Slashdot Mirror
FTC Dismisses Complaint Against Rambus
swordboy writes "A federal judge just threw out the FTC lawsuit against Rambus. This has been discussed at length here before but this changes the landscape yet again. An interesting, possibly coincidental item is that Intel just today announced a new and very powerful DRAM interface that bypasses Rambus IP altogether."
Actually it's more like RAMBUS has *been* dead ever since DDR / DDR2 became competitive in terms of prices.
Not just regurgitating history, though - I wonder if Intel will learn a lesson from RAMBUS's demise in regard to the new fangled transmission scheme*. RAMBUS died because it was 1) not open and 2) charged royalties. DRAM is such a low margin product that royalties will kill any possibility of your product hitting mass market (in RAMBUS's case, even with intel's backing - because none of memory manufactures liked it, so despite playing along they were really thinking of JEDEC and how to get DDR to be more popular / competitive). Intel, though, is probably doing this in a choke move for AMD, so it puts Intel at a tough decision point again: open standard = AMD can use it too, or RAMBUS version 2. That said, Intel isn't stupid, I am guessing their upcoming processors will be designed around a high memory bandwidth architecture to take advantage of it better than what competitors can. The low turnaround time (i.e. no bus turnaround!) is so sexy in a geeky way. circuit board designers are going to get soooo much headache over this though...
* the concept is indeed pretty cool, though you'll need some tough lil drivers that can handle incoming voltage swings while it's driving. The power dissipation on these I/O buffers are key, but in reality these things already exist, of course - just a bit pricy.
My life in the land of the rising sun.
What about if you want to move your data to another box? Even if you think if Flash or whatever, these are slow and die much faster than a HD.
I have no idea how the parent is "Insightful". Moderation hint for parent: Funny + 5
Rambus, long an innovator in memory designs has been virtually sued to death by JEDEC members over their IP rights to the RDRAM designs.
Nice, if it were true. The reason the JEDEC members were sueing was that Rambus was writing down the other companies ideas that were brought up at the JEDEC meetings and having their patent lawyers apply for patents on those ideas the next day. The other companies were not patenting those proposals that they were putting forth at JEDEC while establishing the SDRAM standards, due to a agreement between all members that the SDRAM standard would contain no patent-encumbered technology. When other JEDEC members caught wind of this and complained, RAMBUS left JEDEC, but their patent applications on SDRAM technology continued to change to cover new aspects of the SDRAM spec after each JEDEC meeting! They had a spy (codename: Secret Squirrel) in the meetings who was forwarding the tech to them while the spec was still being determined, and when the spec was published, most of the SDRAM spec was subject to Rambus patents on tech developed by the other members.
Rambus ripped off the JEDEC members and the courts are saying that this is OK. WTF? All is fair in love, war, and business (I guess).
Read, L
Actually, I think you might not be aware of the manner of Rambus' actions. First of all, they're not seeking licensing for the RDRAM designs, but for patents that are infringed upon by the DDR RAM implementation.
The reason other members of the hardware community are so upset, and the reason that Rambus has been the target of so many lawsuits, is that they were on the design commitee which decided upon the spec. for DDR in the first place, and they presented their technology to the standards working group conveniently without mentioning the fact that they owned patents on the implementation.
That's why they deserve the title of 'litigious bastards'-- because that's pretty 'bastardly' behavior.
Within 5 years, I predict that most machines will use RAM memory for all system storage. A backup power system will be required, but system speeds will go through the roof due to faster data access times.
Yeah, but addressibility comes into play. Today's most powerful consumer computers can only address 8GB of memory(Apple G5's, whether or not they are the most "powerful" isn't the point, but they are one of the few consumer level 64 bit machines out there), in order for this to come to fruition, assuming that data storage needs stay constant, we are going to have to go to 128 bit/256 bit chips/busses(busi?) to address this all, and considering intel's reluctance on the 64 bit platform, that's a little hard to believe.
But you do raise a point, maybe if memory gets cheap and fast enough, permanant ramdisk caches of the OS/some of your favorite apps etc could really speed things up, but the non-volatile disk(whether or not it will be a hard drive resembling the ones today will be another debate) isn't going anywhere anytime soon.
Depends on what you mean by "RAM". FRAM and MRAM are rewriteable, random access memory technologies that are also non-volatile. Densities and price points today don't make replacing your current RAM cost effective, but give 'em a chance.
Hard drives will still be around for bulk storage. Dollars-per-bit counts, too.
decide for yourself.
The point is that you may ammend a patent application after it has been applied for and before the decision has been made. They originally filed applications on a rather generic implementation and adjusted it to fit the spec while JEDEC was still in the process of writing it.
At least that's what the other members of JEDEC alleged, and RAMBUS, rather than deny it outright, admitted that it was receiving emails (from someone calling themselves "Secret Squirrel") advising them on how to ammend their technology (and their patent applications), but that they did not know who they were from, and did not know that the information was the same as was being discussed at JEDEC.
Read, L
I wonder if Rambus is being kept afloat by the PlayStation 2. It isn't used in more than a handful of PC motherboards and I imagine there aren't too many Alpha EV7 and Cray X1 systems being sold.
It wasn't. It had a high latency and it's bandwidth only exceeded the bandwidth of SDRAM a bit at first. Then, as DDR ramped up to speed, DDR blew it away in bandwidth and latency.
Rambus was never a great idea. It was very difficult to design a mobo with it. It is rumored that no company ever designed one without the help of Rambus the company.
To be honest, the only reason Rambus went anywhere is because Intel signed an agreement to force bundle it with P4. And this act itself launched Athlon and AMD, because Rambus was unaffordable and didn't provide levels of performance that were unreachable with regular RAM.
If Intel had applied the same level of effort to their SDRAM or DDR motherboards, they would have produced higher performance than Rambus at lower cost. But Intel didn't, they had signed an agreement not to. And they threatened to sue VIA if they brought a (presumably high performance) SDRAM chipset to market for the P4. Only once Intel shipped their own SDRAM-based P4 chipset (the 8200?) did Intel drop this threat against VIA.
RDRAM was mostly marketing. It's performance was never really all it was cracked up to be.
this thing would be more painful to work on chip to chip communications since you don't know if the other chip is Z or the logic state you are receiving simply corresponded with your current driving logic state. (I suppose one can always send a enable / disable signal similar to DQS along with a dataline to indicated if it's active)
You have two misconceptions about the scheme in question:
1) There is no "Z" state. Both sides are ALWAYS driving.
2) You don't have to stop driving the line to receive what the other side is driving toward you.
This is essentially the same hack that lets a telephone send energy at the same band of frequencies in both directions simultaneously, on a single pair of wires:
- You terminate the line at, or near, its characteristic impedence, and so does the device at the far end.
- You inject a current into the line/terminator junction (or, equivalently, shift the voltage at the "cold" end of the terminating resistor) to send.
- You compare the voltage on the pin (or current through the pin, or current through the terminating resistor) to what you expected to see if the far end was at a no-current-injected (or terminator "cold" end at ground) state. The difference is the signal being injected at the far end.
The wire is being driven at both ends at all times (no Zs). You can always tell what the far end is sending, regardless of what you're sending.
If you chose to send by injecting a voltage at the "cold" end of the terminator, you dissipate no power when both ends are sending the same value. You dissipate a significant amount when both ends are sending opposite signals. But you also dissipate the same amount if the transmitting ends of two separate wires are switched - for the time it takes the signal to propagate and the reflection to come back. If the separation between the transmitter and receiver is more than half the length of a bit time, the quiescent state has both sides driving the same value, and the two ends drive opposite about as often as same, it's a wash.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Having been in the business since 1988,I've seen all kinds of ideas like Rambus come and go. Generally, the idea is:
* Create an "essential" technology that is implemented in several large manufacturer's products.
* License the technology to everyone for big $
Most often what happens is that for a year or two, the "essential technology" may actually be very successful. Sometimes it even sticks around for the long haul, but the price becomes a lot lower. Then someone else comes out with "The Next Big Thing" or an open standard with simmilar functionality comes into existence. Some examples that are easy to remember:
* IBM's Microchannel Archetecture (was very cool for about two years, displaced by eisa, bus mastering ISA, then PCI)
* Adobe Postscript, Type 1 Fonts
* Zip drives
Rambus isn't essential any more... but they'll be aroud as much as I don't like them.
-- $G