Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

[izaic3]

Alright, so we've had some leaks so far. I don't know if any of it's been confirmed yet, as it's pretty early, but here is what I've surmised so far (massive grain of salt of course):

• Zen 3 "enhanced"
• DDR5 and LPDDR5 support
• PCIE4
• 6nm
• CVML
• Navi 2/RDNA/RDNA 2
• 12 CU
• Socket FP7/AM5
• Possible USB4
• 2021(unlikely) -2022 (more likely) release date

If if turns out to have RDNA 2 and 12 CU, I could see iGPU performance potentially almost doubling over Cezanne.

If I've made any mistakes or gotten anything wrong, please let me know. I'd also love to hear more knowledgeable people weigh in on their expectations.

 

[Joe NYC]

Laptops costing 2500$ base in a closed ecosystem. AMD can do such a APU, but what is the market for that when the competitor can put a CPU+discrete graphic card for a lot less? Because huge chips on top process node are not cheap.

With AMD still struggling to penetrate the laptop market in meaningful way, I don't think the best approach is to be all the things to all of the people.

Suppose the segment of people who upgrade their memory is 5%.

Suppose AMD decides it is not going serve this 5% but by doing so, it is going to provide 20% better solution to the remaining 95% of the market.

Apple figured out this, as a non-brainer. AMD and Intel are still stuck on this problem.

 

[leoneazzurro]

Apple can because they own the OS and the hardware as a whole.

 

[moinmoin]

With AMD still struggling to penetrate the laptop market in meaningful way, I don't think the best approach is to be all the things to all of the people.

Oh you sweet summer child. For the record: AMD currently enjoys the biggest mobile PC share it ever had.

 

[Joe NYC]

AMD aren't the cartel, they're the ones who are hamstrung by dumb OEM decisions. Who keeps putting single channel memory alongside APUs? Who keeps putting in dGPUs that are worse than the integrated graphics? It's not AMD.

But now, 1 year after M1 was released on the market, it is not like expecting AMD to create a new PC configuration with memory integrated in MCM. M1 already validated this concept.

How much more validation does AMD need? Does AMD really need to wait until Intel releases it, losing any time to market advantage?

One year later, Apple is extending it by by scaling the memory channels (width) for even more bandwidth to serve a more powerful GPU.

AMD has RDNA2 in Rembrandt, hungry for bandwidth, likely with no V-Cache. Integrating DRAM would have elevated it up a segment.

 

[Joe NYC]

Oh you sweet summer child. For the record: AMD currently enjoys the biggest mobile PC share it ever had.

But, isn't it correct that notebooks are Intel's strongest segment?

 

[moinmoin]

But now, 1 year after M1 was released on the market, it is not like expecting AMD to create a new PC configuration with memory integrated in MCM. M1 already validated this concept.

How much more validation does AMD need? Does AMD really need to wait until Intel releases it, losing any time to market advantage?

One year later, Apple is extending it by by scaling the memory channels (width) for even more bandwidth to serve a more powerful GPU.

AMD has RDNA2 in Rembrandt, hungry for bandwidth, likely with no V-Cache. Integrating DRAM would have elevated it up a segment.

This adds the cost for memory to the total cost of the package. Why AMD wouldn't do that is the same reason AMD stopped using HBM in consumer products even though that tech is even AMD's own invention.

But, isn't it correct that notebooks are Intel's strongest segment?

Because Intel was and is traditionally focusing on mobile first. AMD focused on the areas where it could make a big impact the fastest, servers as in cloud data centers and desktops as in the DIY market. APUs were also-rans.

 

[uzzi38]

I have Joe on ignore, because he likes to troll, but I peaked at his comment because I figured it was ridiculous. The thing is, he is comparing a year old architecture on an inferior process to a brand new CPU on N5/N5P.

Based on 3 different rumors, AMD will likely have 35-45W Zen 4 “H” parts launch at some point, and they will likely be faster and more efficient than the M1. Rembrandt is a stopgap for OEMs.

Rembrandt's not a stopgap, it's a good (roughly) doubling of GPU perf, significant battery life play and major I/O overhaul. Really the only thing that doesn't get a proper good touch up are the CPU cores themselves. Minor improvements there.

If anything Phoenix is probably more of a stopgap. Maybe a bit of a boost on the iGPU and it brings Zen 4 which is a very nice improvement, but the next big step up is really Zen 5 APUs.

 

[Joe NYC]

This adds the cost for memory to the total cost of the package. Why AMD wouldn't do that is the same reason AMD stopped using HBM in consumer products even though that tech is even AMD's own invention.

DRAM chips are far cheaper that DIMMs, especially if AMD gets them on discount direct from DRAM makers.

Also, better performance out of the same chip if it is in the same MCM. Less power. Cheaper mobo.

HBM needs interposers. But a single DRAM chip could probably be connected straight through the substrate. Isn't it what Apple did with original M1? I am not sure what's going on with the new ones

 

[Joe NYC]

Rembrandt's not a stopgap, it's a good (roughly) doubling of GPU perf, significant battery life play and major I/O overhaul. Really the only thing that doesn't get a proper good touch up are the CPU cores themselves. Minor improvements there.

If anything Phoenix is probably more of a stopgap. Maybe a bit of a boost on the iGPU and it brings Zen 4 which is a very nice improvement, but the next big step up is really Zen 5 APUs.

The SerDes link being slow and power hungry is the reason why AMD has not been able to partition the GPU into a chiplet, and why APUs are still monolithic.

Once AMD solves this problem (or if Intel beats AMD to it, with Meteor Lake), that will be the end of mobile dGPUs for sure.

Apple is just operating on a different level all together, to be able to put everything on one monolithic die, and not suffer the consequences of low yields.

 

[andermans]

Rembrandt's not a stopgap, it's a good (roughly) doubling of GPU perf, significant battery life play and major I/O overhaul. Really the only thing that doesn't get a proper good touch up are the CPU cores themselves. Minor improvements there.

If anything Phoenix is probably more of a stopgap. Maybe a bit of a boost on the iGPU and it brings Zen 4 which is a very nice improvement, but the next big step up is really Zen 5 APUs.

I think Phoenix might be a good step forward on the GPU side actually, especially if it doubles the number of ALUs per WGP as is rumored for the desktop RDNA3 cards.

 

[LightningZ71]

What's all the hand wringing about? AMD has proved that they are more than capable of building something like this, both with Van Gogh in the steam deck and in the SOCs for the ps5 and the Xbox. At this point they are deliberately choosing not to do it.

 

[LightningZ71]

The SerDes link being slow and power hungry is the reason why AMD has not been able to partition the GPU into a chiplet, and why APUs are still monolithic.

Once AMD solves this problem (or if Intel beats AMD to it, with Meteor Lake), that will be the end of mobile dGPUs for sure.

Apple is just operating on a different level all together, to be able to put everything on one monolithic die, and not suffer the consequences of low yields.

There's not much of a need to partition the gpu into a Chiplet if you can stack a monster SRAM die on it and use it as infinity cache. But that's way off in the future and impossible for AMD to do right now (/s)

 

[soresu]

I think Phoenix might be a good step forward on the GPU side actually, especially if it doubles the number of ALUs per WGP as is rumored for the desktop RDNA3 cards.

Depends on what Phoenix actually is.

Initial rumours painted it as a Zen4/RDNA3 APU, but I saw something recently that had it as Zen4/RDNA2, so basically an ultra light mobile or embedded Raphael.

It does seem odd, but then we got Cezanne/Zen3 this year with Vega instead of RDNA or RDNA2 in its gfx salvo, so it's definitely possible that AMD could do this again, especially if Zen4 is cooking significantly ahead of RDNA3.

AMD already demonstrated a willingness to reuse gfx IP repeatedly from Raven Ridge -> Cezanne, so I do think that Phoenix being Zen4/RDNA2 is very possible.

That being said, if the GPU part of Raphael is indeed part of the rumoured 6nm IOD, and Phoenix is on 5nm, then Phoenix could actually be significantly more efficient than Raphael on the gfx side of things.

 

[leoneazzurro]

Using older architectures for iGPU has several reasons to be. First, it is a well established "building block": if the projects for the new graphical architecture and for the new APU start in parallel, it is unlikely using for the iGPU something new and unproven, whih does not even exist in reality. Second, power optimization:we have seen Vega going from power hungry monster in the form of Radeon VII to a sober mobile incarnation in Renoir and Cezanne, and this even going higher on clocks on the very same production process. This could have been done for RDNA , too, but it would have taken longer and quite probably the launch windows might have been not met. A third reason may be driver stability, which is important for corporate users, and these are one of the main targets of these APUs.

 

[NTMBK]

But now, 1 year after M1 was released on the market, it is not like expecting AMD to create a new PC configuration with memory integrated in MCM. M1 already validated this concept.

How much more validation does AMD need? Does AMD really need to wait until Intel releases it, losing any time to market advantage?

One year later, Apple is extending it by by scaling the memory channels (width) for even more bandwidth to serve a more powerful GPU.

AMD has RDNA2 in Rembrandt, hungry for bandwidth, likely with no V-Cache. Integrating DRAM would have elevated it up a segment.

...how long do you think it takes to produce a new SoC? Honestly- from the "oh, that is what we should build" moment, to the point where they actually put it on a store shelf? It's a lot longer than 1 year.

 

[Shivansps]

So now people realise that Renoir/Carrizo arent premium APUs huh?

Now, AMD can do the same? for sure, but why would they? Thats the big question here, Apple is placing a console killer level APU on a laptop and they can do it because they are Apple, they sell the entire ecosystem, and they dont care about killing their dGPU market because they dont have a dGPU market.

The hard true is that AMD has other things to worry about, first they are keeping the low end dGPU market alive, a thing like the navi 24 should not exist these days, that should be mainstream APU level.

The really good thing about this is that now there is no way AMD would ever dare to call RMB "a premium APU".

 

[uzzi38]

So now people realise that Renoir/Carrizo arent premium APUs huh?

Now, AMD can do the same? for sure, but why would they? Thats the big question here, Apple is placing a console killer level APU on a laptop and they can do it because they are Apple, they sell the entire ecosystem, and they dont care about killing their dGPU market because they dont have a dGPU market.

The hard true is that AMD has other things to worry about, first they are keeping the low end dGPU market alive, a thing like the navi 24 should not exist these days, that should be mainstream APU level.

The really good thing about this is that now there is no way AMD would ever dare to call RMB "a premium APU".

...

Would now be a bad time to say AMD even marketed Van Gogh as a "premium APU"?

 

[Shivansps]

...

Would now be a bad time to say AMD even marketed Van Gogh as a "premium APU"?

And thats why we needed what Apple did, they are now just making themselves look like fools.

 

[jpiniero]

And thats why we needed what Apple did, they are now just making themselves look like fools.

Worth reiterating that Apple is charging an extra $400 to go from the 10+16 to go to 10+32. Plus you are also required to go to 32 GB of memory which is an extra $400 over 16 GB. So the cheapest model with the 10+32 you can buy is $3199. You really cant compare to an AMD APU that goes into sub-1k machines.

 

[Shivansps]

Worth reiterating that Apple is charging an extra $400 to go from the 10+16 to go to 10+32. Plus you are also required to go to 32 GB of memory which is an extra $400 over 16 GB. So the cheapest model with the 10+32 you can buy is $3199. You really cant compare to an AMD APU that goes into sub-1k machines.

if Apple GPU perf claims about the 10+16 version are anywhere near to being correct, that alone makes all APUs, including the still several months away from release RMB to look like cheap toys.

 

[leoneazzurro]

Do you think everyone wants to spend 3K$ only to get "the fastest and most efficient APU available" when a cheaper X86 alternative with discrete graphics will have same/better performances at the cost of worse battery ? Especially when that platform misses the driving force of sales for most of the people(videogames)?

 

[BorisTheBlade82]

Will they be faster in MT? Of course. Even Cézanne can do this. Will they be faster in ST? Maybe. But at the time their competitor will be ADL and probably M2. So who knows? Will they be more efficient? Never. You can mark me on my words. This is simply not a realistic assumption.

 

[moinmoin]

that alone makes all APUs, including the still several months away from release RMB to look like cheap toys.

Somehow this reminds me of someone on this forum who wants APUs to be cheap. Can't recall his name though.

 

[Shivansps]

Somehow this reminds me of someone on this forum who wants APUs to be cheap. Can't recall his name though.

I dont remember either, but i do remember some people telling him that Renoir and Carrizo were premium APUs so it was ok for prices to go up.

 

[jpiniero]

I dont remember either, but i do remember some people telling him that Renoir and Carrizo were premium APUs so it was ok for prices to go up.

Trust me, if AMD does anything like what Apple did, it will be $700 or more.