RX490 Speculation - Vega 10, not Polaris.

Daniel_Chang said:
EMA isn't easy to do. It's not a switch. Tasks must be divided up between the GPUs present, and those tasks must be completed in sync. This takes a lot of planning, so even if an engine has support baked it, it still leaves a lot on the programmer. If this were easy to do, we'd hear of more games using it.
Click to expand...

Being a programmer, I'll second that assessment. Just because a feature exists does not mean it's easy to use. For the software houses that do plan to use this functionality, you'll see in-house libraries built around baseline interfaces because it fits with the skill sets and design methodologies they use to develop software. Each company that does software dev, gaming or otherwise, builds to the skill sets they have or plan to soon acquire. You may very well see certain companies consistently excel at EMA out of the gate while others will slowly catch up.
 
Zion Halcyon said:
See though, that's contrary to what I heard, and more fits an Implicit Multi Adapter, which is what Crossfire and SLI were, based on my understanding. With the DX12 Adapter being Explicit, my understanding is that DX12 then does more of the legwork for the feature.

In fact, the whole marketing thing behind EMA was that it was easier to implement and smoother than the IMA SLI and Crossfire solutions.
Click to expand...

SLI/Crossfire doesn't require too much from the developer (relatively speaking). Both cards share/work the same assets. Sometimes it's as easy as them taking turns rendering a frame. (Easy, again, RELATIVELY speaking). Explicit multi adapter means that that both cards must be programmed to handle independent tasks. And this must be done in in sync since output from both cards will be simultaneously on screen. DX12 doesn't do the work. It allows for the feature, but the developers must still make it work. I cannot explain it any better than that, unfortunately.

To a non-programmer it may sound easy. This isn't a dig ant you or anyone in particular, but for anyone who is a non-programmer that thinks EMA is easy my recommendation would be to learn how to program and show us how it's done :)
 
Daniel_Chang said:
SLI/Crossfire doesn't require too much from the developer (relatively speaking). Both cards share/work the same assets. Sometimes it's as easy as them taking turns rendering a frame. (Easy, again, RELATIVELY speaking). Explicit multi adapter means that that both cards must be programmed to handle independent tasks. And this must be done in in sync since output from both cards will be simultaneously on screen. DX12 doesn't do the work. It allows for the feature, but the developers must still make it work. I cannot explain it any better than that, unfortunately.

To a non-programmer it may sound easy. This isn't a dig ant you or anyone in particular, but for anyone who is a non-programmer that thinks EMA is easy my recommendation would be to learn how to program and show us how it's done :)
Click to expand...

Oh trust me, that doesn't sound easy at all, and I get it. I actually have a little bit of programming background (nothing super serious, but programming was my original major before I ditched it for the hardware side of things) - it's just that up until now, ever since EMA was first mentioned, how EMA was described vs SLI/Crossfire was exactly the OPPOSITE of what you are saying now for the most part.

I appreciate the clarification, trust me.
 
I think this is the article that several people are talking about:
Microsoft To Provide An Abstraction Layer To Help With mGPU Under DirectX 12

So, they are putting out an abstraction layer to make EMA easier to do in DX12, and then some code to demonstrate how to do EMA with the abstraction layer and then without it.


ANYWAYS, to get the thread back on track -- RX 490 speculation .. here is my shot in the dark .. dual Fiji.

Why? Well, we all have seen that screenshot from the sapphire site, nobody seems to be mentioning it but it does specify that card as having 8GB of ram. (I know it says GDDR5 as well but I have seen FuryX's have 4GD5 and such in their product codes)

So, if it has 8GB of ram .. it is either a dual 4GB card or a single GPU with 8GB. We know there isnt a bigger 14nm GPU than Polaris that exists yet, so it is possibly dual Polaris 10, but if that's the case why isn't it a 16GB card? Surely for crossfire you'd want 8GB on each GPU. And don't give me that DX12 thing where the ram can be pooled instead of duplicated, the cards are rendering the same scene, they will need pretty much the same assets anyways, even if their memory can be addressed individually most of the data will need to be on both GPU's anyways. So, maybe it's a dual fiji .. ? I mean I don't know -- it seems like such an expensive part to make (Fiji GPU) so this seems pretty unlikely ... I guess we will see ... nothing seems to fit right now though...
 
Last edited:
extide said:
I think this is the article that several people are talking about:
Microsoft To Provide An Abstraction Layer To Help With mGPU Under DirectX 12

So, they are putting out an abstraction layer to make EMA easier to do in DX12, and then some code to demonstrate how to do EMA with the abstraction layer and then without it.


ANYWAYS, to get the thread back on track -- RX 490 speculation .. here is my shot in the dark .. dual Fiji.

Why? Well, we all have seen that screenshot from the sapphire site, nobody seems to be mentioning it but it does specify that card as having 8GB of ram. (I know it says GDDR5 as well but I have seen FuryX's have 4GD5 and such in their product codes)

So, if it has 8GB of ram .. it is either a dual 4GB card or a single GPU with 8GB. We know there isnt a bigger 14nm GPU than Polaris that exists yet, so it is possibly dual Polaris 10, but if that's the case why isn't it a 16GB card? Surely for crossfire you'd want 8GB on each GPU. And don't give me that DX12 thing where the ram can be pooled instead of duplicated, the cards are rendering the same scene, they will need pretty much the same assets anyways, even if their memory can be addressed individually most of the data will need to be on both GPU's anyways. So, maybe it's a dual fiji .. ? I mean I don't know -- it seems like such an expensive part to make (Fiji GPU) so this seems pretty unlikely ... I guess we will see ... nothing seems to fit right now though...
Click to expand...
I congratulate you on both creating the possibility of a dual fiji and proving in one post why it will not be a dual fiji. Frankly if I was to guess it will be a polaris 10 chip with full rops and better power management so they can increase the clock speeds.
 
Presbytier said:
I congratulate you on both creating the possibility of a dual fiji and proving in one post why it will not be a dual fiji. Frankly if I was to guess it will be a polaris 10 chip with full rops and better power management so they can increase the clock speeds.
Click to expand...

The RX480 is a full Polaris 10 chip.

My guess is this...

Polaris 11 (little Polaris):

RX450 (Speculated)
RX460

Polaris 10 (FAT Polaris):

RX470
RX480

Vega 11 (little Vega):

RX485 (Speculated)
RX490 (Speculated)

Vega 10 (FAT Vega):

RX495 (Speculation)
RXFURY (Speculated)
 
KazeoHin said:
The RX480 is a full Polaris 10 chip.

My guess is this...

Polaris 11 (little Polaris):

RX450 (Speculated)
RX460

Polaris 10 (FAT Polaris):

RX470
RX480

Vega 11 (little Vega):

RX485 (Speculated)
RX490 (Speculated)

Vega 10 (FAT Vega):

RX495 (Speculation)
RXFURY (Speculated)
Click to expand...
So where is the guess there?

EDIT: Never mind I understand your labeling now you have it as a smaller VEGA.
 
Lisa Su said there will be 3 products for 2016 during the earnings call. I assume that to mean that there will not be a 490.
 
Surprised nothing on Vega has been leaked yet for a part that is close to release. This is suppose to be AMD's high end response to the 1080 and all we get is dust blowing in the wind. WTF AMD try harder if you want to at least keep people from buying from the competition by updating us on why we should wait for Vega. Unless you all rdy know that you have lost this round as well. God I swear...AMD frustrates me to no end. I wanna support them....Im trying real hard, but they keep on disappointing me.
 
Spyhawk said:
Surprised nothing on Vega has been leaked yet for a part that is close to release. This is suppose to be AMD's high end response to the 1080 and all we get is dust blowing in the wind. WTF AMD try harder if you want to at least keep people from buying from the competition by updating us on why we should wait for Vega.
Click to expand...

Didn't we get very little leaks on Nvidia's part this year?
 
Spyhawk said:
Surprised nothing on Vega has been leaked yet for a part that is close to release. This is suppose to be AMD's high end response to the 1080 and all we get is dust blowing in the wind. WTF AMD try harder if you want to at least keep people from buying from the competition by updating us on why we should wait for Vega.
Click to expand...

It's August. The roadmap has it coming out early 2017. It's too soon even for credible leaks.
 
I dunno I just feel that they are not aggressive enough with their high end. Not as much as they were with the 480. Im not feeling the vibe...the hype.
 
Guys, Vega isn't coming any time this year. I'm guessing AMD is sh***ing bricks after seeing Nvidia release something as powerful as the 1080, and THEN releasing the Pascal TiX.

Chances are we will see some MAJOR changes to their roadmap next time they unveil one.
 
Amd needs to be much more agressive than Nvidia and Im not seeing that. If they had something worth mentioning, dont you think they would have said something by now. If we go by the initial roadmap and not the rumors of AMD pulling the launch before the end of the year, that would put the launch in a 6 to 8 month window. Yeah.....maybe 6 to 8 months away is a bit too far for leaks....oh well still I feel they fooked up in allowing Nvidia all that leaway in the high end. Its almost starting to feel like colusion. I dont shit on your high end and you dont on mine.
 
KazeoHin said:
Guys, Vega isn't coming any time this year. I'm guessing AMD is sh***ing bricks after seeing Nvidia release something as powerful as the 1080, and THEN releasing the Pascal TiX.

Chances are we will see some MAJOR changes to their roadmap next time they unveil one.
Click to expand...

I dunno....something needs to change soon, very soon. They cant keep on coming out 2nd in the Halo effect field and expect to gain market share. They need to win some too. Not just 1 in 3 or 4 gens.
 
Spyhawk said:
Would be kewl if they could also one up Nvidia across the board. Not just the usual win some lose most scenarios.
Click to expand...

Unfortunately that is impossible: with Nvidia's cash pool, they can afford to drop prices at a moment's notice.
 
KazeoHin said:
Unfortunately that is impossible: with Nvidia's cash pool, they can afford to drop prices at a moment's notice.
Click to expand...

AMD's lack of cash shows with how little their R&D budget is, they're just keeping their head above water, and their parts are just keeping up with nVidia. I don't think we are going to see that magical 48xx series or 79xx series release anytime soon. That being said I still want to play around with 480 Xfire, would be nice if more AIB cards were out.
 
By the time Vega comes out, it'll be fighting against Volta with full-blown ASYNC. :p
Someone please save Radeon from AMD!
 
IDK. If Volta comes out a year ahead, that is a big if. If's because Nvidia must know something about vega that we don't. Unless they expect vega to be faster than what they have. May be what they suspect it being significantly faster in dx12 and by that time we will have quiet a few titles. Nvidia won't accelerate volta for no reason.
 
NKD said:
IDK. If Volta comes out a year ahead, that is a big if. If's because Nvidia must know something about vega that we don't. Unless they expect vega to be faster than what they have. May be what they suspect it being significantly faster in dx12 and by that time we will have quiet a few titles. Nvidia won't accelerate volta for no reason.
Click to expand...

I personally think Nvidia is starting to operate as if in a vacuum, kind-of like Intel: Intel doesn't give two sh**s about what AMD does in the CPU space, in the same way that a tiger doesn't care about the dealing of a fly.
 
Anyone care to do some divination/set odds on when Vega will be released based on when this thread was started? Going by history, you can expect new AMD cards to be released 6-8 months from the first point when rumors start hitting the feeds and when someone starts a thread here (read: when AMD's outsourced marketing teams start dropping breadcrumbs).

We'll see this card sometime around February next year.
 
harmattan said:
Anyone care to do some divination/set odds on when Vega will be released based on when this thread was started? Going by history, you can expect new AMD cards to be released 6-8 months from the first point when rumors start hitting the feeds and when someone starts a thread here (read: when AMD's outsourced marketing teams start dropping breadcrumbs).

We'll see this card sometime around February next year.
Click to expand...

AMD had leaked it on their website as being eligible for some promotion prior to Dec 31st...

With that and the lack of actual info I'm really hesitant to say anything about when it could be here. No idea. AMD is keeping cards tight this year. I'll probably be able to find something out in September via 'industry contacts'.
 
Spyhawk said:
I dunno....something needs to change soon, very soon. They cant keep on coming out 2nd in the Halo effect field and expect to gain market share. They need to win some too. Not just 1 in 3 or 4 gens.
Click to expand...
Let Nvidia users buy their expensive cards with their superior DX11 performance and hope AMD does not make anything any time soon that competes with Nvidia at the high end so prices keep being stupidly high.
Since AMD shown with RX 480 you can buy 2 and get good scaling in DX12. That Halo effect didn't really help the GTX 1060 in DX12 did it ?

KazeoHin said:
I personally think Nvidia is starting to operate as if in a vacuum, kind-of like Intel: Intel doesn't give two sh**s about what AMD does in the CPU space, in the same way that a tiger doesn't care about the dealing of a fly.
Click to expand...

No one is losing any sleep over what Nvidia is doing with the high end they are milking their customers and AMD is just doing what they are best at price/performance cards.

Most of the reasoning behind Halo effect and competing with high end come from people who want a lower price for their high end Nvidia hardware because when AMD competes they might pay less...
 
Spyhawk said:
Surprised nothing on Vega has been leaked yet for a part that is close to release. This is suppose to be AMD's high end response to the 1080 and all we get is dust blowing in the wind. WTF AMD try harder if you want to at least keep people from buying from the competition by updating us on why we should wait for Vega. Unless you all rdy know that you have lost this round as well. God I swear...AMD frustrates me to no end. I wanna support them....Im trying real hard, but they keep on disappointing me.
Click to expand...

Last time AMD had a high-end card to debut they hyped it so much that it fell into the market like a lead balloon. I don't mind if AMD keeps quiet about what they have coming up. The less hype coming from AMD, the better the product tends to be.
 
  • Like
Reactions: N4CR
like this
Algrim said:
Last time AMD had a high-end card to debut they hyped it so much that it fell into the market like a lead balloon. I don't mind if AMD keeps quiet about what they have coming up. The less hype coming from AMD, the better the product tends to be.
Click to expand...

I'm ok with no hype, so long as there is a product. The main problem with the 300 series was that it came out nearly a year after Maxwell. We're about to see that happen again with Vega. The GTX 1060/70/80/Titan are out now. All that's left is the 1050. AMD has one card out. If Vega isn't due until early 2017, that is 6-9 months for GP104 uncontested.

If Vega is merely on par with those cards, AMD has failed again.
 
SighTurtle said:
That shitstorm? It was literally a dart board with a bunch of drunk people throwing the darts.
Click to expand...

what shitstorm? the first leaked rumor is what at the end turned to be true.. R9 390 performance at R9 380 prices... that's where the rumor started the shitstorm came later with people waiting for a 980TI killer at 300$ and the second part of the overclocker's dream with polaris.. the shitstorm was started by their own users wetdreams instead of the web rumors.
 
Daniel_Chang said:
I'm ok with no hype, so long as there is a product. The main problem with the 300 series was that it came out nearly a year after Maxwell. We're about to see that happen again with Vega. The GTX 1060/70/80/Titan are out now. All that's left is the 1050. AMD has one card out. If Vega isn't due until early 2017, that is 6-9 months for GP104 uncontested.

If Vega is merely on par with those cards, AMD has failed again.
Click to expand...

AMD is behind on their product release schedule due to previous CEOs like Hector Ruiz selling off parts of the company instead of investing into the company. If they catch up to Intel or Nvidia by 2025 it would be a triumph.
 
So, a friend and I were doing rough calculations to see WHAT the supposed 'Vega' could be.

Essentially, in order to compete with the PitanX's performance, AMD would have to produce a chip with these specifications:

6400 GCN Cores
1.3GHz clockspeed
8GB+ of HBM2 OR 8Ghz GDDR5 on a 384bit+ bus

This would be a chip coming in at a whopping 600mm squared, that's a square inch of silicon.
 
KazeoHin said:
So, a friend and I were doing rough calculations to see WHAT the supposed 'Vega' could be.

Essentially, in order to compete with the PitanX's performance, AMD would have to produce a chip with these specifications:

6400 GCN Cores
1.3GHz clockspeed
8GB+ of HBM2 OR 8Ghz GDDR5 on a 384bit+ bus

This would be a chip coming in at a whopping 600mm squared, that's a square inch of silicon.
Click to expand...
tbh they could charge exact same price and still sell them. People spend thousands of dollars on a cpu lol.....shit i think people would buy them. God only knows how much power it would draw!!
 
You must log in or register to reply here.
Back
Top