Question Speculation: RDNA3 + CDNA2 Architectures Thread

Page 97 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Platinum Member
Oct 16, 2019
2,705
6,427
146

Tup3x

Golden Member
Dec 31, 2016
1,086
1,085
136
Greymon is doing some incredibly cryptic tweets. Could mean just about anything and can be interpreted positively or negatively, e.g. Nvidia's full AD102 is still faster, the 4090 is still faster, Nvidia is king of power consumption, price, etc. Could even mean nothing at all and is just posting to get people's attention. If so, I guess I was caught lol

I bet many would be more than happy with product that's 90% of what 4090 is if it offers better value for the money.
 
  • Like
Reactions: Tlh97 and prtskg

maddie

Diamond Member
Jul 18, 2010
4,881
4,951
136
As posted here earlier, Angstronomics has leaked an image of the "Navi 31" package, which shows the 7 chiplet design (a single GCD and 6 MCDs flanking it). This makes me think: Why is the GCD just one big chiplet? Why are the compute units not spread over smaller chiplets? I would think the compute units work pretty much independently, perhaps except for the writing to a shared frame buffer. Is the latter the bottleneck, perhaps, making it beneficial to keep them all on the same die? Or is it the overall controller/scheduler that is the bottleneck, and hence needs to be on the same die as all the compute units?

Navi 31 Die Shot - by SkyJuice - Angstronomics
The recent reveal of AMD changing (abandoning ?) the command processor in RDNA3 to a software route a la NVIDIA, might suggest dual a compute die product is more possible. (2) N32 with slightly lower clocks could be a monster.
 

Tuna-Fish

Golden Member
Mar 4, 2011
1,486
2,023
136
I would think the compute units work pretty much independently, perhaps except for the writing to a shared frame buffer.

This is specifically the problem. For traditional rendering, you are not just writing but also constantly reading a shared framebuffer, and all those accesses need to be sorted right. All compute units need to be able to simultaneously access all of the framebuffer in a coherent way.

There is one gigantic crossbar or ring bus or something on the main chip that gets all traffic from any compute unit to any memory-side unit. Then each MCD to GCD link needs to be roughly as capable as the MCD is capable of providing bandwidth. When you split the GCD into multiple ones, then each of them either needs their own link to each MCD (which is a routing nightmare), or there needs to be a link between each GCD to every other GCD that is roughly as capable of all the MCD connected to those GCDs. This means you need to have much more peak interconnect bandwidth for the same result.

This is why the AMD stacking patent where GCDs are stacked on top of a larger MCD that is acting as a bridge between them was so interesting. I wonder if that's an approach they trialed and didn't use, or if it's coming later.
 

In2Photos

Platinum Member
Mar 21, 2007
2,026
2,054
136
If RDNA3 is good enough, performance wise, to compete with Nvidia I just hope they plan to release more than just the flagship model on release day. I've been waiting for AM5 to mature a little before building my next PC. I'd love to pick up a 7700 or 7800 variant if the price is right.
 

gdansk

Platinum Member
Feb 8, 2011
2,966
4,498
136
If RDNA3 is good enough, performance wise, to compete with Nvidia I just hope they plan to release more than just the flagship model on release day. I've been waiting for AM5 to mature a little before building my next PC. I'd love to pick up a 7700 or 7800 variant if the price is right.
Rumors have pointed to those being next year. But we'll see.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,539
3,473
106
As posted here earlier, Angstronomics has leaked an image of the "Navi 31" package, which shows the 7 chiplet design (a single GCD and 6 MCDs flanking it). This makes me think: Why is the GCD just one big chiplet? Why are the compute units not spread over smaller chiplets? I would think the compute units work pretty much independently, perhaps except for the writing to a shared frame buffer. Is the latter the bottleneck, perhaps, making it beneficial to keep them all on the same die? Or is it the overall controller/scheduler that is the bottleneck, and hence needs to be on the same die as all the compute units?

Navi 31 Die Shot - by SkyJuice - Angstronomics

I don't think anyone has managed to break up gaming GPU effectively. Bandwidth and latency requirements are likely the bottleneck.

But TSMC is floating this SoIC_H which would make 2 dies act effectively as a single die. We will see if it comes to fruition soon. Maybe RDNA4

But with RDNA3, I think it may be feasible to extend the architecture to bigger die and 8 MCDs, if AMD wanted to build an ultimate GPU. ~400 to 450 mm2 die would still be within reason for GCD die, and still much smaller than NVidia die
 
  • Like
Reactions: Tlh97 and Vattila

jpiniero

Lifer
Oct 1, 2010
15,223
5,768
136
If RDNA3 is good enough, performance wise, to compete with Nvidia I just hope they plan to release more than just the flagship model on release day. I've been waiting for AM5 to mature a little before building my next PC. I'd love to pick up a 7700 or 7800 variant if the price is right.

If AMD stuck to the RDNA 2 schedule, the N32 model would be released in March or therebouts. And there would only be one desktop model with mobile version getting the cutdown chips. N33 would be August.
 

jpiniero

Lifer
Oct 1, 2010
15,223
5,768
136
But with RDNA3, I think it may be feasible to extend the architecture to bigger die and 8 MCDs, if AMD wanted to build an ultimate GPU. ~400 to 450 mm2 die would still be within reason for GCD die, and still much smaller than NVidia die

It's still an open question as to whether AMD can get people to buy a $1k GPU, regardless of how it compares to nVidia's $1k+ GPUs. That's likely why such a product doesn't exist.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,539
3,473
106
If RDNA3 is good enough, performance wise, to compete with Nvidia I just hope they plan to release more than just the flagship model on release day. I've been waiting for AM5 to mature a little before building my next PC. I'd love to pick up a 7700 or 7800 variant if the price is right.

I resigned to waiting, probably another 3 months, to get what I really want...
 

Joe NYC

Platinum Member
Jun 26, 2021
2,539
3,473
106
It's still an open question as to whether AMD can get people to buy a $1k GPU, regardless of how it compares to nVidia's $1k+ GPUs. That's likely why such a product doesn't exist.

We will see what the reception is.

But just a couple of days ago, the top CPU on AMD.com was 6950XT for $999.

Today, it dropped to $849.

This, to me, implies that 7900XT will be $999

Maybe the XTX ~1099 - 1199.
 
Jul 27, 2020
20,040
13,739
146
My guess is XTX at $1300. They will want to milk the newness of this product as much as they can. Especially if nothing at this price point is able to come close to it.
7900 XT at $1200.
7800 XT $1100.
7800 $1000.
 
  • Haha
Reactions: Kaluan

jpiniero

Lifer
Oct 1, 2010
15,223
5,768
136
We will see what the reception is.

But just a couple of days ago, the top CPU on AMD.com was 6950XT for $999.

Today, it dropped to $849.

This, to me, implies that 7900XT will be $999

Maybe the XTX ~1099 - 1199.

I still think it completely depends on it's performance compared to Ada with maybe a small discount. It becomes foggier if you throw in RT into the mix. Since we don't have a legit source on actual competitiveness, it'd be tough to guess on pricing.

The cut N31 should be easily faster than the 4080 16 GB in raster though.
 
  • Like
Reactions: Joe NYC

Saylick

Diamond Member
Sep 10, 2012
3,532
7,859
136
We will see what the reception is.

But just a couple of days ago, the top CPU on AMD.com was 6950XT for $999.

Today, it dropped to $849.

This, to me, implies that 7900XT will be $999

Maybe the XTX ~1099 - 1199.
The 6950XT would have to drop a little more if the 7900XT is $999, but I do agree with you regarding where RDNA 3 pricing likely falls.

Having the same $999 price going from 6900XT to 7900XT makes a lot of sense from a marketing perspective, while offering 40-50% more performance. The 7900XTX, which offers another 20% on top of that, should be >20% more costly, so $1299 makes sense.
 

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
Although their GPUs will be the main focus of this, I hope they throw in a sneak preview of a Zen 4 v-cache CPU at the end.

Anyone have any bold predictions for the event today? It almost feels a little anticlimactic right now given how fast the hype train is typically rolling through in this thread. The calm before the storm?
 

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
Although their GPUs will be the main focus of this, I hope they throw in a sneak preview of a Zen 4 v-cache CPU at the end.

Anyone have any bold predictions for the event today? It almost feels a little anticlimactic right now given how fast the hype train is typically rolling through in this thread. The calm before the storm?

At least 2 people will be disappointed, but at least 5 will not be.

In all seriousness, AMD undercut NVIDIA only by a small amount (price wise) last gen. While I am not expecting them to be significantly cheaper than the 4xxx series, it would be a pleasant surprise if they kept pricing for halo tier cards somewhat close to what they were last year.

EDIT: On the AT site, this was noted: For example, any “Radeon RX 7000 series” branding is completely absent from all of AMD’s official communiques.

Anyone expect a rebrand? I am curious if they will skip 7xxx or if maybe they rebrand Radeon to something else. My hunch says neither, but it is curious AMD hasn’t said anything about Radeon in reference to next gen.
 
  • Like
Reactions: lightmanek

Saylick

Diamond Member
Sep 10, 2012
3,532
7,859
136
Although their GPUs will be the main focus of this, I hope they throw in a sneak preview of a Zen 4 v-cache CPU at the end.

Anyone have any bold predictions for the event today? It almost feels a little anticlimactic right now given how fast the hype train is typically rolling through in this thread. The calm before the storm?
I have high hopes that RDNA 3 can beat the 4090 given that a TON of stuff under the hood was changed, updated, or improved, but the reasonable man in me says I should temper expectations. As much as we want AMD to strike for the crown, we know historically they tend not to. Instead, they've preferred to offer a balanced product that is more accessible to the masses, even if the performance isn't chart-topping.

Bold predictions would be if we got a 3D cache version of N31 that also launches at the same time as the other two SKUs. Another bold prediction would be if we get a glimpse of Zen 4 X3D with top N31 leveraging latest AMD CPU+GPU tech.
 

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
EDIT: On the AT site, this was noted: For example, any “Radeon RX 7000 series” branding is completely absent from all of AMD’s official communiques.

Anyone expect a rebrand? I am curious if they will skip 7xxx or if maybe they rebrand Radeon to something else. My hunch says neither, but it is curious AMD hasn’t said anything about Radeon in reference to next gen.

That would be really weird considering they launched Zen 4 as the 7000 series in what to me looked like (at least in part) an attempt to align the numbering for their CPUs and GPUs.

I think we'll see a rebrand when they get closer to going beyond a 99XX which would probably be whatever comes after RDNA as the 9xxxx series cards would be RDNA5 and they architecture would be getting quite long in the tooth by that point.

Intel just kept chugging along though and have a 13xxx series now, so it wouldn't be completely unprecedented for it to continue, but by the same token we'd have almost surely have seen some kind of leak if there were new branding. It's even harder to stealth launch something like that than the actual physical product.
 
  • Like
Reactions: Tlh97 and Geddagod

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
Bold predictions would be if we got a 3D cache version of N31 that also launches at the same time as the other two SKUs. Another bold prediction would be if we get a glimpse of Zen 4 X3D with top N31 leveraging latest AMD CPU+GPU tech.

Showing off a v-cache RDNA3 would be even cooler, but I don't know if they'd do that unless they have a launch date for it that's very soon. Otherwise they essentially kill demand for the card they are launching.

It seems more likely they would keep that in the chamber to roll out in response to a 4090 Ti. If we see something like that, I think it would be a product that they intend to launch within 3 months and that we'd get performance figures and price for it so we can see how it fits in the product stack. Even if it won't be available until early next year, I could see them doing it if it lets them secure the performance crown. They would be able to have a 4090 in hand to compare it to, so it's definitely possible.

What they do against the 4080 will be far more interesting in my opinion. I think they can easily beat that and at a much better price. I also wonder if their plans have changed at all due to the 12 GB 4080 getting canceled. Does AMD also hold back whatever they might have launched to compete with that?
 

Rekluse

Member
Sep 16, 2022
36
46
51
While difficult in terms of their current high ASP strategy, I think it's downright silly to release a 1k+ card in a client computing market that has clearly softened. Especially when your main competitor is desperately slashing prices to get rid of old stock.

$999 Serves the purpose of shutting down nVidia's downrange sales while getting word of mouth out on how much of a kingslayer AMD is for the price and subsequently gaining marketshare. Which, given market conditions, would smash team green despite AMD not taking the crown cleanly on performance nor having the brand clout.
 

Karnak

Senior member
Jan 5, 2017
399
767
136
While difficult in terms of their current high ASP strategy,
That's why they won't do it. Su was pretty clear in terms of AMD's ASP and cutting prices in general during their Earnings Call a few days ago.

If they got a GPU which can match the 4090 (even though if it's only in raster) they'll price it accordingly. Not as high as NV - because AMD simply can't to be honest, mainly because of 'mindshare' - but I'd say somewhere around $1400. Still a lot, but $200 less is still a decent amount.
 

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
If nothing else it should indicate that AMD is eventually (no idea when Navi 32 will hit the market) going to offer 6900XT performance for ~$700.