Question Speculation: RDNA2 + CDNA Architectures thread

Page 132 - 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
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,249
136
Im just a passenger on this train now, and I do not know where it is heading.

I'll update you then....Maybe you're dazed and confused? Rumor is you fell asleep listening to Zeppelin.

Nvidia's supply train is stalled on the tracks ahead, couldn't make the grade....Probably didn't allow for the excessive weight of the FE coolers?

AMD's hype train was forced to take an alternative route, grade is much greater, but the momentum is with them. Once they hit the peak it's all downhill from there. According to the latest calculations the train should be hitting dead mans curve around the 28th. The real question is if the brakes will hold (if needed) or if they'll just go balls to the wall and pray the train stays on the tracks!

I'm looking forward to the announcement to see what gains AMD has managed to achieve.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
7,167
7,666
136
David Attenborough's Voice:

We turn our attention now to the RDNA2 thread, where a scarcity of reliable information has manefested itself as a hype train. The hive mind is not helpless however, and has a rather... unsavory... solution to the problem as anti-hype posters begin consuming their breathern to keep the hype down to manageable levels...
 

soresu

Diamond Member
Dec 19, 2014
3,230
2,515
136
David Attenborough's Voice:

We turn our attention now to the RDNA2 thread, where a scarcity of reliable information has manefested itself as a hype train. The hive mind is not helpless however, and has a rather... unsavory... solution to the problem as anti-hype posters begin consuming their breathern to keep the hype down to manageable levels...
I was literally just watching Frozen Planet and spluttered my tea over the keyboard reading that!
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
This just tells me nobody outside of specific people at AMD and AiBs know a damn thing.

Edit to add: this reminds me of Eyefinity where AMD gave partners different code names for the project to see who leaked info early.

It seems entirely possible AMD have sent different AIBs different bioses with different clocks so they can see which AIBs are leaky and which run a tight ship.

It's not like this anymore (different bioses). Only first partner AIBs for AMD gets samples early, particularly Sapphire as they are the ones who have been making reference AMD designs. Sapphire is tight, solid for AMD.

It's only when they plan more custom variants, approaching guys like MSI, ASUS, Gigabyte, Powercolor, XFX, that leaks start to happen. If folks follow the tech scene long enough, they'll know the leakiest are MSI & Gigabyte. XFX & Powercolor has AFAIK, never leaked ES or pre-release samples from AMD.
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
So... What are the odds that after Nvidia released the power guzzling 3000 series, AMD decided to clock the nuts off Navi 2 and blow up the power consumption? If 2.4GHz is true, then they surely use more than 250W.

Total Graphics Power is not TBP or TDP traditionally.

It's just the GPU die itself, minus the PCB + VRAM. Add another 50-60W, easily.

There were earlier rumors, that RDNA2 can clock very high if you allow it more power. David Wang himself seems quite confident when he stated its multi gigahertz capable.

Only problem is do we want reference cards coming out with 350W and on a cooler designed to be quiet at 250-300W max? I'd much prefer a cool & quiet ref card, and then AIB custom models come with +50W OC SKUs.
 

Geranium

Member
Apr 22, 2020
83
101
61
TGP = Nvidia Marketing term.
TBP = AMD marketing term.
No, they are not. TBP = Total/Typical Board Power.
TBP for AMD is how much power that graphics card draws on average. Radeon VII has TBP of 300W and it never draws more than that on average. TBP is GPU Power+Memory Power+Fan Power+Power Stage Power+Everything else.
Average GPU Power Consumption [2020 October 19].png

Here is AMD's Specs Page for Radeon VII : https://www.amd.com/en/products/graphics/amd-radeon-vii
And same can be also true for Nvidia's TGP.
 
  • Like
Reactions: Tlh97

PhoBoChai

Member
Oct 10, 2017
119
389
106
No, they are not. TBP = Total/Typical Board Power.
TBP for AMD is how much power that graphics card draws on average. Radeon VII has TBP of 300W and it never draws more than that on average. TBP is GPU Power+Memory Power+Fan Power+Power Stage Power+Everything else.

This is how it was in the past. Total Board Power is the whole card. Then we started getting silly names for only the graphics package, ie, the GPU die. Now ppl want to change it to TGP = die + vram only.

It's ridiculous.
 

Konan

Senior member
Jul 28, 2017
360
291
106

Leeea

Diamond Member
Apr 3, 2020
3,795
5,549
136
Speculation mode on!:

I have been an AMD fan for a long time. I am used to disappointment.

Will be lucky if it keeps up with a 3080. Hopefully driver issues are not to awful.

But, could be a sweet deal in about a year or so after the issues are worked out. Smaller die size + lower power draw makes a cheaper card to make and cool. May allow AMD to under cut nvidia on $ and put it within reach of us mortals.

If it beats the 3080 that is great for us, but Nvidia will be fine. AMD 5600 series driver issues are still haunting it, it will take a full release cycle before people start trusting AMD.

Which hopefully means I will be able to afford one! :)
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,038
136
As mentioned at beyond3d by Jawed, N23(236mm2) is 50% bigger than Navi14(157mm2). It doesn't make sense, that It should be only 32CU GPU with 128bit GDDR6 when Navi 14 has 24CU and 128bit GDDR6. The difference in the supposed specs is only 8CU and 1 RDNA1 CU is ~2.1mm2, so the difference should be ~17mm2 instead of 79mm2, yet It's not.
 

Saylick

Diamond Member
Sep 10, 2012
3,532
7,859
136
As mentioned at beyond3d by Jawed, N23(236mm2) is 50% bigger than Navi14(157mm2). It doesn't make sense, that It should be only 32CU GPU with 128bit GDDR6 when Navi 14 has 24CU and 128bit GDDR6. The difference in the supposed specs is only 8CU and 1 RDNA1 CU is ~2.1mm2, so the difference should be ~17mm2 instead of 79mm2, yet It's not.
Infinity Cache?
 

Karnak

Senior member
Jan 5, 2017
399
767
136
Yuko is wrong according to ExecutableFix (was the one leaking Cezanne/Lucienne earlier). Besides that doesn't make any sense to me either what Yuko is saying.
 
  • Like
Reactions: Glo.

lixlax

Member
Nov 6, 2014
187
162
116
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.