Apple To Develop Own GPU, Drop Imagination's GPUs From SoCs
by Ryan Smith on April 3, 2017 6:30 AM ESTThe Future: Competition, Secrecy, & the Unexpected
Finally, while Apple developing their own GPU is not unexpected given their interests and resources, the ramifications of it may very well be. There hasn’t been a new, major GPU vendor in almost a decade – technically Qualcomm’s team would count as the youngest, though it’s a spin-off of what’s now AMD’s Radeon Technologies Group – and in fact like the overall SoC market itself, the market for GPU vendors has been contracting as costs go up and SoC designers settle around fewer, more powerful GPU vendors. So for someone as flush with cash as Apple to join the GPU race is a very big deal; just by virtue of starting development of their own GPU, they are now the richest GPU designer.
Of course, once they start shipping their custom GPU, this will also open them up to patent challenges from those other players. While it has largely been on the backburner of public attention, this decade has seen a few GPU vendors take SoC vendors to court. This includes NVIDIA with Samsung and Qualcomm (a case that they lost), and still ongoing is AMD’s case against LG/MediaTek/Sigma/Vizio.
GPU development is a lot more competitive due to the fact that developers and compiled programs aren’t tied to a specific architecture – the abstraction of the APIs insulates against individual architectures – however it also means that there a lot of companies developing novel technologies, and all of those companies are moving in the same general direction with their designs. This potentially makes it very difficult to develop an efficient GPU, as the best means of achieving that efficiency have often already been patented.
What exists then is an uneasy balance between GPU vendors, and a whole lot of secrets. AMD and NVIDIA keep each other in check with their significant patent holdings, Intel licenses NVIDIA patents, etc. And on the flip side of the coin, some vendors like Qualcomm simply don’t talk about their GPUs, and while this has never been stated by the company, the running assumption has long been that they don’t want to expose themselves to patent suits. So as the new kid on the block, Apple is walking straight into a potential legal quagmire.
Unfortunately, I suspect this means that we’ll be lucky to get any kind of technical details out of Apple on how their GPUs work. They can’t fully hide how their CPUs work due to how program compilation works (which is why we know as much as we do), but the abstraction provided by graphics APIs makes it very easy to hide the inner-workings of a GPU and make it a black box. Even when we know how something works, features and implementation details can be hidden right under our noses.
Ultimately today’s press release is a bit bitter-sweet for all involved in the industry. On the one hand it absolutely puts Imagination, a long-time GPU developer, on the back foot. Which is not to spell doom and gloom, but the company will have to work very hard to make up for losing Apple. On the other hand, with a new competitor in the GPU space – albeit one we’ve been expecting – it’s a sign that things are about to get very interesting. If nothing else, Apple enjoys throwing curveballs, so expect the unexpected.
Facebook
Twitter
RSS
144 Comments
View All Comments
Meteor2 - Tuesday, April 4, 2017 - link
What name99 said. Which is awfully like what Qualcomm is doing, isn't it? A bunch of conceptually-different processor designs in one 'platform'. Software uses whichever is most appropriate.peevee - Tuesday, April 18, 2017 - link
It is certainly easier to design your own ISA than to build your own core for somebody else's ISA. And ARM64 us FAR from perfect. So 1980s.quadrivial - Monday, April 3, 2017 - link
Very unlikely. They gave up that chance a couple years ago (ironically, to imagination).Consider, it takes 4-5 years from initial architecture design to final shipment. No company is immune to this timeframe no matter how large. Even more time is required for new ISAs because there are new, unexpected edge cases that occur.
Consider, ARM took about 4 years to ship from the time the ISA was announced. Most of their licensees took closer to 5 years. Apple took a bit less than 2 years.
Consider, Apple was a front-runner to buy MIPS so they could have their own ISA, but they backed out instead. The new ARM ISA is quite similar to MIPS64.
Thought, Apple started designing a uarch that could work well with either MIPS or their new ARMv8. A couple years in (about the time nailing down the architecture would start to become unavoidable), they show ARM a proposal for a new ISA and recommend ARM adopt that ISA otherwise they buy MIPS and switch. ARM announces a new ISA and immediately has teams start working on it, but Apple has a couple year head start. Apple won big because they shipped an extremely fast CPU a full two years before their competitors and even more years for their competitors to catch up.
Maybe imperfect, but its the best explanation I can come up with for how events occurred.
TheMysteryMan11 - Monday, April 3, 2017 - link
Computing still heavily relies on CPU for all the things that matter to power users. ARM is long way away from being powerful enough to actually be useful for power users and creators.It is good enough for consumption and getting better.
But then again, Apple hasnt been doing well catering to creators anyway. Still no refresh for Mac Pro. So you might be right. But that means Apple is Ok with ignoring that segment, which they probably are.
lilmoe - Monday, April 3, 2017 - link
Single purpose equipment aren't mainly CPU dependent. This is my point. Relying on the CPU for general purpose functionality is inherently the least efficient, especially for consumer workloads.Outside the consumer market, for example engineering and video production software, are still very CPU dependent because the software isn't written efficiently. It's so for the sole purpose of supporting the most amount of currently available hardware. I'd argue that if a CAD program was re-written from the ground up to be hardware dependent and GPU accelerated ONLY, then it would run faster and more fluidly on an iPad than on a Core i7 with integrated graphics, if the storage speed was the same.
This leaves only niche applications that are inherently dependent on a CPU, and can't be offloaded to hardware accelerates. With more work on efficient multi-threaded coding, Apple's own CPU cores, in a quad/octa configuration, can arguably suffice. Single-threaded applications are also arguably good enough, even on A72/A73 cores.
Again, this conversation is about consumer/prosumer workloads. It's evident that Apple isn't interested in Server/corporate workloads.
This has been Apple's vision since inception. They want to do everything in-house as a single package for a single purpose. They couldn't in the past, and almost went bankrupt, because they weren't _big enough_. This isn't the case now.
The future doesn't look very bright for the personal computing industry as we know it. There has been talk and rumors that Samsung is taking a VERY similar approach. Rumors started hitting 2 years ago that they were also building their own in-house GPU, and are clashing with Nvidia and AMD graphics IP in the process. It also lead Nvidia to sue Samsung for reasons only known behind the scenes.
ddriver - Monday, April 3, 2017 - link
Yeah let's make the x chip that can only do one n task. And while we are at it, why not scrap all those cars you can drive anywhere there is an open road, and make cars which are best suited for one purpose. You need a special car to do groceries, a special car to go to work, a bunch of different special cars when you go on vacation, depending on what kind of vacation it is.Implying prosumer software isn't properly written is laughable, and a clear indication you don't have a clue. That's the only kind of software that scales well with cores, and can scale linearly to as many threads as you have available.
But I'd have to agree that crapple doesn't really care about making usable hardware, their thing is next-to-useless toys, because it doesn't matter how capable your hardware is, what matters is how much of that lacking and desperately needed self esteem you get from buying their branded, overpriced toy.
Back in the days apple did good products and struggled to make profit, until genius Jobs realized how profitable it can be to exploit dummies and make them even dummer, giving rise to crapple, and the shining beacon of an example that the rest of the industry is taking, effectively ruining technology and reducing it to a fraction of its potential.
lilmoe - Monday, April 3, 2017 - link
Chill bro.I said current software is written to support the most amount of hardware combinations possible. And yes, that's NOT the most efficient way to write software, but it _is_ the most accessible for consumers.
I wasn't implying that one way is better than the other. But it's also true that a single $200 GPU runs circles around $1500 10 core Intel CPU in rendering.
steven75 - Monday, April 3, 2017 - link
How amazing that an "overpriced toy" still shames all Android manufacturers in single thread performance. The brand new S8 (with a price increase, no-less) can't even beat a nearly 2 year old iPhone 6S.I wish all "toys" were superior like that!
fanofanand - Monday, April 3, 2017 - link
How many people are working (like actual productivity) on an S8? Cell phones are toys 99% of the time.FunBunny2 - Monday, April 3, 2017 - link
-- How many people are working (like actual productivity) on an S8? Cell phones are toys 99% of the time.I guess Mao was right.