Jen-Hsun: "We've already built Larrabee 2 years ago"

[AirRaid]

This is from 2008:
http://www.youtube.com/watch?v=_1SEko20vMM

[Ninja Prime]

Hes as bad as or worse than the idiots at the top of sony sometimes...

[CaptNKILL]

LOL

Wow, he's a sarcastic one isn't he...

[pakotlar]

Hes as bad as or worse than the idiots at the top of sony sometimes...

Except he's right this time. Intel is positioning Larrabee as some sort of revolution for GPU/GPGPU, but it really won't be, and no one in the field expects it to be. It's just their GPU entry. Like Jen said, if it was such a revolution, they would've waited until it was ready then blown the competition away. They're revealing it 3 years ahead of time because they know its the same old thing, done in a new, but ultimately roundabout way because of the limits of their expertise with developing fast arrays of DSP's. They're trying to build hype for a product that will quite literally offer nothing new to the end user, take up more die space, and put out more heat.

It's impressive that Intel can even put out a competitive GPU, but thinking that it'll offer something over GT300 or whatever is out at the time (2nd iteration GT300 most likely) is wrong. It'll have some interesting corner case benefits, for those devs who write programs explicetly for the processor, using LRBNI's. Anything that treats Larrabee as a DX11 GPU will give you results that are nothing special compared to the competition.

I certainly don't see him as a hyperbolic worried mess, which is the image most Sony execs since Kutaragi have impressed upon us w/ regard to PS3.

[AirRaid]

Of course, Larrabee will not sit still with the first release, expected to be 32 cores.
It's expected that 48 and 64 core versions of Larrabee 1 will follow in 2010, if not early 2011.

Intel is already working on two generations beyond the Larrabee 1 family: LRB2 & LRB3 with LRB3 supposedly a totally new architecture.

It's no different than Nvidia: I remember clearly Nvidia top brass, perhaps Jen-Hsun himself, stated they were working on NV40 & NV50 (NV50 became G80) at the NV30 / GeForce FX launch event in fall 2002.

There's no doubt that Larrabee and its decendants will make the coming decade very interesting for computer graphics and GPU compute/GPGPU.

[bacon12]

Of course, Larrabee will not sit still with the first release, expected to be 32 cores.
It's expected that 48 and 64 core versions of Larrabee 1 will follow in 2010, if not early 2011.

Intel is already working on two generations beyond the Larrabee 1 family: LRB2 & LRB3 with LRB3 supposedly a totally new architecture.

It's no different than Nvidia: I remember clearly Nvidia top brass, perhaps Jen-Hsun himself, stated they were working on NV40 & NV50 (NV50 became G80) at the NV30 / GeForce FX launch event in fall 2002.

There's no doubt that Larrabee and its decendants will make the coming decade very interesting for computer graphics and GPU compute/GPGPU.

No they won't its just completing the circle for intel now, they have already muscled GMA on so many boards and it was horrible. They had to improve, I don't think its gonna be ground breaking now or five years from now.

[Intel17]

Big companies like Intel tend to learn very quickly. Even if Larrabee isn't a ground-breaker performance wise this time, Intel's next gfx effort will be a huge stride over it, and the effort after that even moreso until they've dominated the market.

Intel is simply able to pour gobs of R&D money into their products, especially if they plan to expand to make it a permanent part of their core business. Make no mistake -- Intel's got the best scientists and engineers in the industry. And they're still doing well in this lousy economy.

[ChrisRay]

To believe Intels Larrabee will be an automatic failure at this point is wishful thinking by some. Or just irrational speculation. Nvidia did build a GPU Compute hardware a long time ago. And they are definately pushing forward with this business model. How good or how bad it will be is really yet to be determined. To suggest that intel is building something revolutionary. Thats an entirely different matter. And they really havent. They are building a GPU based on their own engineering expertise. That will also be used for GPU Computing. Anything thats good for GPU computing is something I'm happy to see happen.

[Blacklash]

I am wondering if Intel will sink the money into having a solid driver team.

If they lag in that department and don't work with game devs at all it will definitely hurt them.

Guess we will see.

[noko]

Well if Intel can't get their GPU off then what the chances of them buying out Nvidia?

[Heinz68]

I think none.

I don't think they care much to have Jen-Hsun Huang on their team, so it would have to be very expensive hostile takeover.