AMD Settles Bulldozer Core Count Lawsuit for 12 Million - 35 bucks per chip
Click here to post a comment for AMD Settles Bulldozer Core Count Lawsuit for 12 Million - 35 bucks per chip on our message forum
Undying
It reminds me of nvidia 970 vram settlement where nvidia agreed to pay up 30$ for every user.
rl66
yes, as in most country, law said that it was 3.5 + 0.5 and so is still 4.
the problem was more a communication problem.
if they win they will earn easy money on the back of AMD and of course the consumer isn't in the center of this lawsuit, but the money is (anyway no one care about the result exept them 🙂 )
anticupidon
Guess that I somehow missed my chance of becoming a lawyer. Everything I hear is settling, lawsuit and money.
Oh cramp, why invest so much in tech education?!
Undying
Kaarme
Lawsuits like this aren't bad, in my opinion. Companies are all about making money, so sometimes they need to be reminded of the customers' rights. I hope Intel will also lose due to the shameful security issues they have left untouched for years and too many generations of products.
If that was indeed tested in the court of law in some other countries and came back with such a conclusion, it would open interesting business opportunities for unscrupulous manufacturers, thus protected by the law. A good thing it didn't fly in the USA, at least.
Neo Cyrus
I said it the moment they started calling their garbage "cores"; they needed to be sued. What they're paying for their BS, in the end, is chump change, just the cost of doing business. But I suppose we shouldn't be too upset since in the grand scheme of things this is, by far, one of the least egregious things a large company has done then ended up paying squat for. I'm just glad AMD lost and had to pay up, even if it is just symbolic... and US only.
schmidtbag
As much as I find it a bit of a douchey move to have sued AMD when this lawsuit first happened (they were already in debt and these lawyers decide to kick them while they're down and unlikely to pay out?), AMD also did a disservice to people in how they were marketing Bulldozer. Not only was it misleading, but it made the products seem a lot worse than they really were. Take the 8350 for example: its performance typically sat between the 3570K and 3770K, which totally makes sense: it was 4c/8t (which was more than the 3570K had to offer) but it had worse IPC (making it generally slower than the 3770K). As a 4c/8t CPU, it was actually decently competitive. It still left much to be desired (especially considering the wattage) but it wasn't a total disaster. But because AMD advertised it as an 8-core CPU, that made the cores seem a lot weaker than they really were. I think we all would've remembered Bulldozer very differently if it weren't for the marketing.
Anyway, I'm a little confused about the 12.1 million. Is that based on how many chips were sold in the US? Is that the maximum amount AMD owes, or literally how much they actually owe?
Also, apparently that comes out to roughly 345,741 chips.
gx-x
up next: AMD settles infinity fabric lawsuit, admitting it is finite, not infinite.
Andrew LB
Celcius
I've gathered from reading the criticism of the Bulldozer-design over the years that a single floating-point pipeline being shared between every two integer pipelines is the overwhelming reason these cannot possibly be considered "true" cores. While well before my time in personal computing, all of the x86 CPUs, from the 8086 through into at least some of the 80486 designs, had no floating-point pipeline whatsoever. So, I guess these were all examples of a "core-less" processor.
The fact that certain portions of the design are shared doesn't alter the fact that it can execute, in the case of the FX-8000 and 9000 chips, eight threads simultaneously, in actual silicon. While having a one-to-one integer and floating-point pipeline arrangement would have almost certainly improved performance, where is that requirement written down and mandated? AMD never kept the particulars of the processor's internals from the public; I haven't actually checked, but I suspect that if I were to go back and peruse Guru3D's review of the Zambezi, it would indicate just what was below the heat-spreader. It wasn't hidden from anyone who took 2-1/2 minutes to learn about it.
I think this was a lousy outcome. That is because you have a judge and twelve jurors, who I seriously doubt any of which have a degree in microprocessor engineering, essentially defining, going forward, what constitutes a processor "core." Or, at least, what does not. So, as an example, if, a decade from now, Intel, VIA or AMD produce a breakthrough design with 256 discrete floating-point pipelines and a single homogeneous blob of special sauce dubbed the "hive-mind" to handle all the integer what-not, it can still only be marketed as a single "core." At least within California.
AMD was wise to short-circuit this nonsense before both the plaintiff and defendant attorneys were forced to be even more creative in divesting income to prevent finding themselves in an even higher tax-bracket then they undoubtedly already are.
Kaarme
Neo Cyrus
rl66
Kaarme