Monday, July 21st 2025
AMD's Upcoming UDNA / RDNA 5 GPU Could Feature 96 CUs and 384-bit Memory Bus
According to one of the most reliable AMD leakers, Kepler_L2, AMD's upcoming UDNA (or RDNA 5) GPU generation will reintroduce higher-end GPU configurations with up to 96 Compute Units (CUs) in the top-end Navi 5X SKU, paired with a 384-bit bus for memory. We still don't know what type of memory AMD will ultimately use, but an early assumption could be that GDDR7 is on the table. When it comes to the middle of the stack, AMD plans a GPU with 64 CUs and a 256-bit memory bus, along with various variations of CUs and memory capacities around that. For the entry-level models, AMD could deliver a 32 CU configuration paired with a 128-bit memory bus. So far, memory capacities are unknown and may be subject to change as AMD finalizes its GPU lineup.
After the RDNA 4 generation, which left AMD without a top-end contender, fighting for the middle-end market share, a UDNA / RDNA 5 will be a welcome addition. We are looking forward to seeing what UDNA design is capable of and what microarchitectural changes AMD has designed. Mass production of these GPUs is expected in Q2 2026, so availability is anticipated in the second half of 2026. A clearer picture of the exact memory capacity and type will emerge as we approach launch.
Source:
Kepler_L2
After the RDNA 4 generation, which left AMD without a top-end contender, fighting for the middle-end market share, a UDNA / RDNA 5 will be a welcome addition. We are looking forward to seeing what UDNA design is capable of and what microarchitectural changes AMD has designed. Mass production of these GPUs is expected in Q2 2026, so availability is anticipated in the second half of 2026. A clearer picture of the exact memory capacity and type will emerge as we approach launch.

201 Comments on AMD's Upcoming UDNA / RDNA 5 GPU Could Feature 96 CUs and 384-bit Memory Bus
As for OEMs, they go with the strong brand that can warranty supply and sales. Intel had the fabs and the name, so AMD was winning in the CPU DIY market, but not the OEMs, Nvidia had the advantage of being the company with the strongest, more premium name in the market, proprietary stuff that tech press was promoting and eventually, with Pascal, clearly the faster hardware. Not always, not in every category, but still they where the premium name that was sitting on the top of the charts. OEMs want premium stickers on their boxes. That's why they are trying so much with Qualcomm. Qualcomm is a premium brand.
But my post you quoted was targeting the audience that for 20+ years plays the same tune. "Please AMD, create good enough and cheap enough hardware so I can spent less when buying hardware from your competitors". And people where always finding excuses to not chose AMD hardware. You see, AMD hardware was considered always for "the poor". We have Threadrippers out there and many still see AMD as the budget option. It is engraved in their minds.You are the one who makes that mistake here, painting an image of me like I am an fanatic idiot who thinks he is the only sentient being on the planet. That's how you describe me here. The fact that you were a red fan 10 years ago and then you "seen the light", doesn't make you objective and me blind. You seen the light 10 years ago, congrats. And because you seen the light and I haven't, in your mind, you are doing here exactly what you are accusing me.
You are just missing my point here, because you are biased and you don't care understanding what I wrote in the post you quoted. 10 years ago you might understood. Today, you have seen the light and the others who still haven't, and insist on "coming to a different conclusion are (obviously) being irrational".
I honestly think people just want nvidia to lower prices so they can buy amd for cheaper.
You know you're getting a top to bottom stack from NV every gen, and that the performance ceiling is moving forward every gen (in the case of Blackwell, the cost of that additional performance was an absolutely Titanic 750mm2 die).
AMD waxes and wanes. They have good gens and placeholder gens, you never really know what you're getting out of them. We all thought AMD was back after the surprise of RDNA2, but then they stumbled with RDNA3 and fell with RDNA4 (in terms of pushing the performance boundary).
Will they be back with RDNA5/UDNA? Is it going to be another placeholder? Who knows. But we all know Nvidia is going to bring the heat with Rubin.
So often times AMD users don't really have an option to upgrade to anyone else except NV. When NV doesn't give them a bad experience, they never have a reason to go back to AMD. If you bought a 7900xtx, AMD doesn't really have an upgrade path for you right now. It's been this way for a long time and it shows in AMD's market share.
It's not the only reason, but it's a big one.
Nvidia is enjoying a performance advantage, so they can do these transitions and still own the gaming market, even when newer products don't have real performance advantage, but mostly a feature advantage. I am talking about the sub $1000 market where the biggest difference from 3000 to 4000 is that we get FG and from 4000 to 5000 we get MFG. Nvidia also enjoys a 90% market share meaning even if they release something that is mostly a refresh, a placeholder, they will again sell greatly. That also affects AMD's decisions about "hi end". If they know that the $1000 target group, doesn't pay for an AMD product, they will direct their resources elsewhere. Of course they are making mistakes, they are not perfect. When RX 7000 was out and i was screaming that RT performance is important, at least for marketing reasons and that 7000 will fail, I was called "Nvidia fanboy". We know how it gone for the 7000 series, even when RX 7900XTX is considered a beast and we all see how RX 9070 got positive reception mostly thanks to it's RT performance improvements.
They stopped making pipe cleaners - they lost the initiative, market presence, and slowly but surely are on the road to exiting the market altogether.
Nope. At launch, Alan Wake 2 ran slightly better on Nvidia, but AMD optimized the game over time to boost performance further. Nvidia didn’t do the same since it had already optimized it. This title is often used to showcase Nvidia' tech. So, clearly a polished game for both, you pointed out a buggy light game that even the 7800XT runs at 4k @ 60fps.
My post is technical and relatively neutral; yours is inflammatory, emotional, and baseless. Do you even understand the difference between spending $500 million and $50 million to develop an architecture? Are AMD’s margins absurd? No, not in products aimed at ordinary consumers. If you have no knowledge to back your opinion, then don’t bother sharing it.
Lets see midrange card should have 256 bit bus and 64 CUs (4096 cores) RX 9070 XT have same 4096 cores so its the same shit, don't think that card with 384 bit bus and 96 CUs will cost as 9070 XT .
So AMD is giving SAME number of GPU cores!!!
It is implied the reality is something like:
128-bit/48 CUs
192-bit/72 CUs
256-bit/96 CUs
384-bit/144 CUs
512-bit/192 CUs
MLID has covered the 192-bit part (Magnus for Xbox). TPU is covering the 256-bit part. The 384-bit part is implied by MLID (which I think he perhaps got incomplete info) as cut-down 512-bit (idk).
If you piece together everything from the different sources you can kind of make some reasonable assumptions.
Perhaps the PS portable the same as a 64-bit part, the PS6 the 128-bit part, and the xbox the 192-bit part? Granted, I'd like to see Sony use the same part as MS for their television console.
Perhaps the 256-bit part similar to a 4090, 384-bit 5090, and 512-bit 6090? Just a guesstimate, and probably not dissimilar to what nVIDIA may do, although perhaps nvidia more units/lower clock.
Power efficiency vs lower cost...no different than almost any other generation AFAICT.
Was at 15.9% in Jan 2024, 17.8% now.
If such card does exist in your universe (it does not in mine, the 9070XT costs 640 Euro in Germany and that includes 19% VAT and what not), what is the price on the similarly performant card by The Filthy Green?
AMD gaining in GPU Steam survey and Intel having an almost 8%, but with Nvidia dominating the Steam charts at the same time, means that those percentages are AMD APUs and Intel iGPUs mostly. AMD and Intel needs to start selling more discreet GPUs. But for this to happen, people and tech press need to stop worshiping everything Nvidia.
ASRock Radeon RX 9070 XT Steel Legend 16GB GDDR6 256bit (RX9070XT SL 16G) is cheapest i can find and it is "only" 730€.
But 730€ or even 640€ is very expensive for a midrange card. And yes, i'm not going to buy NV because they're also expensive and/or slow-ish, the 5070 has only 12 GB vram and starts at 550€, or 5060 TI 16 GB which is expensive, only pcie x8 and slow for what it is and what it costs (450€).
That is, convenient.
geizhals.de/?fs=9070xt
I did not know you meant "the highest price I find for GPU" though. Mea culpa.
As for the "too expansive" in general:
1) Inflation is a thing everywhere
2) TSMC keeps posting record profit after record profit. Guess where that is comnig fromI have missed the bit when you explain where does the gain come from.
About .5% is from Steam Deck.
Only 1.4% left to be explained.
And if AMD is somehow getting bigger market share on the laptop front, a place where OEMs are very obviously strongarmed by blue-green, I might have missed that.
And 9000 series with FSR4 and RT gimmick boost has nothing to do with it?
Why do you think so?
But when going CPU/APU, especially in laptops, it's mostly AMD today. Intel will sell, because of fab capacity and OEMs need Intel's fab capacity, but mostly for that reason today. Intel's financials are so bad that I doubt they can offer big discounts anymore. Especially when they are not the makers of the CPUs, but have to pay TSMC for that. Also their name got trashed last year because of the 14th gen degradation problems and the subpar performance of ARL. On the contrary, all that good publicity that Ryzen gets thanks to X3D models (who ever thought the X3D idea, should get 100 million in bonus and a free house in Hawaii to enjoy the rest of their lines as a "thank you" from Lisa) helps AMD to sell all types of CPUs the same way RTX 5090 helps Nvidia to overprice and keep selling mid range mediocre models. Also Radeon does enjoy a much better name than Intel in GPUs, so that's also a marketing advantage that OEMs will happily take advantage of.
As if that GPP thing came swinging.
Why is there no Strix version of 9070?Yet somehow the best screens are still reserved to Intel/Filthy Green only.
ASUS is no better in GPUs than the old DELL in CPUs.
But things have changed at least the last 1-2 years in laptops. Things are much more balanced.
In my POV things have changed:
First - for the better (3080 punching 6800M in high end setups, even though not all OEMs, and even though still with reservations and crippled)
Then - for the worst, almost complete AMD dGPU ban across the board, most OEMs still striking through premium features, such as superior screens, for AMD variant of the same laptop. (with notebookcheck wondering "oh, why is that", well, the answer is obvious, me thinks)Oh please.
Did ASUS "ban" Strix cards with AMD chips (the last one was with 6000 series) because they have a time machine and new 9000 series would not have cards meant for the less than 10% of the gamers?
Or is it that wonderful EPP thing in full swing, just covert?