Question Speculation: RDNA2 + CDNA Architectures thread

Page 77 - 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
All die sizes are within 5mm^2. The poster here has been right on some things in the past afaik, and to his credit was the first to saying 505mm^2 for Navi21, which other people have backed up. Even still though, take the following with a pich of salt.

Navi21 - 505mm^2

Navi22 - 340mm^2

Navi23 - 240mm^2

Source is the following post: https://www.ptt.cc/bbs/PC_Shopping/M.1588075782.A.C1E.html
 

Viking Warrior

Junior Member
Aug 25, 2020
4
3
41
Yeah, I would also like to see that diagram
Here's the crazy thing about this,there are 160 CU's on this die, 80 x 2 in dual configuration,each CU has 1 TMU,which means a ratio of 1:1,this chip should have at least 320 TMU's and 128 ROP's.I'm definitely missing something.I'm going to try and track down the origins of the diagram,if it is fake someone went to a lot of trouble because the work is as good as it gets.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,522
3,037
136
Here's the crazy thing about this,there are 160 CU's on this die, 80 x 2 in dual configuration,each CU has 1 TMU,which means a ratio of 1:1,this chip should have at least 320 TMU's and 128 ROP's.I'm definitely missing something.I'm going to try and track down the origins of the diagram,if it is fake someone went to a lot of trouble because the work is as good as it gets.
It's enough If you you just provide that diagram.
 

Konan

Senior member
Jul 28, 2017
360
291
106
The 40 CU/192 bit GPU is this year.

So this is Navy Flounder = reported on several times as Navi22 which has already been said to be @ 340mm2 40CU and 192bit therefore technically (with 14-16,000GHz 12GB GDDR6) should only reach a 2080S maximum. Between 2070S and 2080S.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
So this is Navy Flounder = reported on several times as Navi22 which has already been said to be @ 340mm2 40CU and 192bit therefore technically (with 14-16,000GHz 12GB GDDR6) should only reach a 2080S maximum. Between 2070S and 2080S.
Nope.

Three biggest dies are 500, 340, 240 mm2. If 40 CU die is 340 mm2 with 40 CUs, then the largest die cannot be 500 mm2 with 80 CUs.

That 40 CU/192 bit die is 240 mm2 one.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,706
1,233
136
I can see 505 mm2 and 340 mm2 for Arcturus and Sienna Cichlid if they were 5nm pitched.
240 mm2 for Navy Flounder is obviously 7nm.
Navi23 = 24 CUs, similar size to Navi14.
Navi24 = 10 CUs, reduced die size from Polaris 550.

Navi21 = 256-bit
Navi22 = 192-bit (-64-bit)
Navi23 = 128-bit (-64-bit) // Same as Van Gogh/Mero and Xbox Series S. 1032=1033=1040
Navi24 = 64-bit (-64-bit)

(240 * 2) x 0.7 [Average pitch(SerDes/SRAM/Logic)] => 336 mm2
 
Last edited:
  • Like
Reactions: Konan and Summerfun

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
So this is Navy Flounder = reported on several times as Navi22 which has already been said to be @ 340mm2 40CU and 192bit therefore technically (with 14-16,000GHz 12GB GDDR6) should only reach a 2080S maximum. Between 2070S and 2080S.

The leaked specs are very likely from the RX 6600 XT.
 

leoneazzurro

Golden Member
Jul 26, 2016
1,052
1,716
136
So this is Navy Flounder = reported on several times as Navi22 which has already been said to be @ 340mm2 40CU and 192bit therefore technically (with 14-16,000GHz 12GB GDDR6) should only reach a 2080S maximum. Between 2070S and 2080S.

It was supposed that N22 was Navy flounder. There is no official info about that.
 
  • Like
Reactions: Konan

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
So, I am going to risk some speculation here based on available information. I am not all that confident, but we will see.
  1. RX 6500 XT - 36 CUs - small die - $199? - ~2070S perf
  2. RX 6600 XT - 40 CUs - small die - $299? - ~2080S perf
  3. RX 6700 XT - 52 CUs - med die - $399? - a bit faster than the 3070.
  4. RX 6800 XT - 56 CUs - med die - $499? Behind the 3080, well ahead of the 3070.
  5. RX 6900 XT - 72CUs - large die - $599? a bit ahead of the 3080. Around 40%-50% faster than the 2080ti. 275W TDP?
  6. RX 6950 XT - 80CUs - large die (launches later?) - 3090 equiv. Likely $799-$999 depending on yields. 24gb?
Non XT parts slot in between. The 80CU part will not launch with the rest due to the extra time needed for binning. It doesn’t matter, however, as perf gains won’t be big.

On most gaming workloads, AMD will win or at least be tied. On very specific workloads that can take advantage of NVIDIA’s new core configuration, NVIDIA will win.

Rasterization only. No idea about RT.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
So, I am going to risk some speculation here based on available information. I am not all that confident, but we will see.
  1. RX 6500 XT - 36 CUs - small die - $199? - ~2070S perf
  2. RX 6600 XT - 40 CUs - small die - $299? - ~2080S perf
  3. RX 6700 XT - 52 CUs - med die - $399? - a bit faster than the 3070.
  4. RX 6800 XT - 56 CUs - med die - $499? Behind the 3080, well ahead of the 3070.
  5. RX 6900 XT - 72CUs - large die - $599? a bit ahead of the 3080. Around 40%-50% faster than the 2080ti. 275W TDP?
  6. RX 6950 XT - 80CUs - large die (launches later?) - 3090 equiv. Likely $799-$999 depending on yields. 24gb?
Non XT parts slot in between. The 80CU part will not launch with the rest due to the extra time needed for binning. It doesn’t matter, however, as perf gains won’t be big.

On most gaming workloads, AMD will win or at least be tied. On very specific workloads that can take advantage of NVIDIA’s new core configuration, NVIDIA will win.

Rasterization only. No idea about RT.

How 40 CUs , same amount as RX5700XT will reach 2080 Super performance that is 20% faster today ??
52 CUs will not get close to 2080Ti/3070, perhaps close to 2080 Super/3060
72 CUs will not get 40-50% faster vs 2080Ti

Hell the entire stack is way off
 
  • Like
Reactions: Konan

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
So, I am going to risk some speculation here based on available information. I am not all that confident, but we will see.
  1. RX 6500 XT - 36 CUs - small die - $199? - ~2070S perf
  2. RX 6600 XT - 40 CUs - small die - $299? - ~2080S perf
  3. RX 6700 XT - 52 CUs - med die - $399? - a bit faster than the 3070.
  4. RX 6800 XT - 56 CUs - med die - $499? Behind the 3080, well ahead of the 3070.
  5. RX 6900 XT - 72CUs - large die - $599? a bit ahead of the 3080. Around 40%-50% faster than the 2080ti. 275W TDP?
  6. RX 6950 XT - 80CUs - large die (launches later?) - 3090 equiv. Likely $799-$999 depending on yields. 24gb?
Non XT parts slot in between. The 80CU part will not launch with the rest due to the extra time needed for binning. It doesn’t matter, however, as perf gains won’t be big.

On most gaming workloads, AMD will win or at least be tied. On very specific workloads that can take advantage of NVIDIA’s new core configuration, NVIDIA will win.

Rasterization only. No idea about RT.
Nope.
 
  • Haha
Reactions: kurosaki

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
How 40 CUs , same amount as RX5700XT will reach 2080 Super performance that is 20% faster today ??
52 CUs will not get close to 2080Ti/3070, perhaps close to 2080 Super/3060
72 CUs will not get 40-50% faster vs 2080Ti

Hell the entire stack is way off
7% higher IPC and much higher clock speeds. Call me out if you want, but set a reminder for launch day.

EDIT: The 6700XT will have clocks north of 2 ghz, for example.
 

leoneazzurro

Golden Member
Jul 26, 2016
1,052
1,716
136
Hmm... I made some calculations based on the Xbox series S, Xbox X (Scorpio) and the 5500XT. According to Techpowerup, at 1080p both Scorpio and 5500XT GPUs perform like the RX580. There are some difference in the bandwidth, but quite limited. Microsoft declared that Xbox Series S is faster than Scorpio, but let's assume their GPU is performing the same and the difference is due to the better CPU cores in the Series S. This also means that 20 RDNA2 CU @ 1565 MHz perform the same as 22 RDNA1 CU @1717 MHz.
This means that IPC increase from RDNA1 to RDNA2 is around 20%. Let's consider that the Xbox Series X has a 25% more bandwidth (but not quite fully useable due to being divided in two asymmetrical segments). Let's say that this amounts for another 5% of IPC gain. Thsi means that, CU vs CU, RDNA2 should have around 15% better IPC respect to RDNA1.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
Prove it. I will go even further than this and possibly even estimate clock speeds and FP32 TFLOPs. (22 on the biggest chip)
Its simple.

You have no idea which Die will be used for which SKU. AMD will not harvest 40 CU dies, which will have performance of RTX 3070 to go to 6500 SKU range. They will have apropriate price point, if they will perform accordingly.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
This means that IPC increase from RDNA1 to RDNA2 is around 20%. Let's consider that the Xbox Series X has a 25% more bandwidth (but not quite fully useable due to being divided in two asymmetrical segments). Let's say that this amounts for another 5% of IPC gain. Thsi means that, CU vs CU, RDNA2 should have around 15% better IPC respect to RDNA1.
Considering this is speculation thread, I'll bite, even tho Im a complete layman in terms of highly technical stuff.

First of all, RDNA2 appears to schedule more work with each cycle, based on Microsoft's own papers on RDNA2. Secondly, we have all of that Cache rumors that will happen with RDNA2, and most importantly we have leaked recently patents that are discussing new L1 cache design and CU clustering designs.

Both: more work scheduled each cycle, and L1 cache redesign would bring massive IPC increases compared to previous generation.

Important note. In this discussion I use word: "Would". We have no idea if RDNA2 will have that new Cache design, even tho, rumors point into this possibility.
 

leoneazzurro

Golden Member
Jul 26, 2016
1,052
1,716
136
Mine was a speculation based on the projected performance of the Xbox Series S - of course it could be entirely wrong but IIRC rumors of a 10-13% better IPC for RDNA2 were floating around recently.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
Mine was a speculation based on the projected performance of the Xbox Series S - of course it could be entirely wrong but IIRC rumors of a 10-13% better IPC for RDNA2 were floating around recently.
"I'll bite" as in "I'm gonna chime in" with my uneducated speculation ;).

P.S. I believe that you MIGHT be correct, IF the rumors of new cache, and patent materialize in the new architecture, and it really might have 20% IPC uplift over previous generation, after all.
 

DisEnchantment

Golden Member
Mar 3, 2017
1,747
6,598
136
VGH use RDNA2.

C:
+++ b/src/amd/addrlib/src/core/addrlib.cpp
@@ -226,6 +226,9 @@ ADDR_E_RETURNCODE Lib::Create(
                     case FAMILY_NV:
                         pLib = Gfx10HwlInit(&client);
                         break;
+                    case FAMILY_VGH:
+                        pLib = Gfx10HwlInit(&client);

C:
@@ -177,6 +177,7 @@ const char *ac_get_llvm_processor_name(enum radeon_family family)
    case CHIP_SIENNA_CICHLID:
    case CHIP_NAVY_FLOUNDER:
    case CHIP_DIMGREY_CAVEFISH:
+   case CHIP_VANGOGH:
       return "gfx1030";
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
AMD already does this for the 5600 XT and 5700 XT.
5700 XT was priced apropriately - on the same level as slower RTX 2060 Super, and 100$ less that roughly the same RTX 2070. 5600 XT was priced at 280$, 20$ less than RTX 2060, from which RX 5600 XT was faster.

So no, AMD won't price 40 CU die, with performance of RTX 3070 at the price of RTX 3060, or less.

AMD has smaller RDNA2 die, in the pipeline that will fit in 199$ price tag, and 6500 SKUs.