Adding water cooling to the Mac Studio does surprisingly little

Posted:
in Current Mac Hardware
A YouTube video shows how a challenging attempt to add water-cooling to the Mac Studio ultimately doesn't make much of an improvement to performance.

Water-cooled Mac Studio by Linus Tech Tips


Experienced PC builders will be familiar with the benefits of water cooling, including the potential to offer better thermal conditions than air-based methods, the possibility of improved chip performance, and a reduction in noise. In a YouTube video posted on Monday, an attempt was made to install a water cooling system on Apple's Mac Studio with a M1 Max processor.

The plan Linus Tech Tips had was to remove the existing cooling system from a Mac Studio, replace it with a water cooling-based version, and see how it performs. The channel had two identical Mac Studios on hand, allowing for a more direct comparison with like-for-like base units.

Replacing the cooling system in the Mac Studio seemed reasonably straightforward, as the existing system is a heatsink and a large blower section taking up half the internal volume of the enclosure. Removing the component wasn't difficult, but did involve disconnecting a proprietary connection that supplied power.

Holes were made through the Mac Studio's case for the water cooling system (via LinusTechTips)
Holes were made through the Mac Studio's case for the water cooling system (via LinusTechTips)


After removing the heatsink and fans from a large section that comes into contact with the M1 chip using tools ranging from a heat gun to a milling machine, a water block was attached to the remaining plate.

To actually pump the water around the system, the plan involved milling numerous holes into the top of the Mac Studio's aluminum enclosure, allowing cables and pipes to pass through. Due to the lack of space inside, the bulk of the water cooling loop had to be outside the Mac Studio itself.

Since the channel primarily deals with gaming-centric videos, RGB lighting was included in the water-cooling system. Corsair components were chosen for RGB elements, since it had a controller app that works in macOS.





After multiple days, including repeated disassembly and reassembly, the water-cooled Mac Studio did ultimately work. It also was able to run side-by-side with a stock Mac Studio of the same specification.

The results were mixed.

On the positive side, the Mac Studio was cooled down by 30 degrees versus stock. However, in Cinebench R23, the water-cooled Mac Studio scored 12,056 while the regular model managed 12,016. A second test resulted in a score of 12,104, representing a performance improvement of 0.7% -- well within the margin of error of every benchmark.

The low performance change was cited as the M1's "turbo boost" being "really terrible," but that it ultimately doesn't matter "because they sip power and are well cooled."

Obviously, the process of adding water cooling is expensive, difficult, and problems generated by the process won't be covered by either Apple's warranty or under AppleCare. At the very least, questions would be raised by Apple support about the holes in the casing.

While we're glad that the channel tried the mod, given the difficulty for the average user to acquire spare parts and to install them if they break, as well as the whole process and the minimal payoff, AppleInsider strongly advises against attempting what the video demonstrates.

Read on AppleInsider
«1

Comments

  • Reply 1 of 30
    DAalsethDAalseth Posts: 2,197member
    Well whattayaknow, Apple's engineers really know what they are doing. /s
    jas99kurai_kagewatto_cobrawilliamlondonFileMakerFellerolsStrangeDaysjony0
  • Reply 2 of 30
    What about liquid nitrogen cooling?
    watto_cobrawilliamlondonolschia
  • Reply 3 of 30
    I'm not sure what they were expecting, the bulk of the interior is a large purpose-built cooling system...
    DAalsethwatto_cobrawilliamlondonFileMakerFellerjony0
  • Reply 4 of 30
    Based on earlier tests I'm pretty sure that Apple is throttling the system regardless of chip temperature. There's no reason in the world why a M1 Max in a MBP should run at the same speed as a M1 Max in a Studio with a desktop-class cooling system and when not limited by battery life... but it does.
    darkvaderwatto_cobrawilliamlondonFileMakerFeller
  • Reply 5 of 30
    It’s hard to stomach these guys knowing that they despise Apple and its products. 

    And their hyperbolic tone is hard to take. 

    Conclusion: They’re goobers first class. 
    DAalseththtforegoneconclusionwatto_cobrawilliamlondonolsStrangeDaysjony0
  • Reply 6 of 30
    chadbagchadbag Posts: 1,732member
    So if I take a PC and water cool it, without going in and changing clock speeds, RAM timings, perhaps voltages, should I expect to see much of a difference?   Probably not.  

    At least as far as I understand it, not having done water cooling since the very late 90s and early 2000s, is that we water  cool in order to be able to boost clock speeds, adjust timings, and boost voltage (to support the first two).  Water cooling by itself doesn't boost performance.  It gives you expanded headroom to change and boost performance parameters. 

    Unless these guys figure out how to boost the clock speed and/or adjust the memory timings to be more aggressive, I'm not sure what they were expecting from this stunt.  


    DAalsethuraharathtwatto_cobramuthuk_vanalingamwilliamlondonstompyFileMakerFellerolschia
  • Reply 7 of 30
    hmlongco said:
    Based on earlier tests I'm pretty sure that Apple is throttling the system regardless of chip temperature. There's no reason in the world why a M1 Max in a MBP should run at the same speed as a M1 Max in a Studio with a desktop-class cooling system and when not limited by battery life... but it does.
    There is no throttling, it runs at the same speed because the M1 Max doesn’t consume like a desktop class processor. That is its nature, it isn’t designed to behave like Intel CPUs with turbo boosts, with high clock speeds.
    thtwatto_cobrawilliamlondonracerhomie3olsjony0
  • Reply 8 of 30
    DAalsethDAalseth Posts: 2,197member
    chadbag said:
    So if I take a PC and water cool it, without going in and changing clock speeds, RAM timings, perhaps voltages, should I expect to see much of a difference?   Probably not.  

    At least as far as I understand it, not having done water cooling since the very late 90s and early 2000s, is that we water  cool in order to be able to boost clock speeds, adjust timings, and boost voltage (to support the first two).  Water cooling by itself doesn't boost performance.  It gives you expanded headroom to change and boost performance parameters. 

    Unless these guys figure out how to boost the clock speed and/or adjust the memory timings to be more aggressive, I'm not sure what they were expecting from this stunt.  


    That’s a good point. Water cooling itself would do little unless the system were thermally limited. 
    watto_cobraFileMakerFellerolschiajony0
  • Reply 9 of 30
    "low performance change was cited as the M1's "turbo boost" being "really terrible,""
    How can they call something that does not exist 
    terrible? Seems like they missed how the M1 Max is meant to behave.
    Also weird is that when they first tested that machine a few months ago, the Cinebench score was above 12300, just like most other cinebench scores out there, so one would assume that there isn’t really any advantage, on the contrary. 
    watto_cobrawilliamlondonFileMakerFellerthtchia
  • Reply 10 of 30
    No overclocking leads to no performance increase. Who woulda thunk it?
    uraharathtwatto_cobraFileMakerFellerchia
  • Reply 11 of 30
    gatorguygatorguy Posts: 23,423member
    I like seeing articles like this. I'm edging towards picking up the Mac Studio for my photo processing workflow and the report on heat mitigation helps seal it for me. I had fully intended to get one of Apple's refurbished ones for a couple hundred off several weeks ago and got sidetracked. Worst case I would have sold it on for probably no loss. When I came back to the cart they were no longer available. 
    edited September 20
  • Reply 12 of 30
    It would be nice, in the article and in the video, to be clear about whether the Mac Studios are M1 Max or M1 Ultra, since the M1 Ultra have a much beefier cooling system/
    I think the video is about M1 Max, which is the least relevant comparison.
    watto_cobrawilliamlondonFileMakerFeller
  • Reply 13 of 30
    Mike WuertheleMike Wuerthele Posts: 6,516administrator
    jeromec said:
    It would be nice, in the article and in the video, to be clear about whether the Mac Studios are M1 Max or M1 Ultra, since the M1 Ultra have a much beefier cooling system/
    I think the video is about M1 Max, which is the least relevant comparison.
    Paragraph two.

    "In a YouTube video posted on Monday, an attempt was made to install a water cooling system on Apple's Mac Studio with a M1 Max processor."

    The M1 Ultra core temps are often lower under load than the Max is. The Max is the hotter of the two, because of that cooling system choice.
    edited September 20 watto_cobramuthuk_vanalingamDAalsethwilliamlondonFileMakerFellerjeromecolsStrangeDays
  • Reply 14 of 30
    sflocalsflocal Posts: 5,990member
    I watched the video as a curiosity and it didn't disappoint.  If anything, it just reaffirmed for me the silliness that is the PC-fanboy mantra of over-complicating things.  It's silly, and in the end it proved that Apple actually did something right with thermal designs.


    FileMakerFellerolsStrangeDaysjony0
  • Reply 15 of 30
    chadbag said:
    Water cooling by itself doesn't boost performance.  It gives you expanded headroom to change and boost performance parameters.
    Yes, but maybe not.  That was definitely true in the old days when CPUs ran at a fixed clock speed.  Extra cooling would allow you to boost the clocks without overheating the chips, but you don't get more performance unless you actually do boost these frequencies.

    But modern CPUs have variable frequencies, even in their stock configurations.  They run at a baseline frequency, but they will "turbo boost" to higher frequencies when software demands more performance.  They will also run at reduced frequencies when idle.  When temperatures get too high for the cooling solution, they will "thermally throttle", limiting the maximum frequency they can boost to, the amount of time they can run at boosted frequencies, and may (if necessary) even force it to run at frequencies below the baseline.

    With modern chips (at least high-end ones from Intel and AMD), an air-cooled system generally can't keep up with the thermal output of the chip running at its maximum turbo boost frequency.  When you add more/better cooling, the CPU can remain boosted for longer periods of time before throttling, and therefore perform better overall.  With enough cooling so the CPU never throttles, it can remain running at its boost frequency pretty much all the time, which is pretty much as good as it will get without "overlocking" (that is, increasing the maximum boost frequency beyond manufacturer specs).

    And if you've got a separate GPU, all of the above applies to it as well.

    What the Linus experiment proves is that Apple's air-cooling system is sufficient to prevent thermal throttling even when the M1 is running at its maximum performance.  So extra cooling won't benefit it unless you figure out a way to overclock it beyond what the stock air-cooler can handle.  Which we could have predicted, because prior tests show that the Mac Studio doesn't experience thermal throttling under load.

    This is in contrast to (as the video mentioned) a MacBook Air, where the completely passive cooling system isn't sufficient.  If the M1 in there is pushed hard, it will thermal-throttle.  Improving its cooling (e.g. with a fan, as with the MacBook Pro, or with some Frankenstein water cooling system like Linus is fond of) will improve performance, because the CPU will be able to run at maximum load for longer (perhaps indefinitely) before throttling.
    edited September 20 roundaboutnowFileMakerFellerchadbag
  • Reply 16 of 30
    Maybe next time try an actual Lava Lamp. I mean the wax would make a great thermal sink which then bubbles up to take the heat away from the SOC. 
    edited September 20 FileMakerFeller
  • Reply 17 of 30
    shamino said:
    chadbag said:
    Water cooling by itself doesn't boost performance.  It gives you expanded headroom to change and boost performance parameters.
    Yes, but maybe not.  That was definitely true in the old days when CPUs ran at a fixed clock speed.  Extra cooling would allow you to boost the clocks without overheating the chips, but you don't get more performance unless you actually do boost these frequencies.

    But modern CPUs have variable frequencies, even in their stock configurations.  They run at a baseline frequency, but they will "turbo boost" to higher frequencies when software demands more performance.  They will also run at reduced frequencies when idle.  When temperatures get too high for the cooling solution, they will "thermally throttle", limiting the maximum frequency they can boost to, the amount of time they can run at boosted frequencies, and may (if necessary) even force it to run at frequencies below the baseline.

    With modern chips (at least high-end ones from Intel and AMD), an air-cooled system generally can't keep up with the thermal output of the chip running at its maximum turbo boost frequency.  When you add more/better cooling, the CPU can remain boosted for longer periods of time before throttling, and therefore perform better overall.  With enough cooling so the CPU never throttles, it can remain running at its boost frequency pretty much all the time, which is pretty much as good as it will get without "overlocking" (that is, increasing the maximum boost frequency beyond manufacturer specs).

    And if you've got a separate GPU, all of the above applies to it as well.

    What the Linus experiment proves is that Apple's air-cooling system is sufficient to prevent thermal throttling even when the M1 is running at its maximum performance.  So extra cooling won't benefit it unless you figure out a way to overclock it beyond what the stock air-cooler can handle.  Which we could have predicted, because prior tests show that the Mac Studio doesn't experience thermal throttling under load.

    This is in contrast to (as the video mentioned) a MacBook Air, where the completely passive cooling system isn't sufficient.  If the M1 in there is pushed hard, it will thermal-throttle.  Improving its cooling (e.g. with a fan, as with the MacBook Pro, or with some Frankenstein water cooling system like Linus is fond of) will improve performance, because the CPU will be able to run at maximum load for longer (perhaps indefinitely) before throttling.
    Gotta love the triumph of marketing that is "Turbo Boost" - the reality is that the CPU is always throttled but if you can put up with the extra noise (and energy cost) of running the cooling system flat out the throttling gets reduced.

    Apple's approach is that the chip is only throttled when it has to be, but the attitude of LTT and the PC enthusiast crowd is that "the Turbo Boost sucks." Don't try to bend the spoon, instead recognise that there is no Turbo Boost.
    chadbagStrangeDays
  • Reply 18 of 30

    Obviously, the process of adding water cooling is expensive, difficult, and problems generated by the process won't be covered by either Apple's warranty or under AppleCare. At the very least, questions would be raised by Apple support about the holes in the casing.
    "Dude! On the way here I was attacked by some bald guy who had sharks with fricken laser beams! How is that relevant to AppleCare???"
  • Reply 19 of 30
    mpantonempantone Posts: 1,893member
    chadbag said:
    So if I take a PC and water cool it, without going in and changing clock speeds, RAM timings, perhaps voltages, should I expect to see much of a difference?   Probably not.  

    At least as far as I understand it, not having done water cooling since the very late 90s and early 2000s, is that we water  cool in order to be able to boost clock speeds, adjust timings, and boost voltage (to support the first two).  Water cooling by itself doesn't boost performance.  It gives you expanded headroom to change and boost performance parameters. 

    Back when you did it, yes, that was the basic motivation.

    Contemporary PC silicon isn't as overclockable. Much of the overclock capacity is built into the boost clock so overclocking today gives much lower performance gains (as a percentage over stock settings) than 10-15 years ago. This is applicable to GPUs as well as CPUs.

    Today the main benefit of using a liquid cooled loop is better acoustics since the fans run slower due to water's superior thermal capacity as a cooling material. This is more useful when the silicon generates lots of heat which M1-series SoCs do not do. This is one of the advantages of Apple's emphasis on performance-per-watt.

    By contrast, Intel especially had thermal issues. Stuck on their geriatric 14nm process node, they needed to make bigger and bigger dies to keep up in their performance battle with AMD. This is why Apple uses Intel as an example during the performance comparisons with Apple Silicon.

    Also, some people like the aesthetics of a liquid loop, often when it features addressable RGB lighting, something this vlogger did not include.

    Unless these guys figure out how to boost the clock speed and/or adjust the memory timings to be more aggressive, I'm not sure what they were expecting from this stunt.

    Well, they wanted pageviews/ad impressions more than anything else.

    Whether they wanted to or not, they also proved that Apple hardware engineers actually know something about elegantly cooling electronics.

    You can make a high performance PC without a massive tower cooler with two 120mm fans or an expensive custom loop. PCs are really holistic systems and Apple did an admirable job at combining high-performance variants of the M1-series silicon with a thermal solution that is capably matched with it.
    edited September 20 chadbag
  • Reply 20 of 30
    crowleycrowley Posts: 10,360member
    It’s hard to stomach these guys knowing that they despise Apple and its products. 

    And their hyperbolic tone is hard to take. 

    Conclusion: They’re goobers first class. 
    They really don’t, they’re very complimentary of a lot of Apple products, especially notebooks and especially since M1. The MBP is one of their baselines for comparing all other notebooks too. 

    They’re tinkerers, so they don’t like the lack of repairability, expansion, and proprietary technology, but it’s a con rather than a condemnation.
Sign In or Register to comment.