[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[ozzy702]

I was thinking about the “new” RDNA architecture.

Facts:

1. AMD’s older roadmaps showed Navi as a new architecture. This later changed and now Arcturus or whatever the next card is called has the new architecture.

2. A rumor surfaced that the console folks (Sony/Microsoft) wanted the new GPU to be easily backwards compatible to make porting old games to it easy

3. AMD says Navi is a new architecture, call it RDNA

Interpretation:

1. It started as a clean sheet but much of GCN needed to be pulled back in to give it that easy backward compatibility.

2. Interesting name, Radeon DNA. It’s like they're trying to say it’s new, but still possess a lot of what came before.

Conclusion:

When is a “new” architecture really new, and when is it the older architecture with some major/minor updates? Can’t blame AMD for this no matter what the truth is. It’s always a thin line trying to build something new and radical while still retaining the capability to run the old stuff efficiently.

Unfortunately, this is what I expect as well but I am absolutely hoping that I'm wrong and AMD knocks it out of the park. I'm curious for the whole 9900k 2080TI vs AMD setup results, that whole thing has me scratching my head.

 

[RaV666]

As for the whole 3 Fans thing.
These were asrock AIB prototypes!
I had rx 480 with three fans! It doesnt mean anything.

The TDP`s however were leaked , its 180 for PRO variant and 225W for XT, whole card.
So, neither good or bad.But it will get bad when nvidia goes to 7nm.

 

[guskline]

Perhaps Navi is suffering from the drain on AMD's Radeon group of engineers.

 

[JDG1980]

Kinda feel the hype train starting up again. Thread looks a lot like the "there is no reason Polaris should be any slower than the GTX 1070" threads of yesteryear.

Well, to be fair, at the time it wasn't quite clear to any of us why GTX 1070 was so much faster than RX 480, given that their raw processing power is very similar and they both have the same memory bandwidth. It wasn't until August 2016 (a few months after RX 480's release) that it came to light that Maxwell (and obviously Pascal as well) used tile-based rendering. Of course, RTG must have known all along that this was a large part of the answer (why their own efforts at DSBR in Vega weren't nearly as effective is a whole other can of worms).

I think we now have a much clearer idea of the weaknesses of GCN than we did at the time Polaris was released. The question is whether RTG will be able to fix these with Navi or not. Many of us got burned by the hype over Vega and now are defaulting to "I'll believe it when I see it". Hopefully Lisa Su was able to bring some order to the RTG side of the business after Raja finished screwing around, but the lead time of GPUs make me doubtful if there was enough time to get Navi right.

 

[JDG1980]

I feel like more than anything it's the launches of previous cards that tainted a lot of good will. Radeon 7 in all honestly felt like a rushed odd ball last minute decision to me.

Radeon VII wasn't a big deal; it was a low-key release with minimal hype, and probably the only reason it even happened was that Nvidia blew up the prices on Turing.

More than anything else, it was the Vega pre-launch hype that soured fans on RTG, and sowed skepticism about all future releases. Raja intentionally hyped Vega to the moon, bragging (with plausible deniability) that it would be a Volta-killer - and then it turned out that it could barely match AMD's second-best Pascal chip, at far higher power consumption. Fiji had been somewhat of a disappointment; Polaris was OK but nothing special; Vega was a Bulldozer-class failure.

 

[lifeblood]

As for the whole 3 Fans thing.
These were asrock AIB prototypes!
I had rx 480 with three fans! It doesnt mean anything.

The TDP`s however were leaked , its 180 for PRO variant and 225W for XT, whole card.
So, neither good or bad.But it will get bad when nvidia goes to 7nm.

Compared to Radeon VII, 225W is absolutely frigid. Compared to it's competition the RTX 2070, not so much.

Based on the rumors right now, Navi XT is not as fast as 2070, not as power efficient as 2070, and not feature equivalent to 2070, and yet cost essentially the same at $499. Something isn't right.

AMD would not try to sell a card that cost the same as a RTX 2070 unless it was arguably the same value. Those prices from the Sapphire article must be wrong, or the performance is wrong, or something else is wrong.

 

[Glo.]

Based on the rumors right now, Navi XT is not as fast as 2070, not as power efficient as 2070, and not feature equivalent to 2070, and yet cost essentially the same at $499. Something isn't right.

You've just got every single demo and rumor to say that Navi is faster than RTX 2070.

Your post is just an epitome to what AMD's brand perception is, and how it affects their products.

 

[lifeblood]

You've just got every single demo and rumor to say that Navi is faster than RTX 2070.

Your post is just an epitome to what AMD's brand perception is, and how it affects their products.

Every single demo consists of one game. Not exactly a landslide of evidence.

However, AMD are not idiots they know they cannot price it the same as RTX 2070 while it's hotter and missing RT unless the performance is superior to RTX 2070. That's basic economics.

So I will modify my prediction. Assuming $499 and 225W TDP then it's performance will be at least 5% better than RTX 2070.

Of course at 5% better performance it's still a disappointment. It's competitive with the competition, but its competition is badly overpriced as will it be. I guess I need to start shopping for a used 1080 or something.

 

[Glo.]

Every single demo consists of one game. Not exactly a landslide of evidence.

However, AMD are not idiots they know they cannot price it the same as RTX 2070 while it's hotter and missing RT unless the performance is superior to RTX 2070. That's basic economics.

So I will modify my prediction. Assuming $499 and 225W TDP then it's performance will be at least 5% better than RTX 2070.

Of course at 5% better performance it's still a disappointment. It's competitive with the competition, but its competition is badly overpriced as will it be. I guess I need to start shopping for a used 1080 or something.

If Navi is faster than RTX 2070 - yes they can price it at 500$. Simple as that. They will have at least the same, or higher margin, because the die size is much smaller than RTX 2070, so AMD will be a winner(alongside consumers) if both companies will go on a price war in mainstream market.

That, your, last sentence basically proven what I have been writing for a lot of time, about AMD's brand perception.

 

[JujuFish]

If Navi is faster than RTX 2070 - yes they can price it at 500$. Simple as that. They will have at least the same, or higher margin, because the die size is much smaller than RTX 2070, so AMD will be a winner(alongside consumers) if both companies will go on a price war in mainstream market.

That, your, last sentence basically proven what I have been writing for a lot of time, about AMD's brand perception.

If Navi is only marginally faster than the RTX 2070, how are consumers winners?

 

[Glo.]

If Navi is only marginally faster than the RTX 2070, how are consumers winners?

I think you have answer, in that post, when you read whole context, and not just part of it. Price war in the mainstream market.

 

[JujuFish]

I think you have answer, in that post, when you read whole context, and not just part of it. Price war in the mainstream market.

If Navi is only marginally faster, then NV doesn't have any need to change their prices. Brand name alone, along with RT features, will carry them through.

 

[JDG1980]

If Navi is faster than RTX 2070 - yes they can price it at 500$. Simple as that.

If it's only a little faster than the 2070 (5-10%) this won't be a winning strategy, because Navi will be inferior to the RTX 2070 in several other ways (perf/watt, OpenGL support, no tensor cores/RT or DLSS). Yes, brand recognition is a factor as well.

If the uncut Navi card rivals RTX 1080 (say, within 5% of it on average in a broad selection of AAA games) then a $499 price point might be more realistic. That would mean it would have to be about 15-20% faster than the RTX 2070.

 

[JDG1980]

You've just got every single demo and rumor to say that Navi is faster than RTX 2070.

No, we've got one demo on a cherry-picked title that heavily favors AMD showing Navi as being ~10% faster than RTX 2070. (In contrast, the CPU demos showcased titles in which Ryzen had previously experienced worse performance than Intel's chips.) It would have been much more impressive if they had picked, say, GTA V.

 

[Mopetar]

If Navi is faster than RTX 2070 - yes they can price it at 500$. Simple as that. They will have at least the same, or higher margin, because the die size is much smaller than RTX 2070, so AMD will be a winner(alongside consumers) if both companies will go on a price war in mainstream market.

I think the problem with your assumption is that 7nm is far more expensive than the 12nm process that NVidia is using as well as much less mature. NVidia likely isn't getting too many chips that are complete junk, and even some of ones with a lot of defects are probably going into a special bin that will be released later like the GP104 versions of the 1060. You can already tell that TU104 dies are being saved up for a 2070 Ti.

Keep in mind that AMD may well have to price at $500 just to justify using the wafers to make GPUs instead of more Zen 2 chiplets. Those things are only about 80 mm^2 and AMD can slap two defective ones together and sell that for $500.

If 7nm would have been more economical, NVidia would have gone that route and you know that they have more than enough money to afford it. I doubt that AMD making out like a bandit at $500, especially once you factor in the more expensive GDDR6 (at least it's not HBM2 though) and board costs, it doesn't seem like such a sweet deal.

 

[piesquared]

If Navi is only marginally faster, then NV doesn't have any need to change their prices. Brand name alone, along with RT features, will carry them through.

Actually, ray tracing features are a negative. The cost of the added bloat to the die size is passed on to consumers for absolutely zero benefit. It's too weak to be useful and there are only a handful of games that even incorporate ray tracing even if the card could draw them on the screen adequately. Im not sure where all these people are you are talking about that see RTing as a must have feature, but from what ive been reading it's the opposite. Even leatherman himself said the uptake on the cards was like a punch in the gut. Ray tracing might catch on when it addresses more than a niche of a niche of the market and a proper implementation is released that doesn't require consumers to pay 50% extra just because of the ballooned die size. Like doing the majority of work on the CPU for example where it's best suited where there is now an abundance of extra resources with all those cores available.

 

[DarthKyrie]

Actually, ray tracing features are a negative. The cost of the added bloat to the die size is passed on to consumers for absolutely zero benefit. It's too weak to be useful and there are only a handful of games that even incorporate ray tracing even if the card could draw them on the screen adequately. Im not sure where all these people are you are talking about that see RTing as a must have feature, but from what ive been reading it's the opposite. Even leatherman himself said the uptake on the cards was like a punch in the gut. Ray tracing might catch on when it addresses more than a niche of a niche of the market and a proper implementation is released that doesn't require consumers to pay 50% extra just because of the ballooned die size. Like doing the majority of work on the CPU for example where it's best suited where there is now an abundance of extra resources with all those cores available.

Now that's just silly talk wanting to use the right tool for the job.

 

[railven]

Actually, ray tracing features are a negative. The cost of the added bloat to the die size is passed on to consumers for absolutely zero benefit. It's too weak to be useful and there are only a handful of games that even incorporate ray tracing even if the card could draw them on the screen adequately. Im not sure where all these people are you are talking about that see RTing as a must have feature, but from what ive been reading it's the opposite. Even leatherman himself said the uptake on the cards was like a punch in the gut. Ray tracing might catch on when it addresses more than a niche of a niche of the market and a proper implementation is released that doesn't require consumers to pay 50% extra just because of the ballooned die size. Like doing the majority of work on the CPU for example where it's best suited where there is now an abundance of extra resources with all those cores available.

So what you're basically arguing is, consumers will turn down the product with a stronger brand value, with more features [I get this is the crutch of your argument], and (because remember clean sheet design) mature drivers for the same price?

AMD really needs a home run. We're running out of positive spin on this one.

EDIT: Why am I getting dejavu for GTX 680 vs HD 7970. HD 7970 had by far the most advantages and still got roflstopped by the 2GB card. All I'm reading is how Navi will equal or be marginally faster than RTX 2070 at the same price, but "this is the one, this time AMD will nail it!"

 

[alcoholbob]

The GTX 1080 is a 314mm2 die on 16nm. If it got shrunk down to 7nm TSMC it would be around 30% smaller than this the RX 5700 card, and a TDP of around 100W, at the same performance class. This is a card that came out 3 years ago, and Navi is the best AMD could come up with even after substantial funding from Sony.

When Nvidia refreshes launches their 7nm GPUs it's going to be brutal for AMD.

 

[itsmydamnation]

The GTX 1080 is a 314mm2 die on 16nm. If it got shrunk down to 7nm TSMC it would be around 30% smaller than this the RX 5700 card, and a TDP of around 100W, at the same performance class. This is a card that came out 3 years ago, and Navi is the best AMD could come up with even after substantial funding from Sony.

When Nvidia refreshes launches their 7nm GPUs it's going to be brutal for AMD.

What substantial funding from Sony? Chips take 3-5 years beginning to end
so what part of this looks awesome?
https://ycharts.com/companies/AMD/r_and_d_expense
Rory Reid gutted the GPU side to keep the CPU side afloat, its going to take time to fully recover

Second point traditionally AMD/ATI have been well headed in adoption of process then NV. Given AMD in now generating cash flow this could be the case again( like right now) which means how long is NV selling 7nm cards before AMD is selling 6 or 5nm cards?

 

[prtskg]

"this is the one, this time AMD will nail it!"

Is anyone actually saying it though?

 

[railven]

Is anyone actually saying it though?

In those exact words, of course not. Just this page alone I've read consumers will choose the product with less on offer for the same price because reasons and AMD will be on a new node by the time NV gets to [selling high volume] 7nm, thus negating any advantage NV has.

This aren't very sound arguments. The equivalent of me saying tongue-in-cheek "this is the one, this time AMD will nail it!"

 

[Timorous]

I can only see a $500 price point happening if it matches or slightly exceeds 2070 and the supply is low because they want more wafers making Zen2 dies rather than Navi dies. With low supply they would probably sell out regardless, just like R7 sells despite the 2080 being the better card.

OTOH perhaps the PS5 and Xbox Two are going to use a Zen2 die and a Navi GPU/IO die rather than as a monolithic die and the Navi desktop part is just a higher clocked version of the console part meaning they need to produce them anyway so can afford to sell them at $400 because they have enough volume.

255mm^2 + 80mm^2 + a bit for additional IO features is pretty much the same as the die size for the OG PS4 at launch and the Xbox One so the size fits. Reduce the clocks back towards the sweet spot and the GPU + CPU dies should consume 200W or less so the power fits as well.

 

[Ottonomous]

In those exact words, of course not. Just this page alone I've read consumers will choose the product with less on offer for the same price because reasons and AMD will be on a new node by the time NV gets to [selling high volume] 7nm, thus negating any advantage NV has.

This aren't very sound arguments. The equivalent of me saying tongue-in-cheek "this is the one, this time AMD will nail it!"

You're conflating comments on the GPU landscape with a direct attack on the RTX 2070/nvidia. A critique of the usefulness of RT technology and a comment on a possible angle for AMD competitiveness /= attack on nvidia

People are just pointing out best case scenarios for AMD in the hopes that it'll reintroduce competition in the mainstream segment

 

[tviceman]

So what you're basically arguing is, consumers will turn down the product with a stronger brand value, with more features [I get this is the crutch of your argument], and (because remember clean sheet design) mature drivers for the same price?

AMD really needs a home run. We're running out of positive spin on this one.

EDIT: Why am I getting dejavu for GTX 680 vs HD 7970. HD 7970 had by far the most advantages and still got roflstopped by the 2GB card. All I'm reading is how Navi will equal or be marginally faster than RTX 2070 at the same price, but "this is the one, this time AMD will nail it!"

The HD7970 was a superior card long term, but it got stomped by the 680 out of the gate because initially the 680 was faster, cheaper, quieter, and more efficient at release. And it came out only 12 weeks after the HD7970.

Do you see Navi being faster, cheaper, quieter, more efficient, and right on the heels of Turing's release? Nope. As you've argued, Navi has to absolutely be cheaper. The card demonstrated last week is probably 10% slower than RTX 2070 overall. If it's $379-399 then that will be a big win for AMD. If it's $450-499 then they lose.