AMD Ryzen 9 7950X Cooling Requirements & Thermal Throttling

[R0H1T]

I'd argue thats an efficiency preset rather than an optimal one. But yeah, if anything AMD's communication definitely sucked on this.

You can always set your own PPT, TDC, EDC limits & most AMD boards will strictly enforce them. Intel users who claim that their locked chips are more efficient conveniently forget they're comparing locked (Intel) vs unlocked (AMD) chips. When restricted by ECO mode or some of these limits & undervolted, AMD chips can be anywhere between 10~80% more efficient (also depending on the workload) heck look at 7950x results! Not pointing fingers at you but poking holes in their theory that Intel's locked chips are more efficient

 

[Camm]

You can always set your own PPT, TDC, EDC limits & most AMD boards will strictly enforce them. Intel users who claim that their locked chips are more efficient conveniently forget they're comparing locked (Intel) vs unlocked (AMD) chips. When restricted by ECO mode or some of these limits & undervolted, AMD chips can be anywhere between 10~80% more efficient (also depending on the workload) heck look at 7950x results! Not pointing fingers at you but poking holes in their theory that Intel's locked chips are more efficient

Of course. Perhaps poorly articulated by myself, but as with most chips there is a point where temperature/power/heat increases exponentially for further gains, it'd be nice if we could simply set a chip to operate on the optimal curve rather than us having to do it ourselves, and I really dislike the trend from vendors to go smacking up against that wall all the time with users having no simple options to dial them back.

If I was AMD personally, I'd have 3 modes, Xtreme, Optimal & Eco, with X class chips coming by default as the Extreme preset, and non X as Optimal, but it being a simple BIOS toggle to go between them.

As for Intel, it is disgusting how much extra power you use for those last few mhz. It'd be nice to see the same from them personally.

 

[W1zzard]

Ugh, that doesn't sound good. So the ability to throttle has been taken away completely?

Correct, this is intentional, because when dual manual OC you're setting a fixed clock and don't want the chip to mess with your clocks.

W1zzard, i'm curious - is there a dedicated page where you announce your up-n-coming reviews?

Before publishing? No.

When published, we have an RSS feed: https://www.techpowerup.com/rss/
or Twitter / Facebook / Linkedin

me too, i would love to see power usage in relations to which cooler used. Does lower frequency means lower power?

Yes it means lower power, it must. Good question indeed, let me see what I can do

direct die the measured power consumption dropped. If the software measurement is accurate to reality that means the IHS thickness is not only causing higher temps than needed, but also increasing power consumption.

Silicon that runs at lower temperature is more efficient, too, maybe that's what happened. I doubt the power sensors have that much thermal drift, still, that's part of the reason I do proper power measurements with test equipment and don't trust the in-chip power sensors.

Platform Thermal Throttle Ctrl

Just to clarify, that settings only lets you lower the throttle temp, not increase it. Default is 95°C, so you can go 90, 85 .., but not 100, 105 ..

@W1zzard What is the ambient temp for your testing lab?

Temps are based on delta T and normalized to 25°C.

If you sit outside in the cold, your CPU will run cooler at first, and then boost higher, back up to 95°C or as high as possible, given the clock limits, either way you will get more performance. If that's what you are asking

 

[R0H1T]

Of course. Perhaps poorly articulated by myself, but as with most chips there is a point where temperature/power/heat increases exponentially for further gains, it'd be nice if we could simply set a chip to operate on the optimal curve rather than us having to do it ourselves, and I really dislike the trend from vendors to go smacking up against that wall all the time with users having no simple options to dial them back.

If I was AMD personally, I'd have 3 modes, Xtreme, Optimal & Eco, with X class chips coming by default as the Extreme preset, and non X as Optimal, but it being a simple BIOS toggle to go between them.

As for Intel, it is disgusting how much extra power you use for those last few mhz. It'd be nice to see the same from them personally.

Yeah heat is a big issue at my place & no way I'm running anything around 95C, I'd rather lose 10~15% performance & be able to run these 15-20C lower but that's just me!

 

[bug]

Wrong. And and this article proves it.

TechPowerup:

“Our testing in this article shows that the performance losses are minimal, even when pairing the Ryzen 9 7950X flagship with an entry-level cooler that's running at slow fan speed settings. You can simply get started with whatever AM4 heatsink you have lying around and you'll be safe and almost get maximum performance. AMD was wise to make their new Socket AM5 compatible with AM4 coolers, so you have plenty of choice, also on the used market.

The biggest problem is probably psychological. For years we have been trained that "95°C is bad". This is no longer true. 95°C is the new 65°C.”

TechPowerup is basically describing you.

And with the ability to run these at reduced watts - it’s minimal performance drop is astounding.

Temperature drops are fantastic. Nobody who owns these worries about temperature or cooling. Nobody.

Watch that video for proof. Very clear and easy to understand.

Except my beef isn't with the performance. It's with the power dissipated as heat. This is only a tiny bit less demanding than the 12900k, which was labeled right here on TPU as "Very high heat output / power usage".

Plus, the benchmarking here doesn't feel right. There are 4 productivity apps pitted against 4 games (but games are tested several times). So the most you can say is, if you're a gamer, you're good. Which you already knew, because high-res gaming isn't CPU limited anyway. If you buy this for productivity (and work in an office with these installed in several workstations), I'm pretty sure you won't want to run on air.

NB This isn't a critique of the review, it's just that, as it stands, whether you can do with air cooling or not seems to be dependent on your workload.

 

[Taraquin]

Wow! Just wow! Awesome testing!

 

[thewan]

Got my 7700X last week and have been setting up and testing for past few days. I just put my previous ID Cooling SE 224 XT on it, didn't want to buy a new cooler yet. Seeing it jump to 95C every now and then, and the fans ramping up noisily was unsettling at first, but hey I'm slowly getting used to it. I am now running ECO mode because I currently can't use its power to its full potential until I get a new GPU.

I will probably get a new quieter air cooler in the near future. The stock fans on this cooler is way too noisy at higher RPMs. If I'm gonna need to replace the fans, I might as well get a better tower so that the fans don't need to ramp up so much as well under normal loads.

Why not water/AIO? I used to run an 240mm AIO before switching to the SE 224 XT. Right after a year, which is the standard warranty for most things outside of EU, it started leaking. I got back from an outing one day, switched on my PC before heading to the bath so that its fully ready for me to use it. To my horror, after I came out of the bath, my PC casing was emitting smoke from the top exhaust fan and the room smelled of burnt plastic and electronics. I've got phobia towards using an AIO ever since. No please do not try to convince me that todays AIO are safer and better quality. It's a mental thing.

 

[R0H1T]

I don't know about phobias but I don't trust anything with water in it around electronics, that especially includes sugary stuff like Coke & Pepsi or tea

 

[AusWolf]

Got my 7700X last week and have been setting up and testing for past few days. I just put my previous ID Cooling SE 224 XT on it, didn't want to buy a new cooler yet. Seeing it jump to 95C every now and then, and the fans ramping up noisily was unsettling at first, but hey I'm slowly getting used to it. I am now running ECO mode because I currently can't use its power to its full potential until I get a new GPU.

I will probably get a new quieter air cooler in the near future. The stock fans on this cooler is way too noisy at higher RPMs. If I'm gonna need to replace the fans, I might as well get a better tower so that the fans don't need to ramp up so much as well under normal loads.

Why not water/AIO? I used to run an 240mm AIO before switching to the SE 224 XT. Right after a year, which is the standard warranty for most things outside of EU, it started leaking. I got back from an outing one day, switched on my PC before heading to the bath so that its fully ready for me to use it. To my horror, after I came out of the bath, my PC casing was emitting smoke from the top exhaust fan and the room smelled of burnt plastic and electronics. I've got phobia towards using an AIO ever since. No please do not try to convince me that todays AIO are safer and better quality. It's a mental thing.

Hm... maybe a custom water loop then? Only joking.

Anyway, how do you find the Eco mode? How did it improve the temperature? What about clock speeds?

Sorry for the thousand questions. So far, I haven't seen as much media coverage of Eco mode as I'd like to.

 

[Haile Selassie]

The frenzy around the 95°C load temperature of the processor had gotten so bad that some predicted that cheap air coolers could "start fires."

Well, this is what you get when you skip basic physics course in high school.

 

[Dirt Chip]

Very good and informative article but I'm missing a key element: noise.
Not a single word about it.

How do you set fan curve with reasonable noise output when you get 95 degree all time?

As someone who is sensitive to noise, I want to tweek and fine-tune noise to preformance. I can't see how it can be done on zen4 cpu's.
In this regard, zen4 is a step back from past platforms and might be the deciding factor when buying my new system.

This is a "design choice" that turn into desing-flaw, if no matter the workload it causing maximum noise and no, setting 40-60% max noise at 95 degree is out of the question as part of what I do is long rendering. I know that on those occasions I will need to cope wite max noise, but at other cases it`s a no-go.

I suggest doing an iso-noise test or noise to preformance curve to see what is the sacrifice that auto 95 degree carry, because according to the article all is sunshine and flowers while disregarding a very important factor.

 

[R0H1T]

It's not, really they're just allowing the chips to boost higher & going nearly crazy hot in order to match Intel's speeds. Now we can debate ad nauseum who's at fault for this arms GHz race but the fact is these chips can run cooler & a lot more efficiently, you'll just need to tune them accordingly!

 

[TheDeeGee]

This would be 100% a no go as silent enthausiast.

There is a reason i went with a overkill NH-D15 on my Non-K 11700, it means it only has to spin 600 RPM to keep it 45-55C during gaming.

 

[einard]

The coolers arent the issues theese days, it's the damn heatspreader.

Now that LGA is the standard. Why don't they start removing the heatspreader. The bracket holding the cpu down can be used to create a barrier from putting too much preasure from the die.
The LGA socket itself will provide the preasuret to the heatsink.

We had direct die cooling for year in the past, and that was with mounting and supporting big heavy copperblocks directly on a small die. With LGA it would be a lot safer

 

[AusWolf]

This would be 100% a no go as silent enthausiast.

There is a reason i went with a overkill NH-D15 on my Non-K 11700, it means it only has to spin 600 RPM to keep it 45-55C during gaming.

You're not alone with that. I also have a non-K 11700 that I'm water cooling with a 280 mm AIO. 40-50 °C in gaming is nice, but so is 75 °C in Cinebench while maintaining 4.4 GHz across all cores.

If I go with AMD in the future (I might because I'm curious), I'll definitely give Eco mode a try.

 

[W1zzard]

me too, i would love to see power usage in relations to which cooler used. Does lower frequency means lower power?

Power charts have been added, great suggestion, thanks!

Plus, the benchmarking here doesn't feel right. There are 4 productivity apps pitted against 4 games (but games are tested several times).

I picked a few applications that cover the whole spectrum of workloads, this part was easy.

Picking games was harder. I wanted something light that represents MOBA games: CSGO.. I also wanted something that scales across cores: Cyberpunk .. then I wanted Far Cry 6 because it scales very well with CPU perf, and then I felt like I could add one more. Then I realized that things will depends on the resolution, so I tested all of them.

Is your underlying question: what are the app-only averages?

http://imgur.com/uPtqDdG

http://imgur.com/07ivP7b

If you buy this for productivity (and work in an office with these installed in several workstations), I'm pretty sure you won't want to run on air.

Why? Maybe you actually definitely want to run them on air and sacrifice a few percent in perf, but you won't have to worry about maintainability? Which costs a lot of $$ in downtime and support man hours?

 

[AusWolf]

It's not, really they're just allowing the chips to boost higher & going nearly crazy hot in order to match Intel's speeds. Now we can debate ad nauseum who's at fault for this arms GHz race but the fact is these chips can run cooler & a lot more efficiently, you'll just need to tune them accordingly!

A little preview for TLDR (TLDW?):
It looks like Eco mode basically lowers your TDP class from 105 W to 65 W, your temperature by a lot, and your performance by 5-10%.

Now I'm sure that if I end up buying a Zen 4 CPU at some point (the B650 motherboard lineup looks pretty grim at the moment, but we'll see), I'll definitely run it in Eco mode.

 

[R0H1T]

One more thing to note is that if you're going with something like a 6 core, or even 8 core zen4 chip, with a poor cooler like the Wraith's then you should lower the "TDP" to even below 65W for better thermals. Also with CO & default power limits the 7600x will run like a furnace

Basically whatever you get don't run them with sh!t cooling!

 

[bobmeix]

Very informative read!

Some typos that I have noticed:
- the NH-U14S is able to keep temperatures well above the 5 GHz-mark / probably should be frequencies
- and when when all 32 threads are maxed out
- to stay as close possible to 95°C / to stay as close as possible to 95°C
- gains from going water are are relatively small
- Overall this great news for Zen 4 buyers / Overall this is great news for Zen 4 buyers

 

[TheoneandonlyMrK]

Very good and informative article but I'm missing a key element: noise.
Not a single word about it.

How do you set fan curve with reasonable noise output when you get 95 degree all time?

As someone who is sensitive to noise, I want to tweek and fine-tune noise to preformance. I can't see how it can be done on zen4 cpu's.
In this regard, zen4 is a step back from past platforms and might be the deciding factor when buying my new system.

This is a "design choice" that turn into desing-flaw, if no matter the workload it causing maximum noise and no, setting 40-60% max noise at 95 degree is out of the question as part of what I do is long rendering. I know that on those occasions I will need to cope wite max noise, but at other cases it`s a no-go.

I suggest doing an iso-noise test or noise to preformance curve to see what is the sacrifice that auto 95 degree carry, because according to the article all is sunshine and flowers while disregarding a very important factor.

Then re read it.

If a wraith spire at 20% is fine, and it was.

And that's too noisy for you get a MacBook air simple.


But regardless get used to hot chip's, as above hints so AL confirmed, hot chip's are now the normal, unless you do nothing with it.

 

[tony359]

Excellent article with comprehensive and clear testing.
I’m not totally sure about the ‘95 is the new 65’ though. If the manufacturer says it’s fine it doesn’t mean it’s a good thing. AMD can’t possibly say ‘yeah it’s way too high but we had no other choice’
That’s my personal opinion of course.

 

[dj-electric]

This would be 100% a no go as silent enthausiast.

There is a reason i went with a overkill NH-D15 on my Non-K 11700, it means it only has to spin 600 RPM to keep it 45-55C during gaming.

You can run an advanced Ryzen CPU with an NH-D15 on low fan RPM, as shown, and still enjoy its performance with very high clocks

 

[RainingTacco]

No performance comparison between 3200mhz and 5700mhz?

 

[W1zzard]

Very informative read!

Some typos that I have noticed:
- the NH-U14S is able to keep temperatures well above the 5 GHz-mark / probably should be frequencies
- and when when all 32 threads are maxed out
- to stay as close possible to 95°C / to stay as close as possible to 95°C
- gains from going water are are relatively small
- Overall this great news for Zen 4 buyers / Overall this is great news for Zen 4 buyers

Whoops .. big fail .. I've read it several times and missed those .. also other staffers .. thank you!

No performance comparison between 3200mhz and 5700mhz?

Wraith Spire at 20% was slightly unstable in some tests, so I didn't include it, but still wanted to get you the frequencies

 

[bobmeix]

Whoops .. big fail .. I've read it several times and missed those .. also other staffers .. thank you!

It's not a "big fail", as the content quality makes up for the typos.
You're welcome!