May 30th 2000

Computer Component Shortages Worsen.

  An article over at EBN discusses how bad the computer parts shortage is getting.  It's hurting all computer makers, as well as cellular phone companies, and handheld device makers alike.  High demand for technology goods has led to the shortages, and it's getting worse, not better.  DVD drives have been particularly hard hit, due to shortages of Flash Memory chips, digital signal processors, and microprocessor-controllers.  Computer makers are also having trouble obtaining enough Intel processors to meet demand.  AMD has been able to take up some of that slack, but processors at the high-end of the performance scale are very scarce.

  Obviously, there's lots of money to be made here if parts makers can gear up production.  Many companies, including Intel, are in the process of increasing their production capacity, so we can hope that these supply problems will not last too much longer.   But expect many computer parts to remain in short supply through June.

                               Dr. John

May 29th 2000

Voodoo5 Problem Found, but Not Disclosed.

  A post on the Voodoo5 newsgroup (3dfx.products.voodoo5) states that the problem with Voodoo5 boards has been identified, and can be fixed quickly.  The problem appears to involve only a small number of Voodoo5 cards.  Rumor has it that the problem is with the power supply to the boards.  A new shipment date of June 9th has appeared at several distributors.  Let's hope this is really the correct date.

                               Dr. John

Microsoft Trial Puts New Software On Hold.

  According to an article in The Register, Microsoft has canceled it's June 1st announcement concerning their new "Next Generation Windows Services" (NGWS) initiative.  NGWS is supposed to be a new grouping of Win2K services that integrate Windows with the Internet more robustly.  According at an announcement by Steve Balmer about the project; ... "A key set of NGWS will be hosted on the Internet and will be infused into future versions of Windows. ... NGWS ... is the foundation of the company's software services strategy."

  Now we can see why they canceled the announcement; they are in court right now, in part, because they tied Internet Explorer to Windows and gave it away for free.  So obviously, in light of the fact that they may be ordered into a forced breakup into operating system and applications companies, they can not announce that they are about to integrate more Internet stuff into future versions of Windows, at least until this court case is settled. 

  This does not affect gamers and the upcoming version of Windows for them, Windows ME (Millennium Edition).  NGWS will not apply to Windows ME, only to Win2K, the corporate version of Windows. Windows ME, which is really just an improved version of Windows 98SE, should be out late this Fall.

                               Dr. John

May 27th 2000

Microsoft Breakup? (what we need is someone to break up Bill)

  The Microsoft Antitrust case is reaching the penalty stage, and all the buzz on the Internet is about the possibility of a forced breakup into 2 or 3 smaller companies.   Unless the wheels of Justice are suddenly transformed into a well oiled machine, the speculation is a bit premature.  The whole thing is headed for the Appeals court, and any real remedies, if they ever make it to the light of day, will be a long time coming.

  The argument now is whether the proposed breakup into 'Operating System', 'Applications' and maybe a hardware or Internet company, would actually increase competition, and prevent further illegal and anti-competitive behavior by Microsoft.  The answer is clearly, no.  The only thing that would really change Microsoft's behavior would be a corporate decapitation.  Remove the head, and the beast will die.  Let's face it, Bill Gates is Microsoft's worst problem right now.  If he would just step down, and let other people get on with the business of making software, their situation would improve greatly.  The time for a mutiny is ripe.  But I doubt highly anyone at Microsoft is brave enough to face Bill with the truth.  

  Even if Microsoft is broken up, there will still only be one company producing Windows, and very little operating system competition at the consumer level.  Most PCs will still run on Windows.  The only real benefit to the consumer will be that it is fairly certain that various add-ons to Windows, like Internet Explorer, will either have an uninstall option, or will go back to being a real add-on.  This will reduce the size of the operating system, and will also increase it's performance and reliability.  This is very good news, considering how large, slow, and unreliable Windows 98 is right now.  I just wish that Bill would grow up, admit error, and get this all over with.  Can you say, "Snowball's chance in Hell"?

                               Dr. John

May 26th 2000

Voodoo5 Recalled!

   3dfx announced on Wednesday that there will be another delay in the Voodoo5 release. I speculated yesterday that the cards must have already been boxed and ready to ship to big distributors.  It appears that in fact they had already been shipped, and are now being recalled from the distributors.  Here is the blurb I saw:

"After careful consideration, 3dfx has made the decision to bring back the Voodoo5 AGP products that have been shipped. 3dfx has a business model that allows us to maintain a limited number of direct customers, while still being able to maximize availability points to the end user. This means that the boards we have shipped were sent to less than a half dozen distributors, who we maintain a very close working relationship with. We are bringing the boards back to determine if those boards exhibit the same problem. We believe that the overwhelming majority of those boards will test out fine. We are taking this step to minimize any consumer returns through our retailers and to ensure that the products we ship will be of the highest quality. "

  According to 3dfx, the problem is isolated to a few chips, and may not have existed in prior manufacturing runs.  If true, then it sounds like a problem with the VSA-100 chip itself failing on some boards, in some configurations.  This seems incredibly similar to the problem Intel had with the original release of i820 motherboards last Fall, where they recalled them all from warehouses two days before their debut.  In both cases, the product made it past what we assume was months or rigorous testing, only to have a major problem discovered just days before the product introduction.  Let's hope that it turns out to be a few bad chips on one run.  Otherwise, there's going to be a wait.

  3dfx says the delay will only be one or two weeks, but I have a feeling it will be more like a month or more.  It would take a week or two just to get the boards back and tested.  If there is a hardware problem, it will take at least a few more weeks to sort out, since the boards are already assembled and packaged.  If the number of shipped boards is high, then the cost of returning, repairing, repackaging and re-shipping the boards could be considerable. 

  This is the worst news we've heard in awhile.  3dfx may be in trouble if they don't get this worked out quickly and in a cost-effective manner.  The company posted a loss last quarter, so the stockholders have got to be getting worried.  The main thing that 3dfx has going for it now is that many game enthusiasts that did not buy a GeForce card are waiting for the hardware support for Full Scene Antialaising that the Voodoo5 will bring to games.  It is a feature that will significantly improve the appearance of virtually all existing and future 3D games by removing visual artifacts from the rendered images.  But the question is, how long can 3dfx fans hold out?  NVidia has released two new generations of video card since the Voodoo3 cards were introduced.

  I hope 3dfx does not go into quiet mode now.  The entire PC gaming world is watching and waiting, and the PR boys had better earn their keep.  3dfx needs them now more than ever.

                               Dr. John

May 25th 2000

Voodoo5 Delayed Again!

   3dfx announced yesterday that there will be another delay in the release of the new line of Voodoo cards.  This can only be described as a disaster for the company.  All I can say is, if 3dfx goes under, and we are left only with NVidia, the world of 3D graphics cards will be much the poorer, and prices and choices will suffer.  The Voodoo5 cards were originally scheduled for release before last Christmas, and this delay has got to be very bad news for the company.

  The problem is apparently in the hardware, and has to do with failure rates on the cards.  This means to me a bad component, or a bad design, and neither type of problem will be easy to fix properly.  A software workaround may be found, but would be likely to adversely affect performance.  Here is an excerpt from their announcement:

"The company discovered that the Voodoo5 may be experiencing field failure rates at very low levels in certain configurations. The company is conducting further tests to determine whether a problem actually exists. 3dfx anticipates this action will delay product availability between seven and 14 days."

   So there you have it.  No Voodoo cards in May, as we were promised.  If 3dfx does actually identify a real hardware problem, I doubt highly they will have it fixed in two weeks or less.  My guess is that Voodoo5 cards will not go on sale until the end of June at the earliest.  Not good.  Indeed, at this point, I would think that many Voodoo5 cards are already in boxes ready to ship, so unless the fix is with a software patch, we may have a real wait on our hands.

                               Dr. John

More Worthless Chipsets From Intel.

   If you are tired of reading about Intel's problems, then you can stop reading now.  But if you love to kick a guy when he's down, or beat a dead horse, then read on!  First, you can check out the Opinion piece by Andrew Thomas at The Register.  He is saying stuff about Intel, out loud, that I have been saying for months.  Let the engineers take charge, and get those bone-headed marketing guys out of there!!!

  But the chipset lineup from Intel is what is really driving me crazy.  Every upcoming chipset from Intel is a dog, and if they stick to the roadmap, they are certain to send the company into a tailspin.  Integrated graphics on the i815 Solano, marketed towards enthusiast-level customers?  Only Rambus DRAM support on the new i850?  I sure as hell don't want one.  I'd buy a BX motherboard first.  Intel's upcoming chipsets are not targeted at the home-enthusiast, they are a bizarre mixture of cheap junk, like audio-modem riser slots, and overpriced hypeware, like Rambus DRAM.  I am not even sure that business users will want to go the Rambus route, so who does Intel think is going go buy their new motherboards?  They obviously didn't learn a thing from the i820/Rambus disaster, which they are still suffering the torment of the damned over.  

                               Dr. John

May 24th 2000

Intel Processor Price Drops.

   What irony.  As everyone scrambles to find non-existent Pentium III and Celeron II processors, Intel is dropping the price again. This is the second Intel price drop in as many months, but there are still no processors in the supply chain to buy, at any price.  They are all going to Dell to bribe them to not go with AMD.

  But just in case you are interested, Dell will be paying less for those processors starting on Monday.   And if you can find any of the left-overs, you will be able to get them for less too.   The processor shortage is going to continue for at least another month or two, especially if demand remains high.  

  As I've said before, if Intel would let prices slide according to demand, the prices would be higher, but supplies would be much better too.  As long as prices keep dropping while demand is climbing, there will be very few processors available on the open market. 

                               Dr. John

Why Intel is Not Using Rambus On it's Server Platforms.

  It was reported in The Register today that Intel's recent road map reveals why Intel has opted for Double Data Rate DRAM (DDR-DRAM) support on their upcoming  server platform (Foster), even though they are supporting Rambus memory on their desktop platform (Willamette).  The reason is that Rambus is an inferior technology in several key respects.  That's correct, even in Intel's thinking, Rambus has technical deficiencies that make it unsuitable for advanced server applications.

  According to the article at The Register, the Intel slide pertaining to the topic read: "The main issue with Rambus memory in servers is the maximum memory size, volume density and also "chipkill", which is not an issue in the data processing market." So there we finally have it.  Rambus offers less total memory capacity, has a lower memory volume density, and does not support  "Chipkill" technology.  Chipkill is an IBM server technology that is very much like RAID for memory.  If one bank dies, the system drops that bank on the fly, and others are there to take it's place.  Obviously, the 4 sleep-wake modes that Rambus can be in at any time, and it's serial interface nature, do not lend themselves to the Chipkill scheme.  This is interesting, since a report in The Register, and an announcement at Rambus last summer, both suggested Rambus had their own memory fault tolerance technology similar to IBM's Chipkill.

  So Intel has finally admitted that it has always thought of Rambus as a consumer-level product, despite it's extremely high price.  You would think that Intel would be trying to figure out how to make home computers faster and cheaper, not slower and more expensive, but then again, what do I know?

                               Dr. John

May 23rd 2000

Computer Parts Shortage Continues.

  An article at EBN discusses the severe parts shortage that computer makers are still experiencing.  Many computer parts have been in very short supply since last Winter, and recent high demand for certain parts has made the situation worse.  If parts continue to remain in short supply, prices could creep upward.  The shortage means that smaller companies are getting shorted on deliveries as the larger OEMs pressure suppliers to give them first dibs.  It seems hard to believe that manufacturers have maxed out their fabrication capacity, but that's the reason given for the shortages.  Parts we are having extreme trouble getting include, 7200rpm hard drives, DVD drives, high-end video cards (DDR especially), all coppermine Pentium III processors, and high-speed Athlon processors.

NVidia Detonator 5.22 Drivers 'Released'.

  In usual fashion, the newest Win9x Detonator reference drivers for NVidia-based cards has been leaked to the web.  You can find them here.

May 21st 2000

Intel's Backup Plan for the Pentium IV Chipset (and the slow death of Rambus).

  It was reported on the German web site "PC Welt" (PC World) that Intel has a secret chipset plan for the Willamette (upcoming Pentium IV). As reported in The Register, the chipset has the code-name Amador, and includes support for SDRAM! This despite Intel's insistence that the Willamette processor would work only with Intel chipsets that supported Direct Rambus DRAM.  It is not clear how much effort Intel is putting into this chipset design, which goes against everything Intel has said for months about memory support in it's Willamette-compatible chipsets.  But it's always a good idea to be prepared for contingencies (like ridiculously high prices on Rambus RIMMs), and maybe, just maybe, Intel is tiring of getting caught with their pants down.

  It is now abundantly clear, even to Intel, that they pushed Rambus too hard and too fast before it was needed, and before the technology was ready.  The higher bandwidth offered by dual-channel Rambus will not benefit computers until the "Northbridge" bus is more data-saturated than it is in current Pentium III computers.  Willamette (Pentium IV) is the first chip in the Intel line that could have benefited from dual-channel Rambus on a performance basis.  And yet, dual-channel, Double Data Rate DRAM could have filled that bill just as well, with less trouble for everyone involved.  But it would have made Intel less money, so I suppose that the old motto "The customer is always right"  has been superceded by "Greed is good" at Intel.

  If you doubt that Double Data Rate DRAM does indeed offer the performance of Rambus (and then some!), check out this exhaustive review of Dual-channel Rambus on Intel's i840 chipset vs. Double Data Rate DRAM on Micron's new Samurai chipset.  The Double Data Rate DRAM pounds the i840 with a vengeance, in most platforms, and under most tests.  Rambus only won in a few cases by a small margin. Here is a snippet from the article on memory writes: "As seen in the table, the average performance difference for Writes is 45.5% favoring DDR."

  In my opinion, this report crushes the hype put out by the Anandtech article, by doing real world testing, rather than hyping hardware for some freebies.  The article at Simmtester.com is the second nail in the coffin for Rambus DRAM (Tom's hardware review was the first).  The real kicker is that the article at Simmtester compared single -channel Double Data Rate DRAM to dual-channel Rambus DRAM, and Rambus still lost. This means that when dual-channel Double Data Rate memory becomes available, it will crush Rambus soundly.  So the question is, why is Intel foisting high-cost, low performance memory on it's customers?  Good question.

                               Dr. John

May 18th 2000

Memory Makers Tell Intel to Take a Flying... (well, not quite that bad).

  It was reported at EBN today that Intel had a pep talk with major memory makers to cajole them into boosting Direct Rambus DRAM production.  It apparently did not go particularly well for Intel. The memory makers were consolidated in their position that they will only produce as much DRDRAM as they get confirmed, non-cancelable orders for.  Here is a taste of the fun.

The meeting at Intel's Chandler, Ariz., campus, which was to have been a two-day affair, ended early when DRAM suppliers told Intel they would increase Direct RDRAM production levels only if their customers in the PC sector agree to a non-cancellation guarantee on each order placed.

  At current Rambus prices, anyone want to place a large, expensive, non-cancelable order..... anyone?  What do they do to you if you try to cancel an order?  Have you met Guido?  So many questions.

                               Dr. John

May 17th 2000

Tech Report Reports on E3.

  The folks at Tech Report have a nice hardware roundup of stuff they saw at the E3 Expo.  The coolest hardware they saw was probably the new 3dfx cards, which you can read about here.  They may be late, but all the reports suggest they have better image quality than the GeForce 2.  For the hardcore (and well financed) gamer, the V5 6000 will blow the GeForce 2 away.

Rambus vs. DDR-DRAM.

  The pro-Rambus article posted at AnandTech on Monday has sparked an anti-Rambus tirade on the AnandTech message-board.  You can check out all the counterpoints to his article here.  I did not see a single pro-Rambus comment, and many of the posts made excellent arguments about how quad-pumped, dual-channel DDR memory technologies will be easier and cheaper to develop and produce than quad-pumped Rambus memory, not to mention cheaper.  In my mind, the biggest problem with Rambus memory is that it is a proprietary technology, while DDR is an open JEDEC standard.  Going down the proprietary (Mac-like) path is not going to sit well with PC enthusiasts.  It doesn't sit well with me.

  You can check out this article here if you would like to read more about the Willamette (Pentium IV) and it's reliance on Rambus memory technology.  It's a good read on the memory business in general.

                               Dr. John

May 16th 2000

Intel Employees Love Their Jobs.

  Ever wonder how come those guys in the Intel bunny suits are always dancing?  Maybe it's because they're so happy... or maybe they put powerful electrodes in those suits.  It's anyone's guess.  But if the current and ex- Intel employees at FaceIntel.com  are any indication of the mood at Intel these days, then they are going to need electrodes to get them to dance.

  The titles on some of the article's written by these Intel employees tell it all:

 "Intel's sweatshop in Panang, Malaysia" 

 "Corporation or a Dictatorship" 

 "Intel's Lawlessness" 

and the list goes on and on.  We may have a strong cynical leaning when it comes to Intel's marketing and distribution practices, but gosh, even we weren't willing to go all the way and start calling Intel "The Great Satan of Chipmakers" like some other web sites are prone to do (you know who you are!). But my guess is, with all the i820/Rambus problems Intel is having, that there will be more employee layoffs, and more unhappy ex-Intel employees over the coming months.  If their stories are even half true, then life at Intel is apparently very depressing.

  So the next time you see an ex-Intel employee with a cup in hand on the street corner, don't just drop a quarter in the cup, give the poor bloke a hug!

                               Dr. John

May 15th 2000

The Brighter Side of Direct Rambus DRAM.

  A very nice article over at AnandTech discusses the need for higher bandwidth memory systems in future computers.  It dispels some of the rumors that have surrounded Rambus memory, such as the very high operating temperatures, and says Rambus will be needed in computers sooner than people think.  Processors such as the Willamette and Mustang will be "Northbridge bandwidth limited", meaning there isn't enough room on the memory to CPU bus to keep up with CPU speed.  So in a year or less, computers will need Rambus, because even DDR at 266MHz will only have about 2/3 the bandwidth of dual-channel Direct Rambus DRAM.  And even dual-channel Rambus won't be fast enough for the CPU running at 1.5GHz.  So Mustang (Athlon-2) systems will be adversely affected by the use of Double Data Rate DRAM as compared with Willamette (Pentium IV) systems using dual-channel Rambus memory.

  The article almost had me convinced to go buy some Rambus memory right away, until I thought about it.  First off, while people spent $600 for a stick of 128MB of RAM many years ago, it doesn't mean many are willing to go back to those price ranges for memory.  The main reason that people have been willing to pay lots more for something in the past was a significant increase in performance or features.  Rambus offers neither, until memory bandwidth becomes more of a limiting factor in computer performance.

  Also, the article shamefully downplayed the yield problems manufacturers are having producing 800MHz Rambus memory, the only flavor that currently performs better than PC-133 SDRAM.  The vast majority of Rambus chips end up being rated at 600MHz or 700MHz.  

  Finally, the article did not discuss the prohibitive cost of dual-channel Rambus.  Currently, only the Intel i840 chipset supports dual-channel Rambus. What dual-channel means is that in order to get a bandwidth higher than DDR DRAM, Rambus needs to have two channels for memory, with two RIMM banks on each channel.  That's double the number of Rambus RIMMS you will need to buy to get that higher bandwidth.  That means double the cost.

  It's obvious that Rambus could become a major memory option for personal computers if they would just lower the price.  Rambus Inc. insists that the price is high because demand is high, while supply is low.  One of the reasons that supply is low is the poor yield problem, despite the denials from Rambus.  But the main reason Rambus costs so much is that everyone is trying to make back their huge investment in the new equipment required to make and test Rambus memory.  More memory makers are gearing up their production of Direct Rambus DRAM, so let's see if the price drops.  If it gets to the point where it's only twice the price, and noticeably improves performance on high-end systems, then Rambus' day will have come.

                               Dr. John

May 13th 2000

Intel to Computer Makers... "Hang On, We're Working On It."

  Intel i820 motherboard owners wait with excited anticipation as the latest Intel gaff looks like it just might end up getting each of them a free stick of Rambus memory.  According to Intel, only i820 motherboards with the memory translator hub, produced after November 1999, will be eligible for the recall.  All you folks who got systems with Rambus memory, and no translator hub, are out of luck.

  An article at EBN discusses the topic, albeit using some very dubious numbers.  For example, check out this quote in the article.

“From a Rambus standpoint, since supply is very tight and prices have a relative add of 30% over synchronous [DRAM] parts, I don't think there'll be anywhere for prices to go but up,” said Mark Giudici, an analyst at Dataquest Inc., San Jose.

  Sorry Mark, Rambus costs 4 to 5 times as much as SDRAM, not 30% more!  Even for large customers, the price differential for Rambus is at least 2 to 3 times that of SDRAM.  This fiasco will cost Intel $100,000 to $200,000 in the estimates of many analysts.  And it won't garner any increased confidence in Intel.  We have been saying all along that Rambus was going to get Intel into trouble, but we had no idea it would be this much trouble.

  Many computer makers (not us, we wouldn't touch Rambus with a 10 foot, anti-static pole) are now sitting on stocks of bad Intel motherboards, and can not do anything with them until they get the word from Intel on how to proceed with the recall.  Third party motherboard makers who have been using Intel's i820 chipset and memory translator hubs are in the same boat. The funny part is that many motherboard makers were not happy from the beginning with i820 motherboards (remember the first recall in October '99, when the 3 RIMM problem surfaced?). And now here the motherboard makers are, 7 months later, with warehouses full of the same botched stuff from Intel, waiting for instructions from Intel on what to do with it.  It's all making AMD Thunderbirds with Double Data Rate memory sound awfully nice right about now.

                               Dr. John

May 12th 2000

Free Rambus Memory!

  "Stee-ep right up and git yer Freeee Rambus memory, one per customer pleeease!" hawked the Intel representative as he waved glimmering RIMMs of Rambus memory over his head.  The image is enough to bring tears of joy to the eyes of Camino-chipset motherboard owners.

  Intel's recall of i820 motherboards with a defective memory translator hub has the rumor mill churning at full speed.  The first bit of scuttlebutt is that estimates can now be made of the number of Rambus-populated motherboards Intel has been able to sell over the last 6 months.  This gives analysts the ability to determine the relative amount of market penetration that Rambus has made over the last half year (info Intel does not discuss).  The estimates suggest that between 350,000 and 500,000 Rambus-populated motherboards have been sold by Intel.  That means a market penetration in the first 6 months of less than 1%.  That can't be very good news for Intel or Rambus.

  The second, much more popular rumor, is that Intel will fix the problem by giving free Rambus memory modules (RIMMs) to the customers who return their motherboards for repair or replacement.  While this sounds great, and is certain to make the customer support phones ring glorious, the fine print will disappoint.  Rambus has been having trouble producing enough 700MHz and 800MHz Rambus RIMMs, but has an oversupply of 600MHz parts.  Can you see the light bulbs lighting in the busy little heads at Intel? You can be sure that any Rambus giveaway will involve the slowest speed Rambus modules, which do not even perform as well as PC-133 SDRAM on a BX motherboard.  The lackluster performance of the slow-speed RIMMs has made them very unpopular, even though they are by far the most numerous Rambus modules around.

  If you were duped into buying an i820 motherboard system after November 1999, it is your sworn obligation in a capitalistic society to demand your just compensations for a very expensive, defective product.  So relax with a cold one, get on that 1-800 number and wait your turn on line, as you listen to the cheery top 40 music.  And when your turn comes, with just the right amount of righteous indignation, demand your FREE Rambus RIMMs, today!  600MHz notwithstanding, FREE is good!

                               Dr. John

Nature Lovers, Be Heard!

   Since the paperless office is a dream that may never become a reality, we computer-users might want to make sure the Paper Barons don't defoliate the Earth while we're waiting.  You can go here to send an electronic (paperless!) post card to the White House, asking for a permanent ban on any further road-building or logging in the fast receding National Forests.  (Our public service announcement for the week! Happy Friday folks).

Creative to Get Competition in Sound Card Market.

   After Aureal, producers of the popular A3D sound  chips found on cards like the Diamond MX400, went out of business in March, there has been little competition for Creative Lab's Sound Blaster line.  But VideoLogic and Voetra/Turtle Beach are teaming up to stomp on the Sound Blaster.  Videologic is producing a new audio chip with dramatically more capabilities than the Sound Blaster Platinum card, and cards based on it will probably be in the same price range. 

Says Kevin Dale, president, VideoLogic Systems: "This product sets a new benchmark for PC audio, coupling a re-configurable DSP core, audiophile performance and six-channel sound with a feature set you’d only expect to find in expensive pro-audio gear. This affordable and versatile sound card will enhance every major PC audio application from MP3 digital music through gaming and 3D surround sound to PC home theatre..."

  In addition to providing six output channels, the product supports every major 3D audio standard, including EAX™, A3D™, I3DL2, DirectSound3D, as well as Sensaura’s MacroFX™ and MultiDrive™.  

  You can read about the new audio chip here.

May 11th 2000

Intel Recalls Nearly 1 Million "Camino" (i820) Chipset Motherboards.

   Oops, I left one big Intel problem out of the list I posted early yesterday morning....

   It's official.  Intel will be recalling all i820-based motherboards that were produced after November 1999.  At least as long as the customers request it (so if you got one, request it!).  The problem is with the now infamous Memory Translation Hub (MTH) which allows less expensive SDRAM to be used on Intel's Rambus-compatible motherboards.  This was required due to the high cost and low availability of Rambus memory modules. 

  Rumors are circulating that they may populate the replacement boards with Rambus RIMMs at no charge!  That would be the most interesting means I have ever heard of to "promote Rambus as the dominant memory platform for PC's until 2002".  It won't help Intel's bottom line for the quarter, nor will it instill greater confidence in Intel's engineering prowess.  But freebies always help with customer relations.

  If you own Rambus stock, you can read a confidence-boosting article here.  But avoid the discussion forum on the topic, it will burst your bubble.  

                               Dr. John

May 10th 2000

Coppermine Recall? No, But Close!

   A report by Dean Kent over at Real World Technologies states that Intel is having so many problems with the new Flip-Chip Pentium IIIs, that they have gone back to producing older, 0.25 micron Katmai-core Pentium IIIs in volume.  This means even more shortages of Coppermine Pentium IIIs in the future.  But if you want an old Pentium III, just about when AMD is getting the Thunderbird ready, then you're in luck! ;) 

  Below is a list of some of the problems Intel has experienced during the conversion to production of Coppermine Flip-Chips.

1) Serious design flaws in the i820 chipset required for Flip-Chip Pentium IIIs. VIA has a Flip-Chip compatible chipset that is less expensive, and has more features than the i820

2) Over 65 bugs found in the original release of the Coppermine Pentium III (fixed in the next stepping).

3) Unpopularity of i820 motherboards due to RIMM problems, and the high cost of Rambus DRAM

4) Almost no native Flip-Chip (FC-PGA) motherboards available, forcing customers to use Slot adapters

5) Intel's inability to get the materials and parts needed for the mass production of Flip-Chips

6) Flip-Chips cracking due to the design of FCPGA-370 sockets and the excessively strong strong heat sink clamp Intel includes with their retail processors (discussed below).

  The rumors about Coppermines cracking when overclocked turned out to be way off.  The problem with cracked Coppermines (which we have not experienced yet) turns out to be a simple design flaw in Intel's heat sink clamp.  It's way too strong, and puts a huge amount of force on the center of the Flip-Chip when clamped.  This would work fine if the center of the FC-PGA sockets were solid, but they are not.  There is no central support under the Flip-Chips in these sockets, so when an excessively strong heat sink clamp is used, the chips bow downwards at the center, and can eventually crack from the strain.  Intel just needs to ease up on the strength of the clamp.

  Intel has already denied this rumor about Flip-Chip problems, and increased production of Katmai-core Pentium IIIs.  But sources in a position to know say that Intel is increasing 0.25 micron PIII production to fill the gap in Coppermine production.

  It is very difficult to say whether all of these design and production problems will affect the roll out of Intel's upcoming Willamette (32-bit) and Itanium (64 bit) processors this Fall.  My guess is the answer is yes, the debut of these chips will be delayed, and the number of available new chips will be reduced at release time.  At least Intel can say that they have learned a lot, and they have many of their problems mostly behind them.  

  But the bottom line is that Coppermine chips will become even harder to get than they already are, despite the fact that Intel "released" them 7 months ago, and should have worked the kinks out of the system by now.

                               Dr. John

May 9th 2000

New Flip-Chip Cooler Reviewed.

   Neoseeker has a review of the new Alpha PAL 6035 socket 370 cooler.  It's big, and looks pretty darn good.  The problems are the usual ones with Alpha coolers, it's really BIG, and some stuff on your motherboard might get in the way.

Comparison of AGP 1x vs. 2x with GeForce Cards on Athlon 751 Motherboards.

   The folks at Riva3D.com have posted a quick comparison of 1x vs. 2x AGP modes with GeForce2 cards running on Athlon motherboards that have the AMD 751 Irongate chipset.  The surprise is that there is a bigger difference in some video modes than we had been led to believe, but it's still not much of a difference in most modes.  It will be nice when NVidia, AMD and VIA get together and get these things figured out.

May 7th 2000

VIA is Keeping Intel Afloat!

   Them's fightin' words!  But how else can you explain the fact that more Coppermine Pentium III chips are running on VIA 133/133A chipset motherboards than on Intel's i820 motherboards?  In fact, if Intel keeps up this track of product development, VIA and AMD will certainly gain more and more market share, until Intel is no longer the dominant chip and chipset maker.  Let me explain.

  For some great insights on the moves and contingency plans Intel has for it's upcoming processors, check out this article at Electronic Buyers News.  The short of the story is that Intel is still insisting that the Willamette processor, scheduled for an October release, will work on motherboards that support Rambus, but not synchronous memory (SDRAM).  There are rumors that Intel is going to try to foist another "Memory Translation Hub" on us, to allow PC-133 SDRAM to be used on Willamette motherboards.  If true, Intel is going way down the wrong road.  They know damn well that Double Data Rate DRAM offers just as much memory bandwidth as "800MHz" Rambus DRAM, at a fraction of the cost, but there is no extra profit in that deal for Intel.  So basically, greed, and a desire to have proprietary control over PC hardware has led Intel to disregard everything that their customers are telling them, and to plow ahead with Rambus-only processor platforms.  

  This is why Intel is so worried about VIA.  This little chipset maker is making the kind of chipsets that customers want, and that Intel refuses to produce. Only one thing can nip VIA's progress in the bud.  That would be if Intel makes a native Double Data Rate DRAM-compatible chipset for the Willamette processor aimed at consumer-level PCs.  You know they are arguing about it every day that draws them closer to the Willamette's release in the Fall.  Time is running out to make that big a shift in development.  In my humble opinion, it's their only hope for staying on top without loosing significantly more market share to VIA in the chipset business.  When will customers start coming before larger profits in the minds of Intel's management?

  While this kind of customer-unfriendly attitude persists at Intel, VIA and AMD are pushing the Athlon/Thunderbird line in the direction of 266MHz Double Data Rate DRAM, further improving their architecture over Intel's.  But there is one inescapable fact that has really got to rub Intel the wrong way: that without the VIA 133 chipset, Intel would have sold far fewer Coppermine Pentium III processors than they have.  The i820 chipset is a dog that few people want to bring home.

                               Dr. John

Dissecting the "Love Bug".

   While no one loved us enough to send us the Love Bug, a good friend with a knack for programming brought over a printout of the virus.  It's great reading!  The first thing that struck us was the header.  This is the "Big Clue" that led those geniuses at the FBI to announce that the virus originated in the Philippines. Here is the header:

barok -loveletter (vbs) <i hate go to school>

by: spyder / ispyder@mail.com / @GRAMMERSoft Group / Manila, Philippines 

  I love it!  What a clue... we got the guy now!  Everybody who makes viruses puts their real name and email address in the header, don't they?  And don't you love the author's title for the virus?  "i hate go to school".  It Shows!!!  And he's from GRAMMERSoft?  That's not even spelled correctly!  (I hope this guy doesn't make grammar or spell checking software in his day job).  But Grammersoft does apparently exist, and you can read about it here if you hablo Espanol.

  But perhaps the most entertaining, and considerate part of this Visual Basic Script, is the following paraphrased passage (probably lifted from another script):

mIRC Script
  please don't edit this script... mIRC will corrupt.
  if mIRC will corrupt....
  Windows will affect and will not run correctly. thanks

Khaled Mardam-Bey
http://www.mirc.com 

  And he's not just polite, he's modest:

META NAME - Description - "simple, but i think this is good"

  The virus itself is a relatively simple Microsoft Visual Basic Script (corresponding to the vbs extension you see at the end of the email attachment file).  Don't open any attachment that ends in vbs!!!! The virus does a ton of Registry editing, which is what causes many the problems the victims encounter.

  The subheadings in the virus include (among others):

main: creates a shell for executing shell commands

regruns: sets the Visual Basic Script as a service to run at each startup, downloads BUGFIX.EXE from www.skyinet.net, sets it to run at Windows Startup, and resets your Internet Explorer home page

infectfiles: finds file types available on the system for infecting, and gets email addresses from the address book.

html: creates "Love Letter to You.HTM" and writes it to the drive.

  This virus was definitely written to take advantage of some of the many security holes in Microsoft operating systems and email applications.   Just keep in mind that at least for now, you need to execute this type of script by opening the email attachment.  If you don't open it, then it's just a bunch of text in a file.

                               Dr. John

May 5th 2000

Why Did the I-Love-You Virus Work So Well?.

   Those of you who read these pages are not among the unknowing web masses that are the likely targets of the I Love You virus.  I'll bet that any of you that got the email with the attached virus already had heard about it, and were not about to open the attachment. And yet, the affectionately named "Love Bug" has caused more damage in just a few days than most viruses ever do, including infiltrating many corporate web sites.  How did that happen?   Psychology.

  Yup, the whole problem for virus authors is to figure out how to get folks to open that email attachment.  Hmm, let me think.... "I hate you"... no, that won't work,  "Important!"... nah, that's been done,  "You've won!!!".... not bad, maybe that will work, but wait!..... "I Love You!"  Eureka!  That's it!  

  That's all there is to it.  Figure out something that will get all those starry-eyed, love-starved folks sitting at their office computers to unleash the bug onto their corporate LAN.  Obviously, the authors got the email heading just right, by counting on the fact that so many people would actually fall for such an obvious ploy, and open an attachment with that label.  It should make the folks that fall for it truly embarrassed.  Maybe there will be a little less daydreaming at Corporate workstations for the next few weeks, as humiliated office workers call up their network or systems administrators to ask to have the virus removed from their systems. 

                               Dr. John

ABIT Goes Linux!

   Motherboard maker Abit announced today all new motherboards they sell, beginning at the end of May, will come with a free copy of Abit's version of the Linux operating system, called Gentus.  Abit's version of Linux offers many unique advantages including drivers for GeForce cards, as well as ATA/66 and ATA/100 hard drive controllers!  Very nice.  They have overclockers covered too, here is an excerpt from the press release:

The PerMon™ tool set is back, but now in its second version and with several new additions, such as all new graphic icons and easy-to-use scale bars.  These provide information about the CPU and system temperatures, fan speeds, CPU Vcore and other voltage readings. The “Hard Disk Tool,” which allows easy HD parameter adjustment and provides HD performance indexes, is now available for ABIT’s VIA chipset models as well as our BX models.  And for all you “tweakers” out there, the most attractive addition to the tool set has to be G-Point Explorer. This lets you adjust your Front Side Bus speed from Xwindow (KDE or GNOME) without having to go into BIOS!  Just choose your desired FSB speed, and let G-Point Explorer do the rest.  

Coppermine PIIIs Not All They're Cracked Up To Be?

   The cracked coppermine reports continue to flow into Overclockers.com.  Based on what they are saying, it looks like Intel may have skimped a little too much on the silicon packaging, making the chips prone to cracking if uneven pressure is applied with the heat sink clamp.  We can attest to the fact that the heat sink clamp on the retail Intel Flip-Chip processors takes a large amount of force to clip onto the catches.  This puts a tremendous force on the center of the chip, which may lead to microfractures.  If this has happened to you, we are hearing that Intel is replacing them under warranty. 

  Perhaps the more interesting comments in the reports are that many Intel Flip-Chip users say they have experienced general problems getting Flip-Chips to work right with various combinations of motherboards and socket adapters.  Many people say they just don't work right.  We can confirm that some combinations do not work properly. It's a real shame Intel messed up so bad with Flip-Chip coppermine motherboard support.  No one wants an i810e or i820 motherboard, so we are stuck using slot adapters in many circumstances, even though Intel does not recommend using them.  Things like this should drive Intel loyalists over to AMD.

                               Dr. John

Pirates With Attitude Get an Attitude Adjustment.

   It was reported yesterday at the crime reporting web site APB.com that a software pirate group calling itself "Pirates With Attitude" was busted by the FBI.  The article is a good read. These guys had turned software piracy into a well-oiled machine, and it's no wonder they got the attention they did from the Feds.  What a bunch of Bozos.  Even more incredible, the group included ex-employees of NEC and Intel.  Too bad they quit their day jobs!

  Only one person, the alleged leader of the group, has been charged so far, but more arrests are likely.

                               Dr. John

May 4th 2000

 Quake III Arena Internet Security Hole.

   Programmer Robert Duffy of id software has reported a serious security flaw in Quake III Arena's server software which could let a hacker overwrite files on a client system logged onto a Quake III server.  According to Duffy, Internet Security Services alerted id software to the flaw, and now id has a patch ready to fix the problem.  You can read about it here.  You can get the new 1.17 point release of Quake III Arena here.

Cracked Coppermines?.

   We have not had this experience yet, but a few posts at Overclockers.com suggest that some folks have had cracks appear in their overclocked Coppermine Flip-Chips!  This obviously leads to complete chip failure.  It will be interesting to see if these reports increase in number.  If true, it will not bode well for PIII sales to overclockers. But it is just as likely that improper heat sink and cooling system installations were to blame for any such disastrous failures.

                               Dr. John

Copyright 2000, KickAss Gear