12S) that can handle overclocking a 125W part, but unable to handle overclocking a 105W part. That's got to end with unpleasant surprises for the less informed users.

W1zz hasnt left these parts at stock, hence my question about how they'd behave differently on auto settings

AMD poll their chips more often than intel on the newer designs, so they could idle or lower clocks for very brief windows and retain most of the performance with a less penalizing throttle, for example

 

[bug]

Straight from the article below: That said, current testing time, photos, video / editing, troubleshooting, thermal paste spread check, multiple mounts to verify integrity of the data, and finally writing, means each cooler done as specified takes about 22-28 hours. Standardized fans would add another 18 hours to that.

All the above tests are repeated three times in total. The entire process is then restarted for noise normalized testing. For those wondering, this works out to 18 hours of testing excluding the time taken by reboots and swapping from one system to another. That figure also doesn't include all the other work necessary to produce a review, so if you want to ask why we do not test with standardized fans, etc., this is it. There is also the fact that the vast majority of users tend to use a cooler as sold.

Good thing it's not time consuming or anything
Much appreciated.

 

[basco]

Thanks for all your hard work and time !

 

[W1zzard]

W1zz

I’m not involved in this testing

 

[Deleted member 205776]

I shudder to think of the effect on that 3900X after testing 20 coolers 8 times each at 1.4 VSOC......JEDEC timings 3200 should require no more than sub-1.0V VSOC. Someone please tell @crazyeyesreaper that chiplet Zen 2 and Zen 3 should never exceed 1.2V ever, and shouldn't ever need more than 1.1V below 3800MT/s.

That I/O die runs at like 15-20W at Auto 1.1V, and is made for a maximum of something like 30W tops at the max safe of 1.2V. Running VSOC like this I imagine is akin to running a 10900K on ambient with 1.8V VCCSA.

My 3900X runs 4 sticks at 3600 MT/s with 1.0v on the SoC (yes, I recently lowered all those voltages to see what would happen, still stable). 1.4 is insane.

 

[crazyeyesreaper]

My 3900X runs 4 sticks at 3600 MT/s with 1.0v on the SoC (yes, I recently lowered all those voltages to see what would happen, still stable). 1.4 is insane.

It was a software glitch reporting the wrong voltage or maybe the right voltage? BIOS voltages are 1.1v and 1.125 for the system. But i noted oddities in the voltages reported when various apps are running at the same time. Even so if the BIOS says 1.1 the voltage could still very well be 1.2 or higher in reality just food for thought.

 

[Deleted member 205776]

It was a software glitch reporting the wrong voltage or maybe the right voltage? BIOS voltages are 1.1v and 1.125 for the system. But i noted oddities in the voltages reported when various apps are running at the same time. Even so if the BIOS says 1.1 the voltage could still very well be 1.2 or higher in reality just food for thought.

All I know is that ZenTimings sometimes reported my SoC voltage as 1.55 which is obviously a glitch.

 

[Mussels]

I’m not involved in this testing

i screwed that one up

 

[DCB2020]

This is silly every cooler is vastly different, too different for this to have any meaning. How can you make any sense where comparing a 47mm cooler against an 360mm AIO or 16oMM tower cooler.

At least group them into buckets of similar coolers height, orientation, air/vs cool