Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
AlexTheGreatish

This Cooler DRAWS 545W!!? Bad Cooling Ideas #2

Recommended Posts

Posted · Original PosterOP

We got a 545W Peltier cooler... sketchiness ensues.

 

Try out Solidworks Flow Simulation: www.solidworks.com/Linus 
CAD models from this video: https://grabcad.com/library/extreme-tec-1

 

Buy: A Noctua Cooler, don't get a TEC 
On Amazon: https://geni.us/LnK7k6k
On Newegg: https://geni.us/8tp4

 

 

Link to post
Share on other sites

Why oh why did I have to witness that autotune?  lol


Workstation Laptop: Dell Precision 7540, Xeon E-2276M, 32gb DDR4, Quadro T2000 GPU, 4k display

Ryzen Rig 2: ASrock B450 Pro4 ATX, Ryzen 7 1700 @ 4.2ghz all core 1.4vCore, AMD R9 Fury X w/ Swiftech KOMODO waterblock, Custom Loop 2x240mm + 1x120mm radiators in push/pull 16gb (2x8) 3600mhz V-Color Skywalker (or 4x8gb DDR4 2666mhz for large tasks), Corsair HX850 PSU, 128gb Patriot Scorch NVMe Win 10 boot drive, 500gb Samsung 840 EVO SSD, 512GB TeamGroup MS30 M.2 SATA III, CoolerMaster HAF XM Case.  Zalman K600S keyboard, Zalman ZM-GM1 mouse, Viotek GN24C 24" 1080p 144hz curved and Hannspree HF207 as 2nd monitor

https://www.3dmark.com/3dm/37004594?

Ryzen Rig 1: ASUS B350-PRIME ATX, Ryzen 7 1700, Sapphire R9 Fury Tri-X Nitro 4gb HBM, 16gb (2x8) 3200mhz V-Color Skywalker, ANTEC Earthwatts 750w PSU, MasterLiquid Lite 120 AIO cooler in Push/Pull config as rear exhaust, 250gb Samsung 850 Evo SSD, Patriot Burst 240gb SSD, Cougar MX330-X Case.  Zalman K600S keyboard, Zalman ZM-GM1 mouse, Acer XF270HU 2560x1440 144hz IPS monitor

https://www.3dmark.com/3dm/37628874?

Dwight: The Mixed Metals Loop Media Center.  Ask me about it.  Currently decommissioned to move to an mATX setup on a new MOBO once I pick one out (getting its facelift as of June 2020 have new air cooler, drives etc.  About 60% finished.  Will no longer be a closed loop system.  Slight update, upgraded to larger SSD, and air cooling installed Arctic Alpine cooler.

Schrute: ASUS M5A99FX Pro R2.0, FX 8350, Sapphire R9 Fury Tri-X Nitro 4gb HBM, 16gb (4x4) Corsair Vengeance DDR3 1600mhz, Sparkle/FSP 650w PSU, Corsair H100i GTX 240mm AIO w/ 12mm thick fans to fit in top exhaust, 256gb TIMETEC SSD, 1tb WDBlack HDD, Rosewill Nautilus 1.0 case.  DSI 90-Key Mechanical Keyboard w/ Cherry Red switches, Zalman gaming mouse

Micro Form Factor Dell OptiPlex 3040: Dell 0MGK50 A02, i3-6100T, 2x4gb DDR3 1600, Team Group 120gb SSD, 500gb Seagate 7mm HDD attached storage, Windows 10 Pro, Logitech K400+, USB Wifi adapter all vesa mounted to the back of a 37" 1080p TV 

Linux Box: Toshiba Laptop, i7 620M, NVS graphics, 4gb ram tinker toy at the moment.  Running Manjaro XFCE at the moment.

Home Security: ZOSI 8 channel CCTV (4 used at this time, 1080p) DVR H.265+, 3tb HGST Enterprise HDD, ASUS monitor for display

Link to post
Share on other sites

@AlexTheGreatish Why didn't you just go the intel way?

Massive water chiller and off you go!

 

Also I still belive that a TEC cooler can work, but not like this.

What you can do is get 2 loops.

1 loop is connected to the cold side of multiple lower-powered TEC's (more info below) so you don't just have a cold baseplate but an actual cold sub-ambient loop with water that needs anti-freeze because you could easily pull the whole loop into negative C. And another loop that is connected to the hot side with that massive rad (maybe even more) to keep the hot side happy.

So, why get more TEC's? Because you were very likely hitting a thermal transfer limit with your current setup. If you use multiple TEC's that are lower power you can spread out the heat so it can be transfered easier from the TEC to the loops.

 

TL:DR do it again but instead of 1 TEC, use more of them but lower power and instead of whacking a baseplate on it and mounting it straight onto the CPU, make a loop of it.


If you want my attention, quote meh! D: or just stick an @samcool55 in your post :3

Spying on everyone to fight against terrorism is like shooting a mosquito with a cannon

Link to post
Share on other sites
31 minutes ago, samcool55 said:

whole loop into negative C

i assume he made the cold side closer to the cpu as much as possible to avoid heat gain from environment and condensation, which also the PID controller helped with 

 

on a secondary note, i am very interested on how he programmed the PID controller, cause as a chemical engineer, those mathematical modeling and Laplace transforms , along with tuning methods are usually the most tedious for me 

EDIT: i saw he used Auto Tune , lmao Controller theory and thousands of tuning methods thrown through the window 

@AlexTheGreatish the reason the cpu got overwhelmed so quickly is because the controller was unable to respond quickly enough, maybe thats the fault of the Peltier module ( unable to cool/respond to PID quickly enough), but unless you do mathematical modelling you wont be able to know. 

Its essentially why on any plant we design PID controllers and control systems based on exact mathematical models, so the system can behave predictably such that the PID will know exactly how to respond to keep the system steady


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites
Just now, alextulu said:

Why is it that the thermometer showed a different temperature than the one shown on the computer?

Temperature gradient,

the temperature on the cpu die and the cold end of the TEC plate are not the same, so the thermocouple  would have a temperature in between those 


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites
17 minutes ago, Tamesh16 said:

i assume he made the cold side closer to the cpu as much as possible to avoid heat gain from environment and condensation, which also the PID controller helped with 

 

on a secondary note, i am very interested on how he programmed the PID controller, cause as a chemical engineer, those mathematical modeling and Laplace transforms , along with tuning methods are usually the most tedious for me 

EDIT: i saw he used Auto Tune , lmao Controller theory and thousands of tuning methods thrown through the window 

I am not an engineer in any way but to me it seems like putting the TEC as close to the CPU is not a good idea because you don't have a buffer. Well you have one but it's stupidly small. with a complete loop you can actually create a big buffer (cold water) so you can at least run a benchmark without the cpu overheating almost instantly like in the video.


If you want my attention, quote meh! D: or just stick an @samcool55 in your post :3

Spying on everyone to fight against terrorism is like shooting a mosquito with a cannon

Link to post
Share on other sites
1 hour ago, AlexTheGreatish said:

We got a 545W Peltier cooler... sketchiness ensues.

 

Try out Solidworks Flow Simulation: www.solidworks.com/Linus 
CAD models from this video: https://grabcad.com/library/extreme-tec-1

 

Buy: A Noctua Cooler, don't get a TEC 
On Amazon: https://geni.us/LnK7k6k
On Newegg: https://geni.us/8tp4

 

 

Hey !

 

Please revisit this some day.

 

TECs can work ..all be it they are still not a good choice.

 

Best way to get them working is to use a bunch of them, at half power or less, used to chiller a water loop.

 

As im sure ur aware TECs are far more efficient when not used at their highest power.


CPU: Intel i7 3930k w/OC & EK Supremacy EVO Block | Motherboard: Asus P9x79 Pro  | RAM: G.Skill 4x4 1866 CL9 | PSU: Seasonic Platinum 1000w | VDU: Panasonic 42" Plasma |

GPU: Gigabyte 1080ti Gaming OC w/OC & Barrow Block | Sound: Asus Xonar D2X - Z5500 -FiiO X3K DAP/DAC - ATH-M50S | Case: Phantek Enthoo Primo White |

Storage: Samsung 850 Pro 1TB SSD + Samsung 850 Evo 256GB SSD | Cooling: XSPC D5 Photon 270 Res & Pump | 2x XSPC AX240 White Rads | NexXxos Monsta 80x240 Rad P/P |

Link to post
Share on other sites
1 minute ago, samcool55 said:

am not an engineer in any way but to me it seems like putting the TEC as close to the CPU is not a good idea because you don't have a buffer. Well you have one but it's stupidly small. with a complete loop you can actually create a big buffer (cold water) so you can at least run a benchmark without the cpu overheating almost instantly like in the video.

i understand, its exactly why the loop didnt work, as i said a PID controller takes time to respond, and a cpu could hit max heat output almost instantly, so even if the Peltier module can reach cooling as quickly as the cpu can reach its max heat, the controller will take some time to increase the current flow and result in temperature spikes, cause remember most of the cooling is dependent on current through peltier module and its heat transfer rate

and rightfully said in a loop with cool water nothing stops heat transfer, as soon as the temperature rises heat transfer increases instantly due to the temperature gradient 

understandably though, from a pc building perspective, i understand why they might think why this might be better 


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites
3 minutes ago, SolarNova said:

Best way to get them working is to use a bunch of them

as they rightfully said, a bunch of Peltier modules cooling down the water in a loop, would be more inefficient than just using a chiller, but maybe there is better design than i can think of 


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites
1 hour ago, AlexTheGreatish said:

We got a 545W Peltier cooler... sketchiness ensues.

 

Try out Solidworks Flow Simulation: www.solidworks.com/Linus 
CAD models from this video: https://grabcad.com/library/extreme-tec-1

 

Buy: A Noctua Cooler, don't get a TEC 
On Amazon: https://geni.us/LnK7k6k
On Newegg: https://geni.us/8tp4

 

 

Can you please plug it upside down, hot side on the CPU and cold side on the water block. Then you use the electricity from the seebeck effect to power

the pump & fans.

BOOM! Passive cooling baby! 

 

The bigger the CPU load, the more cooling power you get. 

Link to post
Share on other sites
6 minutes ago, Tamesh16 said:

i understand, its exactly why the loop didnt work, as i said a PID controller takes time to respond, and a cpu could hit max heat output almost instantly, so even if the Peltier module can reach cooling as quickly as the cpu can reach its max heat, the controller will take some time to increase the current flow and result in temperature spikes, cause remember most of the cooling is dependent on current through peltier module and its heat transfer rate

and rightfully said in a loop with cool water nothing stops heat transfer, as soon as the temperature rises heat transfer increases instantly due to the temperature gradient 

understandably though, from a pc building perspective, i understand why they might think why this might be better 

Oooh okay I understand now, I have no clue how PID's work or whatever so TIL I guess ?

Makes sense tho, you don't want something that basically goes to panic mode and causes uncontrollable scenario's when it needs to do something.


If you want my attention, quote meh! D: or just stick an @samcool55 in your post :3

Spying on everyone to fight against terrorism is like shooting a mosquito with a cannon

Link to post
Share on other sites
2 minutes ago, tabarnakos said:

seebeck effec

the heat transfer would be terrible, and the power generated would probably not be enough for the fans and pump 


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites
1 minute ago, samcool55 said:

I have no clue how PID's work

PID loops can be very difficult thing to design, but i love it, its like throwing math at an unpredictable world of scenarios 


PC: Alienware 15 R3  Cpu: 7700hq  GPu : 1070 OC   Display: 1080p IPS Gsync panel 60hz  Storage: 970 evo 250 gb / 970 evo plus 500gb

Link to post
Share on other sites

@AlexTheGreatish don't give in to his negativity... you just need more TECs. If one was overwhelmed, then go for 2 or even 4.
Picture something like a copper plate on the CPU with heatpipes rising to the sky with 4 TECs attached, each with individual water blocks and some overkill water cooling.

If you're going to spend his money, build it big, build it beautiful ?


Apple, Piss Off! ~ Linus 2014

No, you're not hallucinating, or maybe you are... either way, I'm back. ~ Linus 2015

Link to post
Share on other sites

Likely you guys know by now but the solid state relay you used can only switch AC loads. It needs to be at 0V for it to switch. You can get alternative relays that would work in DC applications but for what you wanted you could have moved it to the AC side of the power supply.

Link to post
Share on other sites

Next crazy cooling video, hack up an old water cooler or fridge to send refrigerant directly to the CPU on a cold plate. Just don't let the Canadian EPA find out...or let Linus find out what you spent on the vacuum pump (those are always great to have!) and Rwhatever they're using now a days.

 

Ok but seriously, I still think that the root issue is as mentioned above and.....There's no way that CPU was cranking out more than 400W at full load and stock clocks. You're also assuming that whomever you bought the TEC module from wasn't full of shit. If that came from Aliexpress or Alibaba or probably even eBay then those were Chinese watts so the formula to convert from Chinese watts to Standard watts is as follows:

Cw=(.5⋅Sw^2)/4

Where Cw= Chinese watts and Sw= Standard watts. This is a very similar formula used to convert Chinese lumen to Standard lumen when comparing LED's and other lights. Often it's useful to check the spec sheet for the real product they're copying and de-rate it by at least 30% from the original parts lowest figure as most Chinese sellers use the highest rating available for the original and then increase it by 50% to 200%. It's entirely possible that 545W figure is actually a figure derived from a vendor simply assuming that both sides can do the same work at the same time or it's complete and utter made up nonsense. You'd have to actually measure what it's capacity is in the real world as it seems a lot of Chinese vendors live in an alternate universe where the laws of physics are made up and the points don't matter.

Link to post
Share on other sites

I'm gonna put in my 2 cents as I am an enthusiast when it comes to TECs and was a contributor on Overclock.net TEC forms for a long time.

 

There are so many reasons I can see where this could have been great.

 

I think where this concept failed was your approach. First you need to determine the load that you want to cool. I usually go for 200W of load. Next you need to find out what temperature difference (between the hot and cold side) you want to achieve; 20 - 30C is my usual go to. Next you want to find a TEC that gives you 200W of cooling at that Temperature difference. Special note, the lower the temperature difference you want to achieve on a TEC, the higher the Qmax. If you want to have a 0C difference, that is the large number that everyone advertises. The 127XX TECs are absolute garbage. If you want good TECs, go with Custom Thermo Electric (here). You can easily find TECs that give you massive Qmaxs, but if you're going by Pmax, or 0C temperature difference between hot and cold side, you're achieving nothing.

 

People say the sweet spot is 50% power of the TEC, but the sweet spot actually refers to the Max Voltage / 2. If you get a 199XX TEC, you're looking at a max voltage of just under 24 volts. If you run that at 12V you're going to have a ballin' efficient TEC (compared to what you ran in the episode).

 

I just want to say, when you reach 1000W with that TEC, you killed that TEC and essentially overclocked it. You weren't cooling anything and you were heating up the cold side with the TEC itself, plus the heat from the CPU.

 

Now, in regard to your cold plate, it was too thin. You want to use the copper cold plate to use the edges of the TEC to cool the CPU, but it was too thin to accomplish that. I would aim for 3/4 to 1cm thick.

 

You're "Thermostat" was too slow to react and the solid-state relay is very inefficient and generally slower than using MOSFETs. If you want to just run a TEC at 100%, you will never get anything that resembles useful. Let me know if you'd like to learn more as I've spent years on these and I'm very well versed in them.

 

For extra info, some terminology of TECs:

- Delta: Temperature difference between cold side and hot side.

- Qmax: Amount of heat in Watts that can be moved from the cold side to the hot side at a given delta.

- Pmax: Amount of heat generated and power used by the TEC. Pmax = V x A

- Vmax: Defined maximum amount of voltage put into the TEC.

- Imax: Defined maximum amount of current that can be used by the TEC.

- 127XX: 127 couples at XX Amps (Imax) maximum (Vmax = 15.2V)

- 199XX: 199 couples at XX Amps (Imax) maximum (Vmax = 24V)

- There are more TECs out there, but these are generally the best ones to used in these applications.

 

Thanks.

 

PS: This is the peltier you'd wanna use for this project

 

It has a ~150W Qmax at a 20C temperature difference, and uses ~144W to reach that at 12V.

 

I have seen a bigger and better TEC (200W at 30C difference), but I need to track down the link for that TEC.

Link to post
Share on other sites

Needs more machining footage.

 

Also, @AlexTheGreatish, I asked at LTX and you said that the mill was still on a pallet.  Is it on the floor yet?


Resident Mozilla Shill.   Typed on my Ortholinear JJ40 custom keyboard
               __     I am the ASCIIDino.
              / _)
     _.----._/ /      If you can see me you 
    /         /       must put me in your 
 __/ (  | (  |        signature for 24 hours.
/__.-'|_|--|_|        
Link to post
Share on other sites
Posted · Original PosterOP
2 hours ago, EiiiTrone said:

Likely you guys know by now but the solid state relay you used can only switch AC loads. It needs to be at 0V for it to switch. You can get alternative relays that would work in DC applications but for what you wanted you could have moved it to the AC side of the power supply.

Yeah, realized that pretty fast (it was incorrectly labeled in the Amazon webpage).  Switched it around so that the TEC turned the PSU on and off.

35 minutes ago, FezBoy said:

Needs more machining footage.

 

Also, @AlexTheGreatish, I asked at LTX and you said that the mill was still on a pallet.  Is it on the floor yet?

Yeah it's on the floor now and has a beautiful Kurt vise all trammed in!

42 minutes ago, Krow said:

I'm gonna put in my 2 cents as I am an enthusiast when it comes to TECs and was a contributor on Overclock.net TEC forms for a long time.

 

There are so many reasons I can see where this could have been great.

 

I think where this concept failed was your approach. First you need to determine the load that you want to cool. I usually go for 200W of load. Next you need to find out what temperature difference (between the hot and cold side) you want to achieve; 20 - 30C is my usual go to. Next you want to find a TEC that gives you 200W of cooling at that Temperature difference. Special note, the lower the temperature difference you want to achieve on a TEC, the higher the Qmax. If you want to have a 0C difference, that is the large number that everyone advertises. The 127XX TECs are absolute garbage. If you want good TECs, go with Custom Thermo Electric (here). You can easily find TECs that give you massive Qmaxs, but if you're going by Pmax, or 0C temperature difference between hot and cold side, you're achieving nothing.

 

People say the sweet spot is 50% power of the TEC, but the sweet spot actually refers to the Max Voltage / 2. If you get a 199XX TEC, you're looking at a max voltage of just under 24 volts. If you run that at 12V you're going to have a ballin' efficient TEC (compared to what you ran in the episode).

 

I just want to say, when you reach 1000W with that TEC, you killed that TEC and essentially overclocked it. You weren't cooling anything and you were heating up the cold side with the TEC itself, plus the heat from the CPU.

 

Now, in regard to your cold plate, it was too thin. You want to use the copper cold plate to use the edges of the TEC to cool the CPU, but it was too thin to accomplish that. I would aim for 3/4 to 1cm thick.

 

You're "Thermostat" was too slow to react and the solid-state relay is very inefficient and generally slower than using MOSFETs. If you want to just run a TEC at 100%, you will never get anything that resembles useful. Let me know if you'd like to learn more as I've spent years on these and I'm very well versed in them.

 

For extra info, some terminology of TECs:

- Delta: Temperature difference between cold side and hot side.

- Qmax: Amount of heat in Watts that can be moved from the cold side to the hot side at a given delta.

- Pmax: Amount of heat generated and power used by the TEC. Pmax = V x A

- Vmax: Defined maximum amount of voltage put into the TEC.

- Imax: Defined maximum amount of current that can be used by the TEC.

- 127XX: 127 couples at XX Amps (Imax) maximum (Vmax = 15.2V)

- 199XX: 199 couples at XX Amps (Imax) maximum (Vmax = 24V)

- There are more TECs out there, but these are generally the best ones to used in these applications.

 

Thanks.

 

PS: This is the peltier you'd wanna use for this project

 

It has a ~150W Qmax at a 20C temperature difference, and uses ~144W to reach that at 12V.

 

I have seen a bigger and better TEC (200W at 30C difference), but I need to track down the link for that TEC.

At first we were running it 300W and 12V, didn't work at all.  But yeah getting a more efficient TEC might have been able to turn the episode around, in the future will just use a refrigerant based system for chilling.

Link to post
Share on other sites
1 minute ago, AlexTheGreatish said:

Yeah it's on the floor now and has a beautiful Kurt vise all trammed in!

Dope.  Hope to see some content with it soon!


Resident Mozilla Shill.   Typed on my Ortholinear JJ40 custom keyboard
               __     I am the ASCIIDino.
              / _)
     _.----._/ /      If you can see me you 
    /         /       must put me in your 
 __/ (  | (  |        signature for 24 hours.
/__.-'|_|--|_|        
Link to post
Share on other sites

16:51

Quote

[...]but it occurred to us that with so much power we might be running into over current protection limits on the PCI-E connectors

You were using an EVGA T2 1600W. It actually does not have any OCP for the PCIe connectors specifically. It's a single rail PSU. Read: EVGA cheaped out, didn't implement multi rail over current protection, and made it a feature by calling it single rail. You wouldn't run into any issues with a non existant protection. So that thing with rewiring the connector so it worked with an EPS connector... Totally pointless :P


 

Quote

Women. They are a complete mystery.

-Stephen Hawking

 

I think the hoomans put their builds here?

Why do you hoomans give your builds a name? Here's my build, which I shall call "Do as I Say, Not As I Do" (seriously, don't get this build)

Spoiler

Ryzen 1500X @3,925 GHz

Cooler Master Hyper 212 Evo + 2x ML120

MSI B350 Tomahawk Arctic

2x8GB Corsair Vengeance LPX 3000 MHz CL15 (Micron B-die) @2933 MHz

Sapphire Radeon R9 280 Dual-X @1120 MHz / 1450 MHz

120GB 850 Evo

120GB Kingston SSD

500GB WD Blue

Cooler Master Elite 430

Seasonic Prime Titanium 650W

Logitech G710 with Kailh Box Jade

Logitech G502

HyperX Cloud

And my laptop, which I shall call "If It's Stupid But It Works" (It can actually play CS:GO at 50 FPS, and Civ V at 25 FPS)

Spoiler

Lenovo Thinkpad L460

Intel Core i3 6100U

4GB (probably) DDR4 2133 MHz

Intel HD Graphics 520 0.3-1.0 GHz

128GB Samsung MZ7LF128HCHP

Corsair M65 Pro RGB (worst mouse I've ever had)

Sennheiser CX 5.00G

And here would be where I would put a picture of my cat. But apparently, images are not allowed here. So take this instead (*ΦωΦ*)

Hello fellow night theme users

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×