Swiftech Apogee SKF Threadripper Water Block Review @ [H]

[FrgMstr]

Swiftech Apogee SKF Threadripper Water Block Review

Swiftech has long been known for it cooling prowess among computer enthusiasts for two decades. Not to be left behind, Swiftech is now starting production on a water block specifically designed for AMD's Threadripper. Not only is it designed for socket TR4, this new block comes to use with an advanced micro-fin design.

 

[IKV1476]

Great thorough review as always.
It was nice to see that you cleaned the fins out with compressed air and that things got bent.
That is good information for when cleaning the loop so as to not destroy the blocks fins.

It is also interesting that this is the 2nd review where you got better performance using better springs not supplied with the kit.
Hopefully all the manufacturers will step up and put better springs in there kits to beat their competitors in performance.
4-5 or 7-8 degrees better performance is a lot, could mean a slower fan profile for even less noise.

 

[guitarslingerchris]

It is also interesting that this is the 2nd review where you got better performance using better springs not supplied with the kit.
Hopefully all the manufacturers will step up and put better springs in there kits to beat their competitors in performance.
4-5 or 7-8 degrees better performance is a lot, could mean a slower fan profile for even less noise.

My guess is they put springs in that comply with the socket spec so they aren't liable for damage to the board or socket but obviously there are gains to be had going past those artificial limits.

 

[FrgMstr]

My guess is they put springs in that comply with the socket spec so they aren't liable for damage to the board or socket but obviously there are gains to be had going past those artificial limits.

I would suggest that is one aspect of it. I would suggest they are going by socket specifications and/or weighing liability.

Also, our Prolimatech PK-1 is extremely viscous. I have no idea if these companies are testing with some of these "newer" high conductivity TIMs that are so thick.

 

[deeppow]

I can fully understand your report (GOOD work) BUT what components does a user buy? What you have setting around your lab from other setups doesn't mean crap to me. Please, simple links regarding what to buy. Springs are very simple, can you specify the specs on such items? A hardware store might have such items if we know what to look for.

I ain't about to buy one of these without the "right parts."

 

[FrgMstr]

I can fully understand your report (GOOD work) BUT what components does a user buy?

The key questions here are, "What is your budget and what exactly are you cooling?" and "What size restrictions are you working with?"

What you have setting around your lab from other setups doesn't mean crap to me.

We built that system to specifically to test Threadripper water blocks, it was far and away from things I had sitting around. I actually got rid of all that stuff. It is listed in the article with links.

Our cooling loop is comprised of all XSPC components: D5 Photon Reservoir/Pump Combo V2 (run at max RPM); RX480 Radiator V3, and XSPC RayStorm TR4 Neo for comparison.

Please, simple links regarding what to buy.

Refer to question 1.

Springs are very simple, can you specify the specs on such items?

As outlined in the article, I do NOT have the specifications on these springs.

A hardware store might have such items if we know what to look for.

You would be 100% correct. You can build an entire mounting system at Home Depot and a drill.

 

[Elkwood]

Tks for the review!! Its always nice to see with a bit of tinkering how u can improve on stock. I have a Heatkiller tr4 blk and
i wonder if doing something like that would improve its performance. I have reset the blk many times trying to get its thermals
right. Its been a big help when u show the difference tim applications. I went to the single line spread with covered finger/ 5 light dots.

I had a lot of confusion over the tctl vs tdie readings. Till i noticed it was just 27c that separated them. The ryzen oc software shows
tdie and the asrock bios shows it in tctl.

Always liked swifttec products. Glad to see it do so well.

 

[FrgMstr]

Tks for the review!! Its always nice to see with a bit of tinkering how u can improve on stock. I have a Heatkiller tr4 blk and
i wonder if doing something like that would improve its performance. I have reset the blk many times trying to get its thermals
right. Its been a big help when u show the difference tim applications. I went to the single line spread with covered finger/ 5 light dots.

I had a lot of confusion over the tctl vs tdie readings. Till i noticed it was just 27c that separated them. The ryzen oc software shows
tdie and the asrock bios shows it in tctl.

Always liked swifttec products. Glad to see it do so well.

tctl is a offset that AMD uses for some cooling profiles. When it comes to OCing, ignore it in terms of temperatures.

 

[Teenyman45]

Would it be feasible to try one more thermal run with Swiftech's stock springs and stock TIM Mate paste? Unless you were planning on saving the paste for another big round of thermal paste comparisons, it seems like it will otherwise just be collecting dust.

 

[FrgMstr]

Is there a way you could use the same modified mount on the Raystorm?

I did. Even typed it out in the review iirc.

 

[BitMaster]

Nice, those are impressive results.

One question comes to my mind, looking at those tiny fins.
How easy are those micro fins to clean and how would you do that without harming the fins ?

I do know that I have to take mine apart like once a year and clean out the debris that accumulates and I really wonder how user friendly those new micro fin based blocks are when it comes to annual cleaning.

Nice review,I really like the multiple TIMing approach. Good job

 

[FrgMstr]

Nice, those are impressive results.

One question comes to my mind, looking at those tiny fins.
How easy are those micro fins to clean and how would you do that without harming the fins ?

I do know that I have to take mine apart like once a year and clean out the debris that accumulates and I really wonder how user friendly those new micro fin based blocks are when it comes to annual cleaning.

Nice review,I really like the multiple TIMing approach. Good job

As described in the article, I used compressed air. I would not suggest 100psi+ like I used for the spelled out reasons.

With these newer blocks with these tiny fins and channels, I would suggest a filtering system before the block.

 

[magoo]

Really nice article and videos.
It's nice to all of a sudden see a bunch of watercooling reviews.
For a good while it seemed like it was losing some popularity, given the avalanche of AIO cooling.

The entire Swiftech kit is very classy. The block looks really nice, the presentation is top shelf.

Too bad you have to ghetto-rig the thing to get it to show it's true potential.

 

[deeppow]

Kyle, Thank you very much for your detailed reply. The spring constant can be determined and would be useful for both the Swiftech supplied springs and the stronger ones you used. One of the advertisers on the home page is Century Spring Corp and I'm sure with a starting point info the stronger springs are available.

Again, excellent review and info.

 

[Emission]

Thanks a lot for the review Kyle!

I am curious about one thing though. Is prime95 the application that causes ThreadRipper to output the most heat? I've found in my own testing during overclock sessions with my Ryzen 7 1700X that things like CineBench R15 and BOINC will draw more power and cause overclocks to fail more readily than prime95 will. This might be a fault of my prime95 config though, I'm not sure.

Also, how do you get your power usage measurements? The references I use are from HWMonitor, they seem to be pretty accurate.

 

[FrgMstr]

Thanks a lot for the review Kyle!

I am curious about one thing though. Is prime95 the application that causes ThreadRipper to output the most heat? I've found in my own testing during overclock sessions with my Ryzen 7 1700X that things like CineBench R15 and BOINC will draw more power and cause overclocks to fail more readily than prime95 will. This might be a fault of my prime95 config though, I'm not sure.

Also, how do you get your power usage measurements? The references I use are from HWMonitor, they seem to be pretty accurate.

Have not tested Boinc, but you can. Really easy to run HWiNFO and watch total PACKAGE WATTAGE.

Cinebench does not come close to Prime95 Small FFT IIRC. It has to be SMALL FFT though.

 

[FrgMstr]

I'll test out small FFT and double check my numbers. IIRC I run similar settings to small FFT but crank up the memory usage.

I would like to know what you see.

Worth nothing too that running Prime95 FFT, workload is NOT static as you might think. After 30 to 45 minutes of running you should see no higher temp spikes up to 1.5 hours. There IS another spike around 2.1 hours that may be a .1C higher on die temperature. So, just saying, if you are going to use it to log temps, I would suggest using it the same way every time.

 

[atarione]

I think swiftech should tune the spring choice and hopefully they will... for me I'm highly unlikely to buy a product that requires I make a trip to home depot to make it work optimally.

I'm not sure what this means for your marketing / ad revenue.... but (and very limited sample size..) but your videos seem to be quite interesting to my cat Fitz... so maybe show this pic to some cat food companies and see if they want to be sponsors??

 

[Emission]

Hi Kyle, I got a chance to run Prime95 Small FFT overnight and collected all the data using HWiNFO64's logging feature. I took a gander at it and I don't see any peaks that are higher than what I see in BOINC or after repeated Cinebench R15 runs, the package wattage peaks at around 124W.

My settings are as follows:
Bus: 100 MHz
Multi: 38x
Core Voltage: set to 1.275v in BIOS
RAM: DDR4 2400, XMP enabled, CAS 16, 1.2v

All else either left default or the setting isn't available to me in the BIOS (ASRock X370 Fatal1ty ITX board, BIOS is pretty sparse on CPU OC options). Ryzen 7 1700X.

Here's the full log, easiest to view in excel or similar spreadsheet software.

PrimeDataLog.CSV

 

[FrgMstr]

Hi Kyle, I got a chance to run Prime95 Small FFT overnight and collected all the data using HWiNFO64's logging feature. I took a gander at it and I don't see any peaks that are higher than what I see in BOINC or after repeated Cinebench R15 runs, the package wattage peaks at around 124W.

My settings are as follows:
Bus: 100 MHz
Multi: 38x
Core Voltage: set to 1.275v in BIOS
RAM: DDR4 2400, XMP enabled, CAS 16, 1.2v

All else either left default or the setting isn't available to me in the BIOS (ASRock X370 Fatal1ty ITX board, BIOS is pretty sparse on CPU OC options).

Here's the full log, easiest to view in excel or similar spreadsheet software.

PrimeDataLog.CSV

We are running 200MHz faster on the core. (Big difference as the package starts leaking the higher your OC.)
We are running much higher vCore.
We are running much higher Ram rate, 3200MHz.

So yeah, your results would be expected to be much different than ours.

 

[FrgMstr]

I'm aware of the differences in our hardware and settings, I was just trying to see how your Prime95 Small FFT package wattage numbers compare to running any other stress application on your hardware.

Rgrt.