Question Speculation: RDNA3 + CDNA2 Architectures Thread

Page 87 - 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

biostud

Lifer
Feb 27, 2003
18,700
5,434
136
Portugal has a 4% higher VAT than Germany. That €779 price is €655 without German VAT, which becomes €806 with Portugese VAT.

What I did notice is that the Dutch and Spanish Amazon have nearly the same price for the cheapest 6900 XT, but that is ~€900. The much cheaper Dutch prices are from more country-bound sellers. So that suggests that a lack of competition might be the issue.
Denmark has the highest VAT of EU with 25%, but is still cheaper so I guess yay!
 
  • Like
Reactions: Tlh97 and H T C

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
Right now it's almost necessary to test with both AMD and Intel CPUs. Although the average is quite close, each has multiple titles where they perform significantly better than the competition.

Even when we get a 7800X 3D (or whatever it winds up being called) I still think we'll see a few cases where it's not quite at the top.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,537
3,468
106
It's a little more complex than this, being an American company doesn't mean all required transactions are done in US$. Instead it's preferred to do all local transactions in the local currency. The interesting question is what currencies are used when crossing specific borders. In AMD's case all dies nowadays are manufactured in Taiwan (by TSMC), and depending on products assembly and testing afaik happens either in Malaysia (by AMD Global Services) or through third parties in Taiwan and China. When to keep local currency and when to exchange it is a complex business of its own.


Aside differing sales taxes the prices may be more of a reflection of the respective market size (affecting the amount of internal price competition). In Germany RX 6900 XT is available from 779€:

I think most of TSMC contracts are in USD. You will notice, in TSMC earnings report, that their earnings (announced in Taiwanese dollars) fluctuate. When dollar is high, earnings are higher.
 

tajoh111

Senior member
Mar 28, 2005
307
331
136
I think we are going to get something along the lines of 1.6 to 1.7x the performance in real world performance vs a 6900xt. Ray tracing performance will be great than 2x.

If the angstronomics is correct, whatever AMD did to double the number of shaders was very cheap in terms of silicon as Navi 33 die is smaller than navi 23 yet they have doubled the number of shaders.

Combined with the very minor transistor density improvement of TSMC 6nm, I think this architecture is going to be like amphere were compute goes up but gaming performance is going to be quite modest relative to the compute increase.

However power consumption will not increase that much and it will be pointless for the most part as I think AMD has also not increased the pipeline that much to accommodate this wider architecture.

I think Navi 33 is going to be 1.2 to 1.3x a navi 23 but at 120watts instead of 165 watts.

I think navi 31 clocks are going to be more like 2.6-2.8ghz and have power more in the 330 range as the cooler AMD has kind of previewed so far does not look like a 400 watt cooler. That is just something slightly bigger than 2 slots if not a 2 slot card. When you add the ray tracing increase, it helps AMD claim of a greater than 1.5x performance/watt increase.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,537
3,468
106
I think we are going to get something along the lines of 1.6 to 1.7x the performance in real world performance vs a 6900xt. Ray tracing performance will be great than 2x.

If the angstronomics is correct, whatever AMD did to double the number of shaders was very cheap in terms of silicon as Navi 33 die is smaller than navi 23 yet they have doubled the number of shaders.

Combined with the very minor transistor density improvement of TSMC 6nm, I think this architecture is going to be like amphere were compute goes up but gaming performance is going to be quite modest relative to the compute increase.

However power consumption will not increase that much and it will be pointless for the most part as I think AMD has also not increased the pipeline that much to accommodate this wider architecture.

I think Navi 33 is going to be 1.2 to 1.3x a navi 23 but at 120watts instead of 165 watts.

I think navi 31 clocks are going to be more like 2.6-2.8ghz and have power more in the 330 range as the cooler AMD has kind of previewed so far does not look like a 400 watt cooler. That is just something slightly bigger than 2 slots if not a 2 slot card. When you add the ray tracing increase, it helps AMD claim of a greater than 1.5x performance/watt increase.

In other words, N31 below 4080?
 

maddie

Diamond Member
Jul 18, 2010
4,881
4,951
136
I think we are going to get something along the lines of 1.6 to 1.7x the performance in real world performance vs a 6900xt. Ray tracing performance will be great than 2x.

If the angstronomics is correct, whatever AMD did to double the number of shaders was very cheap in terms of silicon as Navi 33 die is smaller than navi 23 yet they have doubled the number of shaders.

Combined with the very minor transistor density improvement of TSMC 6nm, I think this architecture is going to be like amphere were compute goes up but gaming performance is going to be quite modest relative to the compute increase.

However power consumption will not increase that much and it will be pointless for the most part as I think AMD has also not increased the pipeline that much to accommodate this wider architecture.

I think Navi 33 is going to be 1.2 to 1.3x a navi 23 but at 120watts instead of 165 watts.

I think navi 31 clocks are going to be more like 2.6-2.8ghz and have power more in the 330 range as the cooler AMD has kind of previewed so far does not look like a 400 watt cooler. That is just something slightly bigger than 2 slots if not a 2 slot card. When you add the ray tracing increase, it helps AMD claim of a greater than 1.5x performance/watt increase.
I think everything you wrote here is a big miss.
 

exquisitechar

Senior member
Apr 18, 2017
684
942
136
I think we are going to get something along the lines of 1.6 to 1.7x the performance in real world performance vs a 6900xt. Ray tracing performance will be great than 2x.

If the angstronomics is correct, whatever AMD did to double the number of shaders was very cheap in terms of silicon as Navi 33 die is smaller than navi 23 yet they have doubled the number of shaders.

Combined with the very minor transistor density improvement of TSMC 6nm, I think this architecture is going to be like amphere were compute goes up but gaming performance is going to be quite modest relative to the compute increase.

However power consumption will not increase that much and it will be pointless for the most part as I think AMD has also not increased the pipeline that much to accommodate this wider architecture.

I think Navi 33 is going to be 1.2 to 1.3x a navi 23 but at 120watts instead of 165 watts.

I think navi 31 clocks are going to be more like 2.6-2.8ghz and have power more in the 330 range as the cooler AMD has kind of previewed so far does not look like a 400 watt cooler. That is just something slightly bigger than 2 slots if not a 2 slot card. When you add the ray tracing increase, it helps AMD claim of a greater than 1.5x performance/watt increase.
You're underestimating Navi 33, I think. Navi 31 should clock higher than that too. I don't think ray tracing will be necessary for a >1.5x perf/watt improvement.
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
Yeah I know, but I didn't know there were so huge difference within EU.
Yeah, it's kind of sad really considering not only companies but also customers can easily profit of the single market, the latter just doesn't happen often enough as the disparity in prices shows. I can only guess that that's mostly language barriers at work there.

I think most of TSMC contracts are in USD. You will notice, in TSMC earnings report, that their earnings (announced in Taiwanese dollars) fluctuate. When dollar is high, earnings are higher.
Bookkeeping is likely done in USD for most Western customers, payment may still be done in TWD considering plenty ODMs then building products around the chips are Taiwanese (even some big ones with massive factories in China, e.g. Foxconn etc.).
 

Aapje

Golden Member
Mar 21, 2022
1,515
2,065
106
I think it is from igorslab pcb leak.

Yes, but he does say the rather puzzling:
  • In the presumed board design, we see three 6+2 sockets instead of the 12VHPWR. I’ll leave it open whether this PCIe 5.0 connector will make it onto the final cards.
This seems to suggest that we can't take those three connectors in the picture as good evidence that they will choose this.
 

Aapje

Golden Member
Mar 21, 2022
1,515
2,065
106
They've already got that done for CDNA 2 on the MI200 cards. However, I don't foresee them doing that yet for RDNA 3, because of the latency risks and known issues with SLI, which are far less of an issue for workstation workloads.

I would expect that in RDNA 4 at the soonest.
 
  • Like
Reactions: Tlh97 and RnR_au

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
If the angstronomics is correct, whatever AMD did to double the number of shaders was very cheap in terms of silicon as Navi 33 die is smaller than navi 23 yet they have doubled the number of shaders.

Combined with the very minor transistor density improvement of TSMC 6nm, I think this architecture is going to be like amphere were compute goes up but gaming performance is going to be quite modest relative to the compute increase.

However power consumption will not increase that much and it will be pointless for the most part as I think AMD has also not increased the pipeline that much to accommodate this wider architecture.

I think Navi 33 is going to be 1.2 to 1.3x a navi 23 but at 120watts instead of 165 watts.
You are underestimating N33 a bit too much.
Just by having clocks >3GHz would mean >20% better performance.
Ampere gained ~30% better performance by moving from 64 FP32 + 64 INT32 to 64FP32 + 64 INT32/FP32.
RDNA3 should be moving from 64 FP32/INT32 to 128 FP32/INT32.
Unlike Ampere there are 2x more shaders, so I expect higher gains than what Ampere got, I think 50% shouldn't be too unreasonable.
120W is too low, but It should be under 160-170W or something like that according to Bondrewd.
 
Last edited:

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
You are underestimating N33 a bit too much.
Just by having clocks >3GHz would mean >20% better performance.
Ampere gained ~30% better performance by moving from 64 FP32 + 64 INT32 to 64FP32 + 64 INT32/FP32.
RDNA3 should be moving from 64 FP32/INT32 to 128 FP32/INT32.
Unlike Ampere there are 2x more shaders, so I expect higher gains than what Ampere got, I think 50% shouldn't be too unreasonable.
120W is too low, but It should be under 160-170W or something like that according to Bondrewd.
The gain should be nearly linear.

At least for N31 and 32. N33 and PHX should get 80% scaling from increased amount of FP32 per CU.
 
  • Like
Reactions: Leeea

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
The gain should be nearly linear.

At least for N31 and 32. N33 and PHX should get 80% scaling from increased amount of FP32 per CU.
RDNA3 WGP is slightly smaller than RDNA2 WGP on the same process node according to Angstronomics, so having 80-95% increase in performance sounds too unrealistic.
Rumored 2X higher performance for N31 compared to N21 would also be too low, considering 48WGPs would be equivalent to >90 WGPs RDNA2 and N21 has 40WGPs. Then you also have a lot higher clockspeed.
 
Last edited: