Vega/Navi Rumors (Updated)

[Bacon1]

New info/rumors, apparently from official roadmaps:

Vega 11



Vega 10



http://videocardz.com/63715/amd-vega-and-navi-roadmap

Also confirms Vega 10 is the big chip, and it packs 4096 SPs (same as Fiji).

According to AMD, Vega 10 is small chip and Vega 11 is the large chip.

The naming was based on when they created the die, not on size or anything else

So that right there kills the credibility of your post.

 

[Sweepr]

According to AMD, Vega 10 is small chip and Vega 11 is the large chip.

Source?

So that right there kills the credibility of your post.

Still waiting.

 

[zinfamous]

According to AMD, Vega 10 is small chip and Vega 11 is the large chip.

The naming was based on when they created the die, not on size or anything else

So that right there kills the credibility of your post.

I posted this comment earlier, then I went to find one of the AMD slides that "confirmed" this, but I couldn't find it. I did find an old slide, though, that suggested Polaris 11 was big Polaris and Polaris 10 was small Polaris, but we now know that wasn't true (apparently that slide was posted by wcctech or whatever--usually the least-credible site, right).

Anyway, I'm not really sure what the truth is here until we get more info from AMD--because Arachnotronic does make a good point: AMD is actually liable for spreading false information, unlike these other websites. That is why one can really only trust that information when it is made public.

 

[AtenRa]

If im not mistaken it was Raja that said the naming is in order of what Chip was first and what chip was second, Polaris 10 was the first chip and Polaris 11 was the second chip. Same will be with Vega but we dont really know if Vega 10 will be the big or smaller chip.

 

[Bacon1]

Source?



Still waiting.

If im not mistaken it was Raja that said the naming is in order of what Chip was first and what chip was second, Polaris 10 was the first chip and Polaris 11 was the second chip. Same will be with Vega but we dont really know if Vega 10 will be the big or smaller chip.

Yep it was Raja or Robert or someone who said it was based on the development time for when it was created. Best source I can find now are other people talking about it, can't remember who said it or what interview, so feel free to do you own research.

https://www.reddit.com/r/Amd/comments/4kd6os/am_i_the_only_one_confused_by_the_naming_scheme/

 

[Bacon1]

http://www.guru3d.com/news-story/amd-vega-10-vega-20-and-vega-11-gpus-mentioned-by-cto.html

Vega 10
The start of the "Vega" will happen with will be called the Vega 10 (codename) GPU, a series to compete with Nvidias more high-end gear. It's mentioned that this GPU has 64 Compute Units, multiply that with 64 shader processors per cluster and you'll get to 4096 shader processors. It will offer up to 24 TFLOP/s 16-bit (half-precision) floating point performance. You read it right, half-precision. The GPU will be paired with either 8 or 16 GB HBM2 memory at an up-to 512 GB/s memory bandwidth, logic would dictate 8GB though.

Vega 10 would thus be the GPU that replaces the Hawaii/Grenada series. It is mentioned that total board power (not TDP) sits at 225 Watt. Vega 10 should hold HBM2 memory and that's the culprit, it is just not available in volume. Hence even I myself expected a VEGA10 product released in October armed with GDDR5(X) graphics memory). That last bit thus is the uncertain part.

Vega 11
Vega11 - lets call it "Big Vega' which is to replace the FIJI / Fury (X) parts. So as suggested, where Vega10 would get 4096 stream processors and could compete with the GeForce GTX 1080 and GTX 1070 Vega11 on its end would feature 6144 stream processors (=rumor) and would be lined against the GP100/GP102 aka Big Pascal GPU (Titan X). This all is obviously known and discussed many times already. Then there now is some new and intersting info.

 

[Arachnotronic]

http://www.guru3d.com/news-story/amd-vega-10-vega-20-and-vega-11-gpus-mentioned-by-cto.html

Guru3D doesn't have a source, it's just making things up. VideoCardz apparently has seen an internal presentation and Vega10 is clearly positioned as the "big Vega."

 

[Bacon1]

Guru3D doesn't have a source, it's just making things up. VideoCardz apparently has seen an internal presentation and Vega10 is clearly positioned as the "big Vega."

Yes the same creditable source saying that 7nm GPU Vega 20 is coming soon right?

 

[Arachnotronic]

Yes the same creditable source saying that 7nm GPU Vega 20 is coming soon right?

I could see a 7nm TSMC GPU come in late 2018 (though that schedule seems aggressive). GloFo-built, no way.

 

[Sweepr]

So your source is made up stuff by Guru3D (misquoting VideoCardz), not AMD. 'That right there kills the credibility of your post'.

Not to mention it doesn't make any sense either.

- Their 7nm Vega 20 part packs less CUs than Vega 11
- What's the point of a dual Vega 10 card at low clocks months later part if they have a 6000+ SPs SKU coming?
- A huge 6000 SPs+ GPU replaces (mainstream) Polaris 10 in the professional market

 

[Bacon1]

I wasn't able to find which AMD employee said that it was based on development time, but its been known that way for a long time. Why else would Polaris 11 be the slower chip while having a higher #?

I think its amazing you guys are buying the videocardz post and expect to see a 7nm GPU next year. Why would they release it after Navi?

 

[AtenRa]

I could see a 7nm TSMC GPU come in late 2018 (though that schedule seems aggressive). GloFo-built, no way.

2018 and 7nm doesnt go together, even Intels 7nm is for 2019-20.

14/16nm will not be a two year only process.

 

[Det0x]

That is true, but if they update the investor presentation and say "this is our current roadmap" when it is not actually the roadmap, then there is a serious problem.

If they choose to simply pull the roadmap slide from the deck the next time they update it, that's perfectly fine.

One more time.

AMD has plans to ship Navi in 2019 according to the leaked server roadmap. Now you might believe that Navi has been delayed but i don’t find that to be the case. This roadmap is specific to the server platform and these products have the tendency to arrive later as compared to consumer parts. Hence, Navi could still be a planned 2018 launch on the consumer side but could miss that time frame on the server side of things.

I see no conflict

 

[Arachnotronic]

2018 and 7nm doesnt go together, even Intels 7nm is for 2019-20.

14/16nm will not be a two year only process.

Intel's 7nm will be more like what the foundries will call 5nm, so I expect foundries to beat Intel to "7nm."

Anyway, I wouldn't be surprised to see both NV and AMD adopt TSMC 7nm for late 2018/1H 2019 GPUs.

 

[Arachnotronic]

One more time.

AMD has plans to ship Navi in 2019 according to the leaked server roadmap. Now you might believe that Navi has been delayed but i don’t find that to be the case. This roadmap is specific to the server platform and these products have the tendency to arrive later as compared to consumer parts. Hence, Navi could still be a planned 2018 launch on the consumer side but could miss that time frame on the server side of things.

I see no conflict

It's the same exact GPU.

 

[zinfamous]

It's the same exact GPU.

I would assume then that the release for consumer vs industry is based on volume targets and pricing. The two separate dates could reflect the respective timepoints of when they expect to reach the target volume for each initial shipment.

Imagine the room divided into two spaces: one where pallets of schwag will be shipped off to consumers, and the other will be shipped off to industry clients that order in different volume and different pricing contracts.

I learned of this "end-used binning" when I had to order/follow-up on a large, missing piece of a (expensive) equipment in lab. The order was placed before I arrived, but already late by ~2 months leads time when I called. The vendor said that ours was "Still being built" and we only have 3 units in store for our industry clients. Same exact unit, different price structure and inventory. Now, ours wasn't built up properly for our needs with custom parts, apparently, so they we were going to have to add another couple of months (of course they already had our end use needs from the beginning). Well, that was annoying, so I just called another company that offered a similar product and ordered from them, shipped in 2 days and arriving in 6. Done.

 

[Bacon1]

Lets look at all the leaks from Videocardz:

VEGA 10 = Greenland: http://videocardz.com/58665/amds-greenlandvega-10-gpu-to-feature-4096-stream-processors

VEGA 10 - 4096 shaders, 4096 bit 8GB HBM2

VEGA coming in October - http://videocardz.com/59808/amd-vega-gpu-allegedly-pushed-forward-to-october

Can be either 14 Goflo or 16nm TSMC

VEGA 10 != Greenland: http://videocardz.com/62250/amd-vega10-and-vega11-gpus-spotted-in-opencl-driver

Welp Greenland is now different from Vega 10, so first rumor is wrong, or this one is...

Vega coming soon : http://videocardz.com/63194/amd-teases-vega-architecture-launching-soon

Vega coming 1h 2017 : http://videocardz.com/63450/amd-vega-to-launch-in-1h-2017

Vega 10, 11, 20 : http://videocardz.com/63700/exclusive-first-details-about-amd-vega10-and-vega20

So now we have Vega coming in 3 verisons.

14nm 10/11 and 7nm 20 with 32GB of HBM2

Vega 10 also got a bump to 16GB HBM2 from 8GB from their original #s.



So we've gone from:

Vega 10 = Greenland, to not greenland, from 8GB -> 16GB and now adding in a 7nm pre-2019 card too with Navi being pushed back a year.

Can't wait for the next batch of rumors!

 

[Det0x]

Guru3D doesn't have a source, it's just making things up. VideoCardz apparently has seen an internal presentation and Vega10 is clearly positioned as the "big Vega."

I don't see anything which contradicts Vega11 from being the big die ?

As far as i know this is the only information we have so far from that site recently:

We have also learned that AMD is planning to replace Polaris 10 with Vega 11 next year. The specs were not released and we can only expect Vega 11 to share the same fabrication node with Vega 10 (14nm).

http://videocardz.com/63700/exclusive-first-details-about-amd-vega10-and-vega20

Like I said yesterday, full details of VEGA 11 were not yet disclosed. However, I did tell you that Polaris 10 will be replaced by Vega 11 next year. Of course, what I meant was the professional market. It does not mean there won’t be Polaris 10-based Radeons next year. It means that more powerful solutions will be offered in Radeon Pro series with VEGA GPU.

http://videocardz.com/63715/amd-vega-and-navi-roadmap

Vega11 could very well be the big die Vega, as guru3d suggests..

Vega11 - lets call it "Big Vega' which is to replace the FIJI / Fury (X) parts. So as suggested, where Vega10 would get 4096 stream processors and could compete with the GeForce GTX 1080 and GTX 1070 Vega11 on its end would feature 6144 stream processors (=rumor) and would be lined against the GP100/GP102 aka Big Pascal GPU (Titan X). This all is obviously known and discussed many times already. Then there now is some new and intersting info.

http://www.guru3d.com/news-story/amd-vega-10-vega-20-and-vega-11-gpus-mentioned-by-cto.html

Vega10 to fight GTX 1080
Vega11 to fight Titan XP / GTX 1080 TI

- What's the point of a dual Vega 10 card at low clocks months later part if they have a 6000+ SPs SKU coming?

My own speculation:

The reason for them to use the Vega10 as a base for the dual card is simply because Vega11 is a 250/300w TDP part..

Would be much more sensible for them to use downclocked (ala nano) Vega10 as a base for the dual card then, don't you agree ?

 

[AtenRa]

Intel's 7nm will be more like what the foundries will call 5nm, so I expect foundries to beat Intel to "7nm."

Anyway, I wouldn't be surprised to see both NV and AMD adopt TSMC 7nm for late 2018/1H 2019 GPUs.

TSMC 10nm will start production in 2017 for high margin mobile SoCs. How the hell AMD and NVIDIA will use 7nm in 2018 or even in early 2019 ??
Even if TSMC will have 7nm production ready by early 2019(Q1), the process will be expensive and we dont know if the first 7nm will only be for Mobile SoCs like 16nm was in 2015.

I will say that 14/16nm GPUs will be 3-4 years as 28nm was. We havent seen full density products yet, both Polaris and Pascal chips doesnt use the full density available of 14/16nm process.

 

[Arachnotronic]

TSMC 10nm will start production in 2017 for high margin mobile SoCs. How the hell AMD and NVIDIA will use 7nm in 2018 or even in early 2019 ??
Even if TSMC will have 7nm production ready by early 2019(Q1), the process will be expensive and we dont know if the first 7nm will only be for Mobile SoCs like 16nm was in 2015.

I will say that 14/16nm GPUs will be 3-4 years as 28nm was. We havent seen full density products yet, both Polaris and Pascal chips doesnt use the full density available of 14/16nm process.

Mobile SoCs aren't high margin. dGPUs are high margin.

 

[AtenRa]

Mobile SoCs aren't high margin. dGPUs are high margin.

High-End Mobile SoCs are high margin, those chips that get installed in to $600-800 or higher Smartphones. Like Apple A10 and next high-end Samsung and Qualcomm SoCs.

 

[Arachnotronic]

High-End Mobile SoCs are high margin, those chips that get installed in to $600-800 or higher Smartphones. Like Apple A10 and next high-end Samsung and Qualcomm SoCs.

No, they're not. Qualcomm's chip margins are nowhere near what NVIDIA's gross profit margins are.

 

[zinfamous]

No, they're not. Qualcomm's chip margins are nowhere near what NVIDIA's gross profit margins are.

well if nvidia's margins are now the watermark for how we define "high margins", then I think think we need to rewrite a few books.

 

[AtenRa]

No, they're not. Qualcomm's chip margins are nowhere near what NVIDIA's gross profit margins are.

I said the High-End Chips, not the average margins of all Qualcomm SoCs.

 

[maddie]

Lets look at all the leaks from Videocardz:

VEGA 10 = Greenland: http://videocardz.com/58665/amds-greenlandvega-10-gpu-to-feature-4096-stream-processors

VEGA 10 - 4096 shaders, 4096 bit 8GB HBM2

VEGA coming in October - http://videocardz.com/59808/amd-vega-gpu-allegedly-pushed-forward-to-october

Can be either 14 Goflo or 16nm TSMC

VEGA 10 != Greenland: http://videocardz.com/62250/amd-vega10-and-vega11-gpus-spotted-in-opencl-driver

Welp Greenland is now different from Vega 10, so first rumor is wrong, or this one is...

Vega coming soon : http://videocardz.com/63194/amd-teases-vega-architecture-launching-soon

Vega coming 1h 2017 : http://videocardz.com/63450/amd-vega-to-launch-in-1h-2017

Vega 10, 11, 20 : http://videocardz.com/63700/exclusive-first-details-about-amd-vega10-and-vega20

So now we have Vega coming in 3 verisons.

14nm 10/11 and 7nm 20 with 32GB of HBM2

Vega 10 also got a bump to 16GB HBM2 from 8GB from their original #s.



So we've gone from:

Vega 10 = Greenland, to not greenland, from 8GB -> 16GB and now adding in a 7nm pre-2019 card too with Navi being pushed back a year.

Can't wait for the next batch of rumors!

I agree.

At least we have some very diligent, hardworking posters searching the web for these posts to provide some amusement. I thank them for their service.