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

 

[Timorous]

Why would be the margin for N32 higher than N31?
Price: $699 vs $999
That's a difference of $300.

I calculated the cost for making N31 and N33.
Wafer prices are 7nm at that time $8000, 5nm currently $15000, 6nm currently $7000.
MCD is 37mm2 -> 1547 good dies. 7000/1547 = ~ $4.5
N31 GCD is 300mm2 -> 160 good dies. 15000/160 = $94 + 6*$4.5 = $121, with packaging maybe $155.
N31(cut) GCD is 300mm2 -> 160 good dies. 15000/160 = $94 + 5*$4.5 = $116.5, with packaging maybe $148.
N32 GCD is 200mm2 -> 251 good dies. 15000/251 = $60 + 4*$4.5 = $78, with packaging maybe $100.
GDDR6 is $20 per 2GB GDDR6 module included in BOM.

MRSP price will be divided: 8% to shop, 3% shipping, 19% manufacturer margin, you will be left with 70% of MRSP for BOM and AMD's asked price for the chip.

	Chip cost	BOM	MSRP	70% of MRSP
RX 7900 XTX	$155	$330	$999	$699
RX 7900 XT	$148	$290	$899	$629
RX 7800 XT	$100	$230	$649-699	$489

AMD can set the price for manufacturers per chip at:
RX 7900 XTX: $699 - $330 = $369 (36.9% of MRSP)
RX 7900 XT: $629 - $290 = $339 (37.7% of MRSP)
RX 7800 XT: $489 - $230 = $259 (37% of MRSP)

Profit per chip:
RX 7900 XTX: $699 - $330 -$155 = $214
RX 7900 XTX: $629 - $290 -$148 = $191
RX 7800 XT: $489 - $230 - $100 = $159

Both N31 based models are more profitable than N32 for AMD.

159/330 (profit / cost) is 48.2% margin.
214/485 is 44.1% margin for the 7900XTX
191/438 is 43.6% margin for the 7900XT

So $700 7800XT has higher margin than both N31 products which is what I suggested.

Thanks for doing the math to prove my suspicion.

 

[TESKATLIPOKA]

Propose this to AMD lol.

Now when they already launched them? I don't think they will unlaunch them just because I think their naming sense is pretty bad.

There is a difference in TDP range AND clock speed though (game clocks are different between the 7600M XT and the 7700S, and between the 7600M and 7600S). So while similar in specs, they are not the same. Also, in AMD's case one can have a rough estimation of the specs (exact TDP apart) only from the name.

4.5% difference in clockspeed 7600M XT vs 7700S is because that is at their highest TDP 100W vs 120W.
Their TDP range is:
7600M XT: 75-120W
7700S: 75-100W
That means 7600M XT can also be limited to 100W.

If you compared 2 laptops, one with 7600M XT 100W vs 7700S 100W. How will they perform? Won't they perform the same?
The same with 7600M vs 7600S
Unless those S models are binned N33 then these models have no real reason to exist.
For 3 models you can guess the specs, but 7700S has simply wrong name. For no good reason, It's a higher tier.

 

[Kronos1996]

If Navi 33 beats the desktop 3060 by 31% as they claim, that would put it around RX 6800 performance on desktop. If they do plan on selling the 7600 XT for $400-450 then it needs to be faster then the 6700 XT so that would make sense. Otherwise it should be $350 max.

 

[insertcarehere]

@insertcarehere

If the 7800 XT is close in performance to the 7900 XT and priced very close to it, then the entire product is nearly pointless, since it will drive very little extra sales due to being so similar to the 7900 XT. They then also get the bad PR of having to explain why the price for this tier went up so much.

No, close pricing like this would make the existing Navi 31 products seem more attractive, and presumably allow them to get rid of stock for the current 'flawed' chips to clear the path for the 'fixed' N31, assuming there's such a thing, of course.

As for bad PR, AMD threw HEDT users with very expensive platforms under the bus for Zen 3 threadripper, they are not allergic to making business moves with negative PR.

And when it comes to laptops, the number of laptop models with a specific chip shows that the 3050, 3050 Ti and 3060 are the main sellers. There are less than half the models with a 3070 and almost half again, for the 3080. So unless something changes, which seems unlikely, AD107 will be the main seller on laptops for Nvidia and thus the main chip to compete with. You have to keep in mind that laptop makers have limited R&D capacity, so it takes time to make the new laptop models. Any extra model with N32 would for the next period just mean one fewer model with N33.

This implies that there's a fixed proportion of resources allocated per laptop OEM for each vendor, regardless of how many models or SKUs there may be, and that this 'extra' model for N32 wouldn't instead be taking away from 4070/4080 or other Nvidia SKUs. That I find hard to believe.

 

[TESKATLIPOKA]

159/330 (profit / cost) is 48.2% margin.
214/485 is 44.1% margin for the 7900XTX
191/438 is 43.6% margin for the 7900XT

So $700 7800XT has higher margin than both N31 products which is what I suggested.

Thanks for doing the math to prove my suspicion.

We were talking about AMD's margins, right?
Then BOM doesn't have anything to do with AMD, It has to do with manufacturers. AMD sells their chips to manufacturers for a price.

This is the price AMD could charge manufacturers.
X 7900 XTX: $699 - $330 = $369 (36.9% of MRSP)
RX 7900 XT: $629 - $290 = $339 (37.7% of MRSP)
RX 7800 XT: $489 - $230 = $259 (37% of MRSP)

Product margin= (selling price – cost of product) / selling price.
RX 7900 XTX: (369-155)/369 = 58%
RX 7900 XT: (339-148)/339 = 56%
RX 7800 XT: (259-100)/259 = 61%

You calculated It wrongly, but N32 still has higher margin.
BTW why does this margin even matter, when AMD has more $ from N31 than from N32?

edit: This was just an example I came up with. In reality, N31 could have higher margin than N32.

 

[TESKATLIPOKA]

If Navi 33 beats the desktop 3060 by 31% as they claim, that would put it around RX 6800 performance on desktop. If they do plan on selling the 7600 XT for $400-450 then it needs to be faster then the 6700 XT so that would make sense. Otherwise it should be $350 max.

Don't believe everything they claim.
I checked 2 games and It looks different from AMD's chart.

AMD presentation	RX 6700S 8GB (laptop)	RX 7700S 8GB (laptop)	RTX 3060 8GB (desktop)	RXT 3060 12GB (desktop)	RX 7600M XT 8GB (laptop)
Borderlands 3	75	104	69	77	106
Assassin's creed Valhalla	78	102	72	83	100

RX 6600 is the same chip as RX 6700S, but with higher clocks.
In Borderlands 3 RTX 3060 12GB is 20% faster.
In Valhalla they perform comparably.
These results are much different than what AMD claims. RX 6700S performs a lot better than It should.

I would only compare RX 6700S vs 7700S, but keep in mind It's:
RX 6700S vs RX 7700S
TDP: 80W vs 100W
Frequency: 1890 MHz vs 2200MHz
SPecs: 28CU vs 32CU
RX 6800S would be a much better comparison, but then performance would closer. Specs and TDP are the same, only clockspeed is different.

 

[insertcarehere]

Don't believe everything they claim.
I checked 2 games and It looks different from AMD's chart.

AMD presentation	RX 6700S 8GB (laptop)	RX 7700S 8GB (laptop)	RTX 3060 8GB (desktop)	RXT 3060 12GB (desktop)	RX 7600M XT 8GB (laptop)
Borderlands 3	75	104	69	77	106
Assassin's creed Valhalla	78	102	72	83	100

View attachment 74140View attachment 74141
RX 6600 is the same chip as RX 6700S, but with higher clocks.
In Borderlands 3 RTX 3060 12GB is 20% faster.
In Valhalla they perform comparably.
These results are much different than what AMD claims. RX 6700S performs a lot better than It should.

Far more simple than that, the unwillingness in the presentation to compare Navi 33 to anything more performant than 3060 on Nvidia's side and (binned) N23 on their own product stack should suggest plenty that comparisons with anything more powerful don't put N33 in a good light. If N33 did compare well against say 6700XT, AMD loses nothing and gains more interest by showing it outright.

 

[TESKATLIPOKA]

Far more simple than that, the unwillingness in the presentation to compare Navi 33 to anything more performant than 3060 on Nvidia's side and (binned) N23 on their own product stack should suggest plenty that comparisons with anything more powerful don't put N33 in a good light. If N33 djd compare well against say 6700XT, it would well be in AMD's interest to show this.

I don't mind them comparing It only to RTX 3060, but I don't like this desktop vs laptop comparison. They should have used only laptop models along with TDP.
N33 will be slower than N22, It just doesn't have specs or clocks to compare.

 

[Kaluan]

Don't believe everything they claim.
I checked 2 games and It looks different from AMD's chart.

AMD presentation	RX 6700S 8GB (laptop)	RX 7700S 8GB (laptop)	RTX 3060 8GB (desktop)	RXT 3060 12GB (desktop)	RX 7600M XT 8GB (laptop)
Borderlands 3	75	104	69	77	106
Assassin's creed Valhalla	78	102	72	83	100

View attachment 74140View attachment 74141
RX 6600 is the same chip as RX 6700S, but with higher clocks.
In Borderlands 3 RTX 3060 12GB is 20% faster.
In Valhalla they perform comparably.
These results are much different than what AMD claims. RX 6700S performs a lot better than It should.

I would only compare RX 6700S vs 7700S, but keep in mind It's:
RX 6700S vs RX 7700S
TDP: 80W vs 100W
Frequency: 1890 MHz vs 2200MHz
SPecs: 28CU vs 32CU
RX 6800S would be a much better comparison, but then performance would closer. Specs and TDP are the same, only clockspeed is different.

Well, I don't know what settings TPU used, they don't always use the highest settings, AMD did. I see many outlets bench Borderlands 3 w/ Very High or Ultra instead of Badass and AC:V with Very High instead of Ultra.

We don't know the power budgets, or even less so the actual measured TGP, AMD used desktop boards for all the mobile GPU tests, so it's not like we can refer to a specific laptop model. ASUS has a 6700S design that can be configured up to 100W (which it doesn't hit according to a nanoreview review), despite AMD's spec being "up to 80W". So all of this is a mess to correctly frame performance-wise.

AMD didn't even bother to use the same system (to their disadvantage maybe) when it benched the 3060 desktop vs 7600M XT on. 5600X/DDR4-3600 for the Radeon and 7600X/DDR5-6000 for the GeForce
lol


Also, can't seem find the endnote for the 6700S v 7700S test.

 

[biostud]

N33 is supposed to be the big seller for mobile, so they might not want to present it as a lower tier laptop solution right now. Also what TESKA said, it would divulge a lot about desktop.

It's different for AMD than for Nvidia, because all of Nvidia's mobile chips are small monolithic chips, including the mobile '4090' (which is really a 4080 desktop chip, which is really a x070 tier chip). N32 is not a great chip for laptops, I don't think.

@Timorous

If a $650 7800 XT gets close to a 7900 XT, then it will not just cannibalize that card, but a lot of people will probably get it over the 7900 XTX that is nearly twice as expensive. Ultimately, the 7900 XTX is not a halo product and needs to provide value, which it won't.

The 6900XT and 6800XT coexisted @ $649 and $999, so a $999 7900XTX and a $649 7800XT, is more sensible than the the pricing of the motherboards 6xxx series.


But it is pretty weird that N32 is Mia...

 

[Timorous]

We were talking about AMD's margins, right?
Then BOM doesn't have anything to do with AMD, It has to do with manufacturers. AMD sells their chips to manufacturers for a price.

This is the price AMD could charge manufacturers.
X 7900 XTX: $699 - $330 = $369 (36.9% of MRSP)
RX 7900 XT: $629 - $290 = $339 (37.7% of MRSP)
RX 7800 XT: $489 - $230 = $259 (37% of MRSP)

Product margin= (selling price – cost of product) / selling price.
RX 7900 XTX: (369-155)/369 = 58%
RX 7900 XT: (339-148)/339 = 56%
RX 7800 XT: (259-100)/259 = 61%

You calculated It wrongly, but N32 still has higher margin.
BTW why does this margin even matter, when AMD has more $ from N31 than from N32?

edit: This was just an example I came up with. In reality, N31 could have higher margin than N32.

It matters because AMD will sell more N32 than N31 due to there being more buyers at $700 than at $900+

 

[TESKATLIPOKA]

It matters because AMD will sell more N32 than N31 due to there being more buyers at $700 than at $900+

So this was your point. Now I understand.

I also made a comparison for N33.
RX 7600 XT: $449
N33 monolith 204mm2 -> 250 dies. 7000/250 = $28, with packaging maybe $40.
BOM: $140
70% of MRSP is $314.

This is the price AMD could charge manufacturers.
RX 7600 XT: $314 - $140 = $174 (39% of MRSP)

Profit per chip:
RX 7600 XT: $314 - $140 - $40 = $134
That's just $25 less than N32.

Product margin = (selling price – cost of product) / selling price.
RX 7600 XT: (174-40)/174 = 77%

If the MRSP is only $399:
RX 7600 XT: $279-$140=$139 -> 139-40=$99 -> (139-40)/139=71%
If the MRSP is only $349:
RX 7600 XT: $244-$140=$104 -> 104-40=$64 -> (104-40)/104=61.5%
If the MRSP is only $299:
RX 7600 XT: $209-$140=$69 -> 69-40=$29 -> (69-40)/69=42%
If the MRSP is only $269:
RX 7600 XT: $188-$140=$44 -> 44-40= $4 -> (44-40)/44=9%
If the MRSP is only $249:
RX 7600 XT: $174-$140=$34 -> 34-40= -$6 loss

If AMD is really greedy, then RX 7600 XT will cost $449.
If AMD is greedy, then RX 7600 XT will cost $399.
If AMD is sensible, then RX 7600 XT will cost $349.
If AMD wants to make a very good impression, then RX 7600 XT will cost $299.
If AMD wants to start a price war, then RX 7600 XT will cost $269.
If AMD wants to lose money, then RX 7600 XT will cost $249.

 

[Timorous]

So this was your point. Now I understand.
It's better to have higher margin for the cheaper product because you sell more of them and profit more.

I also made a comparison for N33.
N33: $449
N33 monolith 204mm2 -> 250 dies. 7000/250 = $28, with packaging maybe $40.
BOM: $140
70% of MRSP is $314.

This is the price AMD could charge manufacturers.
RX 7600 XT: $314 - $140 = $174 (39% of MRSP)

Profit per chip:
RX 7600 XT: $314 - $140 - $40 = $134
That's just $25 less than N32.

Product margin= (selling price – cost of product) / selling price.
RX 7600 XT: (174-40)/174 = 77%

If the MRSP is only $399:
RX 7600 XT: $279-$140=$139 -> 139-40=$99
RX 7600 XT: (139-40)/139=71%

If the MRSP is only $349:
RX 7600 XT: $244-$140=$104 -> 104-40=$64
RX 7600 XT: (104-40)/104=61.5%

If the MRSP is only $299:
RX 7600 XT: $209-$140=$69 -> 69-40=$29
RX 7600 XT: (69-40)/69=42%

I expect desktop N33 to be more like $350 but I expect AMD are charging less for laptop OEMs to get more design wins which is probably good in the long run.

 

[TESKATLIPOKA]

I expect desktop N33 to be more like $350 but I expect AMD are charging less for laptop OEMs to get more design wins which is probably good in the long run.

You were too fast. I was slow to edit It.

 

[Saylick]

Chips and Cheese posted an article where they microbenched RDNA 3. Enjoy!

Microbenchmarking AMD’s RDNA 3 Graphics Architecture

Editor’s Note (6/14/2023): We have a new article that reevaluates the cache latency of Navi 31, so please refer to that article for some new latency data. RDNA 3 represents the third iteratio…

chipsandcheese.com

 

[Stuka87]

Why would be the margin for N32 higher than N31?
Price: $699 vs $999
That's a difference of $300.

I calculated the cost for making N31 and N33.
Wafer prices are 7nm at that time $8000, 5nm currently $15000, 6nm currently $7000.
MCD is 37mm2 -> 1547 good dies. 7000/1547 = ~ $4.5
N31 GCD is 300mm2 -> 160 good dies. 15000/160 = $94 + 6*$4.5 = $121, with packaging maybe $155.
N31(cut) GCD is 300mm2 -> 160 good dies. 15000/160 = $94 + 5*$4.5 = $116.5, with packaging maybe $148.
N32 GCD is 200mm2 -> 251 good dies. 15000/251 = $60 + 4*$4.5 = $78, with packaging maybe $100.
GDDR6 is $20 per 2GB GDDR6 module included in BOM.

MRSP price will be divided: 8% to shop, 3% shipping, 19% manufacturer margin, you will be left with 70% of MRSP for BOM and AMD's asked price for the chip.

	Chip cost	BOM	MSRP	70% of MRSP
RX 7900 XTX	$155	$330	$999	$699
RX 7900 XT	$148	$290	$899	$629
RX 7800 XT	$100	$230	$649-699	$489

AMD can set the price for manufacturers per chip at:
RX 7900 XTX: $699 - $330 = $369 (36.9% of MRSP)
RX 7900 XT: $629 - $290 = $339 (37.7% of MRSP)
RX 7800 XT: $489 - $230 = $259 (37% of MRSP)

Profit per chip:
RX 7900 XTX: $699 - $330 -$155 = $214
RX 7900 XTX: $629 - $290 -$148 = $191
RX 7800 XT: $489 - $230 - $100 = $159

Both N31 based models are more profitable than N32 for AMD.

This doesn't make sense in reality. You aren't including the amount required to recoup R&D, or marketing, or any of the other costs involved with selling a product. It also assumes 100% chip yields. The profit is likely no where close to what you have listed here.

 

[TESKATLIPOKA]

This doesn't make sense in reality. You aren't including the amount required to recoup R&D, or marketing, or any of the other costs involved with selling a product. It also assumes 100% chip yields. The profit is likely no where close to what you have listed here.

I didn't assume 100% chip yields, but It's true I used only the good ones, the faulty ones were considered unusable, although in reality many should be reusable.

This is pure speculation, It's not like I know how much wafer or packaging cost, the same with BOM per card. How much is transport, cut for manufacturer, shop etc.
This was just an illustration how profit could look per different GPU at different price.
Even If It's not realistic, I think It's still an interesting read.

 

[PJVol]

Chips and Cheese posted an article

Tldr; "wait for the drivers" or in other words "buy the performace in future (61tflops)". Hope driver will learn to pack vopds by the rdna4 launch )

 

[TESKATLIPOKA]

Tldr; "wait for the drivers" or in other words "buy the performace in future (61tflops)". Hope driver will learn to pack vopds by the rdna4 launch )

I think everyone is already accustomed to how mature AMD drivers are at launch.
This is especially true for this generation, where they did a mayor change inside their execution units.
Performance will be certainly better, but we don't know by how much and when.

Nvidia supposedly working on AI-optimized drivers. I think AI could help AMD do faster and better driver updates.

 

[TESKATLIPOKA]

Chips and Cheese posted an article where they microbenched RDNA 3. Enjoy!

Microbenchmarking AMD’s RDNA 3 Graphics Architecture

Editor’s Note (6/14/2023): We have a new article that reevaluates the cache latency of Navi 31, so please refer to that article for some new latency data. RDNA 3 represents the third iteratio…

chipsandcheese.com

Fantastic article. This is why I love Chips and Cheese.

RDNA3 is not a bad architecture, but that dual issue is simply useless If It's left to the compiler.
Games will need heavy optimizations done by AMD's driver team to make this architecture "shine". This doesn't look optimistic.
I really have to wonder why AMD choose this path, when they knew how much work It needs to work correctly.

If the code is optimized correctly for VOPD instructions, then the improvement in TFLOPs is ~100%.

Curiously, RDNA 3 actually regresses in FP64 throughput. We already saw a hint of this earlier with OpenCL, where one RDNA 2 WGP could execute eight FP64 operations per cycle. RDNA 3 cuts throughput in half, meaning a WGP can do four FP64 operations – probably one per SIMD, per cycle.

Naturally, this whole optimization is by no means an easy task, but It should result in significant improvements in games in my opinion.

 

[Aapje]

RDNA3 is not a bad architecture, but that dual issue is simply useless If It's left to the compiler.
Games will need heavy optimizations done by AMD's driver team to make this architecture "shine". This doesn't look optimistic.
I really have to wonder why AMD choose this path, when they knew how much work It needs to work correctly.

Perhaps they are also working on an AI compiler

Seriously though, it might work very well to run games with a dynamic compiler like Java commonly uses and then have that dynamic compiler try out different optimizations and then measure the effect on frametimes. Then they can use this to create 'compiler hints' that are used when compiling the code that is actually shipped.

Unless they already do compilation on the consumer's machine, which the "shader compilation" messages that we commonly see nowadays suggests. If they do that, they can potentially ship the 'compiler hints' with the driver.

 

[TESKATLIPOKA]

Perhaps they are also working on an AI compiler

Or rather AI driver team.

 

[Bigos]

By default shaders are compiled in the driver while the assets are loading. Sometimes they are compiled on the fly (which is usually bad and causes stuttering), the compiled shaders are also often cached locally to speed up future loads.

(You might have heard of "shader compilation" as part of game engine asset creation. These are actually 2 separate stages of shader compilation - the first one compiles from HLSL/GLSL into some IR (DXIL, SPIR-V, ...) and is done by game developers, and the second produces real shader binaries and is done in the driver.)

It is true that VOPD instruction generation is not trivial and requires proper instruction scheduling and register allocation (both of which are standard compiler components). It is however not impossible and does not strictly require hand-written shader binaries prepared by the driver team.

Still, if the AMD driver team has prepared many hand-written shader binaries for RDNA1-2, they will not work with RDNA3. So previous game optimizations might not apply and require rework. This might be what "wait for RDNA3 game optimization" is all about.

Or maybe the current compiler tech does not generate VOPD instructions by itself often enough and the compiler itself needs more work. It would be pretty embarrassing, as the driver team should have been able to work on this for a few years already...

 

[leoneazzurro]

Now when they already launched them? I don't think they will unlaunch them just because I think their naming sense is pretty bad.


4.5% difference in clockspeed 7600M XT vs 7700S is because that is at their highest TDP 100W vs 120W.
Their TDP range is:
7600M XT: 75-120W
7700S: 75-100W
That means 7600M XT can also be limited to 100W.

If you compared 2 laptops, one with 7600M XT 100W vs 7700S 100W. How will they perform? Won't they perform the same?
The same with 7600M vs 7600S
Unless those S models are binned N33 then these models have no real reason to exist.
For 3 models you can guess the specs, but 7700S has simply wrong name. For no good reason, It's a higher tier.

Yes they will perform the same (if no clock limitation on one of the parts will kick in, that is, 7700S may be even more limited than a 7600M in this regard) and no one said two N33 at same TDP and clocks will perform differently lol. What I said is that these two variants have different TDP range settings and slight game clock differences, so while they can be set at the same value and have the same performance, in general they are not so generally they will perform (slightly) differently. And not only, they are in theory aimed at different market segments in the same way Nvidia Max-Q does (with similar issues in some cases as many Max-Q cards are quite TDP limited). Now this for us users may mean nothing, but for OEMs could mean different requirements about cooling solutions, voltage supply characteristics, and so on. This may be enough for a different name? Maybe not, but this is more a commercial issue. Also, it is clear that N33 is the replacement for N23, but N23 was also the 6800S, while here we have only up to the 7700S. This means that if we will have a 7800S, this will be N32 based (and so a 7700M and 7800M, if they will come).

 

[TESKATLIPOKA]

Yes they will perform the same (if no clock limitation on one of the parts will kick in, that is, 7700S may be even more limited than a 7600M in this regard) and no one said two N33 at same TDP and clocks will perform differently lol. What I said is that these two variants have different TDP range settings and slight game clock differences, so while they can be set at the same value and have the same performance, in general they are not so generally they will perform (slightly) differently. And not only, they are in theory aimed at different market segments in the same way Nvidia Max-Q does (with similar issues in some cases as many Max-Q cards are quite TDP limited). Now this for us users may mean nothing, but for OEMs could mean different requirements about cooling solutions, voltage supply characteristics, and so on. This may be enough for a different name? Maybe not, but this is more a commercial issue. Also, it is clear that N33 is the replacement for N23, but N23 was also the 6800S, while here we have only up to the 7700S. This means that if we will have a 7800S, this will be N32 based (and so a 7700M and 7800M, if they will come).

I can agree with this.