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
Raw performance is the highest Metric, that's true, but people are also looking at what visual settings that the extra performance can give them. For example instead of 160FPS with rasterization (that my screen cannot display anyways), I would go for RT/PT enabled at 80FPS (DLSS are always enabled of course).
RT is very much like OLED, once you get used to it it is very hard to let go ;). But yeah right now the cost of entry for RT/PT is still high, but it should lower over time.
We talk alot about pricing but fwiw we're still buying...
I would argue you are in the minority on that statement. Mostly because you have the only card that can do it reasonably well (RTX 4090). I mean using your argument you are sacrificing a huge amount of FPS for better lighting visuals. When I look up the latest reviews, they mostly combine Ray Tracing and DLSS 3/3.5 to get the playable performance which is relying on not only tons of things working together to work perfectly and keep the game functioning, its also hoping that the DLSS is working right with the visuals (Which I am not saying DLSS is not working right, but we have seen screenshots showing differences in games when on or off in the visuals at times). Now I am similar to you as I prefer to have highest visuals I can in my games over 300 FPS which is why I normally buy a high end card (I know my card is aging but life happens LOL), but I cannot stand to be below 120FPS anymore (Well maybe below 100 is better to say) especially in games that involve shooting/driving. Though I also cannot stand potato mode graphics.
Sure, I agree it will but I think its not going to really catch on fully unless the performance hit (At least on Nvidia hardware) is around 10%. 50% is just too high in many cases.
As for what is considered good FPS, i believe most people would be happy with 80FPS
I don't think the 4090 will be worth much on the used market considering that it needs a high-power, quality PSU, which people looking for mid-range/used cards usually don't have. I do, but if you add their price up, maybe you get to the price of a single 4090. Maybe...
Not to mention, I can use my multiple systems to watch films and play low-demanding games in the living room and in the bedroom. I couldn't watch or play anything on just a 4090 sitting on my desk on its own.
Also the market is not determined by one person's subjective hypothesis but more of a collective summation of multiple factors like competing products, sure let's add power requirements, vram size, ai compute demand ( currentco conditions), ebay current used price is for suprim liquid x 4090 $1609. If the 5080 doesn't dethrone the 4090 the prices will sure hold. I personally sold my used 3090 at $950 and 2080ti for $850.
Lastly one bottle of alcohol is fine but the summation of two years ( the rate of some enthusiast upgrade) is sometimes more if you are an ethenolic that is doing more self harm than good in the long run. In the end it's all subjective and the market rarely depends on your opinion due to outside factors for example ai. Let me know if you have any questions
Anyway, regardless of CPU choice, a 450 W TDP graphics card requires a 750-800 W quality PSU at least, which costs north of 100 quid. I am not going to assume that budget builders are going to fork out that cash. It is still not up to you to decide how I want to "harm myself" while enjoying my friends' company. Having a few beers on a weekend hardly constitutes the definition of alcoholism.
Similarly, it's not up to me to judge that your choice of buying a 4090 instead of um... you know... having a life... is a huge waste. You do you, but don't expect everyone to follow your way. The vast majority of people have got many other things to spend their money on besides PC upgrades and games.
except that the US cost for many "services" are multiple times higher, and for no reason, vs cost in EU.
mobile plan in the US cost me ~100$/month, while it costs me 15E/month (limited to 100GB).
so im spending on a year of service, what would barely cover 2 month in the US.
a mobile plan is something i needed, especially with work (upload images/reports etc),
but paying ~100$/m for cable? nope. i rather spend 1K on a gpu and stream stuff.
and thats ignoring that after paying for 1y of cable, i have nothing, while spending +1K on the gpu, i still own the card..
just stating the fact that some ppl (like a friend) are willing to spend money on things that are obviously overpriced in the US (vs EU),
things that will not add to what you "own" (like unlimited mobile/cable plan), while then saying im wasting my money buying a 2080S WB for almost 1000$,
which i now still use, contrary to any service "rented" for a year for the same cost by my friend.
and you can say the same about those spending 2-300$ on a single part (gpu), as there are others having less than that for a whole pc/laptop.
that, i would actually call an investment for life.
or to say it like one of my teachers when explaining want/need:
you need air/food/shelter/tv (as in watching news) and now maybe internet/transportation for some countries,
everything past that is a want, no matter if its 10, 100 or 1000 (of "your" currency)...
edit: ahh, those were the days, when a 460ti was 200 DM :D
I used to go for high-end parts when they were much more affordable. Nowadays, though, I couldn't care less if my games run at 60 FPS on medium graphics instead of 120 on ultra.
only to realize he not only had better coverage (work/home) but also saving a lot of money (what i told him).
and for me, i have seen the benefit of spending more on a LC upper tier, now having a card still good enough to do 4K@30/60,
and with my "new" screen allowing for vrr, making up for not buying a card in years, while giving better experience i had prior using vrr,
even for games i already had enough fps in the past (Siege).
ignoring that especially tmobile shut down all 2/3G sites, to convert to 5G, so cost didnt really go up for them, as the old stuff is off the "budget",
and as you "only" replace the antennas (to say it in simple terms), it saves the cost for survey/plan/setup of a new "spot" to place the hw.
and its even worse for cable/internet, as living in a city center is the same, no matter what 1st world country,
while EU countries offering 3-500Mbit, i had 25/50 over fiber, at multiple times the cost.
i remember almost 20y ago i spend 16E/month for landline/gateway with 16Mbit in germany, while US offered 1Mbit internet only for 50$month.
videocardz.com/newz/amd-rdna5-is-reportedly-entirely-new-architecture-design-rdna4-merely-a-bug-fix-for-rdna3
Haven't released any RDNA 3 refresh yet, haven't released any RDNA 4 card yet, and now started to talk about RDNA 5 ? :peace:
I don't like speculating too far into the future, as it tends to be inaccurate which makes it pointless.
By 2049, there won't be graphics cards in this form and shape, because TSMC will have long been closed, because as we know Moore's law is dead, and you can't shrink the transistors indefinitely.