AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion

"AMD is getting ready to show off the impossible at #SIGGRAPH2017. Sign up here for the spiciest AMD Capsaicin"
They said the exact same thing before Vega FE, referencing Epyc, Threadripper and Ryzen. Maybe some Radeon Pro and Instinct. So yeah, I'd say the impossible is probably one of those.


Experience the art of the impossible in person at AMD Capsaicin SIGGRAPH 2017.

On July 30, learn about our newest additions to the Radeon™ Pro family featuring the cutting-edge Radeon™ “Vega” architecture. Hear from our partners on how AMD’s product lineup – Radeon Pro, Ryzen™ Threadripper™, EPYC™ and Radeon Instinct™ – is revolutionizing the creative pipeline, leading industries from entertainment to engineering into a new era of precision and performance.
https://videocardz.com/70655/amd-announces-capsaicin-siggraph-with-no-word-about-rx-vega
 
Last edited:

The source page:

On July 30, learn about our newest additions to the Radeon™family featuring the cutting-edge Radeon™ “Vega” architecture. Hear from our partners on how AMD’s product lineup is revolutionizing the creative pipeline, leading industries from entertainment to engineering into a new era of precision and performance.

https://www.eventbrite.ca/e/amd-capsaicin-siggraph-2017-tickets-35520337285

And they said they were launch RX a few times:

We can't wait to announce our new Vega products, including RX at this year's #SIGGRAPH - make sure to follow us for more details.


We made a promise to launch RX Vega during SIGGRAPH, Ali. In fact, our CEO made that promise. We’re going to keep it :D


Back on July 1st

Dunno why you are referencing that article when it is super out of date and was flat out wrong.

Finally, During SIGGRAPH 2017 in Los Angeles, USA
We know you’ve been waiting for Radeon RX Vega and we can’t wait to unveil it to you all.

Details on the Radeon RX Vega are coming during SIGGRAPH 2017, so you’ll want to pay attention to what’s happening during this technology summit taking place in the last week of July.

Here’s where you can follow us during our tour and SIGGRAPH for all details on the Radeon RX Vega:

Facebook – @Radeon
Twitter – @Radeon
Instagram – @WeAreRadeon

We’ll be blogging about our meet-ups, so come back to Radeon.com for updates and photos on our tour stops.

https://gaming.radeon.com/en-us/rx-community-meetup/?sf97206526=1
 
Dunno why you are referencing that article when it is super out of date and was flat out wrong.
<ModEdit>
drUZ1VR.jpg

</ModEdit>

The "next post" button at the bottom of that article literally says "AMD confirms Radeon RX Vega is launching at SIGGRAPH". And it was posted the day after.


¯\_(ツ)_/¯




Back to topic:
The next "roadtrip stop" will be tomorrow, this time in Portland.
Maybe our friends in the US will spill some more beans. Or someone will "accidentally" open the AMD tower while no one's looking.
 
Last edited by a moderator:
I think @DavidGraham was merely referenced his bolded part regarding showing the impossible and aligning it to the fact that EPYC and TR will also be part of SIGGRAPH. I don't believe he said anything about RX Vega not being launched.
 
Two HBM2 links, 500GB/s. Plus I think the Greenland codename appears a bunch of times in Vega FE's drivers.
The slightly lower bandwidth offered by HBM2 aside, the most notable shortfall so far for that concept is the 4+ TF count, which in the HPC context would have been double-precision throughput.
Whether Vega physically has enabled IO space for 4 GMI links is unknown.

It seems like that Greenland is what is in the upcoming Vega products, but no 1/2 rate floating point has been indicated. If the supposed leaked slides for future compute cards are accurate, it wouldn't be offered until a 7nm Vega chip with 4 stacks comes out despite Greenland being capable.


And looking at that quad-channel DDR4, it's linked to what we now know as a Threadripper through infinity fabric in a single MCM.
If it ever comes to life, this MCM would be as big as an EPYC CPU.
Perhaps that's what all the spare area in the Vega boards is about... ;)
 
I think @DavidGraham was merely referenced his bolded part regarding showing the impossible and aligning it to the fact that EPYC and TR will also be part of SIGGRAPH. I don't believe he said anything about RX Vega not being launched.
Thank you sincerely, everybody living under a rock knows RX Vega is part of SIGGRAPH by now.
Dunno why you are referencing that article when it is super out of date and was flat
As Malo capably explained, I am making the connection between AMD's "impossible" mantra and their CPU and pro products, a connection they made themselves a month ago.
In fact if you opened the link in that impossible twitter post, you will be directed to Vega's Pro page.
 
The slightly lower bandwidth offered by HBM2 aside, the most notable shortfall so far for that concept is the 4+ TF count, which in the HPC context would have been double-precision throughput.
Whether Vega physically has enabled IO space for 4 GMI links is unknown.

It seems like that Greenland is what is in the upcoming Vega products, but no 1/2 rate floating point has been indicated. If the supposed leaked slides for future compute cards are accurate, it wouldn't be offered until a 7nm Vega chip with 4 stacks comes out despite Greenland being capable.



Perhaps that's what all the spare area in the Vega boards is about... ;)
Considering what AMD revealed in the FAD, I imagine 1/2DP Vega will actually be on 14nm+.
Most likely 7nm was realized as too aggressive a goal for Vega 20, and it was moved to 14nm+, while Navi turned into the first 7nm GPU planned.
 
Considering what AMD revealed in the FAD, I imagine 1/2DP Vega will actually be on 14nm+.
Most likely 7nm was realized as too aggressive a goal for Vega 20, and it was moved to 14nm+, while Navi turned into the first 7nm GPU planned.

It wouldn't mesh with Greenland having 1/2 DP and being the current chip unless the 14nm+ product is a bogus marketing distinction. The latter could be plausible since AMD frequently talks up later steppings of the same chip as a new product, although it would be different in that this "new" product would actually have something different to offer.
 
Considering what AMD revealed in the FAD, I imagine 1/2DP Vega will actually be on 14nm+.
Most likely 7nm was realized as too aggressive a goal for Vega 20, and it was moved to 14nm+, while Navi turned into the first 7nm GPU planned.
So if AMD jump the 14++ to the 7 then Vega 20 for 2018 and Navi for 2019 if not then navi will be 14++ in 2019 or in 2020 on 7nm.
 
So if AMD jump the 14++ to the 7 then Vega 20 for 2018 and Navi for 2019 if not then navi will be 14++ in 2019 or in 2020 on 7nm.
There is no 14nm++ in GF. First products from GF at 7nm will start mass production in late 2018 and products to the market will depend on yields and how late that date is. So most probably expect big products using that node in early 2019. So if everything goes well 14nm+ for 2018, 7nm 2019 and 7nm+ 2020.
 
There is no 14nm++ in GF. First products from GF at 7nm will start mass production in late 2018 and products to the market will depend on yields and how late that date is. So most probably expect big products using that node in early 2019. So if everything goes well 14nm+ for 2018, 7nm 2019 and 7nm+ 2020.
Oh yeah, I confuse the presentation about 7nm to the 14.
 
It wouldn't mesh with Greenland having 1/2 DP and being the current chip unless the 14nm+ product is a bogus marketing distinction. The latter could be plausible since AMD frequently talks up later steppings of the same chip as a new product, although it would be different in that this "new" product would actually have something different to offer.
I mean, having 1/2 DP would explain the size and performance. All the extra transistors went for that. But then why not offer a 1/2 DP card to the HPC market? Hawaii was the last one AMD released, and that's simply not good enough.

I am almost certain that Vega 10 is 1/16 DP. As for Vega 20, the initial leak from Videocardz about Vega in general said that it would be 1/2 DP, so I believe it will be that. But I'm not so certain since the node changed too, so who knows what else changed.
 
AMD clarifies it's position on Vega FE drivers, basically Vega FE has pro drivers that have all the gaming optimizations released up to the point of the pro driver release. The switch to gaming mode is meant to for just the UI indeed. So as Rys said the driver is just old but not gimped.

Typically, and straight from the mouth of AMD, pro drivers do not update as frequently as consumer drivers. When the updates happen, the pro drivers will receive all the same game-level optimizations as the consumer drivers. Let’s say you’re running the pro drivers and Battlefield 6 launches: Those pro drivers don’t need an update (the professional applications haven’t changed), and so don’t receive one on game launch. The drivers are unoptimized for this fictional game. Consumer drivers have launched, though, to support the new title.

As for later gaming optimizations, AMD wished to exploit the switch to enable users to integrate latest optimizations from the consumer branch without the need to update the whole pro driver.

AMD hopes to leverage its toggle switch to permit the user to fetch those consumer drivers for playing the latest games. At a later date, maybe when AutoCAD updates, the professional drivers will eventually launch and include the game optimizations.

And:
For Vega FE, we anticipate no greater than 3% performance uplift for most games with 17.7.2 at this time, based on our conversations with AIB partners. Some specific applications, like 3DMark 11, will see uplift closer to 10%. This is information from GamersNexus contacts and not from AMD officially.
http://www.gamersnexus.net/news-pc/3000-amd-gaming-pro-mode-and-other-driver-updates
 
What happened to the "Dont compare RX vega with FE because they are completely different product!. RX will have 40% more performance!" and all that magical hope.
 
What happened to the "Dont compare RX vega with FE because they are completely different product!. RX will have 40% more performance!" and all that magical hope.
I'm not sure where any of that came from. I think you're just making up stuff. There was certainly some basic discussion about whether or not FE and RX were different in any way other than the HBM2 stacks and speculation regarding certain advertised features working/not working but I don't think anyone here said anything about massive performance deltas between the 2 products.
 
What happened to the "Dont compare RX vega with FE because they are completely different product!. RX will have 40% more performance!" and all that magical hope.
AMD never said anything about how much more performance there would be with RX.
But regardless, AMD has to hold some sort of rabbit in the hat, they can't be stupid enough to think any consumer would buy Vega at ~1000€ if it performs like ~550€ card
 
Back
Top