AMD Zen 5 Chiplet Built on 4 nm, "Granite Ridge" First Model Numbers Leaked

[Minus Infinity]

Good news, the Infinity fabric is now rated for 2400MHz meaning sweet spot for Zen 5 will be 7200MT/s DDR5 up from 6000MT/s. If we can get some CAS32 7200 memory should be a nice uplift vs Zen 4.

 

[atomsymbol]

... And to be clear, the 7030 was a "supercomputer" in 1961.

Yes, it *was* a supercomputer. The 7030 supercomputer chip is roughly equivalent to 80286 (year 1982) or 80386 in *complexity*, and it is less advanced than an 80486 (year 1989). In other words: middle class population in their homes had access to the equivalent of the 7030 supercomputer chip during the 1980-ties, and had access to a chip surpassing the 7030 supercomputer in 1990-ties. It is year 2024 now - what do you think is the complexity of the chips the middle class population has access to in their homes!

 

[john_]

This isn't any different from last gen. X3D cache is also going into server products which generate more money and are a higher priority. Would never see this launch at the same time consumer level is always lower priority.

Which is one of the reasons why they lost their momentum in the retail CPU market.
On the contrary, Nvidia hasn't stopped pushing gaming GPUs because of AI. We are already hearing about an RTX 5090. They could postpone RTX 5090 by a year if they wanted. They are pretty fine and secured with RTX 4090 as the top GPU in the market.

They will not, lol. They have comfortably settled in the consumer market, laptop chips are not yet going to be announced (I think?), so we have no idea if they will have NPUs. And the market that AMD truly cares about doesn’t get hyped by marketing buzzwords. They iterate on what matters - actual computing power and scalability.

Nice FUD right there.
AMD's APU's have had an NPU for over a year, while the upcoming CPU doesn't. You haven't figured out why?

AI is already a marketing advantage, just not on the consumer side. And it won't likely go there either. I still have enough faith in humanity. We're trying it out now, and the early results are not encouraging. Meanwhile, the real cost of these solutions is slowly becoming impossible to hide for us, and they. Are. Immense.

The cost/benefit scenario just doesn't work here. We couldn't do an autonomous car yet without running into the human factor, not even pilot projects work out well. AI? Never. We're trolling it and having fun with it. It ain't gonna work.

When RX 7000 came out I was screaming about the lack of strong RT performance increase in the new GPU cards. People where telling me why raster is still much most important and why RT is still a gimmick. I was insisting that people will swallow the marketing about RT and only consider as future proof options those GPUs that can also offer RT performance. Especially in the hi end segment. I was an Nvidia shill back then.
So, how was RX 7000 sales? How is AMD in the gaming market today?

I see the same thing happening here. No need to repeat my shelf here I guess. No one will understand I guess when I will say that the whole buzzword will be about "AI PC this", "AI PC that". No need to explain that OEMs and Microsoft in particular will have reasons to keep promoting Qualcomm and Intel chips leaving AMD looking like fool and last year's tech, at least on desktops. Because on desktops obviously we don't need AI.

So, Zen 5 introduction, with no AI - just a slide trying to make up some "optimizations" excuses and no X3D.

When you feel unexcited about Zen 5 and start seeing a flood of Snapdragon and Intel AI PCs, you'll might realize the reason.
Marketing is everything today. People don't have time to research. They just throw money on "buzzwords".

 

[JasBC]

Boot time had nothing to do with the CPU, it was due to AM5 being a newer platform.

Mind you there's no need to "hope" for anything, that issue was fixed a long time ago.

Still horrid boots on my B650 Tomahawk. . .

Which is one of the reasons why they lost their momentum in the retail CPU market.
On the contrary, Nvidia hasn't stopped pushing gaming GPUs because of AI. We are already hearing about an RTX 5090. They could postpone RTX 5090 by a year if they wanted. They are pretty fine and secured with RTX 4090 as the top GPU in the market.







When RX 7000 came out I was screaming about the lack of strong RT performance increase in the new GPU cards. People where telling me why raster is still much most important and why RT is still a gimmick. I was insisting that people will swallow the marketing about RT and only consider as future proof options those GPUs that can also offer RT performance. Especially in the hi end segment. I was an Nvidia shill back then.
So, how was RX 7000 sales? How is AMD in the gaming market today?

I see the same thing happening here. No need to repeat my shelf here I guess. No one will understand I guess when I will say that the whole buzzword will be about "AI PC this", "AI PC that". No need to explain that OEMs and Microsoft in particular will have reasons to keep promoting Qualcomm and Intel chips leaving AMD looking like fool and last year's tech, at least on desktops. Because on desktops obviously we don't need AI.

So, Zen 5 introduction, with no AI - just a slide trying to make up some "optimizations" excuses and no X3D.

When you feel unexcited about Zen 5 and start seeing a flood of Snapdragon and Intel AI PCs, you'll might realize the reason.
Marketing is everything today. People don't have time to research. They just throw money on "buzzwords".

The high-end 7000-series cards are the only ones which have sold enough to enter upper crust of the Steam Hardware Survey-rankings. . .

 

[ARF]

View attachment 349471

This leak here is much more interesting.
Lower TDP for ryzen 7 and slightly higher clocks.

This is obviously very fake.
If you think for a second, you will easily realise that those specs are impossible.

 

[atomsymbol]

When RX 7000 came out I was screaming about the lack of strong RT performance increase in the new GPU cards.

Unless you employ yourself at AMD and work on improving the GPU's raytracing performance, your screaming is only waste of your own time. And time (not spice) is the most precious substance in the universe.

 

[Vayra86]

Marketing is everything today. People don't have time to research. They just throw money on "buzzwords".

Thats great but it also directly underlines a crucial difference in how we perceive and approach these companies. You are looking at AMDs bottom line like a shareholder, I look at it as a customer that does exercise due diligence.

And I believe the latter group will win. Always. It just takes time and persistence. We have seen it every time: you can market a tech but eventually people will see and judge it for what it is. I dont subscribe to the idea that following the herd is a good thing for anyone. It just makes a select few rich. Not any of us though.

You said it well 'do we feel excitement'... the real question is why anyone needs to be excited at every product release every gen. You dont benefit from excitement. It just leads you to unnecessary purchases.

TL DR Screw that herd mentality and think for yourself. Nothing of value will go lost.

 

[AMDK11]

Your extreme level of ignorance is on full display there sparky. Speed / performance is relative to many unrelated factors (materials tech, lithography), and has nothing to do with the use of "branch prediction". "IPC" is not and never has been related to "branch prediction". And to be clear, the 7030 was a "supercomputer" in 1961.

You're welcome for the education, welcome to my short but distinguished ignore list.

With this post you have shown that you have no idea what you are writing about.

 

[john_]

The high-end 7000-series cards are the only ones which have sold enough to enter upper crust of the Steam Hardware Survey-rankings. . .

Then why is AMD rumored to be dropping the hi end models from the next series? Maybe because they need to offer features and especially RT performance? It's a different thing selling really well, than selling enough especially after discounts. People are driven by marketing and AMD knows that people are unwilling to pay a significant price if the product doesn't tick enough boxes.

Unless you employ yourself at AMD and work on improving the GPU's raytracing performance, your screaming is only waste of your own time. And time (not spice) is the most precious substance in the universe.

Well, you can say that posting and reading in forums is usually a waste of time. But it is our time to waste and our choice how to waste it. Well, most of the times. Some times we waste time by reading and quoting people who try to play it smart. Anyway....

Thats great but it also directly underlines a crucial difference in how we perceive and approach these companies. You are looking at AMDs bottom line like a shareholder, I look at it as a customer that does exercise due diligence.

And I believe the latter group will win. Always. It just takes time and persistence. We have seen it every time: you can market a tech but eventually people will see and judge it for what it is. I dont subscribe to the idea that following the herd is a good thing for anyone. It just makes a select few rich. Not any of us though.

You said it well 'do we feel excitement'... the real question is why anyone needs to be excited at every product release every gen. You dont benefit from excitement. It just leads you to unnecessary purchases.

TL DR Screw that herd mentality and think for yourself. Nothing of value will go lost.

Not as a shareholder, but as a customer who wishes to keep having choices. AMD's failure in the GPU market limits our choices until they are ready to come up with something competitive.
Doing the same mistake in the CPU market could lead in a more complex situation where our choices could be either going Intel (mostly for desktop), or trying our lack with Qualcomm (laptops) and Windows on ARM.

The herd pay and the select few that end up rich are companies.
AMD invested in creating CPUs with more cores and then improving the idea by going chiplets. They managed to get back market share and today be a healthy company. The herd payed for extra cores. Cores that where still behind Intel's in IPC, but they where more.

Nvidia predicted, invested in AI and now enjoys a success that no one could predict. They created, as they do usually, new "buzzwords". DLSS, Raytracing. The herd payed, AMD's products started looking inferior. The herd kept paying Nvidia.

Intel invested in Hybrid CPUs, just so it can add and advertise more cores. It win back market share. People didn't cared if those where P cores or E cores. At least not the majority. Now Intel is betting it's future in fabs. If they manage this, in 3-5 years they will have parity with AMD in manufacturing node. Add to that their years of experience with hybrid CPUs and AMD will start losing the CPU market.

Qualcomm is betting it's Windows future in AI. Microsoft is helping in that direction. Intel is going full in in AI, fearing losing one more market opportunity. Are they stupid?

TLDR It has nothing to do with how you or I think. We are not part of the herd, but companies need to target that herd, if we wish to keep having choices in the future. If AMD loses it's chances today that Intel is still behind because of it's manufacturing problems, it wouldn't have the power tomorrow to fight against Intel, Qualcomm and probably Nvidia in the CPU market. And guess what. Intel, Qualcomm and Nvidia all consider themselves premium companies, meaning prices will go up. AMD needs to wake up and move now, before it gets crowded in the CPU market.

 

[Denver]

This is obviously very fake.
If you think for a second, you will easily realise that those specs are impossible.

Nah, WhyCry from videocardz has confirmed it with several sources. This must be the only real leak so far.

 

[atomsymbol]

Well, you can say that posting and reading in forums is usually a waste of time. But it is our time to waste and our choice to how to waste it. Well, most of the times. Some times we waste time by reading and quoting people who try to play it smart. Anyway....

Actually, this time participating in this discussion was quite valuable to me (despite the fact that it seems I was communicating with a "confused person" to say it politely): I have learned some things about the old 7030 mainframe architecture, its instruction set, and how terminology can vastly change over time.

 

[john_]

Actually, this time participating in this discussion was quite valuable to me (despite the fact that it seems I was communicating with a "confused person" to say it politely): I have learned some things about the old 7030 mainframe architecture, its instruction set, and how terminology can vastly change over time.

Yeah, figures. Others waste time, you use it quite valuably.
You know what you post, others being confused.
You remind me so so so many people who post online.
Have a nice day.

 

[ARF]

Nah, WhyCry from videocardz has confirmed it with several sources. This must be the only real leak so far.

I doubt it.
Ryzen 5 7600X 105W 5.3GHz --> Ryzen 5 9600X 65W 5.4GHz (higher clocks, lower TDP)
Ryzen 9 7950X 170W 5.7GHz --> Ryzen 9950X 170W 5.7GHz (same clocks, same TDP)

Is it just me who can't calculate, or these people are very wrong, or there is a bug in the hardware ?

 

[atomsymbol]

Yeah, figures. Others waste time, you use it quite valuably.

No comment.

You know what you post, others being confused.

I wrote "confused" in singular, not plural. It is a particular person, not multiple people.

You remind me so so so many people who post online.

That is a false statement.

Have a nice day.

In this Internet free speech era, it is unlikely to have a nice day.

I doubt it.
Ryzen 5 7600X 105W 5.3GHz --> Ryzen 5 9600X 65W 5.4GHz (higher clocks, lower TDP)
Ryzen 9 7950X 170W 5.7GHz --> Ryzen 9950X 170W 5.7GHz (same clocks, same TDP)

Is it just me who can't calculate, or these people are very wrong, or there is a bug in the hardware ?

170 / 16 * 6 = 63.75

 

[Denver]

I doubt it.
Ryzen 5 7600X 105W 5.3GHz --> Ryzen 5 9600X 65W 5.4GHz (higher clocks, lower TDP)
Ryzen 9 7950X 170W 5.7GHz --> Ryzen 9950X 170W 5.7GHz (same clocks, same TDP)

Is it just me who can't calculate, or these people are very wrong, or there is a bug in the hardware ?

Just an optimistic bet; I assume that the specified TDP is PL2 on all CPUs. AMD must have been using a 4nm interaction that was totally geared towards efficiency.

PS: God... There are a lot of fakes being created these days.

 

[Wirko]

From all we’ve heard so far, they are sticking with 8 per CCD for Zen 5 and maaaaaybe Zen 6 goes for 16. There are rumblings of a 12 core CCD, if I remember correctly, but that’s for Zen5c.

There are other limits like L3 size, which is costly and difficult to increase, and RAM bandwidth, which does go up over time, but slowly. 16 cores per CCD (up to 32 per processor) wouldn't make much sense in these circumstances. 12, hardly. 10 might be the optimum number - but AMD seems to stick to powers of 2 wherever possible.

 

[ARF]

16 cores per CCD

Would be perfect.

Ryzen 9 9950X: two disabled CCDs with 24 cores in total.
Ryzen 9 9900X: one full CCD with 16 cores.
Ryzen 7 9700X: one full CCD with lower clocks / lower TDP.
Ryzen 7 9700: one disabled CCD with 12 cores.
Ryzen 5 9600X: one disabled CCD with 12 cores in lower clocks / lower TDP.
Ryzen 5 9600: can be an APU with integrated GPU: 8 cores.
Ryzen 3 9400X: can be an APU with integrated GPU: 6 cores.

(up to 32 per processor)

No. See above.

 

[evernessince]

Still horrid boots on my B650 Tomahawk. . .

No the issue was definitely fixed, it's just that some MSI boards with memory context restore on default tend to not enabling memory context restore sometimes for some reason. You should ensure that your BIOS is up to date and that memory context restore is set to enabled (not just default). Otherwise you can blame MSI for their poor implementation.

It's generally understoodthat when a person makes a broad statement like issue x or issue y is solved it doesn't mean you won't still be able to find examples or outliers of it still occuring. You will always be able to find those no matter the product, especially with the current scale of production. No one would be able to make a definative statement ever if we followed your logic of any example no matter how minute disproving a broad statement with a wide body of evidence supporting it. The bar for disproving such a statement would be proving that a majority of boards still have a long boot time, a far cry from what you've presented here.

 

[ValenOne]

I am not too surprised the processor counts aren't moving up since right now its hard to justify on the consumer level beyond 16 cores. I am very curious on the memory controller upgrade to handle higher memory speeds as that is pretty cool since AMD plays very well with faster memory which may in turn make overclocking memory go further.

Would not surprise me if the names stay the same. Though they could also just be place holders at the moment.

Stagnation.

 

[atomsymbol]

Stagnation.

What alternative do you have in mind? Are you working for a startup that is progressing faster than the rest of chip industry?

 

[GhostRyder]

Stagnation.

Its more that we have made a huge jump already and programmers are not really moving forward as quick in most spaces to go beyond that. If 6 years down the line everything is exactly the same and we are just doing small changes like 10% IPC improvement every year, then we can discuss stagnation.

 

[demirael]

By such logic, the *only* grounds-up x86 architecture was the Intel 4004 CPU (year 1971).

4004 wasn't x86.

 

[atomsymbol]

4004 wasn't x86.

That is disputable. For example, 4004 had the DAA instruction (Decimal Adjust Accumulator) which is still being supported by all AMD and Intel x86 CPUs today. You can literally find "DAA—Decimal Adjust AL After Addition" section in Intel's instruction set reference manual even today. The instruction is valid only in 32-bit mode on x86-64 CPUs, but still, the CPU has to support the instruction in case you start a 32-bit binary in Linux or Windows.

 

[JasBC]

No the issue was definitely fixed, it's just that some MSI boards with memory context restore on default tend to not enabling memory context restore sometimes for some reason. You should ensure that your BIOS is up to date and that memory context restore is set to enabled (not just default). Otherwise you can blame MSI for their poor implementation.

It's generally understoodthat when a person makes a broad statement like issue x or issue y is solved it doesn't mean you won't still be able to find examples or outliers of it still occuring. You will always be able to find those no matter the product, especially with the current scale of production. No one would be able to make a definative statement ever if we followed your logic of any example no matter how minute disproving a broad statement with a wide body of evidence supporting it. The bar for disproving such a statement would be proving that a majority of boards still have a long boot time, a far cry from what you've presented here.

I was just sharing my experience. And you're right - I should really check if MCR is enabled, I currently have no idea.

 

[evernessince]

I was just sharing my experience. And you're right - I should really check if MCR is enabled, I currently have no idea.

Hopefully that fixes the issue for you. Of the reviews I read through for MSI AM5 boards that seems to help most people. That and updating the BIOS.