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

[Rootax]

I agree they're done for 2018 and 2019 for the high end market. I've only Vega FE because of Freesync, VRAM, and a veryyyyyyyy good deal at the time (and my fury X only 4gb were more and more problematic).

It's not the topic, but my only hope now to have some competition is intel by 2020. Not AMD.

 

[Esrever]

Global Foundries said they were taping out a 7nm part by the end of this year before they cancelled their 7nm. I think Navi might have been one the more likely chips that it could have been. The transition to TSMC for everything have probably been under way for a while now but I wonder if it would delay that 1 chip that was originally meant for Glofo and what the chances that chip would be Navi.

 

[Kaotik]

Global Foundries said they were taping out a 7nm part by the end of this year before they cancelled their 7nm. I think Navi might have been one the more likely chips that it could have been. The transition to TSMC for everything have probably been under way for a while now but I wonder if it would delay that 1 chip that was originally meant for Glofo and what the chances that chip would be Navi.

Unlikely, if anything Navi has been brought forward since it's now part of their "initial 7nm products" which definitely wouldn't be possible if it was designed for GloFo and would need to be converted for TSMC. Also at least Patrick Moorehead believes AMD actually had zero relevant designs for GloFo 7nm

https://twitter.com/i/web/status/1034175749249425408

 

[Esrever]

Unlikely, if anything Navi has been brought forward since it's now part of their "initial 7nm products" which definitely wouldn't be possible if it was designed for GloFo and would need to be converted for TSMC. Also at least Patrick Moorehead believes AMD actually had zero relevant designs for GloFo 7nm

https://twitter.com/i/web/status/1034175749249425408

That looks like he was just speculating based on the fact that Vega20 and Epyc2 were TSMC. Nowhere has there been anything said by AMD on Navi being originally for TSMC or Glofo. I don't see how he can be used as a source when he doesn't have any insider information either based on the fact he uses the words "I don't think..".

AMD's only comments have been no public release schedule has changed, which means Navi is still 2019 since that is the only thing they have said about it. Navi could have originally been early Q3 2019 but could now be late Q4 for all we know.

The chip originally meant for Glofo would most likely be a GPU or a APU as those would be within the timeline to be released after 7nm parts of TSMC. Generally GPUs come before CPUs and the APUs that make use of the same arch. So I think Navi would make sense being the first chip at Glofo's 7nm.

 

[Kaotik]

That looks like he was just speculating based on the fact that Vega20 and Epyc2 were TSMC. Nowhere has there been anything said by AMD on Navi being originally for TSMC or Glofo. I don't see how he can be used as a source when he doesn't have any insider information either based on the fact he uses the words "I don't think..".

AMD's only comments have been no public release schedule has changed, which means Navi is still 2019 since that is the only thing they have said about it. Navi could have originally been early Q3 2019 but could now be late Q4 for all we know.

The chip originally meant for Glofo would most likely be a GPU or a APU as those would be within the timeline to be released after 7nm parts of TSMC. Generally GPUs come before CPUs and the APUs that make use of the same arch. So I think Navi would make sense being the first chip at Glofo's 7nm.

You completely disregarded the actual point - if Navi was designed for GloFo it surely couldn't be part of their "initial 7nm products" which now all come from TSMC, surely converting the design over to completely different process would take more time

 

[Esrever]

You completely disregarded the actual point - if Navi was designed for GloFo it surely couldn't be part of their "initial 7nm products" which now all come from TSMC, surely converting the design over to completely different process would take more time

Does AMD actually say "initial 7nm products" in regards to Navi? Even if they did, it could mean anything, how is it even remotely definitive?

AMD could have known about possible cancellation for a while and transferred fab in that time, and my original point was there could be a delay associated which could still impact the products.

 

[Deleted member 90741]

Does AMD actually say "initial 7nm products" in regards to Navi? Even if they did, it could mean anything, how is it even remotely definitive?

AMD could have known about possible cancellation for a while and transferred fab in that time, and my original point was there could be a delay associated which could still impact the products.

From AMD's Investor's relations
http://ir.amd.com/news-releases/new...g-our-high-performance-leadership-focused-7nm

AMD’s next major milestone is the introduction of our upcoming 7nm product portfolio, including the initial products with our second generation “Zen 2” CPU core and our new “Navi” GPU architecture.

To streamline our development and align our investments closely with each of our foundry partner’s investments, today we are announcing we intend to focus the breadth of our 7nm product portfolio on TSMC’s industry-leading 7nm process.

We will leverage the additional investments GLOBALFOUNDRIES is making in their robust 14nm and 12nm technologies at their New York fab

 

[Kaotik]

Does AMD actually say "initial 7nm products" in regards to Navi? Even if they did, it could mean anything, how is it even remotely definitive?

AMD could have known about possible cancellation for a while and transferred fab in that time, and my original point was there could be a delay associated which could still impact the products.

Here's direct quote

AMD’s next major milestone is the introduction of our upcoming 7nm product portfolio, including the initial products with our second generation “Zen 2” CPU core and our new “Navi” GPU architecture.

http://ir.amd.com/news-releases/new...g-our-high-performance-leadership-focused-7nm

 

[yuri]

Reading through recent HotChips Tachyum Prodigy slides it seems HBM3 is gonna be available in 2019. Linking this info with Navi's "Next-gen memory" attribute makes one think. Could the Navi delay be caused by HBM tech availability? If so, it would make the whole AMD's HBM adventure a bit sad.

 

[Esrever]

Reading through recent HotChips Tachyum Prodigy slides it seems HBM3 is gonna be available in 2019. Linking this info with Navi's "Next-gen memory" attribute makes one think. Could the Navi delay be caused by HBM tech availability? If so, it would make the whole AMD's HBM adventure a bit sad.

Next gen memory almost certainly mean GDDR6.

 

[Rootax]

Navi is delayed ?

 

[Kaotik]

Navi is delayed ?

From the original Polaris-roadmap it could be claimed it's delayed, since we already know they'll bring only one 7nm product this year (Vega 20) and it's clearly marked for 2018 same way Vega was marked for 2017.
Other than that, latest comments from AMD suggest early 2019 for Navi, compared to most earlier rumors talking about late 2019 or even later

 

[DrYesterday]

Is there any chance that NAVI is being delayed, so AMD can add hardware ray-tracing support?

 

[Malo]

Is there any chance that NAVI is being delayed, so AMD can add hardware ray-tracing support?

That could depend on when Microsoft started planning on adding DXR I guess.

 

[McHuj]

Is there any chance that NAVI is being delayed, so AMD can add hardware ray-tracing support?

Probably 0 at this point. Hardware design like this takes years. Maybe it got delayed a year or so ago, but not now.

I also think MS had been working on DXR for a long time now as well. They didn't just come up with that idea back in March a week before unveiling it. I'm going to guess that both Nvidia and AMD have known about DXR for a many months if not years prior to the announcement. If AMD chose to build anything for it, is another story.

 

[Alexko]

Probably 0 at this point. Hardware design like this takes years. Maybe it got delayed a year or so ago, but not now.

I also think MS had been working on DXR for a long time now as well. They didn't just come up with that idea back in March a week before unveiling it. I'm going to guess that both Nvidia and AMD have known about DXR for a many months if not years prior to the announcement. If AMD chose to build anything for it, is another story.

I wouldn't be shocked if this turned out to be tessellation all over again, where NVIDIA goes all in right away, while AMD adopts a much more cautious approach, dedicating just enough hardware to DXR as they absolutely have to, until it sees wide enough adoption to warrant the use of more silicon.

That said, DXR would seem to have broader initial support than tessellation did at the time, but that's not necessarily something AMD could have foreseen.

 

[Kaotik]

I wouldn't be shocked if this turned out to be tessellation all over again, where NVIDIA goes all in right away, while AMD adopts a much more cautious approach, dedicating just enough hardware to DXR as they absolutely have to, until it sees wide enough adoption to warrant the use of more silicon.

That said, DXR would seem to have broader initial support than tessellation did at the time, but that's not necessarily something AMD could have foreseen.

Except that AMD went tessellation before it was even in DirectX, you could even argue they've already had tessellation hardware since Radeon 8500 and relatively modern since HD2000

 

[Alexko]

Except that AMD went tessellation before it was even in DirectX, you could even argue they've already had tessellation hardware since Radeon 8500 and relatively modern since HD2000

Yes, but their investment in it, precocious as it was, was initially limited, and only very gradually increased.

 

[bgroovy]

AMD TruForm 4 LIFE, yo!

 

[Kaotik]

AMD TruForm 4 LIFE, yo!

Neverwinter Nights was sooooooo much better looking with the semi-official TruForm (N-Patches) support despite few gaps in models here and there