Slashdot Mirror
Apple Loses Patent Suit To University of Wisconsin, Faces Huge Damages (reuters.com)
An anonymous reader writes: Apple has frequently been in the news for various patent battles, but it's usually against one of their competitors. This time, Apple is on the losing end, and they're losing to the University of Wisconsin-Madison. A jury found that the university's patent on improving processor efficiency (5,781,752) was valid, and Apple's A7, A8, and A8X chips infringed upon it. Those chips are found within recent iPhone and iPad models, which generated huge amounts of money. Because of the ruling, Apple could be liable for up to $826.4 million in damages, to be determined by later phases of the trial.
Here you go.
The IP in question, U.S. Patent No. 5,781,752 for a "Table based data speculation circuit for parallel processing computer," was granted to a University of Wisconsin team led by Dr. Gurindar Sohi in 1998. According to WARF and original patent claims, the '752 patent focuses on improving power efficiency and overall performance in modern computer processor designs by utilizing "data speculation" circuit, also known as a branch predictor.
Oh no... it's the future.
The Wisconsin Alumni Research Foundation (WARF), the people who actually hold the patents (the university isn't allowed to have them directly), did a similar thing to Intel not all that long ago (5 years?). I guess I can't complain since WARF funds my research (and a lot of research on campus -- having an internal funding mechanism is great). However, I recall that a bunch of the Intel money was supposed to go to the college of engineering (COE), which seems to have spent the money redecorating Engineering Hall (which, admittedly, is super ugly) and cutting ECE teaching assistant positions. It would be nice if the COE got a chunk of this and actually put it to good use...
The money goes to the Wisconsin Alumni Research Foundation (WARF). The university isn't allowed to own the patents directly. That means that the university and state can't just raid the money as they see fit, which is probably a good thing.
WARF doles out the money in various ways. Much goes to research on campus, some goes to "faculty retention" (i.e. when another university tries to poach a strong faculty member, WARF steps in with money to keep them here), some is administrative costs (WARF assists university personnel in getting patents), some gets returned directly to the university, some goes to the people who did the research to get the patent, etc.
In short, the answer is "all of the above and more".
The iPod had large "misfeatures" for most at the time it was released: It was firewire and Mac-only.
Mac only limited the market, but I don't think that made it suck. Firewire was a good choice at the time: all macs had firewire. USB2 was only juuuuust out and if you cast your mind back to 2001, it stank. The chipsets were yound and flakey as was the software. Firewire was much faster, and much more reliable.
Eventually USB2 realised it's potential and stopped sucking and actually caught up with firewire in terms of actual speed, but that took years.
SJW n. One who posts facts.
This patent is a bit different to what RISC and MIPS were doing back then. It involves out of order execution, which IIRC ARM didn't do in the Acorn Archimedes era. When doing OoO execution the CPU executes instructions based not on the order in which they are programmed, but on the order in which its resources are available to them (ALUs, FPUs, bus access etc.) This patent also predicts when data produced by one instruction will be needed by another, in a very specific way.
Other manufacturers have either licensed it or used a difference scheme to do data prediction, but the specific method they claim appears to be novel, or at least was back in 1998.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
In terms of endowment per student, that's practically nothing (relative to other universities). The endowment is meant to generate income on interest -- not to be spent directly.
Firewire flopped because Apple charged $1 per port to anyone else wishing to implement it. USB was free. Which brings this discussion back full circle to whether patents are helping or stifling technological progress.
Apple Tax
I'm not an Apple fan: I can't stand 'em, but I'm pretty sure the Apple tax is a myth. Last time I was shopping for laptops, for comparable weight, size, build quality, speed, memory and performance the Macbook Air was pretty close in price to similar machines such as the Zenbook UX11 (when that was current).
SJW n. One who posts facts.
> Mostly because lawyers are typically very dumb, and lawyers are who file the patents.
Patent lawyers are required to have degrees in STEM fields. They don't just let any Tom,Dick, or Harry esquire become a patent attorney. You are not dealing with someone that did underwater basket weaving as their undergrad degree.
(that probably makes the situation all the more sinister)
A Pirate and a Puritan look the same on a balance sheet.
A shame this got modded as Flamebait, since it's exactly the right answer. For some reason, many Slashdotters seem to be completely unaware (or perhaps willfully ignorant) of the distinction between utility patents (i.e. what we think of when we say "patents") and design patents, which are something else entirely. As a result, when they hear that "Apple got a patent on rounded corners", they rightfully think that's utterly ridiculous and an example of a broken patent system, when it's actually nothing of the sort, since design patents more closely resemble a time-limited trademark than they do a utility patent.
The reality of the situation, is that Apple is one of likely thousands of entities with design patents that include a claim for rounded corners. That's because those design patents aren't just making a claim for rounded corners. They're for rounded corners + a long list of additional claims that makes each of those products uniquely identifiable as the product they are. In the case of the iPhone 5 series, the design patent was for something along the lines of rounded corners + chamfered edges + aluminum trim + flat glass front + aluminum back + no adornment on the front + some other stuff I'm forgetting. I've seen a similar design patent filed by Samsung that covers some of their phones, and, as you'd expect, rounded corners were included in their list of claims as well. Again, each claim is considered alongside the other claims, rather than independently of the other claims, and for a competitor to be infringing, they need to be infringing against not just one of the claims, but against many or all of them. After all, rounded corners do not an iPhone make.
In the end, I'm just disappointed that there's a cadre of Slashdotters who appear to be willfully phrasing it as they do so as to confuse the issue, since in most subjects we discuss here, the commenters are seeking the truth of the situation. Unfortunately, when it comes to discussions that elicit fanboy-ish responses, they would rather seek out phrasings that help suit their narrative, rather than accurately convey the facts as they are.
The linked article is just wrong -- this is not in any way a branch predictor. It's predicting when to perform load-store speculation. The idea is that when you execute a load, you may not know whether there are any in-flight stores to overlapping addresses that are earlier in program order. If there are, once you detect it, you will have to squash the load (because it might have gotten a wrong value) and re-execute, flushing the pipeline. That's bad from a performance and energy standpoint (for the same reason branch misprediction is bad). You could always wait until you know that the load won't conflict, but that really hurts performance as well.
The innovation in the patent is that most of the squashes come from a relatively small number of load-store pairs, so by keeping track of them in a prediction table, you can get a large performance benefit for a small area overhead. The patent wasn't terribly useful in 1997 because instruction windows were small, but the authors thought that it would be once chips hit order of 1 billion transistors, which is pretty much what happened.
"This" did not make them public domain in the first place, and ever since 1980 the Federal Government has said that the patents can be owned by the universities themselves. The Federal Government gets to say so because state governments historically do not fund research, only educational activities and infrastructure. The Federal Government funds research (NIH, NSF, DARPA, etc.).
As to why the inventions aren't simply dumped into the public domain: university research does not take a product to commercialization. It is proof of concept, at best. Companies developing truly new concepts to commercialization want IP protection, since the next company entering could otherwise simply reverse engineer and duplicate the functional aspects of the commercial product while claiming license to 'public domain' patent behind the new concept.
Of course you can point to individual products that might have been commercialized anyway, but in aggregate the decision has been that this scheme will deliver more commercial products to market than the 'public domain' route.
If you think differently, then write your Congresscritter.