Slashdot Mirror
Rambus Files Antitrust Suit Against Memory Makers
bender647 writes "Forbes reports: 'Chip designer Rambus sued several major computer memory makers Wednesday, claiming they illegally conspired to limit production and raise prices in an effort to block widespread adoption of Rambus' technology.' Rambus believes that RDRAM was not the success it should have been because chip makers did not want to pay their royalties."
Worse yet, most RAMBUS had to be installed in pairs, while all other memory systems had switched over to single stick technology, doubling the cost.
If you could be told what you can see or read, then it follows that you could be told what to say or think - BoC
Design it? I don't think so...they patented a lot of stuff that was in open discussion at a technology conference.
BTW: 3. Sue
tasks(723) drafts(105) languages(484) examples(29106)
Intel was the company with the existing market clout that tried to force Rambus and their IP down everyone's throats.
I realize that in business these days it is not normal to consider how much of a scum your business partners may or may not be.
But for long term business I think it is worth review. We have to ask, in the end is the world going to be a better place because Intel and Rambus tried lock up a standards process in patents.
Folks need a longer memory then they get from playing XBOX games. These companies have histories.
The chip makers didn't want to pay there royalties oh darn they arent supposed to want to pay them. Rambus died because it was to expensive and it was to expensive because there were huge royalties gee shucks I guess another company with a brain dead business model. Rambus wasent good enough to compete at the prices. Yes at the time it was better but just wasent that much better.
If you want to take over a market you cant just be somewhat better it cant be a evolutionary better it needs to be revolutionary. They wernt and they had stiff compotition and bad business practices with there partners as well.
No sir I dont like it.
Pretty much supply and demand. Here is a good explanation
for it to be an anti-trust issue, the companies would have had to have been purposefully colluding to effect the price (commonly called 'price-fixing') of the good (rambus memory). if they didn't produce large enough quantities to make rambus acceptible as a widely adopteble standard because the royalties made the technology inaccessibly priced (high royalties mean the profit margin shrinks, and can become negative...) , this is not a trust. the claims in the blurb are contradictory, and if they are the claims of rambus, then the case is trying to blame somebody else because a certain someone shot themself in the proverbial foot with excesive royalties...
...illegally conspired to limit production and raise prices in an effort to block widespread adoption of Rambus' technology.' Rambus believes that RDRAM was not the success it should have been because chip makers did not want to pay their royalties."
I can say in all honesty that I didn't buy rambus because the cost was too damn high. I strongly suspect that chipmakers didn't want to sell an overpriced product, and focused on what consumers demanded. From my understanding AMD is outselling intel, and also near as i'm aware rambus isn't even an option for AMD supporting chipsets.
I'm damn sure chip makers didn't want to pay royalities, but this is neither immoral nor illegal when there is a viable cheeper alterantive.
There is no sanctuary. There is no sanctuary. SHUT UP! There is no shut up. There is no shut up.
SCO probabally learned all there is to know about patent litigation as a revenue stream from Rambus. There have been many /. stories, around 1999 - 2000, about Rambus.
"History doesn't repeat itself, but it does rhyme." Mark Twain
Not exactly. They are claiming that the chip makers deliberating conspired to make their chip FAIL because of higher prices because the chip makers weren't getting the profit margin they wanted on them.
It may very well be true, but it could easily be that all the chip makers said, Screw this we don't get the profit we want individually. The catch is that RDRAM really was much better, so the argument is none of them would have neglected it if they hadn't agreed to collectively neglect it, which is illegal.
Never confuse volume with power.
No problem.
> "cheap but good enough" almost always manages to beat "expensive and technically superior."
RDRAM was "technically superior" in theory, but as far as I remember the supposed performance benefits didn't have a significant enough impact on total system performance with the then-current chipsets to justify the huge price difference.
RDRAM looked promising at first, with Intel as the primary backer, but Rambus and Intel thourougly screwed up its introduction. This is how I remember it:
1. RDRAM was hideously expensive
2. Rambus used a "submarine patent" and got the whole DRAM industry up in arms about that
3. Price/performance ratio was bad
4. Chipsets with RDRAM support were expensive and only Intel jumped on the bandwagon initially (and with rather buggy chipsets to boot)
5. As a result, DDR-SDRAM was quickly announced, and RDRAM was history
I suppose the next steps would be:
6. Realize that your product is deader than a doornail
7. Sue the hell out of every major player in the industry
8. PROFIT???
To me, this looks more like the rest of the industry protecting themselves against Rambus' predatory practices and general ineptitude to bring a promising product to market. Perhaps suing Infineon and others wasn't the most brilliant move if they wanted to make RDRAM a success?
"There are already a million monkeys on a million typewriters, and Usenet is NOTHING like Shakespeare." - Blair Houghton
They were in on the talks with the rest of the memory standards organization JEDEC. JEDEC's rules say that members must disclose all IP and licensing terms.
Rambus not only didn't do that, they pimped their own knowledge of ram and techniques to speed up ram AND applied for/lengthened patents crucial to SDRAM and DDR SDRAM. When it started to become obvious that RDRAM was simply not going to make it in the market (Intel's RDRAM chipsets could NOT compete against it's own SDRAM chipsets--i820/i840 vs 440BX), Rambus decided sue anybody and everybody who produced SDRAM and DDR SDRAM but didn't buy licenses.
Intel really didn't have anything that showed off RDRAM's abilities until they went dual channel with the P4's i850. At that time, RDRAM still cost too much and DDR SDRAM went dual channel soon after.
Don't think the high cost of RDRAM was all to blame on the manufacturers and Rambus' license either. A lot of that was in the fabrication and packaging issues. At the time Rambus came out, SDRAM ran at 100/133 MHz while Rambus was at 800 MHz--really 400 Mhz DDR. So there were OBVIOUS electrical characteristics issues that had to be taken care of at the fab and package levels to bring yields up at a time when memory manufacturers were LOSING money per part. Had RDRAM come out sooner, or come out faster at a later date, things probably would have turned out differently.
So, when's lunch?
The PIII was not a good match for RDRAM. RDRAM had a faster transfer rate but more latency than SDRAM. With the chipsets of the day (i815, Apollo Pro 133A, i440BX overclocked to 133) PC133 was faster in most benchmarks. See for yourself, Anandtech still has their article online. By the time the P4 rolled around, it was a better match for RDRAM. The i850 boards were significantly faster than the SDRAM and DDR boards of the day, but by then most people wanted nothing more to do with Rambus.
This, then, is a fairly unsurprising tactic from them. Make no mistake about it, RAMBUS has never been a RAM company. It's an intellectual property company which attempts to make money via licensing and litigation.
Reality has a conservative bias: it conserves mass, energy, momentum...
Scientology
Uhhh, I am pretty sure that the different in expense was converting to a new technology that required more process steps, and required larger die than DRAM.
The array between SDRAM, DDR, DDR2 is nothing, only the periphery changes. For rambus, everything changes, hence the size is larger, larger size mean less die per wafer, and higher cost. The only way Rambus could have worked was if there was no alternative. There was, they lost.
Just like the RIAA, Rambus is blaming others for their own failures. The reason why Rambus (RDRAM) wasn't widely adopted was because it was an inferior product. Sure, in hypothetical and synthetic cases, RDRAM did outperform SDRAM, but in the real world, it got it's ass handed to them. (remember back in the day when Via had to fight Intel to be able to release a P3-compatible chipset using SDRAM? And how Via's low-end chipset was able to pounce Intel's expensive Rambus one and Intel eventually came to the realization that in order to compete, they would have to ditch Rambus). That combined with the expensive cost of RDRAM (even the ones maded by licensed RDRAM manufacturers like Samsung, etc.) and the disadvantage that you had to buy in pairs (talk about antitrust; "Sorry sir, you have to buy TWO copies of Windows for one to work.") or use a dummy stick which adds more cost and lackluster performance across real world appz (including games) lead to its demise. Not to mention, the abandonment by Intel which caused Rambus's stock to be cheaper than the Russian rubble (already used as toilet paper).
If anyone, they should sue themselves for bad business practices. Oh wait, the stockholders did try to sue but later dropped it.
Rambus has nothing to do with Dual Channel. The operating principle behind Rambus seems to be the same basic principle as USB or Firewire vs. SCSI or IEEE 1284 -- a narrow connection is more flexible than a wide one, so you can get much more efficient gains for less trouble if you simply speed up a serial connection (FireWire) rather than trying to enhance a paralell connection (UltraSCSI).
Rambus memory channels are 16 bits wide vs. DDR and SDRAM 64 bit channels. The plan behind RDRAM was to use that 16-bit connection at a far higher speed than SDRAM for the same or better performance. The i820 chipset used single-channel RDRAM; the i850 used dual-channel, essentially dividing the memory into two banks because the Pentium 4 had the speed to access both at once. So you see, it is a chipset thing.
Specifically (and this is all from memory):
Its big advantage was streaming, pumping out a very fast flow of sequential bits. Intel and others were at the time enamored of "convergence" (e.g. playing movies on your computer, which we now know is evil :-),
and it was thought this was a good thing for multimedia.
Except there were of course tradeoffs: you paid several prices in addition to the royalties. The two killers were slow startup time, which made random access speeds bad (which turned out to be critical), and its bit? serial interface, which required something like 600MHz traces on a motherboard, which was well beyond the realistic state of the art at the time. Intel did two separate million part recalls, one for a chipset, and one for a motherboard (just weeks before shipping...). (Note that your fast FSB busses obtain their speeds by using the signal edges cleverly, not by pushing the motherboard trace speeds outrageously.)
(It was also said that Intel higher engineering management forced this down the throats of their engineers who had to try to make it work, and fired anyone who was too persistent in pointing out the Emperor had no clothes. Well, Intel paid a pretty big price for this debacle....)
As far as I'm concerned, it died a well deserved death. The only question I see in this case as others have pointed out is exactly what did the manufacturers say to each other, and can a judge or jury be convinced it was illegal collusion instead of trading gripes. You do have to be careful what you say to your competitors; many companies have strict rules to try to avoid this sort of mess, since it does happen.
So your saying that an effort by independent producers to thwart a semi-monopoly is collusion????
The fact is that JEDEC had been meeting well before Rambus came into being. They had been developing memory standards for a very long time. In fact, Rambus joined JEDEC and attempted to monopolize the memory market via seeding JEDEC processes with patented technologies.
JEDEC is not a cartel. They aren't trying to squeeze or force anyone in or out of a market. They are an open association of companies that work to their common benefit. Anyone can join provided they have the capital. Anyone can license the technology.
Better watch out. If the logic behind this suit is successfull, Microsoft will sue Linus Torvalds for a conspiracy to make Microsoft Windows IP worthless by artificially driving down prices with "givaway" products. Every Linux contributor would be a co-defendant.
Rambus memory ultimately failed because it was ill-tailored for most of the PC market. The vast majority of PC applications rely on massive numbers of low bandwidth memory operations. Rambus memory carried significant penalties for latency. It's only appropriate for high volume web servers and video processing applications.
-------- -------- Support Wesley Clark for president!!!
Gee you wonder why memory prices are high. Just look back a few years when DDR was produced for less than manufacture cost. That was compliments of your three amigos that artificially suppressed prices so that they could call RDRAM expensive. Now that Intel is not using RDRAM the prices just magically increased. That's why Rambus is suing for antitrust. It has now become apparent through previous legal proceedings that IFX, Micron, Hynix and others have conspired to work the memory industry against Rambus and the consumers.
Yes, it failed because it was expensive, but RAMBUS contends it was expensive NOT because of their royalties, but because the memory manufacturers didn't want it to be successful, so the memory manufacturers got together and conspired to MAKE it expensive.
RAMBUS believes that without the alleged price collusion of the manufacturers artificially inflating the price of the RAMBUS memory, RAMBUS's technology would have been more competitive and thus would not have failed.
Illustratively, it would be like you designing this great processor that was better than the processors made by any other vendor that had some great technology in it that you had patented. Unfortunately, you don't have multi-billions of dollars to create a fab, so you need people like Intel and IBM to make your chip for you.
Intel and IBM don't want your new technology to become accepted though, as then they'd have to pay for it, so they take your chip, which costs $100 to make, and which they pay you $2 royalties on, and all agree to sell it for $800 per chip, while selling their own chips for $200.
The price if you were not charging royalties would by $798; the price if the manufacturers had not gotten together and agreed not to sell your chip for less than $800 would have been $202.
paintball
The truth is Farmwald and Horwitz had already applied for patents at the time of the initial JEDEC meeting you discuss.
That's okay.
What comes next isn't. The two of them proposed what became the RAMBUS standard as a proposal to the working group. Still okay, but.... what these two sleezoids did next was to "neglect" to tell everybody that RAMBUS was proprietary would require royalty payments to what became RAMBUS.
So everybody accepted the standard because RAMBUS was a decent idea. After adoption...POOF! "HEY GUYS! WE HAVE A PATENT ON THIS STUFF, AND YOU'RE GONNA PAY OUT THE ASS!"
Funny how that worked.
A judge heard the story and threw RAMBUS out of court on their RAMASS.
And lets look at the trouble with RAMBUS 1) Expensive because (2) Yields were so low (3) because the technology is inherently low-yield (4) Oh, and while it delivers high bandwidth it has (5) Horrible latency.
Please just go away. The RAMBUS company wasn't a company like you might imagine. It was simply a con game that tricked everybody.
DDR 3200 today kicks RAMBUS's ass in performance and price.
And oh, we don't have to pay a ridiculous amount to use it.
Ask Intel. They basically screwed up by supporting RAMBUS. The public didn't want it.
So go away. RAMBUS never had any advantages, and had oh-so-many disadvantages.
>Could someone explain what they did so wrong?
Rambus was formed to commercialize technology developed at Stanford (SLAC) relating to a low-pin count packet serial bus. It's a little sleazy, but done all the time, since there's a lot of technology developed in academia that never sees the light of day.
They developed a use for this serial bus, primarily to increase the throughput (bandwidth) of the interface between the memory controller and the RAM in a computer. They then sold the idea to Intel and convinced them that it was the only way to increase memory bandwidth for future generations of processors (it wasn't, there are many better solutions).
Intel engineers tried the technology, found that it didn't work too well, but were able to improve it considerably from the original design and obtained several patents on it. The Intel design (using reflective signalling) is what is now known as RDRAM, the original Rambus design (using separate clock lines running in opposite directions on a ring) never worked and has long been abandoned.
Intel then tried to standardize the design and find multiple source manufacturers for RAM based on the new design. Several signed on, but many others decided to wait and watch, since there was an alternative design, being developed by the non-profit JEDEC standards group, which promised similar performance without any of the technical drawbacks of RDRAM, and which was also less expensive to manufacture. This was the SDRAM standard.
What Rambus did next was sneaky: they joined the JEDEC standards group for SDRAM, and obtained information on all the design ideas offered by the other members. They then altered some of their existing patent applications (originally written for RDRAM) and modified them to cover some of the ideas presented at the JEDEC group. These patents were subsequently issued, and then Rambus sued the SDRAM manufacturers who used the JEDEC-standard SDRAM specifications.
The manufacturers countersued, saying that Rambus did not disclose any pending patents at the time of joining the JEDEC group, and moreover, the patents did not relate to SDRAM at all, but rather to the earlier Rambus designs (which were no longer in use, by Intel or anybody else). They also argued that the patents should be invalidated, since there was a huge amount of prior art on this subject from the previous 30 years of memory design.
The courts agreed with the memory manufacturers on most counts, and threw out most of the Rambus patents and also found Rambus guilty of fraud. SDRAM continued to be produced, quickly became the dominant standard, and has a significant price and performance advantage over RDRAM which ensures its dominance. Its successor, DDRAM improved the lead significantly, ensuring that RDRAM will forever remain a niche technology in search of a market.
Rambus meanwhile continues its old litigation game-plan, spending far more on legal fees than on actual engineering, of which it does very little. Good digital design engineers are afraid of joing Rambus because it would be a blot on their resume, guaranteeing that they cannot be hired elsewhere after working at Rambus. Bottomline: No further worthwhile innovation from Rambus, but a lot of threats of litigation and actual lawsuits.