Question Speculation: RDNA2 + CDNA Architectures thread

Page 133 - 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
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
There haven't been many GPUs missing a 32-bit part of the memory bus [1]. This means that it has either a misbalanced memory configuration (some memory channels having more memory than others) or a cut-down 160-bit memory bus.

However, if we assume that Navi 22 has 384-bit memory bus then the answer is straight-forward - the cut down version has one double-memory controller less, 320-bit in total. This would also mean that every other rumored bus size, like N21 being 256-bit, should also be multiplied by 2. In my opinion that is a more likely explanation than the "infinity cache" that would magically cut the memory bandwidth needs of a huge GPU in half (the cache might still exist but have a different purpose, e.g. to speed up unpredictable memory access patterns of ray tracing BVH traversal).

Please note though that this is just my assumption that might be completely wrong.

[1] Technically, single GDDR6 channel is 32-bit. Looking at the wiki page, it seems there was at the very least a cut-down GeForce 1050 with 96-bit bus, and 1080Ti/2080Ti use 352-bit bus. For AMD, there is a single cut-down Navi-14. Not many in total.

We know the specs for Navi GPUs.

N21 has 256 bit GDDR6, N22 has 192 bit GDDR6, N23 has 128 bit GDDR6 memory controller.

You can easily get 160 bit bus from N22 cut down version.
 

Konan

Senior member
Jul 28, 2017
360
291
106
So the leaks are suggesting N21= 80CU; N22=40CU; N23=32CU and N24=24CU? Doesn't make sense to me- Navi 22-24 all close together and then huge cap to Navi 21.

Also the leaked clockspeeds are all over the place- anywhere from 2GHz-2.5GHZ.

It could come out something similar to fill the gaps -

Navi 21a = XT = 80CU = boost clock 2050Mhz
Navi 21b = XTL = 72 CU = boost clock 2200Mhz
Navi 21 = XL = 64CU

There should be 2 to 3 SKUs cut from Navi 21.
It was speculated before regarding there was an XTX variant the "secret one" that only AMD has - and that could still come out but "KittyCorgi" (Yuko) above is not saying so today.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
It could come out something similar to fill the gaps -

Navi 21a = XT = 80CU = boost clock 2050Mhz
Navi 21b = XTL = 72 CU = boost clock 2200Mhz
Navi 21 = XL = 64CU

There should be 2 to 3 SKUs cut from Navi 21.
It was speculated before regarding there was an XTX variant the "secret one" that only AMD has - and that could still come out but "KittyCorgi" (Yuko) above is not saying so today.
Navi 22 has XTL.

There is no XTL for Navi 21 GPUs.

Speaking about KittyCorgi, I think at this point in time its pretty clear that his sources have access not to the Big Navi but to Navi 22 GPUs, and if we bare this in mind, everything what he posted makes sense, now.
 

Konan

Senior member
Jul 28, 2017
360
291
106
Navi 22 has XTL.

There is no XTL for Navi 21 GPUs.

Speaking about KittyCorgi, I think at this point in time its pretty clear that his sources have access not to the Big Navi but to Navi 22 GPUs, and if we bare this in mind, everything what he posted makes sense, now.
Yep that could be very true too.
 

lightmanek

Senior member
Feb 19, 2017
429
916
136
Resurrect the ability to turn an XL into an XT!

...thinks fondly of the x1800 XL he bought back in the day that was tricked into being an XT...

XT PE is even better. More letters!
Or My 9500 Non-Pro into 9700 Pro, these were the times! (Yes, it would be nice if something akin to a mythical R300 came out of ATi's labs ... (errr or is it AMD now?)
 
  • Love
Reactions: blckgrffn

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
Digital Foundry just uploaded a video highlighting Xbox Series X power and thermals.

Few quick takeaways are Rise of the Tomb Raider peaked at 151W, Dead or Alive 6 at 165W, and Gears 5 at 211W. Richard even made a nod to Big Navi at the end of the video, which I thought was amusing.
Why is Digital Foundry Hyping up RDNA2 architecture?

Its going to be crap, after all.
 
  • Like
Reactions: Tlh97 and raghu78

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
Ah yes, I remember those days; 9500 unlocked to 9700, DLT3C stepping TBred 1700+ running 2.4+ Ghz, and some BH5 DDR running at 3.5v.

Or unlocking Phenom II 550BE dual core to a 965BE quad core.

But these types of things have not been the case for a while. Cut down chips have physical blocks in the disabled cores to prevent us from doing such thing.
 
  • Like
Reactions: lightmanek

Mopetar

Diamond Member
Jan 31, 2011
8,114
6,770
136
I sh* you not. That's what some Nvidia fans on Twitter are saying and believing.

Why would anyone use what random idiots on social media think for any kind of baseline for expectations? If they aren't regurgitating flat earth nonsense into each other's face holes, they're rambling on about vaccines or any number of other topics on which they're equally ill informed. Your average person on social media is scarcely more intelligent than a baboon and smells twice as bad. Social media is proof that the SETI program should probably direct its sensors down at earth before they worry about any kind of extra terrestrial intelligence.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
In the distant future the extra terrestrials will visit Earth and find that Social media was the devise of mankind.

9 days till the hype train hits dead mans curve. Will it survive?

Has one ever survived? Even releases that were very strong still had hype that was above them. Zen 1 was a strong release, but even at that it did not meet hype. Zen 3 might equal the hype, we will find out in a few weeks.

The original Core architecture was worthy of the hype.

On the GPU side of things though, having trouble finding one. Pascal was a really strong release, and minus initial prices, was well received. But even then not sure it equaled the hype.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
7,166
7,666
136
So the leaks are suggesting N21= 80CU; N22=40CU; N23=32CU and N24=24CU? Doesn't make sense to me- Navi 22-24 all close together and then huge cap to Navi 21.

Also the leaked clockspeeds are all over the place- anywhere from 2GHz-2.5GHZ.

- This has likely been covered somewhere in the 130+ pages of this thread, but I'm curious as well of the logic of having a 80CU chip, then 40/32/24 CU chips.

I understand the huge gap between the 80 and 40, plenty of die harvesting possible to fill that gap, but I would have figured we'd get a 40 CU Chip and a 20CU chip mirroring RDNA 1 (naturally as chips get smaller, you have to artificially segment rather than naturally harvesting dies).

Its been a helluva thread, can anyone refresh the argument for this?
 
  • Like
Reactions: Tlh97 and Elfear

PhoBoChai

Member
Oct 10, 2017
119
389
106
Digital Foundry just uploaded a video highlighting Xbox Series X power and thermals.

Few quick takeaways are Rise of the Tomb Raider peaked at 151W, Dead or Alive 6 at 165W, and Gears 5 at 211W. Richard even made a nod to Big Navi at the end of the video, which I thought was amusing.

DF claims Gears 5 now runs RNDA2 optimized code with an update. No more rushed port.

With ray traced GI and contact hardening shadows. Native 4K60. Ultra settings.

All that from 52 CU... Hype much?