Bringing an end to a saga that has spanned over a decade and most of the life of this site, what’s widely considered the final major legal battle between Rambus and a memory manufacturer has come to an end. Burying the hatchet, Micron and Rambus ended their fight this week with Micron finally agreeing to license Rambus’s technologies and to pay royalties for their use.

According to the Wall Street Journal Micron will be paying Rambus a 0.6% royalty rate on all impacted products, which given Rambus’s wide patent holdings essentially covers all forms of DDR SDRAM and in turn impacts vast majority of Micron’s RAM offerings. The agreement will run for 7 years, with Micron having the option to renew it at that time (as some of Rambus’s patents should still be valid even in 2020). Notably the royalty rates are capped at $10 million per quarter – adding up to $280 million over the period of the 7 year agreement – so the final price tag will depend on Micron’s DRAM revenue if they end up staying under the cap.

This agreement comes just over 2 years after Rambus’s last major fight with Micron, which saw Hynix and Micron successfully defend themselves against claims by Rambus that the two were conspiring against Rambus. That ruling meant that the two firms were not liable for treble damages to Rambus, but it left the matter of patent infringement unresolved. Since then Hynix has settled with Rambus, leaving Micron as the last man standing until now.

Ultimately with the settlement of the Micron fight, Rambus has now signed licensing agreements with all of the major memory manufacturers. This means that although it’s taken the better part of a decade, Rambus has ultimately proven successful in proving that SDRAM and its descendants infringe on Rambus’s patents, allowing them to collect royalties on all of the common forms of DRAM produced today. With the last memory manufacturer now licensing their technology, the only outstanding suits (that we’re aware of) all involve companies who develop memory controllers.

With that said, this does leave the question of where Rambus goes from here. In the PC space RDRAM/XDR has long been dead, and in the console space all of the current-generation consoles are using DDR3 or GDDR5, with the remaining XDR consumption tapering off alongside the last-generation Playstation 3. But there is a very real need for faster memory technologies, especially at the very high end where NVIDIA and other GDDR5 consumers are looking at more exotic solutions such as integrating/stacking DRAM on-chip as GDDR5 reaches its own apex.

Source: The Wall Street Journal

Comments Locked


View All Comments

  • gobaers - Wednesday, December 11, 2013 - link

    Sorry for my ignorance on the topic, but does Rambus actually provide valuable tech to the memory manufacturers, or are they more of a troll? I remember RD-RAM being completely not worth it, and being forced into it by some intel chipsets a decade ago; I'm still harboring some negative feelings toward that company.

    However, if they provide something that moves the tech forward, they should absolutely be compensated for their research.
  • Ryan Smith - Wednesday, December 11, 2013 - link

    In short, DDR wouldn't be possible without RAMBUS's patents. So in a way, yes.
  • emn13 - Thursday, December 12, 2013 - link

    That doesn't mean they provide value - just that they took a potentially "obvious" technical step and registered a patent for it in secret. Then, when another party comes up with a similar solution independently, they can sue. In that scenario, the patent holder hasn't provided any value; to the contrary, they discourage tech development due to the risks they cause.

    That scenario is unfortunately exactly the scenario most encouraged by the patent system because if you can wait to sue until the infringing party is thoroughly dependent on the patent, you have them over a barrel. By contrast, if you tried to actually provide value by seeking licensees in advance - to save them the cost of independently developing the tech - then the potential licensee might decide it's not worth it and implement an alternative that's just different enough not to infringe.

    Given the facts of the case here, where now everyone wanting to produce standard DRAM needs a rambus license, and given the fact they didn't have one initially, it looks like rambus in this instance is clearly a patent troll. Of course, that doesn't mean they always are; and it's a little hard to blame them for doing exactly what the patent system encourages.
  • 3DoubleD - Thursday, December 12, 2013 - link

    The thing is, based on how patents work, Rambus actually provided value by filing the patent in the first place. The reason nations grant patents in the first place is to exchange the details of a novel invention for the right to exclusively own that invention for a period of time. Basically, for giving everyone access to your great idea (putting it in the public domain) you are granted a monopoly on that idea for a period of time. It encourages invention by guaranteeing that no one will copy your product the minute you release it and it guarantees that marching on of innovation.

    As far what patents were meant to do, the system is working perfectly: people are offering their ideas to the public domain. Motivation to actually use the patents should have been more prominently baked into the system. I also believe patents length should be shorter with a more robust extension system, meaning if it takes a decade (or more) to develop the technology to a marketable product, a company should receive a guaranteed minimum period of exclusivity starting at that point. Now if you don't make any appreciable effort to develop that product, then no extension should be awarded. I think that would solve the major too long/too short/patent troll problems with patents in one shot.

    The US congress just made some patent law revisions the other day, I still need to take a look at what the changes were. For some reason I doubt they were as straight forward as I suggest above.
  • HisDivineOrder - Thursday, December 12, 2013 - link

    I think the initial consensus about Rambus was that they had been part of the committees that formed the basis of what DDR wound up being defined as. As a result, they secretly began patenting technologies they then convinced the others to use while not advising them that they were going to be able to claim ownership of said technologies at a later date and try to get a royalty for them.

    That is, Rambus said to them, "Hai guyz, I think we should do THIS!" with THIS! being a thing that they could then claim ownership over. They didn't tell their partners that they were going to be the gatekeepers of said technology when they helped steer them toward those patents.

    It should have come up somewhere along the way that, "Oh yeah, it is a great way to do things, but if you do, we're going to charge you a royalty for doing so." Instead, they waited until the spec was done and it was smacking the crap out of RDRAM and that's when they went, "Oh, remember that cool idea we sold you on..." to the entirety of the DDR making industry.
  • 0ldman79 - Thursday, December 12, 2013 - link

    Honestly from what I remember of the situation Rambus was a part of the standards committee, either patented the ideas during the development of DDR or immediately before.

    Essentially Rambus intentionally screwed the entire memory industry and I hate to see crooked bastards like that get away with it, much less benefit from it.

    Could be I don't have all the facts, but it seems that anything discussed while building an open standard that turns out to be patented by one of those involved is just completely crooked.
  • cbf - Friday, December 13, 2013 - link

    The "they patented the standards we were talking about" story is Micron's version of events, which the US Appeals court didn't buy:

    Rambus says they asked to present their work to the JEDEC committee and were denied. They also say that virtually every major memory on the committee had in fact seen the technology Rambus was filing patents on under non-disclosure.

    In fact, DDR was specifically designed as an alternative to the RDRAM technology Rambus was pushing. The problem is that it turned out that DDR infringed on a handful of the Rambus RDRAM patents anyway (confirmed by multiple juries) -- much to the chagrin of the memory companies.

    Rambus' court setbacks had to do with some dumb, over aggressive document cleanup (done under advice of counsel) that two trial courts and an appeals court found were done in a time period when litigation could be "anticipated" (not commenced mind you). (Judge White in San Jose said no harm was done.) The irony is that Micron and Hynix apparently had no theory about how any hypothetical destroyed documents prejudiced their case. As best as I could read the arguments, they are claiming Rambus destroyed evidence of prior art. One wonders how Rambus could possibly have had the only copy of such prior-art evidence ..
  • A.Noid - Saturday, December 14, 2013 - link

    That is EXACTLY the way I remember it. They were or are part of JEDEC, and talked the committee into using the technology, and patented it just like you said. they are snakes of the worst type.
  • OCedHrt - Monday, December 16, 2013 - link

    According to cbf, Rambus was denied access to JDEC committee.
  • Kevin G - Wednesday, December 11, 2013 - link

    Almost by necessity some of Rambus' technologies will have to be used in future iterations of memory technology. The parallel topology used by SDR/DDR has pretty much reached its limits with DDR4, especially in terms of expansion (one DIMM per channel without buffers). A serial topology would allow for similar bandwidth (though not necessarily latency) as the parallel interface using fewer pins or more channels using a similar number.

    Rambus hasn't been idle since the Pentium 4 days either. They developed the XDR and XDR2 spec. The only major platform to use it was Cell which of course found its way into the Playstation 3. There other minor wins, notably in networking equipment where its high speed serial nature was a key advantage. Speaking of networking technology, Rambus has done some work in the Infiniband area.

Log in

Don't have an account? Sign up now