I really need to upgrade my personal media server but both companies' CPUs just use way too much power. It's kind of ridiculous. Yes, they're fast but power hungry as heck. Makes my dual xeons I run now look energy efficient.
There is some new inefficiency on AMD builds now that the Core Chiplet Die (CCD) and IO Die (IOD) are separate units, not monolithic. DDR5 also taked more power. The fancier X670 chipset (really, two chipset dies) also takes power & could very much use some tuning, but it's also doing a beastly amount of IO work, in some cases for good cause (more low power capabilities/better idle would be appreciated though! One of AMD's first in house chipsets in a long time).
We've barely started exploring how well behaved we can make the just launched 65w ryzen parts... but it wouldnt be shocking to me if there's not really any advantage, if they literally end up being nearly the same part, rebadged & preconfigured to lower power. Is there even really going to be any binning to have selected these? You know maybe not!
Out of the box most chips come tuned for speed, but ask them to be a little nixer, turn on eco mode or power the power point, they are stunningingly efficient. Alas idle could use some tuning; staying under 40w idle is harder than it ought to be.
Just underclock/undervolt, problem solved.
Did you by chance read an Intel i9 review? Yes, Intel's going crazy on the power front trying to match the AMD performance, which of course lead to AMD pushing it a bit as well. But the new non-X AMD CPUs are quite power efficient for the performance.
If that's still to much the M2 or M2 Pro in a mac mini is rather efficient on power and still performance quite well.
only if you disable boost. 88W PPT for AMD's "65W" CPUs at stock.
not a fan of the "PPT" terminology, it's pretty deliberately designed to mislead people and give AMD a mental "cheat factor" in these comparisons.
And no this is not something that "everyone did", in the 5960X/7700K days Intel's TDP values were expected to cover boost clocks and AVX as well, it's only 8700K/9900K/etc where Intel started to play fast and loose with their TDP figures. Which was after the introduction of Ryzen and the "PPT" terminology... in those days AMD pulled more than Intel's chips and they needed a way to hide it. Then they took the lead on efficiency.
kinda like TSMC naming their chips "7nm" and that resulting in huge mindshare problems for Intel even though 10ESF/Intel 7 (third-gen 10nm) was actually better than N7. Or Samung 8nm not even being as good as first-gen Intel 10nm, lol. So Intel had to rename their nodes too. A few bad actors ruin it for everyone.
88W is the number that matters for your load power measurement comparison - again, unless you disable boost.
The "X" chips are also clocked higher... 7950X would be 170W with boost disabled and runs 220W at stock.
So yeah, 100W or 65W TDP is not much at all.
I have a 27" Dell U2719DC from 2 years ago, pretty average monitor, 2560x1440, includes a USB hub, and has a USB-c ports for charging other widgets.
On a bright screen, like full screen tab of HN I get 22 watts. If I have a darker screen like anything full screen in dark/night mode (white text on black background) I get 20 watts. When I touch the screen or top edge it's pretty close to room temperature.
To be fair my work configuration uses two of the 27" monitors.
I'm measuring with a kill-a-watt, which has a decent reputation for accuracy.
https://everymac.com/monitors/apple/studio_cinema/specs/appl... the 30" Cinema Display from 2004 consumed 150W. The IBM "Big Bertha" T220 in 2001 at 20" consumed 111W. So that's the era when I'd expect a 27" to consume 80W or even more.
By 2010 some 27" displays like the Sceptre X270W consumed less than 45W. https://images.anandtech.com/graphs/apple27inchcinemadisplay...