AMD Ryzen 9 7950X Boosts to 5.85 GHz Only if You Can Keep it Under 50°C

[1d10t]

My 5600 never goes above 62C, with air cooling. Configure that thing properly.

Nope I'm good, I'm just live in shithole flat thats all.

 

[vmarv]

Some scientist in the Arctic pole may have a chance. With an open case, sitting over an iceberg.

 

[Jism]

Some scientist in the Arctic pole may have a chance. With an open case, sitting over an iceberg.

Nah.

Ive studied the 2700x boost behaviour. It would lower clocks once you pass 60 degrees (mind you these have a +10 degree difference programmed into it) and the higher the temp the lower the overall clocks.

So when i ramped up a 360mm with 6 fans into push pull, it would sit day and night at 58 degrees and even peaking at 188W power consumption.

I'm sure there's more to it you can do with a AIO such as a washer mod or even Liquid metal. Set the fan profile on aggresive. At least i'm satisfied and due to the split ccd's it should be easier to cool it rather then one big die.

 

[Richards]

Chiplets are bad

 

[JAB Creations]

AIO isnt going to do it either, if these things peak temp like the 5000 series nothing short of LN2 or similar exotic options are gotta cut the mustard.

Sweet! Show us pictures of you with these not-released processors and what their temperatures are!

 

[Bwaze]

Weren't there reports that Zen 4 is even hotter and more difficult to cool than Zen 3? Reports of processors at 95 degrees C while performing simple benchmarks at stock, not some Prime95 torture?

 

[AusWolf]

Weren't there reports that Zen 4 is even hotter and more difficult to cool than Zen 3? Reports of processors at 95 degrees C while performing simple benchmarks at stock, not some Prime95 torture?

Yes. If Zen 3 taught me one thing, it's that the same power concentrated into a smaller die area means your CPU is more difficult to cool. Zen 4 chiplets will be even smaller than Zen 3, so...

 

[Jimmy_]

ummm..interesting!

 

[Solid State Brain]

For comparison, by default on the Intel i9-12900KS the Thermal Velocity Boost temperature for reaching 5.5 GHz is also 50 °C.

Product Specifications

quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.

ark.intel.com

 

[QuietBob]

In all probability, keeping Zen 4 under 50c is going to be a tall order - even with light workloads. It's the same with Zen 3 and most Zen 2 CPUs. My 5800X3D will hit 52c on occasion when simply browsing the net, in a 26c room. Mind you, this is the second lowest boosting Zen 3 SKU in the chiplet desktop lineup, undervolted, and cooled with a top-end dual tower HSF.

ST temperature spikes are instantenous and momentary, which doesn't allow the cooler fan to ramp up to compensate. Has been that way since Zen 2. Unless I keep the fan at a constant maximum, my CPU will be going over 50c in any realistic task.

Apparently, it could take a top performing 360mm AIO or CLC to extract maximum performance from Zen 4. And even that may not be enough when your ambient temperature is more than 20c.

 

[Mussels]

In other words, brief short terms boosts coming right from idle, or low threaded games

(With decent cooling already in place)

My 5600 never goes above 62C, with air cooling. Configure that thing properly.

And I bet we'll find out you did an all core overunderclock at 4.2GHz or something similar.

Don't make claims without facts to back them up.

 

[Chomiq]

Apparently, it could take a top performing 360mm AIO or CLC to extract maximum performance from Zen 4. And even that may not be enough when your ambient temperature is more than 20c.

280mm+ AIO was already a must for Zen 3 if one wanted to reach 5 GHz boost. I'm not surprised that same requirement will have to be met for top Zen 4 SKUs.

If anything, this would be even worse for something like 7800X.

 

[QuietBob]

280mm+ AIO was already a must for Zen 3 if one wanted to reach 5 GHz boost. I'm not surprised that same requirement will have to be met for top Zen 4 SKUs.

@freeagent

What I was referring to is the unrealistically low thermal requirement for peak ST boost on the 7950X (and likely the 7900X, which boosts only 100 MHz lower). Zen 3 would reach its maximum clock with much higher temps.

If anything, this would be even worse for something like 7800X.

Agreed. Cooling a fully populated single CCD on Zen 4 will most likely be even more of a challenge.

 

[freeagent]

280mm+ AIO was already a must for Zen 3 if one wanted to reach 5 GHz boost. I'm not surprised that same requirement will have to be met for top Zen 4 SKUs.

If anything, this would be even worse for something like 7800X.

Nah.. been doing it with an air cooler for many moons now

 

[ymdhis]

The sudden spikes in temperature on minimal workload is "normal" on Ryzen chips. When PBO is enabled, the chip automatically boosts itself to the absolute max voltage and clockspeed it can handle. So you go from something like ~0.7V average idle voltage to ~1.45V in an instant. Hence the sudden temperature boost. You can set up a lower CPU voltage and get vastly improved temperatures; or even disable PBO and set up something like a 1.3V with whatever your chips max clock is. Chances are it will be able to handle its peak clock on all cores if you find the right voltage, and it'll still have lower temps than the insane one that PBO uses.

AMD has been over volting all of its silicon for a long time now, I don't know why, probably they leave things like that as a "safe" margin that works on everything. Vega was legendary for how much extra performance it could get out of undervolting, and from what I've seen the Ryzens are like this too. It's not a bad thing since it can overclock itself to near its maximum with only thermals and power usage limiting it. But the voltage is set up too high and the chip will get very hot. My 3600 would hit 80C+ on its default settings with the base cooler; just setting the right voltages reduced that a lot but I forgot the exact numbers and I don't want to post "how I remember it" bullshit.

Can you please tell me what cooler you use and if you undervolt ot not?

Thanks!

Asrock B550M Steel Legend, CPU voltage is set to the highest negative OFFSET the board allows (-0.1V I think), SoC is set to a quarter of that (-0.025V I think). Setting the SoC too low actually got me random crashes, but the CPU has no issues. Cooler is a Hyper212 Black, with two Noctua A12 fans I managed to get used for a low price, they are set to running the absolute most silent profile (at 60C they hit 1180 RPM). At idle it varies at ~36-40C with the fans going at ~500 RPM only.

And I bet we'll find out you did an all core overunderclock at 4.2GHz or something similar.

Don't make claims without facts to back them up.

dunno what to say, that seems to be the normal all-core clock for this cpu. In single thread it goes up to 4550MHz. Screenshot attached.
I'm not about to start dumping 100+ watts and a custom water loop just to get an extra 100MHz or so. It runs very cool, uses minimal power, and it's fast enough to do everything I use it for (photoshop, emulation, occasional mac VM, movie/audio converting, gaming).

If I disable the turbo clocks (I think the one called Core Performance Boost), then it will cap itself down to 3.8GHz, but on all cores and doesn't even hit 50C. Very ideal for the summer. And it doesn't make any real-world performance change that I can notice during normal usage, only the masturbatory benchmark scores are lower. Maybe audio transcoding is 10% slower but that just means 6 seconds instead of 5 for a longer album.

 

[taka]

Ah you have the 5600G, different beast here 5600x. All BIOS on Auto except XMP and fans on manual. MSI B550A PRO.
Cooler ID Cooling SE-224XT one fan fixed at 800 RPM.
2 x 200mm on intake fixed at 350RPM
2 x 140mm exhaust fixed at 600RPM

 

[ymdhis]

Try setting a negative voltage offset, though from what I can see your chip already uses pretty low values. But it might help reduce the power drain, which may allow for slightly higher performance.

iirc the 5600g is just a 5600 with half the L3 cache and an IGP (which uses negligible power), and the 5600x is just a 5600 with +200 MHz. So the results are sort of comparable.

I do recall getting quite a bit higher temps on my cpu, until I re-arranged the fans so the second a12 was on the cpu heatsink instead of used as exhaust. It helped significantly to have an extra fan on the heatsink.

 

[freeagent]

The sudden spikes in temperature on minimal workload is "normal" on Ryzen chips.

I see people saying that, but on my 5600X and 5900X I see none of these spikes.

To me they are non existent.

In the end I would chalk it up to inadequate cooling.

 

[gruffi]

Not that it matters. Ryzen processors don't actually perform any task at boost frequency, no matter how light and how synthetic it is, and how cold they are.

Nonsense. If I run Cinebench R23 my 5600X boosts to 4.6 GHz single core and 4.3 GHz all core. I use a boxed cooler from first gen Ryzen. PBO is disabled. A Ryzen isn't really hard to cool. At least if you disable PBO. It will still boost to max frequency. I guess the 7950X won't need extraordinary cooling to run it at 5.7 GHz single core and ~5 GHz all core.

 

[mama]

Hope you have an AC and a HUGE rad, cause you'll need like 25°C max water temp to keep the CPU under 50°C... Water-to-Core delta is about 25°C under load even with a decent block.

Guess a MO-RA or two will do the trick quietly in a 20°C room...

I'm looking to install two 60mmx360mm rads plus a 30mmx360mm in a loop to cool both the CPU and GPU (a 6900XT). Dual pump. There is additional room in the HAF for additional rads. It's not a MO-RA I will grant you. No guarantees but I'm hoping the setup I have planned will suffice. And yes, there is air-conditioning.

 

[Jism]

Nonsense. If I run Cinebench R23 my 5600X boosts to 4.6 GHz single core and 4.3 GHz all core. I use a boxed cooler from first gen Ryzen. PBO is disabled. A Ryzen isn't really hard to cool. At least if you disable PBO. It will still boost to max frequency. I guess the 7950X won't need extraordinary cooling to run it at 5.7 GHz single core and ~5 GHz all core.

They do boost and actually hold their boost clocks, based on 3 things:

- Current or power consumption
- Actual Core Temperature
- Motherboard Power delivery design

If any of those seem to fall beyond it's spec the chip will downclock. Its just a OC'ing feature on a automated basis, which can switch up to 50 times a second.

And it's clever, its just constantly extracting more performance out of the box for you rather then having to digg into endless hours of bios OC'ing.

Downside is that there's hardly any manual OC'ing headroom really because they figured out a way to bin the chips even better then before.

If you want longer lasting boost clocks, increase cooling or learn how to insert a negative voltage offset. It's basicly providing the CPU a bit lower voltage which makes the temperature and power consumption go lower, and thus being able to boost longer.

 

[DeathtoGnomes]

Its TDP is rated at 170 W, and its PPT (package power tracking) limit at 230 W.

My 9370 is 220w and the old cooler (280mm) kept it below 50C on full load. Upgrading to a 420 keeps temp below 45C.

@maxfly This would be good for that case and the configuration we were discussing.

 

[YOUDIEMOFO]

AIO isnt going to do it either, if these things peak temp like the 5000 series nothing short of LN2 or similar exotic options are gotta cut the mustard.

A proper custom liquid loop and proper settings in bios will allow for some damn good cooling over an AIO. LN2 needed for this kind of power?!?! Not in a lifetime.

Pushing 250+ watts out of my 5950X and don't see over 75c on my custom loop. And my 3960X crushing along at 350+ watts doesn't see anything near 80c.

If you're cooking with a potato then I guess you get what you put in.

Not anywhere did I read aub zero temperatures in order to attain said frequencies.

At least this company is giving explanations as to why and how things work. Team blue still lets you just go ham with power and go to the moon if that's what you truly want.

I'm looking to install two 60mmx360mm rads plus a 30mmx360mm in a loop to cool both the CPU and GPU (a 6900XT). Dual pump. There is additional room in the HAF for additional rads. It's not a MO-RA I will grant you. No guarantees but I'm hoping the setup I have planned will suffice. And yes, there is air-conditioning.

Have four alohacool nexxxos UT60 480mm (quad 140) and two UT60 360mm (triple 120) radiators in my caselabs STH-10 with pedestal. Triple lian-li DDC overvolted pumps in series.

Keeps 700 watt GPUs at or below 45c and my threadripper that can chug 350+ watts below 80c.

 

[Jism]

My 9370 is 220w and the old cooler (280mm) kept it below 50C on full load. Upgrading to a 420 keeps temp below 45C.

@maxfly This would be good for that case and the configuration we were discussing.

It is a 32nm CPU vs a 7nm or even 6/5nm. The biggest problem is the heat density. A smaller surface that has to be cooled is way more difficult then a 32nm CPU.

 

[DeathtoGnomes]

It is a 32nm CPU vs a 7nm or even 6/5nm. The biggest problem is the heat density. A smaller surface that has to be cooled is way more difficult then a 32nm CPU.

IDK about that, but I expect long periods of run time would reduce that problem.