NVIDIA to Detail New Mainstream GPU at Hot Chips Symposium in August

[kruk]

Polaris was simply a failed architecture that couldn't scale enough to cover the high end.

So Vega is a successful architecture as it can scale from ultra low end to high end? Just trying to understand how this logic works ...

 

[bug]

So Vega is a successful architecture as it can scale from ultra low end to high end? Just trying to understand how this logic works ...

No, you're just playing dumb. I'm not taking the bait.

 

[kruk]

No, you're just playing dumb. I'm not taking the bait.

Why do you always resort to personal attacks when your logic gets challenged?

You don't need to take "the bait", because there is none, but you could stop spreading FUD (thank you very much!). I don't know what your interests are behind this, but ignoring Polaris marketing penetration and calling it failed is something you would not expect from a someone with a Graphics Cards | Expert badge.

 

[bug]

Why do you always resort to personal attacks when your logic gets challenged?

You don't need to take "the bait", because there is none, but you could stop spreading FUD (thank you very much!). I don't know what your interests are behind this, but ignoring Polaris marketing penetration and calling it failed is something you would not expect from a someone with a Graphics Cards | Expert badge.

Show me one link that says Polaris wasn't meant to compete at the high-end and I'll take my words back. And not a link that surfaces one week before Polaris was outed, because by then the failure was set in stone.
Failing to do that, I'll just go back to my belief that, like any other architecture before it, Polaris was meant to compete from high to low end. And it couldn't.

Also "marketing penetration" may be closer to the truth than you'd think.

 

[efikkan]

Polaris didn't come close to delivering what AMD promised. If they had delivered anything close to this it would have knocked out Pascal.

 

[kruk]

Show me one link that says Polaris wasn't meant to compete at the high-end and I'll take my words back. And not a link that surfaces one week before Polaris was outed, because by then the failure was set in stone.
Failing to do that, I'll just go back to my belief that, like any other architecture before it, Polaris was meant to compete from high to low end. And it couldn't.

When in recent years has AMD released a full stack replacement? They have always made incremental updates to their GPUs and rebranded the rest. Why do you think Polaris was meant to be a replacement from bottom to top? Provide links/quotes from credible sources. You accuse, you provide the source.

Polaris didn't come close to delivering what AMD promised. If they had delivered anything close to this it would have knocked out Pascal.
View attachment 102063View attachment 102064

Image 1: I had a OCed 28 nm Cape Verde chip which was rebranded several times and in 2015 you could buy that same chip under the R7 350 name. When I got the Polaris 11 replacement almost two years ago, I measured 3D Mark 11 performance difference on the launch drivers (Win 7) and power pulled from the wall.

I picked the cheapest PCI-E only offer from the available stores, so I think it's random enough. Card had voltage reduced and cores unlocked. I did not touch the core clocks. My measurement has shown that the card is approximately 2.77x faster and it uses approximately 1.15x more power to achieve that. I will say: close enough. I don't feel that I was mislead, but I'm kind of angry that I didn't get the 16 CUs out of the box ... If you believe this measurements or not is up to you. I think my results align well with other benchmarks ...

Image 2: much more controversial and IMO misleading. You don't know the clocks/voltage of the Polaris card, the model of the 950 is unknown, FPS are limited. Therefore I cannot comment on this, sorry. But I have a strong suspicion they used a Pro card with 35 W TBP ...