Final Word on Larrabee

[Viditor]

At a group panel discussion with the Intel Fellows, essentially some of the smartest people at Intel, Intel's head graphics honcho was quizzed about Larrabee.

When asked if they ever expected to see a Larrabee-based graphics part coming out at all, the entire panel looked directly at Piazza, as he hunkered down in his stool.

"I honestly thought i'd get through two days without someone asking me that..." he said, followed by a simple, "I don't think so."

http://www.techradar.com/news/compu...arrabee-was-impractical--716960#ixzz0zv6eMlEq

 

[happy medium]

Oh no Wreckage was right again.......

I personally think they will just buy Nvidia at some point and totally dominate the cpu/gpu on chip market.

 

[tweakboy]

What is Larrabee ?

Well Intel has the Sandy Bridge GPU/CPU chip no ?

Coming next year...gl

 

[Viditor]

Oh no Wreckage was right again.......

I personally think they will just buy Nvidia at some point and totally dominate the cpu/gpu on chip market.

2 problems with that theory...

1. Nvidia would cost far more than it's worth
2. The current market share for graphics (+/-5%) is Nvidia=25%, ATI (AMD)=25%, Intel=50%...yup, those onboard graphics sure do sell ALOT. I don't think the SEC would allow the sale to Intel.

BTW, you should read through Piazza's comments...some interesting points on WHY Larrabee failed.

 

[railven]

BTW, you should read through Piazza's comments...some interesting points on WHY Larrabee failed.

Do you think they'd ever revisit the subject and focus on a balance of programability and set functions?

Or would that require paying a royalty to someone they'd rather not?

 

[Viditor]

Do you think they'd ever revisit the subject and focus on a balance of programability and set functions?

Or would that require paying a royalty to someone they'd rather not?

I think what Piazza is saying is that the concept is not worth the effort.

 

[Anarchist420]

It will be worth it once a small enough manufacturing process comes out.

 

[IntelUser2000]

It would be real ironic if Intel cancelled Larrabbee for gaming because Sandy Bridge's graphics was close.

 

[Viditor]

It will be worth it once a small enough manufacturing process comes out.

I don't think so...remember that Larrabee was a lot of cores designed to use software for all of the graphic functions instead of hardware. The problem wasn't that the hardware didn't funtion properly, but that using software 100% (or even close) proved to be a futile experiment.


When asked why Piazza thought it had failed though he was surprisingly candid. "I just think it's impractical to try to do all the functions in software in view of all the software complexity," he explained. "And we ran into a performance per watt issue trying to do these things."

Even with a smaller process, the performance/watt will still be an issue

 

[Viditor]

It would be real ironic if Intel cancelled Larrabbee for gaming because Sandy Bridge's graphics was close.

LOL
That would be funny, but it's up there with winning back to back Lottos in probability...

 

[soccerballtux]

2 problems with that theory...

1. Nvidia would cost far more than it's worth
2. The current market share for graphics (+/-5%) is Nvidia=25%, ATI (AMD)=25%, Intel=50%...yup, those onboard graphics sure do sell ALOT. I don't think the SEC would allow the sale to Intel.

BTW, you should read through Piazza's comments...some interesting points on WHY Larrabee failed.

Hm. That's a very interesting point. Even though those onboards are practically a completely different market from Nvidia's specialty. I guess it would depend which direction Intel wanted to take it. If they copied over (well, I guess of course they would) all the IP to their integrated side of things, then yeah, SEC would be interested. If they simply applied their manufacturing/process knowledge and capabilities to Nvidia's cards, that would be very interesting.

 

[Idontcare]

I don't think so...remember that Larrabee was a lot of cores designed to use software for all of the graphic functions instead of hardware. The problem wasn't that the hardware didn't funtion properly, but that using software 100% (or even close) proved to be a futile experiment.


When asked why Piazza thought it had failed though he was surprisingly candid. "I just think it's impractical to try to do all the functions in software in view of all the software complexity," he explained. "And we ran into a performance per watt issue trying to do these things."

Even with a smaller process, the performance/watt will still be an issue

(Hi Viditor, haven't seen you posting in a while)

So you think maybe they were thinking they could rely on super-good compilers (much as they do for x86 and itanic) to give larrabee the edge it needed and overcome this deficit?

What I can't get my head around is how is it that they seem to have only come around to this realization so close to the planned/targeted product release date?

Very much like their foray into HDTV and mobile phones.

Hat-trick?

 

[sandorski]

(Hi Viditor, haven't seen you posting in a while)

So you think maybe they were thinking they could rely on super-good compilers (much as they do for x86 and itanic) to give larrabee the edge it needed and overcome this deficit?

What I can't get my head around is how is it that they seem to have only come around to this realization so close to the planned/targeted product release date?

Very much like their foray into HDTV and mobile phones.

Hat-trick?

They seem to gravitate towards the Odd when it comes to Video. Like the i740(IIRC that's what it was called) and low RAM idea. I suspect these are just really smart people(I'm not always convinced of this though) seeing a problem and addressing it, but not really understanding the fundamentals of the thing they're trying to fix.

I highly suspect that both ATI and NVidia were kinda biting their tongue during the whole attempt. Both knowing possibly numerous reasons why it was going to fail, but not wanting to help them.

 

[Viditor]

Hm. That's a very interesting point. Even though those onboards are practically a completely different market from Nvidia's specialty. I guess it would depend which direction Intel wanted to take it. If they copied over (well, I guess of course they would) all the IP to their integrated side of things, then yeah, SEC would be interested. If they simply applied their manufacturing/process knowledge and capabilities to Nvidia's cards, that would be very interesting.

It doesn't really matter what their intentions are...
The goal of the SEC is to prevent unfairness in the market (i.e. avoiding monopolies.) To do this they tend to be very conservative when granting permissions for M&A (Mergers and Acquisitions). Also, remember that Intel is still under active investigation by the FTC and State of New York for anti-competitive actions....
Allowing Intel the chance to increase their marketshare by 50% to have indisputable control of a sector just isn't going to happen...

 

[Viditor]

(Hi Viditor, haven't seen you posting in a while)

So you think maybe they were thinking they could rely on super-good compilers (much as they do for x86 and itanic) to give larrabee the edge it needed and overcome this deficit?

What I can't get my head around is how is it that they seem to have only come around to this realization so close to the planned/targeted product release date?

Very much like their foray into HDTV and mobile phones.

Hat-trick?

G'day mate! (started a new company and I've been flat-out)

I think Sandorski picked it...
While it's quite possible that there was also some marketing or strategic pressure intended, I think that Intel just has this huge blind spot when it comes to high-end graphics...

 

[Dark Shroud]

Intel has more to gain with IGPs and hybred GPU CPUs. Sandy Bridge GPUs are good enough for the masses. As long as Intel keeps this up with yearly performance jumps & adding new APIs when needed (OpenCL) that's all they'll ever have to do.

 

[HumblePie]

Sorry to digress from the OP but...

Wait, the SEC investigates something other than Porn sites?

 

[happy medium]

2 problems with that theory...

1. Nvidia would cost far more than it's worth
2. The current market share for graphics (+/-5%) is Nvidia=25%, ATI (AMD)=25%, Intel=50%...yup, those onboard graphics sure do sell ALOT. I don't think the SEC would allow the sale to Intel.

BTW, you should read through Piazza's comments...some interesting points on WHY Larrabee failed.

This supports my theory. Right from Nvidia's V.P 's mouth.

quote:
"Since we heard similar vibes from some senior people at Intel, we believe that the proverbial can of whoopass might actually get out of the picture so these two companies can start closer together to fight its mutual enemy, Jerry Sanders' legacy company that goes by the name AMD."


http://www.fudzilla.com/home/news/nvidia-aware-of-intel-as-a-tough-competitor

 

[Genx87]

They seem to gravitate towards the Odd when it comes to Video. Like the i740(IIRC that's what it was called) and low RAM idea. I suspect these are just really smart people(I'm not always convinced of this though) seeing a problem and addressing it, but not really understanding the fundamentals of the thing they're trying to fix.

I highly suspect that both ATI and NVidia were kinda biting their tongue during the whole attempt. Both knowing possibly numerous reasons why it was going to fail, but not wanting to help them.

i740 was designed to show off AGPs ability to utilize main memory for the video card. Intel didnt think about the discrete cards that came with their own onboard memory. I think at the time the i740 came out we had 16MB V2s and the Nvidia cards were pushing 32MB. The i740 afaik came with 4MB of on board video.

 

[sandorski]

i740 was designed to show off AGPs ability to utilize main memory for the video card. Intel didnt think about the discrete cards that came with their own onboard memory. I think at the time the i740 came out we had 16MB V2s and the Nvidia cards were pushing 32MB. The i740 afaik came with 4MB of on board video.

Yes, but the reason it had little RAM was largely due to the high Price for RAM that previously existed. AGP and i740 were partially intended to usher in an era of Low RAM Cards that used System RAM. aka, it was thought that Vidcards were running into a RAM Price Wall and that future progress required access to System RAM.

 

[Phynaz]

I thought everyone knew this is Larabee.

 

[bryanW1995]

it might be based upon the technology that they developed for larrabee, but they were pretty clear at idf that larrabee itself is dead.

 

[Viditor]

I thought everyone knew this is Larabee.

Sounds a lot closer...while Larrabee will never be a graphics part, it can still be a nice rendering engine part. That's a lot less complex for the programers, and the large number of cores would suit it very nicely.

 

[Skurge]

I thought everyone knew this is Larabee.

I'm actually quite impressed with that.

 

[sandorski]

I thought everyone knew this is Larabee.

ATI-Crossfire
Nvidia-SLI
Intel-Servers

One of them is doing it wrong.