Friday, May 3rd 2024
AMD to Redesign Ray Tracing Hardware on RDNA 4
AMD's next generation RDNA 4 graphics architecture is expected to feature a completely new ray tracing engine, Kepler L2, a reliable source with GPU leaks, claims. Currently, AMD uses a component called Ray Accelerator, which performs the most compute-intensive portion of the ray intersection and testing pipeline, while AMD's approach to ray tracing on a hardware level still relies greatly on the shader engines. The company had debuted the ray accelerator with RDNA 2, its first architecture to meet DirectX 12 Ultimate specs, and improved the component with RDNA 3, by optimizing certain aspects of its ray testing, to bring about a 50% improvement in ray intersection performance over RDNA 2.
The way Kepler L2 puts it, RDNA 4 will feature a fundamentally transformed ray tracing hardware solution from the ones on RDNA 2 and RDNA 3. This could probably delegate more of the ray tracing workflow onto fixed-function hardware, unburdening the shader engines further. AMD is expected to debut RDNA 4 with its next line of discrete Radeon RX GPUs in the second half of 2024. Given the chatter about a power-packed event by AMD at Computex, with the company expected to unveil "Zen 5" CPU microarchitecture on both server and client processors; we might expect some talk on RDNA 4, too.
Sources:
HotHardware, Kepler_L2 (Twitter)
The way Kepler L2 puts it, RDNA 4 will feature a fundamentally transformed ray tracing hardware solution from the ones on RDNA 2 and RDNA 3. This could probably delegate more of the ray tracing workflow onto fixed-function hardware, unburdening the shader engines further. AMD is expected to debut RDNA 4 with its next line of discrete Radeon RX GPUs in the second half of 2024. Given the chatter about a power-packed event by AMD at Computex, with the company expected to unveil "Zen 5" CPU microarchitecture on both server and client processors; we might expect some talk on RDNA 4, too.
227 Comments on AMD to Redesign Ray Tracing Hardware on RDNA 4
As I pointed out before, 'blaming' company A or B for this isn't going to work out for anyone. Its not AMD's lagging behind that makes RT fail, and its not Nvidia's aggressive push either that makes it succeed. The industry moves when the industry moves, and it will do so when there is a universal belief that everyone can get a nice piece of this new pie. Nvidia's proprietary approach, with both RT, and with its AI acceleration, is not offering those nice pieces of pie for everyone. And therefore, it will not become a thing that way. It will only become a thing when the industry decides upon universal ways to accelerate, or gets the opportunity to carve out their own little niche in the landscape, each for themselves. The latter is probably what will happen to AI because its use cases are so diverse.
But for RT? Unify and democratize it, or it will die. Democratization cannot happen under just a green colored umbrella though, nor just a DX12 ultimate sticker that effectively gets supported by only one player. Strategically, I think AMD has the better play and better position with its console deals and agonizingly slow RT development. This won't happen overnight. Maturity takes time.
Secondly, a lot of 4090/4080 will go into the second hand market, so more and more people will have access to RT/PT as time goes on.
With that said, if it does in fact, increase, it should do so in a way to make it a feasible option with mid-range graphics cards. The kind of "5-10% better" that we've been getting in the last few generations won't do when you still can only use RT properly on a graphics card that costs more than some used cars.
Edit: I used to game just fine on a 1060. I still game fine on any current mid-range card, unless I turn on RT. If RT becomes feasible in my price segment, I'll welcome it. Otherwise, I don't care.I don't think a (used) 450 Watt TDP card is a good idea for an el-cheapo system that probably contains an el-cheapo PSU.
Games always have the option for future proofing that even the fastest 1% of GPU owners cannot play at the highest settings, so they can continue to sell years after released. Just look at Witcher 3 and Cyberpunk 2077, they still have lots and lots of players.
Of the two games you mentioned xx60 class owners won't be able to play it at the same settings and fps as 4090 owners for another 4-5 years at least. Nothing wrong with that, they have something to look forward to. Laughing emojis won't change that fact though.
RTX7060 owners will certainly appreciate how games made in 2024 with PT/RT and that will be selling for 5usd at that time :laugh:
Anyways this discussion is going nowhere, you and I, along with everyone else will just continue voting with our wallets and see how RT/PT will turn out in the future :rolleyes:
The RDNA3.5 iGPUs in Strix Point and Strix Halo aren't getting that improved RT performance.
Just like the PS5 uses a RDNA2.5 implementation, RDNA2 with some features from RDNA3.
My point was if most people I talk to are turning it off or putting it on low, then using it as a club when discussing GPU's as the only thing that matters is where I generally have the problem. If someone is not going to turn it on much, why is it as important as just pure performance? The arguments for RT are similar (At least to me) to the arguments for PhysX back in the day, how it was revolutionary and people talked about having to have it until it just slowly died off. Now I don't think RT is going to die off, just that its still going to be at least 2 more generations of cards minimum (At least in my opinion) before its something we can call 'mainstream'. I do want AMD to improve their RT performance, I just don't think it should be priority one.
RT is similar, you need more expensive GPU in order to drive it.
Saying RT is pointless because the majority cannot experience it is pretty stupid. Every current tech you enjoy today has been prohibitively expensive in the past.
There is no such thing as miracle where AMD would just improve RT performance by tenfold when RT eventually take off in the near future, AMD needs to keep working on improving RT today if they want to catch up in the future.
BTW if AMD has been doing everything right, their gaming division Q1 earnings should not look that pathetic LOL, basically they gave people no incentive to upgrade to RX7000
I stand by my reasoning because the argument used in favor of it does not make sense because of the lack of performance on the cards most people use. When the games dont utilize it too heavily, most of the compared cards from both companies offer good performance (Nvidia still ahead, but not be near that much) and are beyond playable. When its heavily utilized like Cyberpunk, you see a humongous drop in performance to nearly unplayable levels except on the 4090 (4080 can do ok at 1440p ultra everything, but its average is around 60 which means its got alot of drops below that threshold).
Its true that they should work on it, but raw performance should be king because that's what the majority of games are going to care about. Plus it may take off or it may not, we don't know and all the similar techs that start or people say will revolutionize gaming that end up disappearing after a time for one reason or another. There is no telling if this will do the same (I definitely have less doubts about this one disappearing than others but still).
Their earnings in gaming are low, but they are fighting an uphill battle and have been for years. I mean Nvidia has made it very hard to compete and get big press, especially with Nvidia's attitude towards reviewers and such when they don't focus on things they want as we saw when a big reviewer didn't do enough talk about RT in a video. Not to say this generation was flawless from AMD (Far from it), but it should be performing better than it is in my opinion.
I hope they can improve performance with this redesign, will be interesting to see where it is with RDNA 4.