News AMD's Zen 2 CPU for PlayStation 5 is 35% smaller than a Ryzen 7 3800X but sacrifices nothing in terms of gaming performance

[Admin]

It has been discovered that AMD was able to shrink Sony's custom PS5 Zen 2 CPU by 35% compared to normal Ryzen CPUs by chopping off parts of the CPU that aren't needed for gaming.

AMD's Zen 2 CPU for PlayStation 5 is 35% smaller than a Ryzen 7 3800X but sacrifices nothing in terms of gaming performance : Read more

 

[gg83]

What about the Xbox chip? Did Microsoft keep the fpu the same? Or is it so similar there's no sense in mentioning?

 

[usertests]

Great article and everyone should be reading Chips and Cheese (or at least glance at it every week or two to see what's being covered).

The article notes that AMD has not had to do the same thing for Zen 4c, you get everything in the full core but clock speed, and then less L3 cache. I wonder if a future PS6/Xbox would take both approaches at the same time.

 

[thestryker]

What about the Xbox chip? Did Microsoft keep the fpu the same? Or is it so similar there's no sense in mentioning?

The CPU portion of the Xbox Series X/S die is the same as Zen 2 so they didn't have AMD make any changes.

 

[igor_kavinski]

https://forums.anandtech.com/thread...ranite-ridge-ryzen-8000.2607350/post-41178489

The FPU is 35% smaller, not the CPU!

 

[Deleted member 2731765]

The chip review outlet found that Sony's custom-designed Zen 2 chip sports a heavily cut-down FPU enabling the CPU to be shrunk by 35%.

AMD's Zen 2 CPU for PlayStation 5 is 35% smaller than a Ryzen 7 3800X​

Seriously ? Where does the original article state this ? It's the FPU's size they are talking about, NOT the CPU.

Though, size of a quad core cluster only goes down by 5.8%. That’s not enough to enable a dramatic core count increase or a much smaller and cheaper die.

 

[bit_user]

I've got to say it: I feel somewhat vindicated. I predicted there's no way PS5 would have 8x full Zen 2 cores, and I was (technically) right!

Thanks to Brutus, I got to take a closer look at just how AMD got Zen 2’s FPU to drop from 0.91 to 0.59 mm2.

Sony/AMD: Et tu, Brute?

(Sorry, I had to say that too!)
: D

 

[bit_user]

everyone should be reading Chips and Cheese (or at least glance at it every week or two to see what's being covered).

Yes, if you're into that sort of thing. It replaced Anandtech as my go-to for deep architecture-level analysis & microbenchmarking, but they do it even better!

And if you're a regular reader, please chip in a couple $/mo to support C&C on Patreon. They're ad-free, so entirely donation-supported.

 

[bit_user]

Seriously ? Where does the original article state this ? It's the FPU's size they are talking about, NOT the CPU.

Though, size of a quad core cluster only goes down by 5.8%. That’s not enough to enable a dramatic core count increase or a much smaller and cheaper die.

I imagine Aaron Klotz skimmed the C&C article too quickly and got to the part with the die size comparison:

​

I don't see where C&C comes out and says it, but the vast majority of the cluster size reduction seems due to the reduction in L3 cache between the APU-oriented clusters vs. desktop/server clusters. As you state, they do explicitly specify that the cores, themselves, only shrunk by 5.8%. Furthermore, we know Zen 2 APUs have half the L3 cache per core as their desktop cousins.

 

[usertests]

I imagine Aaron Klotz skimmed the C&C article too quickly and got to the part with the die size comparison:

Alternatively he picked up the Zen 4c tidbit. There were two 35% reductions in the article to get confused by!

A 35% reduction in FPU area by itself is impressive. But the size of a quad core cluster only goes down by 5.8%.

A smaller clock mesh and other optimizations let Zen 4c achieve a 35% area reduction for the entire core, not just the FPU.

35% is the core size reduction for Zen 4c from Zen 4, not taking into account different amounts of L3 cache.

 

[gg83]

I imagine Aaron Klotz skimmed the C&C article too quickly and got to the part with the die size comparison:

​

​

I don't see where C&C comes out and says it, but the vast majority of the cluster size reduction seems due to the reduction in L3 cache between the APU-oriented clusters vs. desktop/server clusters. As you state, they do explicitly specify that the cores, themselves, only shrunk by 5.8%. Furthermore, we know Zen 2 APUs have half the L3 cache per core as their desktop cousins.

So the clusters aren't the entire chip?

 

[bit_user]

So the clusters aren't the entire chip?

In this case, we're looking at two APUs and a desktop/server CCD. The PS5 and Steam Deck APUs include an iGPU, whereas the Zen 2 CCD has neither that, nor stuff like memory controller and PCIe.

So, it does make sense for them to focus on just the core clusters.

 

[gg83]

In this case, we're looking at two APUs and a desktop/server CCD. The PS5 and Steam Deck APUs include an iGPU, whereas the Zen 2 CCD has neither that,nor stuff like memory controller and PCIe.

So, it does make sense for them to focus on just the core clusters.

I see, thank you.

 

[DaveLTX]

I was reading it yesterday. As usual tomshw is living up to it's reputation as copy and paste of the first few sentences
35% FPU reduction is only a tiny tiny fraction, the L3 cache takes up the most space since Zen 1
AMD simply didn't do it for client because that reduction didn't make sense to bother with and also Zen 2 for consoles was later on after Zen 2 was frozen for computing
Now they shrink Zen 4 in smarter ways!

 

[Pierce2623]

I’ve read this article before now and I want to point out the CPU is NOT 35% smaller. The FPU is 35% smaller.

 

[TechyIT223]

Typical poor journalism from Tom's hardware. The author should be more careful when writing such tech articles.

Surely he might have overlooked the conclusion written in the original article, but that's not an excuse when you are submitting articles for big tech sites like Tom's. Sounds very unprofessional tbh.

If you read the original article it has clearly been mentioned that they are talking about NPU size. There is no confusion here.

But it seems this Aaron author definitely has reading comprehension issues, or he just hurriedly skipped through the entire article. 😅

No offense, but that's the current trend I'm observing here on Tom's.