The Emphasis is on Streamlining and Market Share, rather than NVIDIA
An interview with AMD’s Jack Huynh at IFA 2024 revealed some intriguing developments, such as amalgamating AMD’s two architectures – the client RDNA and datacenter CDNA – into a fresh UDNA (Universal) strategy, akin to NVIDIA’s CUDA methodology. Unfortunately, there are concrete details indicating that AMD’s forthcoming lineup of desktop gaming cards will not aim to rival NVIDIA in the high-end segment.
Jack Huynh, as quoted by Tom’s Hardware:
“One of the major shifts at AMD is that we currently have a CDNA architecture for our Instinct data center GPUs and RDNA for consumer products. It’s bifurcated. Moving forward, we’re transitioning to UDNA. We will have a singular architecture for both Instinct and client products. By unifying these architectures, we aim to simplify the developer experience compared to the current scenario where selection is necessary and value is stagnant.
We forked the architectures for the sake of sub-optimizations and micro-optimizations, complicating matters for developers, especially with the growth of our data center business. Therefore, unification is imperative. I constantly reiterate to the team that we are looking at millions of developers as our target. We need to scale from hundreds to thousands, tens of thousands, hundreds of thousands, and hopefully, one day, millions. That’s the goal we’re working towards.”
When questioned about whether this strategy implies that future consumer graphics cards will adopt the same architecture as products like the Instinct MI300X, Huynh responded:
“It’s a cloud-to-client approach. This approach also enhances our efficiency. Instead of having separate teams, we will have one unified team. It’s not a groundbreaking concept. We pursued bifurcation for short-term micro-optimizations, but with the current scale, reunification is crucial, and I believe it’s the correct path to take. There may be minor challenges, but they are surmountable.”
The “cloud-to-client strategy” seems to lean more towards virtualization rather than a hardware-centric approach. However, there is no indication that AMD plans to cease producing physical gaming GPUs after transitioning to UDNA.
You can delve into the full interview transcript of Paul Alcorn’s conversation with Jack Huynh on Tom’s Hardware by following this link.
Regrettably for Radeon enthusiasts, another segment of the discussion on consumer graphics at Tom’s Hardware conveys that AMD has no immediate intentions to target the premium gaming GPU market. This decision seems sensible given that NVIDIA currently dominates the high-end segment, which comprises a relatively small portion of the overall graphics market.
Jack Huynh, as relayed by Tom’s Hardware, in response to a query on whether AMD will re-enter the premium gaming GPU segment:
“We might consider it in the future. However, my primary objective presently is to establish scalability for AMD. Without substantial scale, attracting developers is challenging. If I were to inform developers that ‘I am only aiming for a 10% market share,’ they would opt for NVIDIA. I need to present a strategy that demonstrates we can attain a 40% market share. Once that is achieved, developers will align with us, and optimization for AMD will follow suit. After achieving this, we can pursue the top tier.”
For the full interview, visit Tom’s Hardware by clicking here.
