WTF Updated Ryzen 3600 Runs COOL AS SHIT?

[DWD1961]

I just had my 2019 Ryzen 3600 replaced because it was bad. AMD sent me a brand new one.

P95 torture test as hottest setting, default.

I posted a couple years ago about the heat, and it seemed like everyone was having heat issues even without OCing those chips. I never could get it under 80C in a well ventilated case with a 240 AIO:

With the new 3600, I can't get it to bust 70C (a short bust, but runs wide open at 69-70C)
Ambient taken with a heat gun 4 feet from the computer is 22C. That's in a case with ONE 120mm fan and a 120MM AIO. WTF? All I can think of was that AMD must have done something to the newer 3600s? Anyone have any ideas why this is?

 

[toast0]

Maybe some of this is software/firmware updates working for you. But a lot of it is also probably process improvements. Usually, when they first start making the chips, they're binning them close to the limits; towards the end of the run, they are making a lot more great chips, and even mid/lower end skus get good chips that run pretty cool.

 

[MaZa]

IIRC some later 3000 serie Ryzens were actually XT models that just didn't cut the mustard. But they do run cooler and are more overclockable than previous bins.

 

[daglesj]

Hmm I've built several 3600 machines the past couple of years and all were fine thermally. Interesting. I'll keep an eye out. Thanks for the heads up.

 

[jlbenedict]

It's interesting that these 3600's are also popping back up on Newegg and MC for sale, new.
Maybe they are tweaked and refined
Who knows

 

[Legendary Gamer]

Better Binned silicon. It's likely a late generation production run.

 

[DWD1961]

Hmm I've built several 3600 machines the past couple of years and all were fine thermally. Interesting. I'll keep an eye out. Thanks for the heads up.

My first 3600 was bought in early Jan of 2020 and it was defintely a 2019 model. First gen. Maybe they fixed them?

 

[DWD1961]

Well, whatever it is, damn this thing runs cool and nearly at its factory OC too.

 

[freeagentt]

My first AM4 chip was a 3600XT that was the same price as the vanilla. It was such a cool running chip. It would do 4400MHz static with 1.26, and 4500 with 1.33v. Temps were no problem at all. I was running those speed on Le Grand Macho RT with no CPU fan attached and was Linpack stable lol. Who knows, maybe your new 3600 will be similar..

 

[riev90]

The chip replacement you got probably of the newer batch / revision, therefore could run a lot cooler with a lot lower vid / vcore (from 2019's batch 1.35~1.4v default to 2020 up's batch 1.15~1.2v).

I guess that chip is even capable to run 4.5ghz at only 1.2~1.25v with the cooler still can tame it to under 80c.

 

[DWD1961]

The chip replacement you got probably of the newer batch / revision, therefore could run a lot cooler with a lot lower vid / vcore (from 2019's batch 1.35~1.4v default to 2020 up's batch 1.15~1.2v).

I guess that chip is even capable to run 4.5ghz at only 1.2~1.25v with the cooler still can tame it to under 80c.

It's how it should be, you know? Those first gens were way too hot.

Are you saying that I could OC it to 4.5Ghz at the same voltage as factory? Unless I get a new card that requires extra CPU power, I probably would not do it. I just have no reason to OC it. The last time I OCed it was a first gen Intel i7 920, which I clocked from 2660 factory to 3900 on air, nice and cool and stable. I only did that because it was the same speed as the 980 was doing at 3300, but twice the price. So, I got a 980 for a 920 price. But that OC was different because it used a LOT more power to do the OC. But from 2008 up to 2013 I had no bottle necks with the highest end card out at the time, which was a nVidia 295. Point is that I don;t need extra clocks, except to run a video card for gaming at max FPS. In other words, my CPUs last a loooooong time. lol

 

[riev90]

This is my CB R15 score using ryzen 3600 from 2019's batch.

This is with 2020's batch.

Here is the CB R15 summary:
2019 batch:
Single score 189, Multi 1543, vcore auto 1.43v, power cons 100w.
2020 batch:
Single score 213 (12.7% increase), Multi 1804 (17% increase), vcore 1.33v (drop 7%), power cons. max 88w (drop 12%)

I suggest you to try to overclock it, your cooler is sufficient, and the only thing you need to do:
1. Go to bios, then set the vcore to 1.2 or 1.25v
2. Set the cpu multiplier to 45x
Just run cinebench r23 10 minutes loop and if no crash, it shall be stable for daily use. No need to fear of cpu degrade as the 1.2 ~ 1.25v or even 1.3v is still safe for daily usage of ryzen.

 

[DWD1961]

View attachment 534401

This is my CB R15 score using ryzen 3600 from 2019's batch.

View attachment 534402

This is with 2020's batch.

Here is the CB R15 summary:
2019 batch:
Single score 189, Multi 1543, vcore auto 1.43v, power cons 100w.
2020 batch:
Single score 213 (12.7% increase), Multi 1804 (17% increase), vcore 1.33v (drop 7%), power cons. max 88w (drop 12%)

I suggest you to try to overclock it, your cooler is sufficient, and the only thing you need to do:
1. Go to bios, then set the vcore to 1.2 or 1.25v
2. Set the cpu multiplier to 45x
Just run cinebench r23 10 minutes loop and if no crash, it shall be stable for daily use. No need to fear of cpu degrade as the 1.2 ~ 1.25v or even 1.3v is still safe for daily usage of ryzen.

That's interesting. It looks like the 2019 models were dogs. Can you test that 2019 model in default settings with P95 and see what heat it produces?

 

[riev90]

That's interesting. It looks like the 2019 models were dogs. Can you test that 2019 model in default settings with P95 and see what heat it produces?
View attachment 534493

Unfortunately I no longer have those 2 cpus
As for checking the maximum heat produced by CPU, IMHO:
>>> Cinebench R23 with 10~30 minutes looping
>>> y-cruncher with Pi-2.5 and up

 

[Vengance_01]

My guess is purely silicone lottery, better die contact with IHS and better contact with AIO block. Any number of things could be helping.

 

[SmokeRngs]

My guess is purely silicone lottery, better die contact with IHS and better contact with AIO block. Any number of things could be helping.

It's more likely it was an issue with the mounting or contact between the heatspreader and cooler. There's absolutely no way to run that CPU at stock and get the temps he was unless something was seriously wrong with the mounting or cooling system. Hell, it sounds like the plastic on the waterblock to protect it during shipping wasn't removed. My absolute shit 5800x with a cheap Enermax 240 AIO runs cooler and it eats up a hell of a lot more power with a smaller area for the die to transfer heat.

I have no doubt a newer CPU would run cooler than an old one due to better yields in the silicon but nothing like what the OP is seeing.

 

[crazycrave]

My very early Aug 2019 is still running on an MSI B450 Gaming Plus Max upgrade board, it has been under a Corsair 240 AIO must of its life.

 

[Beaflag VonRathburg]

It's more likely it was an issue with the mounting or contact between the heatspreader and cooler. There's absolutely no way to run that CPU at stock and get the temps he was unless something was seriously wrong with the mounting or cooling system. Hell, it sounds like the plastic on the waterblock to protect it during shipping wasn't removed. My absolute shit 5800x with a cheap Enermax 240 AIO runs cooler and it eats up a hell of a lot more power with a smaller area for the die to transfer heat.

I have no doubt a newer CPU would run cooler than an old one due to better yields in the silicon but nothing like what the OP is seeing.

OP also said he RMA'd the 2019 CPU because it was bad. CPU are incredibly resilient so It's certainly easy to suggest something was wrong with it when it ultimately died. What you suggest makes the most sense, but it also could've had a bad internal solder job causing issues with heat transfer.

 

[riev90]

I have no doubt a newer CPU would run cooler than an old one due to better yields in the silicon but nothing like what the OP is seeing.

Let me show you 5 samples of ryzen 3600 with higher max freq with lower vcore requirement:

This is my 1st ryzen 3600 with newer batch Y2020.
Notice it can run HWBot X265 @4475mhz with vcore 1.4v (llc 3) and max heat was 76.9c (I only used Deepcool Gammaxx 400 push - pull fan config)
_________________________________________

This one is 2nd ryzen 3600 I ordered, for my friend's build.
Notice that it run @4300mhz with vcore 1.25v (llc 3), and using wraith stealth (AMD's stock cooler).
_______________________________________

This one is my 3rd ryzen 3600 order, it can do 4400mhz with vcore 1.32v (llc 3), temp max 75.8c using Deepcool Gammaxx 400 push fan only. Notice this one is better than the first two sample (run cooler and more quite, lower vcore).
____________________________________________

This one is my best and ridiculous ryzen 3600, it can do 4450mhz using only 1.225v, has fclk 1900 and has no trouble running trfc 252 with gdm disabled, run a lot cooler at only max 68.8c, require only Deepcool Gammaxx 400 push fan.
______________________________________________

This is my last and worst ryzen 3600 before going to ryzen 5600 in september 2021.
It can do 4250mhz with vcore 1.22v, but still it's max temp is only 63.3c while using stock cooler.
Considering turbo clock (1~2 max core) for ryzen 3600 is 4200mhz requiring 1.4v at default, it's still a lot better and in conjunction with OP's ryzen 3600.

 

[SmokeRngs]

Let me show you 5 samples of ryzen 3600 with higher max freq with lower vcore requirement:

View attachment 539355
This is my 1st ryzen 3600 with newer batch Y2020.
Notice it can run HWBot X265 @4475mhz with vcore 1.4v (llc 3) and max heat was 76.9c (I only used Deepcool Gammaxx 400 push - pull fan config)
_________________________________________

View attachment 539356

This one is 2nd ryzen 3600 I ordered, for my friend's build.
Notice that it run @4300mhz with vcore 1.25v (llc 3), and using wraith stealth (AMD's stock cooler).
_______________________________________

View attachment 539357

This one is my 3rd ryzen 3600 order, it can do 4400mhz with vcore 1.32v (llc 3), temp max 75.8c using Deepcool Gammaxx 400 push fan only. Notice this one is better than the first two sample (run cooler and more quite, lower vcore).
____________________________________________

View attachment 539362

This one is my best and ridiculous ryzen 3600, it can do 4450mhz using only 1.225v, has fclk 1900 and has no trouble running trfc 252 with gdm disabled, run a lot cooler at only max 68.8c, require only Deepcool Gammaxx 400 push fan.
______________________________________________

View attachment 539363

This is my last and worst ryzen 3600 before going to ryzen 5600 in september 2021.
It can do 4250mhz with vcore 1.22v, but still it's max temp is only 63.3c while using stock cooler.
Considering turbo clock (1~2 max core) for ryzen 3600 is 4200mhz requiring 1.4v at default, it's still a lot better and in conjunction with OP's ryzen 3600.

I was referring to the 80C temps the OP was seeing with his 3600 and a 240 AIO. He was claiming the later 3600s were much better regarding thermals and while I have no doubt there's some truth to do that because of process maturation; the difference was way too large for that. While it's possible there was an issue with the soldered heatspreader like another poster mentioned, there had to be something seriously wrong for the OP to see the temps he was getting. A normal 3600 pushed to the limit shouldn't have been anywhere near the temps the OP was seeing no matter when it was made.

 

[riev90]

I was referring to the 80C temps the OP was seeing with his 3600 and a 240 AIO. He was claiming the later 3600s were much better regarding thermals and while I have no doubt there's some truth to do that because of process maturation; the difference was way too large for that. While it's possible there was an issue with the soldered heatspreader like another poster mentioned, there had to be something seriously wrong for the OP to see the temps he was getting. A normal 3600 pushed to the limit shouldn't have been anywhere near the temps the OP was seeing no matter when it was made.

Early ryzen 3600 (old stepping, high VID) sample shown up to 1.5v on short burst / single core and ~1.4v for multicore, so it's understandable that the newer stepping with lower VID will have lower default vcore with temp difference up to 15c.

The same thing happened with ryzen 5000. One of my friend's ryzen 5700g (8C 16T) now can run 4400mhz all core and needed only 1.175v. That's staggering.

 

[aLcATRAZ]

Sure the temp difference isn't just from the newly applied thermal paste?

Hehe, that thread title cracks me up.

 

[DWD1961]

I was referring to the 80C temps the OP was seeing with his 3600 and a 240 AIO. He was claiming the later 3600s were much better regarding thermals and while I have no doubt there's some truth to do that because of process maturation; the difference was way too large for that. While it's possible there was an issue with the soldered heatspreader like another poster mentioned, there had to be something seriously wrong for the OP to see the temps he was getting. A normal 3600 pushed to the limit shouldn't have been anywhere near the temps the OP was seeing no matter when it was made.

Yet if you research those temps from back in 2019-2020, you'll see that many people were seeing the same thing, and even AMD got involved on a Reddit thread and talked about high idle temps with the 3600"

https://community.amd.com/t5/proces...-and-stress-temperatures-too-high/td-p/320911

https://linustechtips.com/topic/136...temps-50c-when-idle-and-up-to-80-when-gaming/

This is where AMD gets involved and explains the high idle temps for the 3600:
https://www.reddit.com/r/Amd/comments/cbls9g/the_final_word_on_idle_voltages_for_3rd_gen_ryzen/

Interestingly enough, with this new CPU, my idle temps are low and cool vs my old 3600, with less cooling.

 

[DWD1961]

Sure the temp difference isn't just from the newly applied thermal paste?

Hehe, that thread title cracks me up.

No. I re-pasted the old CPU with all known possible paste procedures.