AMD: 7nm Navi and Epyc 2 launch in third quarter

I'm looking forward to the Guru3D reviews. It'll be nice to see a really new AMD GPU reviewed after a long while. I didn't really consider Vega VII to be so interesting from a personal point of view as it was obviously not within my price range, nor was it really anything new, per se. Navi, of course, is a different thing. Nvidia RTX also turned out to be a disappointment as far as prices go (and the practicality of RT), which made their introduction fall flat from my pov.

 

Kaarme, indeed. As long as it's a significant jump over my 580 and the noise and temp is kept under control, I can see this being my next card.

 

Navi is GCN's last gasp.
Its not a massive change from Vega, just more efficient, and maybe setting things up for the real next gen, new architecture that comes after (Chiplets, yay).

and I predict they will still clock the balls out of it, thats why it may still consume up to 200+watts, even though its like upper mid-range.

 

AMD's GPU architect has spoken.

 

I hope it have 2560Sps for XX80. It could replace my 580.

Im glad i didnt bought any of those crappy gtx turing cards instead.

 

Did you know that Navi and Vega development started at same date (Finish of Polaris design)? And that Navi got top guns team while Vega got remaining workforce?
Did you know that this remaining work force joined Navi development moment Vega was done?

Can you summarize how much time There was for Vega vs. Navi development and what makes you think that Navi is change from Vega instead of being change from Polaris? (That's if Navi is GCN, which it most likely is not.)
And can you elaborate on how that time investment to Navi resulted in your statement: "Its not a massive change from Vega" ?

 

I've been told that the GCN name is just that... just a name, and that the current GCN architecture in Navi actually is a pretty far cry from the original GCN. I should also note that it's not out of the question for Navi to use chiplets.

And why would clocking it extremely high as far as power consumption be a surprise or even bad for that matter considering the target audience (desktop enthusiasts)? Of course they're going to clock as high as they reasonably can before hitting notably diminishing returns. Despite that I still expect a decent amount of headroom for OCs assuming there isn't some software problem like there was with Vega 7.

The only reason I'm not hyped about Navi is because I'm not expecting it to surpass the 1080 Ti/2080, even in a best case scenario not by enough, to be an upgrade for me. Here's hoping that turns out to be false.

 

200W is pretty optimistic

Realistically, probably 250W+

 

source?

 

Ask @mockingbird2 .

 

Is it just me or does everyone see a pattern where AMD is off with their timing once again... If AMD is only releasing their new cards in the 3rd term then I have a feeling Nvidia will throw them a curveball, because then their RTX line of cards will be exactly a year old and their new cards will be ready for Xmas. My honest opinion is that AMD must get their cards out asap. They have a very good opportunity if it can deliver the performance of the RXT 2060 - RTX 2080 for a lot cheaper. I’m afraid if they wait too long then they will miss this golden opportunity. I hope I’m so wrong with what I just said so we can see AMD get a bit of market share back and Nvidia can lower their prices.

 

@RooiKreef a release in the summer would be sweet although i do not think they do not want too ...for X reasons most likely they can not release em sooner is my best guess ...why ...well i have 0 info other than the logic they do not want to shoot them self's on the foot really

 

They’ve done that so often that some will look for any little chink in the armor to put on blast. Most AMD releases are scrutinized quite heavily. Radeon VII was victim of this. It’s only real fault was it was late to the party but that just started the digging into other bad things that could be said about it.

 

It was kind of glossed over but also in this same earnings report they stated Ryzen is on track for 2nd quarter.

Since navi is coming out in 3rd quarter I wonder if they will be building it on TSMC's 7nm+ ? Considering how long they have had Navi in the works I have a suspicion this is the case so they can eek out another 10% in the same power envelope.

 

from wendel's findings looks like the card is a dream for video editing too ! although Radeon 7 also had a big big big mistake to release with the drivers it did that they fixed .... a week or so later .... at that point postponing the release a week would have been a better option .

 

I'm ready for an upgrade across the board. Installed AC: Odyssey last week and I couldn't get a steady frame rate at 1080p and low settings. My 3570k/GTX 970 setup is just too far behind the curve. I'm ready for an 8c/16t chip and something around a 1080 Ti on the GPU side.

 

It would do them more harm than good to make a paper release out of it. They need reasonable supplies, especially since these are mainstream cards, not enthusiast. 7nm hasn't been ready for real mass production for too long, plus the bloody phone manufacturers are always eating the lion's share. This time AMD also has to get both new CPUs and new GPUs out of that congested TSMC production line everybody and their neighbours are using, with GlobalFoundries out of the picture. All in all I don't think AMD has any choice. They are releasing them as soon as possible, it just isn't very soon due to the realities of the manufacturing capabilities.

 

You are right. On one side I would like to say that I have trouble getting my hands on Snapdragon 855 (~5.6~6.0B? transistors; 73mm^2) with larger storage.
But then there is Kirin 980 (6.9B transistors; 74mm^2) which is available in very large volumes.

=> One of reasons why I do not want Kirin 980 is this transistor density. Apple's A12 (6.9B transistors; 83mm^2) & A12X (10B transistors; 122mm^2) chips have fewer transistors per square mm and same is true about Snapdragon 8180 (8.5B transistors; 112mm^2).
= = = =
This brings another interesting part: Zen+ has 4.8B transistors, but Zen 2 die is around 72nm^2 hinting quite more even while some parts were likely moved into that IO die.

 

There appears to be quite a downturn occurring in new mobile phone uptake, certainly not helped with high saturation levels and £1000+ price tags. Might be an opportunity for some free TSMC space, in time?