So, if Vega 11 use HBM2 what will Vega 10 use, gddrx? AMD didnt specified anything when it comes to Vega.

This will be the 2nd new architecture in the last 3 years. Polaris and vega both amd Great news

Not really a new architecture. It's GCN1.3 or GCN 2.0 like Polaris AFAIK. The only difference is Vega has HBM and larger core count. 300W incoming but I don't care as long as it performs at a reasonable price.

So, if Vega 11 use HBM2 what will Vega 10 use, gddrx? AMD didnt specified anything when it comes to Vega.

All the road maps indicate all Vega GPU's getting HBM2.

So in two months time we will have first HBM2 cards? How can one not hype this? 😀

new rasterizer or GTFO.

Not really a new architecture. It's GCN1.3 or GCN 2.0 like Polaris AFAIK. The only difference is Vega has HBM and larger core count. 300W incoming but I don't care as long as it performs at a reasonable price. All the road maps indicate all Vega GPU's getting HBM2.

No idea how much graphics ip matters to AMD but Vega will be 9.0 while Fiji and Polaris were 8.0. And Hawaii was 7.3. I would guess a lot of other things are changed for Vega also vs what was for polaris. To me it all looks like it's a bigger change then Polaris was as a whole.

So in two months time we will have first HBM2 cards? How can one not hype this? 😀

Didn't Nvidia already release a Pro HBM2 card in the quadro line based on GP100? This would be first consumer level HBM2 though. However I would cool your jets. Let's not repeat the Hawaii or Fiji launch. Just let the card come out and sell itself the less hype the better.

No idea how much graphics ip matters to AMD but Vega will be 9.0 while Fiji and Polaris were 8.0. And Hawaii was 7.3. I would guess a lot of other things are changed for Vega also vs what was for polaris. To me it all looks like it's a bigger change then Polaris was as a whole.

Hadn't seen that. So Fiji and Polaris ar the same graphics ip. Wasn't Fiji GCN1.2 like Tonga?

You so sure 490= Vega? The way AMD was marketing the 480 CF > than 1080. I can see if it happens this yr it would be a 480 X2 card. Vega seems still far out to get to launch if all rumors are right.

So in two months time we will have first HBM2 cards? How can one not hype this? 😀

HBM2 is good, obviously, but I personally don't see why people think it's going to bring some massive improvement. With HBM1 on the Fury X, it helped AMD a lot because die size was near maxed and the GDDR5 controller is massive. Plus GDDR5 @ 28nm uses considerably more power than HBM does. GDDR5x @ 16/14nm though is much closer in power to HBM and these chips are hitting power constraints for their particular "threshold" before they are hitting mm2 limits. For example, a chip the size of a Fury X with GDDR5 would probably be significantly bigger than the ~600mm2 reticle limit on 28nm. HBM made that chip manufacturable. 14/16nm though, the Titan X (P) is only ~480mm2 but it's already at 250w. Generally AMD/Nvidia don't like to go over 300w. So aside from massive bandwidth, HBM2 isn't going to be helping much in the other two ways it helped with the Fury X. Unless of course AMD goes nuts and just builds a massive "**** you" chip and blows past the 300w limit. Which would be awesome. As for that bandwidth, no game, even at 4K is really going use more than what GDDR5x can already offer with a 384bit bus. It's great for compute, especially now with native error checking, which I'm sure AMD will share the Vega design with it's Pro cards, but as a gamer there really isn't anything about HBM that excites me. Like it's nice that Vega 11 and possibly 10 is getting it, but if either card shipped with GDD5x, I don't think it would make a difference to me as a gamer.

Damn the meth heads are gonna have to find a new place to get high if that's the launch venue.:toke:

What I'm curious about is how they're going to name all these cards. If they're doing away with the X suffix then that just leaves them with the RX 490 and possibly the Fury. And I assume there will be more than two cards in Vega.

Didn't Nvidia already release a Pro HBM2 card in the quadro line based on GP100? This would be first consumer level HBM2 though. However I would cool your jets. Let's not repeat the Hawaii or Fiji launch. Just let the card come out and sell itself the less hype the better. Hadn't seen that. So Fiji and Polaris ar the same graphics ip. Wasn't Fiji GCN1.2 like Tonga?

Yeah Fiji was 1.2 like Tonga. That bit baffles me a bit because of all that I've seen so far. I think it will be interesting to see what Vega brings to the table. I would think support for more dx12 features among other things like hbm2 maybe I think possibly gddr5x but it won't be gddr5 for sure.

So realistically, where do you guys think the 490 will end up? Between the 1070 and 1080 @ $499? I don't see them reaching the 1080 in performance, no matter how much they want to charge...

I love how the whole thread is "Don't hype up the card", "AMD dropped the ball", "AMD never promised anything with the 480", and half the thread is "ddrx?" but completely glossing over the one piece of early information we got from AMD (which is naturally one of only a few pictures in the OP) that Vega would be HBM2. So instead of referring to the literal one thing we know about Vega, let's just keep wildly speculating and going through the exact same problem of the 480 launch. Yeah, that'll work out great for everyone.

I don't see them reaching the 1080 in performance, no matter how much they want to charge...

They would reach the 1080 in performance if they could make the GPU clock higher. Based on Polaris, I wouldn't bet on that at all, though. I reckon it will have very moderate clocks and thus be behind the 1080.

So realistically, where do you guys think the 490 will end up? Between the 1070 and 1080 @ $499? I don't see them reaching the 1080 in performance, no matter how much they want to charge...

Depends on what the 490 is As for Vega, I imagine that Vega 11 will be 10-15% faster then a 1080 and Vega 10 will be 10-15% faster then a 1070. You figure they generally are going to want to stay in the ~300w range. Two RX470's is roughly equal to 300w and gives you the rumored 4096 cores. HBM will remove some power, maybe allow a slightly higher clock + small improvements to GCN + memory bandwidth bottlenecks removed. As to what Vega 11/Vega 10 will be named, I have no idea. Maybe we won't even get a 490 and they'll just do a different naming scheme for Vega.

So aside from massive bandwidth, HBM2 isn't going to be helping much in the other two ways it helped with the Fury X. Unless of course AMD goes nuts and just builds a massive "**** you" chip and blows past the 300w limit. Which would be awesome.

I really think this is what AMD should do, build a massive single GPU powerhouse with HBM2, chevy small block style, screw efficiency.

I really think this is what AMD should do, build a massive single GPU powerhouse with HBM2, chevy small block style, screw efficiency.

I would like to see that honestly.

AMD + Rushing = I have a bad feeling about this based on my past experience with AMD.

Don't think they are rushing. With all Polaris/Vega thing it seemed more like AMD had to wait for the right technologies to be properly developed, before it could start selling them (14nm, now HBM2). Everything else was already ready.

new rasterizer or GTFO.

Wasn't that in Polaris? Or at least vastly improved? The 480 does go against 64ROP cards that are running at comparable speeds.

So realistically, where do you guys think the 490 will end up? Between the 1070 and 1080 @ $499? I don't see them reaching the 1080 in performance, no matter how much they want to charge...

I don't believe that they will manage to have enough HBM 2.0. On the other hand, they hold critical patents for HBM, that can give them leverage for having some in the card. The recent professional card with the SSD made me wonder if they will have 4GB of HBM 1.0, and another 4-8GB of GDDR5 on board. It certainly seems that they know how to do mixed memory configurations even with much slower things like SSDs involved. Because of the Furies the driver also seems to have some kind of generic algorithm to handle mixed speed memory, even dealing with much slower system memory, GDDR5 on the board should be an easy task. I don't want to hype this, but architecturally speaking Polaris was quite impressive in some changes, and that was going from IP 8.0 to IP 8.1. Vega is IP 9.0. I'm on the wait and see camp, but I believe that NVIDIA launched the Titan X so early because of Vega, and not despite it. We'll see.

mm I personally dont care if AMD stays away from the top 1% of the market because it makes the top end cards much more expensive. The reason i dont mind is because i dont buy them and it will hurt the people who do so when i see people bragging they have a 1080 ti its not envy i feel but pity and ridicule.:banana: