Where Was RDNA4 at AMD’s Keynote?
Timing is Everything
Before AMD’s keynote at CES 2025 had even concluded, I was already being asked by the press around me in the same room what happened to all the RDNA 4 announcements. AMD 0.00%↑ said that partners are expected to show new RDNA 4 hardware at the trade show. So why didn’t AMD present it?
For those of you unaware, AMD’s graphics architecture is called RDNA, and the fourth generation of this was expected to be announced today, at the CES trade show. It was expected so much that in the pre-briefing before the event, where the press have to sign embargoes to not share the content until the date, the slides that AMD presented even detail new graphics cards based on RDNA4 and their naming. The 45 minute keynote, headed up by EVP Jack Huynh and covering the new CPU and mobile launches, had one mention of an RDNA4 feature, but aside from that, nothing.
The media, and the internet through reddit and social media, have been clutching at straws. Any new launch of a new graphics architecture is a big deal – in time before, it was the biggest launch of a company of the year. For what it’s worth, my reasoning as to why was simply one of timing and alignment and depth – five minutes at the end of a keynote would not have enough information to satisfy the audience, or provide to little detail to make it exciting. A new design like RDNA4 requires depth and nuance, to go into new details for ray tracing, machine learning, graphics performance, power issues – in the past, such as for launches of RDNA3, RDNA2, Polaris, AMD has had dedicated tech days and events to show the info. They’ve even had CPU tech days as part of an event like CES. However nothing like this was planned around CES, so without sequestering the press and analysts for a day or two, they wouldn’t be able to share that information in depth.
In order to get actual answers, rather than my self-analyst informed conjecture, AMD invited a small circle of press for an after-the-keynote open discussion about why RDNA4 was not presented. At the round table were:
Frank Azor, Chief Architect of Gaming Solutions and Gaming Marketing, AMD
David McAfee, CVP and GM, Client Channel Business, AMD
Tim Schiesser, Hardware Unboxed
Balin McKinley, Hardware Unboxed
Ian Cutress, More Than Moore
Dave James, PC Gamer
Michael, TechPowerUp
Here are the salient bullet points.
David and Frank stated that there will be news from partners about cards at CES - however the material shown in the pre-brief is the extent of the news right now. That includes the naming scheme and showing off some of the cards. When it comes to the cycle, AMD is close to disclosing information about RDNA4, the architecture, and the configurations, however for a couple of reasons, they felt CES was not the right place to give that information.
First, the keynote, being an official CES keynote, was limited to 45 minutes. Spending only 5 minutes at the end speaking about RDNA4 wouldn’t do the new architecture justice. AMD felt that the audience would be be left with more questions than answers, and in the past previous graphics architecture launches where met with a day of disclosures, rather than 5 minutes.
Second, is the timing. While AMD is close to launch, they’re perhaps not as close as when a normal launch event would be. AMD stated that they will be planning a dedicated set of contacts around all the features of RDNA4 for press and analysts in an official launch soon, but that date isn’t today.
I asked if the delay was because of any manufacturing or supply issues. AMD said no.
I asked if the delay was due to aligning with what NVIDIA might be launching. AMD said no.
Dave from PCGamer and Tim from Hardware Unboxed asked about the naming alignment scheme - going from 7700 to 9070 for a 7-series - and if this was going to remain consistent going forward.
AMD stated that the move to 9070 where the 7 is the third digit is due to market confusion when comparing against AMD. With a 70 vs 70, AMD argued that users no longer needed a significant investment into the knowledge of the brand to make better like-for-like comparisons. This extended to retail and etail as well. As for the naming scheme changing frequently, while AMD didn’t commit to say it wouldn’t change again (they said it does depend on what happens in the market, and they want to remain agile), they recognise that changing the brand away from conventional wisdom introduces friction to consumers. They stated that part of desktop Ryzen’s success is likely due to the alignment in naming with conventional schemes, and that Radeon wants to replicate that success.
Tim asked if the alignment between 70 and 70 and others would be based on performance, or price. Frank and David said it would be a combination of both. Performance per dollar needs to be a driving factor, but Frank highlighted that the focus point is enabling a structure and feature set for a given price that makes sense to the user. They reiterated going after the major market, the sub-$1000 market.
I asked about the big positive response to Intel’s $250 Battlemage offering. David had positive things to say - highlighting that when you provide the market with a cost effective option that resonates, it will do well. He then reiterated that Ryzen went on a similar journey, which has helped the brand get to where it is today. Frank stated that when it comes to the Radeon side, AMD has had similar success with specific cards in certain price bands, and it’s a case of replicating those for the new generation.
Dave from PC Gamer asked about iGPU naming, and if Strix Halo and derivatives will follow the same naming as RDNA4. David and Frank said that beyond Strix Halo hasn’t been announced yet, so they’re not going to comment at this time.
Tim from Hardware Unboxed asked about an assumption based on a footnote that might have suggested that FSR4 was going to be 9070 exclusive (or minimum) and asked for clarity. David stated that there’s nothing about FSR4 that makes it exclusive to a specific model - the RDNA4 updates include better MLOps that makes technology like FSR4 a lot better than it would have been. We should expect AMD, as it rolls out FSR4, to lean into the Navi 4 architecture capabilities. Tim then asked for confirmation that FSR4 would improve performance, and David confirmed that compared to the analytical approach of FSR3, it’s a significant improvement.
Michael from TechPowerUp asked about the top priorities for RDNA4 were. David gave a top 3:
Improving performance in the areas that gamers care about most - ray tracing, ML Ops for FSR4 and ML Super Resolution
The architecture is designed for efficiency - not just performance per watt, but also manufacturing decisions, design decisions, all optimized to deliver the best value to the end user.
Go to market - driving transparency in terms of positioning, performance to the end user, and then pricing the product so that the Radeon brand delivers.
Dave from PC Gamer asked about AMD’s ability to scale up manufacturing if the product line is a major success. Frank stated that no single GPU launch is going to triple market share over night, but AMD is good at modelling demand and can manage an uptick.
The final question came from AMD’s own, wanting to find out about if the benchmark numbers on the internet were true.
Frank Azor
All the data you have is pretty much inaccurate.
David James – PC Gamer
So all the data that's out there about RDNA4 is wrong?
David McAfee
The performance data that's out there on RDNA4 is completely inaccurate.
Frank Azor
The other thing I will tell you is that nobody has the final driver. So how can the data be accurate? That's why I can confidently say that the data is not accurate.
It was a good talk. I’m looking forward to an RDNA4 architecture disclosure.
thank you for the write-up. i would love to hear more about your thoughts on the matter.
That was informative.