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

[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

They already implemented it in the 6000 series, what is the problem with just c&p'ing that ip block?

This starts getting more into the "black box" thing - we really have no idea if it's that simple or not.

USB4 in a server? it would be more flexible to implement that over pcie.

Which could very well be why, on AM5, USB4 isn't present on the SOC and why it requires an extra chip on the motherboard.

This leads farther credence of desktop Ryzen being more of a "trickle down" from server than server being a "trickle up" from desktop.

It "requires" it as much as Win 11 requires TPM https://www.tomshardware.com/how-to/...pm-requirement

On a Ryzen 6000 device, you straight-up can't boot Linux unless allowing 3rd party certificates is enabled in the BIOS which is the corresponding Pluton BIOS option I was referring to (in theory, enabling that option disables Pluton).

[Anux]

And that's why I lean towards Pluton not being present across the server/desktop/dragon range, uh, range being more likely than Pluton being present, especially since it'd take more work for AMD to implement Pluton than to leave it out for this market segment.

They already implemented it in the 6000 series, what is the problem with just c&p'ing that ip block?

This also kind of goes back to what I said about USB4 not even being integrated into AM5 Ryzen 7000's SOC despite USB4 having much more obvious benefits.

USB4 in a server? it would be more flexible to implement that over pcie.

I would think that Pluton is at least doing something different from TPM if Pluton straight up requires 3rd party certificates to be disabled.

It "requires" it as much as Win 11 requires TPM https://www.tomshardware.com/how-to/...pm-requirement

Maybe, maybe not. Honestly, just like the above regarding differences with TPM, it's pretty difficult to say for sure if it even works like that or not since, by its nature, Pluton is very much a "black box".

That we can definitly agree on.

[Anux]

if you watch sales it looks like nvidia does well on 8nm compared to amd on 7nm...
its also looks like nvidia is able to earn more money per 1mm² chip die compared to amd...
if you benchmark with raytracing it also looks like nvidia is faster...

What does that have to do with process node?

i know only 1 group of people who would prever amd instead of nvidia and thats the linux people who want opensource drivers.

something doesn't add up there, linux market share is 1% AMDs is 20%. I'm not sure why Nvidia is still that dominant because their most sold cards are ones that don't support RT or are much to slow for RT, that rules RT out as it's selling point. I guess it's just like with Intel, consumers need time to realize AMD is competitive again.

my argument also still stands because AMD 7nm cards beat intel 6nm cards...

that could very well be the shitty drivers

the claim that the amd 6900xt is much more efficient than nvidia 2090 becomes a joke as soon as you discover that the amd 6950XT lost all the efficiency benefits.

Its even much more efficient than the 3090 and yes all 6*50 models are overclocked and have faster RAM making them less efficient but the 6950XT is still more efficient than 3090ti. https://www.techpowerup.com/review/m...x-trio/37.html

and Nvidias DLSS has more games who have it in the engine compared to FSR1.0/2.0...

Yes and those that don't have dlss can still use RSR, so it's better to compare native and not skew the results with upscalers.

[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

I'm not sure that it will be 100% monolithic, is there any source for that (it could very well splitt CPU/GPU)?

Current rumors are that it's at least partially monolithic with at least CPU core and I/O on the same chiplet.

But what leads you to assume they would drop Pluton on a notebook chip that will be used with Win 11 99% of times?

First off, I was differentiating between Dragon Range and Phoenix. If we assume that server/desktop truly does lack Pluton then, with Dragon Range basically being a derivative of desktop AM5, I would expect that to also lack Pluton.

Basically I see server/desktop/dragon range as all being in the same boat.

But Phoenix? All bets are off considering that Ryzen 6000, Phoenix's direct predecessor, had Pluton.

I don't think it's because of shared architecture but need for Win server, the webserver space is > 80% linux as is cloud but you have many companies running windows clients and those are typically surrounded by many Win servers. To have 2 different server CPU lineups would not be clever.

And that's why I lean towards Pluton not being present across the server/desktop/dragon range, uh, range being more likely than Pluton being present, especially since it'd take more work for AMD to implement Pluton than to leave it out for this market segment.

This also kind of goes back to what I said about USB4 not even being integrated into AM5 Ryzen 7000's SOC despite USB4 having much more obvious benefits.

What kind of compatibility would there be needed if pluton does nothing different from TPM?

I would think that Pluton is at least doing something different from TPM if Pluton straight up requires 3rd party certificates to be disabled.

the compute die could just ask the I/O die "is pluton available?" and just report that back to the OS.

Maybe, maybe not. Honestly, just like the above regarding differences with TPM, it's pretty difficult to say for sure if it even works like that or not since, by its nature, Pluton is very much a "black box".

Is it really? If I had to implement it, I would just make 2 different APIs for the same hardware.

Again, the black box nature makes things a little bit difficult to know for absolute 100% sure but if the PSP and Pluton were on the same hardware then I really would have expected AMD to have listed that part of the Ryzen 6000 diagram as one larger rectangle labeled as something like "Pluton Platform Security Processor" or "Platform Security Pluton Processor" or the like (e.g. a "PPSP" or "PSPP")

Also, the PSP actually takes care of part of the boot sequence which is why disabling it hasn't really been a thing all this time despite disabling Pluton already being a thing (if the BIOS options are exposed of course).

[Dukenukemx]

the claim that the amd 6900xt is much more efficient than nvidia 2090 becomes a joke as soon as you discover that the amd 6950XT lost all the efficiency benefits.

Anux said 3090, not 2080. There's a big difference.

also to activate FSR1,0 globally really is not the same as activating it in the game Engine.

"Native implementation produce much higher fidelity image as the upscale is applied before any UI elements and post-processing effects are applied. Third-party implementation such as Steam Deck's apply the upscaling at the very end of the rendering pipeline, blurring UI and producing significantly much worse image."
https://www.pcgamingwiki.com/wiki/Fi...per_Resolution

WINE: "

Wine	Can be applied to any Vulkan game (even dxvk and vkd3d-proton), but will look worse than a "real" implementation as it upscales the final image, instead of upscaling before post-effects.

"

the 2D overlay GUI/HUD of the game if you activate it globally is damanged and if you activate it in the game engine the GUI/HUD of the game is not reduced in resolution.

Better than nothing. World of Warcraft surprisingly supports FSR.

and Nvidias DLSS has more games who have it in the engine compared to FSR1.0/2.0...

https://www.pcgamingwiki.com/wiki/Li...lity_upscaling

Yea and DLSS is Nvidia exclusive plus earlier implementations kinda sucked.

[qarium]

It's been done without DLSS and FSR. Also, FSR is really easy to enable in Linux or Windows and works both AMD and Nvidia cards. Probably works on Intel but I haven't tried.
Just add these to the .profile and reboot. When you set the correct resolution you should get FSR working.
export WINE_FULLSCREEN_FSR=1
export WINE_FULLSCREEN_FSR_STRENGTH=2

the claim that the amd 6900xt is much more efficient than nvidia 3090 becomes a joke as soon as you discover that the amd 6950XT lost all the efficiency benefits.

also to activate FSR1,0 globally really is not the same as activating it in the game Engine.

"Native implementation produce much higher fidelity image as the upscale is applied before any UI elements and post-processing effects are applied. Third-party implementation such as Steam Deck's apply the upscaling at the very end of the rendering pipeline, blurring UI and producing significantly much worse image."
https://www.pcgamingwiki.com/wiki/Fi...per_Resolution

WINE: "

Wine	Can be applied to any Vulkan game (even dxvk and vkd3d-proton), but will look worse than a "real" implementation as it upscales the final image, instead of upscaling before post-effects.

"

the 2D overlay GUI/HUD of the game if you activate it globally is damanged and if you activate it in the game engine the GUI/HUD of the game is not reduced in resolution.

and Nvidias DLSS has more games who have it in the engine compared to FSR1.0/2.0...

https://www.pcgamingwiki.com/wiki/Li...lity_upscaling

[qarium]

Really? You know that DLSS is not available in all games but RSR is. Lets just compare native and leave it at that.

if you watch sales it looks like nvidia does well on 8nm compared to amd on 7nm...
its also looks like nvidia is able to earn more money per 1mm² chip die compared to amd...
if you benchmark with raytracing it also looks like nvidia is faster...

i know only 1 group of people who would prever amd instead of nvidia and thats the linux people who want opensource drivers.

my argument also still stands because AMD 7nm cards beat intel 6nm cards...

only because intel orders some 3nm/4nm/5nm wavers does not mean they win a battle.

[Anux]

TPM (part of the PSP) and Pluton are already two seperate parts on Ryzen 6000.

Really? I didn't know that, just assumed they dropped PSP. But wouldn't the most logical way going forward be to keep PSP and Pluton like already done on the 6000? That would solve all you Linux fears. After all the die space for this stuff is negligible.

Compute die = the CCD with the CPU cores and stuff. Zen4 mobile hasn't been announced at all in any official capacity and only Dragon Range

Dragon range is for mobile (55+ W) workstations and gaming rigs.

not Phoenix, uses the exact same CCD. Phoenix and below look to be using a monolithic CCD that includes I/O and, by definition of it being on 4nm (which is information we do know officially), it cannot be the same.

I'm not sure that it will be 100% monolithic, is there any source for that (it could very well splitt CPU/GPU)? And yes it will definitly be a different mask/process. But what leads you to assume they would drop Pluton on a notebook chip that will be used with Win 11 99% of times?

Also, on AMD's previous mobile offerings that have subsequently been used on desktop APUs (e.g. 5700G), even the portion involving the cores are not 100% identical

Jepp they were always half a year behind and a slight reiteration of the core design, it would be logical to assume Phoenix gets the same treatment.

...that's kind of my point though. Pluton doesn't make sense in server, so why would it therefore be present in desktop (or even Dragon Range) due to the shared architecture?

I don't think it's because of shared architecture but need for Win server, the webserver space is > 80% linux as is cloud but you have many companies running windows clients and those are typically surrounded by many Win servers. To have 2 different server CPU lineups would not be clever.

No no, I didn't mean that it'd be on the CCD, I was saying that the CCD would still need to be compatible with any such I/O die

What kind of compatibility would there be needed if pluton does nothing different from TPM? And even if, the compute die could just ask the I/O die "is pluton available?" and just report that back to the OS. Any pluton protocol and communication stuff would be the problem of the OS.

it's actually separate hardware from the PSP.

Is it really? If I had to implement it, I would just make 2 different APIs for the same hardware.

[Dukenukemx]

and let me ask was this with or without DLSS2.x ?

what you say is maybe true and still a loss on people who dont care about powerconsumsion and want the maximum performance.

also the 50w less power consumsion goes away as soon as you activate DLSS2.x...

of course AMD has FSR1.0 and FSR2.0 but most games who support technology like this only support DLss2.x--.-

It's been done without DLSS and FSR. Also, FSR is really easy to enable in Linux or Windows and works both AMD and Nvidia cards. Probably works on Intel but I haven't tried.

Just add these to the .profile and reboot. When you set the correct resolution you should get FSR working.
export WINE_FULLSCREEN_FSR=1
export WINE_FULLSCREEN_FSR_STRENGTH=2

[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

Amusingly, Red Dead Redemption 2 just got announced today with FSR 2.0.

As said its just TPM at its core, either one can just use the TPM standard and ignore the Pluton extentions or there will be a switch to expose a standard tpm. (just guessing)

TPM (part of the PSP) and Pluton are already two seperate parts on Ryzen 6000.

Maybe that's the crux of our disagreement? You seem to be under the impression that it's part of the PSP when it is in fact separate hardware:
https://archive.ph/bFu0n

Zen 4 compute dies are the exact same from notebook to server

Compute die = the CCD with the CPU cores and stuff. Zen4 mobile hasn't been announced at all in any official capacity and only Dragon Range, not Phoenix, uses the exact same CCD. Phoenix and below look to be using a monolithic CCD that includes I/O and, by definition of it being on 4nm (which is information we do know officially), it cannot be the same.

Also, on AMD's previous mobile offerings that have subsequently been used on desktop APUs (e.g. 5700G), even the portion involving the cores are not 100% identical because these have reduced cache and, due to the monolithic nature, then has the cores and I/O moved closer together in order to keep die size down rather than just taking the server/desktop CCD and disabling cache of power/thermal reasons like was done on some Athlon II CPUs.

Are you going to split hairs or do you want to tell me that these are different dies? Because to implement Pluton in X3D they would have to make a new design, extra masks, etc. That would go directly against AMDs strategy behind the chiplett approach. For one SKU?

...that's kind of my point though. Pluton doesn't make sense in server, so why would it therefore be present in desktop (or even Dragon Range) due to the shared architecture?

Heck, there is even AM4 server hardware now as seen by the following being from Asrock Rack rather than plain-old Asrock:
https://www.asrockrack.com/general/p...?Model=X470D4U

There is simply no logical reason why pluton would be on the compute die. Just name one.

No no, I didn't mean that it'd be on the CCD, I was saying that the CCD would still need to be compatible with any such I/O die and, as you stated above, it would seem like making additional work for minimal (if any) gain to have Pluton on anything above a mobile-centric monolithic die since, as I stated above, it's actually separate hardware from the PSP.