Raja Koduri, Radeon Technologies Boss, leaves AMD

Intel to Develop Discrete GPUs, Hires Raja Koduri as Chief Architect & Senior VP

Starting from the top and following yesterday’s formal resignation from AMD, Raja Koduri has jumped ship to Intel, where he will be serving as a Senior VP for the company, overseeing the new Core and Visual Computing group. As a chief architect and general manager, Intel is tasking Raja with significantly expanding their GPU business, particularly as the company re-enters the discrete GPU field. Raja of course has a long history in the GPU space as a leader in GPU architecture, serving as the manager of AMD’s graphics business twice, and in between AMD stints serving as the director of graphics architecture on Apple’s GPU team.



Meanwhile, in perhaps the only news that can outshine the fact that Raja Koduri is joining Intel, is what he will be doing for Intel. As part of today’s revelation, Intel has announced that they are instituting a new top-to-bottom GPU strategy. At the bottom, the company wants to extend their existing iGPU market into new classes of edge devices, and while Intel doesn’t go into much more detail than this, the fact that they use the term “edge” strongly implies that we’re talking about IoT-class devices, where edge goes hand-in-hand with neural network inference. This is a field Intel already plays in to some extent with their Atom processors on the GPU side, and their Movidius neural compute engines on the dedicated silicon sign.



However in what’s likely the most exciting part of this news for PC enthusiasts and the tech industry as a whole, is that in aiming at the top of the market, Intel will once again be going back into developing discrete GPUs. The company has tried this route twice before; once in the early days with the i740 in the late 90s, and again with the aborted Larrabee project in the late 2000s. However even though these efforts never panned out quite like Intel has hoped, the company has continued to develop their GPU architecture and GPU-like devices, the latter embodying the massive parallel compute focused Xeon Phi family.



Yet while Intel has GPU-like products for certain markets, the company doesn’t have a proper GPU solution once you get beyond their existing GT4-class iGPUs, which are, roughly speaking, on par with $150 or so discrete GPUs. Which is to say that Intel doesn’t have access to the midrange market or above with their iGPUs. With the hiring of Raja and Intel’s new direction, the company is going to be expanding into full discrete GPUs for what the company calls “a broad range of computing segments.”
Click to expand...

https://www.anandtech.com/show/1201...ete-gpus-hires-raja-koduri-as-chief-architect
 
TBH I blame Raja for the current Vega situation. Even Fury X etc. I don't think that is the person I would want to run my GPU department.

If anything, this might be a good thing for AMD. IMO I do not see Raja helping intel all that much, not like he take trade secrets he had at RTG.

Either way It is a good move for AMD.
 
Looks to me like the chip will be a low volume experiment product, and that Intel will pump out something with a small Xeon Phi GPGPU strapped into it.
 
TaintedSquirrel said:
I'm excited to see what Raja can do with an actual r&d budget.
Click to expand...



About damned time that Intel got off their ass and did something with their projected massive spare fab capacity to justify capex increases (Their massive cuts to capex are the main reason why they are in such a bind in 10nm currently).

The very fact that Intel has waited 4+ decades to figure out that "Hurr Durr this graphics co-processor thing might be the next big thing!" is about the most "Intel" way of thinking possible. Nvidia is basically eating the entire HPC market while Intel twiddles it's thumbs.
 
Last edited:
Wow I'm surprised amd didn't put a no compete Claus into Raj's contract

Yikes.
 
Chimpee said:
In California, Non-Compete Clause is actually illegal when it comes to working for your competitor, though he probably did sign something regarding trade secrets.
Click to expand...
He is a lead engineer instead of the typical low level engineer where this matters.

He can just make a different GPU architecture if he wants to.

The only place where this matters is if you use something that is essentially identical.

Intel has cross-license agreements with AMD anyways.
 
Wooot 3 companies competing with high end GPUs? (y)
Make it happen!
When was the last time that happened? Back before 3DFX was bought by Nvidia? Like, 20 years ago?
 
If Intel's truly getting serious about GPUs then bye AMD, nice knowing you.

AMD should just sell RTG to Intel and let Intel and Nvidia duke it out. Stay in the CPU market and build back up.

While Intel is distracted with Nvidia they can regain some ground and get healthy.
 
vegeta535 said:
Honestly this doesn't bold well for Navi. Navi was the first product he had full hands on from conception since he joined RTG no? They must not have much faith in Navi if they letting him go.
Click to expand...

Why does this not bode well for Navi. The design is completed. It is ready for tape out. I seriosuly doubt Koduri was forced out as others have suggested. The RTG group has excellent engineers and Koduri is not the only one with design capabilities. I expect Navi to be on time and successful. NAVI is the only design that Koduri is wholly responsible for. The negativity here about him is really not based on fact.
 
trandoanhung1991 said:
If Intel's truly getting serious about GPUs then bye AMD, nice knowing you.

AMD should just sell RTG to Intel and let Intel and Nvidia duke it out. Stay in the CPU market and build back up.

While Intel is distracted with Nvidia they can regain some ground and get healthy.
Click to expand...

Wrong move! Compute capability is the future of pc development and gpu design is key to it. Idle speculation without thought leads to crazy conclusions.
 
Lisa Su apparently got tired of working with someone who would stab her in the back if she's not looking.

The sabbatical was forced upon him and wasn't really his decision.

Raja wasn't quite fired, but the plan was to put someone else in charge, and cut Raja out of any important management decision.

Although he wasn't fired, his decision to voluntary leave was widely expected.
 
_mockingbird said:
Lisa Su apparently got tired of working with someone who would stab her in the back if she's not looking.

The sabbatical was forced upon him and wasn't really his decision.

Raja wasn't quite fired, but the plan was to put someone else in charge, and cut Raja out of any important management decision.

Although he wasn't fired, his decision to voluntary leave was widely expected.
Click to expand...

Lisa Su killed RTG with her server dreams and consoles is the next one they lose.

She should have betted on GPUs instead.
https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-third-quarter-fiscal-2018
 
Last edited:
_mockingbird said:
No.

She put the wrong person in charge.

Now, she is correcting that.
Click to expand...


It wasn't her choice to put him in charge lol, it was the board of directors that gave Raja the position, she was forced to put him in that position.

I don't think AMD's GPU division is even capable of finding a replacement for Raja right now. The amount of potential GPU tech directors out there can be counted on 1 hand. And all of them are empolyed at high level positions with companies that are well quite frankly are capable of paying them and giving them the budgets to create market competitive products.

mockingbird, they could have put a monkey as an engineer in there and it would do the same thing with the budgets they had.

There is no reason Vega mediocre performance increase by clock only from Fiji. NO damn reason. Fiji has so much throughput issues, issues that should have been fixed with Polaris, but didn't come over to Vega? Under the hood they didn't do changes to fix GCN's scaling issues, essentially its the same damn chip with higher clocks. So how much of a budget did they really have dedicated to GPU only with that 400 million a quarter. I woudln't be surprised if its around 100 million the rest goes to CPU. That would explain why Polaris could only match Maxwell in Perf/watt, and Vega also Matches Maxwell too. But that is all from node savings. Architecture wise its the same old GCN.
 
Last edited:
_mockingbird said:
No.

She put the wrong person in charge.

Now, she is correcting that.
Click to expand...

Here is a dollar, now beat Nvidias 1.5B+ R&D investment in graphics. See the issue?

Its so R&D starved you cant even get multiple SKUs because they cant afford the IC design. If you got nothing to work with it doesn't matter how good or bad the person is.

Lisa Su also killed AMDs small cores. And a bread and butter product like the APUs are ~9 months behind in a messed up priority.
 
Last edited:
Shintai said:
Here is a dollar, now beat Nvidias 1.5B+ R&D investment in graphics. See the issue?

Its so R&D starved you cant even get multiple SKUs because they cant afford the IC design. If you got nothing to work with it doesn't matter how good or bad the person is.

Lisa Su also killed AMDs small cores. And a bread and butter product like the APUs are ~9 months behind in a messed up priority.
Click to expand...


Yep there was a reason why r290 series was rebranded to the r390, no money to make anything else, so tweak the existing chips, that's all they could do.
 
razor1 said:
It wasn't her choice to put him in charge lol, it was the board of directors that gave Raja the position, she was forced to put him in that position.

I don't think AMD's GPU division is even capable of finding a replacement for Raja right now. The amount of potential GPU tech directors out there can be counted on 1 hand. And all of them are empolyed at high level positions with companies that are well quite frankly are capable of paying them and giving them the budgets to create market competitive products.
Click to expand...

That's not how it happened.

Raja went up to Su and threatened to leave unless he gets his own group.

RTG was then made from what was then a sprawling of different teams.

razor1 said:
mockingbird, they could have put a monkey as an engineer in there and it would do the same thing with the budgets they had.

There is no reason Vega mediocre performance increase by clock only from Fiji. NO damn reason. Fiji has so much throughput issues, issues that should have been fixed with Polaris, but didn't come over to Vega? Under the hood they didn't do changes to fix GCN's scaling issues, essentially its the same damn chip with higher clocks. So how much of a budget did they really have dedicated to GPU only with that 400 million a quarter. I woudln't be surprised if its around 100 million the rest goes to CPU. That would explain why Polaris could only match Maxwell in Perf/watt, and Vega also Matches Maxwell too. But that is all from node savings. Architecture wise its the same old GCN.
Click to expand...

As I've said before, creating a competitive product not about throwing money at it.

RTG did have a replacement for Fiji: it's Polaris.

The problem was that Polaris performed far below expectation.

As a result, RTG went back to Fiji to develop Vega.
 
RTG was made because it was one of the things Lisa Su was willing to sell off.
 
_mockingbird said:
That's not how it happened.

Raja went up to Su and threatened to leave unless he gets his own group.

RTG was then made from what was then a sprawling of different teams.
Click to expand...

And you think that would MAKE Su put him in the position? A threat means very little unless it has weight behind it. That weight was the board of directors.

As I've said before, creating a competitive product not about throwing money at it.

RTG did have a replacement for Fiji: it's Polaris.

The problem was that Polaris performed far below expectation.

As a result, RTG went back to Fiji to develop Vega.
Click to expand...

Right, and that is why we see so many incarnations of crappy GCN over and over again. Cause they had enough money to make a new architecture but didn't *sarcasm*? So maybe its their lack of understanding GPU bottlenecks that created this problem *sarcasm*, even though we can see the problem as an end user?
 
_mockingbird said:
No, RTG was made at the request of Raja.

It's Raja who wanted the group sold.
Click to expand...


Raja could never sell the group, he knows that, but it was his request to make the group, I agree, but it was so he could have total autonomy of the group, instead of having CPU groups grubby little fingers screwing up the GPU side of things like it has been. You think Su would just hand over that power to him if she wasn't forced to?
 
razor1 said:
And you think that would MAKE Su put him in the position? A threat means very little unless it has weight behind it. That weight was the board of directors.
Click to expand...

Yes. And you don't need to take this from me. In fact, Kyle Bennett wrote about it over a year ago.

razor1 said:
Right, and that is why we see so many incarnations of crappy GCN over and over again. Cause they had enough money to make a new architecture but didn't *sarcasm*? So maybe its their lack of understanding GPU bottlenecks that created this problem *sarcasm*, even though we can see the problem as an end user?
Click to expand...

RTG did have enough money to make a new architecture and did made one, but it didn't meet expectation.

razor1 said:
Raja could never sell the group, he knows that, but it was his request to make the group, I agree, but it was so he could have total autonomy of the group, instead of having CPU groups grubby little fingers screwing up the GPU side of things like it has been. You think Su would just hand over that power to him if she wasn't forced to?
Click to expand...

It literally said right here: "Let’s start with the tension. Koduri was able to wrestle control of the graphics division away during AMD’s last leadership transition after threatening to leave the ship and take a role at Intel, something he's not shy about telling his AMD colleagues. Lisa Su caved and Koduri got the job."
 
_mockingbird said:
Yes. And you don't need to take this from me. In fact, Kyle Bennett wrote about it over a year ago.



RTG did have enough money to make a new architecture and did made one, but it didn't meet expectation.



It literally said right here: "Let’s start with the tension. Koduri was able to wrestle control of the graphics division away during AMD’s last leadership transition after threatening to leave the ship and take a role at Intel, something he's not shy about telling his AMD colleagues. Lisa Su caved and Koduri got the job."
Click to expand...


Just because someone threatens to leave to go somewhere else doesn't force a person to give up their control over 50% of the company, I don't know if you ever were part of an LLC or C corp cause politics play a big way when controlling interest is diversified across many people.and those people are not part of the C level, another words the control interest is outside of the employees.

Wrestled control what does that mean, the control was Su's but it was given to Raja by the party that really had the controlling interest the board of directors. Su MUST go along with their decision.

RTG didn't have enough money, its plainly obvious. Vega is worse than Fiji from a competitive point of view and if we know anything about silicon design they would know that well before the chip even came out. That is how design is done, know the parts and have expectations it might not meat expectations or surpass expectations but always within a certain amount, its like they missed it by more then 10% man, if they did they need to figure out WTF they are doing wrong with their brains.

You make it sound like AMD's engineers are incompetent. They aren't....... What is incompetent was their management of the situations, they should have never came into this situation where they had to sacrifice 50% of their company.
 
Last edited:
You must log in or register to reply here.
Back
Top