Will Intel eventually buy AMD?

[tynopik]

There are more OEMs than Apple.

If one jumps, it's more likely other will too.

And unhappy ? no not really. Would like better performance? Yep.

Semantics

If someone offers a decent CPU core with AMD-level graphics, Apple (and others) will dump Intel in a heartbeat.

 

[Enigmoid]

But there was no substantially faster 15W CPU at the time the Haswell Airs launched.

 

[sm625]

AMD has nothing to offer Intel. AMD APUs do not deliver any FPS/watt improvement vs discrete GPU, which makes them worthless. They do hold an advantage in FPS/watt over Intel, but it is not enough. If AMD had the FPS/watt that Nvidia has, then they might be worth something. Can you imagine a 13 watt die shrunk maxwell 840M stuffed into every intel cpu? It would more than double their performance. An AMD gpu core would only improve performance by 40%. Not worth.

 

[erunion]

AMD's APU graphics are substantially better than Intel's[/B]

.

False dichotomy.

AMD is bleeding GPU market share to Nvidia. Even if we accept that AMDs graphics are better than intel's, why would Intel spend money to buy AMD when better options exist.

 

[frozentundra123456]

If one jumps, it's more likely other will too.



Semantics

If someone offers a decent CPU core with AMD-level graphics, Apple (and others) will dump Intel in a heartbeat.

Iris pro delivers AMD level graphics and a much better cpu core already, and should improve with Broadwell and Skylake.

 

[tynopik]

Even if we accept that AMDs graphics are better than intel's

is this even a question? you say that like you don't actually believe it

why would Intel spend money to buy AMD when better options exist.

well, AMD's graphics have actually been proven in APUs while Nvidia's haven't plus all of AMD's work in HSA may be handy.

in the end it wouldn't surprise me if they bought either

but i'm glad the question has progressed from if they need help to which help would be better.

that said, AMD may be slightly more affordable ($2.4B mkt cap vs $12.1B)

 

[frozentundra123456]

I have repeatedly addressed every single one of these issues, please try reading the thread again.

I *have* read everything you posted. You make a valid point that AMDs igpu is better than Intel's, except for Iris pro. But you stretch that point to claiming without proof that Intel's igpus "suck" and are not "good enough" whatever that means, and that Apple is "unhappy" with them. Obviously, reality paints a different picture. Intel's igpu is obviously "good enough", since they have the entire x86 cpu market for Apple, and the vast majority of the rest of the cpu market. You are entitled to your own opinion, of course, but those are the facts.

 

[tynopik]

More background on Apple and Intel

http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested

Pure economics and an unwillingness to invest in older fabs made the GPU a first class citizen in Intel silicon terms, but Intel management still didn’t have the motivation to dedicate more die area to the GPU. That encouragement would come externally, from Apple.

Looking at the past few years of Apple products, you’ll recognize one common thread: Apple as a company values GPU performance. As a small customer of Intel’s, Apple’s GPU desires didn’t really matter, but as Apple grew, so did its influence within Intel. With every microprocessor generation, Intel talks to its major customers and uses their input to help shape the designs. There’s no sense in building silicon that no one wants to buy, so Intel engages its customers and rolls their feedback into silicon. Apple eventually got to the point where it was buying enough high-margin Intel silicon to influence Intel’s roadmap.

 

[tynopik]

Iris pro delivers AMD level graphics and a much better cpu core already, and should improve with Broadwell and Skylake.

Crystallwell is an expensive and inefficient hack to cover up its poor graphics architecture.

 

[tynopik]

I *have* read everything you posted.

clearly not

Intel's igpus "suck" and are not "good enough" whatever that means

'Not good enough' is always relative to your competition, because if your competition can do it better, then people will go with them.

Intel's igpu is obviously "good enough", since they have the entire x86 cpu market for Apple, and the vast majority of the rest of the cpu market.

There are several factors that matter: CPU performance, graphics performance, power efficiency, price, etc.

OEMs weigh all these factors when making decisions.

Right now Intel's lead in CPU performance is so great that combined with its other strengths, it outweighs the lack of graphics performance.

But that doesn't mean graphics performance is irrelevant.

Because Intel's competitors are fast approaching it in CPU performance, and once they are within striking distance, the great disparity in graphics performance will tip the scales away from Intel.

 

[erunion]

is this even a question? you say that like you don't actually believe it

well, AMD's graphics have actually been proven in APUs while Nvidia's haven't plus all of AMD's work in HSA may be handy.

in the end it wouldn't surprise me if they bought either

but i'm glad the question has progressed from if they need help to which help would be better.

that said, AMD may be slightly more affordable ($2.4B mkt cap vs $12.1B)

The question isn't which product is faster. The question is ROI.

Absolute Performance isn't the most important metric for Intel. Profit is.
Even a cheap option like licensing is going to hurt margins compared to using their own graphics.

 

[frozentundra123456]

clearly not

So now you are a clairvoyant and know what I have and have not read? Trust me, I have read every post. I just disagree with you. It is as simple as that.

 

[tynopik]

Even a cheap option like licensing is going to hurt margins compared to using their own graphics.

1. Their own graphics aren't free
2. Your base assumption is that Intel will retain near monopoly market share indefinitely.

In such a situation, sure, it doesn't make sense to invest anything in it.

But common sense and history tell us that will not continue forever (or even that long relatively speaking).

When a viable competitor appears and Intel's market share starts sliding, you can be sure that will affect their calculations. How valuable is a graphics solution that keeps your market share at ~75% instead of sliding to ~50% or below?

 

[tynopik]

So now you are a clairvoyant and know what I have and have not read? Trust me, I have read every post.

You either didn't read the post or your reading comprehension is kindergarten level, I was being kind and assumed you didn't read it.

Insulting other members is not allowed
Markfw900

I just disagree with you. It is as simple as that.

No, you specifically said you didn't know what I meant by 'not good enough'

You didn't say you disagreed with my definition of 'not good enough', you instead claimed utter ignorance of what I meant even though it was clearly spelled out.

 

[mrmt]

well, AMD's graphics have actually been proven in APUs while Nvidia's haven't plus all of AMD's work in HSA may be handy.

HSA? The AMD vaporware that not even the guys on the foundation want to adopt? I'm sure Intel is eager to have access to it.

 

[tynopik]

HSA? The AMD vaporware that not even the guys on the foundation want to adopt? I'm sure Intel is eager to have access to it.

*shrug* maybe, maybe not, was just throwing it out as a possibility

 

[Madpacket]

HSA is more compelling than anything Intel has offered as of late. Gotta love all those useful AVX extensions so many apps use today ��

Works both ways.

Anyway this thread is way off topic now. Can an administrator please lock the thread? although I think it'll be entertaining to reference this this thread in a few years to see how right or wrong many of us were.

 

[mrmt]

*shrug* maybe, maybe not, was just throwing it out as a possibility

I don't think AMD has much chance in selling their graphics IP to Intel. Unlike Nvidia, who bifurcated its GPU R&D and now has two GPU architectures AMD has only one, which makes it the second-place in both computing and power efficiency and because of that every GCN product will be a value product. That's not something Intel would want on its portfolio, if it was to invest on GPUs, it would have to invest big, otherwise it's better to keep the status quo.

 

[svenge]

HSA is more compelling than anything Intel has offered as of late. Gotta love all those useful AVX extensions so many apps use today ��

Works both ways.

The key difference is that Intel isn't inventing AVX extensions to paper over an abject failure in x86 performance like AMD is with HSA...

 

[DrMrLordX]

HSA exists on more than just paper. You can actually go out, buy an AMD APU, install a complicated set of drivers and other junk, and roll out HSA applications that will only run on another AMD APU-equipped system with mostly the same software stack as the development machine. There are plenty of GPGPU developers, hobbyists, and suchlike that have bought Kaveri APUs (almost universally 7850ks) to evaluate the chips for HSA and OpenCL work, and it appears that many of these people actually enjoy working with the chips.

Unfortunately, that doesn't amount to many sales of Kaveri on their own, and it positions HSA-capable hardware (notably Kaveri) in an awkward position. I don't know how much driver cruft is necessary on a Windows-based machine configured to run HSA apps, but if you want a Linux system that can do it, it can get a little complicated. That's a huge barrier to entry for adoption of HSA outside of limited use in academia or other niches where HSA appears to have generated the most interest to date. Few software developers want to roll out a new version of their software, carefully crafted after months of hard work using funky tools that are not necessarily developer-friendly, only to discover that the software will only run on a tiny number of desktop machines.

With OpenCL apps, generally speaking, you'll be fine as long as you have up-to-date graphics drivers. HSA is more complicated.

To tie this into the main topic, yes Intel is interested in HSA-like functionality. Gen8 and Gen9 graphics bring us iGPUs with hardware features that are supposed to be geared towards GPGPU (among other things). Unlike AMD, Intel produces nothing but iGPU graphics solutions.

 

[JDG1980]

Unlike Nvidia, who bifurcated its GPU R&D and now has two GPU architectures AMD has only one, which makes it the second-place in both computing and power efficiency and because of that every GCN product will be a value product. That's not something Intel would want on its portfolio, if it was to invest on GPUs, it would have to invest big, otherwise it's better to keep the status quo.

Which other Nvidia GPU architecture (besides Maxwell) are you referring to?

Anyway, while it's true that GCN's current iteration is not as efficient as Maxwell, the full node advantage that Intel currently enjoys would almost certainly be enough to make up for it. Besides, there's no reason to think that the GCN architecture is inherently flawed; we're not talking about Netburst or Bulldozer here, but about a graphics architecture that gives competitive performance in real-world apps, and just needs some TLC from R&D to increase efficiency and reduce wasted power usage.

 

[mrmt]

Which other Nvidia GPU architecture (besides Maxwell) are you referring to?

Anyway, while it's true that GCN's current iteration is not as efficient as Maxwell, the full node advantage that Intel currently enjoys would almost certainly be enough to make up for it. Besides, there's no reason to think that the GCN architecture is inherently flawed; we're not talking about Netburst or Bulldozer here, but about a graphics architecture that gives competitive performance in real-world apps, and just needs some TLC from R&D to increase efficiency and reduce wasted power usage.

Nvidia released a "little Kepler" to address the consumer market and a "big Kepler" to address the big compute players and the high end consumer market. They are poised to repeate the recipe with Maxwell, with the Titan Maxwell launching this year. This allows Nvidia to address both markets without worrying about the constraints of those, while AMD must address the compute and the consumer market with the same chip, which means lower compute power when compared with the big Nvidia parts and less efficiency when compared to the mainstream cards.

 

[VirtualLarry]

Nvidia released a "little Kepler" to address the consumer market and a "big Kepler" to address the big compute players and the high end consumer market. They are poised to repeate the recipe with Maxwell, with the Titan Maxwell launching this year. This allows Nvidia to address both markets without worrying about the constraints of those, while AMD must address the compute and the consumer market with the same chip, which means lower compute power when compared with the big Nvidia parts and less efficiency when compared to the mainstream cards.

And for its big HPC customers, it's releasing "Really Big Kepler" too, even though they have Maxwell. Apparently, Kepler is still better for compute than Maxwell. At least, that's the only reason that I can see to re-release an even bigger Kepler for compute customers, when they have Maxwell.

 

[CHADBOGA]

If one jumps, it's more likely other will too.

And what operating system will these others be using? 😵

 

[tynopik]

And what operating system will these others be using? 😵

if it's amd that becomes competitive with zen, windows

if it's arm that becomes competitive, chrome

and windows for arm is still around, ms could work on that if they wanted to