So this first generation won't be for desktop at all? Is this what i should read in the presentation?

asturur:

So this first generation won't be for desktop at all? Is this what i should read in the presentation?

I think we're still going to see desktop hardware based on the first-gen architecture, but server models are going to be released first.

I think it says a bit about what Intel's focus is - it's about winning HPC and everything else comes second to that. That's probably also true of Nvidia to some extent, but they have the advantage of a long history of gaming gpu's so it's not hard to keep that ticking over. I can't see Intel coming at this cold having much hope with gaming if they aren't 100% focused on achieving that over everything else.

Dribble: that is true. Another thing is lots of Nvidia's developments for their HPC-focused cards, the ones with no video outputs, can trickle down to their video cards. For Intel, they don't have this cross-pollination going on. They have a long way to go and just a few very large customers are paying the bill.

asturur:

So this first generation won't be for desktop at all? Is this what i should read in the presentation?

Read: High Performance

Show me Intel's gaming partners, otherwise I'm not even going to take them seriously.

I dont know if anyone one on the market needing this, to wait for Intel to come up with something. Too much already covered by gpu makers.

fry178:

I dont know if anyone one on the market needing this, to wait for Intel to come up with something. Too much already covered by gpu makers.

What do you mean? Having just AMD and Nvidia for discrete GPUs just isn't enough. There needs to be more competition. Nvidia is way too expensive and although AMD is fantastic for compute, their drivers aren't great and they don't have CUDA support (which most servers seem to use). I suspect Intel [at first] is going to be substantially cheaper than Nvidia while having the money to really push developers to use OpenCL and SPIR-V. Intel is pretty good at making drivers too. So, if Intel plays their cards right (and let's face it, they need to in order for this to succeed), they'll be filling a nice gap in the market that AMD can't provide and Nvidia won't. When it comes to gaming, things will likely be a little bit different. I think Intel is still going to be priced between AMD and Nvidia, but they've got a long way to go to get application-specific optimization profiles set up.

angelgraves13:

Intel can post all the slides of 7nm they want. They can’t even get 10nm working. What a joke.

I'm still convinced Intel's 10nm has been perfectly fine for a while, but the problem is it can't handle the high clock rates that desktop users and enthusiasts are expecting of their aging architecture. People want a new CPU that's actually an upgrade, and a 10nm i7 with a much lower clock speed is going to be noticeably slower. More efficient, but slower. That's why most of the 10nm products available are mobile and server chips, where nobody expects base clocks beyond 4GHz. That being said, GPUs normally don't go much higher than 2GHz and I doubt Intel's will be any different. So, I wouldn't be surprised if they will actually achieve 7nm.

if clockspeeds are the problem, then Intel have forgotten the things they have learned before.

@schmidtbag Meant for commercial stuff Those that would have the need for a relevant amount (as in sales), already have something as no one will stop and wait for intel to release something, as they most likely will lose more money delaying a project, than the higher cost Nv/Amd will have

Stormyandcold:

Show me Intel's gaming partners, otherwise I'm not even going to take them seriously.

Being that the first release of Xe is for HPC/AI....there would be no "gaming partners" involved. The first Xe based products are going into "Supercomputers".... Not everything revolves around gaming....

sykozis:

The first Xe based products are going into "Supercomputers"

You know when you've been around too long and start stating the obvious? Then people look at you like "err, wtf are you on?".

angelgraves13:

Intel is going to be like Supermicro if they keep focusing only on servers and HPC.

Intel isn't focusing only on servers and HPC. Xe is still expected to be released as a consumer graphics card. They are just taking the most logical approach. Start with HPC and scale down.

angelgraves13:

I'll believe it when I see it. We'll probably see AMD's RDNA2 with ray tracing before we see a GPU by Intel.

Given past history, I have my own doubts as to Intel actually releasing a consumer graphics card. Rumors put launch in summer 2020 though. Given how much hype was built up around Larrabee and claims from Intel in regards to how advanced it would be....and it's final launch as an HPC only card.... I'm reserving my excitement for when Hilbert can post an actual review.

The problem with Larrabee was that the performance previews were so-so, but, within a very short time it became totally out-dated. Intel knew there'd be no consumer market for it so it was scrapped. My gut feeling is that if Intel actually had something they'd be screaming about it and hyping it up right now. The fact they aren't doing that gives me zero confidence. The Xe might as well not exist.

There will be desktop too. Initial slide shows Entry Mainstream Performance Server

Stormyandcold:

The problem with Larrabee was that the performance previews were so-so, but, within a very short time it became totally out-dated. Intel knew there'd be no consumer market for it so it was scrapped. My gut feeling is that if Intel actually had something they'd be screaming about it and hyping it up right now. The fact they aren't doing that gives me zero confidence. The Xe might as well not exist.

The problem with Larrabee was that it was never a graphics architecture to start with. It was based on P54C and was overly software dependent. Given that Intel has already announced sales of Xe to both the Argonne National Laboratory and the US Department of Energy, I think it's safe to say that Xe actually exists. Intel also announced that the first release of Xe would target HPC and not consumers. From a marketing standpoint, it makes no sense to "hype up" a product that isn't expected to launch for 6 months or more, when you're competing against companies that are capable of responding as quickly as NVidia is. You want competitors to have as little time to react before a product launch as possible. Especially if you're intending to disrupt a highly competitive market. The best thing for Intel to do, is keep reminding people that Xe is coming, and let the product speak for itself by sending it out to as many reviewers as possible. There was a lot of hype built up for Larrabee....and in the end, it never launched. A lot of people are going to be skeptical of Intel because of that. Hyping up Xe, will only increase that skepticism. If Intel starts to hype up Xe, and fails to live up to the hype, Xe will be dead on arrival.

I see Intel coming out with a 10nm gpu platform for the masses. Don't look for their desktop gpu until 2021 I believe.

NewTRUMP Order:

I see Intel coming out with a 10nm gpu platform for the masses. Don't look for their desktop gpu until 2021 I believe.

Current rumor is summer-2020.... A 2021 release wouldn't be a surprise though. I just want to see Intel actually release a product this time.