AMD plans major changes for Radeon with united UDNA GPU architecture for data centers and gamers
AMD has a new GPU roadmap – UDNA is replacing both RDNA and CDNA
The GPU market has changed a lot since AMD introduced their domain-specific RDNA and CDNA graphics architectures. For starters, the growth of AI has changed what gamers and developers expect from GPUs. Additionally, gaming workloads have become increasingly compute-focused. Now, AMD has confirmed its plans to move to a new GPU architecture, one that will span both gaming and enterprise. AMD has announced their unified UDNA GPU architecture.
In the past, AMD created RDNA and CDNA as GPU architectures focused on gaming and compute, respectively. Both architectures benefitted from micro-optimisations for each market, but this has led to disunity within AMD’s GPU hierarchy. The GPU market has changed, and data-center and gaming GPUs need not be dissimilar, especially now that AMD has an established presence in the data-center GPU market.
Both RDNA and CDNA have been successful within their markets. That said, the split is now causing troubles for AMD. With UDNA, developers will have one unified architecture to work with, making it easier to optimise software for all AMD GPUs.
Below is what AMD’s Jack Huynh had to say about UDNA.
AMD: So, part of a big change at AMD is today we have a CDNA architecture for our Instinct data center GPUs and RDNA for the consumer stuff. It’s forked. Going forward, we will call it UDNA. There’ll be one unified architecture, both Instinct and client [consumer]. We’ll unify it so that it will be so much easier for developers versus today, where they have to choose and value is not improving.
We forked it because then you get the sub-optimizations and the micro-optimizations, but then it’s very difficult for these developers, especially as we’re growing our data center business, so now we need to unify it. That’s been a part of it. Because remember what I said earlier? I’m thinking about millions of developers; that’s where we want to get to. Step one is to get to the hundreds, thousands, tens of thousands, hundreds of thousands, and hopefully, one day, millions. That’s what I’m telling the team right now. It’s that scale we have to build now.
AMD’s Jack Huynh to Tom’s Hardware
RDNA and CDNA made sense for AMD at the time, but UDNA makes sense for today’s GPU market
Both RDNA and CDNA have been successful. Today, AMD has strong sales in the datacenter GPU market with its latest versions of CDNA. As for gaming, AMD is in a stronger position today than it was with its pre-RDNA Vega architecture. Regardless, the disunity of AMD’s GPU architectures has its problems. For starters, AMD’s RDNA gaming GPUs tend to fall behind their competitors in workstation workloads (like video editing). Beyond that, the difference between AMD’s gaming and compute-focused chips makes it hard for developers to optimise for both. Moving forward, AMD is working on a unified GPU architecture, which should help it gain developer support. Additionally, it should unify AMD’s GPU engineering resources, which could result in a stronger product lineup for users.
Sadly, it’s unknown when AMD’s first UDNA GPUs will come to market. It will likely be years before we see UDNA products. That means that we could still see a few more generations of both RDNA and CDNA. Today’s AMD has a cloud-to-client strategy, just like Nvidia.
You can join the discussion on AMD’s new UDNA GPU architecture plans on the OC3D Forums.
