Zen 4 vs Zen 3: AMD Ryzen 9 7950X manages crucial performance efficiency gains over Ryzen 9 5950X despite huge TDP rise

[Redaktion]

The AMD Ryzen 9 7950X has been spotted on the SiSoftware Sandra database where it has produced very encouraging benchmark results. Some comparisons with the Zen 3-based Ryzen 9 5950X have shown the Zen 4 chip to be up to +118% quicker; however, in terms of performance efficiency the Ryzen 9 7950X still offers an edge, albeit a slight one.

https://www.notebookcheck.net/Zen-4-vs-Zen-3-AMD-Ryzen-9-7950X-manages-crucial-performance-efficiency-gains-over-Ryzen-9-5950X-despite-huge-TDP-rise.652913.0.html

[Anonymousgg]

For lovers of percentage differences, this places the AMD Ryzen 9 7950X an acceptable +9.82% ahead of its predecessor.

I, too, love percentages.

[NikoB]

I predict that everything will end up with governments forcibly limiting the TDP race. Energy efficiency is trending. Apparently, electricity prices for crowds of miners and gamers are not yet high enough in Europe and especially in the USA. It's time to raise the price per kWh even higher... ;)

[NikoB]

And I'm dreaming of a silent laptop and PC with the performance of R9/i9 and 4090Ti...

[_MT_]

I predict that everything will end up with governments forcibly limiting the TDP race.

AMD's TDP is largely meaningless. And a big factor here is always a motherboard. There is a natural limit as high TDPs require more exotic cooling and there is limited apetite for that. Datacentres are a different story and we could see a migration towards ever higher densities and liquid cooling. In practice, power supply can be more limiting than cooling.

[Rob Stan]

TDP, PPT (or even Intel's bizarre PL1, PL2 and PL1=PL2 shenanigans) DON'T = actual power usage.

You need actual power measurements during a task (preferably bursty AND sustained) to make any kind of worthwhile claim about "efficiency" about a setup/chip.
Making such claims about chips that aren't even released yet is even more ridiculous.

[Slane]

It's known that it's only the PPT (max 142 Watts for 5800X, 5900x and 5950X) that represent the REAL power usage during benchmarks or heavy loads. What his PPT on the on the 7950X during the bench?

[_MT_]

It's known that it's only the PPT (max 142 Watts for 5800X, 5900x and 5950X) that represent the REAL power usage during benchmarks or heavy loads. What his PPT on the on the 7950X during the bench?

That's before you consider motherboards that try to trick CPUs into drawing more than they should. It can give them advantage in motherboard benchmarks. And it complicates CPU benchmarks as you really have to know what you're doing to ensure that the motherboard behaves. Dilettantes need not apply.

PPT is 135 % of TDP in AMD's world. AM5 is reported to support 230 W PPT (= 170 W TDP).

[_MT_]

or even Intel's bizarre PL1, PL2 and PL1=PL2 shenanigans

Why bizarre? PL1 is sustainable limit, PL2 is unsustainable limit, tau is how long it can be sustained and PL1 = PL2 means that the headline boost can be maintained indefinitely (in other words, there is no second-stage, time-limited boost). Of course, assuming a cooling system can take it. TDP is just a guidance for cooler selection. Again, motherboards can mess with it. If you see different numbers in practice, it's probably the motherboard.