Discussion RDNA4 + CDNA3 Architectures Thread

Page 172 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,749
6,614
136





With the GFX940 patches in full swing since first week of March, it is looking like MI300 is not far in the distant future!
Usually AMD takes around 3Qs to get the support in LLVM and amdgpu. Lately, since RDNA2 the window they push to add support for new devices is much reduced to prevent leaks.
But looking at the flurry of code in LLVM, it is a lot of commits. Maybe because US Govt is starting to prepare the SW environment for El Capitan (Maybe to avoid slow bring up situation like Frontier for example)

See here for the GFX940 specific commits
Or Phoronix

There is a lot more if you know whom to follow in LLVM review chains (before getting merged to github), but I am not going to link AMD employees.

I am starting to think MI300 will launch around the same time like Hopper probably only a couple of months later!
Although I believe Hopper had problems not having a host CPU capable of doing PCIe 5 in the very near future therefore it might have gotten pushed back a bit until SPR and Genoa arrives later in 2022.
If PVC slips again I believe MI300 could launch before it

This is nuts, MI100/200/300 cadence is impressive.



Previous thread on CDNA2 and RDNA3 here

 
Last edited:

marees

Senior member
Apr 28, 2024
578
639
96
imagine if the RX 7900 XT had launched at its current street price of about $680.

So, here's a game plan for AMD. Take it on the chin with RDNA 4, its next-gen GPU architecture. Launch that new Radeon RX 8800 XT with RX 7900 XT levels of raster performance, plus a bit of a ray tracing upgrade, for $400 and absolutely blow everything else away at that price point.

Maybe AMD wouldn't make any money doing that. But so long as it's not a case of massive losses, it would reignite interest in AMD graphics. Meanwhile, move heaven and earth to get RDNA 5 working properly with what I'm expecting to be a true chiplet architecture, and cash in on the new-found appetite for Radeon graphics with a top-to-bottom family of GPUs around two years from now. Price them low, too. Not as low as RDNA 4, but still low enough to have them hammer Nvidia for raster at a given price point and be in touch for ray tracing.

Then build from there with subsequent generations. That isn't going to get AMD to 50% market share in two generations. But I bet it would move the needle, which categorically isn't happening now. AMD has barely managed a dent over the last decade.

 

marees

Senior member
Apr 28, 2024
578
639
96
imagine if the RX 7900 XT had launched at its current street price of about $680.

So, here's a game plan for AMD. Take it on the chin with RDNA 4, its next-gen GPU architecture. Launch that new Radeon RX 8800 XT with RX 7900 XT levels of raster performance, plus a bit of a ray tracing upgrade, for $400 and absolutely blow everything else away at that price point.

Maybe AMD wouldn't make any money doing that. But so long as it's not a case of massive losses, it would reignite interest in AMD graphics. Meanwhile, move heaven and earth to get RDNA 5 working properly with what I'm expecting to be a true chiplet architecture, and cash in on the new-found appetite for Radeon graphics with a top-to-bottom family of GPUs around two years from now. Price them low, too. Not as low as RDNA 4, but still low enough to have them hammer Nvidia for raster at a given price point and be in touch for ray tracing.

Then build from there with subsequent generations. That isn't going to get AMD to 50% market share in two generations. But I bet it would move the needle, which categorically isn't happening now. AMD has barely managed a dent over the last decade.

My guess is if AMD replicates the RDNA 1 naming strategy then that could indicate lesser launch price

5700xt ($400) == 8700xtx ($500)
5700 ($350) == 8700xt ($440)
5600xt ($280) == 8600xt ($360)

5500xt ($200) == 8500xt ($250)
5500xt 4gb ($170) == 8500 ($200)
 

RnR_au

Platinum Member
Jun 6, 2021
2,113
5,097
106
It's not RDNA3's WMMA engine but RDNA4's, which doubles rates per cycle and adds sparsity on top of that. That's how you get to 300 (well 288 actually) TOPs.
Just curious... has the RDNA3.5 received these WMMA engine improvements... ie will Strix Halo sport awesome TOPS?
 
Reactions: Tlh97

ToTTenTranz

Member
Feb 4, 2021
182
313
106
My guess is if AMD replicates the RDNA 1 naming strategy then that could indicate lesser launch price

5700xt ($400) == 8700xtx ($500)
5700 ($350) == 8700xt ($440)
5600xt ($280) == 8600xt ($360)

5500xt ($200) == 8500xt ($250)
5500xt 4gb ($170) == 8500 ($200)

$500 for the full fledged Navi 48 with 16GB of 20Gbps GDDR6 would be a massive increase in performance-per-dollar, and one we haven't had since 2016.
 

marees

Senior member
Apr 28, 2024
578
639
96
I would count on it. I assume the main features NV will hype will be AI and DLSS4.

18 GB is possible but later.
The super versions could have the new gddr7 versions

Hence

5060 - 8gb
5060 super ‐ 12gb

5060 ti & 5070 - 12gb
5070 super 18gb

5070 ti ???

5080 16gb
5080 ti?/5080 super 24gb
 

Mahboi

Golden Member
Apr 4, 2024
1,035
1,900
96
That is big question. What can bring DLSS4? Tripple frame generation for 1000 fps?
MLID said at some point that the framerate alley is sort of closed, so instead they'll go for image quality. Basically dump a lot of AI to make stuff look better. HWUB followed up on that too at some point.

I don't know how true the whole "framerate is a dead end" part is true, but it does make sense. Most decent cards can reach 240 FPS in a competitive game. My old RX 6600 used to give me 144 FPS in Overwatch 2 with dips in the 100s, with FG and upscaling it would definitely handle 240 FPS on VSync with no issues.

Even a 8700 XTX can probably reach a 4K 120 FPS in most games with FSR 4 + FG.
Trying to make the image look extra special seems like the logical next step, nobody's going to care about framerate within 3-4 years. It's already of questionable value past a certain point, like those stupid 360 Hz/480Hz monitors. I think people will instead start yapping about frametimes and variance if AMD can push that narrative (lol, like AMD's got a marketing team even remotely capable of that). Their FG on Vsync thing seems like a fat argument for it.
 
Reactions: Tlh97

CastleBravo

Member
Dec 6, 2019
127
273
136
imagine if the RX 7900 XT had launched at its current street price of about $680.

So, here's a game plan for AMD. Take it on the chin with RDNA 4, its next-gen GPU architecture. Launch that new Radeon RX 8800 XT with RX 7900 XT levels of raster performance, plus a bit of a ray tracing upgrade, for $400 and absolutely blow everything else away at that price point.

Maybe AMD wouldn't make any money doing that. But so long as it's not a case of massive losses, it would reignite interest in AMD graphics. Meanwhile, move heaven and earth to get RDNA 5 working properly with what I'm expecting to be a true chiplet architecture, and cash in on the new-found appetite for Radeon graphics with a top-to-bottom family of GPUs around two years from now. Price them low, too. Not as low as RDNA 4, but still low enough to have them hammer Nvidia for raster at a given price point and be in touch for ray tracing.

Then build from there with subsequent generations. That isn't going to get AMD to 50% market share in two generations. But I bet it would move the needle, which categorically isn't happening now. AMD has barely managed a dent over the last decade.


Radeon has been moving the needle quite a bit recently, but there is a nuance...
 

tajoh111

Senior member
Mar 28, 2005
315
335
136
imagine if the RX 7900 XT had launched at its current street price of about $680.

So, here's a game plan for AMD. Take it on the chin with RDNA 4, its next-gen GPU architecture. Launch that new Radeon RX 8800 XT with RX 7900 XT levels of raster performance, plus a bit of a ray tracing upgrade, for $400 and absolutely blow everything else away at that price point.

Maybe AMD wouldn't make any money doing that. But so long as it's not a case of massive losses, it would reignite interest in AMD graphics. Meanwhile, move heaven and earth to get RDNA 5 working properly with what I'm expecting to be a true chiplet architecture, and cash in on the new-found appetite for Radeon graphics with a top-to-bottom family of GPUs around two years from now. Price them low, too. Not as low as RDNA 4, but still low enough to have them hammer Nvidia for raster at a given price point and be in touch for ray tracing.

Then build from there with subsequent generations. That isn't going to get AMD to 50% market share in two generations. But I bet it would move the needle, which categorically isn't happening now. AMD has barely managed a dent over the last decade.


Not worth it.

Getting into a bleeding match with Nvidia is pointless at this point because Nvidia has too much money and professional products to subsidize a spending match. If they were going to do this strategy, it would have needed to be done before the AI boom.

The relative large die along with the rest of the components to make a videocard make it difficult to lower the price too much. GPU's are a low margin product considering the relatively large die size next to CPU's and the added components needed to make a videocard.

Plus the videocard market is not valuable enough. Pumping out videocards at potentially the expenses of CPU production(cash flow) is not a good idea. On the other hand getting into a spending war with Intel would be a much better idea.

CPU are cheap to produce and there is no interference from partners. In addition, Intel is already pushing negative margins and are getting broker by the second. Intel's been amortizing IP at such a crazy rate that they cannot afford to lower prices, particularly when they are relying on TSMC 3nm.

So if AMD wants to have a bleeding match, it would be best to do it with Intel. Not nvidia. The stakes are better(more total addressable market) and against a much weaker opponents with factors they can control.

AMD can likely still make money off a 300 dollar 16 core 9950x. These chips consist of 2x 70mm2 4nm dies. Imagine how wrecked intel would be if AMD entire line up consisted of products from 300 to 100 dollars.

On the other hand, if Navi 48 has 64 CU's and a 256 bit bus, It will be this 300+mm2 monolithic die, which has to be sold to partners for 200 dollars to make a $400 dollar price point possible to account for the rest of the components to make a videocard along with the profits for distributers and board partners. Just not as feasible. Plus if Nvidia followed suit and participated in the bleeding match, they could outlast AMD. AMD should lower prices to get more marketshare but not enough to make trigger a war with Jensen Huang.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |