AMD Announces Ryzen 7000 Series "Zen 4" Desktop CPUs - Linux Benchmarks To Come

[qarium]

That was in response to https://www.phoronix.com/forums/foru...29#post1343729 where he claimed all gaming benchmarks are unfair and fraud.

but it is fact that these gaming benchmarks are unfair and fraud!...

compile the benchmark native to ARM 64bit and come up with a WebGPU or Metal solution to benchmark.

then we will get real numbers and not some x86 legacy fraud comparison.

[Dukenukemx]

Not a question I can answer, but it seems like they probably should.

The answer is maybe, but that depends on the developers.

[Anux]

but it is fact that these gaming benchmarks are unfair and fraud!...

Dude, you remind me of my 3 year old niece.

compile the benchmark native to ARM 64bit and come up with a WebGPU or Metal solution to benchmark.

I never came up with those gaming benchmark stuff, that was your claim therefore you are in charge of providing benchmarks. I'm totaly fine with the ones Michael provided.

[qarium]

Dude, you remind me of my 3 year old niece.

yes of course i know this world is so evil from ground up that any idea of fairness is fixed to the dream of a 3 year old children.

I never came up with those gaming benchmark stuff, that was your claim therefore you are in charge of providing benchmarks. I'm totaly fine with the ones Michael provided.

its not me who do want the focus on gaming benchmarks its: Dukenukemx

[qarium]

The answer is maybe,

"“3nm will cost $4 billion to $5 billion in process development, and the fab cost for 40,000 wafers per month will be $15 billion to $20 billion.”"

https://www.extremetech.com/computing/272096-3nm-process-node

5nm costs 542million dollars to design a chip
4nm is like 1-2 billion dollars to design a chip

looks like intel did take a lot of risk by ordering a lot of 5nm/4nm/3nm from TSMC...
and they did the order before the crypto mining bubble did burst ...
if they did order to much and can't sell them they will pay a high price for it.

and history shows lower node does not garantee a win... nvidia RTX3090 is on 8nm and wins agaist amd on 7nm and 6nm and amd wins agaist intels 6nm gpus... this means even if intel has 5/4/3nm they can still lose.

a chip company can go bankrupt with shit like that.

[Anux]

and history shows lower node does not garantee a win... nvidia RTX3090 is on 8nm and wins agaist amd on 7nm and 6nm and amd wins agaist intels 6nm gpus... this means even if intel has 5/4/3nm they can still lose.

Last time I checked the 6900XT traded blows with 3090 maybe 3% behind on average at 4K. And that is with 50W less, wouldn't exactly call this loosing.

[Mahboi]

Forgive me for being late to this thread...



I've been keeping tabs on this subject every once in a while for the last couple of months now (I did an OCD-level search on subject just two night ago) and, even now, there's been literally (not figuratively!) nothing stated anywhere that Ryzen 7000 / Zen4 actually has Pluton outside of some random internet commentors on places like Reddit proclaiming it does simply based on the fact that Ryzen 6000 / Zen3+ has it, or because places like TomsHardware keep stating things like "nothing has been confirmed but Pluton is expected."

The problem with all of these assumptions is that Ryzen 6000 / Zen3+ also has integrated USB4 right on the SOC and yet Ryzen 7000 / Zen4 does not and requires the motherboard manufacturer to include a separate additional chip on the motherboard for USB4 support - this was confirmed back at Computex. Therefore, something being implemented into Ryzen 6000 / Zen3+ very much does not mean that it's implemented into Ryzen 7000 / Zen4 - and something like USB4 would be much more of a "marketable bullet point", yet even that is not included.

So let's not get ahead of ourselves when literally (not figuratively!) nothing has been stated regarding Pluton considering that, in terms of how companies tend to advertise product feature-set (e.g. AMD did not advertise on previous generations of Ryzen/Zen that they did not support AVX-512 and basically just straight-up pretended that such a function didn't exist), it's looking more and more likely that Ryzen 7000 / Zen4 doesn't have Pluton.

Very interesting...I do seem to recall Lisa Su saying they would move forward with Pluton starting Zen3+. Have they changed their mind, are they not advertising it, or is there some sort of problem? First I heard that this wasn't put in, the statements were quite clear from MS and AMD/Intel that it would be in every CPU moving forward from Zen3+.

[Anux]

Very interesting...I do seem to recall Lisa Su saying they would move forward with Pluton starting Zen3+. Have they changed their mind, are they not advertising it, or is there some sort of problem? First I heard that this wasn't put in, the statements were quite clear from MS and AMD/Intel that it would be in every CPU moving forward from Zen3+.

Pluton is just hardware TPM integrated in the CPU with an even more fancy marketing campaign around it. Mostly because most external hardware and internal firmware implementations are vulnerable to certain attacks. But don't be fooled there will be attacks and backdoors for it like before.
Edit: and yes it will be in all future CPUs from AMD, Intel and Qualcomm, basically every CPU manufacturer that wants to run Windows 11 will have it.

[NM64]

Oct. 2022 UPDATE: Welp, if Anandtech's Ryzen 7950X review is to be believed, then desktop Zen4 / Ryzen 7000 does indeed have Pluton:

• https://www.anandtech.com/show/17585...the-high-end/6

————————————————————————————————

ORIGINAL POST

Have they changed their mind, are they not advertising it, or is there some sort of problem?

I imagine that the vast difference in SOC topology between server/desktop and mobile has a lot to do with it - anything like Pluton would not be part of the compute die(s) (PSP, which handles TPM, is currently part of the I/O die), and mobile does not use chiplets at this time (AMD's upcoming Zen4 "Dragon Range" is rumored to basically just be the two-chiplet desktop repackaged for laptops though).

Furthermore, consider that the desktop core chiplets are the exact same compute die chiplets used in server platforms, and we know that there are a lot of servers running Linux rather than Windows (I would think that the likes of Google for example wouldn't be too impressed if they were forced to run Microsoft's Pluton security solution, especially since they already have their own "Google TPM" security solution, at least in Chrome OS, that simply leverages TPM).

One thing in particular is that there's rumors from Moore's Law is Dead that Zen3 and Zen4 is basically designed so that, if AMD wanted to, they could pair Zen3's I/O die with Zen4 compute die(s) on AM4 or pair Zen4's I/O die with Zen3 compute die(s) on AM5, and obvious Zen3 non-plus does not support Pluton.

On the subject of I/O dies, it's worth noting that the server I/O die is different from the desktop I/O die despite the compute dies being identical. So, much like the mixing-and-matching of Zen3 and Zen4 compute and I/O dies, you still need the Zen4 compute dies to be compatible with both the desktop and server I/O dies.

Pluton is just hardware TPM

"Just"... except that Microsoft mandates that Microsoft requires the 3rd Party Certificate to be disabled by default if you ship Windows pre-installed on a device with Pluton.

Now you may say "just toggle the corresponding BIOS setting" but that's assuming such a setting is even exposed - have you seen how locked down most OEM systems are? Heck, even enthusiast-focused gaming handhelds, devices ideally suited for the future release of SteamOS 3, can have even more locked down BIOS settings (I have a 4800U OneXPlayer and I've never seen a BIOS with so few user-changable options, and I'm a PC hardware enthusiast that literally, not figuratively, has more used motherboards laying around than I know what to do with).

It will be in all future CPUs from AMD, Intel and Qualcomm, basically every CPU manufacturer that wants to run Windows 11 will have it.

You're doing the very thing I already stated that too many armchair experts are doing - proclaiming definitive statements when we have no definitive proof.

If you meant all CPUs period, then we can already say that's false since there have been several desktop CPUs launched after Ryzen 6000 (e.g. the 5800X3D) that lack Pluton. If you meant CPU architecture, then that goes back to what I said above in that Pluton is not something that would be part of the actual compute die(s) and therefore we have a divergent architecture between server/desktop and mobile.

[Mahboi]

(AMD's upcoming Zen4 "Dragon Range" is rumored to basically just be the two-chiplet desktop repackaged for laptops though).

The grapevine (aka extremely trustworthy youtube leakers, ofc) says that it won't even be that. The goal with Dragon Range may be a simple no-chiplet design with the intent of aiming for sub 55W power consumption. Which if they do succeed, might actually make me replace my near-dead laptop.

A low-ish power CPU with RDNA3 is gonna be a strong sell I feel, even in the upcoming economic crisis. AMD has put its balls in the right holes, and may yet reap rewards for all of the 2020s...