[Fudzilla] Apple 2016 laptops will have AMD GPUs

[sirmo]

I'd say CUDA is more likely to disappear if it doesn't get open sourced than anything.

 

[renderstate]

This is the main page for Metal. It's supported on all their OSes now, with the exception of watchOS. They highlight compute abs graphics and OpenCL is completely absent from this new effort. I haven't spent much time checking but I couldn't even find if Metal supports interop with OpenCL.

https://developer.apple.com/metal/

 

[ZGR]

Also if Apple were to move to AMD APUs for their non-discrete-graphics devices, they could simplify their driver development as they would only be developing for one architecture instead of 2.

That would be huge. There has yet to be an Apple computer running on an AMD CPU yet, besides Hackintoshes. The Intel-Apple relationship has been quite solid overall.

 

[renderstate]

Based on what? Metal uses OpenCL for compute pipeline exactly in the same way as Vulkan will use OpenCL 2.0/2.1 for compute pipeline rendering in games.

Metal doesn't use OpenCL. You can read it in the pages I linked above.

 

[Glo.]

This is the main page for Metal. It's supported on all their OSes now, with the exception of watchOS. They highlight compute abs graphics and OpenCL is completely absent from this new effort. I haven't spent much time checking but I couldn't even find if Metal supports interop with OpenCL.

https://developer.apple.com/metal/

https://streamcomputing.eu/blog/2015-06-08/apple-metal-versus-vulkan-opencl-2-1/

Apple Metal in one sentence: one queue for both OpenCL and OpenGL, using C++11.

 

[KompuKare]

It soured the relationship. Because for awhile Nvidia wouldn't own the issue. But yeah I've mentioned OpenCL among other things.

Not sure what you mean by own? Yes, eventually they had to tell their OEMs something. But for retail customers of their cards and motherboards containing their chipsets there was nothing.
Well nothing voluntary anyhow: in North America there was a class-action but it was a miserly settlement ("sorry we broke your $1000+ laptop, here's a $250 one, now go away"). In Europe there was nothing.
I lost or personally saw:

• a BFG 8800GT (coincidental BFG went bankrupt around this time although in Europe they only had a short 2-3 year warranty AFAIK: anyway by the time my card broke they were gone);
• my brother lost an Immo3D 8800GT;
• I also have seen at least 3-4 dead nForce 6100 mobos (the last one died only a few months ago so not so bad but similar Core2 mobos with Intel chipsets are still going strong);
• a 8400GS MXM whose demise made this AiO unusable. Luckily was able to salvage most of those parts and rebuild though. Ironically that machine had a perfectly good Intel chipset mobo but no way to enable it and disable the MXM card thanks to the way (Nvidia) designed MXM.
• a bunch of laptops which I saw.

So yes, not good and I've mostly avoided Nvidia since then.

 

[sirmo]

Not sure what you mean by own? Yes, eventually they had to tell their OEMs something. But for retail customers of their cards and motherboards containing their chipsets there was nothing.
Well nothing voluntary anyhow: in North America there was a class-action but it was a miserly settlement ("sorry we broke your $1000+ laptop, here's a $250 one, now go away"). In Europe there was nothing.
I lost or personally saw:

• a BFG 8800GT (coincidental BFG went bankrupt around this time although in Europe they only had a short 2-3 year warranty AFAIK: anyway by the time my card broke they were gone);
• my brother lost an Immo3D 8800GT;
• I also have seen at least 3-4 dead nForce 6100 mobos (the last one died only a few months ago so not so bad but similar Core2 mobos with Intel chipsets are still going strong);
• a 8400GS MXM whose demise made this AiO unusable. Luckily was able to salvage most of those parts and rebuild though. Ironically that machine had a perfectly good Intel chipset mobo but no way to enable it and disable the MXM card thanks to the way (Nvidia) designed MXM.
• a bunch of laptops which I saw.

So yes, not good and I've mostly avoided Nvidia since then.

By not owning [the issue] I mean not admitting there was an issue with how GPUs were packaged (the IC packaging, not the product box packaging).

Early life failure rate was as high as 40% on some products and they refused to acknowledge it for a long time.

 

[Gikaseixas]

Plus AMD is willing to sacrifice other opportunities so that they can secure the Apple contract, something Nvidia is less inclined to do given it's market position

 

[caswow]

yea its a good thing not beeing in any of apples products.

 

[Bacon1]

Plus AMD is willing to sacrifice other opportunities so that they can secure the Apple contract, something Nvidia is less inclined to do given it's market position

Sacrifice other opportunities? Since when is securing a contract with one of the most profitable companies around get turned into having to sacrifice? Why would you turn down good known money instead of going with 3rd party sellers?

 

[Elixer]

Even if AMD chips are found in just 25% of Macs, AMD has a potential of banking 5M+ AMD Polaris chips just from the Apple contract.

That would require AMD to have multiple fabs for polaris would it not?
They would need to pump out 250K+ chips per day to meet demand for back to school PC from Dell, HP, & Lenovo, and then you got Apple, and on top of that, you also have the 'upgrade' crowd, and consoles.

That would mean, they need at least 2-3 fabs...

 

[VulgarDisplay]

That would require AMD to have multiple fabs for polaris would it not?
They would need to pump out 250K+ chips per day to meet demand for back to school PC from Dell, HP, & Lenovo, and then you got Apple, and on top of that, you also have the 'upgrade' crowd, and consoles.

That would mean, they need at least 2-3 fabs...

They have two fabs don't they?

Glofo and Samsung?

Sent from my XT1575 using Tapatalk

 

[sirmo]

That would require AMD to have multiple fabs for polaris would it not?
They would need to pump out 250K+ chips per day to meet demand for back to school PC from Dell, HP, & Lenovo, and then you got Apple, and on top of that, you also have the 'upgrade' crowd, and consoles.

That would mean, they need at least 2-3 fabs...

We don't know how many wafers GloFo can produce. Back in AMD's hay day, they produced tons, and GloFo is bigger today then they were under AMD.

But they do also have a contract with Samsung as well.

 

[Elixer]

They have two fabs don't they?

Glofo and Samsung?

Sent from my XT1575 using Tapatalk

They are under contract with Gloflo to make X chips with them, but, I haven't heard anything about AMD dual sourcing any chips.
While Gloflo is using Samsung tech, it hasn't been mentioned anyplace that I can find that they are indeed using multiple fabs at this point.

 

[Enigmoid]

That's a key point right there. People keep criticizing AMD's strategy but when we look back at historical market share (post #174), contrary to what's being repeated over and over, AMD lost a lot of market share when 750/750Ti launched, well before 970/980 even showed up.

2015 Mac Sales = 20.6M units (it was almost 19M in 2014)
http://www.statista.com/statistics/263444/sales-of-apple-mac-computers-since-first-quarter-2006/

Even if AMD chips are found in just 25% of Macs, AMD has a potential of banking 5M+ AMD Polaris chips just from the Apple contract.

Some people seem to be [H] confused about AMD's strategy this time. They are trying to launch products without having direct competition from NV. This is why we should expect Vega to either blow the doors off 1080 in price/performance or flat out beat it. I am sure NV is loving this too since it means AMD has forfeited the entire $300+ dGPU market for the next 5-6 months, allowing NV to milk GP104 uncontested with ludicrous marketing gimmicks like the $70-100 FE cards. Both companies are benefiting by not directly competing with each other. Let's just hope Vega is the next X1900XT series, not X1800 series.

The vast majority of macs do not have discrete GPUs.

The only ones with discrete are the very high end laptops (upgraded 15") and upgraded imacs. Needless to say the vast majority of apple's sales are their 13" notebooks. It is quite likely that less than 5% of macs ship with a discrete GPU.

I think that it is nice that AMD will get the contracts but I do not expect this to have a tremendous impact to sales.

 

[Yakk]

If TRG need to reach these kinds of production volumes, good thing Polaris are small chips then.

 

[sirmo]

The vast majority of macs do not have discrete GPUs.

The only ones with discrete are the very high end laptops (upgraded 15") and upgraded imacs. Needless to say the vast majority of apple's sales are their 13" notebooks. It is quite likely that less than 5% of macs ship with a discrete GPU.

I think that it is nice that AMD will get the contracts but I do not expect this to have a tremendous impact to sales.

I think it's more than 5%. Mac Pros are pretty popular. And so are iMacs. You have to understand all the macs with retna display had an AMD GPU. And retna was a pretty big selling point.

 

[Yakk]

Is there still a rumor of new standalone 5K screens coming with an integrated GPU?

 

[Krteq]

They would need to pump out 250K+ chips per day to meet demand for back to school PC from Dell, HP, & Lenovo, and then you got Apple, and on top of that, you also have the 'upgrade' crowd, and consoles.

That would mean, they need at least 2-3 fabs...

So, what do you thing AMD is doing since last autumn. Yes, they had working chips in autumn and yes, they are producing lots of chips since then.

They clearly said at the end of the year that they have a lot of design wins with Polaris and they will launch cards in the midyear before back to school season.

 

[Elixer]

We don't know how many wafers GloFo can produce. Back in AMD's hay day, they produced tons, and GloFo is bigger today then they were under AMD.

But they do also have a contract with Samsung as well.

We have a rough idea...

Today, Samsung has roughly 11,000 wafer starts per month (wspm) of 14nm capacity, which represents about 10% of its total 300mm fab capacity, according to Pacific Crest Securities. Over time, Samsung is expected to convert some of its 28nm capacity, giving it a total of 46,000 wspm of 14nm capacity, according to the firm.

Another foundry vendor, GlobalFoundries, licensed Samsung’s 14nm finFET process some time ago. In its New York fab, GlobalFoudries is capable of ramping up around 30,000 wspm of 14nm finFET capacity, according to Pacific Crest Securities. Later this year, GlobalFoundries will move into finFET production.

Not to be outdone, TSMC will begin volume production for its 16nm finFET process by the middle of this year. By the end of 2016, the company plans to have an installed capacity of 100,000 wspm for 16nm finFET technology, according to J.K. Wang, vice president of 300mm fab operations at TSMC.

Samsung has 3 14nm fabs, 2 in south Korea, and 1 in the USA, and Gloflo has the NY Fab.

Not sure where you found out about a contract with Samsung for AMD to use their fabs, I can't find any evidence besides using GloFlo.

 

[sirmo]

There was a news story circulating while back: http://www.techpowerup.com/218578/samsung-to-fab-amd-zen-and-arctic-islands-on-its-14-nm-finfet-node

 

[Elixer]

So, what do you thing AMD is doing since last autumn. Yes, they had working chips in autumn and yes, they are producing lots of chips since then.

They clearly said at the end of the year that they have a lot of design wins with Polaris and they will launch cards in the midyear before back to school season.

Source of that information?
If indeed they had been mass producing chips, it would have been in their financials, and it wasn't mentioned in either quarters that they have had since then.
So, I don't buy they have been producing chips since last autumn.

 

[LTC8K6]

So which AMD gpu is going to get full Tonga-ed out of our hands?

 

[sirmo]

So which AMD gpu is going to get full Tonga-ed out of our hands?

If Apple refreshes the iMac it might be the rx480x or whatever its name will be. But it hasn't even been a year since 5k iMac was launched so they might wait on the refresh till next year.

 

[Elixer]

There was a news story circulating while back: http://www.techpowerup.com/218578/samsung-to-fab-amd-zen-and-arctic-islands-on-its-14-nm-finfet-node

Yes, they don't site sources either, so it is a rumor.
http://www.businesskorea.co.kr/engl...ung-electronics-benefits-14nm-finfet-contract

Samsung Electronics will earn considerable profits from the foundry contract between AMD and GF. Last year, Samsung Electronics established a strategic partnership with GF to provide the 14nm FinFET process technology in order to secure clients in North America, including Qualcomm and Apple.

Meanwhile, some say that Samsung Electronics can manufacture some AMD products, which are to be produced by GF, at the plant in Austin, depending on GPU productions. If there is a problem with the production yield rate of 14nm chips of GF, Samsung Electronics can produce them. Since its competitor NVIDIA uses TSMC’s 16nm process to produce the Pascal architecture-based next-generation GPU, AMD is expected to stick with the 14nm process of GF and Samsung Electronics in order to put differentiation.

So, there is no solid evidence of this happening yet.