AMD: Navi Speculation, Rumours and Discussion [2017-2018]

Status
Not open for further replies.
I'd love for this to be a role reversal where previously ATI/AMD have talked up a feature only for Nvidia to release a design that makes AMDs version look second rate.

I'm thinking of the unified shaders of R600 and G80, where Nvidia were rather quiet on whether G80 would have unified shaders until they dropped 8800 GTX, and ATI talked quite a lot about them.
Tessellation is a more recent example - AMD had it in hardware from R600 all they way up to Evergreen, but Nvidia came out with Fermi and took the lead.

It could also just be Nvidia doing the same as the examples above, except this time AMD have barely shown up to the party. Lets hope not.
 
Hasn't AMD mentioned that Navi was going to be their first major GFX revision since GCN was introduced? IE - marking a significant move away from the derivative GCN architectures that we've seen since the 7970?

It's certainly possible that AMD knew about Microsofts plans for DXR for quite some time now. Whether Microsoft has been hitting up NV, AMD, and possibly Intel on whether they are capable of hardware that can accelerate DXR/RT back far enough for it to have influenced Navi's design is a key question that no one outside of Microsoft, NV, AMD and maybe Intel would know.

The safe assumption would be that AMD doesn't have dedicated hardware to specifically accelerate DXR in Navi. But the possibility is certainly there that there "might" be or that they'll have hardware that might be able to partially accelerate some features in DXR.

But until AMD announce something, I'm not going to assume that they have anything specifically targeting DXR in Navi.

In other words, it's possible that NV's design for the RT cores was influenced by Microsoft hitting up all the major players to see what was possible. If so, then AMD likely had a similar amount of time to try to implement something, albeit with far less capital resources to dedicate to R&D.

Regards,
SB
 
Hasn't AMD mentioned that Navi was going to be their first major GFX revision since GCN was introduced? IE - marking a significant move away from the derivative GCN architectures that we've seen since the 7970?
I'm pretty sure they haven't said anything like that, just that it has new ISA (gfx10) which doesn't really say much, Vega had new ISA too.
 
The wildcard for Navi is what market AMD is primarily aiming it at. The oldest architecture in AMD's current stack isn't 2017's Vega, it's 2016's Polaris. Which itself was so heavily derived from GCN3 that the shader ISAs are identical. AMD may not intend on replacing "youthful" Vega in 2019, given that it will have taken them 3 years just to replace Polaris. It's going to depend very heavily on the conditions that led to them launching two half-stacks rather than one whole stack, and if those conditions are still in play for 2019.

(And keep in mind that Vega 20 isn't even out yet)
 
Well that's the problem if Navi is to replace Polaris then it is going to be expected to be faster than it. But the performance differential between a RX 580 and Vega doesn't leave a tremendous amount of room for improvement without eating into Vega's performance segment.

So would they want to continue making these larger more expensive products when their own mid range parts are cannibalizing their sales I dunno.
 
It might be similar situation as with Fiji and Polaris. So we might see Navi in 100-300$ market, and Vega will be there if someone needs it :) Though since we know Vega20 will be ready for HPC in next few months, i would expect entire "7nm Vega" lineup before Navi
 
The wildcard for Navi is what market AMD is primarily aiming it at. The oldest architecture in AMD's current stack isn't 2017's Vega, it's 2016's Polaris. Which itself was so heavily derived from GCN3 that the shader ISAs are identical. AMD may not intend on replacing "youthful" Vega in 2019, given that it will have taken them 3 years just to replace Polaris. It's going to depend very heavily on the conditions that led to them launching two half-stacks rather than one whole stack, and if those conditions are still in play for 2019.

(And keep in mind that Vega 20 isn't even out yet)


Vega 20 will be oonly a pro/research card, no ?

Even if Navi is targeting mid range, mid range by mid/end of 2019 will be a nvidia chip maybe faster than curent Vega10...

I real feel like they're to far behind nVidia to catch them, and if nothing drastic happens, well... Maybe amd can sell rtg to someone who care about it...
 
I think they do care about RTG as far as making half-decent IGPs and console GPUs go

Yeah but do they care more than that ? Like, eh that's why most of us are interested in gpu, innovation and fighting in the high end niche too ? We'll see next year I guess.
 
Well, the Navi lineup was originally targeted/developed as a replacement for both Vega 10 and Polaris. The Navi 10 was to replace Vega 10 and Navi 11 to replace the unfortunate Vega 11. Originally AMD wanted to launch the Polaris replacement as mid-range Vega 11 in 2017 along with Vega 10 and Vega 10x2.

The question is, what out of Navi plans survived to this day. It feels like none would stand a chance vs 2019/2020 nV models. Given the current RTX line being already 40-80% faster than Vega and Navi will most probably face the nV's 7nm counterparts in H2 2019/H1 2020...

Following the path of Fiji and Vega 10 => Navi 10 ~ 4 SE (4096 x 256 x 64), 2+GHz, IPC +5%, HBM3, 300W. Meh
 
I'm pretty sure they haven't said anything like that, just that it has new ISA (gfx10) which doesn't really say much, Vega had new ISA too.
there is talk navi not being gcn based.
https://wccftech.com/exclusive-amd-navi-gpu-roadmap-cost-zen/
whether their right or wrong is another matter, i wouldn't bet house their right personally.
If it is the case then what is 'next gen', that sounds like the one that shouldn't gcn based to me.
 
Wasn't Navi supposed to be the "Next-Gen" arch, completely different from GCN? I seem to remember so from slides back in early-mid 2016. Did something change?
 
yes you're right. I wonder how much they gain from that, vs R&D needed...

I think consoles at least is also an insurance of sorts. When they have a competitive architecture (a la Zen) around, prolly consoles (and semi-custom) margins aren't too great comparatively.
But when they don't at least it's something that they can reliably sell. So they'd want to keep RTG arround even if a cost/benefit analysis on the mid term would imply they're better off without. If they can
 
I strongly believe the Navi-ISA will be more different from the Vega-ISA than the Vega-ISA was from the Polaris-ISA. But I'm not sure yet if it is a GCN Generation 6, or something else. I'm pretty sure they already taped out some first Navi10 silicon, but so far they have been sparse on the open source driver patches. I highly suspect we will see something in the next quarter, if they want to release something in mid-2019. Currently there are much more questions than there are answers. Navi10 could be anything for me right now, I'm only pretty sure the previous PS5 rumors are totally bogus.
 
Status
Not open for further replies.
Back
Top