Bingo. The Steam Deck OLED’s SoC seems to have cut out the Cadence DSPs, which we can see with the significantly smaller SoC (especially when combined with the new process node).

Pretty sure VGH also has more PCIe lanes on die as well, alongside full size Zen 2 cores (all console variants have a cut FPU).

Same peak, lower average is what I’ve understood so far. At least as far as Valve has described it to tech press.

How it fits into your report I’ll leave to you so I won’t answer your question directly, but I’ll give you some context on what the acquisition is for.

So when you look at product markets, Xilinx and AMD are very different. They’re both focused on totally different markets. However, there’s a lot of technical expertise at Xilinx and AMD that both companies can use - there’s a lot of overlap.

Firstly: engineering talent. While Xilinx specialise in FPGAs, they are very leading edge when it comes to working with advanced packaging techniques, which we’ve also seen AMD take advantage of significantly (X3D, MI300). So there’s an overlap here which both companies can take advantage of.

Second up: product IP and software development. You’ve probably recently heard of Ryzen AI, right? That block on the newest mobile Ryzen APUs called the AI Engine is developed by Xilinx - or well it’s their IP I mean. There are instances like this where the two companies can utilise the others IP to improve upon their own products as well.

Probably because it hits a higher frame rate.

The higher you go, the more CPU you need, and the more CPU you need, the less power and memory bandwidth you have for the iGPU.

That’s not an issue, that’s intentional to reduce artifacting. Unlike FSR3 and DLSS3, AFMF doesn’t have access to motion vectors to try and preserve image quality.

This is the fourth preview driver now. First was the initial one, second added RDNA2 support, third drastically improved frametime consistency and this latest one supposedly just improves driver stability (it’s likely just rebased on a more recent 23.30 branch, which is yet to go mainstream AFAIK).

Frametime consistency was fixed with the last driver already.

None of the better YT GPU reviewers seem to have looked at that improved release though, so nobody’s really judged it properly.

It also exists to clear out any more stock of the OG parts. While the Deck OLED looks mostly the same, most of the internals were reworked/refined. Even down to the joystick PCBs.

Apple’s E-cores were never useless, they’re easily best in class by a large margin. They’ve got the best perf/W in the industry by a country mile really, the things sip power, and while they’re not the fastest little cores, they are still extremely potent. I don’t see them changing away from that core structure any time soon.

As for the GPU, idk off the top of my head, but the IP is likely similar to A17. I wouldn’t expect much - the main advantage is the addition of hardware RT support, but from what we’ve seen the RT capabilities aren’t hugely improved over shaders. Definitely going to be a more modest improvement than prior iterations here.