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

[Bouowmx]

Radeon RX 5000: good on not calling Navi "RX 3000"

 

[maddogmcgee]

New architecture boys.

 

[JDG1980]

Lisa Su held up a Navi die... I'm sure it won't be long before someone runs some calculations to figure out exactly how big it was. She said it was a relatively small chip.

 

[Bouowmx]

Strange Brigade: AMD Radeon RX 5700 = 1.1 * NVIDIA GeForce RTX 2070
Good on that Sapphire rep

 

[sandorski]

Small with incredible Performance.....she knows how to talk to a guy....

 

[maddie]

Strange Brigade: AMD Radeon RX 5700 = 1.1 * NVIDIA GeForce RTX 2070
Good on that Sapphire rep

RX 5700 sounds like the lower end version.

 

[Saylick]

RX 5700 sounds like the lower end version.

Gotta leave room for a Radeon VII replacement....

 

[crazzy.heartz]

WOW, a RX 5700 is competing against RTX 2070

So, that implies a RX 5800 vs RTX 2080

Maybe, an RX 5900 vs RTX 2080Ti

& New RDNA Architecture !!

Good Times ahead

 

[Saylick]

Lisa Su held up a Navi die... I'm sure it won't be long before someone runs some calculations to figure out exactly how big it was. She said it was a relatively small chip.

I can't say it looked like sub-200 mm2. Looks closer to 300mm2 ?

 

[JDG1980]

I can't say it looked like sub-200 mm2. Looks closer to 300mm2 ?

Vega 20 is 331mm^2... it would be surprising if this wasn't quite a bit smaller, since it has worse performance than Radeon VII. I agree it looked bigger than I thought it would be, but I'm betting somewhere around 250mm^2.

 

[alcoholbob]

Strange Brigade: AMD Radeon RX 5700 = 1.1 * NVIDIA GeForce RTX 2070
Good on that Sapphire rep

Hmm well Vega 56 is normally 10% faster than 1070 but is 34% faster than 1070 in Strange Brigade. So if RX 5700 is 10% faster than RTX 2070 at Strange Brigade, it might still be about 10% *slower* on average than a RTX 2070.

 

[itsmydamnation]

if it really is a new uarch, then trying to guess performance across the board is out the window.

 

[Saylick]

Think anyone can estimate the die size from this picture? Haha.

 

[IntelUser2000]

@Saylick Assuming my fingers are the same size as the hand model, I got about 300mm2, very roughly speaking. Perhaps its actually smaller, but still quite a bit bigger than Polaris with 232mm2.

if it really is a new uarch, then trying to guess performance across the board is out the window.

Yea, newer generations also focus on weaknesses of the previous generation too. That throws such arbitrary comparisons out the window.

At least AMD is claiming its 25% faster per clock across sets of games, not "IPC" as they claimed with Polaris, and then Vega which turned out to be barely faster than predecessors.

 

[gdansk]

I'm worried RDNA is simply branding since people kept complaining about GCN being the problem.
It may be a new design, but they always say that.

 

[Head1985]

WOW +25% IPC and +50% performance/watt To be honest i was more interested in ryzen 3000, but i am more interested in navi now.

 

[GodisanAtheist]

Glad AMD dropped some meaningful info to chew on.

Using TPU's numbers for 1440P from their most recent performance update:
Assuming a 1.2x IPC increase translates perfectly into a 20% performance boost over existing parts (huge if here), it would put a Radeon VII with those kinds of IPC improvements at 89% (up from 74%) vs 2080 82% and 2080ti 100%. That's not bad and would bring AMD within rough performance parity in standard rasterized scenarios with NV's Turing arch.

Now so long as the performance per watt and scalability of the arch is up to snuff they might have some legitimate growth potential in terms of gaming performance and may actually be able to make some inroads into the laptop space.

NV or course isn't going to be sitting still and if RDNA is going to be AMD's next 10 year arch then we could be talking about how RDNA is a garbage arch in 2028 and AMD should never have wasted their time etc etc etc

 

[JasonLD]

WOW +25% IPC and +50% performance/watt To be honest i was more interested in ryzen 3000, but i am more interested in navi now.

Well, Polaris touted 2.8x perf/watt....
They already showed it hangs around 2070, so that is that.

 

[IntelUser2000]

Assuming a 1.2x IPC increase translates perfectly into a 20% performance boost over existing parts (huge if here),

Well, Polaris touted 2.8x perf/watt....
They already showed it hangs around 2070, so that is that.

Good thing is they are not talking about not-relevant-in-real-world IPC or perf/watt but gains in gaming,

Altogether, a Navi core/CU should be 25% faster than a Vega core on a clock-for-clock basis. The devil is in the details of course – AMD's data is based off of their internal testing, taking the geomean of 30 games tested at 3840x2160 with Ultra settings and 4x AA

Only question is whether the 25% number is realistic, but at least its much more practical than "IPC is up xxxx%!!!"

 

[DrMrLordX]

They already showed it hangs around 2070, so that is that.

Exactly. Now it comes down to prices. If the Sapphire rep that dropped the $499 price is right then we have no reason to celebrate.

Don't get me wrong: I'm glad AMD is still trying to do something to serve the consumer dGPU market. Let's hope they do better than a $499 2070 replacement.

 

[mohit9206]

New architecture boys.

Is that any reason to get excited? Someone here earlier told me GCN is just an ISA so even if it was GCN that wouldn't make any difference.

 

[JasonLD]

Exactly. Now it comes down to prices. If the Sapphire rep that dropped the $499 price is right then we have no reason to celebrate.

Don't get me wrong: I'm glad AMD is still trying to do something to serve the consumer dGPU market. Let's hope they do better than a $499 2070 replacement.

Even if that is true, can't really blame them since 7nm appears to be considerably more expensive than 14nm on equilvalent size.

 

[DrMrLordX]

Even if that is true, can't really blame them since 7nm appears to be considerably more expensive than 14nm on equilvalent size.

Why not? What's the margin on the card? We're not talking about an HBM2 behemoth like Radeon VII. The BoM on the card should be much, much lower than for Radeon VII, RX Vega, or Fiji. VRMs shouldn't have to be overbuilt either, since the expected power use should be more on Polaris levels.

So why price it like it's RX Vega? It makes no sense.

 

[JasonLD]

Why not? What's the margin on the card? We're not talking about an HBM2 behemoth like Radeon VII. The BoM on the card should be much, much lower than for Radeon VII, RX Vega, or Fiji. VRMs shouldn't have to be overbuilt either, since the expected power use should be more on Polaris levels.

So why price it like it's RX Vega? It makes no sense.

I really don't expect $499 to happen, but I was saying in terms of what if scenario. If AMD wants to sell 2070 class GPU and hope to steal some marketshare away from Nvidia, they have to price it around $399-$449.

 

[DrMrLordX]

I really don't expect $499 to happen, but I was saying in terms of what if scenario. If AMD wants to sell 2070 class GPU and hope to steal some marketshare away from Nvidia, they have to price it around $399-$449.

Agreed. I think the leak prices of $399 for Navi14 and $499 for Navi10 are both about $100 too high, though lowering them both by $50 wouldn't be a bad idea either.

Somewhere along the line, they were supposed to replace Polaris. RX480 launched at $239 for the 8GB version. RX470, $179. RX580, $229 for 8GB version. RX570, $169. RX590 was the most expensive Polaris at $279.

Where is the Polaris replacement?