• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD 7nm GPU News and Rumours 2018/2019

IF (and it's a big If) Navi is still GCN, then it is the final iteration...

When was it ever not GCN? All the AMD roadmaps put "Next Gen" in 2020, with Navi listed as 2019. Have you seen something to the contrary? The "next gen" memory bit for Navi on the roadmaps would indicate GDDR6 if that still holds true. And I also can't see Navi not beating Vega.

The speculation continues...

AMD also need some low-end parts...

A full RX 600 series based on Polaris 30 would fill that need, but then how much more grunt can they squeeze out to top the RX 590, unless they strap on GDDR6 too. Polaris 30 can't just be a single product to satisfy GloFo, surely they must do something else with it?

Can it be H2 2019 yet? I wanna know what's going on :D
 
Navi is officially "targeting the mid-range" so where would Polaris 30 even fit in? A full 600 series below the mid-range? Doesn't make any sense.

It makes perfect sense if you look at the broader landscape, not just AMD's little corner of the world.

I've said this before: the very phrase "mid-range" alludes to there being a "bottom range" and a "top range". And consider the performance spread of the gaming GPU market, not just AMD's products. If you consider the RTX kit's performance to be the "top range" then that will put the GTX 1080 and GTX 1080 Ti into the "mid-range" performance bracket. Target Navi at that performance level and lo and behold, Navi is the mid-range product. And by extension then, the bottom range products are populated by Polaris 30.

Just because Polaris right now is AMD's mid-range product doesn't mean it will or should remain AMD's mid-range product.
 
When was it ever not GCN? All the AMD roadmaps put "Next Gen" in 2020, with Navi listed as 2019. Have you seen something to the contrary?

Every other week it seems to flip/flop between being completely new, or being an improved GCN. Even googling "is Navi GCN" brings a whole load of mixed articles from different times, all with different opinions :D

Polaris 30 can't just be a single product to satisfy GloFo, surely they must do something else with it?
I don't think so - whilst it is a "GloFo satisfaction project", it also seems to be a placeholder with Vega if I understand being EOL very soon, and possibly a case that RX570/580 is as well.

I think the separate branding e.g. RX5xx and VegaXX branding will disappear - it's a legacy of Raja (both Fury&Vega parts). Polaris30 may well find it's way into the RX600 range, but I can't see it scaling down any further e.g. to a "esports" part
 
The only thing that will make Navi worthwhile is if 7nm is as good as it promises to be. If they can get equivalents of Vega 56 & 64 but using half the power whilst also being cheaper to produce (due to not having the extra compute stuff, using GDDR6 instead of HBM2, and a smaller die), then they'll have a good product to hold them over until next-gen. I'm just not convinced the stars will align in that way, especially given their current focus on the CPU market.
 
Every other week it seems to flip/flop between being completely new, or being an improved GCN. Even googling "is Navi GCN" brings a whole load of mixed articles from different times, all with different opinions :D

Ain't that the truth! I did the same thing to check I wasn't talking out of my rear and got exactly the same.

If they can get equivalents of Vega 56 & 64 but using half the power whilst also being cheaper to produce...

Assuming Navi is 6th gen GCN, I see no reason for it to end up being worse than Vega, unless some major ball-dropping occurs. Excluding the new compute stuff added to Vega 20, the 7nm shrink alone gave it a performance boost. Even leaving in Vega 10's compute stuff we'll get a smaller die, and I'd like to think AMD would be more particular with their binning and testing so the utter dog chips that need a bazillion volts won't ruin the baseline power requirements for the entire range (like we saw with Vega). So that gives us better power and thermals.

All-in, it'll be superior to Vega on that alone.

The biggie is going to be HBM: you cannae make cheap cards with that stuff welded to a die. The roadmap slides have "next gen memory" tacked next to Navi and David Wang has said they'd be looking at both HBM and GDDR6 for products. But I don't think they have the cash to develop Navi with HBM and GDDR6 controllers, so it's probably gonna be HBM. Unless Navi is never going to be a compute product so development shifted to a GDDR6 controller.

I think we all know what we'd like to see, but it's probably not going to happen because GCN is still a thing for one last hurrah and AMD still don't have a lot of cash.

Like I said, is it H2 2019 yet?
 
Every other week it seems to flip/flop between being completely new, or being an improved GCN. Even googling "is Navi GCN" brings a whole load of mixed articles from different times, all with different opinions :D

It is possible that the rumours talk about Navi & Arcturus - Navi could be GCN 6.0, while Arcturus Next-gen architecture 1.0.
 
It makes perfect sense if you look at the broader landscape, not just AMD's little corner of the world.

I've said this before: the very phrase "mid-range" alludes to there being a "bottom range" and a "top range". And consider the performance spread of the gaming GPU market, not just AMD's products. If you consider the RTX kit's performance to be the "top range" then that will put the GTX 1080 and GTX 1080 Ti into the "mid-range" performance bracket. Target Navi at that performance level and lo and behold, Navi is the mid-range product. And by extension then, the bottom range products are populated by Polaris 30.

Just because Polaris right now is AMD's mid-range product doesn't mean it will or should remain AMD's mid-range product.
If you re-use Polaris at all perf points below 1080 level then you are basically re-re-re-releasing the 480 (again).

That would be monumentally stupid, when you have a Navi architecture which represents two generations of GCN evolution, and can scale downwards.

The reason Vega didn't get cut down to fill the mid-range of the 500 series was due to HBM (expensive). Not only does Navi not use expensive interposers and HBM, we think, but it will also incorporate lessons learned from Vega and all the further refinements since then, developed for PS5.

Re-using Polaris doesn't make sense whichever way you look at it.

And I very much doubt they are going to have both Navi and Polaris at 7nm. It's double the effort when Navi is all you need.

Also Navi was specifically developed for 7nm, so you'd imagine it will make a lot better use of the particulars of that node than Polaris could. "12nm" on the other hand is basically still "14nm".
 
If you re-use Polaris at all perf points below 1080 level then you are basically re-re-re-releasing the 480 (again).

That would be monumentally stupid, when you have a Navi architecture which represents two generations of GCN evolution, and can scale downwards.

The reason Vega didn't get cut down to fill the mid-range of the 500 series was due to HBM (expensive). Not only does Navi not use expensive interposers and HBM, we think, but it will also incorporate lessons learned from Vega and all the further refinements since then, developed for PS5.

Re-using Polaris doesn't make sense whichever way you look at it.

And I very much doubt they are going to have both Navi and Polaris at 7nm. It's double the effort when Navi is all you need.

Also Navi was specifically developed for 7nm, so you'd imagine it will make a lot better use of the particulars of that node than Polaris could. "12nm" on the other hand is basically still "14nm".

Yeah, they could have, at the very least, tried to at least update the multimedia engine with hardware acceleration of newer video formats.
 
@FoxEye, as you rightly imply, this entire thing will depend on Navi's memory type. If AMD stick with HBM then it'll be a bust because you just can't make HBM-based cards cheaply enough to hit a mid-range price bracket. If Navi comes with GDDR6 then we're onto a winner potentially with fast, small dies that sip power and are cheap to produce, and yes the Vega lessons will (should) be learned.

I'm not quite sure why you think continuing Polaris 30 is "monumentally stupid" though, and nobody ever said anything about Polaris on 7nm. Sticking with the assumption that Navi does get strapped with GDDR6 AND it can be chopped down sufficiently to hit the budget market, then yes there's no need for Polaris 30. However, if Navi can't be cut down that far and/or is equipped with HBM, are you suggesting AMD just abandon the sub £250 market?

GloFo 12nm isn't "just" 14nm because there is some performance gain to be had - the RX 590 is, what, 15% faster than the 580? That means AMD could potentially refresh the entire RX 500 line with 10-15% boost across the board and populate the sub £250 market. And if it's a simple transition to GDDR6 as Micron suggest, the IMC could be updated too and bring a further performance uplift.

So this is the deal: if Navi can scale all the way down to a £100 card then we can finally lay Polaris to rest. If Navi can't scale or AMD won't scale Navi down below £250 then Polaris 30 is still be needed for another year or so until Arcturus lands.
 
Polaris at £250 would be utter fail. Because we know Polaris tops out at 580 perf +/- a few %.

You can't keep the same perf at the £250 price point for 5 generations.

If they re-use Polaris for the 600 series then the 580 better become the 660 and move down to £150. If they do that they could re-use Polaris and, although an old, old, old (power-hungry) card, moving it sufficiently down the price points would be just about acceptable.

A 580 rebranded as the 680 and still occupying the £250 price point? Utter fail. Monumentally stupid. I stand by that.[/user]
 
Polaris at £250 would be utter fail. Because we know Polaris tops out at 580 perf +/- a few %.

You can't keep the same perf at the £250 price point for 5 generations.

If they re-use Polaris for the 600 series then the 580 better become the 660 and move down to £150. If they do that they could re-use Polaris and, although an old, old, old (power-hungry) card, moving it sufficiently down the price points would be just about acceptable.

A 580 rebranded as the 680 and still occupying the £250 price point? Utter fail. Monumentally stupid. I stand by that.[/user]

Stand by it then, because I'm not sure you're understanding what I'm saying or just being a little belligerent because you think Polaris should just vanish. Answer this instead:

If Navi is not cheaper than £250, what do AMD sell?
 
Stand by it then, because I'm not sure you're understanding what I'm saying or just being a little belligerent because you think Polaris should just vanish. Answer this instead:

If Navi is not cheaper than £250, what do AMD sell?
In the post you just quoted I said Polaris 580 perf would be OK at £150. So they could sell top-end Polaris at £150 and that wouldn't be too terrible (despite being very old and power-hungry, which might not be a great situation for low-end rigs with cheap PSUs).

If Navi isn't cheaper than £250 then they have nothing between £150 and £250.

But answer me this... do you think Polaris can occupy price points up to £250 and people won't be ridiculing AMD for it?
 
Next year Nvidia will have the GTX2060/GTX2060TI,and I am sure the RTX2070 will probably drop down in price a bit in USD,once the GTX1060 and GTX1070 are EOL. The GTX2060/2060TI will probably be GTX1070 level or a tad faster.
 
I might be wrong but where can I find this official announcement? Ask D.P also other day no reply.
I think you might be right it's based on info/rumours from two separate sources, but no official AMD announcement.

However it makes complete sense for them to target Navi to GTX 1080 perf (which will for sure be "mid-range" in late 2019). For one we know Navi is being designed very much for PS5. Power consumption will matter every bit as much as performance. PS5 won't be able to have a 200W GPU (PS4 uses ~140W full tilt).

Second 1080 perf for the masses would be a huge, huge win for AMD. Atm if you want that perf it will cost you £500+. If AMD can bring this perf to the £300 ish price bracket, they will have a runaway success on their hands.

Lastly we know from AMD's own slides that 7nm can bring their GPUs 1.25x perf at the same power, if they really want to release the fastest GPU they can. However it will still be slower than nV's best cards and will still use more power. Focusing on delivering 1080 perf at the lower power possible not only fits in with their PS5 partnership, but also will give them millions of sales. It's win/win really. It just makes so much sense.

e: It's worth pointing out that the info isn't that Navi won't perform better than a 1080. It's that 2019 Navi will be at that level, and "bigger/biggest" Navi will be a 2020 product.

To my mind that either means Little Navi is a late 2019 product, or that big Navi really is 6-12 months behind Little Navi in development.
 
Last edited:
But answer me this... do you think Polaris can occupy price points up to £250 and people won't be ridiculing AMD for it?

I don't give 2 hoots if AMD are ridiculed for it, they've had nothing but grief since Vega showed up and blew everybody's minds with its sheer mediocrity. And that has zero to do with the discussion at hand. So yeah, I can see Polaris 30 existing between £150-£250; chopping £40 off current prices would put the RX 590 in around £230 and give the GTX 1060 a good kicking. Doesn't mean I think it should but that doesn't mean it won't; my personal desires have zero to do with how things will shape up, I'm theorising based on logic and pragmatism.

And since I'm theorising, here's something to suggest that Navi will be GDDR6-based AND fully scalable: Polaris 30 exists as a single product in the RX 500 series.

If AMD knew that Navi couldn't get chopped down sufficiently to fit the budget market, or the use of HBM2 would be prohibitive to low-end budgets, then that's another 2 years before a fully scalable architecture is available. However old and hungry it is, the Polaris 30 respin could have incorporated a GDDR6 memory controller and maybe trade the clock speed boost for a couple more shader units or something. That would see a chunky uplift in performance across a full RX 600 range, targeting the markets from GTX 1070 and down.

But that hasn't happened. Are AMD just going to keep the RX 500s limping along until Arcturus? Or will Navi fill the entire gamer product stack?
 
Polaris 30 looks to also be made at Samsung:

https://www.techpowerup.com/249669/samsung-amds-second-foundry-partner-for-polaris-30

So it appears AMD is trying to free up 14NM capacity?? It might make sense since the IO chips for Ryzen 2 are made on 14NM,and Ryzen+ is already made on GF 12NM.

I do wonder if Navi could also go the chiplet way,ie,some of the logic like the memory controllers,and other blocks made on 14NM or 12NM,and shader blocks on 7NM??
 
Last edited:
Polaris at £250 would be utter fail. Because we know Polaris tops out at 580 perf +/- a few %.

You can't keep the same perf at the £250 price point for 5 generations.

If they re-use Polaris for the 600 series then the 580 better become the 660 and move down to £150. If they do that they could re-use Polaris and, although an old, old, old (power-hungry) card, moving it sufficiently down the price points would be just about acceptable.

A 580 rebranded as the 680 and still occupying the £250 price point? Utter fail. Monumentally stupid. I stand by that.[/user]

Well AMD seem to have done a lot better than few percent, but AMD could easily hold with a Polaris based card while Nvidia are trying to get thier act together.
 
Back
Top Bottom