Question Speculation: RDNA2 + CDNA Architectures thread

[uzzi38]

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

 

[eek2121]

I am really starting to wonder If the info about 505mm2 and 80CU is actually correct. For 80CU, 128ROPs and 512bit Rdna1 GPU I got only 461mm2 and that is with the same transistor density.
I personally don't believe density will stay the same so either die size is wrong or the number of CU.

RDNA 2 features a slight rework to include stuff like RT.

There have now been multple trustworthy sources saying that:

Navi 21 - 80CU
Navi 22 - 60CU
Navi 23 - 40CU
Navi 24 - 20CU?

 

[TESKATLIPOKA]

That doesn't necessarily mean they can't be wrong about the actual CU count, unless they have some insider info, which is in my opinion not very likely considering how tight lipped AMD is.
Or the CU count is correct, but die size is wrong.

 

[uzzi38]

I am really starting to wonder If the info about 505mm2 and 80CU is actually correct. For 80CU, 128ROPs and 512bit Rdna1 GPU I got only 461mm2 and that is with the same transistor density.
I personally don't believe density will stay the same so either die size is wrong or the number of CU.

More than one person has corroborated the die size. AquariusZi is not the only person who says ~500mm^2 for Navi21. If he was, we'd wouldn't be so sure of it being this size.

 

[Olikan]

RDNA 2 features a slight rework to include stuff like RT.

XBOXSX CU is actually slight smaller than n10

Or the CU count is correct, but die size is wrong.

Yes, and if AMD uses HBM2, they could fit a full 8-core zen2 there... and would still have some space left

 

[TESKATLIPOKA]

More than one person has corroborated the die size. AquariusZi is not the only person who says ~500mm^2 for Navi21. If he was, we'd wouldn't be so sure of it being this size.

Then is 80CU wrong? Or you think density is the same and the missing 40mm2 is used up for better IPC and rt functionality?

 

[dzoni2k2]

Lower density and higher clock speed. Speed demon baby.

 

[TESKATLIPOKA]

Lower density and higher clock speed. Speed demon baby.

So instead of using high density libraries to increase density by ~25-50% to Renoir levels which can be used for more CUs, ROPs... they use high performance libraries to increase the frequency by ~20%? That's not a very good tradeoff in my opinion.
For example:
RDNA1 with 80CU, 128ROPs and 512bit GDDR6 is 461mm2.
If I increase density by 30% I end up with 120CU, 192ROPs and 512bit GDDR6 and the size would be ~451mm2. If I set 2Ghz for the first GPU then for the second GPU I would need 1.4-1.5Ghz for the same performance and at those clocks I think It would consume less power than the first one.

 

[eek2121]

XBOXSX CU is actually slight smaller than n10


Yes, and if AMD uses HBM2, they could fit a full 8-core zen2 there... and would still have some space left

My understanding is that the GPU portion of the Xbox SoC is not directly comparable to RDNA2. They will be mostly similar, but there are also a number of differences.

 

[Glo.]

So instead of using high density libraries to increase density by ~25-50% to Renoir levels which can be used for more CUs, ROPs... they use high performance libraries to increase the frequency by ~20%? That's not a very good tradeoff in my opinion.
For example:
RDNA1 with 80CU, 128ROPs and 512bit GDDR6 is 461mm2.
If I increase density by 30% I end up with 120CU, 192ROPs and 512bit GDDR6 and the size would be ~451mm2. If I set 2Ghz for the first GPU then for the second GPU I would need 1.4-1.5Ghz for the same performance and at those clocks I think It would consume less power than the first one.

"Large Caches"

 

[Saylick]

Take this with a pinch of salt as always, but here's yet another MLID video about how RDNA 2 has better perf/W against Ampere.

 

[sontin]

This guy is making money with clickbait and fakes. Dont know why anybody believes him after the Ampere fiasco. He was totally wrong on everything which was not leaked by Twitter.

 

[Glo.]

This guy is making money with clickbait and fakes. Dont know why anybody believes him after the Ampere fiasco. He was totally wrong on everything which was not leaked by Twitter.

Considering that he has photos of Quadro 6000 GPU that have not been released previously I would not post things like this, especially considering the amount of things he was correct on in terms of Ampere info .

 

[sontin]

A photo is nothing. We got leaks from the 3080 cooler in june. nVidia has sent Quadros to partners. It will be released in 1 1/2 weeks.

He was wrong with every "non leaked" point about Ampere. Go back and watch his videos again.

Dont trust such clickbait youtubers.

 

[Glo.]

He was wrong with every "non leaked" point about Ampere. Go back and watch his videos again.

Which are...?

Its on you to substantiate this.

 

[sontin]

Last posting, but you can skip through the video and look at the slides:

Both slides have 100% fake "leaks" and nothing has come reality. There is a difference between a leaked roadmap and photos and to have real insider information about architectures.

 

[A///]

That guy is a moron.

Anyone catch Frank Azor's tweet replies this morning? All words or actual real talk? Something that's bugging me. It was assumed AMD may not have a good contender, but Frank's post alluded to AMD stockpiling cards to avoid the NVidia fiasco. Would that imply a hot item is in our midst, lads?

 

[Glo.]

That guy is a moron.

Anyone catch Frank Azor's tweet replies this morning? All words or actual real talk? Something that's bugging me. It was assumed AMD may not have a good contender, but Frank's post alluded to AMD stockpiling cards to avoid the NVidia fiasco. Would that imply a hot item is in our midst, lads?

It was considered that AMD has no contention this time around by who, exactly?

 

[Kenmitch]

That guy is a moron.

Anyone catch Frank Azor's tweet replies this morning? All words or actual real talk? Something that's bugging me. It was assumed AMD may not have a good contender, but Frank's post alluded to AMD stockpiling cards to avoid the NVidia fiasco. Would that imply a hot item is in our midst, lads?

Lisa's gonna have his jewels if he can't back up what he tweets.

The comment about scrapping the 16k gaming was amusing.

 

[Kenmitch]

It was considered that AMD has no contention this time around by who, exactly?

Debbie Downer's?

 

[blckgrffn]

That guy is a moron.

Anyone catch Frank Azor's tweet replies this morning? All words or actual real talk? Something that's bugging me. It was assumed AMD may not have a good contender, but Frank's post alluded to AMD stockpiling cards to avoid the NVidia fiasco. Would that imply a hot item is in our midst, lads?

Well, if AMD has 20% market share AND the limitless, insatiable, nay infinite demand for the latest nvidia products is what is causing their supply woes - unless AMD has 20% of infinity ready I think they are screwed. /s

I never go on Twitter, but when I do it’s because I follow links here. I did read his latest comments about how there were listening to feedback, etc. Realistically, what could the they really do? Either the containers are essentially already at customs/domestic ports now allowing for delivery in the next four weeks or they are allocating more resources to air freight because they think there is blood in the water. Flipping 2080ti type consumers for at least a year with Big Navi would be quite a coup and worth a ton of advertising bucks in terms of influencers on forums like these. IDK what to read into that.

I would think that if they thought they would be more likely to be competitive in the lower price tiers there isn’t as much margin to go crazy...

Here’s to hoping they have cards worthy of the hype, and more than a handful of them to sell!

 

[blckgrffn]

Lisa's gonna have his jewels if he can't back up what he tweets.

The comment about scrapping the 16k gaming was amusing.

Honestly I felt his tweets were really measured and the troll factor wasn’t crazy. I guess most tweets that I end up hearing about are borderline insane 🤔

 

[Glo.]

Honestly I felt his tweets were really measured and the troll factor wasn’t crazy. I guess most tweets that I end up hearing about are borderline insane 🤔

Or simply they are genuinely excited about RDNA2 GPUs and cannot wait to show them to the public?

Think about how this tech community reacts.

Herkelmann shows "thinking Face" on twitter in response to Jeremy Soller's twitt. Its being retweeted by Hardware Canucks official profile suggesting that face is a signature of white flag, that they can't compete.

Later, FA twitts that tommorow something will be said. It creates hype that AMD FOR SURE WILL SHOW NEW CARDS!!!ONETWOTHREE!

People are just overreacting to everything there is on Twitter.

Why do we always have to try to read between the lines, instead of seeing clearly - without projections?

 

[blckgrffn]

Or simply they are genuinely excited about RDNA2 GPUs and cannot wait to show them to the public?

Think about how this tech community reacts.

Herkelmann shows "thinking Face" on twitter in response to Jeremy Soller's twitt. Its being retweeted by Hardware Canucks official profile suggesting that face is a signature of white flag, that they can't compete.

Later, FA twitts that tommorow something will be said. It creates hype that AMD FOR SURE WILL SHOW NEW CARDS!!!ONETWOTHREE!

People are just overreacting to everything there is on Twitter.

Why do we always have to try to read between the lines, instead of seeing clearly - without projections?

Because audiences are monetized. I am not walking into that bleak topic, but if you want ad revenue you need views. To get views you need followers and you see where I am going here. It’s FUD culture and the noise drowns out the legitimate concerns too. Let’s just say that the current state of social media/media online does not impress me one bit.

Edit:
Re-read my post and I walked into it. Sigh.

 

[eek2121]

Then is 80CU wrong? Or you think density is the same and the missing 40mm2 is used up for better IPC and rt functionality?

I don’t care about die sizes. Right now we have actual proof of specs from AMD regarding 3 of the 4 chips.

EDIT: AMD leaked the firmware of the actual cards via ROCm and a Linux PR and then later tried to retract (via git, rofl). Thus happened just a couple days ago.

@Glo. I ask again, do you have a hat?

 

[NostaSeronx]

Right now we have actual proof of specs from AMD regarding 3 of the 4 chips.

We have...
gc_num_se * gc_num_sh_per_se * gc_num_cu_per_sh
4 * 10 * 2 => 80 CUs for Sienna Cichlid (Navi21=GFX1030)
2 * 14 * 2 => 56 CUs for Arden (Arden=Anaconda=GFX1020)
2 * 10 * 2 => 40 CUs for Navy Flounder (Navi22=GFX1031)
1 * 12 * 2 => 24 CUs for Dimgrey Cavefish (Navi23=VN=MR=LH=GFX1032=GFX1033=GFX1040)
1 * 5 * 2 => 10 CUs for Navi24 (GFX1034)