I'd like to note that the temperature graph for 13700K at 125W is a duplicate of R7 7700X in error. You can find all the graphs, including the correct 13700K graph in written review here: reviewer.co/coolers/thermalright-si-100
Thank you for making this. Low profile coolers and use on CPUs over 65W... I always get them confused. I hope you'll be able to make a vid w/ the upcoming TR AXP90-52 FULL copper. Have a good one
BeQuiet wouldn't send one when I enquired, so cannot help. I can say that manufacturer wattage ratings are meaningless and it certainly cannot handle 100W on AM5 CPU.
Do you think you can make a video showing how to setup the black ridge on AM5? I've seen a lot of people say it's possible but no tutorials or other sources on how to do it
Thanks, i would buy thermalright axp-90 x53 white for my itx build, i would use 12100f and i would upgrade to 13600k after, i would use him in stock with 1 92mm in the case. Thanks
Zen 4 CPUs have very poor thermal transfer, due to very thick IHS. Testing was also carried out in the worst-case scenario use case for given CPU. Wattage doesn't mean anything if the CPU has poor thermal properties. One CPU generation or platforms thermal properties aren't universal either.
Thank you for the tests. Can you please provide CPU package power readings for the different coolers? For the ones at the bottom, it is likely the CPU is throttling. Ideally, if we can have the CPU package power for an unlocked CPU in a sandwich case (or something like the fractal ridge), we can get an approximate TDP for the different coolers in an SFF setting, which is highly valuable information.
Package power, also known at PPT for AMD, is locked as stated on grapths. AM5 CPUs have abysmal performance/tuning out of the box and should never be ran with unlocked limits as they simply perform worse on lower end cooling solutions that cannot constrain them. No CPU cooler shown are thermal throttling, but lowest-end solutions are running slightly lower all core clock speeds as demonstrated. There is no one case fits all for all of these coolers. In order for things to be as fair as possible, variables needed to be limited as much as possible, power packages targets being one of the most important
@@L91 Thanks for the reply. I think practically, a user just wants a cooler to make full use of his CPU. I don't mean you should use a CPU unlocked. I mean you should test the CPU unlocked, to see actual power limitations of the cooler. Since you are already controlling for noise, temperatures do not matter as much (as long as they are below 90). A better metric is power consumption, because different CPUs use different power.
@G19rom both AM5 CPUs tested provided higher Cinebench R23 scores in their tuned configurations tested, vs. running stock. I don't think you understand how controlled testing works. I've explained the inherit issue with your methods/request and why it will produce bad data.
@@L91 Again, I am suggesting that you test with CPU unlocked, only for testing cooler power limits, not for actual use. CPU temperatures are no longer the best metric for cooler performance, since a CPU will OC itself until it reaches the temperature limit (usually 90 degrees). At fixed temperature (Tj max or for example 90), fixed noise (your control), each cooler will be able to cool a certain power (in Watts). This power value is useful for evaluating cooler performance. You will only need to test with one power hungry CPU for all configurations. In other words, for example: I would like to know that an L9a can cool 72W in a sandwich case, while the APX90-36 can do 77W. For example, in your data, if these coolers are used with an 88W CPU, temperature is not a good indicator of performance, since the CPU clock drops. Both of them reach 93 and 94 degrees (the temperature limit), but they are running at different power (for example, 72W and 77W, which is not shown).
@G19rom the power limits are fixed, all are running at set power limit of 88W for 7600X. In layman's terms, all are pulling 88W for 7600X & 105W for 7700X during testing. Unlocking the power limit will attually make performance worse. I'm not sure how much clearer I can be.
Any chance you could run through how you got the Alpenfohn BR to be compatible with AM5? Last I heard, board manufacturers changed the socket layout so the AM4 mounting kit was no longer compatible even with the Thermal Grizzly short backplate. Cheers mate, great video 👌👌
I modified the included brackets by removing the struts on them so they are flat. Alpenfohn is working on the AM5 kit, though. I used hand saw to cut then off and just sanded the surface smooth/flat
@@L91 Thanks for the reply, I was hoping there was a hardware fix over altering the existing hardware, but if Alpenfohn is engineering an AM5 socket, then I'll wait for that :D
The Ryzen 7 7700X temp chart is the same as the Core i7-13700K temp chart. Is this a mistake? Or do the temps for all the coolers happen to be exactly the same?
@@L91 yeah I get that, but the 7700x and the 13700k graphs are *the same*, like 7700x and the 13700k have the same temperatures for each and every cooler. I'm asking this because the 13700k graph is not whatnI've seen in your Thermalright SI-100 review.
@CsQ_RandomRepository oh shit, you're right. Thanks for pointing this out. Yeah, I've inserted the wrong graph in for 13700K. I'm not sure how I missed that or how everyone else did. The SI-100 review does have the correct info for 13700K.
So Thermalright's $25 SI-100 outperforms their own AXP120-X67 that costs twice as much? Damn Edit: nvm I thought they were the same size but the SI-100 is a good bit larger overall so that makes sense.
Is it possible for you to do an open-air bench where the motherboard is mounted vertically? I've heard that it will affect the efficiency of heat pipes.
When I test the R7 78003D, I'll do some testing vertically for reference. This won't be for while, though, as I'm working on other content for next month.
@@L91 One more stupid request: is it possible to add is-60 to your test or just add a 25mm top fan to is47k? So that there is the "best air cooling solution under 74mm" namely is-60 with 25mm top fan, x67, l12s, is55 with t30 top fan. I only know that black ridge with 12025 top fan is currently the best option(it out performs the L12s by a measurable margin but it doesn't come with am5 bracket and it's a little bit pricey)
A good test but it would have been better to show the performance vs the height of each cooler. The graph suggests the ones near the bottom are bad but in reality the best performers are much larger coolers so will always be better.
@@L91I mean Asus motherboard, not gpu. If your mobo is silent maybye you can share manufacturing date? Heard a man with mobo from March 2023 doesn't have this issue
@MrKony98 yes, I'm referring to the motherboard. I get VRM noise during gaming with 6900XT, but not 3060 Ti. Honestly it's wierd. My board is from November 2022
I notice the IS-55 performs much worse here than in your previous test, compared to the X47 and X53: ruclips.net/video/P6isZN2_HCQ/видео.html In that test, it was able to do 79C for a 130W solo run (albeit with the fan at 100%). Any idea what the reason for this might be?
I changed the orientation of the X47, X53 coolers vs. previous test. I explained that the X47, X53 perform much better in orientation shown at the start of the video.
@@L91 Oops, that's what I get for going straight to the graphs instead of watching the video! However, it still seems to be the case that the IS-55 is performing worse than in the previous test. If it's doing 95C at 105W, how was it doing 79C at 130W before? Multicore vs. singlecore?
@archagon honestly, there could be a number of factors, including motherboard changes, GPU installed vs. not installed, different platforms, different CPUs, using included fan vs. 3rd party fan. It's honestly been so long since I did the original test. IMO, I just perform the tests and present the results, and I'm still learning myself. If you're using AM5, then these results are most accurate results. I also believe these results to be more accurate, as the testing was more controlled vs. previously.
There's no errors, coolers have been tested across range of CPUs, and im also currently testing 7800X3D which lines up with existing data. At a certain point, air coolers have diminished returns, the larger they get, and difference margins decrease with lower end CPUs for most part. The SI-100 is an excellent cooler. Thermal rights make great coolers with extremely competitive pricing. Unfortunately, only you can decide what to make of it.
I'd like to note that the temperature graph for 13700K at 125W is a duplicate of R7 7700X in error. You can find all the graphs, including the correct 13700K graph in written review here: reviewer.co/coolers/thermalright-si-100
Incredible how the AXP90-x47 copper can hold its own against the larger coolers.
Straight to the point, thank you. Very hard to find a proper test suite for the Thermalright axp90 series.
Thanks, very quick yet detailed and informative. Picked up an SI-100 (black) on sale for $22 and counldn't happier
Thank you for making this. Low profile coolers and use on CPUs over 65W... I always get them confused. I hope you'll be able to make a vid w/ the upcoming TR AXP90-52 FULL copper. Have a good one
The SI-100 is only 30 dollars and comes in all black??? This company is smashing it out of the park.
I would like to see this cooler vs be quiet! Pure Rock LP 100w CPU cooler. Thanks for the video.
BeQuiet wouldn't send one when I enquired, so cannot help. I can say that manufacturer wattage ratings are meaningless and it certainly cannot handle 100W on AM5 CPU.
@@L91 Send me your adress i will send you a Pure Rock LP, I am also sure that it cannot cool an 100 TDP CPU
Thanks for your videos really useful info
very informative, thanks
Would have loved to see comparison with stock cooler. And why the x37 disappeared in other tests? 😢
so getting the axp90-x53 is not worth it? what about axp90-x47 with 25mm fan swap?
Curious if you have any data on the AXP90-X47 Copper vs Non-copper version. :)
if only SI-100 can fit to fractal terra with a 7900xt reference card that would be awesome
Just to clarify - the fans are set at different speeds to normalize for noise, correct?
Correct. Outside of running these fans at full speed, slight variations of speeds tested have no impact on cooling performance.
Do you think you can make a video showing how to setup the black ridge on AM5? I've seen a lot of people say it's possible but no tutorials or other sources on how to do it
SI-100 outperform Noctua C14S! is this because of the low profile of C14S?
The SI-100 is 15mm shorter than C14S, both are low profile designs. The SI-100 just better cooler.
@@L91 I meant the fan in low profile mode, the fan below the fens
@HD Master so did I, the SI-100 is only 100mm tall, including the fan
how is it compare to JF13K
which one would you say will do better x53 with stock/slim or x47 with 25mm fan? the cpu is r7600 :)
Thanks, i would buy thermalright axp-90 x53 white for my itx build, i would use 12100f and i would upgrade to 13600k after, i would use him in stock with 1 92mm in the case. Thanks
That's why I think using low profiles coolers for CPU above 65 W TDP is not optimal 🙂
Zen 4 CPUs have very poor thermal transfer, due to very thick IHS. Testing was also carried out in the worst-case scenario use case for given CPU. Wattage doesn't mean anything if the CPU has poor thermal properties. One CPU generation or platforms thermal properties aren't universal either.
Where is noctua l9x65?
Thank you for the tests. Can you please provide CPU package power readings for the different coolers? For the ones at the bottom, it is likely the CPU is throttling. Ideally, if we can have the CPU package power for an unlocked CPU in a sandwich case (or something like the fractal ridge), we can get an approximate TDP for the different coolers in an SFF setting, which is highly valuable information.
Package power, also known at PPT for AMD, is locked as stated on grapths. AM5 CPUs have abysmal performance/tuning out of the box and should never be ran with unlocked limits as they simply perform worse on lower end cooling solutions that cannot constrain them. No CPU cooler shown are thermal throttling, but lowest-end solutions are running slightly lower all core clock speeds as demonstrated. There is no one case fits all for all of these coolers. In order for things to be as fair as possible, variables needed to be limited as much as possible, power packages targets being one of the most important
@@L91 Thanks for the reply. I think practically, a user just wants a cooler to make full use of his CPU. I don't mean you should use a CPU unlocked. I mean you should test the CPU unlocked, to see actual power limitations of the cooler. Since you are already controlling for noise, temperatures do not matter as much (as long as they are below 90). A better metric is power consumption, because different CPUs use different power.
@G19rom both AM5 CPUs tested provided higher Cinebench R23 scores in their tuned configurations tested, vs. running stock. I don't think you understand how controlled testing works. I've explained the inherit issue with your methods/request and why it will produce bad data.
@@L91 Again, I am suggesting that you test with CPU unlocked, only for testing cooler power limits, not for actual use. CPU temperatures are no longer the best metric for cooler performance, since a CPU will OC itself until it reaches the temperature limit (usually 90 degrees). At fixed temperature (Tj max or for example 90), fixed noise (your control), each cooler will be able to cool a certain power (in Watts). This power value is useful for evaluating cooler performance. You will only need to test with one power hungry CPU for all configurations.
In other words, for example: I would like to know that an L9a can cool 72W in a sandwich case, while the APX90-36 can do 77W.
For example, in your data, if these coolers are used with an 88W CPU, temperature is not a good indicator of performance, since the CPU clock drops. Both of them reach 93 and 94 degrees (the temperature limit), but they are running at different power (for example, 72W and 77W, which is not shown).
@G19rom the power limits are fixed, all are running at set power limit of 88W for 7600X. In layman's terms, all are pulling 88W for 7600X & 105W for 7700X during testing. Unlocking the power limit will attually make performance worse. I'm not sure how much clearer I can be.
Any chance you could run through how you got the Alpenfohn BR to be compatible with AM5?
Last I heard, board manufacturers changed the socket layout so the AM4 mounting kit was no longer compatible even with the Thermal Grizzly short backplate.
Cheers mate, great video 👌👌
I modified the included brackets by removing the struts on them so they are flat. Alpenfohn is working on the AM5 kit, though. I used hand saw to cut then off and just sanded the surface smooth/flat
@@L91 Thanks for the reply, I was hoping there was a hardware fix over altering the existing hardware, but if Alpenfohn is engineering an AM5 socket, then I'll wait for that :D
@AxleLotl I had spare kit, so used that, but since it's not official, decided to drop it from testing with 7800X3D charts
Do you have any tests of the Scythe Mugen 5 Rev. C with their Kaze Flex II fan?
I don't sorry, Scythe wouldn't even send the BS3 for testing/comparison, and had to purchase it myself for inclusion.
The Ryzen 7 7700X temp chart is the same as the Core i7-13700K temp chart. Is this a mistake? Or do the temps for all the coolers happen to be exactly the same?
13700K is only running at 125W
@@L91 yeah I get that, but the 7700x and the 13700k graphs are *the same*, like 7700x and the 13700k have the same temperatures for each and every cooler. I'm asking this because the 13700k graph is not whatnI've seen in your Thermalright SI-100 review.
@CsQ_RandomRepository oh shit, you're right. Thanks for pointing this out. Yeah, I've inserted the wrong graph in for 13700K. I'm not sure how I missed that or how everyone else did. The SI-100 review does have the correct info for 13700K.
So Thermalright's $25 SI-100 outperforms their own AXP120-X67 that costs twice as much? Damn
Edit: nvm I thought they were the same size but the SI-100 is a good bit larger overall so that makes sense.
Is it possible for you to do an open-air bench where the motherboard is mounted vertically? I've heard that it will affect the efficiency of heat pipes.
When I test the R7 78003D, I'll do some testing vertically for reference. This won't be for while, though, as I'm working on other content for next month.
@@L91 One more stupid request: is it possible to add is-60 to your test or just add a 25mm top fan to is47k? So that there is the "best air cooling solution under 74mm" namely is-60 with 25mm top fan, x67, l12s, is55 with t30 top fan. I only know that black ridge with 12025 top fan is currently the best option(it out performs the L12s by a measurable margin but it doesn't come with am5 bracket and it's a little bit pricey)
A good test but it would have been better to show the performance vs the height of each cooler. The graph suggests the ones near the bottom are bad but in reality the best performers are much larger coolers so will always be better.
Thanks for the feedback. Please see here: www.caselabs.org/coolers/low-profile-coolers
Does your board also suffer from VRM whine? Many people complain about ASUS AM5 boards VRMs being noisy on idle
Not at idle, or during heavily multithreaded applications like Cinebench R23. I've only noticed it during gaming gaming and only certain GPUs
@@L91I mean Asus motherboard, not gpu. If your mobo is silent maybye you can share manufacturing date? Heard a man with mobo from March 2023 doesn't have this issue
@MrKony98 yes, I'm referring to the motherboard. I get VRM noise during gaming with 6900XT, but not 3060 Ti. Honestly it's wierd. My board is from November 2022
have you tried the 7700 and 7900 non X with this coolers?
7700 non-x won't be much different from X, haven't got any cooling info on 7900 non-x
@@L91 the 13700k that you tried. have you undervolted that or just Power limit? how was the performance when you PL it to 125W?
@@blindman8093 just power limited, didn't check Cinebench scores since it was just for comparison and data validation
@@L91 ohh i see. thank you very much.
I notice the IS-55 performs much worse here than in your previous test, compared to the X47 and X53: ruclips.net/video/P6isZN2_HCQ/видео.html
In that test, it was able to do 79C for a 130W solo run (albeit with the fan at 100%).
Any idea what the reason for this might be?
I changed the orientation of the X47, X53 coolers vs. previous test. I explained that the X47, X53 perform much better in orientation shown at the start of the video.
@@L91 Oops, that's what I get for going straight to the graphs instead of watching the video!
However, it still seems to be the case that the IS-55 is performing worse than in the previous test. If it's doing 95C at 105W, how was it doing 79C at 130W before? Multicore vs. singlecore?
@archagon honestly, there could be a number of factors, including motherboard changes, GPU installed vs. not installed, different platforms, different CPUs, using included fan vs. 3rd party fan. It's honestly been so long since I did the original test. IMO, I just perform the tests and present the results, and I'm still learning myself. If you're using AM5, then these results are most accurate results. I also believe these results to be more accurate, as the testing was more controlled vs. previously.
How is Si-100 so powerful? Are you sure there is no error?
There's no errors, coolers have been tested across range of CPUs, and im also currently testing 7800X3D which lines up with existing data. At a certain point, air coolers have diminished returns, the larger they get, and difference margins decrease with lower end CPUs for most part. The SI-100 is an excellent cooler. Thermal rights make great coolers with extremely competitive pricing. Unfortunately, only you can decide what to make of it.
I appreciate the effort, but it's kinda useless without noise levels...