AMD Ryzen 7000 Undervolting Yields Great Results with Temperatures

[R0H1T]

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 you can't train that AI to set such voltage using literally billions of applications & gazillion of workloads/combinations with multiple apps running in the background & other hardware configurations/cooling to also take into account. That's more than the number of stars in all our universe!

Oh, did I Mention OS & its impact?

 

[Vader]

This is useless without benchmarks. Remember Ryzen 3000 series and clock stretching? Because this smells like that to me. . There is always some headroom to undervolt, but to do it to this extent without losing performance would be extremely rare

 

[zlobby]

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

Are you reading the CVE lists for those?

 

[Cippo95]

This is useless without benchmarks. Remember Ryzen 3000 series and clock stretching? Because this smells like that to me. . There is always some headroom to undervolt, but to do it to this extent without losing performance would be extremely rare

Exactly!
Very strange that other "hardware enthusiasts" commenting here don't recall clock stretching tbh.

 

[R0H1T]

I don't because I never owned one, what was it about btw

 

[Operandi]

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.

This.

AMD has to set a value to make all CPUs stable under wide variety of conditions under all workloads. By all means tweak away, cause its cool and fun but randoms on Twitter and forums didn't discover something that AMD engineers didn't already know.

 

[R0H1T]

Why is that bad thing, considering most of them improve performance/compatibility or fixes previous bugs? Did you also forget AM4 socket is now 5 years old? And now tell me which Intel MB from 5 years back supports 12th gen

Think I misunderstood what you're saying

 

[AM4isGOD]

Can we do nuclear power station and lava jokes about AM5 now? Or will it be frowned upon by the AMD crew that did it RE Intel ADL /s for @thegnome

 

[thegnome]

Can we do nuclear power station and lava jokes about AM5 now? Or will it be frowned upon by the AMD crew that did it RE Intel ADL

Well not unless AM5 uses more power than Raptor/Alder Lake. Can't exactly make those jokes when the chips run lower power, can you?

 

[dir_d]

Not a lot of details here, i will wait for some reviews. It might just be this workload that the voltage can be lowered for.

 

[AM4isGOD]

Well not unless AM5 uses more power than Raptor/Alder Lake. Can't exactly make those jokes when the chips run lower power, can you?

See post no. 2, or did you comment without reading any posts? Maybe i will edit my post with a /s for sarcasm just for you.

 

[TheoneandonlyMrK]

Can we do nuclear power station and lava jokes about AM5 now? Or will it be frowned upon by the AMD crew that did it RE Intel ADL

So long as you wake up to the reality that raptor lake's pushing Intel's power limit up too, and we know what the world thought of alderlakes power use.

Or await verified proof perhaps that could work!.

And anyway, like system reporting software isn't ever wrong, they likely need to update hwinfo64 ASAP.

 

[spnidel]

undervolting a PC component yields lower temperature and potentially boosts clockspeeds?
woah... I had no idea...

 

[zlobby]

Exactly!
Very strange that other "hardware enthusiasts" commenting here don't recall clock stretching tbh.

I though you were talking about c(l)ock stretching! There are lots of informative videos on the subject.

 

[mahirzukic2]

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

I was thinking the same thing. Why not CPU stress test? It's right there above the FPU stress test.

 

[R0H1T]

Maximum heat, just the CPU test won't produce that.

 

[Aretak]

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

The information presented here is based on an engineering sample using an unfinished BIOS, as the Twitter user in question notes.

https://twitter.com/x/status/1565389117319708672

I feel like that should really be mentioned in the article. The final silicon may have a completely different voltage curve.

 

[ARF]

Because you can't train that AI to set such voltage using literally billions of applications & gazillion of workloads/combinations with multiple apps running in the background & other hardware configurations/cooling to also take into account. That's more than the number of stars in all our universe!

Oh, did I Mention OS & its impact?

You are speaking absolutely nonsense. There are not so many popular apps and their load on the CPU is not different in order to claim different voltage level for each of those applications.

 

[zlobby]

The information presented here is based on an engineering sample using an unfinished BIOS, as the Twitter user in question notes.

https://twitter.com/x/status/1565389117319708672

I feel like that should really be mentioned in the article. The final silicon may have a completely different voltage curve.

Yes! I too was baited by this. And reading diagonally, too. But I'm willing to blame it on sloppy journalism; it's easier.

 

[TheoneandonlyMrK]

You are speaking absolutely nonsense. There are not so many popular apps and their load on the CPU is not different in order to claim different voltage level for each of those applications.

If we look back through recent history, and of course fact, it's clear your wrong.

Of course different applications draw different levels of power and stress CPU in different ways.

And if you USED your pc for more than occasional gaming and 24hr trolling you would have seen it.

Would community grid, folding at home, at the same time, WORK, etc, does not equal light browsing and a bit of gaming.

 

[R0H1T]

You are speaking absolutely nonsense. There are not so many popular apps and their load on the CPU is not different in order to claim different voltage level for each of those applications.

Who said different voltage levels for each app? Are you thick or what or do you not know how varying workloads, with single or multiple applications running can affect every system differently? I'm not even counting the silicon lottery, either you're very new to this or you think we have this AI a billion times better than our current capabilities

 

[ymdhis]

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

I'd rather apply a new BIOS update for a bugfix than to have to delid the CPU and apply liquid metal on the die so it doesn't run at 80-90C.

Also on the last Intel chip I had, the USB3 drivers were full out of whack, sometimes they completely stopped working (all of the ports!) and required a hard reboot. That was back on a Z77 board though, and eventually I realized that it was the Intel drivers fault.

 

[AM4isGOD]

I'd rather apply a new BIOS update for a bugfix than to have to delid the CPU and apply liquid metal on the die so it doesn't run at 80-90C.

Well that is crap too, as you don't have to, i have a 12700k and did not have to delid it at all. Knee jerk response to a comment you did not like i guess.

Also, because you had a problem with Intel USB 3 drivers, does not mean all Intel boards or drivers are crap either.

 

[Bloax]

True bliss on AM4 would be retiring to a Cezanne processor (Ryzen 5500 / 5300/5600/5700G) on some cheapo 2-DIMM board while clocking your RAM to the moon, and not having to deal with all the wonders of having to find three voltages (SOC/IOD/CCD) within +/- 0.01v that function properly without shitting the bed with fabric error corrections (microstuttering):

as for the topic of modern furnace CPUs:

Alder Lake runs surprisingly cool when you disable hyperthreading
probably going to be the same for Zen4, we'll see soon enough


Now, if AM5 was to be better than AM4 - then it would let you set SOC, IOD and CCD voltage while the CPU is still running.
As currently, the main thing that makes LGA1700 "better" than AM4 besides mild performance improvements - is the fact that to get that performance, you don't need to watch paint dry for several hours.
Several hours of boredom on top of already having to do memeory overclocking - but that part is the same for both AM4 and LGA1700, hee hee.

 

[Mussels]

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

Ryzen dont either.
The BIOS updates always added new CPU support, and occasionally a bugfix.
Intel just require you to buy a new motherboard, free BIOS update included.

Can we do nuclear power station and lava jokes about AM5 now? Or will it be frowned upon by the AMD crew that did it RE Intel ADL /s for @thegnome

For using 125W?
I'm not sure you thought that one through.