Wednesday, September 13th 2017

Raja Koduri On a Sabbatical from RTG till December, AMD CEO Takes Over

Raja Koduri, chief of AMD's Radeon Technologies Group (RTG), has reportedly taken an extended leave from the company, running up to December 2017. Ryan Shrout, editor of PC Perspective stated that he got confirmation from the company about this development. Company CEO Lisa Su has taken direct control over RTG in the meantime.

Formed in 2015 after a major internal reorganization, RTG handles a bulk of AMD's graphics IP, developing and marketing products under the Radeon brand, including Radeon RX series consumer graphics chips, Radeon Pro series professional graphics chips, and Radeon Instinct line of GPGPU accelerators. This move is of particular significance as Q4 tends to be the biggest revenue quarter, as sales rally on account of Holiday.

Raja wrote the following letter to his team

"RTG Team,

You haven't heard from me collectively in a while - a symptom not only of the whirlwind of launching Vega, but simply of the huge number of demands on my time since the formation of RTG. Looking back over this short period, it is an impressive view. We have delivered 6 straight quarters of double-digit growth in graphics, culminating in the launch of Vega and being back in high-performance. What we have done with Vega is unparalleled. We entered the high-end gaming, professional workstation and machine intelligence markets with Vega in a very short period of time. The demand for Vega (and Polaris!) is fantastic, and overall momentum for our graphics is strong.

Incredibly, we as AMD also managed to spectacularly re-enter the high-performance CPU segments this year. We are all exceptionally proud of Ryzen, Epyc and Threadripper. The computing world is not the same anymore and the whole world is cheering for AMD. Congratulations and thanks to those of you in RTG who helped see these products through. The market for high-performance computing is on an explosive growth trajectory driven by machine intelligence, visual cloud, blockchain and other exciting new workloads. Our vision of immersive and instinctive computing is within grasp. As we enter 2018, I will be shifting my focus more toward architecting and realizing this vision and rebalancing my operational responsibilities.

At the beginning of the year I warned that Vega would be hard. At the time, some folks didn't believe me. Now many of you understand what I said. Vega was indeed hard on many, and my sincere heartfelt thanks to all of you who endured the Vega journey with me. Vega was personally hard on me as well and I used up a lot of family credits during this journey. I have decided to take a time-off in Q4 to spend time with my family. I have been contemplating this for a while now and there was never a good time to do this. Lisa and I agreed that Q4 is better than 2018, before the next wave of product excitement. Lisa will be acting as the leader of RTG during by absence. My sincere thanks to Lisa and rest of AET for supporting me in this decision and agreeing to take on additional workload during my absence.

I am looking to start my time-off on Sept 25th and return in December.

Thank you, all of you, for your unwavering focus, dedication and support over these past months, and for helping us to build something incredible. We are not done yet, and keep the momentum going!

Regards, Raja"
Source: Ryan Shrout (Twitter)
Add your own comment

50 Comments on Raja Koduri On a Sabbatical from RTG till December, AMD CEO Takes Over

#26
Vayra86
FordGT90ConceptNot really. The fact they can't keep it in stock is proof of that. Polaris has the same problem.
Time will tell. I can see where you're coming from, perhaps the numbers actually DO look good for AMD's expectations, because Vega is definitely more than just a gaming card, but still. Its not a great product in terms of positioning, profit margins, and yield / stock problems. The most plausible explanation for Vega's lack of availability is IMO still that AMD doesn't really want to sell a huge lot of them at all, and is purposely squeezing down production.

That said, you can't really defend Raja's public appearances though, right?
Posted on Reply
#27
FrustratedGarrett
Good riddance! Whoever hired him should be fired too, along with low-energy Lisa.
Posted on Reply
#28
Totally
No one is being fired. Simply PR damage control.
Posted on Reply
#29
FrustratedGarrett
TotallyNo one is being fired. Simply PR damage control.
He's leaving on his own because he knows that he's the least qualified person in that company to hold that position.
Posted on Reply
#30
Totally
FrustratedGarrettHe's leaving on his own because he knows that he's the least qualified person in that company to hold that position.
He's coming back at an unknown capacity(not as the poster child for RTG at least) at the end of the year.
Posted on Reply
#31
FordGT90Concept
"I go fast!1!11!1!"
TotallyNo one is saying there is anything wrong with the product...
I was replying to Prima.Vera whom said exactly that. :p
Vayra86That said, you can't really defend Raja's public appearances though, right?
Enlighten me. What I have seen, I can't say Koduri did anything outstandingly wrong.
Posted on Reply
#32
rtwjunkie
PC Gaming Enthusiast
FrustratedGarrettHe's leaving on his own because he knows that he's the least qualified person in that company to hold that position.
You like making up your own news as you go along, ehh? o_O
Posted on Reply
#33
HisDivineOrder
I'll always remember Raja as the guy Anand built a shrine to in text form...

www.anandtech.com/show/6907/the-king-is-back-raja-koduri-leaves-apple-returns-to-amd

Look at that giant picture. Look at that guy leaving Apple, where Anand wound up going. It's just chock full of amusing elements.

Back then, the part that was formerly ATI was the big hope for enthusiasts. Today? Raja has run that in the ground so far that we mostly shrug when someone brings up Polaris or Vega, then a fanboy will change the subject to Ryzen.
Posted on Reply
#34
Slizzo
I'm not sure that Vega was Raja's fuck up. Wasn't that project started before he took the helm?

I can understand AMD wanting to reign in his tweeting about their products though. Let a marketing team handle that.
Posted on Reply
#35
dorsetknob
"YOUR RMA REQUEST IS CON-REFUSED"
AquinusPeople don't seem to understand what sabbatical means..
low country Saturday Clog thrower ?
Not Someone you really want in your company :)
There will be a job waiting at the $1000 phone company for him
Posted on Reply
#36
wiyosaya
Personally, I see this as a good thing. Su has a brain.
Posted on Reply
#37
Frick
Fishfaced Nincompoop
FordGT90ConceptNot really. The fact they can't keep it in stock is proof of that. Polaris has the same problem.
Has there been any Vega stock to begin with though? Serious question.
Posted on Reply
#38
LogitechFan
they needed to ditch this idiot without loosing face. Well, this is the way to do it apparently, before RTG goes on a chopping block and sold for parts to samsung
Posted on Reply
#39
EarthDog
FrickHas there been any Vega stock to begin with though? Serious question.
Not since day one has there been any sold at MSRP. They are not stocking that SKU and it then supply/demand take over from there.




Its weird to call Raja an idiot while in the same breath not using the proper term, losing. I never heard that Raja was a loose fellow. :p
Posted on Reply
#40
evernessince
Vayra86'Back in last year's high performance' :) there fixed that for you Raja

Also, this, is telling to me:
"As we enter 2018, I will be shifting my focus more toward architecting and realizing this vision and rebalancing my operational responsibilities. "

He is definitely not going to be the public face anymore.
Technically that's last year for Nvidia too so yeah... Your point kind of falls flat. Nvidia doesn't have volta planned for anytime soon either. Last year's high performance is still this year's high performance.
Posted on Reply
#41
evernessince
TotallyNo one is saying there is anything wrong with the product, it is the launch and handling of it afterwards that was abysmal. Without any actual numbers the fact that they can't keep stock can not be but in a positive nor negative light because I can easily say they can't keep stock because they didn't have much to begin with.
I agree that the launch was bad but no one can keep stock right now for any card appealing to miners.
Posted on Reply
#42
rtwjunkie
PC Gaming Enthusiast
dorsetknobThere will be a job waiting at the $1000 phone company for him
LOL, you mean Apple, I'm guessing!
Posted on Reply
#43
Steevo
FrustratedGarrettHe's leaving on his own because he knows that he's the least qualified person in that company to hold that position.
And you are qualified to make that judgement call? Or you are the worst troll I have ever seen.
Posted on Reply
#44
Th3pwn3r
LogitechFanthey needed to ditch this idiot without loosing face. Well, this is the way to do it apparently, before RTG goes on a chopping block and sold for parts to samsung
EarthDogNot since day one has there been any sold at MSRP. They are not stocking that SKU and it then supply/demand take over from there.




Its weird to call Raja an idiot while in the same breath not using the proper term, losing. I never heard that Raja was a loose fellow. :p
LOL, was going to say the same thing.
Posted on Reply
#45
Vayra86
evernessinceTechnically that's last year for Nvidia too so yeah... Your point kind of falls flat. Nvidia doesn't have volta planned for anytime soon either. Last year's high performance is still this year's high performance.
Actually no. GV100 is already a product on the marketplace. Not a gaming-GPU, but direct competition for the Vega architecture. If you keep the rose tinted glasses and try to forget that reality then yes, Pascal is 'this year's high performance'... in the small mind of a gamer.

In fact, even Pascal's GP100 is a direct competitor to Vega and that by itself would already have been enough performance wise. Vega really is a couple years too late, seeing as Navi won't be scaling up the performance on one die any further, while Nvidia's Volta does increase single-die performance. AMD actually forced itself to jump on Navi a full gen earlier than Nvidia, and the only reason for doing so, is because they have nowhere else to go. Nvidia on the other hand can push Volta as a single die solution, and THEN still double its potential through the MCM-solutions they have planned. The potential performance gap will then be doubled - a scary thought IMO.

Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU.

And about cards being in stock... I have seen the majority of cards perfectly available for the past few months. Inflated price points, yes. But available, in stock, 24/5 deliveries everywhere. Meanwhile the Vega cards are on hold, out of stock, or to be delivered within a week or two weeks - which remains to be seen.
Posted on Reply
#46
Prima.Vera
FordGT90ConceptNot really. The fact they can't keep it in stock is proof of that. Polaris has the same problem.
Because the yields are crap, so they are actually wasting a lot of money and time with useless semiconductor rebuts.
Posted on Reply
#47
evernessince
Vayra86Actually no. GV100 is already a product on the marketplace. Not a gaming-GPU, but direct competition for the Vega architecture. If you keep the rose tinted glasses and try to forget that reality then yes, Pascal is 'this year's high performance'... in the small mind of a gamer.

In fact, even Pascal's GP100 is a direct competitor to Vega and that by itself would already have been enough performance wise. Vega really is a couple years too late, seeing as Navi won't be scaling up the performance on one die any further, while Nvidia's Volta does increase single-die performance. AMD actually forced itself to jump on Navi a full gen earlier than Nvidia, and the only reason for doing so, is because they have nowhere else to go. Nvidia on the other hand can push Volta as a single die solution, and THEN still double its potential through the MCM-solutions they have planned. The potential performance gap will then be doubled - a scary thought IMO.

Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU.

And about cards being in stock... I have seen the majority of cards perfectly available for the past few months. Inflated price points, yes. But available, in stock, 24/5 deliveries everywhere. Meanwhile the Vega cards are on hold, out of stock, or to be delivered within a week or two weeks - which remains to be seen.
Lol, people on the internet. Your original comment wasn't about volta in the enterprise, it was about consumer volta. Your just grasping at straws /gg

"Nvidia on the other hand can push Volta as a single die solution, and THEN still double its potential through the MCM-solutions they have planned."

Lol, you really have no understanding of this do you? Why in the world would Nvidia continue to make pricey large dies if they could make multiple smaller and cheaper ones? The answer: they can't. AMD is the only CPU manufacturer on the market right now that managed to stitch together multiple higher performance CPUs, something not even Intel can do yet with an R&D budget bigger than the value of the entirety of AMD. Nvidia has even less experience with interconnects then either AMD or Intel and you just suddenly expect them to be able to have the same groundbreaking achievement as AMD did? You are delusional. Nvidia doesn't even have an MCM on their roadmap so your looking at volta and then the next architecture AT LEAST still being single.

This is about as far as Nvidia has gotten

research.nvidia.com/publication/2017-06_MCM-GPU%3A-Multi-Chip-Module-GPUs

of which, Intel has had MCM processors in the past that ultimately failed due to latency between dies. Notice the article's publication date, after Ryzen's launch. This means that Nvidia didn't think MCMs were feasible until after Ryzen. This also means they have zero progress into actually implementing this into their GPUs for real. Ryzen comes out and all of a sudden Nvidia starts talking up MCMs like they are the future. I remember the same thing happening when the original iphone came out and it took others YEARS to catch up.

"Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU."

I don't even know what card you are talking about. WIthout even referencing the card or sources you are just talking out of your ass. Vega 56 = 1070 and Vega 64 = 1080. Their performance are on par with each other. I know you can be talking about the 1080 Ti, because it isn't a year old nor is it mid tier. All of this doesn't change the fact that these cards are still the best Nvidia's got right now. If you didn't realize, your own logic can easily be used against you. For example, The R9 390 was the same price as the GTX 970 and yet the 390's GPU was over 2 years old and still had double the RAM.
Posted on Reply
#48
Slizzo
evernessinceLol, people on the internet. Your original comment wasn't about volta in the enterprise, it was about consumer volta. Your just grasping at straws /gg

"Nvidia on the other hand can push Volta as a single die solution, and THEN still double its potential through the MCM-solutions they have planned."

Lol, you really have no understanding of this do you? Why in the world would Nvidia continue to make pricey large dies if they could make multiple smaller and cheaper ones? The answer: they can't. AMD is the only CPU manufacturer on the market right now that managed to stitch together multiple higher performance CPUs, something not even Intel can do yet with an R&D budget bigger than the value of the entirety of AMD. Nvidia has even less experience with interconnects then either AMD or Intel and you just suddenly expect them to be able to have the same groundbreaking achievement as AMD did? You are delusional. Nvidia doesn't even have an MCM on their roadmap so your looking at volta and then the next architecture AT LEAST still being single.

This is about as far as Nvidia has gotten

research.nvidia.com/publication/2017-06_MCM-GPU:-Multi-Chip-Module-GPUs

of which, Intel has had MCM processors in the past that ultimately failed due to latency between dies. Notice the article's publication date, after Ryzen's launch. This means that Nvidia didn't think MCMs were feasible until after Ryzen. This also means they have zero progress into actually implementing this into their GPUs for real. Ryzen comes out and all of a sudden Nvidia starts talking up MCMs like they are the future. I remember the same thing happening when the original iphone came out and it took others YEARS to catch up.

"Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU."

I don't even know what card you are talking about. WIthout even referencing the card or sources you are just talking out of your ass. Vega 56 = 1070 and Vega 64 = 1080. Their performance are on par with each other. I know you can be talking about the 1080 Ti, because it isn't a year old nor is it mid tier. All of this doesn't change the fact that these cards are still the best Nvidia's got right now. If you didn't realize, your own logic can easily be used against you. For example, The R9 390 was the same price as the GTX 970 and yet the 390's GPU was over 2 years old and still had double the RAM.
If you think that MCM is something that NVIDIA magically pulled out of their ass after Ryzen dropped, then you've got your head in the clouds. It's likely that this has been in R&D for a while now.

Look, MCM is nothing new, neither in the CPU space (Intel had done it before and got knocked hard for it), nor the GPU space (3DFX was doing it, among others). Remember Core 2 Quad? Yeah, MCM, kicked AMD's ass. That's not to say that previous to that AMD's processors weren't kicking Intel's ass.
Posted on Reply
#49
Totally
evernessincemultiple higher performance CPUs, something not even Intel can do yet
*Spits out coffee* What? Remember Core 2, brosef(ina)? iirc those dual cores used a similar approach and were just two single-core processors on a chip.
Posted on Reply
#50
Vayra86
evernessinceLol, people on the internet. Your original comment wasn't about volta in the enterprise, it was about consumer volta. Your just grasping at straws /gg

"Nvidia on the other hand can push Volta as a single die solution, and THEN still double its potential through the MCM-solutions they have planned."

Lol, you really have no understanding of this do you? Why in the world would Nvidia continue to make pricey large dies if they could make multiple smaller and cheaper ones? The answer: they can't. AMD is the only CPU manufacturer on the market right now that managed to stitch together multiple higher performance CPUs, something not even Intel can do yet with an R&D budget bigger than the value of the entirety of AMD. Nvidia has even less experience with interconnects then either AMD or Intel and you just suddenly expect them to be able to have the same groundbreaking achievement as AMD did? You are delusional. Nvidia doesn't even have an MCM on their roadmap so your looking at volta and then the next architecture AT LEAST still being single.

This is about as far as Nvidia has gotten

research.nvidia.com/publication/2017-06_MCM-GPU%3A-Multi-Chip-Module-GPUs

of which, Intel has had MCM processors in the past that ultimately failed due to latency between dies. Notice the article's publication date, after Ryzen's launch. This means that Nvidia didn't think MCMs were feasible until after Ryzen. This also means they have zero progress into actually implementing this into their GPUs for real. Ryzen comes out and all of a sudden Nvidia starts talking up MCMs like they are the future. I remember the same thing happening when the original iphone came out and it took others YEARS to catch up.

"Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU."

I don't even know what card you are talking about. WIthout even referencing the card or sources you are just talking out of your ass. Vega 56 = 1070 and Vega 64 = 1080. Their performance are on par with each other. I know you can be talking about the 1080 Ti, because it isn't a year old nor is it mid tier. All of this doesn't change the fact that these cards are still the best Nvidia's got right now. If you didn't realize, your own logic can easily be used against you. For example, The R9 390 was the same price as the GTX 970 and yet the 390's GPU was over 2 years old and still had double the RAM.
Nvidia has been marketing the large dies and taken good margins on it. From the Titan, to all of its offspring that came after. The 1080ti is just a continuation of that, I'm not sure why you think Nvidia prefers to stop building gaming GPUs above the Gx104 SKU in the future. Nvidia is not bumping into power constraints on the die either, unlike AMD. Basically that's just you applying AMD's GCN limitations to Nvidia's business strategy. It makes no sense at all, Nvidia has headroom everywhere, up to the point that they even cut down their GP100 and build us a GP102 for gaming.

And that is also why it matters that GV100 is already in existence, I'm not even going to spell this out further for you. If you cannot or do not want to see this, what can I say. Do you really think Nvidia is going to accelerate R&D on MCM so they can toss Volta aside for gaming entirely? How is that efficient? These companies are busy with the next best thing, always, and prefer having a new performance bump in the pipeline. That means you suck everything dry until you move on further. Case in point: the 1070ti release, the 9/11GBps updates, the 650ti Boost, 780 and 780ti (hey look, big dies being released in no less than two competitive, volume products) these things happen all the time.

So, I'm delusional and have no understanding (nice, btw), because Nvidia obviously doesn't want to market Titans. Okay, let's move on.

My original comment was about Volta just like yours, except you fail to see how easy it is for Nvidia to cut down a Volta and push it as a gaming GPU. Meanwhile, AMD has NOTHING to cut down, they just have Vega rumors that they are going to feed off for another couple of years, to then release an underwhelming product. You go boy, keep rooting for that.
evernessince"Surely you realize that Vega is ALSO still trailing a full GPU tier behind in terms of gaming performance, right? It's more than 30% behind, so in fact it is last year's high performance of the mid tier SKU."

I don't even know what card you are talking about. WIthout even referencing the card or sources you are just talking out of your ass. Vega 56 = 1070 and Vega 64 = 1080. Their performance are on par with each other. I know you can be talking about the 1080 Ti, because it isn't a year old nor is it mid tier. All of this doesn't change the fact that these cards are still the best Nvidia's got right now. If you didn't realize, your own logic can easily be used against you. For example, The R9 390 was the same price as the GTX 970 and yet the 390's GPU was over 2 years old and still had double the RAM.
First you say you don't know what card, then you confirm the 1070 and the 1080. If you want to act tough, you need to actually make a point. Try it sometime. What you just did here was quote me and then burn me with my own argument, as if that somehow would ever work out well. My own logic indeed. Thanks for repeating it, I guess? And the 1080ti... that one doesn't even need to come out to play, because Vega cannot remotely touch it. So yes, Vega offers last years' high performance bracket, today, at inflated price points.

Then you go ahead and harp on about the 970 versus a rebranded AMD card like it has any relation to the topic, and even though the 970 was a sales record in gaming GPUs and AMD bleeds market share since forever - most notably when they started rebranding to R9 product lines. Meanwhile, AMD had big, expensive dies versus a super lean, heavily cut down failed GM104 - failed even to the point that the cut down was sloppy and Nvidia had to settle a court case for it. And still, it sold more.

Love makes blind, right?
Posted on Reply
Add your own comment
Dec 22nd, 2024 12:44 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts