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

 

[moinmoin]

I recently received an information (third party source, loosely affiliated with a board partner) that some models would release on Nov. 9th. Not claiming this to be accurate at all, but the timing seems right.

AMD did state before that RX 6000 would be out before the consoles launch. Nov. 9th is the very last possible day to keep that promise.

 

[Heartbreaker]

AMD did state before that RX 6000 would be out before the consoles launch. Nov. 9th is the very last possible day to keep that promise.

I think that is just one interpretation of a vague statement.

 

[moinmoin]

I think that is just one interpretation of a vague statement.

The statement that Big Navi will be AMD's first RDNA 2 based product not next-gen consoles isn't exactly vague.

 

[Midwayman]

The statement that Big Navi will be AMD's first RDNA 2 based product not next-gen consoles isn't exactly vague.

Well, it could be a paper launch depending on exact wording and how weasely you want to get.

 

[Kuiva maa]

Well, it could be a paper launch depending on exact wording and how weasely you want to get.

In that case AMD will be in trouble. Frank Azor already publicly reassured that it will not be a paper launch.

 

[Heartbreaker]

The statement that Big Navi will be AMD's first RDNA 2 based product not next-gen consoles isn't exactly vague.

True that isn't vague, but that isn't what was actually said. He never said it was launching before consoles. Just that Big Navi would be their first RDNA2 product. Many have interpreted that they are launching before consoles. But he might see consoles as Sony/Microsoft products...

 

[Konan]

Frank Azor already publicly reassured that it will not be a paper launch.

No, not necessarily -

https://twitter.com/x/status/1309469727383748609

https://twitter.com/x/status/1309465638751137793

 

[amrnuke]

It's highly likely that at launch, not everyone who wants a Zen3 or RDNA2 product will get one at the exact moment they want.

This is true also of the Subaru BR-Z, Ford GT, movie releases, and brisket at Lakewood Smokehouse.

wccftech should be ashamed at this ridiculously idiotic article. Just further sinks their reputation, which is hard to do recently.

 

[Hitman928]

I'd be shocked if AMD's top end GPU doesn't sell out on launch day and probably quickly. The question will be just how fast it sells out, how many end up in actual gamer's hands, and how quickly can they get restocked. Nvidia already said Ampere stock will be limited until 2021 so if AMD can supply enough cards to keep replenished items in stock for 30 minutes or even an hour, that would be a win at this point in my book though others may feel different.

 

[Konan]

Less than a month to go now. Soooo looking forward to seeing how this plays out. At the end of it I’ll get a new card (doesn’t matter who from) but looking forward to price and performance reviews...

that’s a few now who are letting out what AIBs are saying - IMO I think they aren’t seeing the big Kahuna to come

https://twitter.com/x/status/1313609282940801035

 

[dzoni2k2]

No, not necessarily -

https://twitter.com/x/status/1309469727383748609

https://twitter.com/x/status/1309465638751137793

Frank Azor saying it wont be a paper launch and WCCFT interpreting that as infinite supply... Not a paper launch and infinite supply are two very different things.

 

[Konan]

Frank Azor saying it wont be a paper launch and WCCFT interpreting that as infinite supply... Not a paper launch and infinite supply are two very different things.

What I took from it was that Frank is not guaranteeing Or publicly reassuring that there will be sufficient volume for everybody. At the same time he was denying that would be infinite supply which I don’t think WCCF tech was actually saying to begin with.

Either you are confident that there will be supply and you will say so oh you’re not confident they will be supply and you’ll be quiet.

 

[Heartbreaker]

Less than a month to go now. Soooo looking forward to seeing how this plays out. At the end of it I’ll get a new card (doesn’t matter who from) but looking forward to price and performance reviews...

that’s a few now who are letting out what AIBs are saying - IMO I think they aren’t seeing the big Kahuna to come

https://twitter.com/x/status/1313609282940801035

I seriously doubt AMD would only target performance equal to the previous generation. Everything seems to indicate an 80 CU part. IIRC that was even found in some submitted Apple Mac code.

So just assuming that 80 CUs is correct. Given the same IPC and Clock speed as 5700 XT, and you would pretty much match the 3080. Couple that with AMD having a fair bit of time, to tweak clocks/power to match up with 3080.

One way or another they should be trading blows with 3080, or exceeding it if they really had a perf/watt breakthrough.

 

[dzoni2k2]

What I took from it was that Frank is not guaranteeing Or publicly reassuring that there will be sufficient volume for everybody. At the same time he was denying that would be infinite supply which I don’t think WCCF tech was actually saying to begin with.

Either you are confident that there will be supply and you will say so oh you’re not confident they will be supply and you’ll be quiet.

I'll repeat again, Azor took a bet it wont be a paper launch. That's all. Hard launch does not mean sufficient volume for everyone. That's putting words into his mouth.

 

[Olikan]

Less than a month to go now. Soooo looking forward to seeing how this plays out. At the end of it I’ll get a new card (doesn’t matter who from) but looking forward to price and performance reviews...

that’s a few now who are letting out what AIBs are saying - IMO I think they aren’t seeing the big Kahuna to come

https://twitter.com/x/status/1313609282940801035

Ah yes, Ampere 7nm confirmed! videocardz...

 

[dzoni2k2]

I seriously doubt AMD would only target performance equal to the previous generation. Everything seems to indicate an 80 CU part. IIRC that was even found in some submitted Apple Mac code.

So just assuming that 80 CUs is correct. Given the same IPC and Clock speed as 5700 XT, and you would pretty much match the 3080. Couple that with AMD having a fair bit of time, to tweak clocks/power to match up with 3080.

One way or another they should be trading blows with 3080, or exceeding it if they really had a perf/watt breakthrough.

Videocardz just craving attention.

 

[Karnak]

Less than a month to go now. Soooo looking forward to seeing how this plays out. At the end of it I’ll get a new card (doesn’t matter who from) but looking forward to price and performance reviews...

that’s a few now who are letting out what AIBs are saying - IMO I think they aren’t seeing the big Kahuna to come

https://twitter.com/x/status/1313609282940801035

"AMD now targets 3070"

big lol at videocardz for that. I mean, just use your brain. According to TPU, avg. clock for the 2080Ti FE is 1824MHz. That's ~15.9TFLOPs. We know that Turing and RDNA perform similiar with identical TFLOPs within a few percent here and there.

So assuming the fastest Navi21 comes with 20TFLOPs raw power, that alone would be a ~26% increase in TFLOPs and that doesn't include things like higher IPC or other architectural improvements. I mean, the 3070 will be a bit slower than the 2080Ti even according to some of NV own benchmarks. How is a (most likely) >20TFLOPs RDNA2 GPU only a 2080Ti/3070 competitor? Just no.

It will be at least in 3080 territory. The rest remains to be seen.

 

[dzoni2k2]

"AMD now targets 3070"

big lol at videocardz for that. I mean, just use your brain. According to TPU, avg. clock for the 2080Ti FE is 1824MHz. That's ~15.9TFLOPs. We know that Turing and RDNA perform similiar with identical TFLOPs within a few percent here and there.

So assuming the fastest Navi21 comes with 20TFLOPs raw power, that alone would be a ~26% increase in TFLOPs and that doesn't include things like higher IPC or other architectural improvements. I mean, the 3070 will be a bit slower than the 2080Ti even according to some of NV own benchmarks. How is a (most likely) >20TFLOPs RDNA2 GPU only a 2080Ti/3070 competitor? Just no.

It will be at least in 3080 territory. The rest remains to be seen.

More like 22.5 TF.

 

[Konan]

I'll repeat again, Azor took a bet it wont be a paper launch. That's all. Hard launch does not mean sufficient volume for everyone. That's putting words into his mouth.

Maybe you missed the statement I was replying to?

In that case AMD will be in trouble. Frank Azor already publicly reassured that it will not be a paper launch.

And again my point is that he isn’t

 

[dzoni2k2]

Maybe you missed the statement I was replying to?



And again my point is that he isn’t

I didn't miss anything. You replied not necessarily to a statement that it won't be a paper launch.

 

[Konan]

Videocardz just craving attention.

Well that makes
- Igors Lab
- Coreteks
- VideoCardz

All sharing the same AIB information. Don’t think it’s attention craving.
Only Coreteks has shared his perf. opinion on this info so far.
And VideoCardz following up with

https://twitter.com/x/status/1313613983027466240

 

[linkgoron]

Less than a month to go now. Soooo looking forward to seeing how this plays out. At the end of it I’ll get a new card (doesn’t matter who from) but looking forward to price and performance reviews...

that’s a few now who are letting out what AIBs are saying - IMO I think they aren’t seeing the big Kahuna to come

https://twitter.com/x/status/1313609282940801035

So AMD targeted big Navi (500+mm^2 7nm card) to fight a 3070 (390mm^2 8nm) which itself is slower than the 2080ti card? They also only gained 50% performance with a new architecture and double the size card? I find that hard to believe. This would put RDNA2 at Vega levels of fail.

I don't see how AMD could've expected anything less than 2080ti + 30-40% at least, for a die shrink.

 

[Karnak]

Funny how whycry thinks he is talking about N21 while it's actually N22.

 

[Hail The Brain Slug]

By all rights, GA104 top card should have been 30% faster than a 2080ti given the node shrink and architectural change.

If AMD actually has been targeting the top GA104 card, they have been doing it for far longer than it was known that Ampere was not a significant improvement over Turing SKU for SKU.

If the top Navi21 card is targeting 2080ti + 30%, which is what the top GA104 card should have achieved, that's still fantastic.

 

[Zstream]

Going direct to clients means higher margins. I don’t care who builds it as long as it does it’s job and cost appropriately.