Wednesday, August 28th 2024
AMD RDNA 4 GPU Memory and Infinity Cache Configurations Surface
AMD's next generation RDNA 4 graphics architecture will see the company focus on the performance segment of the market. The company is rumored to not be making a successor to the enthusiast-segment "Navi 21" and "Navi 31" chips based on RDNA 4, and will instead focus on improving performance and efficiency in the most high-volume segments, just like the original RDNA-powered generation, the Radeon RX 5000 series. There are two chips in the new RDNA 4 generation that have hit the rumor mill, the "Navi 48" and the "Navi 44." The "Navi 48" is the faster of the two, powering the top SKUs in this generation, while the "Navi 44" is expected to be the mid-tier chip.
According to Kepler_L2, a reliable source with GPU leaks, and VideoCardz, which connected the tweet to the RDNA 4 generation, the top "Navi 48" silicon is expected to feature a 256-bit wide GDDR6 memory interface—so there's no upgrade to GDDR7. The top SKU based on this chip, the "Navi 48 XTX," will feature a memory speed of 20 Gbps, for 640 GB/s of memory bandwidth. The next-best SKU, codenamed "Navi 48 XT," will feature a slightly lower 18 Gbps memory speed at the same bus-width, for 576 GB/s of memory bandwidth. The "Navi 44" chip has a respectable 192-bit wide memory bus, and its top SKU will feature a 19 Gbps speed, for 456 GB/s of bandwidth on tap.
Another set of rumors from the same sources also point to the Infinity Cache sizes of these chips. "Navi 48" comes with 64 MB of it, which will be available on both the "Navi 48 XTX" and "Navi 48 XT," while the "Navi 44" silicon comes with 48 MB of it. We are hearing from multiple sources that the "Navi 4x" GPU family will stick to traditional monolithic silicon designs, and not venture out into chiplet disaggregation like the company did with the "Navi 31" and the "Navi 32."
Yet another set of rumors, these from Moore's Law is Dead, talk about how AMD's design focus with RDNA 4 will be to ace performance, performance-per-Watt, and performance cost of ray tracing, in the segments of the market that NVIDIA makes the most volumes in, if not the most margins in. MLID points to the likelihood of the ray tracing performance improvements riding on there being not one, but two ray accelerators per compute unit, with a greater degree of fixed-function acceleration for the ray tracing workflow (i.e. less of it will be delegated to the programmable shaders).
Sources:
Kepler_L2 (memory speeds), Wccftech, VideoCardz (memory speeds), Kepler_L2 (cache size), VideoCardz (cache size), Moore's Law is Dead (YouTube)
According to Kepler_L2, a reliable source with GPU leaks, and VideoCardz, which connected the tweet to the RDNA 4 generation, the top "Navi 48" silicon is expected to feature a 256-bit wide GDDR6 memory interface—so there's no upgrade to GDDR7. The top SKU based on this chip, the "Navi 48 XTX," will feature a memory speed of 20 Gbps, for 640 GB/s of memory bandwidth. The next-best SKU, codenamed "Navi 48 XT," will feature a slightly lower 18 Gbps memory speed at the same bus-width, for 576 GB/s of memory bandwidth. The "Navi 44" chip has a respectable 192-bit wide memory bus, and its top SKU will feature a 19 Gbps speed, for 456 GB/s of bandwidth on tap.
Yet another set of rumors, these from Moore's Law is Dead, talk about how AMD's design focus with RDNA 4 will be to ace performance, performance-per-Watt, and performance cost of ray tracing, in the segments of the market that NVIDIA makes the most volumes in, if not the most margins in. MLID points to the likelihood of the ray tracing performance improvements riding on there being not one, but two ray accelerators per compute unit, with a greater degree of fixed-function acceleration for the ray tracing workflow (i.e. less of it will be delegated to the programmable shaders).
104 Comments on AMD RDNA 4 GPU Memory and Infinity Cache Configurations Surface
Unfortunate but perhaps expected by this point.
Maybe there is a major problem in AMD's management, and they have to sit around that Board of directors, and decide what actions are necessary to fix the abnormally poor execution of the graphics division.
The 6800 XT which is faster in some games has those 128 MB.
Not intending to start up the debate about the worth of ray tracing but it is the future one step at a time. AMD knows this and that's why they are focusing on improving ray tracing performance. The bottom line is that to compete with Nvidia they have to compete with Nvidia so that is what they are going to do. I hope.
So in terms of RDNA4, AMD "just" has to build a midrange product that comes in earlier, cheaper and faster in ray-tracing than NVIDIA, while also having drivers that don't fail at basics like multi-monitor power draw. This would allow them to claw back marketshare (and mindshare) for a potential return to the high-end with RDNA5.
Looks like external memory controllers are gone too which will also increase the size of the monolithic die.
www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+Serie/RX+7900+XTX.html
www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+Serie/RX+7600.html
2. The EU is the most receptive market to AMD, by far
3. The 7900 XTX's 24 GB is attractive for AI use. There's a high chance many if not most of these cards sold for this reasonFab allocation which can be used to manufacture more profitable, high-margin Epyc processors indeed.
1. Consumer Ryzen;
2. Consumer Radeon;
3. Semi-custom Playstation and Xbox chips.
2) I'd love to see the numbers to back this up.
3) Given the issues Tiny Build had with 7900s for AI builds, I highly doubt that "most" 7900xtx or Xt sales went to AI use. Running those commercial loads on nvidia is a lot easier.It's funny, we've heard this argument before, the last time AMD snubbed the high end and focused on mid range chips.
When AMD did this with polaris, what was the result? Oh yeah, nvidia grabbing up huge chunks of the market as high end AMD buyers migrated to nvidia as AMD stagnated on 290x performance until the disaster that was Vega. Meanwhile, Nvidia leverages that position to sell record numbers of GPUs at higher prices then anything AMD made. Then AMD released the RX 5000 series, still failing to compete at the high end, and failing to make an impact.
AMD finally got back on track with the 6000s, and the 7000s became a repeat of the RX 200s where only the high end saw improvement and prices stagnated. Now we're doing RX 5000 again, dropping the high end to favor only mid range and lower.
Imma predict this is going to go just as well as it did last time, with complete stagnation and consumers moving on to nvidia to find some semblance of improvement.
Exactly the same cache configuration as RDNA3 cards with same memory bus width.
Now we're seeing rDNA4 will have the same cache and memory config as rDNA3, with a similar core count.
Where is this improvement coming from?
Navi 31: Die size: 529 mm^2, Transistors: 57.7 Billion, VRAM: 24 GB, Pixel rate: 480 GPix/s, Texture rate: 960 GTex/s
AD103: Die size: 379 mm^2, Transistors: 45.9 Billion, VRAM: 16 GB, Pixel rate: 280 GPix/s, Texture rate: 760 GTex/s
So, again, which is a competitor to which?Mindfactory is not a "single store". It is the largest retailer in Europe, and sells tons and tons of inventory. Be respectful.
Ok so from TPU, but the meaty part is just ~
304.35 mm² (GCD Die) on 5nm TSMC.
7900 xtx enabled raytracing +/-5% to a 3090
Adding another raytracing unit does not seem wise as RDNA3 has a hardtime filling up its improved RT units with BHV transversal additions. Beside that it also doesn't even both using it 2 issue per-clock addition either unless specifically coded for it.
RDNA3 largely matched RTX 3090 in RT, while RDNA2 lagged behind it a fair bit.
Core counts and cache isn't everything that affects RT performance..
Which rDNA3 are you referring to? Because the 7800xt has roughly the same core config as a 6900xt, and near identical raytracing performance in most games.
Each CU contained raytracing hardware. more CUs means more raytracing performance. Per CU, rDNA3 had almost no performance improvements. Go look at TPU's 7800xt performance review, average of a whopping 3% faster in RT then the 6900xt.
www.techpowerup.com/review/amd-radeon-rx-7800-xt/34.html