Ryzen 3700X bad temps because of bad IHS?

[Alvin853]

My main system is running a 3700X cooled by a Galahad 240 AIO, and my temps always seemed a bit on the high side, considering this is only a 65W TDP CPU. I'm seeing idle temps in the high 30s to low 40s and full load temps in high 70s hitting 80°C sometimes. Not really what I was expecting from a high quality 240mm AIO. I did remount/repaste everything, to make sure it wasn't a bad mount, but the results are the same.

 

I also have a Ryzen 5600G for a secondary system, 65W TDP, so I decided to do a test and put the 5600G into my system, and the temps are much lower now, idling in the low 30s, full load around 70°C, more reasonable values for a 240mm AIO, as I have a SFF PC and not expecting crazy cooling capabilities. And this is even with the fans running slower, because I didn't bother to change the fan curve. So clearly the AIO is capable of performing much better than it does on the 3700X, and the only difference is the CPU itself.

 

When taking out the 3700X I notice a lot of thermal paste accumulating in some "pools" on the IHS, can an uneven IHS affect cooling performance this much?

[Bombastinator]

29 minutes ago, Alvin853 said:

My main system is running a 3700X cooled by a Galahad 240 AIO, and my temps always seemed a bit on the high side, considering this is only a 65W TDP CPU. I'm seeing idle temps in the high 30s to low 40s and full load temps in high 70s hitting 80°C sometimes. Not really what I was expecting from a high quality 240mm AIO. I did remount/repaste everything, to make sure it wasn't a bad mount, but the results are the same.

 

I also have a Ryzen 5600G for a secondary system, 65W TDP, so I decided to do a test and put the 5600G into my system, and the temps are much lower now, idling in the low 30s, full load around 70°C, more reasonable values for a 240mm AIO, as I have a SFF PC and not expecting crazy cooling capabilities. And this is even with the fans running slower, because I didn't bother to change the fan curve. So clearly the AIO is capable of performing much better than it does on the 3700X, and the only difference is the CPU itself.

 

When taking out the 3700X I notice a lot of thermal paste accumulating in some "pools" on the IHS, can an uneven IHS affect cooling performance this much?

Can it?  Yes.  GN keeps a special cooler around to demonstrate what a bad cold plate can do.

[Alvin853]

6 minutes ago, Bombastinator said:

Can it?  Yes.  GN keeps a special cooler around to demonstrate what a bad cold plate can do.

Well yeah for sure it can, and I realize coolers have huge variety, but I'd expect IHSs to be much more standardized, a temp difference of 10 or more doesn't seem like good quality control on the manufacturing side at AMD. For sure IHSs aren't always perfectly flat, that's why lapping is a thing, but I always thought that was just a minor improvement. Maybe there are other factors, such as die size, location of the temp sensors, contact between die and IHS. 

[Biohazard777]

42 minutes ago, Alvin853 said:

My main system is running a 3700X cooled by a Galahad 240 AIO, and my temps always seemed a bit on the high side, considering this is only a 65W TDP CPU. I'm seeing idle temps in the high 30s to low 40s and full load temps in high 70s hitting 80°C sometimes. Not really what I was expecting from a high quality 240mm AIO. I did remount/repaste everything, to make sure it wasn't a bad mount, but the results are the same.

 

I also have a Ryzen 5600G for a secondary system, 65W TDP, so I decided to do a test and put the 5600G into my system, and the temps are much lower now, idling in the low 30s, full load around 70°C, more reasonable values for a 240mm AIO, as I have a SFF PC and not expecting crazy cooling capabilities. And this is even with the fans running slower, because I didn't bother to change the fan curve. So clearly the AIO is capable of performing much better than it does on the 3700X, and the only difference is the CPU itself.

 

When taking out the 3700X I notice a lot of thermal paste accumulating in some "pools" on the IHS, can an uneven IHS affect cooling performance this much?

You are placing too much trust in AMD TDP ratings... 

https://www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/19

https://www.anandtech.com/show/16824/amd-ryzen-7-5700g-and-ryzen-5-5600g-apu-review/2
Also when you post temperatures... they are kinda meaningless without knowing what the ambient temperature is.

[SavageNeo]

seems normal

[Alvin853]

39 minutes ago, Biohazard777 said:

You are placing too much trust in AMD TDP ratings... 

https://www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/19

https://www.anandtech.com/show/16824/amd-ryzen-7-5700g-and-ryzen-5-5600g-apu-review/2
Also when you post temperatures... they are kinda meaningless without knowing what the ambient temperature is.

Ambient temperatures sure aren't constant (usually in the 21-24°C range in my room), but they didn't vary by 10 degrees between the 3700X and the 5600G tests. And while TDP is not very accurate, the anandtech tests show the 3700X maxing out at 90W package power, and the 5600G at 81W, so slightly lower, but that shouldn't account for a 10 degree difference in temps. 

[Bombastinator]

1 hour ago, Alvin853 said:

Well yeah for sure it can, and I realize coolers have huge variety, but I'd expect IHSs to be much more standardized, a temp difference of 10 or more doesn't seem like good quality control on the manufacturing side at AMD. For sure IHSs aren't always perfectly flat, that's why lapping is a thing, but I always thought that was just a minor improvement. Maybe there are other factors, such as die size, location of the temp sensors, contact between die and IHS. 

They are, partially because there are only two manufacturers. Metal is still metal though.  It could be the ihs, it’s apparently not the cold plate.  If you have a bumpy ihs lapping is an option.  It has issues though, and there is a lot of not-so-great advice on how to do it. I suggest doing lots of reading on the subject before it is attempted.  Lapping is a very very old methodology that peaked about the first quarter of the last century like a lot of analog and machining methods. It’s still done, as it’s the only way to achieve flatness of a level where actual physics properties change in some weird ways.  Competitive overclockers have been known to use it.  It’s a lot of work, and if you go too far into the metal it can make delidding and bare die the only option to make the chip work.  Even a tiny amount of it destroys resale value. It’s not something you want to screw up.

[Jeppes]

Look up the real power consumption. I would guess stock 3700X with decent cooler is 90W or more and 5600G is 65W.

[Bombastinator]

1 minute ago, Jeppes said:

Look up the real power consumption. I would guess stock 3700X with decent cooler is 90W or more and 5600G is 65W.

I heard mid to low 80s.  A multi fan watercooler should handle that though if it’s any good

[Alvin853]

19 minutes ago, Jeppes said:

Look up the real power consumption. I would guess stock 3700X with decent cooler is 90W or more and 5600G is 65W.

The Anandtech reviews were already linked above, 3700X seems to max out right at 90W, the 5600G at 81W, so not that much of a difference.

[Jeppes]

43 minutes ago, Alvin853 said:

The Anandtech reviews were already linked above, 3700X seems to max out right at 90W, the 5600G at 81W, so not that much of a difference.

One can only guess what the numbers are in your system and if you tested in a way that provides same numbers every time. Would also be nice to know how you run it as you face about the same temps that GN got with a 200W 3950X with fans 1000rpm on their review.

[Alvin853]

5 minutes ago, Jeppes said:

One can only guess what the numbers are in your system and if you tested in a way that provides same numbers every time. Would also be nice to know how you run it as you face about the same temps that GN got with a 200W 3950X with fans 1000rpm on their review.

My BIOS is running all default/auto except XMP enabled, my full load test is running CB R23 multi-thread looped for 10 minutes, see where the temps end up, idle test is after load test close everything, let it sit for a couple minutes and see where the temps bottom out. Yes some background programs will still be running, but they're the same in all tests, so my results might not be comparable to other people's results (especially since my SFF case might be a limiting factor), but they have the same conditions between my 2 CPUs. The GN review was one of the main sources why I got the Galahad AIO, so I was expecting much better cooling performance than what I'm seeing. At first I was worried my AIO was bad, but it performs pretty much as expected with the 5600G, just not with the 3700X.

[Biohazard777]

5 hours ago, Alvin853 said:

Ambient temperatures sure aren't constant (usually in the 21-24°C range in my room), but they didn't vary by 10 degrees between the 3700X and the 5600G tests. And while TDP is not very accurate, the anandtech tests show the 3700X maxing out at 90W package power, and the 5600G at 81W, so slightly lower, but that shouldn't account for a 10 degree difference in temps. 

TDP is different from power draw, Thermal Design Power is meant as a measure of how much heat a chip can produce (max) so that appropriate cooling solutions can be easily picked.
I don't find it unreasonable that a chip with 65W TDP is actually drawing 90W under full load, that could be fine... What is unreasonable is to have the same architecture (Zen2) Ryzen 5 3600 with the same 65W TDP which comes with Wraith Stealth (tiny, all aluminium, piece of sh*t) in the box while Ryzen 7 3700X comes with Wraith Prism (decent cooler with copper heat pipes) and is somehow also 65W TDP. 
TDP rating is kinda pointless when it comes to AMD (and Intel heh), serves more as a marketing pitch:

Here, I did a quick test on one of the desktops in my house, it has 3700X with stock Wraith Prism cooler, stock paste inside a mediocre quality full size tower (3x120mm intake fans, but it doesn't have a mesh front, instead it is solid front and has some side holes for intake):
Idle (kinda pointless, but why not):

Spoiler

Full Load on CPU (3 minutes, stopped rising and wouldn't have gone much higher than that):

Spoiler

85.5°C with ambient of 21°C, so there you have it, your AIO is doing a better job than (my) stock Wraith Prism . You also probably don't have your fans ramping up like I have, so less noise. IMO your AIO is doing a solid job if you are putting synthetic (not normal) loads on it.

 

3 hours ago, Alvin853 said:

The Anandtech reviews were already linked above, 3700X seems to max out right at 90W, the 5600G at 81W, so not that much of a difference.

That is the peak power draw, 5600G has an APU, if that is utilized or not will also have an impact (not familiar with tests used by Anandtech where it peaked to 81W).
Even if the efficiency is the same and that difference is just ~11% more  (81W + 8.91W = 89.91W) , it is not negligible.

But when it comes to heat dissipation what you really want to look at is sustained, over time, draw:
For 5600G:

What I'm seeing here is ~77W.
And for 3700X:

It is 88W sustained.
Let's compare those two, increase of ~14% (77W + 10.78W = 87.78W).
Now how about your temps, well to have meaningful results we need be looking only at the delta (above ambient):
70°C with ambient of 21°C is equal to 49°C delta. 14% of that delta is 6.86°C , let's round that up to 7°C.
And you are getting something like 10°C difference, with your ambient fluctuating 3°C (21-24), if you weren't concerned about ambient when you took those temps then you shouldn't be concerned about 3°C diff. 
* Before someone calls me out for not using Kelvins:

Spoiler

1 unit increase in the Kelvin scale is the same as a 1 degree increase in the Celsius scale. The only difference between these two scales is where the zero points are.
Meaning you'd get the same result:

Ambient: 21°C = 294.15K
Load 5600G: 70°C = 343.15K
ΔT:  343.15K - 294.15K = 49K
14% of that ΔT = 0.14* 49K = 6.86K
Load 5600G + increase over delta: 343.15K + 6.86K = 350.01K = 76.86°C ~= 77°C

There is a lot I haven't covered, keeping it simple. Like what is the efficiency of each chip, how is the air moving inside the case (it can differ based on ambient inside the case and outside), is the extra power draw also causing power delivery circuitry on the mobo and the PSU to give out excess heat and thus raise the ambient inside the case, etc.

TL:DR
- 3700X and 5600G are not the same, the difference in heat output is not negligible.
- AMD TDP ratings are crap.
- If you are satisfied with the temps your are getting with 5600G you should more or less expect what you are getting with 3700X. Maybe you can shave off a couple of of degrees but I personally wouldn't be lapping the IHS for maybe 3°C difference.
- Getting a bigger AIO is also quite pointless.
- Be happy with what you've got, use your PC and don't let it use you / your time. 

[Bombastinator]

4 hours ago, Biohazard777 said:

TDP is different from power draw, Thermal Design Power is meant as a measure of how much heat a chip can produce (max) so that appropriate cooling solutions can be easily picked.
I don't find it unreasonable that a chip with 65W TDP is actually drawing 90W under full load, that could be fine... What is unreasonable is to have the same architecture (Zen2) Ryzen 5 3600 with the same 65W TDP which comes with Wraith Stealth (tiny, all aluminium, piece of sh*t) in the box while Ryzen 7 3700X comes with Wraith Prism (decent cooler with copper heat pipes) and is somehow also 65W TDP. 
TDP rating is kinda pointless when it comes to AMD (and Intel heh), serves more as a marketing pitch:

Here, I did a quick test on one of the desktops in my house, it has 3700X with stock Wraith Prism cooler, stock paste inside a mediocre quality full size tower (3x120mm intake fans, but it doesn't have a mesh front, instead it is solid front and has some side holes for intake):
Idle (kinda pointless, but why not):

  Reveal hidden contents

Full Load on CPU (3 minutes, stopped rising and wouldn't have gone much higher than that):

  Reveal hidden contents

85.5°C with ambient of 21°C, so there you have it, your AIO is doing a better job than (my) stock Wraith Prism . You also probably don't have your fans ramping up like I have, so less noise. IMO your AIO is doing a solid job if you are putting synthetic (not normal) loads on it.

 

That is the peak power draw, 5600G has an APU, if that is utilized or not will also have an impact (not familiar with tests used by Anandtech where it peaked to 81W).
Even if the efficiency is the same and that difference is just ~11% more  (81W + 8.91W = 89.91W) , it is not negligible.

But when it comes to heat dissipation what you really want to look at is sustained, over time, draw:
For 5600G:

What I'm seeing here is ~77W.
And for 3700X:

It is 88W sustained.
Let's compare those two, increase of ~14% (77W + 10.78W = 87.78W).
Now how about your temps, well to have meaningful results we need be looking only at the delta (above ambient):
70°C with ambient of 21°C is equal to 49°C delta. 14% of that delta is 6.86°C , let's round that up to 7°C.
And you are getting something like 10°C difference, with your ambient fluctuating 3°C (21-24), if you weren't concerned about ambient when you took those temps then you shouldn't be concerned about 3°C diff. 
* Before someone calls me out for not using Kelvins:

  Reveal hidden contents

1 unit increase in the Kelvin scale is the same as a 1 degree increase in the Celsius scale. The only difference between these two scales is where the zero points are.
Meaning you'd get the same result:

Ambient: 21°C = 294.15K
Load 5600G: 70°C = 343.15K
ΔT:  343.15K - 294.15K = 49K
14% of that ΔT = 0.14* 49K = 6.86K
Load 5600G + increase over delta: 343.15K + 6.86K = 350.01K = 76.86°C ~= 77°C

There is a lot I haven't covered, keeping it simple. Like what is the efficiency of each chip, how is the air moving inside the case (it can differ based on ambient inside the case and outside), is the extra power draw also causing power delivery circuitry on the mobo and the PSU to give out excess heat and thus raise the ambient inside the case, etc.

TL:DR
- 3700X and 5600G are not the same, the difference in heat output is not negligible.
- AMD TDP ratings are crap.
- If you are satisfied with the temps your are getting with 5600G you should more or less expect what you are getting with 3700X. Maybe you can shave off a couple of of degrees but I personally wouldn't be lapping the IHS for maybe 3°C difference.
- Getting a bigger AIO is also quite pointless.
- Be happy with what you've got, use your PC and don't let it use you / your time. 

Wow that was a long post.  Even 95 watts should still be handleable by a 4 pipe air cooler though.  Those things should even be able to handle 125w without much trouble.  Given of course that it is a decent 240.  There are some that aren’t.

[Biohazard777]

51 minutes ago, Bombastinator said:

Wow that was a long post.

Yeah, I got carried away heh.
 

51 minutes ago, Bombastinator said:

Even 95 watts should still be handleable by a 4 pipe air cooler though.  Those things should even be able to handle 125w without much trouble.  Given of course that it is a decent 240.  There are some that aren’t.

My point was just that he shouldn't be putting = between 5600G and 3700X, that will impact his max temp even with a perfectly flat IHS.
He mentioned he is running it inside an SFF case, which makes it really hard to compare to tests online (like the GN tests @ 200W load). So I honestly don't know if his AIO is underperforming or not, I just know this isn't something I would spend time troubleshooting since he is way bellow problematic temps.

 

Edit: some random reddit post I stumbled upon now:

Spoiler

Yeah not Galahad but similar enough... again so many variables to get to the bottom of this, sure 23°C for ambient but what kind of case RiftBlade has, has he tweaked voltages in bios or not, did he mount it with enough pressure, what kind of thermal paste he used, what kind of load he considers full load. Again IMO not worth spending time on.

[Frenky]

12 hours ago, Alvin853 said:

My main system is running a 3700X cooled by a Galahad 240 AIO, and my temps always seemed a bit on the high side, considering this is only a 65W TDP CPU. I'm seeing idle temps in the high 30s to low 40s and full load temps in high 70s hitting 80°C sometimes. Not really what I was expecting from a high quality 240mm AIO. I did remount/repaste everything, to make sure it wasn't a bad mount, but the results are the same.

 

I also have a Ryzen 5600G for a secondary system, 65W TDP, so I decided to do a test and put the 5600G into my system, and the temps are much lower now, idling in the low 30s, full load around 70°C, more reasonable values for a 240mm AIO, as I have a SFF PC and not expecting crazy cooling capabilities. And this is even with the fans running slower, because I didn't bother to change the fan curve. So clearly the AIO is capable of performing much better than it does on the 3700X, and the only difference is the CPU itself.

 

When taking out the 3700X I notice a lot of thermal paste accumulating in some "pools" on the IHS, can an uneven IHS affect cooling performance this much?

I have a 3700X with a NH-D15 with 2 fans and a high airflow case, and the cpu still gets toasty. Even now when I'm watching a movie it jumps around between 40c and 50c constantly. When gaming it easily reaches 70c+. I think what you're describing is normal and nothing to worry about.