Apple's ongoing use of ImgTec PowerVR GPU IP

They're obviously serious about this...

 

Most developers use additional compression on top of GPU file formats (such as LZMA, ZLIB, etc) and their own custom data packaging formats. It would be hard to find all textures from a big binary blob (custom package directory format that is now known to Apple), especially if some of that data is compressed with third party compression tools. And even if you found that data, it would be hard to replace it, because you'd have to be able to modify the developer's custom data package format. Unless of course the new compression format produces exactly same size images (ASTC compression ratio is different from PVRTC). Not an easy task.

Also, I am unsure how transcoding on a server would avoid the PVRTC license. Do you only need a license for local encoding (on device), but not when you are offloading encoding to a server? There's too much speculation. Nobody seems to have read the patent, so nobody knows what it covers.

 

Probably because they couldn't get to an agreement on how much IMG would be worth.


Maybe because they already have something lined up with Mediatek.

 

Considering MediaTek's general low GPU performance, someone at MediaTek probably wish that had come through too.

It's not too late for MediaTek to buy IMGTec though.

Cheers

 

It's entirely possible (and has happened before) for the offer to be materially different from the market cap. In fact, it usually is. The value is still abstracted into a share price, but it doesn't have to be based on it. There is no current offer on the table for Imagination, and I don't expect there to be an acquisition in the immediate future. For this, we blame leadership.

 

By "they" I meant apple and IMG's shareholders.

I mentioned mediatek because they may have approached IMG's management with a better purchase deal, and this information was passed on to key shareholders.

 

As far as I know there were no other offers for IMG. From what I've seen in talks with prospective new clients (us included), they should be able to quickly make up most of the lost revenue within the next two years.

 

There's speculation that whatever GPUs Apple develop, they will touch on all the current buzzwords, i.e. machine learning, computer vision, etc.

 

Well if it's truly in-house they'll prioritise accordingly. If AR is really as important as they say, they'll be optimising their GPU for that. I'm still having a hard time picturing Apple AR hardware...

 

Presupposition that their stock value recovers slowly.

There's nothing missing considering machine learning, computer vision etc. in IMG's GPU IP; au contraire they're well ahead their direct competition in the ULP mobile space. As for anything VR/AR I have the same hard time as you for it. In fact I'll be even more bold here and will claim extremely surprised if Apple has developed anything that is well ahead in terms of efficiency compared to their current licensed GPU IP.

I quickly searched for Mc Combe since he had left IMG, but his current job doesn't seem to trace any rays at all https://www.linkedin.com/in/jamesmccombe/

 

Once upon a time, Imagination shrewdly gained a head start in a market with massive potential, GPUs for pocket computers, wielding their best-in-class processor designs. They quickly dominated this rapidly growing market, appearing in the SoC implementations of most of the major players like Samsung and TI and even serving as ARM's own GPU reference design for a while.

As Imagination looked out onto the exploding landscape of pocket and embedded processing technology and took account of the advantageous position they had earned, they decided to chart a course to own a broad expanse of it. They would offer hardware IP, associated software IP, tool sets, ecosystems, platforms, and implementation services for every kind of processing from display algorithms to imaging to communication protocols to VOIP to microcontrollers and even a grab at CPUs.

The fundamental understanding for targeting efficiency that their engineers possessed in designing processors unfortunately didn't extend to their executive's approach to growing the company. They were always stretching a little too far a little too fast, never booking the profits a company with such a large piece of a such a strong market should've. They had justifiably high expectations and valuations for themselves, but they didn't adapt those to the inevitable consolidation that retrenched the opportunity available in the market as fiercely as it had once expanded. Whereas the prevailing logic once was to not sit out the multitude of opportunities offered by the wider, open market of licensees, the game of musical chairs that suddenly established itself within the GPU IP market necessitated grabbing the best seat available and not leaving any room to be squeezed off.

The greatest benefit of Apple's success is that they can afford to remain beholden to no one. They not only know this, they know all the long term benefits to competitiveness, flexibility, and differentiation it brings outweighs any short term savings they'd get from settling on patent disputes, licensing arrangements, and the like. Ownership of the GPU technology on which they depend is the driving motivation of their actions here, not the idea that some future architecture they have in mind can be done best by some team they've pieced together in recent years compared to the design expertise of their long time, world-class partner.

Once upon a time, the openness of the market was such that Apple couldn't have expected a complete and total commitment from Imagination, but Apple eventually represented not only the best possible showcase for their technology but also the best market opportunity. Letting Apple take ownership for a fair price would've been a no-brainer.

Apple should still buy Imagination for whatever price they were offering before. I don't think Apple's internal GPU has much of a chance at being competitively efficient, yet I don't think they care at first. I think their first goal is simply to own their technology, and they feel they can iterate as long as they need until they've got a competitive offering.

 

I agree - history has shown that technical metrics (perf/w, perf/mm2) matter less to the mobile SoC space than bundling, integration, and market position. Consumers aren't in a position to differentiate between mobile devices based on GPU or even SoC. The software isn't there for them even to notice much about the efficiency or technical sophistication of a component like a GPU. People will buy the next iPhone regardless of its technical components.

 

Having seen IMG's IP and roadmap for the next 4 years, I'd agree with the sentiment that they have the best in-class technology in terms of efficiency and performance. Whether anyone uses it outside of Apple is a different story. However, I would bet on Apple at least being competitive in that regard.

But that begs the question - why not acquire IMG if their in-house solution is <10% better (just a random number)? It just seems like a ton of effort just to make something slightly better. Not buying Imagination speaks volumes about their belief in their own solution's competitiveness. Also, I don't think they can really buy IMG anymore (regulatory oversight in EU + the share price drop/antitrust).

 

It's hard for me to imagine they would notify IMG before getting their own solution at least to a competitive point.

What do they gain from acquiring IMG, even if they aren't massively better? Patents basically? Would they have ongoing obligations to other licensees?

 

I don't think IMG ever wanted to get bought in the first place. Apple's attempt to acquire them was officially mentioned after IMG's CEO resigned. Aren't the times of hostile takeovers long gone or could it be there are legal issues that keeps Apple from forcing an acquisition? (yes now as you say obsiously but I mean before that....).