Question Speculation: RDNA2 + CDNA Architectures thread

Page 99 - 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
 

lixlax

Member
Nov 6, 2014
187
162
116
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??
 

kurosaki

Senior member
Feb 7, 2019
258
250
86
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??
3! My first thought screams 3! 😁
 

deathBOB

Senior member
Dec 2, 2007
569
239
116
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??

They already listed the price of the 3070.

It's obviously 4, both AMD and Nvidia will sell every new GPU they make between now and the holidays thanks to covid.
 

blckgrffn

Diamond Member
May 1, 2003
9,301
3,442
136
www.teamjuchems.com
They already listed the price of the 3070.

It's obviously 4, both AMD and Nvidia will sell every new GPU they make between now and the holidays thanks to covid.

Pretty much this for 2020 sales. I feel like AMD as the relatively weaker brand has more to worry about with a weak launch than nvidia. There are so many who will only consider nvidia solutions regardless of their competitiveness. They need to simply see enough value against the previous generation/two generations in benchmarks and you can leave AMD right off the graphs.

If AMD really delivers a solid right hook AND a body blow the only way it will really matter is if they can just flood the shelves with stock (ideally from the top to the bottom of the stack) and nvidia runs out. I would imagine that most people buy and sit on their GPUs for 1-3 years at a time and are most likely to stick with the brand they previously purchased. That's just "typical" customer behavior.

That could be something that actually changes the pie chart.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,340
5,464
136
It's a show, drowning us in (conflicting) details and commentary to keep the audience watching and increase credibility. He did the same with Ampere in the Nvidia Forced Scarcity video.
  • He claimed Nvidia is trying to increase margins by forcing AIBs to cut down on BOM and also increase pricing over MSRP. Except we now know cards such as the Asus TUF Gaming took no shortcuts when it comes to BOM and will sell at MSRP.
  • He claimed FE cards would be loss leaders with the purpose of getting better results in reviews over AIB cards, except AIB cards have proven to be faster and cooler.
Again, I won't fault him for claiming Samsung 8nm yields were poor and then reconsidering them as good. That's understandable for someone working with leaks and rumors. The problem is adding false juicy information on top to increase... conversion. That's the Mars fertilizer.

MLID is nonsense to a ridiculous level, and he keeps getting proved wrong, and yet a bunch of people keeping treating it like something reliable.

I remember people getting proactively angered with NVidia, because MLID said that DLSS 3.0 in Ampere would just work with any game that has TAA and that NVidia was considering defaulting it to on, so they would score higher in benchmarks.

So people watched the video, then got mad because NVidia was going to cheat by turning on DLSS 3.0 by default. :rolleyes:

This was all such obvious nonsense, from nearly every angle. He also claimed Ampere RT performance was 4 Times higher than Turing, possibly in the same video. Some people were angry about this to, since NVidia was wasting too much effort on RT performance.

So Ampere comes out, all the stuff is shown to be total nonsense, and people just go back to lapping up his new nonsense without a thought, about how inaccurate he has been in the past.
 

blckgrffn

Diamond Member
May 1, 2003
9,301
3,442
136
www.teamjuchems.com
MLID is nonsense to a ridiculous level, and he keeps getting proved wrong, and yet a bunch of people keeping treating it like something reliable.

I remember people getting proactively angered with NVidia, because MLID said that DLSS 3.0 in Ampere would just work with any game that has TAA and that NVidia was considering defaulting it to on, so they would score higher in benchmarks.

So people watched the video, then got mad because NVidia was going to cheat by turning on DLSS 3.0 by default. :rolleyes:

This was all such obvious nonsense, from nearly every angle. He also claimed Ampere RT performance was 4 Times higher than Turing, possibly in the same video. Some people were angry about this to, since NVidia was wasting too much effort on RT performance.

So Ampere comes out, all the stuff is shown to be total nonsense, and people just go back to lapping up his new nonsense without a thought, about how inaccurate he has been in the past.

His bank account thanks them all. Seriously, this has to be way easier than many of the other jobs available.
 

Vixis Rei

Junior Member
Jul 4, 2020
13
16
41
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??

I feel like it's probably a few things at once.
They probably feel like since RDNA2 launches so late they can probably sell more 3080's / 3090's before AMD meteor showers on their parade.

They may have already listed the price of the 3070 but releasing after the AMD showcase allows them to adjust prices if needed and not be so caught out of position. stamping any silly amd tricks.

Could also be building up stock of 3070's to have so the launch doesn't seem so anemic for a potentially higher volume selling card. cause i don't doubt that AMD and Nvidia will see through all of them.

So maybe mostly #3? maybe.... I dunno if i was an evil cooperation i'd be thinking all this...
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136

stebler

Member
Sep 10, 2020
25
75
61
Green Sardine is very likely Lucienne because Cezanne has newer ip blocks like VCN 2.2 and GC 9.3.

Code:
ACP       #0  3.1.0 
ATHUB     #0  1.5.0 
CLKA      #0 10.0.2 
CLKA      #1 10.0.2 
CLKA      #2 10.0.2 
CLKB      #0 10.0.2 
DBGU_IO   #0  3.0.0 
DBGU_NBIO #0  3.0.0 
DF        #0  3.5.2 
DIO       #0 127.127.63 
DMU       #0  2.1.0 
FCH       #0  7.0.0 
FUSE      #0 10.1.1 
GC        #0  9.3.0 
DAZ       #0 127.127.63 
HDP       #0  4.1.2 
IOAGR     #0  6.0.0 
IOAPIC    #0  6.0.0 
IOHC      #0  6.0.0 
ISP       #0  2.1.0 
MMHUB     #0  1.5.0 
MP0       #0 12.0.1 
MP1       #0 12.0.1 
MP2       #0 11.0.0 
NBIF      #0  2.5.0 
OSSSYS    #0  4.3.0 
PCIE      #0  5.1.0 
PCS       #0  5.1.0 
PCS       #1  5.1.0 
PCS       #2  5.1.0 
PCS       #3  5.1.0 
SDMA0     #0  4.1.2 
SMUIO     #0 10.0.2 
SST       #0  1.1.0 
SST       #1  1.1.0 
SYSTEMHUB #0  2.5.0 
THM       #0 10.1.1 
UMC       #0  7.6.0 
UMC       #1  7.6.0 
USB       #0  4.6.0 
UVD       #0  2.2.0
 

Midwayman

Diamond Member
Jan 28, 2000
5,723
325
126
  • He claimed FE cards would be loss leaders with the purpose of getting better results in reviews over AIB cards, except AIB cards have proven to be faster and cooler.

Those aren't exactly mutually exclusive statements. It can still be horribly expensive to build. Like most blower style coolers, it seems most at home in tight cases without a lot of airflow. Reviewers OTOH are often using open air benches or enormous top end cases. JayzTwoCents did some testing on temp and found the FE actually making the CPU cooler, which is really weird.
 

Konan

Senior member
Jul 28, 2017
360
291
106
Thoughts?
It is possible that a large 128MB cache could occupy 120-150mm2 of space (based on Zen2 estimate slices) .... so if N21 is 505mm2 that leaves a pretty small area to pack 80CU, no?
 

uzzi38

Platinum Member
Oct 16, 2019
2,705
6,427
146
Green Sardine is very likely Lucienne because Cezanne has newer ip blocks like VCN 2.2 and GC 9.3.

Code:
ACP #0 3.1.0 
ATHUB #0 1.5.0 
CLKA #0 10.0.2 
CLKA #1 10.0.2 
CLKA #2 10.0.2 
CLKB #0 10.0.2 
DBGU_IO #0 3.0.0 
DBGU_NBIO #0 3.0.0 
DF #0 3.5.2 
DIO #0 127.127.63 
DMU #0 2.1.0 
FCH #0 7.0.0 
FUSE #0 10.1.1 
GC #0 9.3.0 
DAZ #0 127.127.63 
HDP #0 4.1.2 
IOAGR #0 6.0.0 
IOAPIC #0 6.0.0 
IOHC #0 6.0.0 
ISP #0 2.1.0 
MMHUB #0 1.5.0 
MP0 #0 12.0.1 
MP1 #0 12.0.1 
MP2 #0 11.0.0 
NBIF #0 2.5.0 
OSSSYS #0 4.3.0 
PCIE #0 5.1.0 
PCS #0 5.1.0 
PCS #1 5.1.0 
PCS #2 5.1.0 
PCS #3 5.1.0 
SDMA0 #0 4.1.2 
SMUIO #0 10.0.2 
SST #0 1.1.0 
SST #1 1.1.0 
SYSTEMHUB #0 2.5.0 
THM #0 10.1.1 
UMC #0 7.6.0 
UMC #1 7.6.0 
USB #0 4.6.0 
UVD #0 2.2.0
Thought so, but thanks for the confirmation dude!
 

Dribble

Platinum Member
Aug 9, 2005
2,076
611
136
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??
5) Marketing 101 the last big news is what stays in the news.
 
  • Like
Reactions: Tlh97 and Krteq

sandorski

No Lifer
Oct 10, 1999
70,240
5,810
126
5) Marketing 101 the last big news is what stays in the news.

I think that is basically option 2, which seems very likely. Maybe NVidia has my Clock wiggle room for the 3070 specifically that it can use depending on AMDs competing part.

There is also the possibility that it identified a design flaw in the 3080 Crashing issue which it wanted fixed prior to the release of the 3070.
 

Tup3x

Golden Member
Dec 31, 2016
1,086
1,085
136
I think that is basically option 2, which seems very likely. Maybe NVidia has my Clock wiggle room for the 3070 specifically that it can use depending on AMDs competing part.

There is also the possibility that it identified a design flaw in the 3080 Crashing issue which it wanted fixed prior to the release of the 3070.
Or they are confident in its performance and expect to sell a ton even after RDNA2 announcement. If they knew it sucked in comparison it would make sense to launch it before. That way people would at least buy them. In any case, it's possible to interpret this in so many different ways...
 

dzoni2k2

Member
Sep 30, 2009
153
198
116
What's your take on the RTX 3070 delay regarding RDNA2?

1) Nvidia knows RDNA2 is not impressive and they are going to steal all the thunder after AMD presentation?
2) They know 3070 isn't very competitive vs RDNA2 and Nvidia wants to avoid comparisons on AMD slides?
3) They have no idea what the GPUs from AMD are like and how they perform (so they can cut the price if needed)?
4) It has nothing to do with AMD and they're just improving drivers and stock?
5) ??

They could delay it by a month and it wouldn't make a difference because it's a paper launch with 0 stock either way.
 

dzoni2k2

Member
Sep 30, 2009
153
198
116
Or they are confident in its performance and expect to sell a ton even after RDNA2 announcement. If they knew it sucked in comparison it would make sense to launch it before. That way people would at least buy them. In any case, it's possible to interpret this in so many different ways...

You don't move around launch dates like a headless chicken if you're confident in your product.
 
  • Like
Reactions: Tlh97 and Mopetar