Imagination Technologies up for sale [2017]

[tangey]

Apple seems to make an assertion that they no longer have to pay licenses, period. Not using anything new doesn't mean not using anything old.

That wasn't the point I was making. The point I was making was that any improvements in GPUs inside their socs since that time, were not coming from IMG. I note that in response, IMG have not contested this, only that Apple should have told anyone (breaking confidentialty).

That would be consistent with a view that far from just some customisations and driver development, Apple have been making significant changes to the GPU hardware over that time,entirely independant of the design path that IMG was taking for it's Rogue / Furian line of GPUs, A10 and A10x would fall into that category.

Given the more detailed response from Apple, i.e. the request in reduction of payments due to using less IMG IP, one wonders has apple already totally designed out IMGs video encode/decode IP blocks, that they also used since the earliest A series chips.

 

[tangey]

Apples response was quite reveailing, and gave some insight into their motivations for moving to their own solution.
Regarding PVRTC, you gave a couple of options already, but omitted the simplest - that Apple already have paid for an unlimited license for its use.

I had not thought about that option. However, although it would certainly be a 'fit', it's not entirely in keeping with the statement from Apple, in which they assert they won't be using any IMG IP. The scenario would be that they would still be using IMG IP, just that ongoing payments would not be required, which I admit is a subtle difference.

 

[Lodix]

That wasn't the point I was making. The point I was making was that any improvements in GPUs inside their socs since that time, were not coming from IMG. I note that in response, IMG have not contested this, only that Apple should have told anyone (breaking confidentialty).

That would be consistent with a view that far from just some customisations and driver development, Apple have been making significant changes to the GPU hardware over that time,entirely independant of the design path that IMG was taking for it's Rogue / Furian line of GPUs, A10 and A10x would fall into that category.

Given the more detailed response from Apple, i.e. the request in reduction of payments due to using less IMG IP, one wonders has apple already totally designed out IMGs video encode/decode IP blocks, that they also used since the earliest A series chips.

The fact that they haven't licensed anything "new" since 2015 doesn't mean anything. IMG releases new architectures every few years and the improvements since then doesn't mean Apple has changed the architecture itself but just taking advantage from the improvements of new process nodes, packaging, better desing of the SOC... I am just giving another point of view I could be wrong.

 

[tangey]

EEtimes has today published a long article from Jon Peddie, on the IMG/Apple dispute, and more widely on the perils of being an apple supplier.
http://www.eetimes.com/document.asp?doc_id=1332007

His view is that:-

Imagination is looking into the possibilities of patent infringement, and it probably will find it.

Which is an interesting quote from someone who has been a high-profile observer of the graphics industry for a long time.

He also concludes similarly to me, in that one possible outcome is that a buyer with good finanical resources will be doing so with a view to pursue any such infringement.

 

[Entropy]

I hoped that someone else would respond.
That is a terrible article. It deals more with Apple than anything else, but doesn' t even really discuss either Apple or vertical integration, which its title says is the subject.
And yes, Jon Peddie is a high profile observer of the graphics industry in terms of market. Shares and trends, IGPs vs. discrete, and such. Not graphics technology.

That quote you found interesting refers to nothing but Andrew Heaths' statement, that led to Apples rather revealing rebuttal.
It doesn't deal with those facts that are unassailable in that rebuttal, doesn't make any suggestion as to what any IP-infringement would actually be, doesn't ask the question why Apple wouldn't simply license any such technical specifics, nor of course what monetary value might be attached to any such IP. Nor does it mention that Apple is in a unique position in that it not only would control the processing elements, but also the API accessing them, making compatibility mostly a non-issue. He seems to think Apple hiring ImgTech personnel is some novelty, when the press was awash in articles five years ago when they hired entire groups of engineers from AMD.

Instead he goes into some rather meandering and superficial discussion about the perils of vertical integration, but doesn't keep it focussed on suppliers, but rather makes some vague statements about it not supporting industry development, but ironically brings up examples where Apple has driven the industry as a whole towards better integrated graphics, fails to recognize how they have driven the lithographic process spending of the semiconductor fabs benefitting the entire industry and so on. It recognizes that other players have their own graphics IP, but doesn't discuss why it makes sense for Apple to do the same. Read for instance this blog post about Bosch building their own fab on SemiWiki for a better one minute paragraph of how the industry dynamics have been shifting around the last decades.

So - nothing new, and rather confused and shallow at that.

 

[Entropy]

Also in the interest of explaining where Apple is heading, I've seen a number of forum reports from Apple developer conference attendees that say the same things. A short example would be this

Apple doesn't want a GPU, they want a throughput engine to complement their existing latency engine. This is obvious when you see everything they were pushing at WWDC --- all about computation photography, machine learning, pouring data through image filtering to recognizers to renderers that display the data that's been acquired in new ways and aggregated with streams of data acquired from other sources like the camera.
They're way beyond interest in engines whose primary skill is in processing OpenGL!

A throughput engine done right, and that they control, means among other things:

- code can be shifted from the latency engine to the throughput engine and back as easily as it is shifted from the performance to the low-energy core. This would, in the best case, require the throughput engine to implement some (subset of) AArch64, but could be done at a little more hassle with a different instruction set and just sharing address space and coherency.

- being able to ramp the frequency of the throughput core up and down, like that of the latency core, so that power can be balanced across both the way Intel now does.

Now this contains a bit of speculative context, but the examples brought up are common ones, coupled with how Apple wants to maximize local processing in order to increase security and keep the need for cellular network traffic from ballooning.
Also it doesn't bring up the problems of being beholden to another company for a critical component in terms of processing, power draw, API construction and so on. Nor the security risks, nor competitive aspects and so on.
Apple has very good reasons to go their own way.

 

[Shifty Geezer]

What's a 'latency engine'? I guess 'throughput engine' is GPU compute type wide workloads. 'Latency engine' is some super low latency, lower performance processing? Are we talking parallel DSP sort of thing alongside a CPU? How does this fit in with anything anyone is making?

 

[tangey]

I hoped that someone else would respond.
That is a terrible article. It deals more with Apple than anything else, but doesn' t even really discuss either Apple or vertical integration, which its title says is the subject.
And yes, Jon Peddie is a high profile observer of the graphics industry in terms of market. Shares and trends, IGPs vs. discrete, and such. Not graphics technology.

That quote you found interesting refers to nothing but Andrew Heaths' statement, that led to Apples rather revealing rebuttal.

The single quote I used was quoting Jon Peddie's narrative, the full paragraph reads as:-

Apple's abandonment of Imagination Technologies' GPU, in favor of an in-house design is bold, and questionable. Imagination is looking into the possibilities of patent infringement, and it probably will find it.

The paragraph which followed in the article WAS an excerpt from an IMG PR on the subject, but was clearly presented as such. The quote I used is the author's own words, and so I take it as his view (i.e. it is also his view that it is bold and questionable, which is also not commentary from IMG), not relaying an IMG view.

Regardless of his overall standing in the graphics community, which I suggest is subjective, it is the first quote I have seen from outside of IMG, from someone in the industry at any level,that suggests that IMG will likely find infringement. You could read the quote as suggesting they'll find it whether it's there or not, but that's not how I read it.

 

[digitalwanderer]

Dumb question, but why doesn't Apple just do an out of court settlement and be done with it? I was under the impression that had quite a bit of monies at the moment and don't get why they don't just settle since whatever the amount it'll probably be a pittance to them?

 

[psurge]

What's a 'latency engine'?

A 'latency engine' is basically a high performance CPU - a processor optimized for low execution latency of one task. So for example, instead of switching to another thread when the current one is stalled, like a GPU would, a latency engine might choose to speculatively execute past an unresolved branch, and/or to spend significant power on finding independent instructions from the current thread that can be executed (OOOE). It's what you want to run the serial bits of computation graphs on, so that their performance impact is minimized.

 

[Entropy]

Dumb question, but why doesn't Apple just do an out of court settlement and be done with it? I was under the impression that had quite a bit of monies at the moment and don't get why they don't just settle since whatever the amount it'll probably be a pittance to them?

Easy answer - because there is no reason for them to have any out of court settlement.

They have been working on this for a good long while and was probably in development/planning before the hiring spree of AMD graphics engineers. (At which time for instance Raja Koduri was already working at Apple. Apple hired him and AMD GPU CTO Bob Drebin back in 2009. What do you think these people have been doing?)

 

[BRiT]

Dumb question, but why doesn't Apple just do an out of court settlement and be done with it? I was under the impression that had quite a bit of monies at the moment and don't get why they don't just settle since whatever the amount it'll probably be a pittance to them?

Pride.

 

[digitalwanderer]

Pride.

To that I give the only appropriate answer.

Respect.

 

[manux]

It's not necessarily pride. If apple truly doesn't infringe there is nothing to settle.

 

[nutball]

They also have a duty to their shareholders not to throw money away.

 

[tangey]

That quote you found interesting refers to nothing but Andrew Heaths' statement, that led to Apples rather revealing rebuttal.

Jon peddie's contribution in the comments of the article confirms the point from my previous posting.

GPUs are tricky things, and the companies that did the foundation work (ATI, 3Dlabs, Nvidia, and Imagination), know that and can safely assume a new company (Apple in this case) would have to viloate some IP. But it makes the lawyers rich in teasing out what IP has been viloated, and very often (85% of the time is my guess) its an innocent hapenstance - there are just so many ways to design a shader, etc.

 

[Dominik D]

That wasn't the point I was making. The point I was making was that any improvements in GPUs inside their socs since that time, were not coming from IMG.

Sure, you wrote that, but then you jumped to the following conclusions:

IMG knew well in advance of their public statement that they were not in Apple's plans going forward.

This depends on what you call "plans". Not licensing new IP is one thing, arguing that you owe no royalties on old designs is something else completely.

Also, unrelated, this article you've mentioned:

http://www.realworldtech.com/apple-custom-gpu/

Man, that brought back some memories of how cringe-worthy articles like these written by people who have no clue are.

 

[tangey]

Sure, you wrote that, but then you jumped to the following conclusions:

When I quote your message, previous quotes (i.e. what you were referencing) get dropped, but basically you were responding to my:-

IMG knew well in advance of their public statement that they were not in Apple's plans going forward.

The point being made was that, (and this of course assumes you take Apple's public statements at face value), Apple had, prior to the IMG announcement, asked for a reduction in royalty rate as they told IMG that they had designed out some of the previously used IP in later chips. Also, IMG knew that Apple had not taken any of their new IP, and clearly Apple were not going to standstill. Looking at the improvements in the A10 (excluding process/clock), one assumes IMG knew that whatever Apple was doing thus far, was working pretty well. They knew Apple was drifting in a direction that no longer included IMG new technology, that would lead them less and less reliant on IMG IP.

This means they already knew, a long time prior to the 2017 announcement, that a) they were not going to get any new licensing revenue from Apple once staged payments for the existing deals were completed, and b) Apple wanted to (we don't know if it actually happened) reduce the existing per unit payment to IMG.

In summation IMG knew, if you beleive Apple, there was a major pending income reduction on the horizion, from a customer that supplied circa 50% of company revenue. They clearly did not know that Apple was going to assert they were going "IMG-free", but given what they did know, they should have been 1) extrapolating just how far Apple potentially could go in designing them out (where would the GPU in the A11 be...the A12 be) and 2) reduce GPU resources accordingly.

But as I recall, IMG during this period was hiring more GPU people (was 50 mentioned ?). I have absolutely zero insight into the internal operations, but again, taking Apple at face value, it would be reasonable to think that IMG should have seen significant GPU resources freed up over recent times, unless of course IMG been hands off from Apple for a long time now at this point.

 

[Nebuchadnezzar]

They knew Apple was drifting in a direction that no longer included IMG new technology, that would lead them less and less reliant on IMG IP.

https://uk.practicallaw.thomsonreut...ontextData=(sc.Default)&firstPage=true&bhcp=1

Listing Rules. The key Listing Rules affecting business updates are in the box "Key UK Listing Rule obligations". In summary, issuers must notify without delay major new developments that may affect their business if the development may lead to a substantial share price movement. Issuers must also notify without delay information concerning a change in financial condition, performance or expectation of performance if the change would be likely to lead to a substantial share price movement.

Now ask yourself why this applied to the recent April 4th announcement and why it didn't happen in 2015 as Apple so claims.

 

[Entropy]

https://uk.practicallaw.thomsonreut...ontextData=(sc.Default)&firstPage=true&bhcp=1


Now ask yourself why this applied to the recent April 4th announcement and why it didn't happen in 2015 as Apple so claims.

I think you should state your hypothesis, because I can easily see more than one explanation.
My guess is that Img simply held off any public announcement. The wording of a recent statement where they claimed that Apples intentions didn't become completely clear until... was rather specific in its wording.
(Apart from the fact that Apple for years had recruited graphics professionals with the clearly stated intent of creating proprietary "world class graphics IP", which had somehow managed to escape the notice of IMG. Yeah. Right. This entire we-had-no-idea spiel is nothing but financial market posturing.)