AMD Vega Architechture

jjgreenwood · 14 Jun 2016 at 17:12

Did a quick search but couldn't find a thread on this. I'n not overly knowledgeable about the technical details of graphics cards but thought it may be fun to ask the question of those that know more. Apparently Vega is due in October I was wondering if it's possible to estimate the probable performance of the card given what we know of Polaris.

This has been doing the rounds suggesting the specs of the cards:

And this has been posted as suggesting the performance of Polaris:

Given the performance delta of last gen's FuryX vs the older 290X:

Is it possible to make a relatively safe assumption that performance is likely to have a similar delta for the high end Vega part due later in the year.

Presumably AMD will want this part to hit 1080 levels of performance at least, and on the surface of things it appears they will need quite a significant improvement in clock speeds & architecture to get them there.

Seanspeed · 14 Jun 2016 at 17:18

AMD themselves have said Vega is 2017.

Comments about it coming in Oct this year are just rumors so far.

We also are not sure whether there will be one or two Vega GPU's.

meldarthx · 14 Jun 2016 at 17:26

we have a couple sources saying there are two Vega Chips -

http://www.anandtech.com/show/10145/amd-unveils-gpu-architecture-roadmap-after-polaris-comes-vega

Anandtech usually pretty trustworthy on these - but Oct is still a rumor for smaller Vega - I could see possible......

Seanspeed · 14 Jun 2016 at 17:55

meldarthx said:
we have a couple sources saying there are two Vega Chips -

http://www.anandtech.com/show/10145/amd-unveils-gpu-architecture-roadmap-after-polaris-comes-vega

Anandtech usually pretty trustworthy on these - but Oct is still a rumor for smaller Vega - I could see possible......

Cool, hadn't seen that confirmation before.

It does make the most sense. I imagine a baby Vega, even at 300mm^, would easily rival a 1080. And if it's bigger, 350-400mm^, it will outclass it by a decent margin.

Jono8 · 14 Jun 2016 at 18:24

Certainly looking like smaller vega will be competing with the 1070/1080 and the larger vega will be left to compete with the GP102.

Shame AMD wont have anything competitive with the 1070/1080 till October at the earliest though

Orangey · 14 Jun 2016 at 18:27

Vega shines brightly despite being far away.

GFX-Kid · 14 Jun 2016 at 18:35

Orangey said:
Vega shines brightly despite being far away.

Haha. Your comments are usually depressing to read, but this one had the opposite effect on me

r22snapper · 14 Jun 2016 at 18:37

Orangey said:
Vega shines brightly despite being far away.

Jono8 · 14 Jun 2016 at 18:39

Orangey said:
Vega shines brightly despite being far away.

r7slayer · 14 Jun 2016 at 18:44

So does the sun but hopefully vega is not as hot ehh

tavtavtav · 14 Jun 2016 at 18:55

Vega is far brighter than the Sun

Nutella33 · 14 Jun 2016 at 19:42

Orangey said:
Vega shines brightly despite being far away.

Polaris star to follow, leading to vega through the darkness
like sun and moon in life.

Troezar · 14 Jun 2016 at 21:02

Waits for flopper to appear...

JanM · 14 Jun 2016 at 21:09

The Vega 11 enthusiast looks like a beast on paper.

Nutella33 · 14 Jun 2016 at 21:26

Yeah. I hope it doesn't take too long after V10 to arrive on our shelves

Seanspeed · 14 Jun 2016 at 21:29

I hope they dont make Polaris 11 the smaller Polaris and then Vega 11 the bigger Vega.

Knowing AMD's tendency for confusing naming schemes, though...

kalniel · 14 Jun 2016 at 22:20

Seanspeed said:
I hope they dont make Polaris 11 the smaller Polaris and then Vega 11 the bigger Vega.

Knowing AMD's tendency for confusing naming schemes, though...

They already said the polaris number is chronological rather than an indication of performance. The consumer numbers (480 etc.) are the ones that reflect performance.

Seanspeed · 14 Jun 2016 at 22:26

kalniel said:
They already said the polaris number is chronological rather than an indication of performance. The consumer numbers (480 etc.) are the ones that reflect performance.

Cool, thanks.

I find that a bit ridiculous, but eh.

drunkenmaster · 15 Jun 2016 at 00:43

Seanspeed said:
Cool, hadn't seen that confirmation before.

It does make the most sense. I imagine a baby Vega, even at 300mm^, would easily rival a 1080. And if it's bigger, 350-400mm^, it will outclass it by a decent margin.

Based on other sizes I would be very surprised if AMD isn't basically bigger at every price point in die size. Nvidia is what 200mm^2 for GP106(i forget the number, people assuming from the seemingly GP106 shots from the drive px2) while 314mm^2 for GP104. These sizes make sense as does the rumour of 498mm^2 for Gp102. As Polaris is already 120/232mm^2, it would make much more sense for small Vega to be in the 350mm-400mm^2 range and big Vega to be 500-600mm^2 range.

Too small a size difference makes very little sense so I full expect small Vega to be quite a bit faster and bigger than Gp104. Big Vega is certain for HBM2, if smaller Vega has it, look at the performance/W improvement from HBM1 on Fury, HBM2 is another small step in power efficiency over HBM1. If small Vega is HBM2 it will be able to have significantly higher performance/W imho over GP104, if Gp102 doesn't have HBM2 then the bigger bus and clock speeds will take so much power I can't see how it would be competitive with an HBM2 based chip.

Maxwell was for Fury because AMD stuck HBM with an older architecture with very few updates. Vega will be a architecture step and HBM2.

drunkenmaster · 15 Jun 2016 at 00:45

Seanspeed said:
Cool, thanks.

I find that a bit ridiculous, but eh.

In reality 99.999% of end users have no clue what the code name of their cores are, Hawaii, Tahiti, etc. Most users know they have a 280x and nothing else. Nvidia have GP100, 102, 104, 106, smaller number is faster, honestly I've never cared about naming schemes, for end user bigger number = faster, internal names can be whatever you want, it has no effect on your sales, only the 0.003% of enthusiasts who read about it.