Though I might add, 65 W on an AMD product means means as much as all the +es on 14 nm+++ Intel CPUs. We only know if it's a direct comparison or not when it comes out and somebody tests it for power consumption and heat.
Well, 65W means 88W PPT (at least in all SKUs up until now), so it's likely this is the same. But it doesn't act the same way as Intel's chips do, that's for sure. Much more aggressive and opportunistic boost, and this is sustained indefinitely (balanced by thermals) without ever falling back to a PL1-like power limit.
I’m a little disappointed that AMD does not have an IGP on all its consumer-facing products. Even if it’s just a base unit to drive displays, it seems like they are missing out here. Now that it’s near impossible to buy a GPU at a decent price, how long before it cuts into CPU sales because people need something to drive the display? I know the G-line is monolithic, so I’m sure that brings about challenges, but I would think they could have added GPU functionality in there with the chiplet concept. Hard to believe we arrived at a time where most consumer Intel CPUs have an IGP, while most AMD CPUs do not.
This has actually been the default for quite a while. Intel introduced ubiquitous iGPUs with ... Sandy Bridge, IIRC? AMD FX never had an iGPU, and that was the precursor to Ryzen CPUs. Of course AMD realized ca. 2015 that updating FX further was useless and a waste of their very limited R&D money at the time, and instead focused on A-series APUs for several years, which had great iGPUs for their time. But Ryzen was as such a return to normal for AMD in the CPU space.
Chiplet-based APUs have been a dream of many since the first chiplet CPUs, but they still seem a ways out. Either they'd need to put the iGPU in the IOD, which would make it
huge (it's already quite sizable, after all, and adding at least 3 CUs plus all the media blocks, display output blocks and the like would need some space), and would require it to be made on a cutting-edge node rather than GloFo 12nm like today. The other alternative is to put a GPU die (GCD?) in place of the second CCD, though that would require AMD to have a suitably sized GPU die to use - which they don't, and haven't for quite a few years. Lexa/RX 550 at 110mm² is still a bit too large, and is way too old at this point. Navi 23 at 236mm² is far too large to sit in place of a ~80mm² CCD. There would also be issues with memory bandwidth holding back a larger iGPU, etc., though those can be overcome with very fast DDR4 to some extent. The revised WSA recently seems to indicate that AMD wants to move to 7nm IODs though, which might make room for the inclusion of a small iGPU across the board. I would certainly welcome this - if nothing else, it would be great for troubleshooting without a dGPU. DDR5 will also be of great benefit to future iGPUs of course, so perhaps we'll actually see a proper chiplet APU on AM5? I'm hopeful.
Yeah, I think Zen was a huge pivot that really brought AMD back from futility, so I guess they made the right decisions. Now that they are basically a significant player in the semiconductor market, I’d love to see them bring back the APU for most of their lineup again.
Just looking at their presentation today, AMD is putting a lot of thought in the mobile (notebook) gaming segment. To me, this seems ripe for them making APUs like they have in the PS5 and XSX, but without the budget constraints. Making an entire class of APUs to go into gaming laptops priced in the 1000-1500 range would be a big hit.
That would be interesting, but OEMs are unlikely to bite - remember how few OEMs made use of KBL-G? I think the main reason is cooling - cooling a separate CPU and GPU in a notebook is always going to be easier than the increased thermal density of a high powered (100W+) APU. Keeping the APU cool might be fine, but it would inevitably lead to a significant hot spot in the chassis. Laying out the PCB might also be difficult, as you'd need a very large VRM setup for the APU, and possibly a semi-exotic RAM layout (256-bit (LP)DDR5?). Still, it could make for some cool designs if they made it and helped OEMs engineer products around it. I think a more realistic APU approach would be gaming APUs for the >$1000 range though, targeting mid-to-high 1080p gaming. Much easier to implement, much easier to cool, more flexible in terms of form factor, and maybe even usable as a socketed desktop product
