Was Intel's ARC "Failure" Raja's fault?

Of course the ARC problems were not all the fault of Raja Koduri.

Managers can have a big indirect effect, affecting who is hired, how development is organised, how much employees feel empowered, what backup plans there are, etc.

It's impossible to say what kind of beneficial or detrimental effects Raja Koduri had on ARC's development and release without quite a few first hand reports of what it's like to work under him.
 
I attribute Intel's "failure" to two key decisions:

1-Deciding to jump staright to discrete instead of developing ever greater integrated graphics and then scaling it up to discrete once they reached good enough architecture and specially drivers.

2-Deciding to launch it so soon instead of giving it the time it clearly needed to mature.

I don't believe money would be a problem for intel and I really can't say how much of this is Raja's fault but I honestly guess it jsut isnt, this just smells to me like board of directors forcing the GPU division's hand.

Ultimately, I do think Raja could have done a better job but we also have to ask if it is reallistic that he did so, we can't know the real situation inside the dicrete GPU division and can only guess. In reality the ability of one person can be severely impaired when if they are being ordered around to do things they wouldn't otherwise, and we don't know how much of that he and his team had to endure, and/or from who the orders came from. I just hope intel insistis for a while longer on making their discrete GPUs a success, and actually manage to do so.

Also, honestly I really thing ARC is a strong first architecture, and if given time and investment their second generation ARC GPUs could very well be great. It's really a shame they didn't just keep it internally and released only ARC 2.
 
Adored TV made a very good video some time back about alot of things that go on behind doors at Intel so Raja was only one part of the issue.

Overall I feel progress was made and in the end not a total disaster.

I just feel they will hold until they now have something good run it on a lower size team, cause in all honesty if they don't get something going really well soon they are going to be in a much worse situation.
 
1-Deciding to jump staright to discrete instead of developing ever greater integrated graphics and then scaling it up to discrete once they reached good enough architecture and specially drivers.

2-Deciding to launch it so soon instead of giving it the time it clearly needed to mature.

1. Intel iGPUs surpassed AMD's Vega ones. I don't think that Intel would have gotten anywhere with faster iGPUs. They are still much weaker than dGPUs, don't have their own RAM and no need to communication over PCIe, and are just different beasts altogether.

2. Intel might have done even worse then. In retrospect, yes, NVIDIA and AMD haven't bothered even with mid-range on the current lineups yet. But Intel was already very late. Postponing release even further wouldn't have done it any favours.
 
1. Intel iGPUs surpassed AMD's Vega ones. I don't think that Intel would have gotten anywhere with faster iGPUs. They are still much weaker than dGPUs, don't have their own RAM and no need to communication over PCIe, and are just different beasts altogether.

2. Intel might have done even worse then. In retrospect, yes, NVIDIA and AMD haven't bothered even with mid-range on the current lineups yet. But Intel was already very late. Postponing release even further wouldn't have done it any favours.

1- most of intel's problems are related to the way their drivers handles the calls or order the cores, etc that's not discrete GPU related, that's universal, they would have found and fixed that if they ever cared enough about it and would be in a much better place to adapt the driver to a dGPU, additionally, while yes, iGPU is quite different from dGPU you're so so wrong in undermining the value an experience with building actually competitive iGPUs would have in developing a competitive dGPU, a lot of the experience is applicable and AMDs iGPUs that perform so well are based on dGPU architectures.

2-I don't think you understand what I meant, I meant NOT launching ARC at all, it was simply not ready for prime time, ARC issues while mostly related to drivers are Not all driver related, there are at least one critical design issue that cannot be overcome by drivers, they had already waited 5 years developing it, I do understand they trying to hit the perfect window for launching around when GPUs from AMD and NVIDIA were crazy expensive but they still had to have a good product, and they didn't, they still don't! And ARC will never be good thanks to the architecture issues, it might get somewhat decent, but not good. it needs a second generation to fix all issues and if they missed their perfect window, then the ability to recognize that and take a step back to make a good product would be definitely welcome, they would not fare worse, with a good product they would obviously fare better launching a bad product when gamers have so much good options for cheap on the used market would never be a good idea, launching a good product at competitive prices when the market is in a stable position would at least warrant some sales aside from SIs, in That case they would not launch ARC, they would only launch ARC 2 (even if they needed up calling it just ARC), and that would happen only next year assuming the market does settle down next year.
 
Back
Top