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

 

[JustMe21]

Quoting a post from Ampere thread as it was about this Coreteks video:

Claiming RDNA2 only reaching RTX 3070 speeds (and the only selling point being supposedly 16GB of GDDR6).



I'm a bit sceptical about Coreteks claims (though dissapointing rumors about AMD's products have became true in the past). Why?

1. AMD obviously knew what their Big Navi performance target was going to be from the start (as they set it).
2. They knew where 2080 Ti stands for about 2 years now.
3. They also knew the next Nvidia card will be on Samsung 7nm EUV (as it was wildly reported at the time)
4. They still decided to market Big Navi as "high-end!" knowing it will release somewhere around 2020Q4 / 2020Q1.

If roughly 2080 Ti's perfomance is all they planned for, they must have had rocks in their head thinking this will be "high end" in 2021. There is no way Nvidia was gonna bring less than 30% performance uplift. They could have gotten most of that from the shrink.

Considering historic trends it would have been absolutely reasonable to model Ampere as at least 40-50% performance uplift vs Turing. Yes, Nvidia could have underdelivered, but you don't design your products on the premise that the competition will screw it up - see how well it turned out for Intel, or AMD in the past to that matter. On the other hand, expecting competiton to bring the best usually ends up with great products. See AMDs 64 Core "Rome" processors as an example, which were originally supposed to compete with Ice Lake Server from day one.

To top it all off while Ampere is slightly better than expected (mostly due to higher TDPs and lower prices), there is nothing that unique about 2070 reaching 2080 Ti perf.

• 970 was roughly as fast as 780 ti.
• 1070 was quite a bit faster than 980 ti.

Turin generation was the only outlier in this regard and considering Nvidia essentially went with the same node, it was somewhat to be expected (the only main difference between 16nm and 12nm is recticle limit).

TL;DR:
I hope Coreteks is dead wrong. If he's right AMD's project managers must have been dumb as a rock to market Big Navi as "High end"

Well, the only known numbers at the time were the numbers for the 2080 Ti and if AMD could beat that, then it is still pretty good as everything else is still speculation.

The Nvidia presentation mentioned DLSS way too much for my happiness as I'd rather have high FPS at native 4K rather than high 4K FPS achieved using AI, 1/4 upscaled renderings. One of the reasons I have preferred AMD cards over Nvidia, despite typically not being as high of FPS, is that their image quality always looked better to me and DLSS represents an even further reduction in image quality on Nvidia's side. I also don't like how Nvidia is trying to work in another of their proprietary technologies into games. Make it an open standard if you want people to truly embrace it.

And the 2080 Ti in the Doom Eternal demo Nvidia showed had much lower numbers than the the other 2080 Ti Doom Eternal video on YouTube. Also the 2080 Ti was using an older driver version and you would think Nvidia could have used the same 455 driver version for their own older product, right?

 

[leoneazzurro]

Big Navi Radeon 6000 Branding Unearthed In AMD Custom Fortnite Map And How To Find It

With Big Navi not so far away, it does not come as a surprise that AMD is trying to garner some attention.

hothardware.com

Teasing is starting

 

[Tup3x]

After Coreteks' Ampere rumours I wouldn't put any weight on his RDNA2 rumours/speculation.

 

[Kenmitch]

Sony has already said that in order to be one of the first ones with the console, you have to prove your a sony fan, whatever that means. But, it could be due to limited quantities early on.

Well maybe they're going to have a giveaway?

 

[blckgrffn]

Well maybe they're going to have a giveaway?

Thank goodness I already activated my email address!

 

[Kenmitch]

Looks like AMD got sidetracked. Maybe it explains the delays?

From Big Navi to Big Wheels? AMD Sells Mountain Bikes Now

Lisa Su's Got a New Ride

www.tomshardware.com

 

[CastleBravo]

Looks like AMD got sidetracked. Maybe it explains the delays?

From Big Navi to Big Wheels? AMD Sells Mountain Bikes Now

Lisa Su's Got a New Ride

www.tomshardware.com

I'm predicting a lot of crashes due to bad drivers.

 

[GoodRevrnd]

Big Navi Radeon 6000 Branding Unearthed In AMD Custom Fortnite Map And How To Find It

With Big Navi not so far away, it does not come as a surprise that AMD is trying to garner some attention.

hothardware.com

Teasing is starting

I'm concerned the cadence of teases isn't higher. If they have something competitive AND available I would think there'd be a spoiler event a few days before 3000 launch. I guess there's still time but would think we'd hear more by now.

 

[blckgrffn]

Looks like AMD got sidetracked. Maybe it explains the delays?

From Big Navi to Big Wheels? AMD Sells Mountain Bikes Now

Lisa Su's Got a New Ride

www.tomshardware.com

21 speed mountain bike?!? Gross $299 though, that's value pricing 🤔

That said, I saw a bike on a carrier the other day (can't remember which day exactly, this week though) with tons of AMD graphics on it. Road bike. I was surprised - it must either have been a prize or that is a whole lot of fandom to have coordinated colors stickers (but looked like painted on) on a bike like that

 

[Kenmitch]

21 speed mountain bike?!? Gross $299 though, that's value pricing 🤔

That said, I saw a bike on a carrier the other day (can't remember which day exactly, this week though) with tons of AMD graphics on it. Road bike. I was surprised - it must either have been a prize or that is a whole lot of fandom to have coordinated colors stickers (but looked like painted on) on a bike like that

I wouldn't buy one. People love cheap bikes these days.

Just thought I'd drop it to ease some tension during the wait and see.

Rumor has it there's a PS5 event on the 9th.

 

[dacostafilipe]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

 

[raghu78]

He's not bad. Just mistaken I think.

https://twitter.com/x/status/1301839482287796224

8GB VRAM SKU is the big giveaway.

I'll just say I talk to him on Discord... it's not the first time he's gotten GPU dies completely confused. For the longest time he thought Navi10 had both GDDR6 and HBM2 controllers, insisting that a HBM2 SKU would launch for Apple at some point.

For somebody who makes a living off youtube channel talking tech and saying he has sources he is really bad at even a basic understanding of technology. There is enough information to realize RDNA2 is much more efficient that Ampere from a perf/sq and perf/watt perspective. The only question remains is scalability of performance which I think has been addressed with the architecture of RDNA2 which scales resources by using the Shader engine as a building block just like Nvidia does with Graphics Processing Cluster.

 

[CastleBravo]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

I don't trust Coreteks alone, but with Igor reporting similar things, I suspect there is some truth to it. Both of the leaks are likely from an AIB, so one possibility is that AIBs don't know anything about the full spec Navi21 card (no AIB models until next year?), and the leaks are about the cut down Navi21, or maybe even Navi22.

 

[Konan]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

Well if his perf. speculation is true it lines up with a few others including Coreteks and others. Also seems like AMD have to prioritize the consoles first (which makes sense) and links to what @uzzi38 said the page back re: https://www.pttweb.cc/bbs/PC_Shopping/M.1599147022.A.4CF and also Tweaktown.

 

[Kenmitch]

I don't trust Coreteks alone, but with Igor reporting similar things, I suspect there is some truth to it. Both of the leaks are likely from an AIB, so one possibility is that AIBs don't know anything about the full spec Navi21 card (no AIB models until next year?), and the leaks are about the cut down Navi21, or maybe even Navi22.

I go with the X-files way....Trust no one.

 

[CastleBravo]

Igor's leak says that Big Navi somehow goes from being just barely above 3070 to being a 3080 competitor by bumping the power up from 275w to ~300w, but this is supposed to be really bad for efficiency? That doesn't add up IMO.

 

[uzzi38]

Igor's leak says that Big Navi somehow goes from being just barely above 3070 to being a 3080 competitor by bumping the power up from 275w to ~300w, but this is supposed to be really bad for efficiency? That doesn't add up IMO.

I don't get this either, especially considering the 3080 is a 320W TGP card.

 

[Hitman928]

How would his AIB source know the performance of Big Navi if they don't even have chips from AMD yet? Sounds like a rumor of a rumor type of thing. I would be shocked if top model Big Navi is only able to edge out a 3070 at a higher wattage, that means something went terribly wrong with scaling the RDNA2 architecture.

 

[Konan]

Maybe it is Navi22 and they are AIB only? but then 275W in that bracket sounds bad.

 

[uzzi38]

How would his AIB source know the performance of Big Navi if they don't even have chips from AMD yet? Sounds like a rumor of a rumor type of thing. I would be shocked if top model Big Navi is only able to edge out a 3070 at a higher wattage, that means something went terribly wrong with scaling the RDNA2 architecture.

Well from that translation apparently it's Nvidia's numbers... when AIBs don't even have cards?

Honestly, there's a whole lot that doesn't make sense.

Unless you menat assuming the 3080 is as powerful as Nvidia state. In which case, we don't really know if AIBs are being told the truth if AIBs are only going to launch them next year. Why bother?

 

[tajoh111]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

If AMD has the same scaling issues due to not increase ROPS, I can see these rumors being true. Fury X had these issues as these issues as well as Vega.

Vega in particularly showed this issue. From Rx 580 to Vega 64 only resulted in a 50% increase in performance

AMD Radeon RX Vega 64 8 GB Review

Our AMD Radeon RX Vega 64 review confirms that the company achieved major performance improvements over their last-generation Polaris and Fiji cards: Vega is faster than the GTX 1080. We tested six different performance configurations of the Vega 64, with surprising results.

www.techpowerup.com

2304 to 4096 shaders = 1.77x increase.

On top of this, TDP increased from 185watts to 300watts.

During that generation, Vega was promised to be a new architecture and also improved performance per watt by 50% as seen in the chart on the bottom.

AMD achieving 50% more performance than a rx 5700xt while only consuming 50watts more power but would be an improvement to their track record of recent big chips.

I.e

RX580 --> Vega 64
1x performance --> 1.5x performance
185watts --> 300watts
AMD got to double ROPS during this generation too.

5700xt --> Big Navi
1x performance -->1.5-1.6x performance
225watt -->275watt.

This isn't that bad considering this is not really a new node and AMD again is using 64 ROPS. People should set their expectations lower to be pleasantly surprised. Igor definitely has some connections in the industry and generally has been on the pro AMD side with his overclocking videos, while at one point working with AMD directly.

 

[exquisitechar]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

Definitely not true.

 

[uzzi38]

If AMD has the same scaling issues due to not increase ROPS, I can see these rumors being true. Fury X had these issues as these issues as well as Vega.

Vega in particularly showed this issue. From Rx 580 to Vega 64 only resulted in a 50% increase in performance

2304 to 4096 shaders = 1.77x increase.

On top of this, TDP increased from 185watts to 300watts.

During this generation, Vega was promised to be a new architecture and also improved performance per watt by 50%.

AMD achieving 50% more performance than a rx 5700xt while only consuming 50watts more power but would be an improvement to their track record of recent big chips.

I.e

RX580 --> Vega 64
1x performance --> 1.5x performance
185watts --> 300watts

5700xt --> Big Navi
1x performance -->1.5-1.6x performance
225watt -->275watt.

We don't know yet if ROP count hasn't increased. We know the number of RBEs etc, but what we don't know is how ROPs would be layed out. It's clear that AMD have reworked the rasterisation pipeline from RDNA. Especially considering the GPixel/sec count that MS gave for the Series X.

 

[Samwell]

Igor's leak says that Big Navi somehow goes from being just barely above 3070 to being a 3080 competitor by bumping the power up from 275w to ~300w, but this is supposed to be really bad for efficiency? That doesn't add up IMO.

He's not saying it's barely above 3070 and need 300W for the 3080. He said it's above 3070 (might be right in the middle between 3070 and 3080), but they need more than 300W to get near 3080. He didn't say how much more than 300W they need for the 3080. But anyway, he also mentions it's very unsure and a lot of speculation. So we should take it with a grain of salt.

More important are his mentions about AIB not having any infos on the cards yet, as he's having excellent contacts to AIBs. So probably no AIB RDNA2 this year. That's disappointing.

 

[linkgoron]

Some RDNA2 rumours from Igor's Lab : (German)

- Actual "big navi" has 275W and just above a 3070 performance (if the Nvidia charts are true)
- Possibility that AMD pushed the TDP to +-300W to get around 3080 performance but that would be really bad for the efficiency
- Board partners don't yet have chips or even BOM (bill of materials) what will lead to only reference design to be available this year

Just above a 3070 would be a total failure for a 500mm^2 card and a new architecture. The 5700xt is around 2/3 of a 2080ti performance-wise in 4K. Getting a double the size card with a new arch only gain ~50%-60% would be pretty terrible.