AMD Ryzen Threadripper Waterblock Comparison #1

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,534
AMD Ryzen Threadripper Waterblock Comparison #1

We have been waiting for AMD Threadripper CPU custom cooling parts to make their way to us. We have our first two purpose-built Threadripper waterblocks from XSPC and Bykski. We put both these coolers to the test with our 4GHz overclocked Threadripper in hour long stress tests to see how our temperatures fare.
 
Great article.
I *love* the picture of the SSD overlaid with the CPU - gives a great feel for how big that thing is.
I may just have to take the plunge on a TR setup here shortly...
 
Sorry to kind of be off topic but how are the quick disconnects going? Do you notice any flow rate reduction?
 
Sorry to kind of be off topic but how are the quick disconnects going? Do you notice any flow rate reduction?
No. These flow greatly. You can hook these together and run them under a faucet stream and not overflow the coupling. Been using these for years. Great product.
 
I really like my Raystorm Pro (Intel) for essentially the same reasons. Well thought out mount, easy to clean, solid build quality, and holy crap does it dump heat.

Sorry to kind of be off topic but how are the quick disconnects going? Do you notice any flow rate reduction?

With the 1/2" ones it should be negligible. But the pump they have is adjustable, so with a D5 you can just turn it up if needed.
 
I'll never need this CPU but your article was very well done and nicely complete.
We don't see too much in the way of good watercooling discussion and TIM application anymore, anywhere. (I suppose it has to do with all the AIO coolers available now.)

As I said, even if I never buy that ginormous CPU, the write-up was a very good read.

Thank you.
 
Excellent write up...especially on the TIM application and the TIM material you recommend! I was going to use Prolimatech PK-Zero paste...is the PK-1 a better option?
 
I really like my Raystorm Pro (Intel) for essentially the same reasons. Well thought out mount, easy to clean, solid build quality, and holy crap does it dump heat.



With the 1/2" ones it should be negligible. But the pump they have is adjustable, so with a D5 you can just turn it up if needed.
I have been running it wide open.
 
Excellent write up...especially on the TIM application and the TIM material you recommend! I was going to use Prolimatech PK-Zero paste...is the PK-1 a better option?
I have not used the PK-Zero, so I cannot say. I just bought 30 grams more of the PK-1 to use here on the test bench, but on the new HSF/AIO testing platform I moved to PK-3. It has a bit higher thermal conductivity, but the 3 is a bit harder to spread than the 1. And trying to get pk-3 even across that whole damn Threadripper repeatedly was about to piss me off. For smaller CPUs it was OK to use. PK-1 is 10.2w, and PK-3 is 11.2w. The PK-Zero is 8w. Higher is "better."
 
Gawd damn Kyle, thank you so much for this write-up, and the videos you always produce. Extremely useful information, always very interesting, and fascinating like a muthafucka. Shit like this not only helps a ton of us out with the info and data provided, but is among our highest-valued sources of entertainment. I don't know if it's subliminal messaging or some kind of indetectable crack you guys are feeding us through our screens, but I've been an addict for way too long, and I'm sure when I finally keel over and give up the ghost, I'll still be all up on tha [H]. It's also a lot of fun seeing AMD CPUs getting used for serious testing of all types again. Quite a journey this has been, dealing with Threadripper's multiple dies and awesomely large socket footprint. OC'ing a ridiculous number of cores, and dealing with the heat output. AMD has made things a lot more interesting and fun again, as Kyle and many others have said over these past weeks. It's an exciting time. I've been waiting eagerly to see how the different waterblock and AIO cooler companies responded to Threadripper, and how their Threadripper products fared. It's gonna be fun seeing where we go from here.
 
"I did check for temperature deltas between our MG and Pk-1 TIM. ... I actually thought it would be bigger, ..."
I expected it to be less. The gap between the two surfaces should be so small (ideally nil, mostly) that the TIM used should make very little difference.
With a distance of 10um the conductivity is 10kW/K vs 100kW/K. At 100W that means a temperature difference of ~0.1K.
You must have squeezed more than 0.3mm of TIM between the surfaces...
 
"I did check for temperature deltas between our MG and Pk-1 TIM. ... I actually thought it would be bigger, ..."
I expected it to be less. The gap between the two surfaces should be so small (ideally nil, mostly) that the TIM used should make very little difference.
With a distance of 10um the conductivity is 10kW/K vs 100kW/K. At 100W that means a temperature difference of ~0.1K.
You must have squeezed more than 0.3mm of TIM between the surfaces...
Yeppers. Apparently I have gotten reall really good at mounting those blocks after 20 years. ;)
 
Great review on these water blocks.

I just don't have a feel for what size radiator for stock clocks and OC. Is 480mm (how thick?) mandatory for sustain 4ghz? I have a 360mm radiator 60mm thick, not sure where that would land with the 1950x. I hope some of that will be further explored if possible.

Then of course any ideal cases that would support those combinations. I am just too much a noob on custom loops.
 
Great review on these water blocks.

I just don't have a feel for what size radiator for stock clocks and OC. Is 480mm (how thick?) mandatory for sustain 4ghz? I have a 360mm radiator 60mm thick, not sure where that would land with the 1950x. I hope some of that will be further explored if possible.

Then of course any ideal cases that would support those combinations. I am just too much a noob on custom loops.
My triple fan Koolance rated at 900w would not handle it.
 
Great review on these water blocks.

I just don't have a feel for what size radiator for stock clocks and OC. Is 480mm (how thick?) mandatory for sustain 4ghz? I have a 360mm radiator 60mm thick, not sure where that would land with the 1950x. I hope some of that will be further explored if possible.

Then of course any ideal cases that would support those combinations. I am just too much a noob on custom loops.

I always used to hear 120mm per 100 watt for decent performance on custom loops. There's a lot of radiator reviews out there that will tell you exactly what a radiator will disappate at a certain fan speed.

Or just buy this behemoth like I did and be done with radiators forever.

http://www.xoxide.com/watercool-mora3-9x140-ltblackradiator.html
 
I always used to hear 120mm per 100 watt for decent performance on custom loops. There's a lot of radiator reviews out there that will tell you exactly what a radiator will disappate at a certain fan speed.

Or just buy this behemoth like I did and be done with radiators forever.

http://www.xoxide.com/watercool-mora3-9x140-ltblackradiator.html
Radiator thickness should also be a consideration, number of tubes, fin count, passes, how well it flows air, thickness of the tubes, material as in heat transfer ability or coefficient - in short wattage rating for a given flow rate, air flow and temperatures of fluid and the air. Anyways bought an EK X360 kit, been sitting awhile, really wondering if sufficient for future upgrades.

I would most likely chill the water and case with an air conditioner. Very huge radiators just comes to a point of just being to big and if really needed, means dumping a lot of heat into the room.
 
Last edited:
Radiator thickness should also be a consideration, number of tubes, fin count, passes, how well it flows air, thickness of the tubes, material as in heat transfer ability or coefficient - in short wattage rating for a given flow rate, air flow and temperatures of fluid and the air. Anyways bought an EK X360 kit, been sitting awhile, really wondering if sufficient for future upgrades.

I would most likely chill the water and case with an air conditioner. Very huge radiators just comes to a point of just being to big and if really needed, means dumping a lot of heat into the room.

Dude, everything dumps heat into the room. Nothing is magical, there's always a cost, and physics.
 
Dude, everything dumps heat into the room. Nothing is magical, there's always a cost, and physics.
Yep. If you have 500 watts in a room, you have 500 watts in a room. Just depends how quickly it gets to other places besides your computer case. :)
 
Thanks for the article. Still can't justify buying a TR for my needs though... I wish I could.

BTW -- typo on the 2nd page link... "2 - AMD Ryzen Threadripper Coo<l>ing"
 
Wouldn't it actually decrease?



This.

The delta between thin and thicker radiators. Increase the fan speed and it'll increase. At low fan speeds the air saturates so additonal thickness doesn't typically help.

Thermal is actually really easy to predict and calculate.

Google a radiator roundup and look at the chart for low fan speeds. There's not a lot of variance. If you know your target GPM, fan speed, temp delta all the data is out there.
 
Dude, everything dumps heat into the room. Nothing is magical, there's always a cost, and physics.
Agree in most home systems, now I had an air condition case for years (not now, may again though) in that unusual configuration the heat is dumped outside and not ever inside. In Fact the case was cooler then the room so running the computer would cool the room. I think Justreason is doing this now so not too strange. The advantages I've found was:
  • Great consistent temperature control of system
  • Significant cooling potential - I could go down to -7c except that caused problems with rotating devices like hard drives and fans
    • Going from 30c case temp to 10c will give you virtually a 20c decrease temperature across the board (if you keep the same fan speeds and pump speeds)
    • CPU running at 70c would now run at 50c, VRMs, GPU's the same
    • I use to run at 5c to 10c
  • Dust free
  • Reduced noise except A/C unit does add some noise but is rather muffled since it is an enclosed system with the computer case
  • Of course heat free influence or very little due to monitor to your room
  • It just does not cool the CPU but everything - VRMS, GPU, Power supply etc.
The trick if you go cold is:
  • Keep the A/C loop closed, no way for moisture or very little to enter the setup
  • Keep the coldest zone or place (where moisture will come out if there is any) always in the A/C unit
  • Slightly preheat the air coming from the A/C unit to the PC, so the temperature is always above the dew point
  • Insulate the case on the inside to prevent sweating on the outside which also makes the case even more quieter, also nothing is open to allow noise to easily come out. Case wise it becomes very quiet
I can see if I build a TR system (possibly next year) this will be the route I will go. I may do this anyways with the Ryzen 1700x system.
 
Last edited:
True, I always look through my biased perspective. If you want slow and quiet, the big reason I do custom loops personally, it doesn't take much thickness to saturate the air with heat so area (LxW) is the most important factor. For the thick ones you need to crank the fans. A thick rad will only get you 10% more cooling than a slim at 1000RPM. For example an AlphaCool 45mm vs 60mm only has 1% (3W) difference at 1000 RPM because the air is fully saturated. Dissipating around 275W for 360mm, 10deg delta and 1GPM. A thin rad will get you around 240W.

If you're willing to crank the fans that delta will increase.
A thicker radiator can allow for more tubes, slower flow rates which allow more time or interaction for heat transfer, also larger surface area for heat transfer. The increase in heat transfer rate can be achieved by many factors. Really the wattage rating with temperatures and flow is really the bottom line on how effective the radiator is for cooling PC components. If your coolant temperature = air temperature a.k.a 25c is when you would not have heat transfer, in other words never really saturated. For temperature leaving the cooler and water temperature same thing. Two pass heat exchangers (not sure if they are used for this application) with a cross flow design (hottest water sees the hottest air first and the colder section or water sees the air from the fan last increases heat transfer from one fluid medium to another or to air.

Now I can take a small radiator and push 1000w or 10,000w of heat transfer but is utterly worthless for cooling a CPU. For example steam at high temp being cooled with air, a lot of heat transfer but means nothing for cooling a CPU.
 
The delta between thin and thicker radiators. Increase the fan speed and it'll increase. At low fan speeds the air saturates so additonal thickness doesn't typically help.

Thermal is actually really easy to predict and calculate.

Google a radiator roundup and look at the chart for low fan speeds. There's not a lot of variance. If you know your target GPM, fan speed, temp delta all the data is out there.

Not going to get into it and hijack this thread, but you clearly are not making any sense.


Agree in most home systems, now I had an air condition case for years (not now, may again though) in that unusual configuration the heat is dumped outside and not ever inside. I think Justreason is doing this now so not too strange. The advantages I've found was:
  • Great consistent temperature control of system
  • Significant cooling potential - I could go down to -7c except that caused problems with rotating devices like hard drives and fans
    • Going from 30c case temp to 10c will give you virtually a 20c decrease temperature across the board (if you keep the same fan speeds and pump speeds)
    • CPU running at 70c would now run at 50c, VRMs, GPU's the same
    • I use to run at 5c to 10c
  • Dust free
  • Reduced noise except A/C unit does add some noise but is rather muffled since it is an enclosed system with the computer case
  • Of course heat free influence or very little due to monitor to your room
  • It just does not cool the CPU but everything - VRMS, GPU, Power supply etc.
The trick if you go cold is:
  • Keep the A/C loop closed, no way for moisture or very little to enter the setup
  • Keep the coldest zone or place (where moisture will come out if there is any) always in the A/C unit
  • Slightly preheat the air coming from the A/C unit to the PC, so the temperature is always above the dew point
  • Insulate the case on the inside to prevent sweating on the outside which also makes the case even more quieter, also nothing is open to allow noise to easily come out. Case wise it becomes very quiet
I can see if I build a TR system (possibly next year) this will be the route I will go. I may do this anyways with the Ryzen 1700x system.

I don't know if that's all worth it as TR is not a heat monster nor is it very difficult to cool. That said, I'd like to see your chilled build.
 
Not going to get into it and hijack this thread, but you clearly are not making any sense.




I don't know if that's all worth it as TR is not a heat monster nor is it very difficult to cool. That said, I'd like to see your chilled build.
I have images of the old setup, previous chill build - just currently without power at home (IRMA), maybe for an extended period of time so stuck in a motel room for now.
 
I have images of the old setup, previous chill build - just currently without power at home (IRMA), maybe for an extended period of time so stuck in a motel room for now.

Wow, good luck out there and with Irma.
 
Wow, good luck out there and with Irma.
IRMA is history except what it left behind. Thanks

Looks like TR has good water blocks to choose from, just how much you can push it beyond stock looks to be where it can become very problematic on the rest of the components where big is not big enough, talking about radiators.
 
IRMA is history except what it left behind. Thanks

Looks like TR has good water blocks to choose from, just how much you can push it beyond stock looks to be where it can become very problematic on the rest of the components where big is not big enough, talking about radiators.

4.1-ish is the typical process maximum. You need LN2/Dice to get to above that, or maybe phase can get you part way, dunno. The max with TR and Ryzen is a silicon issue and not a cooling issue so there's very little return with extreme cooling, like chillers and phase.
 
A thicker radiator can allow for more tubes, slower flow rates which allow more time or interaction for heat transfer, also larger surface area for heat transfer. The increase in heat transfer rate can be achieved by many factors. Really the wattage rating with temperatures and flow is really the bottom line on how effective the radiator is for cooling PC components. If your coolant temperature = air temperature a.k.a 25c is when you would not have heat transfer, in other words never really saturated. For temperature leaving the cooler and water temperature same thing. Two pass heat exchangers (not sure if they are used for this application) with a cross flow design (hottest water sees the hottest air first and the colder section or water sees the air from the fan last increases heat transfer from one fluid medium to another or to air.

Now I can take a small radiator and push 1000w or 10,000w of heat transfer but is utterly worthless for cooling a CPU. For example steam at high temp being cooled with air, a lot of heat transfer but means nothing for cooling a CPU.


All I was saying is that if the air is saturated at 30mm of your 45mm (so water temp is approximately air temp) rad at low fan speeds the last 15mm doesn't benefit.

If you look at reviews the thicker rads don't shine until you have high flows.

IMG_3236.PNG


Look at this chart from http://www.xtremerigs.net/2015/02/11/radiator-round-2015/5/

The top 20 radiators are within +/- 5%. You guys are acting like it's some kind of black art. It's an easily measureable metric. I've designed everything from custom chilled loops to 500kW industrial systems. They've always been spot on with calcs. Heat is by far the easiest thing to predict.
 
  • Like
Reactions: Meeho
like this
All I was saying is that if the air is saturated at 30mm of your 45mm (so water temp is approximately air temp) rad at low fan speeds the last 15mm doesn't benefit.

If you look at reviews the thicker rads don't shine until you have high flows.

View attachment 36473

Look at this chart from http://www.xtremerigs.net/2015/02/11/radiator-round-2015/5/

The top 20 radiators are within +/- 5%. You guys are acting like it's some kind of black art. It's an easily measureable metric. I've designed everything from custom chilled loops to 500kW industrial systems. They've always been spot on with calcs. Heat is by far the easiest thing to predict.

I can totally see that 4 of the top 6 rads are all 60mm or thicker, the 5th which is 50mm and the last being the only thin rad. Doh?
 
I can totally see that 4 of the top 6 rads are all 60mm or thicker, the 5th which is 50mm and the last being the only thin rad. Doh?

And they are all neglible in performance...

Look at the ST30 vs UT60 in the first chart (low fan speed) vs them in this chart.

IMG_3247.PNG


TLDR at low fan speed they are 0.5% apart and high fan speed 23%.
 
Last edited:
Talking about 100w difference with no change in footprint except height or thickness. I think I wold take the thicker cooler, just my preference. I have the EK-CoolStream XE 360, which looks like a 500w radiator. The EK-CoolStream SE 360 is a 26mm thick or height and it is about 165w less than the thicker XE 360 being 60mm thick.

Pushing the volts/clocks of TR 1950X it would not be enought to keep below 75c from Kyles data even using the thicker XE 360 (I think that is when TR starts to throttle limiting the OC benefits). I do believe if you lower the temperature 20c for the air cooling the radiator that would do the trick. At stock speeds or more mild OCs/voltage it should be OK without cooling the air going to the radiator.

Starts with waterblock that can transfer the heat out of the CPU to the water and then the radiator has to transfer it out of the water while keeping the CPU less than 75Fc. If CPU throttling was at 95c vice 75c that would allow smaller radiators to work better.

I wish EK would have sent Kyle a block, I really don't understand why EK wouldn't since a review in itself is already free advertising, unless they have an inferior block and didn't want that exposed.

I wonder if [H]ardOCP will do some chill water system reviews and if even it would give a significant boost at that for 24/7 use.
 
Last edited:
I wonder if [H]ardOCP will do some chill water system reviews and if even it would give a significant boost at that for 24/7 use.

That would be interesting. Repeating the test on a cold night would be the same (and save equipment costs and avoid concerns of condensation.)

It's almost a postive of AMD using multiple dies and creating a massive CPU. Intel's single die 18 core will be a bitch to cool.

I use the Raystorm block. They make good stuff.
 
That would be interesting. Repeating the test on a cold night would be the same (and save equipment costs and avoid concerns of condensation.)

It's almost a postive of AMD using multiple dies and creating a massive CPU. Intel's single die 18 core will be a bitch to cool.

I use the Raystorm block. They make good stuff.
The saving grace maybe with Intel 18 thread version is that it is a bigger die than the 10 core version. We just have to wait and see.
 
Talking about 100w difference with no change in footprint except height or thickness. I think I wold take the thicker cooler, just my preference. I have the EK-CoolStream XE 360, which looks like a 500w radiator. The EK-CoolStream SE 360 is a 26mm thick or height and it is about 165w less than the thicker XE 360 being 60mm thick.

Pushing the volts/clocks of TR 1950X it would not be enought to keep below 75c from Kyles data even using the thicker XE 360 (I think that is when TR starts to throttle limiting the OC benefits). I do believe if you lower the temperature 20c for the air cooling the radiator that would do the trick. At stock speeds or more mild OCs/voltage it should be OK without cooling the air going to the radiator.

Starts with waterblock that can transfer the heat out of the CPU to the water and then the radiator has to transfer it out of the water while keeping the CPU less than 75Fc. If CPU throttling was at 95c vice 75c that would allow smaller radiators to work better.

I wish EK would have sent Kyle a block, I really don't understand why EK wouldn't since a review in itself is already free advertising, unless they have an inferior block and didn't want that exposed.

I wonder if [H]ardOCP will do some chill water system reviews and if even it would give a significant boost at that for 24/7 use.

Nay im using an EK block right now and it cools just fine. I can run full hammer down at stock speed and it pegs at around 61-64c OC its in the low 70's just like the other two blocks.

I am going to bet it was a new employee that given they had some authority to make a decision probably didn't know who Kyle was and assumed he was just some random internet wannabe reviewer with 200 subs. Its a shame it didn't go higher in their company than the lowbie that made the decision for Kyle.
 
Well I called out EK on their FB for making an inadequate block for TR as I have the block and the xpsc one. My temps in prime 95 small went from throttling @ 1.35v to a max of about 69c-70c with the xspc block. They posted they contacted hardocp for a sample to review per their FB response.
 
Well I called out EK on their FB for making an inadequate block for TR as I have the block and the xpsc one. My temps in prime 95 small went from throttling @ 1.35v to a max of about 69c-70c with the xspc block. They posted they contacted hardocp for a sample to review per their FB response.
Screen shot? I would love to see that. Also, sure there was not something wrong with the EK block? That is crazy.
 
I own 2 of the EK blocks. I wanted the XSPC one but the wait was too long and I needed a block so I got the EK on shortly after launch. I then read the XSPC prototype review on hardocp and I saw the 75c in prime 95 with 1.4v while I was throttling @ 1.35v in prime 95 going over 85c @ 4ghz I felt something was up. I then tested my other rig with the other EK block and found it hitting 85c as well. I swapped to the XSPC block afterwards and my prime 95 dropped to 70-72 area under the same conditions.
image.jpg

image.jpg

And checking my email this morning, I responded to EK, they did reach out. They did not like me questioning their change of heart on sampling.

https://hardforum.com/threads/ek-threadripper-waterblocks-available.1941926/#post-1043273912
 
Back
Top