Quote Originally Posted by BenchZowner View Post
We shall see...

Now our issue is the release date & quantity, performance is something we don't talk about anymore is it ?

Don't get lost
Release date isn't really an issue for those that don't immediately jump into a worst case scenario...
Same with quantity, the ~10k units at launch has been restated on almost every single high-end launch since Fermi. It is nice that they changed it from 15k to 16k for this launch.
Performance has never really been questioned if you have been actually reading the solid info we have on Vega. Worst case is ~10% slower than 1080Ti FE (going by TPU charts).

Quote Originally Posted by Darakian View Post
If they're supply constrained then limiting vega to high margin areas makes sense. However note the background render at this timestamp
https://youtu.be/590h3XIUfHg?t=4401
Vega logo on an HBM-less package. Maybe it's just art, but maybe we also have gddr variants coming.

https://i.imgur.com/fWkIZqN.png
Render seems more or less accurate to me. I believe they will be using an IHS due to the different heights of the 4hi and 8hi HBM2 stacks.
I don't see any GDDR ICs on the PCB in that render.

Quote Originally Posted by Tim View Post
What the hell is going on with consumer Vega?! Still no news.

Ugh just stop it man, you're worse then AMD. Every time with these 'I know more then you but I'm not saying anything' rubbish. Even Raja can't top your nonsense.
What is your problem?
An expanding high margin market that should double its revenue in the next 2-3 years is more geared to Financial Analyst Day than AMD trying to claw back some of the highend enthusiast desktop market.
Maybe there is some sort of consumer tradeshow coming up where they have scheduled a press event?

Quote Originally Posted by zanzabar View Post
no one is shipping bulk hbm2. samsung has some test batches they shipped out for testing recently, but micron and hynx have not shipped test batches recently and no one has a schedule for bulk. amd needs a gp102 equivalent so they can stick gddr on a part made for hbm, but since they went all in on hbm with nothing gddr on the same architecture i dont see it happening.
Samsung went into mass production of HBM2 in Jan '16. Hynix said they weren't more than a quarter behind at that time due to bringing up a new line.
Supposedly both are fully ramped as of Q1 '17, so I'm not sure where you are getting this.