Intel and AMD roadmap for 2021 and 2022 reveal interesting developments: Zen 4 Raphael to change status quo by featuring an RDNA2 iGPU

[Redaktion]

Leaked roadmaps point to a busy 2021 and 2022 for both Intel and AMD. In 2021, we can expect to see Intel offering Rocket Lake-S and Comet Lake-Refresh parts alongside Tiger Lake-H. This will be followed by corresponding Alder Lake parts in 2022. AMD will introduce Cezanne, Lucienne, and Van Gogh APUs next year. Raphael in 2022 will be a Zen 4 mainstream part with integrated RDNA2 graphics and will be complemented by Rembrandt, Barcelo-U, and Dragon Crest on the laptop side.

https://www.notebookcheck.net/Intel-and-AMD-roadmap-for-2021-and-2022-reveal-interesting-developments-Zen-4-Raphael-to-change-status-quo-by-featuring-an-RDNA2-iGPU.510997.0.html

[vertigo]

Xe Gen12? Did they just skip 2-11?

Was hoping RDNA2 would be implemented sooner, hopefully AMD moves it up a bit, but looks very interesting nonetheless. After years of barely any significant improvement in CPUs, things appear to finally be moving along.

[Mate]

So AMD is not switching to 5nm anytime soon?  It means Apple already won . Time to go back to Mac world..

[vertigo]

So AMD is not switching to 5nm anytime soon?  It means Apple already won . Time to go back to Mac world..

There's so little difference anymore with architecture size changes that it's not really that significant, and actual architecture often matters more. And any modern CPU is powerful enough that they're all going to be more than enough for most work and most people. So choosing/switching ecosystems simply based on the fact one has a 5nm CPU vs a 7 or 10nm one makes no sense.

[TruthIsThere]

So AMD is not switching to 5nm anytime soon?  It means Apple already won . Time to go back to Mac world..

There's so little difference anymore with architecture size changes that it's not really that significant, and actual architecture often matters more. And any modern CPU is powerful enough that they're all going to be more than enough for most work and most people. So choosing/switching ecosystems simply based on the fact one has a 5nm CPU vs a 7 or 10nm one makes no sense.

So, that's why Apple is buying up all the 5N and 3N because advanced architecture "makes no sense?"

Stop sniffing that glue, bud.

[xpclient]

If RDNA2 is going to be this late to AMD APUs, then Intel now has more updated graphics. Unless AMD laptops are paired with Nvidia GPUs and get Thunderbolt, the choice still remains Intel for me. Cezanne will probably only bring PCIe 4.0 and a CPU performance boost.

[Padmakara]

Well, looks like AMD will move to 5N when apple will be on 3N. AMD should force TSMC to build a factory for them, otherwise they will always be second tier. And you never know when Intel wakes up, like 14 years ago when they destroyed AMD with core2 duo.
This is a shame, you can't find new AMD gen cpus, gpus or laptops outside of US, because TSMC doesn't have the capacity... So AMD should be smart and make TSMC to build more for them

[Mate]

#6 "This is a shame, you can't find new AMD gen cpus, gpus or laptops outside of US" - let me correct you - you cant find new AMD products outside China. Probably more than 50% Renoirs are in chinese homes and offices

#3 "There's so little difference anymore with architecture size changes that it's not really that significant," - you are joking , right? AMD doubled performance and improved battery  when they jumped  from 12nm to 7nm. Difference between 7 and 5 is not that big, still significant. 30% lower energy consumption is huge. Also chip would be 2x smaller so production costs should be lower or you can just add few cores.  Now manufacturing cost matters a lot, thats actually one of major reasons why M1 is that powerful and can be cooled off even without fan.

[_MT_]

Xe Gen12? Did they just skip 2-11?

Obviously, they're referring to the processor generation and use it as iGPU generation (like they improve the iGPU in every generation). It was the same with 10th and 11th generation.

"There's so little difference anymore with architecture size changes that it's not really that significant, and actual architecture often matters more."
At the very least, it changes transistor density. You can fit more in the same die. Or you can make more processors from a single wafer. Assuming similar yields, it would significantly improve supply for the same wafer allocation. Something AMD could really use. Availability of many of their products is simply laughable.

[_MT_]

Difference between 7 and 5 is not that big, still significant. 30% lower energy consumption is huge. Also chip would be 2x smaller so production costs should be lower or you can just add few cores.  Now manufacturing cost matters a lot, thats actually one of major reasons why M1 is that powerful and can be cooled off even without fan.

A significant chunk of it is the different architecture. M1 has a relatively low maximum frequency. It compensates by processing more instructions simultaneously (this is referred to as width). And this is something where ARM has an advantage as it's easier to implement (you can do things that would be prohibitively expensive if at all possible on x86). It's pretty clear now that Apple has been planning this for a long time and has designed their mobile processors to be capable of scaling to desktop (after all, they referred to the design as desktop class back when they released the first generation IIRC). They just needed to catch up in performance. And they tripled it in five years. That's an astonishing pace. Even if AMD was currently on the 5nm node, they wouldn't be able to match M1's efficiency.

[Brollan]

"Comet Lake-Refresh parts alongside Tiger Lake-H"
"Enhanced SuperFin architecture"

Oh look, it's a yet another news editor with poor knowledge of the terminology. Where do they find them?

[neblogai]

If RDNA2 is going to be this late to AMD APUs, then Intel now has more updated graphics. Unless AMD laptops are paired with Nvidia GPUs and get Thunderbolt, the choice still remains Intel for me. Cezanne will probably only bring PCIe 4.0 and a CPU performance boost.

RDNA2 is coming to Van Gogh- a competitor to Intel's UP4 APUs, and it will likely trounce it in games in ultraportable/fanless. For gaming laptops- H-series will use dGPU, U-series can use dGPUs like MX350/450 too. And that iGPU in TigerLake is nothing special- Intel was showing wonders their slides, but in real life laptop tests- even current Renoir with Vega iGPU goes toe to toe with them. Lucienne/Cezanne, with higher enabled clocks and numbers of CUs, is likely to be somewhat faster.
The only real area where Intel leads, is Thunderbolt. Cezanne is still PCIe 3.0, and no thunderbolt or USB4. So for external dGPUs- Intel is the only option.

[JayN]

TGL-U memory controller already has support for LPDDR5, but the current round of laptops are using LPDDR4x. Any reason we wouldn't see a refresh of TGL-U laptops with LPDDR5 in 2021?

[neblogai]

Intel representative already said on Tiger Lake launch (on High Pro Tech channel), that laptops with LPDDR5 are not coming on launch, but in 2021. Not sure why it is so.

[vertigo]

There's so little difference anymore with architecture size changes that it's not really that significant, and actual architecture often matters more. And any modern CPU is powerful enough that they're all going to be more than enough for most work and most people. So choosing/switching ecosystems simply based on the fact one has a 5nm CPU vs a 7 or 10nm one makes no sense.

So, that's why Apple is buying up all the 5N and 3N because advanced architecture "makes no sense?"

Stop sniffing that glue, bud.

#3 "There's so little difference anymore with architecture size changes that it's not really that significant," - you are joking , right? AMD doubled performance and improved battery  when they jumped  from 12nm to 7nm. Difference between 7 and 5 is not that big, still significant. 30% lower energy consumption is huge. Also chip would be 2x smaller so production costs should be lower or you can just add few cores.  Now manufacturing cost matters a lot, thats actually one of major reasons why M1 is that powerful and can be cooled off even without fan.

"There's so little difference anymore with architecture size changes that it's not really that significant, and actual architecture often matters more."
At the very least, it changes transistor density. You can fit more in the same die. Or you can make more processors from a single wafer. Assuming similar yields, it would significantly improve supply for the same wafer allocation. Something AMD could really use. Availability of many of their products is simply laughable.

I didn't say there wasn't a difference, just not significant, certainly not enough to switch over to an entirely different ecosystem for it. As architecture size gets smaller, there's less difference between steps, which results in diminishing returns. As _MT_ pointed out, and as I mentioned, a big part of the improvements are with the architecture itself. Yes, a 5nm chip will be better than a 7nm one with the same architecture, but a 7nm chip with a better architecture can beat a 5nm one with a worse one. Size matters here, but it's not the only factor. I was simply stating that, and that it's not a good reason to switch from Windows or Linux to Mac.

And of course Apple is going to buy up 5nm and 3nm production, because they need it, simple as that. They're currently producing chips at that level, so they need the production to do so, and they're closer to 3nm than anyone else, so of course they're going to buy production for that, to work on moving to it. As the others will, or should, be buying production in their next step(s).

Smaller is of course better, and they should all be working on achieving that, but they should focus just as much, if not more, on improving their architecture itself. It's getting increasingly harder to go smaller, with increasingly smaller payoff, and while doing so can result in better yields, at first they will almost certainly be worse, until the process is mastered. Keep in mind, too, that during the past several years of Intel's tick-tock cycles, there was typically much more improvement when they changed the architecture itself whereas the improvement from a die shrink was less pronounced.