The AMD Execution Thread [2007 - 2017]

[iMacmatician]

I understood the orange coloring in said roadmap to mean Zen-cores for all around, including APUs

Each of AMD's three roadmap slides of that type (desktop graphics, mobile graphics, and desktop and mobile CPUs and APUs) shows separate colors for different years. Given that the M300 parts so far are known to use existing chips, I'm not assuming that the orange color represents anything more than an 2016 introduction for those parts (and a new "generation" at least for marketing).

 

[A1xLLcqAgt0qc2RyMz0y]

Nvidia (which only sells Discrete GPUs) now has 14.9% market share or 2% more market share than AMD which has 12.9% market share and sells both APUs (that contain GPUs) and discrete GPUs. This is a gain of 0.6% in share for Nvidia compared to AMD in the just completed quarter.

http://www.jonpeddie.com/publications/market_watch

Also of interest are these paragraphs:

AMD’s ... AMD’s discrete desktop shipments decreased -14.55% from last quarter, and notebook discrete shipments decreased -13.6%.

Nvidia’s desktop discrete shipments were down -6.96% from last quarter; and the company’s notebook discrete shipments decreased -20.8%.

These numbers seem to imply that Nvidia will again gain in discrete market share again. The current split is 76% for Nvidia and 24% for AMD.

The Discrete GPU report comes out a little later than this overall one.

 

[Deleted member 13524]

How long until discrete notebook GPUs die off at every range but the higher end of gaming desktop replacements?

 

[Alexko]

It would seem that HBM will become mainstream on APUs (AMD's and Intel's, presumably) around 2017. So I'd expect discrete GPUs in notebooks to take a serious hit in 2018 and to be reduced to very low volume by the end of 2019.

 

[A1xLLcqAgt0qc2RyMz0y]

It would seem that HBM will become mainstream on APUs (AMD's and Intel's, presumably) around 2017. So I'd expect discrete GPUs in notebooks to take a serious hit in 2018 and to be reduced to very low volume by the end of 2019.

A number of assumptions in that paragraph.

You imply that Intel will have improved their GPUs and drivers. That in itself is a very big assumption.

Also who knows if AMD will still be around in 2017 much less 2019.

Oh and somehow Nvidia will be sitting idle during these periods.

On the subject of when Discrete GPUs will bite the big one (you state 2019) who can forget the Charlie/Spigzone prediction that discrete GPUs will be gone by May 2015:

http://www.semiaccurate.com/forums/showpost.php?p=48497&postcount=10

May 11, 2010

I would ask the question in a more general sense. Will GPUs exist in 5 years. The answer there would be no.

The low end dies this year, or at least starts to do a PeeWee Herman at the end of Buffy the Vampire Slayer (the movie, not the show). There goes the volume. 2012 sees the same happening for the high end. The middle isn't enough to sustain NV.

They have 2 years to make compute and widgets profitable. Good luck there guys.

-Charlie

 

[Alexko]

A number of assumptions in that paragraph.

You imply that Intel will have improved their GPUs and drivers. That in itself is a very big assumption.

Intel's GPUs have been steadily improving and I'm told their drivers have as well, although I haven't had the opportunity to confirm this myself. This assumption seems perfectly reasonable to me.

Also who knows if AMD will still be around in 2017 much less 2019.

AMD has a good bit of debt, but none of it has to be paid back before 2019. Therefore, yes, I expect them to survive at least until then, unless of course someone buys them off.

Oh and somehow Nvidia will be sitting idle during these periods.

No, but that's beside the point.

 

[3dilettante]

Intel's GPUs have been steadily improving and I'm told their drivers have as well, although I haven't had the opportunity to confirm this myself. This assumption seems perfectly reasonable to me.

Skylake's GPU is scaling up a decent amount, so at least on the physical side it should be expected to improve.

AMD has a good bit of debt, but none of it has to be paid back before 2019. Therefore, yes, I expect them to survive at least until then, unless of course someone buys them off.

If there were a goal to strive for, it would be for Zen and its range of products to earn enough revenue such that AMD can retire a significant amount of that 2019-2020 debt. Perhaps there's a magic threshold where it would be worth salvaging the company, like if it can retire enough debt and maintain cash reserves sufficiently high enough that a buyer could weigh them favorably with the long-term obligations in the next decade.

 

[A1xLLcqAgt0qc2RyMz0y]

AMD has a good bit of debt, but none of it has to be paid back before 2019. Therefore, yes, I expect them to survive at least until then, unless of course someone buys them off.

AMD is again burning CASH. If they continue having losses and burning CASH they will run out of money for day-to-day operations long before 2019. So yes they might not survive past 2016.

No, but that's beside the point.

So you remove the discrete vendor who currently has 76% of the market just to make your assumption more plausible.

 

[Alexko]

AMD is again burning CASH. If they continue having losses and burning CASH they will run out of money for day-to-day operations long before 2019. So yes they might not survive past 2016.

They've lost a bit of money this quarter, and forecast further losses for Q2, but profits for H2. Overall 2015 may turn out to be a wash.

So you remove the discrete vendor who currently has 76% of the market just to make your assumption more plausible.

I'm not removing anything, just saying that bigger integrated GPUs with lots of bandwidth will make discrete ones unnecessary for the vast majority of the notebook market.

 

[silent_guy]

AMD is again burning CASH. If they continue having losses and burning CASH they will run out of money for day-to-day operations long before 2019. So yes they might not survive past 2016.

They're not profitable, but their cash burn is minimal and they have plenty of reserves. They'll make it to 2017 without breaking a sweat. Companies like AMD can survive in this kind of purgatory for many years.

 

[Triskaine]

They're not profitable, but their cash burn is minimal and they have plenty of reserves. They'll make it to 2017 without breaking a sweat. Companies like AMD can survive in this kind of purgatory for many years.

The company can survive this, but can their product lineup? The effects of the layoffs and budget cuts at AMD in the past years are becoming increasingly visible. The last decent remains of their core IP are in acute danger of becoming uncompetitive for good.

 

[silent_guy]

If they couldn't afford a new 28nm generation, they may have skipped a cycle and focussed directly on 16/14nm. That makes them less competitive today, but gives them a chance to be competitive tomorrow.

 

[3dilettante]

If one believes a random assortment of rumor mill stories, from the tentative attempts at porting GPUs, to Skybridge, to what would have been 22nm before the dream of having a CPU-dedicated process completely died, etc., it would seem that it's less skipping than having plans and in-progress projects being blown out of the water.
I'm not saying that it would have been possible to fully divine the torturous stagnation that would have resulted ahead of time, but imagine if AMD had given up on a process node until 14/16nm (you know, if it could tell which foundry will stop delaying it first and reach some level of manufacturing maturity). Perhaps a Maxwell-like iteration to the GPU line could have occurred, or an APU on AMD's roadmap might have come in less than a year behind initial promises. Maybe it could have afforded to retain some of the teams it lost.
Rather than have designs or tech that have in-built assumptions about a node transition that makes the changes impractical to backport but still an engineering cost that might not port to the new processes, AMD could have legitimately done very little versus accomplishing little for all the effort expended.

 

[Deleted member 13524]

If one believes a random assortment of rumor mill stories, from the tentative attempts at porting GPUs, to Skybridge, to what would have been 22nm before the dream of having a CPU-dedicated process completely died, etc., it would seem that it's less skipping than having plans and in-progress projects being blown out of the water.

You mean Reverse Hyperthreading isn't coming to life?

 

[Davros]

Analysts claim AMD will be bankrupt by 2020
http://www.fudzilla.com/news/processors/37823-analysts-claim-amd-will-be-bankrupt-by-2020

 

[Rys]

I know it's probably been covered in this thread already, but what happens (if anything!) to Intel as a result of AMD completely dissolving? I've never been truly clear what the impact is to Intel for the only other x86 ISA licensee leaving the market, resulting in no competition.

 

[Florin]

I don't believe anything would happen to Intel at all.

The way I see it is AMD came to prominence because they were once needed by Intel to meet the requirement of availability of a second source as mandated by their contract with IBM. But that contract has long expired.

Being the only competitor in a given space is not necessarily a problem. What is troublesome is maintaining a monopoly through 'exclusionary conduct' or 'predatory acts' and/or some specific ways of leveraging an existing monopoly to gain an advantage when entering a related market.

In other words, Intel would be sitting pretty with their 96% server CPU market share for as long as they behave.

 

[Rys]

I frame the query given their previous anticompetitive behaviour. If they stay clean then I guess it's no issue.

 

[Deleted member 13524]

Erm... how ridiculous can a 5-year-old prediction be for an IHV that will launch a top-to-bottom range of new products within a year?

I know it's probably been covered in this thread already, but what happens (if anything!) to Intel as a result of AMD completely dissolving? I've never been truly clear what the impact is to Intel for the only other x86 ISA licensee leaving the market, resulting in no competition.

When that happens, VIA will rise from the ashes with an APU carrying a Zen-like 8-core multithreaded CPU, HSA'ed with an iGPU callled Savage 4K capable of ~10TFLOPs and 16GB of 1TB/s HBM2.
They haven't done anything in 5 years, so they're definitely onto something big


IMO, if AMD ever goes down, Intel should be forced to sell x86 licenses to other companies.
From the top of my head, IBM, nVidia, apple, Samsung, etc. all of them could be very interested in making their own APUs with x86 cores.

 

[Alatar]

Q1:

http://jonpeddie.com/publications/add-in-board-report