Wednesday, November 8th 2017
Intel Hires Raja Koduri, to Develop Discrete GPUs, This Time for Real
Intel hired Raja Koduri, who resigned as head of AMD's Radeon Technologies Group (RTG), earlier this week. Koduri has been made Senior Vice President and Chief Architect of Intel's future discrete GPUs. That's right, Intel has renewed its dreams to power high-end graphics cards that compete with AMD and NVIDIA. Intel's last attempt at a discrete GPU was "Larrabee," which evolved into a super-scalar multi-core processor for HPC applications under the Xeon Phi line.
This development heralds two major theories. One, that Intel's collaboration with AMD RTG on graphics IP could only go further from here, and what is a multi-chip module of Intel and AMD IP now, could in the future become a true heterogeneous die of Intel's and AMD's IP. Two, that the consolidation of AMD's graphics assets and IP into a monolithic entity as RTG, could make it easier to sell it lock, stock, and barrel, possibly to Intel.Intel places great faith in Koduri's ability to either develop a major product line from scratch, or to effect tectonic shifts in the industry, such as Apple's transition to the x86 architecture for its Mac product line.
Intel could have one of three approaches to build a discrete GPU from scratch. The first and most obvious one would be to scale up its current gen 9.5 architecture. The trouble is, that Intel's SIMD parallelism is more transistor-heavy than even NVIDIA. It takes roughly 400-500 million transistors (coarse estimate) on the "Kaby Lake" die to build a GPU with 24 execution units. With 10 billion transistors, we're looking at around 480 execution units, if not more.
The second approach would be to build a new graphics architecture from scratch. Something like this, even with Intel's deep R&D pockets, could take a team led by Koduri 3-4 years. The resulting architecture has to relevant to the market of the time, or end up missing the bus like Larrabee. The third approach would be to either license or acquire GPU IP from AMD. Koduri has the reputation of a tech business strategist as much as an IP guru to effect such a change.
These are strange times in the vale, as silicon giants Broadcom and Qualcomm look to coalesce into the world's third largest chipmaker, and Marvell with Cavium follow on. Stranger things are currently happening between past industry rivals, than the possibility of Intel acquiring RTG from AMD in exchange for cash, and allowing AMD's merger with another chipmaker without affecting its x86 license. This is just a really audacious theory.
This development heralds two major theories. One, that Intel's collaboration with AMD RTG on graphics IP could only go further from here, and what is a multi-chip module of Intel and AMD IP now, could in the future become a true heterogeneous die of Intel's and AMD's IP. Two, that the consolidation of AMD's graphics assets and IP into a monolithic entity as RTG, could make it easier to sell it lock, stock, and barrel, possibly to Intel.Intel places great faith in Koduri's ability to either develop a major product line from scratch, or to effect tectonic shifts in the industry, such as Apple's transition to the x86 architecture for its Mac product line.
Intel could have one of three approaches to build a discrete GPU from scratch. The first and most obvious one would be to scale up its current gen 9.5 architecture. The trouble is, that Intel's SIMD parallelism is more transistor-heavy than even NVIDIA. It takes roughly 400-500 million transistors (coarse estimate) on the "Kaby Lake" die to build a GPU with 24 execution units. With 10 billion transistors, we're looking at around 480 execution units, if not more.
The second approach would be to build a new graphics architecture from scratch. Something like this, even with Intel's deep R&D pockets, could take a team led by Koduri 3-4 years. The resulting architecture has to relevant to the market of the time, or end up missing the bus like Larrabee. The third approach would be to either license or acquire GPU IP from AMD. Koduri has the reputation of a tech business strategist as much as an IP guru to effect such a change.
These are strange times in the vale, as silicon giants Broadcom and Qualcomm look to coalesce into the world's third largest chipmaker, and Marvell with Cavium follow on. Stranger things are currently happening between past industry rivals, than the possibility of Intel acquiring RTG from AMD in exchange for cash, and allowing AMD's merger with another chipmaker without affecting its x86 license. This is just a really audacious theory.
71 Comments on Intel Hires Raja Koduri, to Develop Discrete GPUs, This Time for Real
Unless AMD can easily find its own tech and its patented, would this not be grounds for some sort of lawsuit?
Lisa Su must be pissed.
It's like rats fleeing a sinking ship.
What I find more than a little curious is out of all the youtube channels covering tech and the computer biz none of them are talking about this aspect of these events. It would seem to me one of Raja's most perishable values right now for Intel is that of liason with the staffs of the RTG and Semi Custom Division. That is a real important asset now that AMD will be supplying hardware to Intel. Raja has the AMD Telephone Directory and Organization Layout of the place. He is just the guy for Intel to go to if they experience any problems as Raja will know who exactly to call. I was really surprised when Intel blew the trumpets and read from the scroll Raja's new title and responsibilities in such a public fashion so soon. Obviously Raja will now have to take it easy with the phone calls to his old employer.
With Intel I would imagine its first things first and that means get this new project rolling which should take too long. Maybe 6 to 9 months. Then they go big guns on the Discrete GPU Project.
If any operation is going to have its staff raided its going to be NVIDIA. Both Intel and AMD have their eye on them at this point. I would expect some major doings in January.
No doubt they will change plans where feasible.
Is it inconvenient ? Yes
Is it essential ? Also yes
These things happen. Lisa and the Staff will just have to deal with it. It's called strategic planning.
Guys like Raja and Jim Keller are Hard Chargers. They are extremely focused on a task and when that task comes to a conclusion they move on. Their term of service in any one gig is rarely more than 5 years and can be as short as 2 years. The fact that Raja left so abruptly is viewed by Senior Management as an opportunity. This is why Lisa was appointed as Interim Department Head. You can bet she is assessing what is going on in that department and making notes about what needs to be addressed. You can also expect that as time progresses Lisa will step away and some young hot shot who is gunning for a promotion will be given a 'Delegation of Authority" to evaluate whether he can handle the job. The way they can rescind the delegation without causing to much of a mess.
I would not be surprised if AMD is eyeing up Jim Keller again. He has a very good record with AMD and is 2 years into his current gig. I doubt he will last more than 3 more years with Tesla. So things may have to cook for a year or two. Maybe someone gets promoted to fill the hole left by Raja. Jim becomes available a year or two down the road and the AMD Technical Staff is restructured.
Who knows ?
Let no mistake be made the Really Money in the computer business lies in Enterprise/Data Center Sector. And right now only AMD can offer both an x86 CPU Platform (Epyc) along with a Compute Card GPU that can rival anything available. (Vega) Intel snatched up Raja and no doubt will develop their own discrete GPU but that will take time. In the mean time AMD will release Navi accelerating NVIDIA's departure. As of Q2 the sale of GPUs accounted for 60% of NVIDIA's revenue. AI and Autonomous Driving was about 15% each. Gaming hardware sales amounts to almost nothing and even that is fading like a tan with the growth of Online Gaming.
Raja leaving AMD is'nt bad news at all. AMD still has a few key players such as someone who was on lead on the development of the Cell CPU which was known in the PS3. That CPU could push big numbers compared to X86 and was pretty decent on a console too. Widely used into folding (complete DC's stacked with PS3's). They set the base for Navi, now it's time for raja to go.