Larrabee - Update

[drunkenmaster]

You missed the point of what I was saying... even if they make an absolutely superb piece of hardware - and being intel I'm pretty sure the will in the end - they absolutely won't make it in the gaming GPU market... intel simply lack the mentality to make it work and no amount of R&D, money or time will fix that. They should stick to what they are good at like CPUs.

You said Larabee is dead, thats 100% inaccurate, I also didn't miss what you said.

You keep saying they don't have the mentality, and thats just the daftest thing I've ever heard. They have the money to kill the current Larabee, buy the entire developement team off Nvidia and have them build them a AMD/IBM/PS3/Arm style core, whatever they want, if they want.

You've decided that Larabee isn't what YOU think it should be and are having a go at Intel.

In all likelyhood Intel will end up incredibly competitive in high end gaming within 3-4 years but their first goal as a intergrated GPU thats absolutely 100% usable as extra FPU processing power as standard for ANY app that wants it.

THey have exactly the right mentality about it because their FIRST move is to make an on die gpu that can be FULLY utilised as part of the CPU first and foremost, their current design will be incredibly optimised for just that purpose, making a larger derivitive with more cores and more functionality is simply the next step and I have literally no doubt Intel can do it, because at worst, they can afford to through out a several billion pound project, start again or make secondary performance GPU project as and when they like.

Larabee, as it is intended to be, is much closer to a x86 core, offering easier programming and more gpgpu like structure, that is the exact way Nvidia is going, just on a much more optimised platform with literally decades of experience, it was never going to happen overnight. The only mistake Intel made with Larabee is giving the impression it might happen overnight to people who really don't get its going to take years.

Current gpu's are moving towards a gpu-gpgpu-cpu cross, and Larabee is just moving into the middle from the otherside, it will still end up in the same place.

AMD/Nvidia will be pushing for better compability with general processing usage, being natively c+ programable, being more accessible for normal computing, and Intel are simply and understandably starting from the other end, and moving the other way, to add more acceleration and graphics appropriate hardware into it.

You seem to miss that Intel has FAR more experience with normal software and has a huge advantage in terms of being able to converge its own cpu/gpu plans together as IT see's fit, in a way Nvidia simply can't and also move and design a plan that will end up with their cpu/gpu encased in the same package and eventually in a way that programs can be compiled as normal but take advantage of the on die gpu.

Nvidia will literally NEVER have that level of intergration and will never be able to compete large scale on GPGPU in the future.

Why, because in 4 years at 16-32 cores there will be multiple types, a octo cpu with 24 gpu cores that can be seemlessly coded to use all the power easily.

EDIT:- Its also worth noting Nvidia being bought by anyone is incredibly unlikely, they aren't making good mobile products, with all their design wins for Tegra(most of them) actually ending up in production with other cores as they miss targeted specs constantly, its a fair point, how many Tegra 1 design wins were there, and how many products, loads of "design wins" smeg all products. Their GPU line is, well, on course for a catastrophic failure without a massive rethink, their CEO is pushing the desing for one goal and its contrary to the manufacturing industry, for 3 years production has been getting worse with no sign of a change from Nvidia.

They are an idea's company, nothing more or less anymore, if their idea's fail to the tune that they'd be worth buying in terms of price, the value of the company without a decent gpu design, will have simply evapourated. The only value they'd have is some very good engineers, who can simply be lured away with jobs and cash. The entire value of the company is its design team, you don't need to buy the company, just hire the team. If they made their own gpu's and couple provide an entire supply of a new design and had the process tech to go with it, they'd be worth billions, they aren't.

 

[Rroff]

You said Larabee is dead, thats 100% inaccurate, I also didn't miss what you said.

You keep saying they don't have the mentality, and thats just the daftest thing I've ever heard. They have the money to kill the current Larabee, buy the entire developement team off Nvidia and have them build them a AMD/IBM/PS3/Arm style core, whatever they want, if they want.

You've decided that Larabee isn't what YOU think it should be and are having a go at Intel.

Larrabee as a gaming GPU is dead, intel might not know it yet. They can kill the project off and restart it as many times as they like but the whole approach thats inherent to intel is wrong and until that changes it will not succeed in that market. Look at their products, they make a solid no frills bit of hardware, with adequate no frills, get it right first time, software support and it never moves on, everything else has to work around it. This simply won't work in the gaming market. Works great for servers, embedded devices, CPUs, storage devices, etc. infact is absolutely whats required in those markets which is why intel has done so well there.

Don't get me wrong, I hope intel can get their heads around this we badly need a 3rd player in the market, I just don't see it happening without a huge shift in focus from intel.

 

[jakspyder]

only time will tell. i for one hope intel bring our yet another fantastic product. and i hope it takes the GPU market by storm forcing both AMD and Nvidia to really pull the finger out to compete with intel. I know im being hopeful but we can have our dreams

either way i have no doubt that eventually, as has been said, intel will bring out a great product and i also believe they will eventually make it big time in the GPU area. It just wont happen next week.

 

[D.P.]

There was no gtx480 around when larabee was being targeted for release so no need to mention it in the article.

There was no 5870 then either.

More dog turd from the master fanboy. How pathetic charlie is.

 

[iviv]

There was no 5870 then either.

More dog turd from the master fanboy. How pathetic charlie is.

How I see it:

Larrabee 1, aka the part you have heard of, was beset by delays. By the time it was deprecated, it was well over a year and a half late. The first official release date to come our way was mid to late 2008.

The B0 stepping did fix the vast majority of the problems, but it didn't come back until late 2009. This pushed the release date out to Q2, basically now. For a chip that was targeting the HD4870/GTX285 generation of GPUs, a year's delay meant that it would be facing the HD5870/GTX285, one of them moving the bar up substantially.

It was designed to be released mid to late 2008. Instead, it was pushed back a year, up to mid to late 2009. The 5870 was released September 23rd. Which appears to be mid to late 2009 to me.

However, the ambiguity does come in where he says the release date being pushed back to Q2, since Q2 of 2009 wouldn't make sense, its too early, and Q2 2010 does mean competing with the 480.

 

[RizlaKing]

It's only on the software driver side that intel need to come upto speed really they are pretty lousy there but nothing i don't think they can change. Larabee isn't dead of course it isn't intel know cpu+gpu is the future and they will be working hard on getting a competitive gpu ready. Will they come into the gaming end of the market i don't think so at the start but eventually they will end up there with whatever changes need to be made in terms of mentality or approach as gaming is the place to showcase your progress and intel as well as everyone else is well aware of that.

While a third player would definately be good i don't see 3 players being around very long if intel do hit the gpu section fully commited i can see nvidia being squeezed out and eventually bought by intel bringing us back to a two player situation. I just cannot see long term how nvidia can compete with amd and intel unless they manage to develop a kick ass cpu of their own which i think is about as likely as george bush being voted humantarian of the year.