Discussion RDNA4 + CDNA3 Architectures Thread

[DisEnchantment]

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

History for llvm/lib/Target/AMDGPU - llvm/llvm-project

The LLVM Project is a collection of modular and reusable compiler and toolchain technologies. - History for llvm/lib/Target/AMDGPU - llvm/llvm-project

github.com

Or Phoronix

More AMD "GFX940" Enablement Work Landing In LLVM - Phoronix

www.phoronix.com

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

Question - Speculation: RDNA3 + CDNA2 Architectures Thread

Man I have been dying to make this one for a while now. First rumours for RDNA3 are here so new thread time! Just going to start off with this one for now: kopite7kimi on Twitter: "@VideoCardz Ah, I mean a simple mcm design with 10240 cores is not enough. Because the lift from RDNA2 to RDNA3...

forums.anandtech.com

 

[TESKATLIPOKA]

You seem forgotten of my questions about RDNA4 that I asked three months ago: - Link with table

I am looking and looking, yet you aren't asking any questions in that link.

I could see why you don't believe at that time; hey the whole Youtube community is hyping RDNA4 as well. Also thanks to adroc giving some misinformation about RDNA4 also won't help. Anyhow, you should be able to see clearer picture of positioning of RDNA4 now... Actually if you throw GDDR7 out of equation, you can calculate the how many CU N43 is having....

In fact, I can estimate rasterization performance from the table I linked. RX7600XT (N44) should perform 5-15% better than RX7600 with upper range to be OC version.

RX7700 (I will stick with N43 for now) should perform 25-35% better than RX7600, the OC version's performance will approach RX7700XT but not RT performance.

I didn't ask about positioning or how It should perform.

Yeah, as I explained here, mobile RDNA4 will most likely to be model as 8000M series due to different market conditions. OTOH, Unless AMD renamed whole dGPU to RX8000 series, then RDNA4 dGPU will be model RX7600XT and RX7700...cause as I said, RDNA4 is a die shrink of RDNA3, the changes on micro-arhicteture is due to lower clock speed of GPU. As clock speed go lower so does TF, AMD has to pump up some changes on front end scheduler to make it slightly powerful to compensate of lower clock speed. So that, RX7600XT will perform better than RX7600 with lower clock speed and same CU...

If RDNA4 was just a die shrink of RDNA3, then It wouldn't be called as RDNA4.
From when do you call changes in architecture as a die shrink? That's some news to me. I also don't see a reason why RDNA4 based GPUs should have lower clock than what RDNA3 has.

Well, at least you start thinking of mobile RDNA4 but clearly you miss the reason why RX7600 won't have cut down version on desktop lineup. It is cause of cut down version is available on mobile 7600M series. The same also apply to RX7600XT. I hope you won't ask me why there is no cut down version of RX7700 in the future

Where is It written that every cutdown GPU has to go to mobile? Nowhere.
It's enough If you explain why both of RDNA4 GPUs should be called like the older generation.
Or why RX 7600XT 16GB should be RDNA4 based. Thanks.

 

[adroc_thurston]

If RDNA4 was just a die shrink of RDNA3, then It wouldn't be called as RDNA4.
From when do you call changes in architecture as a die shrink? That's some news to me. I also don't see a reason why RDNA4 based GPUs should have lower clock than what RDNA3 has.

You really should waste no time on a schizo.
It's pointless, he has voiced in his head.

Personal attacks are not allowed. This needs to stop. -AT Mod Shmee

 

[TESKATLIPOKA]

AMD Close to Launching Radeon RX 7800M Series Based on "Navi 32"
I think we will hear more info at CES 2024.

Likelihood of RDNA4 showing up at CES2024 as @Tigerick thinks is lower and lower.

 

[misuspita]

It would really be interesting to see some ext GPU based on more powerful chips, such as GPD G1, but with muuuch bigger fans, as the small ones are whiniy on higher rpm with more audible frequencies. I'd buy some miniPC like those Beelink GTR 7 Pro or Minisforum 780XTX and couple them (preffertably with occulink) with a powerfull e-GPU when I need to play games or use it with a professional program that needs GPU acceleration, when needed.

 

[Rekluse]

So the conclusion of this thread being, we have to wait past 2025 for AMD to release a Halo GPU that can touch whatever nVidia has coming down the pipe ?

 

[GodisanAtheist]

So the conclusion of this thread being, we have to wait past 2025 for AMD to release a Halo GPU that can touch whatever nVidia has coming down the pipe ?

-Threads, threads never end...

 

[Kepler_L2]

So the conclusion of this thread being, we have to wait past 2025 for AMD to release a Halo GPU that can touch whatever nVidia has coming down the pipe ?

Yes pray for Navi50 to not be cancelled (and start saving up money because it will make the 4090 look like a great deal).

 

[TESKATLIPOKA]

Yes pray for Navi50 to not be cancelled (and start saving up money because it will make the 4090 look like a great deal).

Only If 4090 has better perf/$, but in ultra highend that never mattered.

 

[SolidQ]

Yes pray for Navi50 to not be cancelled (and start saving up money because it will make the 4090 look like a great deal).

N50 3k$ GPU?

 

[blckgrffn]

N50 3k$ GPU?

It will be interesting to watch where the real high end goes.

People spend ludicrous money on their hobbies, whether that's golf clubs, guitars, cars or liquor. Even at $3k will the true "ultra high end" flinch? I doubt it. $4k for 4K gaming?

Especially if its an actual halo part that can be put into a "normal" case and powered by a "normal" (less than 1500W?) PSU.

Seeing it just sit cheerfully at the top of every review of every GPU will be all the marketing it needs.

nvidia is free to go this route too, obviously. I thinks arguable they've been walking this line since they launched the Titan. Man, I hope AMD gives it a cool name, like Zeus.

 

[soresu]

Yes pray for Navi50 to not be cancelled (and start saving up money because it will make the 4090 look like a great deal).

Also praying for HIP-RT to improve a lot, still seems to be way behind OptiX which isn't helping them get commercial RT/PT renderers like Renderman and Arnold to bite on an AMD backend.

 

[Tigerick]

Guys, here is the final table of speculation of upcoming RX7600XT; once you read through the specs you should have better understanding of why AMD wants to hide the codenames:-

	RX 7600	RX 7600XT	RX 7600XT OC	RTX 4060	RTX 4060Ti 16GB
Current Price	$269			$299	$449
Launching Price	~ $199 ?	$249 - $269 ?	$299 - $329 ?	?	?
GPU	RDNA3 N33	RDNA4 N44	RDNA4 N44	AD107	AD106
Process Node	N6	N4P	N4P	4N	4N
Die Size	204 mm2	~ 160 mm2	~ 160 mm2	159 mm2	188 mm2
CU / SM	32	32 ?	32 ?	24	34
Base Clock	1720 MHz	?	?	1830 MHz	2310 MHz
Boost Clock	2655 MHz	Lower	?	2460 MHz	2535 MHz
FP32	21.75 TF	Lower	?	15.11 TF	22.06 TF
Memory Size	8GB GDDR6	16GB GDDR6	16GB GDDR6	8GB GDDR6	16GB GDDR6
Memory Clock	18 Gbps	18 Gbps ?	?	17 Gbps	18 Gbps
Memory BW	288 GB/s	288 GB/s	?	272 GB/s	288 GB/s
TDP	165 W	~ 130 W	~ 165 W	115 W	165 W
TPU Perf	Base	~ +5%	~ +10%	+ 0%	+ 21%
RT Perf	Base	~ +38%	~ +43%	+ 119%	+ 188%

• Basically, RDNA4 is a die shrink of RDNA3 (more likely N32), but due to N4P process and AMD's intention to lower the TDP to save PCB/VRM/MOSFET costs, the clock speed of the GPU would be lower than N33. You can compare to RTX4060's TDP with similar process. That's why AMD has to upgrade the graphics engine to make it more powerful so that 7600XT will perform better than 7600. In CPU term, we refer that as IPC, in GPU term....well that one have to ask GPU specialist
• That's why we should expect two versions of 7600XT with 16GB at launch, standard version with lower TDP (maybe single fan like RTX4060) and OC version which should use same PCB as current 7600.
• Thanks to @adroc_thurston early leaking about how cheap GDDR6 would be, I should expect N44 to come with 16GB not 8GB. My bad , with lower BOM and cheap GDDR6, AMD would most likely offer RX7600XT with 16GB at current RX7600's pricing. We should expect RX7600 to drop to $199 price point also. Now I know what the new Polaris meant for N44: with lower TDP, doubling RAM and lower price point; this is AMD's once again attack at nVidia's RTX4060. (PS: And RTX4060Ti 16GB)
• Don't expect much faster rasterization performance but at least they would be slightly faster than RTX4060.
• Ray tracing wise: 7600XT should perform 40% better than RX7600, unfortunately still far slower than RTX4060, I wish I am wrong on these figures but if you click the link above, you will know why...NV really shines on RT's performance...
• Again kudos to @adroc_thurston for leaking die size of 250mm2, $399 price point and mid end position of upcoming N43, I could speculate two RDNA4 models. Speaking of codenames, N44 has remained for a while cause as you see N44 supposed to sit below N33 based on last number. Turn out 7600XT (N44) is positioned above 7600 (N33).
• Now you know why AMD wants to change N43 to N48 cause N43 is obviously sitting below N32. And that's the position of upcoming RX7700....no wonder @adroc_thurston want to pretend he doesn't know about N43, my bad . Guys, remember if you want to discuss about upcoming RX7700 please refer code name as N48 not N43....

 

[TESKATLIPOKA]

Guys, here is the final table of speculation of upcoming RX7600XT; once you read through the specs you should have better understanding of why AMD wants to hide the codenames:-

	RX 7600	RX 7600XT	RX 7600XT OC	RTX 4060
Current Price	$269			$299
Launching Price	~ $199 ?	$249 - $269 ?	$299 - $329 ?	?
GPU	RDNA3 N33	RDNA4 N44	RDNA4 N44	AD107
Process Node	N6	N4P	N4P	4N
Die Size	204 mm2	~ 160 mm2	~ 160 mm2	159 mm2
CU / SM	32	32 ?	32 ?	24
Base Clock	1720 MHz	?	?	1830 MHz
Boost Clock	2655 MHz	Lower	?	2460 MHz
FP32	21.75 TF	Lower	?	15.11 TF
Memory Size	8GB GDDR6	16GB GDDR6	16GB GDDR6	8GB GDDR6
Memory Clock	18 Gbps	18 Gbps ?	?	17 Gbps
Memory BW	288 GB/s	288 GB/s	?	272 GB/s
TDP	165 W	~ 130 W	~ 165 W	115 W
TPU Perf	Base	~ +5%	~ +10%	+ 0%
RT Perf	Base	~ +38%	~ +43%	+ 119%

• Basically, RDNA4 is a die shrink of RDNA3 (more likely N32), but due to N4P process and AMD's intention to lower the TDP to save PCB/VRM/MOSFET costs, the clock speed of the GPU would be lower than N33. You can compare to RTX4060's TDP with similar process. That's why AMD has to upgrade the graphics engine to make it more powerful so that 7600XT will perform better than 7600. In CPU term, we refer that as IPC, in GPU term....well that one have to ask GPU specialist
• That's why we should expect two versions of 7600XT with 16GB at launch, standard version with lower TDP (maybe single fan like RTX4060) and OC version which should use same PCB as current 7600.
• Thanks to @adroc_thurston early leaking about how cheap GDDR6 would be, I should expect N44 to come with 16GB not 8GB. My bad , with lower BOM and cheap GDDR6, AMD would most likely offer RX7600XT with 16GB at current RX7600's pricing. We should expect RX7600 to drop to $199 price point also. Now I know what the new Polaris meant for N44: with lower TDP, doubling RAM and lower price point; this is AMD's once again attack at nVidia's RTX4060.
• Don't expect much faster rasterization performance but at least they would be slightly faster than RTX4060.
• Ray tracing wise: 7600XT should perform 40% better than RX7600, unfortunately still far slower than RTX4060, I wish I am wrong on these figures but if you click the link above, you will know why...NV really shines on RT's performance...

1.) RX 7600 8GB -> release price is $269, current price is likely lower.
2.) RX 7600XT at $249-269 is already very expensive compared to RX 7600 8GB at $199, when you get only 8GB extra Vram and 5% extra performance.
3.) RX 7600XT OC is unsellable at that price.
4.) 160mm2 at N4P vs 204mm2 N6. It's clear which one is cheaper -> N6. There is no reason making a similarly performing chip based on RDNA4, which is costlier to make.
5.) 130W vs 165W is only 21% lower, there is no reason for lower boost If It was using a better process.
6.) Your expected RT performance is based on nothing.
7.) Your claim about RX 7600XT using RDNA4 is based on nothing.
8.) RX 7600XT using a better process is also based on nothing.

Thanks to reading your table, I know It's just pure speculation with many contradicting points and without any basis, just your expectations.
Yet you make It sound like It's a fact or you know some insider info.

 

[SolidQ]

5% extra performance

Tha from Single Source

 

[TESKATLIPOKA]

Tha from Single Source

And now you crushed @Tigerick's fantasy about RX 7600XT using RDNA4.

 

[SolidQ]

RX 7600XT using RDNA4

RDNA4 have difference WGP + RT, it's impossible make 7600XT on RDNA4 and have faster RT, than 7700XT

 

[Krteq]

I don't know where was that "N4x" for RX 7600XT rumors came from, but from Linux kernel driver sources we have confirmed that RDNA 3 refresh (GFX1150 aka "RDNA3.5") is coming soon

For example:

Mesa 23.3 Released With Initial NVK Vulkan Driver, AMD RDNA3 Refresh & Raspberry Pi 5 - Phoronix

www.phoronix.com

AMD RDNA3 Refresh Graphics Support May Be In Good Shape With Linux 6.7 - Phoronix

www.phoronix.com

 

[Tigerick]

Tha from Single Source

Hoho, if you check Gigabyte 7600 OC and Acer 7600 OC reviews, you would not quote RGT source at all cause both cards are available at around 180W TGP. So tell me if 7600 OC are available at 180 W, 7600XT OC needs how many TDP???

 

[jpiniero]

I don't know where was that "N4x" for RX 7600XT rumors came from, but from Linux kernel driver sources we have confirmed that RDNA 3 refresh (GFX1150 aka "RDNA3.5") is coming soon

For example:

Mesa 23.3 Released With Initial NVK Vulkan Driver, AMD RDNA3 Refresh & Raspberry Pi 5 - Phoronix

www.phoronix.com

AMD RDNA3 Refresh Graphics Support May Be In Good Shape With Linux 6.7 - Phoronix

www.phoronix.com

Isn't that just Strix Point?

 

[Tuna-Fish]

Hoho, if you check Gigabyte 7600 OC and Acer 7600 OC reviews, you would not quote RGT source at all cause both cards are available at around 180W TGP. So tell me if 7600 OC are available at 180 W, 7600XT OC needs how many TDP???

There is not going to be a reference 7600XT, only AIB models. They are basically taking the AIB 7600 OC models, slapping more and faster ram on them, and shipping that as 7600XT. That's literally all it is.

 

[blckgrffn]

There is not going to be a reference 7600XT, only AIB models. They are basically taking the AIB 7600 OC models, slapping more and faster ram on them, and shipping that as 7600XT. That's literally all it is.

So you are saying its like the 4060Ti and 4060ti 16GB?

Makes sense. Honestly, I am fine with that. It I wish nvidia had done something different with their parts - RX 4060/RX 4060 GT (8GB)/RX 4060Ti (16GB) but oh well. The 16GB variant of that was relatively short lived anyway, I wonder if this XT part will be too.

 

[Tigerick]

There is not going to be a reference 7600XT, only AIB models. They are basically taking the AIB 7600 OC models, slapping more and faster ram on them, and shipping that as 7600XT. That's literally all it is.

Then it should be call 7600 OC 16GB cause there are no new features on it like RTX4060Ti and 4060Ti 16GB. If AMD name 7600XT, then there must have some new features

 

[Joe NYC]

Then it should be call 7600 OC 16GB cause there are no new features on it like RTX4060Ti and 4060Ti 16GB. If AMD name 7600XT, then there must have some new features

16 GB of memory is the new feature.

 

[Tigerick]

16 GB of memory is the new feature.

Beside 4060Ti example, you should check out 7700/7800 and 7900 XT/XTX...plus how about OC version of 7600XT if you guys think they are using N33, this is N6 process we are talking

 

[Joe NYC]

Beside 4060Ti example, you should check out 7700/7800 and 7900 XT/XTX...plus how about OC version of 7600XT if you guys think they are using N33, this is N6 process we are talking

As far as 7600 XT, I think this is just N33 with extra memory, and maybe a couple of extra MHz thrown in.

I think it is unlikely AMD is going to be investing any more into RDNA 3 desktop line. It is selling ok as is.

I did a quick check and inventory of RDNA 2 has dried up. Very few models left at Micro Centers, and none are left on AMD.com web side. Lack of RDNA 2 cards improves the sales of RDNA 3 cards further...