AMD Ryzen 7000 Undervolting Yields Great Results with Temperatures

[btarunr]

AMD Ryzen 7000 "Zen 4" processors can hit up to 95 °C at stock settings, with cooling most appropriate to the TDP level. This is because the PPT (package power tracking) limits for the 170 W TDP processors is as high as 230 W, and for the 105 W TDP models, it's 130 W. After reaching this temperature threshold, the processor begins to downclock itself to lower temperatures. Harukaze5719 discovered that higher than needed core voltages could be at play, and manually undervolting the processors could free up significant thermal headroom, letting the processors hold on to higher boost multipliers better.



View at TechPowerUp Main Site | Source

 

[phanbuey]

That's a ton of voltage at stock for no reason.

 

[dyonoctis]

That’s a massive difference, wtf ?

 

[neatfeatguy]

Kind of reminds me of my Phenom II x4 940 - default voltage getting pushed to it was .5V higher than it needed to be to run stock settings. Under volting it helped drop the temps by almost 10C and the CPU ran just fine. Maybe the issue is just with the BIOS that's being used and things will prove to be better upon the official release?

 

[TheoneandonlyMrK]

Kind of reminds me of my Phenom II x4 940 - default voltage getting pushed to it was .5V higher than it needed to be to run stock settings. Under volting it helped drop the temps by almost 10C and the CPU ran just fine. Maybe the issue is just with the BIOS that's being used and things will prove to be better upon the official release?

Or perhaps it's the guy setting the shit up wrong.

How many times are we going to again find out that motherboard makers like to automatically oc when all settings are default or auto as is usual.

This guy's leaks are more like plops.

 

[JustBenching]

That’s a massive difference, wtf ?

Not really, every CPU in existence acts like that. High stock voltage is to cover every single case, some CPU's wont have the best silicon quality.

It is the same with alderlake, you could undervolt it to drop wattage by as much as 70w in CBR23.

 

[john_]

Kind of reminds me of my Phenom II x4 940 - default voltage getting pushed to it was .5V higher than it needed to be to run stock settings. Under volting it helped drop the temps by almost 10C and the CPU ran just fine. Maybe the issue is just with the BIOS that's being used and things will prove to be better upon the official release?

Yeah we had that with Phenoms. My 1055T was running at 3.5GHz from 2.8GHz default, with undervoltage. 700 extra MHz with lower voltage than the default.

AMD seems to keep pumping voltage to it's chips, to warranty that ALL of them will be stable, no matter the power consumption and temps. And while today this could be an option considering all the internal mechanisms modern CPUs have, looking at Phenom going over 70C while the specs where saying to stay under 65C, wasn't very nice.

Anyway, it's nice to see that 7000 series are probably not going to end up as little pieces of volcano lava.

 

[Asni]

Which workload? Looks meaningless to me.
You can't halve the power draw at the same MT boost clock, that'a single or light threaded workload.

 

[Deleted member 185088]

That's a ton of voltage at stock for no reason.

Maybe to make sure all CPUs can be stable.

 

[Denver]

The difference is insane... impossible not to think it's a bios issue.

 

[Chris34]

Holy. That's quite a difference. Probably because the voltages settings aren't fully optimized yet?

 

[defaultluser]

That's a ton of voltage at stock for no reason.

its there for one purpose only: to factory-overclock the chips to catch-up with Raptor Lake!

Oh, and also likely they will be favoring the low voltage chips for mobile?

 

[Wirko]

High stock voltage is to cover every single case, some CPU's wont have the best silicon quality.

Stock voltage used to be encoded in every single processor, is it different now? (I'm only certain about Core 2 Duo and its VID voltage).

 

[windwhirl]

Which workload? Looks meaningless to me.
You can't halve the power draw at the same MT boost clock, that'a single or light threaded workload.

AIDA64 FPU stress test. Which doesn't really tell me much, but that's what's in the screenshot

 

[ExcuseMeWtf]

I guess they will refine this in later steppings.

 

[zlobby]

That's a ton of voltage at stock for no reason.

I willing to bet AMD put it there for a reason.

On a side note, I can't wait to see what these beasts can do with heavy water-cooled setups or even LN!

Heck, I wasn't planning on Zen4 swap for any of my stuff but this thing just looks better and betrer with each leak! Any idea on the TR/EPYC/PRO launch dates?

 

[Kapone33]

I am hoping that these are more like the 3000 series than the 5000 series. Though only at most a 300 MHz boost on clocks but running it at 1.25 volts max was really nice.

 

[ARF]

That's a ton of voltage at stock for no reason.

Absolutely. Why doesn't AMD use smart AI self training in BIOS to train it to automatically adjust the voltage curve, while the user can simply set a target maximum temperature, for example 65-70°C?

 

[TheinsanegamerN]

Yeah we had that with Phenoms. My 1055T was running at 3.5GHz from 2.8GHz default, with undervoltage. 700 extra MHz with lower voltage than the default.

AMD seems to keep pumping voltage to it's chips, to warranty that ALL of them will be stable, no matter the power consumption and temps. And while today this could be an option considering all the internal mechanisms modern CPUs have, looking at Phenom going over 70C while the specs where saying to stay under 65C, wasn't very nice.

Anyway, it's nice to see that 7000 series are probably not going to end up as little pieces of volcano lava.

I remember them doing the same thing with Llano, which was great for laptop users because you could install k10stat, undervolt the chips, and turn a 4 hour battery runtime into a 7 hour runtime, or crank the clocks from 1.4 GHz to 2.7. Those were fun days.

Absolutely. Why doesn't AMD use smart AI self training in BIOS to train it to automatically adjust the voltage curve, while the user can simply set a target maximum temperature, for example 65-70°C?

Because that is unnecessary. AMD doesnt need an expensive complicated AI, they just need to stop overvolting the piss out of their CPUs for 0.1% perf gain. Same with nvidia and intel chips.

 

[phanbuey]

Or perhaps it's the guy setting the shit up wrong.

How many times are we going to again find out that motherboard makers like to automatically oc when all settings are default or auto as is usual.

This guy's leaks are more like plops.

I think this too... I get the going past diminishing power returns for % gain on all bins, but this is to a level that's extreme - it's basically a 100% increase in power draw overvolt from the manual settings.

Seems like wrong settings.

 

[user556]

A better cooler helps muchly when combined with undervolting. The lower the ultimate temperature ceiling is, the more undervolting will work. And in turn comes improved power efficiency as well.
So, that same cooler you buy to handle hot overclocking, works just as effectively for cool undervolting with heavy workloads.

 

[ymdhis]

They overvolted them on purpose so they can compete with Intel in the "highest wattage cpu" category.

 

[thegnome]

Good lord, that is one heavy overvolt at stock. Seems like Ryzen 7000 might actually be cooler than expected with that UV.

 

[TheDeeGee]

They overvolted them on purpose so they can compete with Intel in the "highest wattage cpu" category.

Atleast Intel CPUs don't require a BIOS update every week

 

[Bloax]

Is this news to people?
The 5800x clocks higher when you reduce its power limits and force it to use a more reasonable voltage/frequency.
Even more so if you then nudge the curve offset into slight negatives, but it's still a hilarious chip.