AMD Announces Q4 2016 Earnings [Anandtech]

[Bacon1]

http://www.anandtech.com/show/11092/amd-announces-q4-2016-earnings

AMD attributes their increase in revenue primarily due to increased GPU sales. AMD’s Computing and Graphics segment had revenues for the quarter of $600 million, which is up 28% year-over-year. Polaris seems to be doing quite well, which is great to see. The segment still had an operating loss of $21 million, but that is a big improvement from the operating loss of $99 million a year ago. CPU average selling price fell year-over-year, but GPU average selling price increased year-over-year, thanks to higher desktop and professional graphics pricing. We eagerly await the launch of Ryzen, which AMD showed off last quarter, and expect to hear more about it this quarter.

Sounds like Polaris is doing very well.

Can't wait for Vega

 

[tviceman]

It's crazy that in October-December of 2015 AMD had the 390/390X and 3 Fury cards, yet had LOWER ASP than October-December 2016 with the RX 480 and everything below.

It apparently must show how little Fury lineup and the 390/390X sold during that time frame.

 

[Qwertilot]

Well the fury stuff at least had some real production quantity issues, and for a good while wasn't really even priced to try and sell it.

 

[AtenRa]

http://seekingalpha.com/article/404...-2016-results-earnings-call-transcript?page=2

Desktop GPU shipments and revenue increased by double-digit percentages from a year ago, based on growing OEM and channel adoption of Polaris GPUs. Polaris processor sales were particularly strong in the performance and enthusiast portions of the market, resulting in our highest channel GPU sales in more than two years.

Mobile GPU sales growth outpaced desktop in the quarter, and we believe we gained further share in this part of the market as the first Polaris-based notebooks launched.

 

[raghu78]

AMD will have Vega GPUs top to bottom complete stack according to Lisa Su. This pretty much confirms there is no Polaris refresh. Its all Vega with GFX IP 9.0 and NCU going forward. My guess is we will have a next gen Rx 6xx series as Rx 5xx is already a OEM rebrand of Rx 4xx (Polaris).

https://videocardz.com/65538/amd-radeon-rx-570-spotted-in-samsung-notebook

http://seekingalpha.com/article/404...-results-earnings-call-transcript?part=single

Lisa T. Su - Advanced Micro Devices, Inc.

"Yes. So, in terms of our strategy, I think on both the CPU and the GPU side, we have been on a fairly deliberate path to ensure that we got to a very competitive roadmap. So, on the CPU side, with Ryzen and Naples, we believe we will be quite competitive. On the GPU side, as we launch Vega, we will have a full stack, sort of top to bottom with new hardware. We continue to invest in software, and our approach to software is really around open source and using the ecosystem and using the community and focused on sort of the new APIs. So in gaming, we are very focused on DX12 and Vulkan and on the professional graphics and on the GPU server side, really using our GPUOpen. So, we'll continue to invest in software. No question that that's really critical for the graphics market, but we feel we are making good progress."

 

[CatMerc]

AMD will have Vega GPUs top to bottom complete stack according to Lisa Su. This pretty much confirms there is no Polaris refresh. Its all Vega with GFX IP 9.0 and NCU going forward. My guess is we will have a next gen Rx 6xx series as Rx 5xx is already a OEM rebrand of Rx 4xx (Polaris).

https://videocardz.com/65538/amd-radeon-rx-570-spotted-in-samsung-notebook

http://seekingalpha.com/article/404...-results-earnings-call-transcript?part=single

Lisa T. Su - Advanced Micro Devices, Inc.

"Yes. So, in terms of our strategy, I think on both the CPU and the GPU side, we have been on a fairly deliberate path to ensure that we got to a very competitive roadmap. So, on the CPU side, with Ryzen and Naples, we believe we will be quite competitive. On the GPU side, as we launch Vega, we will have a full stack, sort of top to bottom with new hardware. We continue to invest in software, and our approach to software is really around open source and using the ecosystem and using the community and focused on sort of the new APIs. So in gaming, we are very focused on DX12 and Vulkan and on the professional graphics and on the GPU server side, really using our GPUOpen. So, we'll continue to invest in software. No question that that's really critical for the graphics market, but we feel we are making good progress."

You can interpert it in multiple ways. Technically a revision of Polaris 10 like the Polaris 10 XT2 spotted in Mac drivers could mean new hardware.

I long suspected that Vega will be 500/600 series, and Polaris will move down a peg to make room for Vega at the top, but still stick around.

 

[tviceman]

AMD will have Vega GPUs top to bottom complete stack according to Lisa Su. This pretty much confirms there is no Polaris refresh. Its all Vega with GFX IP 9.0 and NCU going forward. My guess is we will have a next gen Rx 6xx series as Rx 5xx is already a OEM rebrand of Rx 4xx (Polaris).

https://videocardz.com/65538/amd-radeon-rx-570-spotted-in-samsung-notebook

http://seekingalpha.com/article/404...-results-earnings-call-transcript?part=single

Lisa T. Su - Advanced Micro Devices, Inc.

"Yes. So, in terms of our strategy, I think on both the CPU and the GPU side, we have been on a fairly deliberate path to ensure that we got to a very competitive roadmap. So, on the CPU side, with Ryzen and Naples, we believe we will be quite competitive. On the GPU side, as we launch Vega, we will have a full stack, sort of top to bottom with new hardware. We continue to invest in software, and our approach to software is really around open source and using the ecosystem and using the community and focused on sort of the new APIs. So in gaming, we are very focused on DX12 and Vulkan and on the professional graphics and on the GPU server side, really using our GPUOpen. So, we'll continue to invest in software. No question that that's really critical for the graphics market, but we feel we are making good progress."

That doesn't mean Polaris won't see a refresh; it just means that AMD will be competing in the high end again, in addition to the mid range and low end. That said, I have no idea if Polaris will get a refresh, but if it does I highly doubt it will be the fantasy-version with 33-50% improved perf/w that people have latched on to around here. AMD hasn't "revised" an existing GPU for a long, long time. They don't have the resources to split up like that.

 

[CatMerc]

That doesn't mean Polaris won't see a refresh; it just means that AMD will be competing in the high end again, in addition to the mid range and low end. That said, I have no idea if Polaris will get a refresh, but if it does I highly doubt it will be the fantasy-version with 33-50% improved perf/w that people have latched on to around here. AMD hasn't "revised" an existing GPU for a long, long time. They don't have the resources to split up like that.

Process improvements and tighter binning could do that. There are already samples in the wild that are far more efficient than what we saw at launch. Buildzoid's and J2C's samples being examples of that.

 

[raghu78]

That doesn't mean Polaris won't see a refresh; it just means that AMD will be competing in the high end again, in addition to the mid range and low end. That said, I have no idea if Polaris will get a refresh, but if it does I highly doubt it will be the fantasy-version with 33-50% improved perf/w that people have latched on to around here. AMD hasn't "revised" an existing GPU for a long, long time. They don't have the resources to split up like that.

There is no reason to stick with Polaris as Vega brings improved IPC , higher clocks and better power efficiency. Vega should deliver significantly better perf/watt and perf/sqmm over Polaris. I consider Polaris more of a process pipe cleaner like the HD 4770 and Vega as a new architecture which will power AMD's entire GPU stack from top to bottom.

 

[PPB]

You cant make a full stack of products with 2 dies. Either they have a third die planned to replace 460, or it will be as we think that Polaris and Vega would only overlap a little in the middle (P10)

 

[AtenRa]

There is no reason to stick with Polaris as Vega brings improved IPC , higher clocks and better power efficiency. Vega should deliver significantly better perf/watt and perf/sqmm over Polaris. I consider Polaris more of a process pipe cleaner like the HD 4770 and Vega as a new architecture which will power AMD's entire GPU stack from top to bottom.

After HD4770 we had ..................

 

[PPB]

After HD4770 we had ..................

HD 4770 was only 1 SKU and wasnt hyped as Polaris was. Stop trying to correlate both products.

 

[AtenRa]

HD 4770 was only 1 SKU and wasnt hyped as Polaris was. Stop trying to correlate both products.

I was talking about Cypress (HD5870).

 

[Headfoot]

I just cant see them jettisoning Polaris already. I agree that bottom is polaris and polaris refresh, mid is overlapping, and high is Vega. At least until Vega 20 and the die shrinks. For a company strapped for R&D money, I can't think of a more foolish thing to do than push out a brand new modified arch die then drop it in 6 months for another, newer arch die. If they were planning on dropping it they would have been much more efficient to base it all off of a single more scalable architecture.

 

[AtenRa]

Why do you people believe AMD will introduce both entry, middle and High-End VEGA at the same time in Q2 2017 ???

AMD could easily launch High-End VEGA in Q2 2017, then launch Middle-End VEGA in Q3 and entry VEGA in Q4.
This way Polaris 10 (RX 470/80) will be replaced after 5 quarters (June 2016 to September 2017), same with Polaris 11 (RX 460) will be replaced after 5 quarters after official launch (August 2016 to November 2017).

Im not saying this will happen but its not impossible for AMD to have a top to bottom VEGA GPUs until the end of 2017.

 

[raghu78]

I just cant see them jettisoning Polaris already. I agree that bottom is polaris and polaris refresh, mid is overlapping, and high is Vega. At least until Vega 20 and the die shrinks. For a company strapped for R&D money, I can't think of a more foolish thing to do than push out a brand new modified arch die then drop it in 6 months for another, newer arch die. If they were planning on dropping it they would have been much more efficient to base it all off of a single more scalable architecture.

We are likely to see the high end Vega 10 by mid-late Q2 2017. Vega 11 will probably launch in early Q3 and the replacements for Polaris 10 and 11 by late Q3/early Q4. So Polaris 10 and 11 would have been in the market for 4-5 quarters. More importantly Polaris served as a 14nm process pipe cleaner and as a testbed for few features like primitive discard accelerator.

Why do you people believe AMD will introduce both entry, middle and High-End VEGA at the same time in Q2 2017 ???

AMD could easily launch High-End VEGA in Q2 2017, then launch Middle-End VEGA in Q3 and entry VEGA in Q4.
This way Polaris 10 (RX 470/80) will be replaced after 5 quarters (June 2016 to September 2017), same with Polaris 11 (RX 460) will be replaced after 5 quarters after official launch (August 2016 to November 2017).

Im not saying this will happen but its not impossible for AMD to have a top to bottom VEGA GPUs until the end of 2017.

Exactly. It would take 6 months to roll out a full stack based on Vega. If we assume a late May-June launch for the high end Vega 10 it would be Q4 by the time AMD releases all the GPUs.

 

[tviceman]

There is no reason to stick with Polaris as Vega brings improved IPC , higher clocks and better power efficiency. Vega should deliver significantly better perf/watt and perf/sqmm over Polaris. I consider Polaris more of a process pipe cleaner like the HD 4770 and Vega as a new architecture which will power AMD's entire GPU stack from top to bottom.

Vega are bigger, more expensive chips that can't serve the <$200 market segment. And insofar as Vega's performance characteristics are concerned, lets just wait and see. AMD said Polaris was 2.8x more efficient than their 28nm offerings, but that was absolutely, positively, undeniably, nowhere close to being remotely true AT ALL. AMD talks a big talk prior to chip releases and often ends up under-delivering on their promises and/or comes with stipulations.

We are likely to see the high end Vega 10 by mid-late Q2 2017. Vega 11 will probably launch in early Q3 and the replacements for Polaris 10 and 11 by late Q3/early Q4. So Polaris 10 and 11 would have been in the market for 4-5 quarters. More importantly Polaris served as a 14nm process pipe cleaner and as a testbed for few features like primitive discard accelerator.

After all the cuts to R&D and engineers over the years, where is AMD coming up with the resources to release more than 4 new GPU chips in a 16 month time span? Case in point - look how long AMD stuck with the last ~230 mm2 chip they had, let alone most of their other 28nm chips which relived to see rebadge after rebadge.

 

[PPB]

I was talking about Cypress (HD5870).

Dont play dumb, you know what I mean. You are desperately trying to make look as Polaris was HD 4770 and Vega is 5870. The context is completely different.

 

[CatMerc]

Dont play dumb, you know what I mean. You are desperately trying to make look as Polaris was HD 4770 and Vega is 5870. The context is completely different.

Is it not?
Polaris looks to be a rather small refinement of GCN ported over to 14nm. Vega is claimed by AMD to be the biggest change in GCN since its inception.

 

[rainy]

AMD said Polaris was 2.8x more efficient than their 28nm offerings

Quite obviously, you've have missed a little however important phrase "up to".

 

[PPB]

Is it not?
Polaris looks to be a rather small refinement of GCN ported over to 14nm. Vega is claimed by AMD to be the biggest change in GCN since its inception.

Lets see:

- HD 4770 was a product inserted into the precedent product stack on bigger node. Polaris is neither 1 product only, it was not inserted in a precedent product stack, and is a product stack on it's own (3 different bins of P10, 3 different bins of P11, for a total 6 total different skus for consumer-gaming).
- HD 4770 came 10 months later than 4850/70, and 5 months earlier than 5850/70. Polaris 10/11 came 1 year later than Fury X, and 8 to 10 months earlier than suggested VEGA launch.
- While the IP version wasn't updated with Polaris, AMD insiders already claimed that was to make Polaris more transparent to earlier GCN optimizations already made on the ecosystems and we know there are IPC and uncore changes between Polaris and the latest product before it (Fiji). HD 4700 was 100% the same uarch as 5870, just TMUs/Shaders/ROPs/bus width were different/bigger.
- Marketing around both products were totally different.

For Polaris to be said pipecleaner product, P10 for example it must have been a R9 385/X inserted in the rehash of the rehash of a product stack that R300 series was, and VEGA just must be a 2-2.5x of everything P10 had, without new uarch changes. Thank god this is not the case and we are getting an architecture revamp.

 

[tviceman]

Quite obviously, you've have missed a little however important phrase "up to".

https://www.techpowerup.com/reviews/AMD/RX_480/25.html Not a single AMD product from last gen that the RX 480 achieved a 2x efficiency over, let alone 2.8x.

I expect this kind of response from a kool aid drinker. Are you one of those? If AMD achieved 2.8x efficiency in one instance, and one instance only (and the jury is out if that ever happened) it's a fluke and shame on them for continuing to perpetuate the cycle of failed expectations over and over again.

Personal insults or attacks are not allowed
Markfw
Anandtech Moderator

 

[AtenRa]

https://www.techpowerup.com/reviews/AMD/RX_480/25.html Not a single AMD product from last gen that the RX 480 achieved a 2x efficiency over, let alone 2.8x.

*See Endnotes for details.

 

[Bacon1]

https://www.techpowerup.com/reviews/AMD/RX_480/25.html Not a single AMD product from last gen that the RX 480 achieved a 2x efficiency over, let alone 2.8x.

 

[tviceman]

*See Endnotes for details.

Those are cute end notes....oh wait the RX 270 doesn't draw nearly 180 watts on those tests and the rx 470 draws more than 110 watts. Simple math shows AMD's footnotes are either lies or are incredibly misleading and still untrue.