Why is this no longer popular?

Discussion in 'Extreme Cooling Solutions' started by M76, Nov 14, 2019.

  1. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    I remember in the 2000s everyone was raving about extreme cooling solutions, like vapochill and peltier devices. What happened to this genre of the overclockers club?

    Why is the scene dead?
     
  2. Ehren8879

    Ehren8879 [H]ardness Supreme

    Messages:
    4,261
    Joined:
    Sep 24, 2004
    Once the novelty and newness wore off, the hassle wasn't really worth the expense or the performance.
     
    Gideon, Keljian, ChoGGi and 2 others like this.
  3. jmilcher

    jmilcher [H]ardness Supreme

    Messages:
    4,387
    Joined:
    Feb 3, 2008
    Back then, there was also a ton more overclocking headroom. So these extreme solutions were fun.
     
  4. Mega6

    Mega6 2[H]4U

    Messages:
    2,199
    Joined:
    Aug 13, 2017
    Now u get nothing on OC, maybe less. What's the point?
     
  5. N4CR

    N4CR [H]ardness Supreme

    Messages:
    4,101
    Joined:
    Oct 17, 2011
    I'm very interested to see how 7nm fares with a direct WC pelt setup and a custom spreader. Or custom phase.
    252w on LN2, -80c, 5GHz all core in blender at 1.42v on a damn 3950x.
    Now that would be even more of a rendering and gaming monster.

    Maybe just ignore the shitty chiplet and treat it like an overpriced 3800x.. Wonder what it could do then. With a very special heatspreader there could be some interesting results to be had, thermal gradient becomes a bit of a workaround when you are constrained due to density and material.

    What I'm banking on is undervolting enough to get it as close as possible to -140c and see what it does then.. The efficiency of 7nm is astounding. Getting heat out when running at higher temperatures is the hard part.
     
    Neapolitan6th likes this.
  6. Nenu

    Nenu [H]ardened

    Messages:
    19,039
    Joined:
    Apr 28, 2007
    Peltier devices (heat pumps) died because it took at least a magnitude more power to cool a device and you had to get rid of the original heat plus that of the peltier or the damn thing would melt.
    Phase change suffered problems with complexity, size, initial cost, longevity and cost to repair - if you could find someone to do it.
    Simplicity has proven key so far.
     
    kirbyrj and drescherjm like this.
  7. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    That's why it's extreme, because it is complex and not efficient. But hardcore OC was never about efficiency. If power efficiency was a concern then you wouldn't OC at all.

    As for phase change coolers I don't think it's more complex than water cooling. Only instead of a pump you have a compressor. Otherwise it is the same basic setup, some tubes, a radiator and a cooling block.
     
  8. N4CR

    N4CR [H]ardness Supreme

    Messages:
    4,101
    Joined:
    Oct 17, 2011
    And phase change is incredibly reliable if done right, fridges and AC is just some of the examples. Some crazy fuckers use cascade phase change.. 2kW to cool a PC lol.
    Water chillers are pretty cheap these days though.


    This company says they are out of production but I bet if you got in contact they could make something for you.
    http://shopeu.dimastech.com/en/cooling-technologies-dimastech
    upload_2019-11-16_3-51-18.png
     
    M76 likes this.
  9. Nenu

    Nenu [H]ardened

    Messages:
    19,039
    Joined:
    Apr 28, 2007
    With CPU power use increasing it became infeasible.
    A much higher power PSU, huge heatsink and fans.
    GPUs were already infeasible.
    It lost popularity for good reason.

    Its way more complex.
    It requires an exotic coolant, pressurised pipes, placement of the radiator above the heat sinks, more space, ...
    The Inability to change the sinks without de-pressurising, fitting the new block and re-repressurising.
    To add another cooling block (to cool your GPU / CPU as well, or to add more GPUs) will need a full re-spec and replacement.
    If you exceed the cooling capacity its a much bigger pita than water to add a better compressor, fatter pipe or add a radiator.
    They are impractical to work on unless you are skilled.
    They arent quiet, ...
     
  10. Kardonxt

    Kardonxt 2[H]4U

    Messages:
    3,026
    Joined:
    Apr 13, 2009
    For extreme OCers LN2 is much cheaper and gives much better results. There never has been a huge market for people who care so much about OC that they need to run sub ambient temp cooling 24/7 and they are willing to pay a huge premium for it. The price to performance lands on custom loops or LN2 depending on your goals. Phase change and peltier are a weird middle ground that are way more expensive.

    CPUs have become so powerful and so inexpensive anyone who really needs that level of power can just buy a monster cpu that fits their needs for less than a phase change setup costs and none of the hassle.
     
  11. N4CR

    N4CR [H]ardness Supreme

    Messages:
    4,101
    Joined:
    Oct 17, 2011
    It is possible to design a removable system, there are flexible ones these days.
    upload_2019-11-17_1-0-1.png

    Personally it's more the 'smearing dielectric grease all over a $500 mobo' part that puts me off.

    CPU power has not increased much for a decade or more. Yes with pelts the overhead is high, but it's also much easier to deal with if power cost isn't an issue.
    These 7nm cpus are so efficient that it may become feasible again to do this.. 5GHz, -80c, 252W and 16 cores ffs. Almost halve that with 8 cores... now you are 125W!!! That's not a huge heat load. Wonder what 65W TDP cpus could do....?
    Newer AC compressors are also pretty quiet but agreed they are a little louder than a good water setup. If you really are [H] you'd just mount it remotely anyway or put it in an isolating, sound deadening box.. I know how to make a 750Kg 2 cylinder WW2 diesel engine run silent, so an AC compressor is peanuts.

    Where I'm going with this, is it is still feasible to daily drive if you are bored enough to do it.
    Current water loops are barely handling the heat density of 7nm when running it hard. So Ambient cooling might just be ready for a comeback...

    Water loops, lighting and other rice are so 2004-7 and done to death, meanwhile phase is a unicorn for most newer players and 'gamer rgb' types. If it can give better bench e-peen or 10fps higher in 480p CS, it's worth it for some.
     
    Neapolitan6th likes this.
  12. N4CR

    N4CR [H]ardness Supreme

    Messages:
    4,101
    Joined:
    Oct 17, 2011
    Sure, LN2 is much easier. But also much more babysitting to keep it going. Phase is continuous once setup. -130- -150 is about where Zen2 will run anyway, you are not using super crazy temperatures... phase can get close enough and probably can be daily driven.
    The other counter is that for people who want to push the best/mega CPUs... this is for them.
    It's been my dream to do a phase setup and I will do it one day. And I'll post it on here for you to all enjoy and/or laugh at my stupidity.
     
  13. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    You can't daily an LN2 setup. I'm not talking about going for records, just a system where you push the envelope. Frankly Air and water are boring and don't give that great results anyway. I was always a fan of unique and unusual solutions, damned be the work necessary, if it's 10% better it was worth it.

    I just feel like doing a custom water loop gives negligible results over an AIO, and already a hassle to do. So why not spice things up with adding a peltier to it. With the heat density of current cpus, I think peltier devices could be used to transfer the heat quickly from the die to a larger cold plate that is water cooled.
     
  14. Nenu

    Nenu [H]ardened

    Messages:
    19,039
    Joined:
    Apr 28, 2007
    You must take it to a licensed workshop to have it de-gassed and re-gassed.
    Its illegal to do yourself unless you are licensed and have the equipment.

    It may be possible to use it on 125W, but much higher becomes impractical due to the amount of heat that needs releasing from the peltier and it is released locally.
    You could fashion an air duct round the peltiers cooler to draw hot air elsewhere. I use this method to get cold air to my GPU.
    But you need a HUGE cooler with a 125W capable peltier, with a LOT of air flow to keep the peltiers hot side cool.
    It will be noisy.
    (and this presumes the problem below isnt present, in reality its not simple or easy to use a 125W peltier on a CPU long term)

    The peltier base will need adapting (ie a tapered metal block) to fit the much smaller 35mm square CPU die and will limit the amount of cooling it can do.
    The 135W peltier I saw last had a 5cm square base. Higher power have a quantitatively larger base.
    The CPU die lid area remains the same on the higher power models meaning heat transfer will be worse throttled resulting in the need for an even larger peltier to get a higher deltaT.
    Power use and noise skyrocket while effectiveness reduces drastically.

    Note:
    This doesnt take into account breaking the TDP of a 125W CPU.
    To do that will require starting with a larger peltier, compounding the problems.

    https://www.tweaktown.com/reviews/6...-thermoelectric-cpu-cooler-review/index7.html
    This review using the 84W TDP 4770K cooled by an unknown wattage peltier/heatsink combo produced ok results at low load (close to the best air coolers) but at high load it dropped back to middle of the pack air cooler.
    In both cases it made more noise vs the Noctua D15 despite not besting it, 4dB at low load and 17dB overclocked!
    This doesnt directly correlate what can be achieved but it demonstrates its not easy to provide decent cooling and keep quiet, even with only 84W TDP and a modest overclock.


    ps
    When I say 125W peltier, I mean one capable of sinking a 125W load.
    The peltier will likely be at least 2.5x the power of that, the total heat that needs removing is 2.5x or higher than the original load.
    Assuming the peltier is running in its most efficient voltage range.

    edit
    After a lot of editing my post because I have a truly sore head today, not feeling well and a bit hungover on top, perhaps you can help decide.
    I saw this years ago, it took a while to find.
    https://forums.evga.com/TECPeltier-definitions-formulas-temp-estimations-m76115.aspx
    I'm not sure if the final amount of heat given off the hot side (from the calculation) includes the load and the peltiers power use, or whether the load heat must be added to the (calculated) peltiers on the hot plate.
    ie working out how much power needs to be cooled.
     
    Last edited: Nov 16, 2019
    GoldenTiger likes this.
  15. CaptainClueless

    CaptainClueless n00b

    Messages:
    37
    Joined:
    Jun 2, 2016
    You don't think a refrigeration system is more complicated than a water pump, a radiator, and some tubing? Ok...
    Ignoring the whole issue with legal regulations around most refrigerants, there's quite a bit more to it. You don't just open up a reservoir, pour in some refrigerant, and turn on the compressor. That's not how this works at all.
     
  16. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    People have been using refigerators for the better part of the last century. And they are quite reliable. An AIO Water cooler has the same number of components as an AIO phase change cooler would have. There are no additional points of failure. And current refrigerants are environmentally safe. It's not freon anymore.

    Why do you think the first is less complicated?

    Water pump, radiator, and some tubing
    -or-
    Compressor, radiator, and some tubing?
     
    Last edited: Nov 16, 2019
  17. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,151
    Joined:
    Sep 13, 2008
    OC'ing er system can make it more power efficient. OC is not mutuality exclusive with power efficiency.

    -- edit --
    I realized now you used the term "hardcore OC" and thereby by not about all OC's
    I retract my correction as even though my point is valid its not applicable to your statement
     
  18. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    I've never seen a system draw less power with OC than without it. Power draw seems to be linear with increasing frequency, so if you OC at best you can get the same efficiency, but better? You'd need to show actual tests for me to believe that.
     
  19. Kardonxt

    Kardonxt 2[H]4U

    Messages:
    3,026
    Joined:
    Apr 13, 2009
    Like I said in my post, the scene is dead because there is practically no market for people who need to run sub zero temps daily. Im not suggesting anyone run LN2 for daily use. Just that none of these setups are sold anymore because no one was buying them.... basic economics.

    Yes you can build one, Yes it would be cool. But that wasnt the question asked in the OP.
     
  20. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,151
    Joined:
    Sep 13, 2008
    1: Power draw is not the same as efficiency. that you first mistake
    CPU power draw increased close to linear with frequency you are correct
    Howeveer cpu is not the only unit pulling power, if you get 10% more cpu performance for 10% more draw and you taks is 100% cpu speed connstrained. you are using 10% time on the entire system for 10% more power on the cpu.
    aka saving ~10% on the power of the system minus the CPU.

    2: also it is possible to overclock and lower voltage. Voltage has a bigger impact on power draw than Frequency. makign you even able in some cases to reduce power draw and increase performance.
    Lowering voltage was often used to overclock certain Slot A AMD athlon to get them to overclock



    The above is also why it's idiotic when when people lower their clock to save power if they don't lower their voltage as well.
    Lowering clocks to reduce power usage should only be done if it makes you able to lower the voltage as well

    Cause PPL are confusing powerdraw with powerefficiency



    P.S.
    I've never seen a system draw less power with OC than without it
    You are not omniscient. just because you haven't seen does not make it impossible.
    This is why we don't prove a negative.
     
    Last edited: Nov 16, 2019
  21. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    What? You're very confused. It is I that would have to prove a negative in this scenario. Your task is to prove a system does exist that is more power efficient when overclocked, and that's proving a positive.
    And make it something newer than Slot A, as a 20 year old design would be quite irrelevant here when discussing extreme cooling in current year. And no anecdotal evidence, let's see numbers or it never happened.
     
  22. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,151
    Joined:
    Sep 13, 2008
    Yes that is a try at proving the negative. You are drawing towards a conclusion of negativity based on the negative presence on your test sample (experience)
    Aka: I have never seen it so it can't be true.

    Not irrelevant. You statement was ultimate with no exception implied, and is only an example of the to support the argument. Not the argument itself.

    Fact is still OC can in certain cases improve energy efficiency of a system.


    You have still not addressed the fact you mistook energy draw as energy efficiency. Can I assume we agree on this point ? or is you debate going to continued to be skipping argumentative points just to avoid talking about it?
     
    Last edited: Nov 16, 2019
  23. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,816
    Joined:
    Apr 5, 2016
    Sidebar, but of curiosity, is there an established unit for efficiency when it comes to processors? Something like FLOPs/watt or similar? To make any efficiency comparison you'd need a unit like that, wouldn't you?
     
  24. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,151
    Joined:
    Sep 13, 2008
    TBH I am really not sure but I don't see a universal way to do so, because different software will behave differently in differently systems.
    and no you would not an establish units to measure efficiency you can use (time on a task(/(Watt hour used)

    Speeding up the CPU to make the task faster can save you power, as you use the power in a shorter time for the entire system. But only increased the power draw on the CPU
    aka 10% save on entire system power used for a task can be more than the 12% increased power usage on the CPU.
     
  25. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,816
    Joined:
    Apr 5, 2016
    Well, you've still made a unit there - the task in question is just part of the definition of your unit. You have to have a unit to measure improvement. I was just curious if there was a standardized one.

    I'm not aiming to contest anything or make a point here, I've just enjoyed reading the conversation and it got me thinking about efficiency measurements in terms of processing power. You can't rate it like an electromechanical device, because it doesn't do any "work" by the classical definition. So I thought that the data manipulated per watt-hour would be the next way to go.
     
  26. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,151
    Joined:
    Sep 13, 2008
    The issue i think is that benchmarking is always very task relient. that why we don't just do one game and think every other games it going to show the same difference.
    it doesn't help real world that out CPU become 50% faster if we are waiting in storage I/o anyway

    But now we are going don the old benchmark debate about synthetic/to the metal benchmarks vs realworld applications.
    Both have its merits
     
  27. M76

    M76 [H]ardForum Junkie

    Messages:
    9,766
    Joined:
    Jun 12, 2012
    Proving a negative is proving god does not exist.
    Proving a positive is proving god does exist.

    why would it be any different with cpus? You need to prove that a cpu exists that is more energy efficient when overclocked.

    You said some people overlclock for better efficiency, no they undervolt for better efficiency. Big difference.
    I will admit that you might get minor OC from some cpus with less than stock voltage, but that is not what I was thinking when you straight up say overclocking.
    Overclocking to me means getting max possible performance out of the cpu.
    Undervolting is an entirely different art, and I've never heard anyone refer to it as overclocking. And it doesn't automatically mean it is more energy efficient. So I still need real world examples to accept it as fact.


    No, I did not confuse power draw with efficiency, unless you take half my sentence out of context, while not seeing the whole picture. Let me recap:

    I said overclocking increases power draw exponentially, while increasing performance linearly. Which makes it impossible for an overclocked CPU to be more power efficient than a stock one. And I also added: UNLESS you can significantly under-volt it at the same time.

    I'm still waiting on at least one real world example, supported by actual measurements.
    If you cannot cite one, I'll continue under the assumption that it does not exist. As I live under the assumption that none of the thousands of human gods ever existed either.