June 26th 2000

Fight Back, Boycott Rambus!

  I had not gotten a chance to see the Hitachi counter-suit against Rambus back in March when it was filled, but The Register got a hold of it and posted the entire document here.  It's huge, but I read the whole thing.  The document answered numerous questions I had, but prompted new questions.  Rather than quoting the document, I'll just summarize.

  To set the scene, remember that Rambus sued Hitachi in early 2000, claiming that all Hitachi's SDRAM, DDR DRAM and logic controllers infringed on Rambus patents.  Rambus singled out Hitachi among semiconductor makers because they were the largest foreign memory manufacturer that was not producing Direct Rambus DRAM.  Hitachi then counter-sued in March 2000, claiming that Rambus was attempting to control the computer memory market.  In the last two weeks, Toshiba and Hitachi settled with Rambus out of court by agreeing to pay undisclosed royalties on all covered memory and memory controller products. This is all recent news, but it's what happened during the previous nine years that is so interesting, and points out the deceitfulness of Rambus' entire business model.

  The Joint Electronic Devices Engineering Council (“JEDEC”) Solid State Technology Association is a semiconductor engineering standardization body of the Electronic Industries Alliance (“EIA”).  JEDEC was formed to open up the semiconductor industry and provide for standards in semiconductor manufacturing, permitting interchangeability of parts, and promoting competition.  JEDEC had rules that stated that all participating members must disclose any patents they have pending on the design of the semiconductor products covered by JEDEC's standards setting committee.

  Rambus not only failed to disclose their patent applications to the committee while a member, but actually may have acquired competitors intellectual property, which it then used to amend and revise it's pending, undisclosed patent applications.  In order to cover themselves, Rambus sent a letter on September 11, 1995 to JEDEC stating:  "At this time, Rambus elects to not make a specific comment on our intellectual property position relative to the Synclink proposal. Our presence or silence at committee meetings does not constitute an endorsement of any proposal under the committee’s consideration nor does it make any statement regarding potential infringement of Rambus intellectual property." 

  If I were JEDEC, I would have thrown them out on their ears right then for participating in bad faith. Rambus completely subverted the entire purpose of an open standards committee, to their benefit, and the detriment of all the other members who acted in good faith.  Less than a year later, Rambus announced in a June 17, 1996 letter that it was leaving JEDEC: 

"Recently at JEDEC meetings the subject of Rambus patents has been raised. Rambus plans to continue to license its proprietary technology on terms that are consistent with the business plan of Rambus, and those terms may not be consistent with the terms set by standards bodies, including JEDEC".

The letter listed certain issued Rambus patents, but failed to disclose any information about pending applications in the Related Family or the Additional Patent Rights, other than a statement that “Rambus has also applied for a number of additional patents in order to protect Rambus technology.” 

  Rambus technology?  If there ever was a case of a company participating in a Standards Committee in bad faith, this is it.  Hitachi claims that Rambus used information obtained from JEDEC and it's honest members to amend their ongoing patent applications in order to make the applications cover existing memory and memory controller technologies.  One of the most astounding aspects of the whole case is that the patents were granted from mid-1999 on, years after some of the covered products had gone to market!

  This is the worst behavior by a company that I have ever observed.  It makes Microsoft seem honest and trustworthy.  But without a company like Hitachi bringing a countersuit against Rambus, there is little that can be done at this point.  I am at a loss as to why JEDEC has remained silent on this issue.  If Rambus' conduct while a member of JEDEC clearly violated the standards body's rules, then Hitachi had a valid countersuit.  There are probably numerous legal and business reasons why Hitachi gave up the fight, but it's a real shame, since they were fighting for us consumers, as well as for themselves.

  There is only one thing we can do right now as consumers.  That is to get the word out on Rambus' horrific business practices, deceitfulness, and attempted monopolization of the memory market.  Further, we can boycott Rambus' products... to speak with our wallets, as it were.  This means that we should not buy products from Dell, or other computer makers, that are based upon Intel's i820 or i840 chipsets.  The problem with this approach is that Rambus may eventually get money no matter what you buy, because they are attempting to skim money off of every memory sale in the world.  If you currently do not buy Hitachi, Toshiba or Rambus DRAM, then Rambus will not get any royalties, so keep that in mind when buying a system or memory. 

  I do not know what will happen next.  No other companies besides Toshiba and Hitachi have settled out of court with Rambus, but they may also decide that it is cheaper to cave in than litigate.  But for our part, KickAss Gear states that it will not use Rambus products, or purchase any products that will afford royalties to Rambus.  Let the boycott begin!

                              Dr. John

June 24th 2000

Rambus Plays a Dangerous Game.

  Encouraged by their recent out-of-court wins against Hitachi and Toshiba, Rambus has come out swinging.  They are pushing hard to leverage money from all memory and logic controller manufacturers, which can only serve to drive prices up.  The entire situation is quite ugly, and I am prone to wonder if the strong-arm tactics will come back to bite Rambus.

  A report at EBN last night outlines some comments recently made by Avo Kanadjian, VP of marketing for Rambus.  Avo is obviously feeling very good about the turn of fortunes for Rambus.  Here are some excerpts.

  Kanadjian did downplay any strategy aimed at reaping large revenues from other components that might use the synchronous interface, such as logic controllers. "Our overriding goal is to build our Rambus business," he said.

  Got that?  They just want to build their business. Rambus doesn't make anything, they just skim money off the top of everyone else's sales, driving prices up, and making themselves rich with the work of lawyers, not engineers.  How much more will we have to pay for memory and processor chipsets so that Rambus can sit back and get rich?  They won't say, but here is another excerpt form the article.

 Kanadjian responded to analysts' speculation that Hitachi and Toshiba received "sweetheart" royalty terms that will likely pressure other DRAM manufacturers to the bargaining table. "There is no such thing as a standard licensing agreement," he said. "Each company's case is different. Hitachi and Toshiba have different licensing terms, and new agreements may well have different terms." 

  This is exactly the kind of thing that got Microsoft in trouble with the DoJ.  Indeed, part of the conduct remedies in the Microsoft case, which have been put on hold for now, was that MS could no longer give sweetheart deals to players who did their bidding (not selling competing products, etc.).

  Part of being successful in life is learning how to not make enemies.  This is a Kindergarten lesion that Rambus officials slept through.  The public is mad (and will be madder if memory and CPU prices go up), and the industry is mad.  You can't just keep making people mad at you, and not suffer negative consequences.  The Rambus method of using intellectual property patent claims to force an entire industry to pay them royalties is certain to leave a bad taste in everyone's mouths.  It may drive their stock prices up for now, but "Karma" has a way of turning around the worst meaning plans of mice and men.

                              Dr. John

June 23rd 2000

Rambus Wins, Hitachi Crumbles!

  It was reported at numerous web sites today, including EBN and The Resister, that Hitachi and Rambus (RMBS) have ended their suit-counter-suit litigation proceedings.  This is huge news folks.  If you buy stocks, buy Rambus now, before more bad news comes out and it goes down again.

  Basically, Rambus has claimed that it holds patents on all current forms of memory, even memory it did not design.  Rambus does not produce any memory itself, it licenses it's "intellectual property" to memory makers.   They in turn pay Rambus royalties. In the last year, Rambus has filed suit against virtually every memory maker, claiming that Rambus holds patents covering memory types they did not even design. Rambus designed so-called "Direct Rambus DRAM", but apparently they have patents on underlying timing schemes that pertain to all forms of modern computer memory.  Here is a quote from a Rambus official: "Our patents are pretty fundamental," said Rambus chief executive, Geoff Tate, during a press conference Friday to announce Rambus' next-generation memory interface technology for the consumer market. "I think it's likely to say that most companies will violate these patents."

  Now, both Toshiba and Hitachi have settled the cases out of court, and agreed to pay Rambus royalties on all synchronous memory and memory controller products they make.  They have in fact agreed to pay Rambus higher royalties on the memory types that Rambus did not design, than on the ones that it did.  Obviously, this is an attempt to reduce the huge price differential between the Rambus-designed products, and standard memory types.  

  Hopefully, Rambus will get hauled before the ITC when the other memory makers get organized.  To me, this seems to violate the Sherman Antitrust Act, by trying to control memory prices world wide.  Rambus can charge any royalties they want, and there is no competing product.  So buy RMBS now, while the buying is good. Expect prices to plummet (again) if Antitrust allegations are formally filed against Rambus. 

                              Dr. John

3dfx Woes.

  3dfx announced yesterday that they will not meet earnings expectations for Q2, due to the continuing world wide shortage of many electronics components.  Here is the meat of the announcement.

"This shortfall is extremely frustrating, as the company has been on course for returning to operating profitability," Dr. Alex Leupp said. "We believe this temporary component shortfall will result in deferred rather than lost revenues, and, therefore, we remain confident about our third quarter."

"We feel very positive about our overall progress to date and believe additional improvements are on the horizon," Leupp continued. "It is unfortunate that the availability of components has undermined the prospects for improved earnings in the second quarter, but we believe it has not affected the longer term outlook for achieving profitability."

  The parts shortage is affecting manufacturers from DVD drive makers to Cell Phone makers.  3dfx changed it's business model last year when it purchased STB Systems, a component manufacturing company.  The idea was to become the sole supplier of 3dfx video cards, from design, though manufacturing, to final distribution.  Since that is how ATI and Matrox make their video cards, that leaves NVidia as the only major 3D accelerator chip provider for third party video card makers like Creative, Asus, Elsa and others.

  The end result is that 3dfx now suffers from parts shortages, while NVidia does not.  NVidia just makes the graphics chips and the reference driver set.  Other companies have to do all the work of buying all the parts, and assembling GeForce boards.

  In the long run, there's no telling whether the 3dfx acquisition of STB will turn out to help or hurt the company, but the lack of competition for NVidia in the graphics chip market has helped NVidia's bottom line.  

                              Dr. John

June 22nd 2000

Intel Says Rambus Doesn't Have Memory Monopoly.

  Mike Magee at The Register reports that Intel CEO Craig Barrett answered questions at a Stockholm press conference on Tuesday.   One question was about the Toshiba-Rambus deal, where Toshiba agreed to pay Rambus royalties for all it's memory products, not just Rambus products.  They did this because Rambus claims to have patents covering all modern forms of PC memory.

  Mr. Barrett responded by saying that Rambus did not have an "intellectual collar" on the memory market.  I would have asked if he thought Rambus wanted such a collar on the market, since it appears they do.   He went on to say that memory makers were resourceful, and would come up with their own intellectual property.

  To be honest, I've getting sick and tired of of hearing wealthy suits talk about "intellectual property" (something they can patent and make lots of royalties on).  As an ex-scientist, I wonder how Mr. Barrett would feel if every single medical discovery by scientists was patented, and he and his family were charged huge sums for the medical benefits (it's happening more all the time, and look at medical costs!).  Drug companies do that, and that's why drugs are so damned expensive.  Scientists work just as hard at finding out how nature works, but most of them do not patent the results of their efforts.  I want more open standards, like the JEDEC memory committee.  I'm tired of everyone wanting to make the next PC platform proprietary, just like those really expensive Macintosh's.

                              Dr. John

Intel Says Processors Will Be Scarce Until Late July.

   Another story at the Register says that Intel has informed customers that there would be CPU shortages at all speeds until the end of July, at the earliest.  If this keeps up, AMD will have a de facto processor monopoly soon!  

June 21st 2000

Intel Steps Up.

  According to an article at EBN yesterday, Intel is moving to the next Coppermine stepping. The transition is from stepping B-0 to C-0. The new stepping reduces "errata" (bugs) as well as cutting die size.  Intel said that the changes will permit higher clock speeds, and will lead to a 5% reduction in the die size (the size of the actual silicon) of the "Coppermine" Pentium III.  That translates to reduced manufacturing costs.

  Intel also stated in response to an EBN query that the stepping process is a normal part of microprocessor manufacturing.  We agree, but rarely do simple CPU steppings change the layout, size, and top-end of a processor line.  It sounds like a minor-upgrade to the PIII to me, to let Intel hit the 1GHz mark with a larger yield.   They will be limited at first to the top-end, Slot-1 PIIIs, and should start showing up in July.

  And while we're on the subject of Intel, it was reported at The Register today that Intel's CEO, Craig Barrett, has predicted a world-wide shortage of PC part for at least the next year and a half!  He blames the parts shortages on under-estimations of demand several years ago when new plant construction was started. He says additional construction is the only solution. If true, expect PC prices to go up as supplies tighten.

                              Dr. John

June 20th 2000

Celeron.... R.I.P.

  With the launch of the AMD Duron processor yesterday, we got to hear all sort of goodies about this new low-cost processor.  The benchmarks show it beating the new Celeron-2 from Intel by as much as 25% at the same clock speed, for less money.  In fact, at Tom's Hardware, a review shows the Duron 700 beating a new Celeron-2 700 by almost 40% in Quake III using a GeForce2 card.  Intel has some real catch-up to do now.  They either need to up the 66MHz bus speed for the Celeron-2 to 100MHz, or they need to drop prices substantially, now that they have the slower low-end CPU line.

                              Dr. John

June 17th 2000

Rambus In the News!

  The buzz on the Net about Rambus has really heated up over the last two days.  Since the story broke at The Register yesterday, stories have popped up everywhere.  There are good reads at The Tech Report, and EBN, and you can check my rant below.  The basic upshot is that Rambus is using litigation and threatened litigation in an attempt to drive up the price, and reduce the availability of competing memory products.  If Rambus wins these court fights with other memory makers, and in the case of Hitachi, forces a ban on imports of their memory products, then the reduced availability of memory chips will drive SDRAM and DDR DRAM prices way up, like last Fall when 128MBs of PC-100 SDRAM was going for over $300.  Rambus obviously wants to reduce the price differential between their overpriced product, and competing memory types.  If you thought Rambus had enemies in the industry before, just wait.  It's almost as though they were shooting for "most hated hardware manufacturer" status.  Is there a category for that in the Guinness Book of World Records?

  Here is an onerous quote from Rambus: "Our patents are pretty fundamental," said Rambus chief executive, Geoff Tate, during a press conference Friday to announce Rambus' next-generation memory interface technology for the consumer market. "I think it's likely to say that most companies will violate these patents." This kind of thinking may lead to a Microsoft-like antitrust case, or at least it should.  Any company that tries to corner an entire market deserves the DoJ.

                               Dr. John 

Article: Rambus Attempts to Corner the Memory Market

June 16th 2000

Rambus Inc: Making Friends in the Memory Industry.

  Under continuing pressure from Rambus lawyers, Toshiba has decided to pay Rambus Inc. royalty fees for their SDRAM and Double Data Rate DRAM (DDR DRAM) products.  That's correct, for SDRAM and DDR DRAM, not Rambus DRAM!  You can read about this extremely unpleasant occurrence here. This has been brewing for awhile, since Rambus had filed suit claiming that all DRAM manufacturers were infringing on Rambus patents for SDRAM and DDR DRAM products.  I am not privy to the details of the complaint, but apparently, Toshiba decided that it would be better to pay than continue with a court case.  If other memory manufacturers follow suit, SDRAM and DDR DRAM prices will go up.

  Memory makers and consumers alike have not been too happy with Rambus and their high-priced products.  These strong-arm tactics, which can only serve to drive up memory prices to the benefit of no one but Rambus stockholders and corporate suits, will undoubtedly make even more folks unhappy with Rambus Inc.  They have already convinced me to avoid all Rambus products like the plague.  Rambus has announced that they will charge higher royalties for SDRAM and DDR DRAM products in order to drive the price up closer the the cost of Rambus DRAM.  What a pleasant bunch.  

                               Dr. John

V5 5500 Driver Dilemma.

  We have seen posts at 3dfxgamers.com that suggested that the default drivers that get installed when you run the V5 installation program are not the best drivers for the 5500 model.  We can confirm that the default drivers will report the card as a "Voodoo series" video card in the device manager, while other drivers on the CD report the card as a Voodoo 5.  The posts at 3dfx gamers said the new drivers greatly increased the speed of their new video cards, but we did not find this to be true, at least under Direct 3D.  There are several V5 drivers on the CD, but the latest one is not installed by default when you run the setup program.  To update your V5 card to the latest version, follow these instructions with the 3dfx V5 CD in the CD drive:

1) Go to the device manager
2) Select the V5 video adapter
3) Select the "driver" tab
4) Click on update driver
5) Click next
6) Choose "display a list of all the drivers in a...."
7) Click on "show all hardware"

Under the 3dfx heading, there should be two drivers available, one made on 5/16, and one dated 6/10.  Select the 6/10 driver, and click on the Next button.  It may warn you that the driver is not written specifically for the device (even though it is), but continue with the installation process.  Now, after a reboot, if you go to the device manager, your video adapter will be reported as a Voodoo5, rather than "Voodoo Series".

                               Dr. John

V5 - GeForce2 Showdown.

  SimHQ.com has posted a comparative review of the V5 5500 and Gladiac GeForce2 card.  They come down on the side of the V5 for various reasons, which you can read about here.  This is a telling excerpt: "Installation on each card was very different.  The V5 5500....  It just went in and ran. The GLADIAC, on the other hand, was not so easy.  On 2 of the 3 systems, the system booted into a Blue Screen of Death (BSOD) immediately.  After entering SAFE MODE and doing some driver clean up and reboots, I was able to get the systems to come up...  On the GLADIAC, frequent BSODs, lock ups and Crash To Desktops (CTDs) made this a time consuming and tedious process.  Hours were spent on each of 4 machines I tested this card in.  None of the installs were ever complete in my opinion, and the machine this card resides in now still has the occasional issue."

  That about sums up our GeForce experiences quite nicely.

                               Dr. John

AMD To Continue Slot-A.

  Mike Magee at The Register reported today that AMD has promised continued production of Slot-A Athlon processors.  Motherboard makers have been upset that many KX-133A motherboards (which had just been produced) had become obsolete overnight after the introduction of the Socket-A "Thunderbird" Athlon.  AMD has now stated it will maintain Slot-A Athlon production for the time being, to help motherboard makers reduce KX-133A inventories.  Of course they did not promise to make Slot-A "Thunderbirds" available to the retail market, which would be the most customer-friendly policy.  Let's hope they do.

                               Dr. John

June 15th 2000 (late edition)

UUNet strikes again.

  UUNet service was finally restored this evening, after about 16 hours of outage in our area.  Sorry for any inconvenience while we were down.

June 14th 2000

Microsoft Heads Upstairs to the DC Court of Appeals.

  If Microsoft's Lawyers were getting tired of going to the DC Court building, they aren't getting any relief soon.  The DC Court of Appeals, which is located upstairs in the same building, agreed to hear the case yesterday.  This means that the case will drag on for longer than if the Supreme Court had gotten the case directly.  The Supreme Court usually prefers to hear cases that have been argued to death, and boiled down to the basics, so this route will probably suit them just fine. 

  The next big question is if Microsoft gets "relief" from the proposed remedies in the form of a court injunction.  If so, then it's business as usual for the Software bully for the next two years.  If not, then it's going to be a wild ride at Microsoft as they scramble to attempt to comply within the next 90 days.  Windows ME (millennium edition) is scheduled for release in July.  How it may be affected if the court injunction sought by Microsoft is not forthcoming is anyone's guess.

                               Dr. John

June 12th 2000

KX133A Mobos Dead Meat?

  There is quite a buzz on the Net about VIA and AMD messing up with the switch-over from Slot-A to Socket-A chipsets.  And while motherboard makers are still fuming over Intel's motherboard mess-ups. now they are turning their anger toward VIA and AMD.  Apparently, motherboard manufacturers have hundreds of thousands of KX133A (Slot-A) motherboards sitting in warehouses ready to ship.  But with the advent of the Thunderbird Athlon, all these motherboards become useless, since the Thunderbird requires a new chipset and socket (KT133A/Socket-A).

  There is one way for AMD and VIA to fix this mess, but I doubt they will do it.  It would be possible for AMD to continue producing the Slot-A version of the Athlon, until KX-133A motherboard supplies are reduced.  They could either produce the older 0.25 micron Athlons, and sell them at a big discount, or make more Slot-A, 0.18 micron Thunderbird Athlons for retail distribution, so that end users and small computer makers can buy them and install them on KX-133A motherboards.  Let's hope that from now on, Intel and AMD will work out the logistics of motherboard production with motherboard makers well before new processors are debuted.

                               Dr. John

Voodoo5 Cards Very Stable, but Not Too Fast.

  We have been putting Voodoo5 cards through their paces over the weekend, and while we love all the features and hardware anti-aliasing, we were a tad disappointed with the cards speed.  The Voodoo5 card seems to perform about on par with the original GeForce card, which puts them below the performance level of GeForce 2 cards, although at about the same price range.  The Vodoo5 6000, which has not debuted yet, will beat the GeForce 2 cards, but at a very high price (about $600).  I think that 3dfx has some price cutting to do.  It may not please their stock holders, but they need to knock $50 off the 5500, and $100 off the proposed price for the 6000 model.

  The very best thing we noticed about the new V5 5500 is that is installs much more easily, and has far fewer driver problems than GeForce cards.  3dfx has always had better driver support than NVidia, and they prove it again with the V5 drivers.  The V5 5500 is a rock solid card, that will not have you ripping your hair out night after night as you plow through installing and uninstalling boat loads of leaked drivers.  What a refreshing change!

                               Dr. John

June 11th 2000

Voodoo5 Stuff.

  3dfx officially announced the release of the Voodoo5 cards late last week.  We managed to get our hands on a few, and will be completing our testing shortly.  They look great, but they are not as fast as GeForce2 cards.  Also, they are so big, that some folks will have trouble fitting them into their computer cases!  For an interesting article on the upcoming Level of Detail slider that 3dfx will be incorporating into future driver releases, check out this article.

June 9th 2000

Voodoo5 Cards Hit Store Shelves!

  We have heard reports from several sources that Voodoo5 cards have begun arriving in stores, including Electronics Boutique, and CompUSA.  This is excellent news for 3dfx fans.  We have not seen them show up at distributors yet, but ETA dates at several distributors range from today, to June 19th.  So if you want one, it looks like you should be able to get your hands on it shortly (if you've got the $300!).

June 7th 2000

Microsoft Broken Up in Antitrust Ruling.

  Judge Jackson handed down his decision at 430pm Eastern time today, ordering the division of Microsoft into two distinct companies; an applications company and an operating system company.  No surprises here.  Microsoft behaved illegally, and now a remedy is being ordered by the Court.  If Microsoft had cooperated with the Department of Justice, they probably could have gotten a better deal.

  This begins the long appeals process, which may strip away some of the ordered remedies, but will probably not reverse the breakup.  It is not in Microsoft's interest to let this drag on, since it puts much of their upcoming software on indefinite hold.  If they were smart, they would begin the process of reorganizing. It will be necessary at some point, so they might as well get started.  Dragging this out will hurt Microsoft much more than any inconvenience the breakup will entail.

  Bill Gates says his appeals case is strong, but that suggests to me he didn't pay much attention to the details of the trial.  Microsoft's case is weak, and they offered no credible alternative remedies to those put forward by the court to address Microsoft's anti-competitive business practices.  In a post-ruling news conference, Bill Gates said several times his company would continue to make integrated software, and that he was confident that the rulings would be overturned on appeal.  

  The best part about this in my mind is that Internet Explorer and Windows will be produced by different companies, and won't have an artificial ball and chain holding them together.  Windows will be a better operating system for it.

                               Dr. John

Thunderbird and Duron Overclockability?

  Posts are springing up on the Web that suggest that overclocking Thunderbird and Duron-based AMD processors may be possible by jumpers or BIOS settings on Slot-A motherboards.  Check out the links at JC News on the subject.  Here is the first post, and here is the second.  It sounds plausible to me, but we'll have to wait to see what happens when real KT133 motherboards hit the market.

AMD and Intel on Different Tracks... Memory Market Split In Two.

  Based on information coming from the Computex meeting in Taipei, it seems certain that Intel is going to stick with Direct Rambus DRAM, and AMD is going with Double Data Rate DRAM (DDR DRAM), on their future processor platforms.  This is a fateful, and potentially disastrous course for one of the two, and only time will tell who chose most wisely.  On the surface, AMD seems to have taken the safe route, and Intel the rocky road, but time and circumstance have a way of upsetting even the best-packed apple carts.

  The Willamette is Intel's first 7th generation chip (P7), and will be followed by a more powerful variant, code named Tulloch, by the middle of 2001.  Both chips are slated to run on motherboards with Intel chipsets that support only Rambus DRAM.  The cancellation of the updated Memory Translation Hub (MTH) indicates that Intel is dropping any SDRAM support in it's upcoming chipsets.  AMD is going straight for 266MHz DDR SDRAM, a memory scheme that Intel has said it will not support.

  So there you have it, an Intel/AMD split of the memory market.  DDR SDRAM for AMD/VIA systems, and Rambus for Intel systems. This possibility has been looming for a year now, but back then, the problems with Rambus were not known.  It is interesting that Intel has decided, even after all the Rambus-related problems. that Rambus will be the only memory technology supported in the chipsets for Willamette and Tulloch.  It's a big decision.  But Intel seems committed to it, and their company's future may be riding on this one decision alone.  If AMD and VIA successfully create 266MHz DDR-capable chipsets with other advanced features, which perform as well as Intel's parts, but at a much lower cost, then Intel will loose this fight fairly quickly.  If, on the other hand, Intel and Rambus manage to beat the upcoming AMD Mustang processor in benchmarks by a decent margin, and Rambus prices fall dramatically, then Intel may luck out.  But right now, it's anybody's guess what will happen over the next year.

                               Dr. John

June 6th 2000

More Intel Problems With the MTH... It's Canceled!

  An article at EBN says that Intel is delaying the release of their new computer-on-a-chip, code named Timna, until 2001.  The problem is again with the memory translation hub (MTH), which allows the chip to work with both Rambus memory, and less expensive SDRAM. This delay decision highlights Intel's entrenched problems with Rambus memory.  Virtually every platform that Intel has been working on, from the low end i810, to the mid range i820, to the high end i840, are having problems that boil down to Rambus-related.  I can't imagine that this is not starting to sink in with Intel.  What seems most inexplicable is Intel's adoption of Rambus for the Timna project.  The entire point of Timna is to reduce the cost of computer hardware dramatically, so that Intel can reclaim mastery of the low-end computer market.  So why did they couple Timna with the most expensive memory type available?  It just doesn't make any sense.

  The article at EBN states that Intel has canceled the MTH altogether due to all the problems with it.  Here is a quote from the article:

 "...the company discovered that certain combinations of temperature and signals caused errors in the MTH, the component that has currently caused a recall of all Intel and third-party-manufactured motherboards which integrate it. Previously, Intel said it was designing a replacement, due sometime in the third quarter. Now, that chip has been canceled, pushing out the Timna chip in its wake. 

  The article says this means that Intel is scrapping SDRAM support altogether, which seems hard to believe.  If true, then Intel is abandoning SDRAM for Rambus, come what may.  Intel motherboards are very unpopular now due to the high cost and low availability of Rambus DRAM.  This has been a loosing proposition for Intel over the last 8 months, and I am wondering how long they can keep Rambus in tow. Perhaps Intel has decided it will give Rambus one big, final push to see if it catches on.  Good luck boys, you'll need it.

  This is another big Rambus-related disaster for Intel, which have been coming at a regular rate of about 1 a month for the last half year.  It can't go on forever.

                               Dr. John

Direct X 8.0 Preview.

  If you are interested in reading an article on the details of the upcoming game API from Microsoft, check out this article over at Reactor Critical.

Thunderbirds Are Go, But What About KT133 Mobos?

  The 'release' of the new socket-version of the Athlon, with on-die L2 cache, occurred yesterday, but so far, we have been unable to find any KT133 socketed Athlon motherboards for sale.  According to many motherboard makers, KT133 motherboards are in production, and should be ready for shipment in the next few weeks.

  The new Athlons retain the same name, rather than getting an updated moniker, and will be distinguished from the older versions with a logo on the box that states "Performance Enhancing Cache Memory".

June 5th 2000

T-Birds Everywhere!

  If you ever watched the old Puppet and model show called Thunderbirds when you were a wee lad, then you'll want to say "Thunderbirds are Go!" too.  That's the only news story on half the Internet today.  Why is it such a news item?  Because AMD now has a new processor, code-named Thunderbird, based on 0.18 micron technology, with 256KB of on-die cache, and it fits in a new socketed motherboard.  Sound familiar?  Well, it's about as close to an Intel Pentium III Coppermine as AMD can muster. Benchmarks are coming out, and if you were hoping that AMD was going to blow Intel out of the water, you're in for a disappointment.  The Coppermine PIII and Thunderbird Athlon are pretty close to one another in speed at the same clock setting.

  But anyone who thinks that it's no big deal since the T-bird is not faster than the PIII is missing some major points.  First, the T-bird costs less at the same speed, second, they will probably be more plentiful than Coppermine PIIIs, and third, AMD will have 950MHz and 1GHz models ready in quantity before Intel does.  If you suggested all of this two years ago, when AMD was struggling, folks would have laughed at you.  Now it's a reality, and it's occurring at a time when Intel is having severe problems with new motherboard design and production.

  Two plants are producing T-birds, one in Germany and one in Texas.  The German plant will be using the newer "Copper Interconnect" technology in the chip manufacturing process, so rumors have been circulating that the German chips may be more overclockable.

  Just like Intel's new Flip-Chip Pentium III, the T-bird will go into a new processor socket, known as the Socket-A format.   It is very hard to say how long it will be before Slot-adapters and Socket-A motherboards will become generally available.  But considering that many of Intel's current problems relate to the lack of good motherboard support for their new processors, AMD would be wise to avoid the same mistake of releasing a processor before motherboard support was ready to go.  So we know the Thunderbirds are Go!, but are the motherboards?  Let's hope so.

                               Dr. John

Reviews: 
Sharky Extreme
Tom's Hardware
CPU Review
Gamer's Depot

June 4th 2000

Intel Flukes Up Again.

  An internal Intel document, leaked to our beloved Mike Magee at The Register, has us nearly in fits of laughter, if it weren't so tragic, that is.  Indeed, Intel's recent history reads like a poor Greek tragedy (et tu, AMD?).  It is now clear that Intel's troubles are internal, structural, and deeply entrenched, and they aren't going away any time soon.  The argument can be made that AMD is driving Intel to do stupid things (like 'release' non-existent 1GHz Pentium III chips), but most recent Intel debacles have been truly Intel-inspired.  Take, for example, the adoption of the most expensive memory technology on the planet (long before it's ready, or needed), the botched i820 engineering, and the embarrassing memory translation hub incident, to name just a few.

  The leaked Intel document has so much interesting stuff, you may just want to go over to The Register and read it here. But here are the good parts, with my translations of Intel-speak:

1) "Intel expects that both Pentium® lll and Celeron™ processor supply will remain tight through the remainder of Q2'00. This is not a product or process health issue." (Doc's translation: go buy AMD CPUs for the next few months)

2) "Intel will continue to ship Pentium lll processors based on .25-micron technology (prior code name: Katmai) into the distribution channel during May, June, and July '00....This is meant to ease the transition to the FC-PGA form factor and to aid in reducing distribution channel inventory on older technology boards."  (Doc's translation: No FC-PGA motherboards available, but plenty of BX boards!)

3) "On March 8, 2000 Intel announced the introduction of the Intel® Pentium® lll processor 1GHz. Broader quantities of Pentium lll processors 1GHz will be available in the Q3'00 time frame. Shipment of Boxed Intel Pentium lll processors 1GHz are expected to begin at that time." (Doc's translation: we announce stuff 6 months before we can deliver!)

4) "On May 24, 2000 Intel introduced the Intel® Pentium® III processor 933 MHz.... Pending completion and validation of a new thermal solution, the boxed version of this processor is expected to begin shipping in late June." (Doc's translation: The Coppermine runs hot at 933MHz, and our heat sinks ain't good enough!)

5) "Intel is transitioning the Pentium lll processor line from SECC2 to FC-PGA packaging. Intel expects that about 1/2 of the Pentium lll processors shipped to authorized distributors in Q2'00 will be in the FC-PGA package. This is expected to increase to about 75% in Q3'00. Customers who have not already done so, should transition their Pentium lll processor-based desktop system offerings to the new FC-PGA form factor now.... Intel is now shipping CA810E desktop boards to support Pentium lll processors in FC-PGA package and is planning additional board offerings. Other motherboard vendors are also shipping socketed boards that support Pentium lll processors in the FC-PGA package based on 810E, 440BX and 3rd party chipsets. For more information about boards and slot-to-socket adapters go to the Technical Center on program.intel.com" (Doc's translation: We are so desperate to sell Flip-chips, for which almost no motherboards exist, that we will even mention slot-adapters and 3rd party chipsets!!!) 

6) "As previously indicated, Intel has decided to temporarily suspend shipments of Boxed Intel® Pentium® lll Xeon™ processors 933, 866, 800, 733, 667, 600 MHz with 256k on-die L2 cache (prior code name: Cascades).... This suspension is in response to delays in the availability of motherboards that support these processors in the distribution channel" (Doc's translation: It's hard to sell CPUs without any motherboard support, and who's fault is it that there are no motherboards?)

  I would estimate that about 80% of Intel's current problems stem directly from their ill-advised adoption of Rambus memory as their platform of choice for RAM. This has created the shortage of functional motherboards that will take their new processors. The other 20% can be attributed mostly to their rush to one-up AMD. Both of these problems were perpetrated by Intel's marketing team, perhaps with the backing of Intel's upper management.  I am certain that the engineers at Intel are not that naive, and have probably been pressured into cobbling together the mismatched junk on recent Intel motherboards.  It seems more likely they told marketing to go with Rambus or SDRAM, not both, but were quickly overruled. 

  It is inexcusable that Intel 'released' the Flip-Chip Coppermine PIII in October of 1999, and in June of 2000, there are still almost no native Flip-Chip capable motherboards to be found.  I have a feeling that AMD will make certain that socketed Athlon-Ultra motherboards are available much sooner after tomorrow's debut of their new processors, code named "Thunderbird".

                               Dr. John

June 3rd 2000

Anand Backpedals on Rambus Endorsement.

  Anand Shimpi over at AnandTech.com has been doing a series on Rambus memory, and why it will be a good technology for future computers.  He has gotten lots of criticism for his relatively pro-Rambus stance, and his lack of discussion of Double Data Rate DRAM when talking about the benefits of, and future need for Rambus memory.  Well, the criticisms appear to have struck a nerve with Anand.  His latest installment does some backpedaling, and makes the admission that Double Data Rate SDRAM will fulfill most of the same requirements, at a much lower price.  While he does not say that Rambus is down for the count, he admits that Double Data Rate DRAM will almost certainly be the memory of choice for the vast majority of computer enthusiasts for the foreseeable future.

                               Dr. John

Computer Parts Prices to Rise?

  Another article over at EBN suggests that the shortage of certain electronic parts is having an inflationary effect on electronic device and computer manufacturers.  Some computer parts prices are increasing, but for the most part, most are still going down, or remaining steady.  The parts shortage is having more of an effect on mobile phone and hand held device prices and availability than on prices in the computer industry.  But components like DVD drives, which rely on parts found in mobile phones and hand held devices, are in very short supply.  If manufacturers do not gear up production of the required parts, the shortages may begin to have a severe effect on the computer and electronics industries.  It is interesting that the current CPU shortage has not led to processor price increases, due to the ongoing price war between AMD and Intel.

June 2nd 2000

Abit Releases a Flip-Chip Motherboard, the VH6.

  Abit announced yesterday that they would soon be releasing a Flip-Chip socket motherboard for Coppermine Pentium III and Celeron-2 processors.  The new motherboard, designated the VH6, is Abit's first board with a Flip-Chip (FC-PGA) socket which can accept the socket-version Pentium IIIs and Celeron2s.  Previously, if you wanted to overclock a "Coppermine" core Flip-Chip, you needed an FC-PGA socket adapter.  These adapters often cause problems of their own, so a motherboard that takes Flip-Chips without an adapter will simplify system setup, as well as reduce cost.  The VH6 should be available this month.

AMD Athlon-Ultra and Duron Processors to Debut.

  AMD is set to release the Thunderbird version of the Athlon (Athlon Ultra) next week.  The Duron, which is meant to compete with the Celeron line, will be released the week after.  So far, AMD has been much better at getting chips to market after they are announced, while Intel often announces a release one to two months before the new chips actually become available.  Very few motherboards capable of taking the new processors are available now, but motherboard manufacturers have them in the works.  The Athlon-Ultra and Duron processors will be mostly produced in the new socket-A format, which is why new motherboards are required.

  More chipsets are coming out that will support the socketed Athlon Ultras and Durons, including a single-chip designed by SiS.  This small chip manufacturing company will produce very-low-cost chipsets for the AMD line, creating some competition for VIA, who has the lion's share of the Athlon chipset market currently.  The new single-chip design has been designated the 'SiS730S'.  Motherboards based on the 730S should be available this Fall.

                               Dr. John

June 1st 2000

Voodoo5 Cards Shipping Again.

  3dfx has announced that it will resume shipping Voodoo5 cards immediately.  They expect that the cards will reach retail stores for sale on Friday, June 9th.  It's good to know that the recall was just an example of the company being cautious.  3dfx did not identify what the problem was, but apparently, it was not serious.  Let's hope the extra 2 weeks gave 3dfx enough time to get even more product ready to ship.

  In related news, Kyle over at HardOCP got some email from Brian Burke at 3dfx, in which Brian discusses the V5 delay in the context of "shipping vs. non-shipping" products.  It makes a good read.

                               Dr. John

Intel Delays the i815 Chipset Debut by Two Weeks.

  An article at EBN says that Intel has delayed the release of their chipset with PC-133 SDRAM support.  Poor sales, and the Memory Translator Hub problems with the i820 chipset led Intel to push forward with mid-level chipsets, the i815 and i815E.  These chipsets have support for PC-133 SDRAM, an open-standard technology that Intel said it would not support.  The E version also has ATA/100 hard drive support.

  We are eagerly waiting to see if Intel got it right with these new chipsets, or if they are going to be as disappointing as the i810 and i820 chipsets.

                               Dr. John

Copyright 2000, KickAss Gear