Intel reorganizes its GPU division, remains committed to Arc

[Daniel Sims]

Why it matters: The troubled launch of Arc Alchemist has invited rumors that Intel might abandon its attempts to become the third pillar of the discrete graphics card market. The company denies these rumors, even as internal reorganization intensifies speculation over the fate of Intel's future GPUs.

Intel recently announced that it's splitting its Accelerated Computing Group (AXG) in two to focus better on the consumer and data center graphics markets. Raja Koduri, AXG's founder, will once again be Intel's Chief Architect to coordinate the CPU, GPU, and AI divisions.

The group's consumer graphics team – in charge of the Intel Arc discrete graphics cards – will join Intel's Client Computing Group. The accelerated computing teams will join the data center and AI group – another important enterprise-side market for graphics processing. Nvidia dominates both markets.

It's easy to interpret this reorganization and reassignment as a bad sign for Arc GPUs, but Intel maintains it has no plans to sunset the brand. The company says it will keep supporting the Alchemist line and still plans to launch its successors – Battlemage and Celestial – in the coming years.

Intel had to delay the Alchemist graphics cards multiple times before shipping them in the middle of 2022, and they haven't significantly impacted Nvidia's and AMD's market share. Intel's Arc 380 proved disappointing compared to other low-end GPUs like AMD's Radeon RX 6400 or Nvidia's GTX 1650. Meanwhile, Intel's A750 and A770 compared somewhat favorably to Nvidia's mid-range cards from one and two years prior, suggesting the delays put Intel behind the curve.

Upon launch, one of Arc Alchemist's most prominent weaknesses was its disappointing performance in DirectX 11 and DirectX 9 games. Intel mainly focused on DirectX 12 performance for playing the latest titles, but many recent games still use DirectX 11. Counter-Strike: Global Offensive – the most popular game on Steam despite its age – still uses DirectX 9. The company has had to play catch-up to improve performance in older games, but its extensive driver updates to that effect suggest Intel isn't giving up yet.

The chip giant told Tom's Hardware that while it's committed to shipping Battlemage, it won't set any firm release dates yet. The delays that plagued Alchemist taught Intel to be more cautious with promises as it enters a new arena.

Permalink to story.

https://www.techspot.com/news/97081-intel-reorganizes-gpu-division-remains-committed-arc.html

 

[OortCloud]

I still think for a first pass, the Intel GPUs are a good effort. It was always going to be a bit of a clusterf**k.
There's just so much that goes into a GPU and its drivers.
Per-game optimizations, Per-engine optimizations, DXn/Vulcan optimisations, Ray Tracing, Rasterisation, Pixel Shaders, Video Playback, Motherboard compatibility, Monitor Compatibility, HDMI, DP versions, HDCP etc etc etc

Having overcome a large part of that initial hurdle it seems madness they would give in now when they are in sight of the opposition and so close to becoming a genuine player in a market that seems to be gaining so much importance and relevance in so many fields. Not just for playing games, but also as a modern-day math co-processor for modelling, AI, etc etc.

 

[Tom Yum]

I still think for a first pass, the Intel GPUs are a good effort. It was always going to be a bit of a clusterf**k.
There's just so much that goes into a GPU and its drivers.
Per-game optimizations, Per-engine optimizations, DXn/Vulcan optimisations, Ray Tracing, Rasterisation, Pixel Shaders, Video Playback, Motherboard compatibility, Monitor Compatibility, HDMI, DP versions, HDCP etc etc etc

But very little of that is a first time for Intel, as they've been developing iGPUs with nearly all those elements (bar ray tracing and Vulkan) since at least GMA X4500 in 2008. Vulkan/full DX12 support has existed since Skylake in 2015.

Now one could argue that optimising game performance was never a priority for Intel with their iGPUs, leaving all those lessons learned on game optimisation for Alchemist. But to my mind, that is an Intel business decision (not a technical one) and something fully in their control, which means the consequences fully theirs to own. Personally, that doesn't earn them a pass.

 

[kiwigraeme]

They have deep pockets
Best they concentrate on low end to mid consumer gpus and special purpose mathematics , matrices, AI etc
going after the high end you have the AMD problem.

If I was AMD I would go after the middle bigtime - turn Nvidia into a niche expensive high end.
yes big margins on high end - but like airlines you need those economy class to cover costs ( production , drivers etc ) - plus these cards will be the PS6 etc ( top of midclass for that year )
Also if bulk of people own midclass AMD cards developers will favour them - Nvidia 5090 will look after itself

 

[Watzupken]

But very little of that is a first time for Intel, as they've been developing iGPUs with nearly all those elements (bar ray tracing and Vulkan) since at least GMA X4500 in 2008. Vulkan/full DX12 support has existed since Skylake in 2015.

Now one could argue that optimising game performance was never a priority for Intel with their iGPUs, leaving all those lessons learned on game optimisation for Alchemist. But to my mind, that is an Intel business decision (not a technical one) and something fully in their control, which means the consequences fully theirs to own. Personally, that doesn't earn them a pass.

I don't feel iGPU and a proper dedicated GPU is the same. Intel's real effort in gaming grade iGPU is the XE graphics. Even so, because its iGPU, the effort is still not the same because games not running properly in iGPU is still somewhat forgivable, as compared to a GPU that's targeting gamers. In my opinion, Intel did well for a first attempt, especially when you considered previous effort in the form of Larabee did not even make it into production. And performance of the first gen ARC is pretty good. The main issue is the timing, which basically means their first time effort is going head to head with AMD and Nvidia's next gen GPUs. The same will happen if Battlemage is going to be released in 2024. So not telling people the release date of Battlemage is not going to help Intel become more competitive.

 

[Warrior247]

There was a lot of bad decision making, and oversight here with Intel. I wanted to buy this card, but there is no real reason to buy it. It’s a day late and a dollar short.