After an extended wait, the way forward for AMD’s perfect graphics playing cards is in any case right here. The corporate simply introduced the RX 9000 collection, constructed from two new GPUs: The RX 9070 XT and the RX 9070. Each are set to release on March 6. Whilst the GPUs sound exciting, I’m left with a number of questions even if AMD really stuffed within the gaps on RDNA 4.
AMD first of all teased RDNA 4, or RX 9000 collection, right through CES 2025. Then again, it took just about two months for extra concrete data to reach, and nowadays, it’s all in any case right here.
AMD’s focal point all through the presentation was once the truth that players purchase mainstream GPUs extra frequently than behemoths alongside the traces of the RTX 5090. To that finish, AMD cited its personal analysis, pronouncing that 85% of players purchase GPUs that price lower than $700. On the identical time, upper resolutions are slowly changing into mainstream, with increasingly other people purchasing 1440p and 4K displays. The desires of players have long past up, now together with forged efficiency in ray tracing, however preserving it reasonably priced is vital too.
Get your weekly teardown of the tech in the back of PC gaming
That’s the place my first query is available in: How a lot those GPUs will price. AMD didn’t expose that key piece of knowledge right through the clicking briefing, nevertheless it’ll arise right through the development, at which level we’ll replace this newsletter. Till we all know the cost, it’s exhausting to invest about the place RDNA 4 will land amongst competing playing cards, however no less than we have now a large number of specifications and a few benchmarks to head over right here.
The specifications of RDNA 4 were leaked months in the past, and it seems that leakers were given a large number of it proper. The flagship RX 9070 XT serves up 64 compute devices (CUs), 64 {hardware} ray tracing (HW RT) accelerators, and 128 AI accelerators. The spice up clock just about hits 3GHz, maxing out at 2.97GHz, this means that that overclocked GPUs will without a doubt pass previous that threshold. The cardboard comes with 16GB of VRAM throughout a 256-bit bus, and this reminiscence configuration additionally applies to the RX 9070 non-XT. The flagship has a complete board energy (TBP) of 304 watts.
The RX 9070 cuts again on each spec, scaling right down to 56 CUs and a most clock velocity of two.52GHz. However, the facility intake may be considerably decrease, now at 220 watts.
Each GPUs will use GDDR6 reminiscence. AMD selected to stick with GDDR6, clocked at 20Gbps, which it discovered to provide the most efficient stability of efficiency and value. Nvidia’s RTX 50-series already made the transfer to GDDR7, having used GDDR6X within the earlier technology, however AMD offsets for the lack of bandwidth by way of including a lot of VRAM to its GPUs.
AMD printed some benchmarks for each GPUs, and so they appear lovely nice. Then again, when evaluating to Nvidia, it used an RTX 3090 versus one thing from a present technology. AMD defined that this was once as a result of individuals who purchased the RTX 30-series can be having a look to improve now, however it could’ve been great to peer some ratings for that.
That is every other query I used to be left with right through the presentation — how will the brand new playing cards stack up in opposition to Nvidia equivalents? Thankfully, AMD printed that it expects the RX 9070 XT to be on par with Nvidia’s $750 RTX 5070 Ti.
Regardless, the RX 9070 is alleged to provide as much as 21% quicker efficiency than the RX 7900 GRE at 4K, which is a sizeable uplift. In the meantime, the RX 9070 XT can be as much as 42% quicker at 4K, whilst additionally appearing forged ray tracing enhancements.
With the brand new compute unit (CU) in RDNA 4, AMD was once in a position to spice up potency, clock speeds, and sign in allocation. The brand new graphics playing cards will come provided with third-gen ray tracing accelerators, which is able to be offering 2x ray tracing throughput in line with CU. Traditionally, AMD trailed in the back of Nvidia in ray tracing, nevertheless it’s indisputably making enhancements in that regard.
The brand new GPUs will even include second-gen AI accelerators, which AMD guarantees will lead to as much as 8x quicker AI efficiency when in comparison to RDNA 3. The die dimension is at about 350mm2, and the transistor rely at round 44 billion. AMD additionally printed that the GPU was once constructed on a 4nm procedure. The RX 9000 collection makes the transfer to PCIe 5.0, and it additionally provides DisplayPort 2.1a and HDMI 2.1b toughen.
The corporate may not be making its personal variations of the playing cards, sometimes called MBA (Made Via AMD), nevertheless it says it’ll paintings carefully with its companions to make sure availability, and extra importantly, availability at MSRP.
Along the RX 9000 collection, AMD additionally introduced FSR 4, which is able to stay an RDNA 4-exclusive in the meanwhile. AMD has prior to now stated that it wish to open up FSR 4 to different playing cards, nevertheless it’s unclear when or if that would possibly finally end up going down.
