Tech

Raja Koduri and his adventure at Intel, an in-depth look

It was undoubtedly the most important news of yesterday, Raja Koduri confirmed his departure from Intel, and with it put an end to an adventure of almost six years in which he held positions of great importance in the chip giant, in fact he became executive vice president and the head of Intel’s graphics division.

Raja Koduri’s arrival at Intel was no accident. His departure from AMD to join the ranks of the chip giant came at a time when the importance of the GPU was increasingand in which Intel was fully aware that it had to get to work to return to a market in which it had never succeeded.

We can say that the chip giant was not wrong in its new commitment to the graphics core. The importance of the GPU is greater today than everboth in the general consumer sector and in the professional sector, where it is used to accelerate many workloads and to support the development and evolution of artificial intelligence, inference and deep learning, and this reality will only grow in the next years.

Raja Koduri and his goals at Intel

intel xe

The truth is that from the beginning it was not easy at all, and it was not due to a question of resources or human talent, quite the contrary, Intel is a company that has the necessary means to face any project with guaranteesbut the truth is that this was especially complicated.

To understand it better, we must analyze the key objective that Raja Koduri had to fulfill at Intel. This was develop a GPU architecture, As many of our readers will know, but it did not work with any architecture, it had to meet a series of very specific keys to be able to truly compete with NVIDIA and AMD both in the professional sector and in the general consumer market. I summarize those keys:

  • High performance, at the level of the latest generations of NVIDIA and AMD, which at that time were Ampere and RDNA2.
  • Easily scalable, to be able to design more or less powerful GPUs depending on the needs of the moment, and to be able to create a varied product catalog in a simpler way.
  • Efficient and optimized, with a good performance per watt ratio and adequate support at the software and driver level.
  • Compatible with the latest generation technologies and equipped with hardware specialized in ray tracing and artificial intelligence.

design from scratch A GPU capable of meeting those four objectives was not easy, that’s for sure, and I’m sure that with this simple breakdown of keys you will have been able to better understand the delays and problems that Intel has faced and overcome since it was introduced. embarked on the adventure of returning to the field of GPU-based accelerators and graphics cards.

Arc Alchemist: goal achieved, but halfway

The development of the Intel Xe architecturewhich is the foundation of both professional GPUs and those aimed at the chip giant’s general consumer market, was very complicated and ended up suffering significant delays. These delays meant that the launch of the first commercial products occurred late, in a limited way, and that the support at the driver level was not really optimal.

I think it’s obvious that those delays also played a role in the marketing of Intel Arc Alchemist graphics cards, and had a negative impact on its sales. It is completely normal, because they were released when NVIDIA and AMD already had their new graphics generations ready, the GeForce RTX 40 and Radeon RX 7000, and this reduced the interest and value of the Arc Alchemist series.

Raja Koduri he was very late, this cannot be denied, and when you work for a company of the level of Intel it is clear that there comes a point where it is not tolerable. All in all, and despite the delays suffered by the Arc Alchemist, it must be recognized that the former vice president of Intel’s graphics unit was able to lead his team on the path that made it possible to create an architecture that met many of those targets we’ve seen.

Intel Xe is an architecture highly scalable, since it is possible to join Xe blocks to create graphics cores with higher power. It has also proven to be a powerful architecture and capable of competing with Ampere and RDNA2both for raw performance and for support of advanced technologies, since it has specialized hardware for speed up ray tracing and with specialized units in artificial intelligence, known as XMX arrayswhich are precisely what Intel XeSS intelligent rebuild and rescale technology accelerates.

With Arc Alchemist, important goals were met, but Koduri left things behind along the way. Intel Xe architecture it didn’t scratch a good level in terms of performance per watt consumedcame to the market with strong limitations in games based on deprecated APIs (before DirectX 12 and Vulkan), it lost a lot of performance with processors that didn’t allow ResizableBAR to be enabled, and the Intel XeSS implementation wasn’t as good as expected.

Intel has known how to play its cards and has taken the necessary measures to improve the value of the Arc Alchemist A750 and A770, focusing mainly on improving performance with outdated APIs and reducing the price to adjust to the latest movements of NVIDIA and AMD in the mid-range, and in the end we can say that it has managed to save a release that for some clearly smacked of disaster. No, it hasn’t been a success, that’s clear, especially since that generation of graphics cards came very late, but it’s not fair to call it a failure either, and I think it’s a worthy first step, as far as possible, in the return of Intel to the graphics sector.

Realignment of the graphic division, departure of Raja Koduri and legacy

Raja Koduri Intel

That Intel did not like the results of the graphics division headed by Raja Koduri was made clear when the giant decided to reorganize that division, a move that was accompanied by a demotion in rank that saw Koduri step down from executive vice president. From that moment there was a very marked silence that ended up leading us to the announcement of his departure from Intel, all in a period of just three months.

The official version given by both Intel and Raja Koduri is that it is a friendly outing, and that the latter will embark on a new adventure at the head of his own startup, whose objective will be the development of AI solutions for gaming, multimedia and entertainment. However, it is clear that despite everything the main trigger for that exit has been that objective half-fulfilled that we have told you in the previous section.

In case anyone doubts this, I would remind you that the Intel Xe architecture is also used in high-performance graphics accelerators for the professional sector, known as Ponte Vecchio, and that at this level Koduri was not able to meet all its objectives or the expectations it had generated. Nor can we speak of an absolute failure, but it is clear that did not rise to the occasion, nor the resources and time that Intel put at your disposal.

Intel’s new graphics division will go ahead, and for now it seems to cHe will continue his work based on the legacy left by Raja Koduri. This means that, in principle, there shouldn’t be any major changes in the short- and medium-term future of the chip giant’s GPU strategy, and that the release of Arc Battlemage and Arc Celestial should be assured.

It’s important to note that both of these generations of graphics cards are targeting the general consumer market, and They will use the Intel Xe architecture as a base. On that basis they will apply revisions and changes that should improve performance and efficiency, but there will be no architecture change, so in the end the differences in terms of raw power could be relatively modest.

Arc Battlemage should hit the market between the end of 2023 and the first half of 2024and Arc Celestial would be available at some point in 2025, as long as there are no delays and that Intel can meet its roadmap. In principle we can also expect the same hardware division at the GPU level (shaders, geometry and texturing engines, cores for ray tracing and XMX matrices for AI).

Those two new generations of graphics cards will be the legacy of Raja Koduri at Intel, because as I said they will use the Xe architecture with some modifications and improvements. The big change at the architectural level will not come until the launch of the Arc Druid, scheduled for 2026 at the earliest. Details are still very scarce, as we’re talking about a graphics generation that’s still several years away, so I’m sorry I can’t share more information with you at the moment.

Will Intel be able to succeed in the graphics sector?

Intel-Arc-Battlemage

Focusing on the general consumer market, which in the end is the one that interests the reader of MuyComputer, I think that has achieved very positive things with Arc Alchemist, and that with that generation it has shown that it could really create a GPU capable of running current games and supporting next-gen technologies such as ray tracing and intelligent upscaling.

However, deep down it’s not easy at allAnd I think that it all depends on what you do with Arc Battlemage. This second generation of graphics cards is going to be key for Intel, and its success or failure will decisively determine whether the company should keep working to improve its position in the general consumer graphics space or if it should throw in the towel.

Both professionals and users can understand that a first step is always complicated, and that not everything can be perfect. That’s why I’m so understanding with Arc Alchemist, But now that Intel has experienced that first release, it can’t afford certain things anymore, and we can’t accept certain mistakes either.

Arc Battlemage has to hit the market in a very good state of maturity, both at the level of hardware and drivers and support, and it is imperative that the chip giant cover a broader spectrum of the market and close new deals to bring XeSS to a greater number of titles.

It is also necessary that this new graphic generation be cable to compete with what NVIDIA and AMD have on the market, either by matching or exceeding them in terms of raw performance and technologies, or by offering attractive value for money in performance if it falls short of both (selling well-performing products at a lower price to compensate for their lower performance) .

If you are wondering about the possible abandonment of Intel, I can assure you that this will not happen in the short or medium term, although some rumors pointed to it at the time. There is a compelling reason, and that is that the giant from Santa Clara has invested a lot of moneyresources and time in its adventure within the graphics sector, and throwing in the towel would be the equivalent of wasting all those resources.

On the other hand, you should keep in mind that the GPU sector is extremely important, and that Intel needs this component at different levels, ranging from general consumer dedicated graphics cards to integrated graphics cores and graphics accelerators for the professional sector. In the end there are many fronts, and a company like Intel cannot afford to be left out of the race in the GPU sector. However, this does not mean that you cannot take a break to face certain challenges or new development cycles in the long term, obviously.

Related Articles