Review Arctic Liquid Freezer III AIO Review: To put it bluntly, the Liquid Freezer III is unimpressive

[Admin]

Unfortunately, the Liquid Freezer III fails our tests and melts under the heat of Intel’s i7-13700K.

Arctic Liquid Freezer III AIO Review: To put it bluntly, the Liquid Freezer III is unimpressive : Read more

 

[-Fran-]

Come on Albert, I'm sure you're mad they bundled a contact frame and didn't let you use yours!

Joking aside, quite shocking to see these numbers. I'm guessing since you've always been testing with the contact frame, the fact they're bundling it gives them no real advantage over the others which rely on the normal ILM instead.

Would you have time or energy to re-test them in an AMD platform using whatever they bundle for it?

Regards,

 

[helper800]

Very disappointed ...

 

[Albert.Thomas]

Would you have time or energy to re-test them in an AMD platform using whatever they bundle for it?

Unfortunately, no.

I don't think that the results would be significantly different and I'm behind on testing other products because of how much extra retesting this review took.

 

[UnforcedERROR]

This is a spectacular letdown, especially considering it's attractive relative to the previous version. It's somewhat baffling that a company who's been in the enthusiast space as long as Arctic would do so poorly on the follow-up to one of the highest recommended products in years. I'd gladly take a 3rd-gen "ugly" version of the performance was there, and obviously other people felt the same way about the 2nd-gen.

 

[thestryker]

I'm a little disappointed in the results, but they're not really shocking to me. Hardware Canucks had it middle of the pack on their Intel testing, but towards the top on AMD. TPU also had it middle of the road on LGA1700. After seeing the GN review I can't help but wonder if the issue is the contact frame, but there's no way to find out (could try a LFII and if it performs better would answer that question to some degree). I understand why they went with the contact frame design, but I think it was a mistake just the same.

The LFIII really strikes me as the best AMD choice and an average Intel choice. The discounted pricing would probably make it an okay Intel buy, but I don't think I'd buy one with regular pricing.

 

[Giroro]

It can't even freeze liquid.

 

[MoxNix]

It would've been nice to add the Freezer II variations to the comparison charts.

 

[RubberMallet]

I have the 420mm version of this installed on a 14900K. It is able to cool 5.7ghz multicore cinebench runs @ 290W in mid 80s, without any thermal throttling. (This is 15C better than my old AIO which would hit 100C under this load even with a contact frame.) By my estimation, it will cool 325W before throttling and do it quietly. It can cool 200W gaming loads silently.

I dont understand how TH is getting thermal throttling at 230-235W with the 280/360 versions.

As you can see, they are the quietest in the test, which indicates the slowest fans. The best performing AIOs are also loudest, indicating higher power fans. If he tested the RGB versions, those fans are known to have lower rpm than the non-rgb and probably explains the results.

Could the stock fans be that bad? Is the goal to get buyers to purchase additional upgrade fans?

I have always trusted TH; this is the first article that leads me to question that trust. Even with the P12 argb fans, it is impossible to believe that a 360mm thermal throttles at 235W. Even the lowest, spec 360 AIOs from two generations ago did much better than that.

The noise/performance is as expected, but the low headroom is what is hard to believe. Did he have the fans/pump on 100% at max cpu load? The only way to explain this is tester error.

(The cooler a cpu is running, the less watts it will draw at a given clock speed. If the 13700K is maxed out and drawing 235W, that is better than a maxed out 13700K drawing 240W, which indicates it is running hotter. )

They don't list ambient temp (no calibration), so my guess is that their whole database of cooling results are not temp normalized, and worthless.

 

[35below0]

I have always trusted TH; this is the first article that leads me to question that trust. Even with the P12 argb fans, it is impossible to believe that a 360mm thermal throttles at 235W. Even the lowest, spec 360 AIOs from two generations ago did much better than that.

Because your cooler didn't get a nice review?

A few other reviews also showed LF III is underwhelming. TH methodology exposed it further.
There are limits to every test.

I'm not sure how you can use one result to completely lose trust. There's lots to criticise TH articles and reviews over, but you're burying results as worthless even though you have not tested the cooler in other configurations than your own.
The paranoia or accusations of steering users to upgrade packs or away from Arctic are werid.

In addition, the contact frame, price, socket support ar other things listed as cons. And those are not performance related.

Why am i even defending TH ?

 

[RubberMallet]

Because your cooler didn't get a nice review?

A few other reviews also showed LF III is underwhelming. TH methodology exposed it further.
There are limits to every test.

Its not that the results are bad, it is that they are so bad to defy explanation.

Even a cheap 360 from two generations ago outperforms what they showed here.

I suspect their data is fragmented from different testers, different locations, and different understanding of what they are testing.

What I think happened is this:

A 360mm aio (even the worst) is not going to thermal throttle (100C) at 235W. It just won't. Reporting otherwise is misinformation.

What is probably happening is that the 13700K is maxed out at 5.2 multicore in their benchmark, and with the arctic LF3 is drawing 235W, and the "best" cooler is drawing ~240W.

At the same workload, a cooler CPU will consume less watts. So the LF3 might only need 235W to complete the workload, where another cooler might require 240W for the same workload.

At the bottom of the chart, you have junky single tower air coolers that are legitimately maxed out at 100C and unable to provide more.

Their assertion that it thermal throttles at 235W is like asserting that the earth is flat.

 

[35below0]

I didn't do the test so i'll leave that for @Albert.Thomas to answer.

 

[thestryker]

Its not that the results are bad, it is that they are so bad to defy explanation.

Even a cheap 360 from two generations ago outperforms what they showed here.

I suspect their data is fragmented from different testers, different locations, and different understanding of what they are testing.

You're wrong all of the data in that chart is pulled from the same setup.

A 360mm aio (even the worst) is not going to thermal throttle (100C) at 235W. It just won't. Reporting otherwise is misinformation.

What is probably happening is that the 13700K is maxed out at 5.2 multicore in their benchmark, and with the arctic LF3 is drawing 235W, and the "best" cooler is drawing ~240W.

At the same workload, a cooler CPU will consume less watts. So the LF3 might only need 235W to complete the workload, where another cooler might require 240W for the same workload.

At the bottom of the chart, you have junky single tower air coolers that are legitimately maxed out at 100C and unable to provide more.

Their assertion that it thermal throttles at 235W is like asserting that the earth is flat.

You're getting too caught up in the number values and not paying attention to the overall performance characteristics. TPU reviewed the LFIII 240 and found it to be worse than the Fuma 3 and AK620 with unlimited power just like it is on these charts as well. I'd imagine that the reason for the wattage values being different comes down to silicon quality.

Also if you'd bothered to read the article you'd see that it was specified every cooler being listed in that unlimited chart thermal throttled.

 

[Albert.Thomas]

Even a cheap 360 from two generations ago outperforms what they showed here.

DeepCool's LS720 was the first "cheap" AIO that I've tested capable of passing my maximum stress test

I suspect their data is fragmented from different testers, different locations, and different understanding of what they are testing.

My testing methodology is very clearly explained. Each cooler has been tested with a Intel i7-13700K CPU on a MSI Z690 A Pro motherboard with a Thermalright LGA 1700 Contact frame installed. The case is a BeQuiet Silent Base 802 with system fans set to the level 1 setting. Ambient temperatures are strictly enforced at 23 degrees C.

A 360mm aio (even the worst) is not going to thermal throttle (100C) at 235W. It just won't. Reporting otherwise is misinformation.

Many 360mm AIOs fail my maximum strength stress test, that finding was exactly why I chose this setup for testing.

At the same workload, a cooler CPU will consume less watts. So the LF3 might only need 235W to complete the workload, where another cooler might require 240W for the same workload.

This is true due to electrical leakage, but Arctic's offerings aren't capable of this. They hit TJMax in my tests, and throttle to varying degrees. If they were capable of keeping the CPU under TJMax, I would have shown the CPU temperature instead.

I have the 420mm version of this installed on a 14900K. It is able to cool 5.7ghz multicore cinebench runs @ 290W in mid 80s, without any thermal throttling. (This is 15C better than my old AIO which would hit 100C under this load even with a contact frame.) By my estimation, it will cool 325W before throttling and do it quietly. It can cool 200W gaming loads silently.

Unless you are running the exact same test configuration as mine at the same ambient temperature, your results are not comparable.

That said, I've tested the 14900K on a few MSI motherboards and I have yet to test any AIO that's capable of running Cinebench without thermal throttling to some extent. I will add that I've only tested a single 420mm AIO, so there's a chance that a good one could handle it. I highly suspect your 14900K is tuned, which can't be compared to a "stock" setting.

They don't list ambient temp (no calibration), so my guess is that their whole database of cooling results are not temp normalized, and worthless.

With all due respect, I very clearly label the ambient temperature of 23C. Unlike certain of my competitors, that ambient temperature is strictly enforced.

 

[helper800]

DeepCool's LS720 was the first "cheap" AIO that I've tested capable of passing my maximum stress test

My testing methodology is very clearly explained. Each cooler has been tested with a Intel i7-13700K CPU on a MSI Z690 A Pro motherboard with a Thermalright LGA 1700 Contact frame installed. The case is a BeQuiet Silent Base 802 with system fans set to the level 1 setting. Ambient temperatures are strictly enforced at 23 degrees C.

Many 360mm AIOs fail my maximum strength stress test, that finding was exactly why I chose this setup for testing.

This is true due to electrical leakage, but Arctic's offerings aren't capable of this. They hit TJMax in my tests, and throttle to varying degrees. If they were capable of keeping the CPU under TJMax, I would have shown the CPU temperature instead.

Unless you are running the exact same test configuration as mine at the same ambient temperature, your results are not comparable.

That said, I've tested the 14900K on a few MSI motherboards and I have yet to test any AIO that's capable of running Cinebench without thermal throttling to some extent. I will add that I've only tested a single 420mm AIO, so there's a chance that a good one could handle it. I highly suspect your 14900K is tuned, which can't be compared to a "stock" setting.


With all due respect, I very clearly label the ambient temperature of 23C. Unlike certain of my competitors, that ambient temperature is strictly enforced.

Considering the GN review of the contact plate pressure patterns, there is clearly an issue with uneven pressure using a leaf spring tensioner on a plastic frame causing a lot of flex. When you use a lot of force to sandwich a spherical tensioner object that is steel against a flat plastic frame you will get warping on that flat plastic frame. I wonder if you place an appropriately sized flat steel spacer between the frame and the leaf spring tensioner if there would be less warping in the frame and a significantly better contact patch on the IHS. I hope that makes sense lol.

@Albert.Thomas do you have and suspicions as to why the LF III coolers are underperforming compared to even cheap air coolers? I cannot imagine that Arctics LF III pump/micro-fin/radiator design is so badly engineered that with such a modification as above it would not improve in performance at least a little.

 

[thestryker]

@Albert.Thomas do you have and suspicions as to why the LF III coolers are underperforming compared to even cheap air coolers? I cannot imagine that Arctics LF III pump/micro-fin/radiator design is so badly engineered that with such a modification as above it would not improve in performance at least a little.

FWIW my guess is that it's almost entirely up to mounting mechanism. If you look at how they perform on AM5 they are very good, but they also have an offset mount. It's going to be very interesting to see how the LFIII performs when Intel brings tiles to desktop. If they perform comparatively better there then we'll know it is about the mounting.

 

[Albert.Thomas]

@Albert.Thomas do you have and suspicions as to why the LF III coolers are underperforming compared to even cheap air coolers? I cannot imagine that Arctics LF III pump/micro-fin/radiator design is so badly engineered that with such a modification as above it would not improve in performance at least a little.

Users in my Discord server have mentioned that better fans have improved their thermal performance, but I can't really say.

Keep in mind that *all* of my results are performed with a contact frame. Someone who tests without a contact frame might have worse comparison results for the other coolers.

 

[RubberMallet]

DeepCool's LS720 was the first "cheap" AIO that I've tested capable of passing my maximum stress test

My testing methodology is very clearly explained. Each cooler has been tested with a Intel i7-13700K CPU on a MSI Z690 A Pro motherboard with a Thermalright LGA 1700 Contact frame installed. The case is a BeQuiet Silent Base 802 with system fans set to the level 1 setting. Ambient temperatures are strictly enforced at 23 degrees C.

Many 360mm AIOs fail my maximum strength stress test, that finding was exactly why I chose this setup for testing.

This is true due to electrical leakage, but Arctic's offerings aren't capable of this. They hit TJMax in my tests, and throttle to varying degrees. If they were capable of keeping the CPU under TJMax, I would have shown the CPU temperature instead.

Unless you are running the exact same test configuration as mine at the same ambient temperature, your results are not comparable.

That said, I've tested the 14900K on a few MSI motherboards and I have yet to test any AIO that's capable of running Cinebench without thermal throttling to some extent. I will add that I've only tested a single 420mm AIO, so there's a chance that a good one could handle it. I highly suspect your 14900K is tuned, which can't be compared to a "stock" setting.


With all due respect, I very clearly label the ambient temperature of 23C. Unlike certain of my competitors, that ambient temperature is strictly enforced.

There is womething wrong with your understanding.

My "tuned" 14900K has an undervolt, but that doesn't influence how much the cooler is able to handle. With my undervolt, 5.6 multicore scores 41200 in cinebench @ 260W. Without the undervolt, it scores same 41200 in cinebench, but draws 275W. It will run hotter yes, but the LF3 is able to handle the additional heat without throttling.

It handles 260 watts in low 70s, and 275 watts in mid 70s. Throttling happens at 100C, so there is still a ton of headroom.

(Just for a sanity check so you understand what I'm saying, I have removed the Intel stock power limits and am running 7-22W higher, and still have roughly 28C headroom before my 14900K will even think of throttling. And you are claiming that the 13700K throttles before even reaching the Intel power limit.)

While GN and other sites think the performance is underwhelming, they aren't claiming it throttles at 235W where a typical air cooler does, that is just ridiculous. Like I said, when you see it drawing 235W for a particular workload, and another cooler draws 240W, that just means the cpu is operating more efficiently. Unless you are talking about a cheap, single tower, single fan cooler.

The results for the LF3 are midpack because it ships with mild fans.

The reason you think it throttles at 235W has to be a problem with setup, or fundamentally not understanding what you are testing.

This is just crazy, bad, impossible wrong. You need to retract the articule until you figure out what is happening. I don't think your error is intentional, I don't work for Arctic or give a damn about them, but your reputation is more likely to take a hit when you figure out the mistake you made.

Whats worse about this, is it lends credence to the "air coolers are almost as good as AIOs" when that is absolutely never the case unless you are talking about a 240aio. The people reading this will find their confirmation bias and hit the reddits armed with this new misinformation.

 

[Albert.Thomas]

My "tuned" 14900K has an undervolt, but that doesn't influence how much the cooler is able to handle. With my undervolt, 5.6 multicore scores 41200 in cinebench @ 260W. Without the undervolt, it scores same 41200 in cinebench, but draws 275W. It will run hotter yes, but the LF3 is able to handle the additional heat without throttling.

If you are only drawing 260-275W in Cinebench R23, something is funky. I am just as skeptical of this report as you are of my Liquid Freezer results, unless your motherboard is applying very conservative settings (maybe using the "air cooler" setting?) An i9-14900K will easily use over 320W in Cinebench R23 when power limits are removed.

This is just crazy, bad, impossible wrong. You need to retract the articule until you figure out what is happening. I don't think your error is intentional, I don't work for Arctic or give a damn about them, but your reputation is more likely to take a hit when you figure out the mistake you made.

Absolutely not. There is no error here, I tested 3 different units and remounted them multiple times. I firmly stand by these results.

 

[35below0]

The reason you think it throttles at 235W has to be a problem with setup, or fundamentally not understanding what you are testing.

You should think about your own reputation before attacking others who are valued and respected members of the community. Even if you're a cooling system tester or expert yourself.

 

[helper800]

There is womething wrong with your understanding.

My "tuned" 14900K has an undervolt, but that doesn't influence how much the cooler is able to handle. With my undervolt, 5.6 multicore scores 41200 in cinebench @ 260W. Without the undervolt, it scores same 41200 in cinebench, but draws 275W. It will run hotter yes, but the LF3 is able to handle the additional heat without throttling.

It handles 260 watts in low 70s, and 275 watts in mid 70s. Throttling happens at 100C, so there is still a ton of headroom.

(Just for a sanity check so you understand what I'm saying, I have removed the Intel stock power limits and am running 7-22W higher, and still have roughly 28C headroom before my 14900K will even think of throttling. And you are claiming that the 13700K throttles before even reaching the Intel power limit.)

While GN and other sites think the performance is underwhelming, they aren't claiming it throttles at 235W where a typical air cooler does, that is just ridiculous. Like I said, when you see it drawing 235W for a particular workload, and another cooler draws 240W, that just means the cpu is operating more efficiently. Unless you are talking about a cheap, single tower, single fan cooler.

The results for the LF3 are midpack because it ships with mild fans.

The reason you think it throttles at 235W has to be a problem with setup, or fundamentally not understanding what you are testing.

This is just crazy, bad, impossible wrong. You need to retract the articule until you figure out what is happening. I don't think your error is intentional, I don't work for Arctic or give a damn about them, but your reputation is more likely to take a hit when you figure out the mistake you made.

Whats worse about this, is it lends credence to the "air coolers are almost as good as AIOs" when that is absolutely never the case unless you are talking about a 240aio. The people reading this will find their confirmation bias and hit the reddits armed with this new misinformation.

There are so many variables between any two given setups its impossible to get continuity between testing. You have the individual silicon and ASIC quality between chips, flatness of the IHS, and slight differences in TIM between the chip and the IHS. Then you have to come to gripes that every mounting of a cooler is uniquely a bit better or worse than the last. Slightly more TIM used in a slightly different distribution between the IHS and the cold plate. Then you have cold plate flatness and quality of production of the AIO itself which will have very small defects all over the place compared to its 3d design on a computer. Ambient air temperatures, airflow of case or testbench, differences in fill between each individual AIO, motherboard choice and all of its implications, et cetera, et cetera. You can only control so many variables like using the same exact chip, and test bench between testing, but even with all the control in the world there are tons of small variables that you cannot keep exactly the same between testing.

With all that said, what we can derive from @Albert.Thomas testing is with three different 360mm LF III AIO's on a 13700k with his testing methodology, we get said results. If he got 3 bad AIOs and you have a 'good' one then we can still say there are some major problems with the cooler on the consistency side of production, but more than likely its just a mediocre cooler that runs very quiet.

 

[Albert.Thomas]

You should think about your own reputation before attacking others who are valued and respected members of the community. Even if you're a cooling system tester or expert yourself.

I appreciate your intention here, but I absolutely welcome any and all criticism - as long it is tactfully given. No one, no matter how much of an "expert" they may seem, should be considered above making mistakes. I am just as human as anyone else.

 

[Viking2121]

I still wonder what it would do with other fans on it that can move more air, I feel like the fans are mostly the blame on these usually for the sake of being quiet.

 

[helper800]

I still wonder what it would do with other fans on it that can move more air, I feel like the fans are mostly the blame on these usually for the sake of being quiet.

I do wonder if Arctic instead decided to use three p12 max fans instead of the ones it came with if it would have performed significantly better...

 

[35below0]

I appreciate your intention here, but I absolutely welcome any and all criticism - as long it is tactfully given. No one, no matter how much of an "expert" they may seem, should be considered above making mistakes. I am just as human as anyone else.

Yeah, criticism. But calling the results worthless, ALL test results worthless, and questioning whether you fundamentally understand what you are doing goes beyond criticism. Accusations of publishing misinformation as well.

That's where it went too far. Questioning results is fair, esp when coolers underperform or strongly outperform rivals (Peerless Assasin, Phantom Spirit EVO).