Intel Architecture Day 2021 & The Linux State

[phoronix]

Phoronix: Intel Architecture Day 2021 & The Linux State

Intel this week hosted a virtual Architecture Day where they talked up their latest efforts from Alder Lake and Sapphire Rapids to their next-generation discrete graphics capabilities as well as other new offerings around IPUs and more. Here are the highlights from Intel Architecture Day 2021 and with a particular focus from our Linux angle.

https://www.phoronix.com/vr.php?view=30468

[spykes]

In my opinion, it is a drawback that P-cores in Alder Lake desktop CPUs have AVX-512 units disabled.

I think AVX-512 is a waste of silicon space on desktop CPUs, AVX2 covers all major use cases on a desktop and I would prefer to have more CPU cores than having AVX-512.

[avem]

The most important announcement is probably XeSS : vendor-neutral AI upscaler using motion vectors which promises to rival DLSS 2.0.

Too bad Michael hasn't paid attention to it.

FSR can be safely buried.

[Setif]

The most important announcement is probably XeSS : vendor-neutral AI upscaler using motion vectors which promises to rival DLSS 2.0.

Too bad Michael hasn't paid attention to it.

FSR can be safely buried.

Too bad, you haven't read the full article (page5):

One of the new bits on the Intel graphics software front is the company working on Xe Super Sampling as their alternative to NVIDIA DLSS and AMD FidelityFX Super Resolution. Intel will be releasing a XeSS (Xe Super Sampling) SDK this month so will be interesting then to learn more about it and the Linux support.

[numacross]

I would prefer Intel to release two versions of desktop CPUs: with and without AVX-512.

Haven't they been doing that already? Skylake on desktop didn't support AVX-512, but the HEDT offerings based on Skylake-X did.

The interesting part is they supported it for both desktop (Rocket Lake) and laptop (Tiger Lake) in the current generation, and now it'll be disabled on both?

[avem]

Too bad, you haven't read the full article (page5):

This deserves a separate article, not a fifth page most people will never read because the first four are dull and barely interesting.

Your attempt at being sarcastic and smart has failed spectacularly.

[aufkrawall]

The most important announcement is probably XeSS : vendor-neutral AI upscaler using motion vectors which promises to rival DLSS 2.0.

Too bad Michael hasn't paid attention to it.

FSR can be safely buried.

And hopefully DLSS too, if XeSS really can compete and Nvidia doesn't release a stubstantially better version. Though it's weird that AMD, Microsoft and Sony all have failed so hard, and of all vendors, it will actually be Intel who seem to be industry's open source white knight (I don't count FSR, it's a too primitive attempt).
Who knows, maybe they will also implement a mining brake, unlike AMD who don't care if they sell to miners or gamers/consumers...

[avem]

And hopefully DLSS too, if XeSS really can compete and Nvidia doesn't release a stubstantially better version. Though it's weird that AMD, Microsoft and Sony all have failed so hard, and of all vendors, it will actually be Intel who seem to be industry's open source white knight (I don't count FSR, it's a too primitive attempt).
Who knows, maybe they will also implement a mining brake, unlike AMD who don't care if they sell to miners or gamers/consumers...

A vendor-neutral solution is of course better but we'll see how it competes with DLSS. And more importantly Intel must have game developers on board - this upscaler just like DLSS mustn't be straightforward to implement. FRS on the other hand is basically a simple shader you can insert into your pipeline with next to zero effort.

[jaxa]

The Gracemont cores reaching +40% performance over Skylake at the same power usage is a big highlight. It helps explain how a Core i9-12900K could outperform a Ryzen 9 5950X, while using a lot of energy. Those results probably use DDR5 while Skylake is stuck with DDR4.

Edit: They are comparing 1C1T for both while the same Skylake core is capable of 1C2T, for another 30% or so multi-threaded performance. Still a good result.

[bridgman]

it's weird that AMD, Microsoft and Sony all have failed so hard, and of all vendors, it will actually be Intel who seem to be industry's open source white knight

I don't think that is fair - our slides were at least as good as Intel's slides