Ye, go ruin another company.

There goes Tenstorrent......

Is Jim trying to destroy his own company by allowing Raja in? What was he thinking? Jim, if you see this message, hire me to take Raja's place. I have 0 experience in the field, but I won't sink the company like Raja will.

I dont know why Raja was seen as someone singularly capable of helping companies develop and launch miracle products on his own. Its not a one man show. He must work within a companys development structure, design teams, allocated resources, deadlines and multiple other steps and processes that can impact the way a final product turns out. Its a collaborative effort that depends on so many factors to get things right, not just one man. Pretty sure Tenstorrent know best as to what value he may provide them. And I would think he's relieved not to be thrust into a similar situation as with AMD or Intel where too much was expected of him in terms of timely product releases.

alanm:

I dont know why Raja was seen as someone singularly capable of helping companies develop and launch miracle products on his own. Its not a one man show. He must work within a companys development structure, design teams, allocated resources, deadlines and multiple other steps and processes that can impact the way a final product turns out. Its a collaborative effort that depends on so many factors to get things right, not just one man. Pretty sure Tenstorrent know best as to what value he may provide them. And I would think he's relieved not to be thrust into a similar situation as with AMD or Intel where too much was expected of him in terms of timely product releases.

How long have you been his press secretary? https://media3.giphy.com/media/98jU7NxuNSSZ2/200w.gif?cid=ecf05e47c2vfqqbibj38mndsmpsj3tbbqzkp872l3znyme0e&rid=200w.gif&ct=g

He will hire all his friends like he always do, and "quit" in 3 to 4 years

NewTRUMP Order:

How long have you been his press secretary?

How long have you been towing the line and not thinking independently?

alanm:



How long have you been towing the line and not thinking independently?

You are the only one here who defends him that I have seen, and that is funny. He left AMD, and bam, GPU division takes off. He sucks. Intel GPUs literally are Vega 2.0 after him. He screws up any company he goes too. The dude literally sold himself as the savior of AMD gpus, and then to be the fore bring of Intel GPUS to be a mainstay. He brought all the expectations upon himself. Your sad assed simping for him is cringy, and hilarious.

Agonist:

You are the only one here who defends him that I have seen, and that is funny. He left AMD, and bam, GPU division takes off. He sucks. Intel GPUs literally are Vega 2.0 after him. He screws up any company he goes too. The dude literally sold himself as the savior of AMD gpus, and then to be the fore bring of Intel GPUS to be a mainstay. He brought all the expectations upon himself. Your sad assed simping for him is cringy, and hilarious.

Honestly, while I'm not impressed with his work and is not someone I would regard successful, he's struck quite a lot of bad luck. With AMD, Vega never became what he wanted, because this was pre-Ryzen days and AMD was nearly bankrupt. They diverted resources toward PS and XB, while hoping Vega's raw performance would be appealing to workstations and servers (it wasn't, really). Had AMD committed to finishing Vega's development, it probably would've turned out a lot better. When it came to Intel, he was doomed to fail. Arc would not have been successful regardless of who did it. It could have been a financial success if it were released a half year earlier (due to miners) but otherwise it was an overwhelming undertaking. Contrary to what a lot of people here believe, Intel has had decent GPUs since Sandy Bridge, but they didn't bother to optimize anything and [on Windows] they would abandon support the moment a new chip came out. For an unoptimized architecture, Xe was pretty damn good compared to AMD's APUs. Arc is derived from Xe, so it's not exactly the greatest engineering challenge. The problem came down to Intel trying to compete in a market where they just didn't have polished drivers. They could have just kept throwing money at the problem, but not if investors had anything to say about it. So in both situations, I don't see how he was ever supposed to look good. At least with Tenstorrent, he would have an easier time. I don't think the architecture for an AI processor is as complex as a GPU, and there isn't decades worth of optimization to do. So - as long as Tenstorrent isn't starving for funding (which honestly, it might if knowledgeable investors know Raja is part of the team) this ought to be a true test of what he can actually do.

Agonist:

You are the only one here who defends him that I have seen, and that is funny. He left AMD, and bam, GPU division takes off. He sucks. Intel GPUs literally are Vega 2.0 after him. He screws up any company he goes too. The dude literally sold himself as the savior of AMD gpus, and then to be the fore bring of Intel GPUS to be a mainstay. He brought all the expectations upon himself. Your sad assed simping for him is cringy, and hilarious.

I have nothing for or against Koduri because I am in no position to ascertain what value he provides to whomever employs him. Obviously those that do employ him see far greater value in him than random forum posters do.

alanm:

I dont know why Raja was seen as someone singularly capable of helping companies develop and launch miracle products on his own. Its not a one man show. He must work within a companys development structure, design teams, allocated resources, deadlines and multiple other steps and processes that can impact the way a final product turns out. Its a collaborative effort that depends on so many factors to get things right, not just one man. Pretty sure Tenstorrent know best as to what value he may provide them. And I would think he's relieved not to be thrust into a similar situation as with AMD or Intel where too much was expected of him in terms of timely product releases.

track record in recent years hasn't been very good, Its not so much as a "one man show" as him impeding products via judgement errors. Used to think maybe he was just unlucky after what happened with vega, but it seems more like he was causing problems via poor decisions/ taking too much risk, Vega obviously wasn't a complete failure since its still the basis for line of CDNA accelerators, but it definitely had some serious shortfalls likely due to management issues, for instance Primitive shaders, Vega hardware supports it, and it supposedly works, but the cost of implementing it on the software side is high on vega due to some hw bug/quirk/poor implementation( I remember reading on the amd drm(linux) mailing list some time ago, that switching between NGG and the legacy geometry causes vega 10 to hang or something to that effect) , from the outside it looks like he expected the software team to "figure it out", without assessing the full impact, since he was still tweeting about enabling features on vega after launch, which of course never ended up happening , and as a result it hurt amd in the consumer gpu space. At that level his most important job was basically to "not make costly mistakes", and it seems he made similar mistakes at the arc/xe division, given the state they are in. we know the people at amd and intel are very competent and capable and that both companies have tons of IPs to pull from, so if they aren't able to work together properly and bring a product to market that is satisfactory, fault generally lies at the leadership . Its important to note that people who are technically excellent and competent, do not always fill leadership roles well. IE , He may fit perfectly on the tenstorrent board of directors, but not so well as chief architect for a gpu division.

It's cool to make the guy a scapegoat but really Arc isn't a failure - it's a first gen gpu you can't expect it to equal Nvidia/AMD. It works, it's got enough performance to play games, drivers not surprisingly were very shaky but improving, and it's selling. Sure they didn't hit it out of the park but it's a good basis to build on if Intel don't shelve it to do AI instead. I don't suppose real companies see him as a failure at all - that's why he's go this job.

Dribble:

It's cool to make the guy a scapegoat but really Arc isn't a failure - it's a first gen gpu you can't expect it to equal Nvidia/AMD. It works, it's got enough performance to play games, drivers not surprisingly were very shaky but improving, and it's selling. Sure they didn't hit it out of the park but it's a good basis to build on if Intel don't shelve it to do AI instead. I don't suppose real companies see him as a failure at all - that's why he's go this job.

In the case of ARC , the consumer gpus are ok, but they cancelled a generation of their compute accelerators (rialto bridge), which isn't very good especially since they were close to launch. and now they have effectively delayed next gen xe (compute and consumer) for a whole extra year , this all comes shortly after raja left , the timing of which probably isn't a coincidence. it is possible that this is entirely due to the delay of intel's 4n process, and has nothing to do with raja, but I would be skeptical, since this would be like the 4th or 5th time a process node delay caused a cancellation/delay at intel in the last 10 years. raja himself has a skillset that few people have , he is very difficult to replace, I'm sure intel would have rather kept him on in a different role as they tried to do. he isn't a failure, but that doesn't mean he can't make mistakes. one can look to the fall of 3dfx, SGI, sun microsystems, where business mistakes are made by very competent people, mistakes are easy to make in new industries, modern graphics/compute (programmable pipeline) is <25 years old afterall.

I am curious to see what development will be done on RISC-V.