Question Speculation: RDNA3 + CDNA2 Architectures Thread

[uzzi38]

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 is much bigger than from RDNA1 to RDNA2. We should expect many big improvements in GFX11. 🤔" / Twitter

 

[Mopetar]

Below? I think it will mainly be aimed directly at 4070 perhaps slightly above. Maybe extending as low as 4060 Ti Competitor.

7800 16 GB - Full fat everything. Out perform the 4070 by a bit, for about the same price (~$600).

7700 XT 12 GB - Similar to 4070 for $100 less (~$500)

7700 12 GB - Aimed slightly above 4060 Ti for better price. (~$400) - The one I find most interesting.

7600/XT are Navi 33 aimed at 4060 and below.

We already saw that last generation that AMD isn't going to get many sales even if they are slightly better at the same price. The only reason to keep prices high is to maximize their profit. But they probably have a minimum amount they need to sell just to pay for the hardware design and software support for all of it. Some of that gets covered by APU and console SoC designs, but investors probably wouldn't be happy if AMD weren't trying to gain more of the market.

 

[Heartbreaker]

We already saw that last generation that AMD isn't going to get many sales even if they are slightly better at the same price.

We have seen this for multiple generations. Why would you suddenly expect change?

 

[Heartbreaker]

This is probably going to be an unpopular opinion, but IMO AMD cares little about the dGPU consumer market beyond doing what work is necessary to keep their chips reasonably competitive and in consoles. It just doesn't make them the money that CDNA and data center chips do. So they'd rather price their chips high (e.g. RDNA3) and get good margins even if it means they're 1) not making that many of them and 2) not gaining market share.

I'd be happy to be proven wrong with RDNA4.

Where can we see the revenue comparison between dGPU and data center GPUs?

Because I get the impression that AMD is doing even worse in data center GPUs, than it is in dGPUs.

 

[Ajay]

According to TPU full N32 is 64 CU. 5 cuts seems like a lot for a 200mm² GCD. If they cut down the same percentage as N31 you'd get 64, 56, and 46 CU SKU's although that's probably not the right way to look at it. 6950XT performance seems a bit optimistic to me for full N32. The 7900XT is only 10-15% faster than the 6950XT in raster.

The way prices are moving, the 6950 XT will be $600 soon. So, the 7800 XT better have the the same performance or it’ll be pointless to release it. The 7800 XT will have the advantage of lower power consumption.

 

[TESKATLIPOKA]

The way prices are moving, the 6950 XT will be $600 soon. So, the 7800 XT better have the the same performance or it’ll be pointless to release it. The 7800 XT will have the advantage of lower power consumption.

I think with RT enabled, It should perform very close.
RDNA3 is pretty disappointing, but I don't think choosing RDNA2 over It is such a great option unless there is a significant perf/$ difference.
Dual-issue will likely bring additional performance in newer games in the future.

 

[leoneazzurro]

Most of those costs are minimal. The most expensive part is when they need a bigger PCB with more wiring, but that is still not that much more costly.

It depends. Testing definitely is not so minimal. But yes, BOM is the biggest part of manufacturing cost.

 

[Jaskalas]

We have seen this for multiple generations. Why would you suddenly expect change?

Because NVIDIA shorts on VRAM and as of this year, 2023, it matters a great deal.
RT may as well not even exist with 12GB. It is a huge opening for AMD that did not exist until now.

 

[Aapje]

Indeed. Nvidia dropped the ball. However, AMD didn't deliver a strong generation so far.

 

[Rigg]

We have seen this for multiple generations. Why would you suddenly expect change?

Because suddenly 4000 series Nvidia cards don't appear to be selling particularly well at their current asking prices.

The way prices are moving, the 6950 XT will be $600 soon. So, the 7800 XT better have the the same performance or it’ll be pointless to release it. The 7800 XT will have the advantage of lower power consumption.

The 7800 XT would be pointless to sell because it basically already exists. They just stupidly named it the 7900 XT instead and tried to charge too much money for it. If they want to make a $600-$650 7800 xt card they should offer a more cost effective N31 with a 256 bit bus and 16 GB of VRAM. Even if that means they need to phase out the 7900 XT. Doubling down on bad sales and marketing strategies isn't going to help. I think their best strategy at this point is to save the 7800 XT name for an N31 refresh at the end of the year. I doubt they can clock N32 high enough to reach 6950xt raster performance without destroying its power and thermal efficiency.

 

[Heartbreaker]

Because suddenly 4000 series Nvidia cards don't appear to be selling particularly well at their current asking prices.

Newsflash: Nothing computer related is selling particularly well.

 

[Rigg]

Newsflash: Nothing computer related is selling particularly well.

Newsflash:

 

[Heartbreaker]

Newsflash:
View attachment 80186

Newsflash: Irrelevant.

Overwhelming likely that AMD will repeat what it did for several generations now. Which is offering slightly better perf/$.

 

[moinmoin]

This is probably going to be an unpopular opinion, but IMO AMD cares little about the dGPU consumer market beyond doing what work is necessary to keep their chips reasonably competitive and in consoles.

Pretty sure AMD does care about matching/exceeding Nvidia's consumer dGPU top performance and planned to do that with RDNA3 as well, but wasn't prepared for the now known failure.

 

[Ajay]

The 7800 XT would be pointless to sell because it basically already exists. They just stupidly named it the 7900 XT instead and tried to charge too much money for it. If they want to make a $600-$650 7800 xt card they should offer a more cost effective N31 with a 256 bit bus and 16 GB of VRAM. Even if that means they need to phase out the 7900 XT. Doubling down on bad sales and marketing strategies isn't going to help. I think their best strategy at this point is to save the 7800 XT name for an N31 refresh at the end of the year. I doubt they can clock N32 high enough to reach 6950xt raster performance without destroying its power and thermal efficiency.

That's a fair point. I don’t know what the sales distribution is between the 7900 XT and the 7900 XTX. That would show whether or not the marketing delivered for AMD or not.

Not sure what this refresh is - 7950 XTX with slight process gains and better binning?

 

[DooKey]

Because NVIDIA shorts on VRAM and as of this year, 2023, it matters a great deal.
RT may as well not even exist with 12GB. It is a huge opening for AMD that did not exist until now.

Oh, you are one of those that thinks the vast majority somehow reads tech forums and will suddenly believe that 12gb is DOA in RT.

Mindshare is everything and only one company has earned it. Business 101.

 

[Rigg]

That's a fair point. I don’t know what the sales distribution is between the 7900 XT and the 7900 XTX. That would show whether or not the marketing delivered for AMD or not.

I'm sure the sales distribution between them is better now that the 7900 XT can be purchased for as low as $760. The MSRP was clearly a marketing failure since it's only been 4.5 months since these cards released and it's already received a significant price cut. I think the naming scheme is bad marketing since the 6950 XT, 6900 XT, 6800 xt, and 6800 were actually competitive in raster performance with their 3090 TI, 3090, 3080 ti, and 3080 counterparts. The 7900 XTX gets curb stomped by the 4090. Based on the precedent that was set by the 6000 series cards, the current 7000 cards have no business being marketed as 90 tier cards when they actually compete with the 4080 and 4070 ti (AKA 4080 12 GB). The 4090 was on the market for a month before the 7900 cards were officially announced. They had time to call an audible and they didn't change course. They knew the 4090 was going to destroy the 7900 XTX and they didn't adjust the marketing to compensate. These cards should have been branded as 7800 XT and 7800 and priced a bit more aggressively IMO.

Not sure what this refresh is - 7950 XTX with slight process gains and better binning?

Assuming that it would be financially feasible, there seems to be silicon level issues that might be helped with a new stepping. All they did for 6950 XT was give it more clock speed via power budget and some slightly faster memory.

 

[Jaskalas]

Oh, you are one of those that thinks the vast majority somehow reads tech forums and will suddenly believe that 12gb is DOA in RT.

Mindshare is everything and only one company has earned it. Business 101.

No, I think there is a lot of ill will building among those of us who purchased 8GB Ampere.
Imagine the mindshare if NVIDIA does it again with Ada.

Tech forums? No, Youtube.
Through multiple tech channels, the news is spreading that 8GB is DOA, and anyone who purchases above entry level is likely to hear and take notice as every single new title shreds GPUs for a sole reason, and AMD already provided the solution last generation. That's more mindshare, more customers for RDNA3 if/when it comes down to the mid range.

 

[insertcarehere]

They're packing the same number of CU. RDNA3 has already shown it doesn't clock high on N5. It has also shown the dual issue capability amounts to about 0% IPC increase. Another case of RDNA3 being a bit of a dud as usual.

And finally despite 780M having ~25% higher boost clock it doesn't seem to have a 25% higher typical clock rate. I guess that's the disappointing part.

The 780m has been shown to clock at 2.8ghz and N33 at 80-ishW clocks at 2.45ghz or thereabouts. Which is pretty impressive given the former is an iGPU and the latter is on N6 and power-constrained.

The problem is that all this extra clock speed along with the other supposed "improvements" don't seem to be helping RDNA3 performance nearly as much as they should on paper.

 

[Kaluan]

That assumes that customers are dumb and don't understand that Nvidia is better at RT. For AMD, any mention of RT is at best a defensive maneuver, to convince customers that the card is not as bad as people may think. Any attempt to convince customers that the 7600 XT than the 6600 XT is better due to its raytracing ability, automatically pushes customers to cards that are better at raytracing, which are Nvidia's cards. After all, if RT is a reason to pay more for a 7600 XT over a 6600 XT, then RT is also a reason to pay more for a 4060, because the card will be better at that than a 7600 XT.

So you're telling me the 7900 XTX announcement slide that showed it's up to(TM) 82% faster in ray tracing than 6950 XT pushed people into buying 4080s?

lol what a bizarre argument

 

[Kaluan]

You just gave the answer. 6700 XT was more expensive than 5700 XT. Why do you expect 7700XT to be suddenly cheaper than 6700XT?
A cutdown N31 launched at $899 and between It and full N32 is not that much difference in performance to warrant 2x difference in price.

You're using SKU nomenclature for comparison in the first and die segmentation in the second and act like it's the same argument.

That's extremely flawed reasoning.

I could just as easily say the $999 MSRP 7900XTX (top N31 bin) was clearly priced lower than the $1099 6950XT (top N21 bin)

But none of that matters, like some other ppl said, we're not in the same market conditions and neither is AMD in a "as good" a competitiveness position as last gen. Additionally, N33 should be clearly cheaper to make than N23, other BOM factors aside.

 

[Aapje]

So you're telling me the 7900 XTX announcement slide that showed it's up to(TM) 82% faster in ray tracing than 6950 XT pushed people into buying 4080s?

lol what a bizarre argument

On the one hand, that slide can reduce the perception that AMD sucks at RT, but it also pushes the idea that RT performance is important, which indeed is bad for AMD.

Anyway, my argument was not that it is a very bad idea to have one slide about RT, but rather that it's a very bad idea to take the weak spot of your product and put your marketing focus on that. You want:

"Buy this product because it is better than the competition at X"

Not:

"Buy this product because it is better at Y than our previous generation, but worse at Y than our competition"

 

[coercitiv]

So you're telling me the 7900 XTX announcement slide that showed it's up to(TM) 82% faster in ray tracing than 6950 XT pushed people into buying 4080s?

lol what a bizarre argument

It's not bizarre, it falls in the same category of not acknowledging your competition when you already won the fight. It's the underdog that needs the direct comparison, needs to show the world their product matches or beats the leader. So if I see AMD loudly advertising RT performance while shying away from comparing with the competition, I deduct two things: AMD acknowledges RT is the future, AMD is weaker than Nvidia at RT.

AMD discreetly reassuring people that the 7000 series brings better RT performance is one thing, AMD telling the masses that 7000 series is an upgrade over 6000 series based on the RT performance is a different story with a different outcome.

 

[Timorous]

N32 will be cheaper than N21 and might be a touch more expensive than N22 but I don't think there will be much in it. N33 is cheaper than N23.

With the current prices AMD could offer an N32 based 7800XT with 6950XT performance at $600 with more margin.

They could offer a $500 7700XT with 6800XT performance which would also have more margin than the current 6800XT parts.

I think even a really cut $380 N32 based 7600XT with 6750XT performance would probably have more margin than the current 6750XT does.

A $280 7500XT with 6650XT-6700 performance would be a similar or better margin than the top N23 perts as well.

So at every tier AMD can improve their margin position Vs current RDNA2 products and gain a lot of mindshare.

Also when you consider the 4070 should really be a 4060 at around $400 it is not like the above stack is fantastic, better than NV but nowhere like offering 6800XT perf at $400.

 

[Timorous]

On the one hand, that slide can reduce the perception that AMD sucks at RT, but it also pushes the idea that RT performance is important, which indeed is bad for AMD.

Anyway, my argument was not that it is a very bad idea to have one slide about RT, but rather that it's a very bad idea to take the weak spot of your product and put your marketing focus on that. You want:

"Buy this product because it is better than the competition at X"

Not:

"Buy this product because it is better at Y than our previous generation, but worse at Y than our competition"

See the problem is that while the NV RT hardware is far better than what AMD offer NV gimped their cards with low VRAM.

The 3070Ti has far better raw RT performance than the 6800 but in any game that exceeds the 8GB VRAM buffer the 6800 can offer playable framerates with RT turned on and the 3070Ti cannot. You see the same happing with the 12GB 4070Ti and 4070. I would expect them to fall down the RT ranking as more games that exceed the 12GB buffer get added to RT test scenarios. If you look at the newest games like Jedi: Survivor. RE4, Forspoken etc the 7900XT is around on par with the 4070Ti in RT.

The other issue is that HUB showed how in some games that run out of VRAM rather than tanking frame rate they actually tank the LOD and texture swap all the time. So you can run a benchmark like Forspoken and get what look like normal FPS for the tier of card you are running but unless the reviewer shows you that it was not texture swapping I have no way of knowing if the FPS attained was at a like for like IQ which totally muddies the waters.

 

[PJVol]

3SEs with 20 CUs per SE seems like the config AMD will go with for N32 and if you look at an N31 die shot you can see exactly how it will be laid out.

Can you clarify? N31 has two rows of three SE's. How exactly are the 3 SE's arranged in N32, all in line?