Radeon Technologies Group boss Raja Koduri exits

Shawn Knight

Posts: 15,306   +193
Staff member

Raja Koduri, head of the Radeon Technologies Group at AMD, took a leave of absence in late September with the intent of returning to the job sometime in December. Those plans have since changed.

In a memo to AMD staff obtained by Hexus, Koduri reveals that he is leaving the Radeon Technologies Group and AMD.

Koduri says he wholeheartedly believes in what the company is doing with Vega, Navi and beyond and is incredibly proud of how far they have come and where they are going. In thinking about how computing will evolve, however, Koduri notes that he feels the urge to pursue his passion beyond hardware and explore broader solutions.

For now, it’ll be business as usual. Sources tell Hexus that AMD CEO Lisa Su will continue to oversee the Radeon Technologies Group for the foreseeable future. A search for a permanent replacement is already underway with an appointment expected within the next few months.

Also worth noting is that AMD is reportedly putting additional resources behind the Radeon Technologies Group moving forward with the goal of making the division bigger and better. Just this week, for example, it was announced that AMD would be putting its Radeon graphics inside future Intel mobile CPUs.

Any bets on what Koduri does next in his career?

Permalink to story.

 
Not surprised, especially after his constant missteps with Fury, Polaris, and now Vega. Guy couldn't cement all his ideas into a single product, and he was the one very against any RX490 Polaris chip (which subsequently allowed Nvidia to capture a massive amount of lucrative upper mid-range gaming PCs with the unchallenged 1070).

He will not be missed by AMD.
 
Not surprised, especially after his constant missteps with Fury, Polaris, and now Vega. Guy couldn't cement all his ideas into a single product, and he was the one very against any RX490 Polaris chip (which subsequently allowed Nvidia to capture a massive amount of lucrative upper mid-range gaming PCs with the unchallenged 1070).

He will not be missed by AMD.

Outside of the initial power delivery problems of Polaris, I think noone would call it a misstep; Vega on the other side...
 
Not surprised, especially after his constant missteps with Fury, Polaris, and now Vega. Guy couldn't cement all his ideas into a single product, and he was the one very against any RX490 Polaris chip (which subsequently allowed Nvidia to capture a massive amount of lucrative upper mid-range gaming PCs with the unchallenged 1070).

He will not be missed by AMD.

AMD actually had two different engineering teams at RTG. Raja focused on Fury and Vega while a separate team did Polaris so his screw ups are actually worse then you make it out to be because the only competent architecture, polaris, had little involvement from Raja. Polaris was actually a nice upgrade for AMD achieving about a 100% performance per watt improvement. Fury and Vega on the otherhand tried to be far too forward looking and ended up falling on their face. First with Fury, which has already hit a huge memory bottleneck in many games. HBM didn't really add anything and the GPU never really uses all it's cores due to the piss poor scheduler. Vega, which touted a bunch of new features, still doesn't have half of what was advertised enabled and at this point I doubt most of it ever will be. The only good thing with Vega is it's higher clocks but then again nothing is stopping Vega from chugging electricity. If it wasn't for the Polaris improvements, Vega would be guzzling 600w.

At least a bit of good news, Navi is also being designed by that alternate team so at least you can expect some straight up performance improvements.
 
Not surprised, especially after his constant missteps with Fury, Polaris, and now Vega. Guy couldn't cement all his ideas into a single product, and he was the one very against any RX490 Polaris chip (which subsequently allowed Nvidia to capture a massive amount of lucrative upper mid-range gaming PCs with the unchallenged 1070).

He will not be missed by AMD.

Outside of the initial power delivery problems of Polaris, I think noone would call it a misstep; Vega on the other side...
The misstep with polaris isnt with the cards themselves IMO, it is with AMD stopping at the 480. They left huge swaths of the GPU market with no competition, because raja wanted vega 56 to fill that part. Cue the delays, and AMD lost out on a ton of money because they refused to make a 3072 core 384 bit polaris chip.

Huge lost potential there. Polaris should have had 3072/4096 core parts. The 1070 sold like gangbusters and showed exactly what AMD lost out on. Didnt help that the lack of a 490 made the polaris chips look slow as molasses because the first 3-5 GPUs on a graph were all nvidia models.
 
The misstep with polaris isnt with the cards themselves IMO, it is with AMD stopping at the 480. They left huge swaths of the GPU market with no competition, because raja wanted vega 56 to fill that part. Cue the delays, and AMD lost out on a ton of money because they refused to make a 3072 core 384 bit polaris chip.

Huge lost potential there. Polaris should have had 3072/4096 core parts. The 1070 sold like gangbusters and showed exactly what AMD lost out on. Didnt help that the lack of a 490 made the polaris chips look slow as molasses because the first 3-5 GPUs on a graph were all nvidia models.

Well, he's moving to Intel, so whatever we may think, to Intel he didn't fail in the balance sheet, but I understand your point. And just because he's doing so, AMD will either hate him, miss him, or thank him for bridging Intel + AMD (Kaby Lake-G); but definitely won't forget him.
 
If I weren't an Nvidia owner looking for a second 1080,on the cheap.I would find this funny.it's not.

The peops saying the 480 was good.must be those that bought the 4 gig card that could flash to the 8 gig card,any that bought the original 8 gig card were cheated and lied too.donations accepted here.I wonder what his severance package is like.
 
Last edited:
Not surprised, especially after his constant missteps with Fury, Polaris, and now Vega. Guy couldn't cement all his ideas into a single product, and he was the one very against any RX490 Polaris chip (which subsequently allowed Nvidia to capture a massive amount of lucrative upper mid-range gaming PCs with the unchallenged 1070).

He will not be missed by AMD.

Outside of the initial power delivery problems of Polaris, I think noone would call it a misstep; Vega on the other side...
Vega wasn't a misstep. It delivers on the server side quite well which was clearly the focus of the architecture. It has great compute performance and it's also good when it comes to power usage if voltage and clocks are not at Vega 64 levels.

It just came too late for the consumer side and with the very high prices of memory chips, especially HMB2 memory, they could not undercut Nvidia in price. I think they expected a price drop for memory, but it kinda doubled since 2016 :D
 
Back