Father of Larrabee Failure Back at Intel

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,630
Tom Forsyth, the father of the Larrabee "GPU," is back at Intel working with Raja Koduri. He explains that he took the job at Intel, but does not know what he will be working on yet. Insider rumors point to him being "Executive Director of What Not to Do This Time." While we are big fans of Raja here at HardOCP, we do have to believe this position is likely redundant.
 
I wish they had of released Larrabee. So much R&D into something never released.
 
Welp, so much for having a third competitor in the dGPU market...
 
Since we don't know what direction this Intel GPU is going to take it is weird to comment on where it is going. Unless he was hired to keep him away from the competition but that seems unlikely..
 
katanaD said:
how do you take a job that you dont know what your doing?
Click to expand...

Well, he could know what his actual position is, and just not know which project he'll be working on yet. Or, maybe they're ramping up for a project, but haven't fully set up their organization yet. I've taken jobs before for the general position, then added to a specific project afterwards. This actually happens a lot at tech companies.
 
Hi Kyle. I was under the impression that you did not like Raja after he tried to get RTG separated from AMD. He and Lisa did not get along as I understand it. He also was not transparent about Vega at all. I do not like him one bit.

Maybe you can also help us understand something. How important is the leadership of somebody like Raja? Can AMD for instance recover from him leaving? Was he the main creator of Vega - in other words, how indispensable was Raja at AMD? Are they just completing what he already did and then they are in trouble?

Sorry for the many questions...
 
i bet Raja is regreting leaving AMD, knowing how he loved to be on the spot lights doing interviews presenting products, etc, and i have a feeling he wont be doing this at intel.
he probably have so much work that h doesn't have enough time left to drink.
i don't know why i picture raja as a lazy guy who delegates everything.
 
Johan Steyn said:
Hi Kyle. I was under the impression that you did not like Raja after he tried to get RTG separated from AMD. He and Lisa did not get along as I understand it. He also was not transparent about Vega at all. I do not like him one bit.

Maybe you can also help us understand something. How important is the leadership of somebody like Raja? Can AMD for instance recover from him leaving? Was he the main creator of Vega - in other words, how indispensable was Raja at AMD? Are they just completing what he already did and then they are in trouble?

Sorry for the many questions...
Click to expand...


Raja was more worried about capsaicin and cream and dancing around with a stogie while presenting something off topic about alcohol with a lab coat on. Then he throws up some random AOTS numbers that were head scratching and when the card released everyone was like WTF. The guy acted all excited like these releases were industry leading, they were 2 years behind Nvidia every damn launch after Fiji. From Lisa Su's standpoint, he was an embarrassment and I am willing to bet the talent did not start with him. They now have David Wang... look him up. They lost nothing in my opinion.

Scott Herkelman
David Wang
Mike Rayfield


Sounds like a good team to me. We'll see what happens.
 
Johan Steyn said:
Hi Kyle. I was under the impression that you did not like Raja after he tried to get RTG separated from AMD.
Click to expand...
Don't confuse "liking" someone with reporting. This is such a fallacy in the industry that in order to report something "bad" you have to dislike that person. Raja and I are still in communication. Bottom line he is a good guy, and I do like it. However, that has nothing to do with my analysis of his actions.

Johan Steyn said:
How important is the leadership of somebody like Raja?
Click to expand...
Extremely. He is the person leading the project engineers and he is requiring them to meet his visions of what the tech should be.

Johan Steyn said:
Can AMD for instance recover from him leaving?
Click to expand...
Assuredly. I would suggest that it is likely that AMD is better off without him due to politics inside the organization.

Johan Steyn said:
Was he the main creator of Vega - in other words, how indispensable was Raja at AMD?
Click to expand...
Vega was pretty much Raja's baby, but I think it did not meet the targets that he was focused on.

Johan Steyn said:
Are they just completing what he already did and then they are in trouble?
Click to expand...
Not necessarily completing what he did, but they are certainly having to work inside the project confinements that are his legacy.
 
It seems to me Intel is getting serious about entering the GPU world.

Its too early to know if they will succeed, and I don't expect them to beat nvidia on the first try, but they could put a lot of pressure en AMD.
 
Considering Nvidia absorbed all of the other companies that had anything valuable in the GPU spectrum over a decade, it's not surprising Intel couldn't just deliver an enthusiast GPU on the first go. They have however been silently the number 1 gpu in the world since I don't know, forever, so eventually even a broken clock will be right 2 times a day blah blah blah.

It really is surprising that INTEL neglected the gpu importance in transition off the CPU for almost every industry moving ahead for so long, allowing nvidia grow from a 1 billion dollar 'accessory' company to a beast of a 155B various market leading stock , (Intel hovers around 255B, for now. . .)

I wouldn't be surprised if Intel shoots for the 199$ market on their first go of a consumer GPU, (they their marketing and claims are pure bullshit), and over time works on becoming competitive in the GPU sector. Every business and system Nvidia gets a foothold in, is a lost customer for Intel.
 
J3RK said:
Well, he could know what his actual position is, and just not know which project he'll be working on yet. Or, maybe they're ramping up for a project, but haven't fully set up their organization yet. I've taken jobs before for the general position, then added to a specific project afterwards. This actually happens a lot at tech companies.
Click to expand...

Is that where they put you on the roof of Hooli until your contract is up?
 
oldmanbal said:
Considering Nvidia absorbed all of the other companies that had anything valuable in the GPU spectrum over a decade, it's not surprising Intel couldn't just deliver an enthusiast GPU on the first go. They have however been silently the number 1 gpu in the world since I don't know, forever, so eventually even a broken clock will be right 2 times a day blah blah blah.
Click to expand...

Probably because around that time, every analyst was spelling the death of the PC. So management wisely decided to spend their headcount and capital on such fruitful ventures like Phones, Tablets, and Wearables.

I mean why on earth would you make a gaming GPU if PC gaming was dead, and all the revenue was coming from the likes of Angry Birds?

oldmanbal said:
I wouldn't be surprised if Intel shoots for the 199$ market on their first go of a consumer GPU, (they their marketing and claims are pure bullshit), and over time works on becoming competitive in the GPU sector. Every business and system Nvidia gets a foothold in, is a lost customer for Intel.
Click to expand...

Idk about price point, but make no mistake the primary objective is about defending the compute market, where GPGPU is making strides. And the new buzzword is AI.
 
Last edited:
Zarathustra[H] said:
Back when Vega first launched and people were a little disappointed with it, the word was that the Vega architecture was still derivative of the Islands architecture, which is why it disappointed, and that Navï would be the first truly new "Raja" architecture when we could see what he could do.

Was this just fanboy nonsense, or does it have some truth to it?
Click to expand...

I think anyone who looked at the original Vega 64 driver capabilities upstreamed to Linux noticed a very uncanny resemblance to Fiji. And I think anyone that was remotely realistic could've predicted Vega performance as a linear function of Fiji based on core clock. Just some people didn't like where the numbers ended up and held out 'hope'.
 
oldmanbal said:
Considering Nvidia absorbed all of the other companies that had anything valuable in the GPU spectrum over a decade, it's not surprising Intel couldn't just deliver an enthusiast GPU on the first go. They have however been silently the number 1 gpu in the world since I don't know, forever, so eventually even a broken clock will be right 2 times a day blah blah blah.

It really is surprising that INTEL neglected the gpu importance in transition off the CPU for almost every industry moving ahead for so long, allowing nvidia grow from a 1 billion dollar 'accessory' company to a beast of a 155B various market leading stock , (Intel hovers around 255B, for now. . .)

I wouldn't be surprised if Intel shoots for the 199$ market on their first go of a consumer GPU, (they their marketing and claims are pure bullshit), and over time works on becoming competitive in the GPU sector. Every business and system Nvidia gets a foothold in, is a lost customer for Intel.
Click to expand...

Not just Nvidia. Apple has been quietly absorbing major GPU technology for decades..
 
  • Like
Reactions: DocNo
like this
Larrabee was an interesting research project which should not have gotten exposure outside the company - it was early work on a concept.

You do research on a lot of things, not all of them pan out. That's why it is Research and not Development. If I had a buck for every "let's probe this path" project I worked on which never saw the light of day, I'd have... well, a fistful of dollars. Normally, you don't hear about them though.
 
PhaseNoise said:
Larrabee was an interesting research project which should not have gotten exposure outside the company - it was early work on a concept.

You do research on a lot of things, not all of them pan out. That's why it is Research and not Development. If I had a buck for every "let's probe this path" project I worked on which never saw the light of day, I'd have... well, a fistful of dollars. Normally, you don't hear about them though.
Click to expand...

It wasn't a failure. KC/KL/Phi has incredible GFloP performance in a 200 watt package. But it's not optimized or streamlined for graphics. It's backend only had a very simple texture mapper as dedicated hardware.
 
DigitalGriffin said:
It wasn't a failure. KC/KL/Phi has incredible GFloP performance in a 200 watt package. But it's not optimized or streamlined for graphics. It's backend only had a very simple texture mapper as dedicated hardware.
Click to expand...

Agreed. I'm saying that "pasta at the wall" research projects themselves don't always get productized themselves, but that research is worth doing and needs to be done.
 
Riccochet said:
Wish I could get hired for a job at ridiculous pay and not know what the fuck my job will actually be.
Click to expand...

Right on man! I was going to say this. Shit man, I wish I was at the level of getting a life-changing career opportunity and not knowing wtf is it.
 
You must log in or register to reply here.
Back
Top