Question Speculation: RDNA2 + CDNA Architectures thread

Page 129 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
uzzi38

uzzi38

Platinum Member
Oct 16, 2019
2,621
5,872
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
 
  • Like
Reactions: MagnusTheBrewer, lightmanek, Tlh97 and 2 others
Sort by date Sort by votes
R

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Glo. said:
Yes, guys. Kepler_L2 is correct.

Historically XT and XTX are full dies. They just differ by clock speeds, and power targets.

So its not 72 CUs. Its full, 80 CUs, with 2.4 GHz game clock, at 255W of power drawn(hence the TGP).

Im lost. Its GG. Its beyond EVERYTHING we thought it could possibly be.
Click to expand...

https://twitter.com/x/status/1300438803421618179

80CU vs 82SM 72CU vs 68SM 72CU vs 68SM

What we will see.

So if kopite is right and I tie the info VCZ said about XTX having more CU than XT then its obvious that

RX 6900 XT = Navi 21 XTX = 80/16GB /128 ROPs (fully enabled)
RX 6900 = Navi 21 XT = 72,/16GB/128 ROPs (4 WGP disabled)
RX 6800XT = Navi 21 XL = 72/16GB/112 ROPs (1 SA fully disabled)
RX 6800 = Navi 21 XE = 60/12GB/96 ROPs (1 SE disabled with 64 bit G6 MC disabled)

https://twitter.com/x/status/1317588927193165825
Navi21 XT > Base clock 2000MHz to 2100MHz > Boost clock around 2400MHz

https://twitter.com/T4CFantasy/status/1316618720148676608
2160 / 2410
 
Last edited:
  • Like
  • Love
Reactions: spursindonesia, Tlh97, SinaptiX and 1 other person
Upvote 0 Downvote
A

andermans

Member
Sep 11, 2020
151
153
76
Wouldn't the game clock be pretty much equal the boost clock if you give it enough cooling & power? I've looked at the clock speed graphs of some Rx 5700 XT AIB reviews on TPU and most stay pretty damn near the advertised boost speed.
 
Upvote 0 Downvote
Glo.

Glo.

Diamond Member
Apr 25, 2015
5,704
4,548
136
raghu78 said:
https://twitter.com/x/status/1300438803421618179

80CU vs 82SM 72CU vs 68SM 72CU vs 68SM

What we will see.

So if kopite is right and I tie the info VCZ said about XTX having more CU than XT then its obvious that

RX 6900 XT = Navi 21 XTX = 80/16GB /128 ROPs (fully enabled)
RX 6900 = Navi 21 XT = 72,/16GB/128 ROPs (4 WGP disabled)
RX 6800XT = Navi 21 XL = 72/16GB/112 ROPs (1 SA fully disabled)
RX 6800 = Navi 21 XE = 60/12GB/96 ROPs (1 SE disabled with 64 bit G6 MC disabled)

https://twitter.com/x/status/1317588927193165825
Navi21 XT > Base clock 2000MHz to 2100MHz > Boost clock around 2400MHz

https://twitter.com/T4CFantasy/status/1316618720148676608
2160 / 2410
Click to expand...
Or...

80 CU/High clocks - XTX, 80CU/Low clocks - XT, 72/high clock - XL, 72/low clock - XE.

And Navi 24 is actually 64 CU die.

Because that way it also makes absolute sense.
 
  • Like
Reactions: Tlh97 and misuspita
Upvote 0 Downvote
R

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Glo. said:
Or...

80 CU/High clocks - XTX, 80CU/Low clocks - XT, 72/high clock - XL, 72/low clock - XE.

And Navi 24 is actually 64 CU die.

Because that way it also makes absolute sense.
Click to expand...

N24 is not 64 CU. N24 is a small die with 16-20 CU and < 150 sq mm. You have to remember N23 according to Acquariuszi is 240 sq mm.
 
Upvote 0 Downvote
T

Timorous

Golden Member
Oct 27, 2008
1,605
2,746
136
80 CUs @ 2.4Ghz game clocks.

With 80% CU scaling, 50% clock scaling and 0% IPC increase vs 5700xt you get, 1.8 × 1.125 = 2.025x 5700XT. This is 3080 performance.

This is conservative.

72CUs with 80% CU scaling, 50% clock scaling and 10% IPC gets 1.64 × 1.125 × 1.1 = 2.0295x 5700XT.

Even with conservative CU scaling and clock scaling numbers a 72 CU card with a 10% IPC gain will be trading blows with the 3080.

If clock scaling and or CU scaling is better than these numbers the top 80CU card should easily beat the 3090.
 
  • Like
  • Love
  • Wow
Reactions: spursindonesia, Tlh97, moinmoin and 1 other person
Upvote 0 Downvote
Glo.

Glo.

Diamond Member
Apr 25, 2015
5,704
4,548
136
Vope45 said:
Thats a really high clock but what important is performance. If its on par with 3080 I'd call that a victory for AMD.
Click to expand...
Which SKU in particular you are talking about, being on par with RTX 3080?

Because maybe you missed, but there is no way in hell top SKU is only on par with 3080. No way, in hell.
 
  • Love
Reactions: spursindonesia
Upvote 0 Downvote
T

Timorous

Golden Member
Oct 27, 2008
1,605
2,746
136
Glo. said:
Which SKU in particular you are talking about, being on par with RTX 3080?

Because maybe you missed, but there is no way in hell top SKU is only on par with 3080. No way, in hell.
Click to expand...

Agreed. If these clocks hold in games then it is looking 3090ish minimum.
 
Upvote 0 Downvote
V

Vope45

Member
Oct 4, 2020
114
168
86
Glo. said:
Which SKU in particular you are talking about, being on par with RTX 3080?

Because maybe you missed, but there is no way in hell top SKU is only on par with 3080. No way, in hell.
Click to expand...

The Xt variants, XTX at this point will most likely be a higher clock version of the XT, I really hope not though.
 
Upvote 0 Downvote
L

linkgoron

Platinum Member
Mar 9, 2005
2,293
814
136
If what I've seen is correct, AMD have been using the XT moniker on GPU code-names for years to signify the fully enabled chip (found one example in the last ten years where it wasn't). Why would they change it now? XTX has been used a few times recently (Navi 10, 14, and less recently in 2013...) to signify a more highly-clocked fully-enabled chip. They might change it, but IMO it's unlikely.

What we've seen from AMD has been a card slower by around 10% than the 3080, and people here are talking about beating the 3090 by 10-15%. I'll be happy if AMD can produce a 5% winner vs the 3080. People here are leading themselves to disappointment. People should go re-read the Vega thread. The same user, using the same tactics, overhyping yet another AMD launch.

Vega/Navi Rumors (Updated)

Vega 10 - 64 CUs / 4096 SPs - 24 TFLOPs 16-bit (VideoCardz) / 10 TFLOPs single-precision (Techfrag) - Not as huge (6000 SPs+) as some expected, but will have a 300W dual GPU version to go up against NVIDIA's best next year Vega 11 - Smaller chip, could replace some Polaris 10 products Vega 20...
forums.anandtech.com forums.anandtech.com


Glo. said:
Which SKU in particular you are talking about, being on par with RTX 3080?

Because maybe you missed, but there is no way in hell top SKU is only on par with 3080. No way, in hell.
Click to expand...

I will leave you with this spoiler to the Vega thread:
1602981002814.png
 
  • Like
  • Haha
  • Wow
Reactions: Cableman, spursindonesia, zinfamous and 9 others
Upvote 0 Downvote
Glo.

Glo.

Diamond Member
Apr 25, 2015
5,704
4,548
136
linkgoron said:
If what I've seen is correct, AMD have been using the XT moniker on GPU code-names for years to signify the fully enabled chip (found one example in the last ten years where it wasn't). Why would they change it now? XTX has been used a few times recently (Navi 10, 14, and less recently in 2013...) to signify a more highly-clocked fully-enabled chip. They might change it, but IMO it's unlikely.

What we've seen from AMD has been a card slower by around 10% than the 3080, and people here are talking about beating the 3090 by 10-15%. I'll be happy if AMD can produce a 5% winner vs the 3080. People here are leading themselves to disappointment. People should go re-read the Vega thread. The same user, using the same tactics, overhyping yet another AMD launch.

Vega/Navi Rumors (Updated)

Vega 10 - 64 CUs / 4096 SPs - 24 TFLOPs 16-bit (VideoCardz) / 10 TFLOPs single-precision (Techfrag) - Not as huge (6000 SPs+) as some expected, but will have a 300W dual GPU version to go up against NVIDIA's best next year Vega 11 - Smaller chip, could replace some Polaris 10 products Vega 20...
forums.anandtech.com forums.anandtech.com




I will leave you with this spoiler to the Vega thread:
View attachment 31918
Click to expand...
The GPU that they have showed might have not been neither XTX, nor XT, but XL, per this:
https://twitter.com/x/status/1317575449325744131
and this:
https://twitter.com/x/status/1317537065127628806

So basically, low clocked version of 72 CU die, with this level of clocks: https://twitter.com/_rogame/status/1317584994156617729

I perfectly understand where you guys coming from with disbelief about Navi 2.

But NOTHING so far points to reality, that this time AMD underperformed in any way shape or form. This is the real deal.

P.S. Shoot the message, not the messenger. Your post will age really badly.
 
  • Love
Reactions: spursindonesia
Upvote 0 Downvote
L

linkgoron

Platinum Member
Mar 9, 2005
2,293
814
136
Glo. said:
The GPU that they have showed might have not been neither XTX, nor XT, but XL, per this:
https://twitter.com/x/status/1317575449325744131
and this:
https://twitter.com/x/status/1317537065127628806

So basically, low clocked version of 72 CU die. https://twitter.com/_rogame/status/1317584994156617729

I perfectly understand where you guys coming from with disbelief about Navi 2.

But NOTHING so far points to reality, that this time AMD underperformed in any way shape or form. This is the real deal.
Click to expand...
1602982113483.png

1602982299093.png
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom