Question Speculation: RDNA2 + CDNA Architectures thread

Page 165 - 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
2,698
6,393
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
 

soresu

Platinum Member
Dec 19, 2014
2,933
2,156
136
Hmm, there's some more info about the Infinity Cache on AMD's RDNA 2 webpage in the footnotes:


This is in reference to AMD's claim of the following:
- 256-bit 16 Gpbs GDDR6 = 512 GB/s
- 384-bit 19.5 Gpbs GDDR6x = 936 GB/s
- 256-bit 16 Gpbs GDDR6 + 128MB Infinity Cache = 1664 GB/s Effective

A shot directly across the bow of Ampere.
Considering how reliant RT is on memory I wonder how well this would play out at equal TFLOPS?
 

Martimus

Diamond Member
Apr 24, 2007
4,488
153
106
Will also be interesting to see 6800 XT vs 3080 on an non-Zen3 CPU. The value of the 6800 XT might not look as good if you have to spend another $300+ on a CPU and maybe even a motherboard for it to match/beat the 3080.
SAM wasn't enabled on their testing with the 3080 and 6800XT. They only enabled it on the 3090 to 6900XT chart.
 

sze5003

Lifer
Aug 18, 2012
14,184
626
126
For the 6900xt I assume it's full potential will be better if you have an amd card and board right?

Although for pure gaming use I'm not sure if that would matter much.
 

soresu

Platinum Member
Dec 19, 2014
2,933
2,156
136
So do we know what this 128MB infinity cache is? If I am not mistaken, they said something in the lines of 256bit+infinity cache=Twice the performance of standard 384bit bus. How is this possible? wtf?
The presenter/engineer mentioned that it was modelled on Zen3's L3 cache.
Yeah, I think AMD will shine in DXR 1.1 titles. Performance will be inferior to Ampere's in DXR 1.0 titles.
Makes me wonder if CP 2077 delay has DXR 1.1 zero day goodies in store.
 

Greyguy1948

Member
Nov 29, 2008
156
16
91
So do we know what this 128MB infinity cache is? If I am not mistaken, they said something in the lines of 256bit+infinity cache=Twice the performance of standard 384bit bus. How is this possible? wtf?

Also do the consoles have it? Will it be present on lower tier cards?

It looks like eDRAM cache. Intel and IBM used/are using it. Transistor count is "only" 28 billion so any transistor cache is smaller.
EDRAM on Wiki
 
Last edited:

moinmoin

Diamond Member
Jun 1, 2017
4,993
7,763
136
Looks like there's another video from AMD. My apologies if this has been posted already.


Some explanation on AMD Smart Access Memory, Anti-Lag, Radeon Boost, etc.
First time I saw it posted, good video.

Infinity Control Fabric sounds like the Scalable Control Fabric is being expanded outside the device (I imagine that's also part of Infinity Architecture), allowing different devices to know of each other's state and sensors. This may explain why SAM requires Ryzen 5000. This also raises the possibility that RDNA2 reuses the Infinity Fabric hierarchy already known from the Zen CPUs (with Scalable Control Fabric monitoring and controlling everything while Scalable Data Fabric connects all the parts for I/O), with Infinity Cache working the same way as L3 cache in Zen CPU, that is as essentially a crossbar between all cores (= CUs in RDNA2's case).
 

MrTeal

Diamond Member
Dec 7, 2003
3,584
1,743
136
Looks like 128MB Infinity Cache is true afterall and looks like you dont need to specially program for it.
Also this 'big Navi' has 128 ROP not 64 as rumor suggested.
Hopefully those cards will not have the fate of R9 290X.

Another note: Why no 720p benchmark AMD?? That is the most popular resolution that most people use.
Maybe in 2003. The only people playing at 720 are people trying to play shooters at 350FPS and people stuck with 768p laptop screens.
 

Hitman928

Diamond Member
Apr 15, 2012
5,567
8,717
136
Oh... using Rage mode voids the warranty

View attachment 32529

I wouldn't worry about it. As long as you're giving the card decent ambient air, there shouldn't be any issues. Pretty sure this statement is there to cover people using Wattman to crank voltages to crazy levels. Also not sure rage mode even counts as overclocking but maybe AMD can clarify.
 

Vope45

Member
Oct 4, 2020
114
168
86
Looks like 128MB Infinity Cache is true afterall and looks like you dont need to specially program for it.
Also this 'big Navi' has 128 ROP not 64 as rumor suggested.
Hopefully those cards will not have the fate of R9 290X.

Another note: Why no 720p benchmark AMD?? That is the most popular resolution that most people use.

Is this a joke ?
 

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
Heh, all those shades...

+1 for edge connectors. WIth these larhe cards having connectors at back would just be annoying.

To something completely different: Anyone else thought that die shot looks like you could just "cut-it" in the middle and have 2 40 CU dies, erm, chiplets?

With all the trade-off the cache has, it's probably needed for chiplet gpus and solves (or at least softens) the iGPU memory bandwidth issue at the same time. While it's a trade-off got these high end GPUs, the tech itself make a ton of sense in this bigger picture.

Else pricing as expected, 6800 actually a bit high. Supply will probably be very limited else the 6800 at $500-530 would have made a bit more sense. Thats probably the downside of the small bus, can't cut it even on the 6800 and hence need the full 16GB vram.

Anyway, these are all out of my league anyway.
 
Reactions: lightmanek

n0x1ous

Platinum Member
Sep 9, 2010
2,572
248
106
I don't like the naming. XT has always meant full die for Radeon in the past.

Would have preferred 6900 XT, 6900 PRO, 6900 (9 signifying Navi 21 biggest die)

then maybe a 6900 XTX with AIO cooler
 
Reactions: Elfear

Saylick

Diamond Member
Sep 10, 2012
3,372
7,104
136
To something completely different: Anyone else thought that die shot looks like you could just "cut-it" in the middle and have 2 40 CU dies, erm, chiplets?
It makes the most sense in my mind to have each Shader Engine be its own chiplet when AMD moves RDNA to a chiplet architecture. The Shader Engine has all of the complete parts to run the graphics pipeline so its the most basic building block that gets duplicated. The memory controllers, last-level cache (Infinity Cache), display connectors, codec engines, etc can be on an IO die. This approach pretty much mirrors how it's done on Zen 2/3.

 

randomhero

Member
Apr 28, 2020
184
251
136
Very good showing AMD! I would count this as their zen2 moment in graphics.
And inf cache! I knew right from the moment RGT leaked that(their source is definitely AMD) it was true. It was so way out there, typical AMD!
 
Reactions: lightmanek
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |