Eric van Fast Graphics heeft wat info over de nieuwe S3 Savage 2000 chip op zijn voorpagina gedumpt:

Diamond is now officially a part of S3, the merger has completed
Diamond Multimedia will keep it's brand name.
Diamond Multimedia "might" make consumer level graphic cards bases upon non-S3 chipsets, although that's not currently in the planning.
Diamond will be the first to make a card based upon the new Savage 2000 chipset, other cards with S3 Savage 2000 chipset will most probably be in the OEM sector.
Cards with the Savage 2000 chipset will be available before Christmas.
The faster Savage 2000+ chipset (200 MHz core / 200 MHz mem) will be available in Q1 2000.
The earthquake in Taiwan did not damage the S3 plant, but the power has been cut for some time. Large volumes of Savage 2000 chips are currently in stock though.
Savage 2000 cards with 64 Mb of memory are unlikely to appear, also due to the sky high SDRAM prices.
At the moment there are only a few working Savage 2000 cards, and word is that even the press will have to wait another month before S3 starts to hand out samples.
The Savage 2000 chipset DOES support Environment Mapped Bumpmapping, 100% compatible with the hardware bumpmapping on the Matrox G400 chip.
The T&L engine is supposed to be comparable to the one in nVidia's GeForce, but S3 doesn't claim to free up 50% of you CPU time, though T&L will be very benefitial to (most of all) systems with slower CPU's. No word on professional Open-GL performance yet.
S3 predicts, contrary to nVidia, that the polygon count will not raise to drastically in the next year. Polygon counts in the low 10K area are expected, certainly not in the 100K area.
With the Savage 2000 chipset S3 aims at fill-rate. The Quad-texel engine is able to spit out 700 MPixels per second.
I was shown an early card (Alpha Silicon) running at undisclosed speed. Unreal Tournament ran really nice with S3TC support as well (great looking textures)... I was also shown Quake 3 1.08, but unfortunately no framerates. Everything looked really cool though....
Diamond's Savage 2000 card will be called "Viper 2"

Thanks JanKees voor de tip

Anand heeft een review in elkaar geschroefd van de net uitgebrachte Matrox Marvel G400TV. Hier heb je wat info over de hardware MJPEG encoder, lees de rest hier.

The major selling point for the Marvel G200 was the fact that it didn’t simply include a software MPEG2 encoder for video. Instead it used a hardware MJPEG encoder for true to source video capture. This Hardware MJPEG support is once again provided for on the Marvel G400 courtesy of the on-board Zoran ZR36060PQC.

For a card retailing at $300 there is no way to provide support for hardware MPEG2 video capture and editing with the current technology. Instead of offering a software MPEG2 video capture/editing solution Matrox decided to stick to their guns and continue, in their opinion to use the next best thing, hardware MJPEG. Hardware MJPEG is not as efficient as MPEG2 compression but a hardware MPEG2 codec on a sub $300 enthusiast's card is not a reality we can enjoy.

In terms of efficiency, MJPEG loses out in the amount of disk space needed to hold a single second of video at NTSC or PAL resolutions at 30 fps with no visible loss in quality or drop in frames. At 704 x 480 (NTSC) at ~30 fps (29.9 fps) -- otherwise known as D1 resolution (full NTSC or full PAL resolution) -- the data rate required by the MJPEG codec as approximately 3MB per second. To capture a 30 second commercial would require 90MB of disk space, and a 30 minute TV program would require a hefty 5.4GB of space. For a product directed at home users, there are some steep storage requirements to get the quality Matrox is boasting for the Marvel G400. Of course you don't have to opt for the highest image quality (which you probably won’t considering that 50GB hard drives aren't common place just yet) in which case the MJPEG codec isn't too bad but it is still quite a disk space hog.

Meer info over al de video-editing features staan hier.

Bij Overclocker Australia staat een heftig artikel online van een gast die zijn Celeron 366 watergekoeld op 643MHz heeft draaien (117FSB).

I ran the CPU at 2.2v because that was what it needed with the peltier by itself. As I've already said, it would boot 98 at 616MHz but would crash pretty soon thereafter. Also, due to my dodgy Maxtor hard drive, I had to turn off UDMA because it didn't like the overclocked PCI bus at 616MHz. So, I filled the bucket with water, let the pump run for a minute or so (to make sure the water-jacket was nice and cold) then cranked it up at 616MHz. It booted as expected, and cruised happily along for about 30 minutes of Unreal flyby+SETI before crashing, by which time the mobo temp sensor was reporting 37C. I put my hand into the bucket of water and it was quite warm.

Never satisfied with what I have, I put some fresh cold tapwater into the bucket and tried for 117MHz FSB, which is 643MHz core speed. The great thing about this speed is that the AOpen AX6BC-Pro has 2 options for your PCI speed at that FSB speed - one of which is 29.75MHz! This means the PCI is actually underclocked, which let me turn the UDMA back on for my hard disk. To my surprise it booted, loaded Unreal, ran fine for about 20 minutes until crashing with a bucket of very warm water as before.

Needless to say, I liked the taste of this speed. To prove it was really the heat and not the CPU just failing to cope with the speed, I got another bucket of fresh water and put 2 trays of ice-cubes into it. This ran at 643 for almost an hour, until the ice-cubes melted, the water heated way up and bang.. you get the picture. During that hour I unzipped files, ran Prime95, SETI@Home, the Unreal flyby, played some Sports Car GT, copied a 288mb file from one directory to another and did a File Compare on them.. all fine. No errors. No crashes. Woo!

119MHz FSB, or 670MHz core, was a no-go. Locked up at the Win98 login screen, even at 2.4v with UDMA turned off.

Nadeeltje is dus dat als zijn emmertje water warm wordt het zaakie crasht

Planet Riva heeft een review van de RealMagic Hollywood Plus MPEG2 decoder online gemikt. Hier heb je vast een gedeelte, lees de rest hier.

The decoder looks like a standard PCI card and is fairly small in size. On the decoder one will find the following : Stereo Out, S/P DIF connector, TV Out, VGA Out and VGA in. The Stereo Out is for connecting speakers etc for audio, S/P DIF is for digital audio, TV Out is for watching DVDs on your TV, VGA Out is for connecting the decoder to your monitor and VGA Out is for the pass-through cable from your video card to the decoder. Along with the decoder I received a cable that let's you connect your sound card's LINE-IN port to the decoders Stereo Out port (I personally don't use it), a pass-through cable for you to connect the decoder to your video card, an S-Video to Composite cable, a demonstration CD with clips from the Star Trek : First Contact movie, a very well written manual and finally two disks containing the drivers and playback software.

Het heeft niet zoveel met hardware of computers te maken, maar als studenten gaan protesteren vinnik het altijd wel interessant worden . Via Slashdot kwam ik bij dit artikel van WorldNetDaily, over een groep van 350 studenten uit Ruston, Louisiana, die in protest zijn gegaan tegen het dragen van ID badges waarin hun social security number wordt gebrandmerkt. Ten eerste vinden ze het niet prettig om als vee behandeld te worden, ten tweede blijken de nummers erg eenvoudig met een barcode scanner uit gelezen te kunnen worden:

The badges are worn on a lanyard with the Pepsi logo on it. The badge has a photo of the student, the school name, the student's name, and a barcode which represents the Social Security number. Although administrators claim the number is protected from unauthorized use through encryption in the barcode, the students know that is not true.

To prove how easy it is to read the barcodes, Jonathan Washington, 16, reads any barcode in about 15 seconds. He tells other students their Social Security number and then asks them to sign his petition to have the cards changed.

His methods are convincing. Over 350 students have signed in just the first few days. Rachel Winchel, 16, another Ruston student, is helping to spread the protest and circulate the petition. [break] Dit is nou een koele actie : [/break] Since the faculty and administration also wear the cards, students are busy taking down their Social Security numbers with plans to publish the results on the Internet if the cards are not discontinued.

"Actually, it's not difficult to look at it and know what it means," explained Winchel. "There's narrow bars and there's wide bars. Each number zero through nine has a code. By memorizing the code for the number zero through nine, you can just glance at someone's card and the numbers just pop out.

[...] Washington's complaints to Dr. Scriber [directeur van die school] produced a sudden change in policy in the past few days. The librarian was told to remove Washington's Social Security number from the library computer which reads the barcodes. A new card has not been issued, and the cafeteria computer has not been changed.

"He thought that would get me to leave him alone about it," said Washington, who has vowed to continue his fight. He will not be satisfied until the cards are completely changed for all students. If the change is not made he expects to take the school to court for civil damages.

Check hier het volledige artikel.

Sharky Extreme heeft een vergelijking van de Saitek Cyborg 2000 en Thrustmaster Top Gun USB pretpookjes neer gemikt. Van deze twee ziet de Cyborg er absoluut het coolst uit, toevallig wordt deze stok ook als beste bevonden:

Overall the excessive stiffness of the Top Gun USB forces us to shoot that Maverick down. Any flight stick that takes enough effort to cause soreness is a pain in the butt. We don't want it and we can't recommend it for anyone but a WWF pro-wrestler. As for the Cyborg 2000, we find it very compelling, but the hat switch problem stops us from recommending it whole-heartedly. If you don't care about a hat switch go ahead and buy the Cyborg, it should make you very happy. It is better than many joysticks in the $80 range. But if you need a hat switch, we can't recommend the Top Gun USB or the Cyborg 2000, so you're on your own.

Het gebruik van de Apple AirPort (waarmee je files kunt overbeamen naar je iBook) zal niet toegestaan zijn in Frankrijk omdat de 2,4GHz band waarop het apparaatje werkt is voorbehouden aan het franse leger, aldus Liberation. Wil je dat ding toch gebruiken, dan zul je er specifiek toestemming voor moeten vragen. Die idiote fransen weer... Hier een stukje uit de (best wel grappige ) engelse vertaling van Babbelvis:

An apparatus using the frequency of 2,4 Ghz can scramble an apparatus of the army or vice versa. " Contrary to other countries of Europe which do not have this constraint, in France this frequency band is a soldier. APPLE probably did not think there. But we will not put all our apparatuses at breakage because the iBook arrives ", launched the admiral Bizard.

" the problem was not taken enough in time ", it added. For the moment, the admiral should treat only 10 requests per week but it is likely to be found submerged if the iBook is a success.

On his side, the person in charge marketing produced of APPLE France, Hughes Asseman, remains serene, by recalling that the range of the iBook does not exceed 50 meters and should not pose problem of authorization.

" to receive the Internet without wire, the iBook communicates by radio waves with a terminal, Airport, connected on the telephone wire or a cable, in a radius of 50 meters. Unless being in a joint part with a military HQ, that cannot pose problem ", commented on Hughes Asseman. " We have zero concern on marketing ".

" We will have to examine the design features of the apparatus, for example to check that the range does not exceed 50 meters ", concluded from his side the head of the military national office of the frequencies.

The requests for authorization will have in any case to be sent individually by each holder of iBook using the Airport terminal, delivered in option and which must be launched about mid-October.

Techweb heeft een artikeltje gepost met meer info over de delay van de Intel Camino chipset. De release van de i820, oorspronkelijk gepland voor maandag, zou tot 3 maanden vertraagd kunnen worden:

Analyst Peter Glaskowsky of MicroDesign Resources, Sebastopol, Calif., said the error could push out Camino's launch anywhere from one to three months. In the meantime, OEMs will virtually be forced to purchase Intel's 440BX chip set, or select Via's Apollo Pro Plus 133.

Industry sources said earlier this week that they had discovered a close proximity of Direct Rambus inline memory modules (RIMMs) created too much heat in certain areas. Existing signal-integrity problems apparently were aggravated by the additional heat, analysts said.

[...] The quickest way to solve the timing problem with three RIMM slots is to shorten the memory bus line, industry sources said; and they now say they believe Intel is doing exactly that. But if Intel cannot design a work-around as expected, then existing boards will have to be scrapped pending a redesign.

Planet Riva heeft een review gepost van de nieuwe GA-660 Plus TNT2 kaart van Gigabyte. Deze .22 micron TNT2 draait standaard op 183/183MHz, maar echt veel hoger kan ie niet, hooguit 190MHz. Begin oktober komt deze kaart op de markt.

The GA-660 Plus is a nice looking card. It is laid out on cool looking blue PCB. There are four 6ns SDRAM modules, each 8MB. There is an active cooling fan and heat-sink on the top side of the card, which is a sign that this is no "plain vanilla' offering.

The capacitors are not "oil drum sized" clunkers, the card has a neat and clean layout. The TNT2 chip hidden beneath the fan is .22 micron fab, which is more than likely the last design change for the TNT2 chip.

Flipping the card over reveals a profile which is as smooth as a baby's butt, plus a rear heat-sink. That's right RIVA fanatics, the GA-660 Plus comes with a bright blue, over-sized heat-sink on the back of the board! And it's not just for looks either, it really does a great job of pulling additional heat from the bottom of the TNT2 chip.

I went out to my local computer parts store and spent three dollars on a 466 CPU fan which I attached to the rear heat-sink. Attaching a rear fan on this card is a good idea, because this heat-sink does generate a lot of heat. Of course we all know that a hot heat-sink is a good thing, because it is sign that the heat-sink is pulling heat off the video chip itself.

Attaching a fan on this heat-sink insures that you will have a much cooler card than most of the other Ultra TNT2 cards which come with only one heat-sink and fan. The fact that it uses .22 micron fab for the chip, as opposed to .25 micron fab should also serve to keep it a little cooler than some of the .25 micron TNT2 Ultra versions which are on the market. All of the bench tests were performed without the additional fan attached, so as to give a true to life product review.

The GA-660 Plus is put together well, it implements the spring tension method of attaching the main heat-sink, which is much better than the "glob of glue" method.

Lang leve de onzekerheid! Is het nog zo dat we in een van de andere postings (GeForce 256 preview melden dat de GeForce niet gedelayed zal worden door de aardbeving in Taiwan (bron was GamePC), is het zo dat Daiki het volgende meld:

"...maar ze weten wel te melden dat de tragische aardbeving in Taiwan geen negatief effect zal hebben.."

Some sourse said that GeForce256 would be delay because of taiwanese huge earthquake.

This sourse said that current semicon including GeForce256 chip were all spoiled, so GeForce256 chip shippnent will be early November in best case.

Ik heb eigenlijk wat meer vertrouwen in Daiki. Hij heeft nogal vaak gelijk.

Bij Sharky Extreme is een review te vinden van de Elitegroup P6BXT -A+ moederbord. Opmerkelijk aan deze plank is dat hij zowel een Socket370 als een Slot1 aan boord heeft. Hier heb je een gedeelte over de overklokbaarheid:

Overclocking the P6BXT was fine thanks to its BIOS controllable CPU settings, but we missed the flexibility that more front side bus speeds would offer for users seeking to push the speed of their CPU to higher levels.

Celeron overclockers particularly would be much better off with boards like the MSI-6163Pro, or the Soyo SY-6BA+ III, both of which offer settings in-between the long jump from 83 to 100MHz.

Strangely the P6BXT does offer manual voltage control for the CPU, which can assist overclockers in achieving that next step higher as long as the front side bus speed they've selected is able to be handled by their CPU/SDRAM.

The P6BXT performed well under duress, it's a very stable board whether overclocked or not.

Koen van Hardwareinfo.net heeft een zeer uitgebreide review in elkaar gezet rondom de Soyo SY-6VBA133 moederplank met VIA Apollo Pro 133 chipset. Dit plankje heeft naast 5 PCI-sloten, 2 ISA-sloten en 4 DIMM-sloten ook gelijk een ATA/66 controller en ondersteuning voor 133MHz geheugen. Voor meer info, lees de review hier.

Een interessant gegeven is dat bij het 6VBA moederbord de geheugen kloksnelheid niet noodzakelijk gelijk moet zijn aan de FSB kloksnelheid. Het moederbord geeft de mogelijkheid om de FSB-instelling indien gewenst te verhogen of te verlagen met de PCI kloksnelheid (normaal 33 MHz). Dit houdt bijvoorbeeld in dat bij een 100 MHz FSB instelling het geheugen niet alleen op 100 MHz kan functioneren, maar ook op 66 MHz (100 - 33) en 133 MHz (100 + 33). Hoe dit systeem precies werkt is voor mij een raadsel: ieder die me de werking van dit systeem kan uitleggen moedig ik zeker aan dit te doen.

Het nut van deze extra functionaliteit is twee ledig. Ten eerste is het zo bij upgrades mogelijk om het oude geheugen te blijven gebruiken. Als je bijvoorbeeld van een 66 MHz FSB Pentium II overstapt naar een 100 MHz FSB Pentium III, dan kan het oude PC66 nog gebruikt worden. Zo ook als een 100 MHz FSB Pentium II of III wordt ingeruild voor een 133 MHz FSB Pentium III: in ruil voor een beetje minder performance maakt het moederbord het ook hier mogelijk om PC100 geheugen te blijven gebruiken. Het twee nut is juist andersom: door bij een 100 MHz FSB CPU het geheugen op 133 MHz te laten werken (met PC133 geheugen) krijg je volgens SOYO een redelijke snelheidswinst.

Aangezien ik helaas geen PC133 geheugen modules beschikbaar had voor deze test heb ik dit niet kunnen uitproberen. Persoonlijk denk ik dat je niet veel zult kunnen merken van een dergelijke snelheidwinst: wat heb je namelijk aan 133 MHz geheugen, als de CPU de data maar op 100 MHz kan in- en uitlezen? Uiteraard zijn deze extra instellingen allemaal via de BIOS te configureren.

FPS3D.com heeft een review online gezet van de Maxtor DiamondMax 6800 harddisk. Deze schijf draait slechts op 5400rpm maar heeft wel een opslag capaciteit van maar liefst 27.2GB

Maxtor has put together a fairly nice hard drive here. The only thing holding it back from pure excellence is the low RPM and high seek time. Other than that, this drive rocks. You get a very wide range of capacity choices so you aren't stuck with too much or too little. If you are looking for a good size, average hard drive, the DiamondMax should be your #1 pick. But, if you are looking for something with more speed and more features, go with the Quantum Fireball Plus KA. Maxtor has put out a good product, though, and I suggest you consider it when purchasing your next hard drive.

Bij Silicon Investor hebben ze een e-mail van 3dfx ontvangen waarin ze wat vragen hebben beantwoord over de delay van de Voodoo4 / Napalm naar Q1 2000. Lees het hele verhaaltje hier.

[Q] At the press conference it was said that they were adding some features. That would delay it into the fourth quarter. We know that the next card will have 32/24 bit 3D, compression, T-buffer, glide support, a lot of memory. What feature(s) was added that caused the next card to be late.

[A] Well, strictly speaking, there was no single feature that made this product late. As I have explained in other forums, the delay was principally in the "physical back-end", literally the last stage before the product goes to the fab and after the design phase has taken place. The delays were not due to feature additions per se but more to the overall complexity and density of the chip.

BTW, we have never before experienced this kind of delay in the backend and believe that we will never again experience such a delay. It really appears to have been a very unfortunate anomaly.

[Q] Due to the low stock price, how is the staffing of engineers for the Current projects; is there much turnover?

[A] We have been very successful in recruiting recently, having substantially increased the number of engineers working on future products. We have had some turnover, but not an unexpected amount. Our engineers in particular are very driven. They have been more negatively affected by the disappointment of the Napalm slip than by the stock price. They don't like losing the lead at any point. They are committed to getting it back.

On the other hand, the engineers I have spoken to this week were noticeably more upbeat now that we know more about GeForce. The part is much less impressive than we had feared, and they are now quite confident about our leadership with Napalm and other chips planned in the near to mid term.

Voor de rest, klik hier.

Bij GamePC hebben ze een preview in elkaar gedraaid over de nVidia GeForce 256. Helaas geen nieuwe info, maar ze weten wel te melden dat de tragische aardbeving in Taiwan geen negatief effect zal hebben op de GeForce 256 productie. Hier heb je een hap uit de preview.

I was taken into the NVIDIA demo room and shown their current demo programs. What I saw was truly impressive. On screen models with well over 100 thousand polygons (most games use models in the 3000 polygon range) running smoothly with several realistic lighting effects like environment map cube mapping. This is similar to bump mapping, but allows for per pixel lighting and looks extremely realistic. The demos were as (if not more) impressive than those I've seen of the Playstation2. Rippling water, facial animation and high-poly models are all possible with this technology and all done with true polygons. Currently, graphics processor use lighting, texturing and shading techniques to achieve similar effects, but the difference in night and day.

If you buy a GeForce based product this season, there will be titles available to take advantage of the hardware T&L. Firstly, OpenGL-based games fully support HW T&L. So Quake 3 fans should see performance increases, plus the possibility of some special maps designed with a higher polygon count. Activision's BattleZone II, Acclaim's Shadowman and Halo by Bungie were all mentioned as titles that will be supporting hardware T&L in their graphics pipelines. The upcoming title, The Whole Experience, looks to be an Unreal-killer in terms of graphics and fully supports HW T&L. A quick demo included full trees (not just the sprite or cross-style ones) and a very lush enviroment that really has a sense of being far more real than anything else we've seen on a computer desktop so far.

Overall, hardware T&L is definitely the way computer games will be heading (after gamers see some of the demos, we will expect no less) and the GeForce256 will be leading the way. I have been told that developers are whole heartedly support the new techniques and we will be seeing more and more titles hit the shelves ready for DX7 and HW T&L in the very near future.

AGN Hardware heeft een kudde webmasters van nVidia fan sites bij elkaar getrommeld en gevraagd wat zij van de GeForce 256 vinden. Uiteraard ontbreekt ook de onvermijdelijke Fill-rate vs T&L vraag niet:

Chris Gordon-Planet Riva: As if I didn't see this question coming =) Having seen the various demos in person at NVIDIA, I have to say I'm completely convinced that T & L is the next big step when it comes to enhancing realism in games. The detail level that you can have in models/levels using the GeForce is just scary. Take the infamous tree demo, for example (http://www.nvidia.com/products.nsf/htmlmedia/treemark.html ) and you'll see quickly that fillrate is not going to be as much the limiting factor in new games, but triangle rate. Heck, I've already heard a couple complaints from my readers that their brand-new PCs and high-end graphics boards don't do much for this benchmark... and they're right. Current systems which rely on the CPU to do the triangle transformation calculations really aren't up to the task when scene complexity is increased. Moving the system CPU out of the picture, as GeForce does, is a big departure from the current way game graphics are rendered.

Meer reakties vind je in dit artikel van AGN Hardware.

3DWars heeft een artikeltje over internet browser wars neer gemikt. Voor de verandering worden hier niet IE en Netscape tegen elkaar afgezet, maar brengt-ie buitenbeentje Opera in de familie. Zelf heb ik Opera een tijdje gebruikt voor nieuws posten en uiteindelijk kwam ik tot de conclusie dat het een instabiele bout browser is, die alleen handig is wanneer je achter een super trage bak met weinig RAM zit. Laat ik ook maar meteen richting zijn conclusie skippen...

So will I change my browser? No, sorry Opera, I like you, but only as a friend. The war may have passed Opera by in the sense that consumers are already used to a certain interface. I found it difficult to use the address bar on the bottom of the screen. I am so used to it being on the top, I was actually frustrated by that (Quick note: the stop button is really small and to the left of the address bar-hard to find. Perhaps Opera may be trying too hard to make itself different, by sacrificing small cosmetic features that IE and Netscape learned from consumers, in order to establish itself as something different). It may be too late for Opera to win consumers who are already satisfied. But never underestimate the power of the consumer to like something that's better. If Opera can outclass the two Goliath's, it will win. But as is, Opera doesn't have a product that will enable it to get more than a foot in the door. It's a start, and I like to see the new guy, because they always bring a fresh perspective to the table. So if Opera comes out with Opera 5.0 for free, you can bet it will make more than a crack in the enemy's armor. But don't expect the competition to sit back and let its years of huge investments go down the drain. If anything, the great browser wars aren't over and the product we get will only improve as the battle turns red-hot.

De New York Times heeft een artikeltje gepost over Billies nieuwe internet strategie. Microsoft gaat MSN nieuw leven in blazen en Bill blijkt erg goed naar meneer McNealy van Sun geluisterd te hebben, want Microsoft gaat network computers bouwen...

The Microsoft Corporation disclosed several pieces of its new Internet strategy Thursday, saying that it would offer the public a piece of its most successful online effort, the Expedia travel site; would significantly redesign its MSN portal, and would introduce a network computing device.

Anticipating a world in which software will be sold as a service instead of a packaged good, the company said it would re-introduce MSN, shifting its focus to online communications, improved information searching and electronic commerce.

The company also demonstrated its new low-priced network computer, the MSN Web Companion, which will deliver the MSN service beginning in mid-2000. The company said the device was part of an evolving business strategy -- going beyond its stranglehold on the PC industry to develop new ways to connect consumers to the Internet.

The company called the new approach its "everyday Web" strategy and, in a complete turnabout, said it no longer had a dogmatic commitment to the PC platform.

"The vision of the company is about software as a service," said Richard Belluzzo, group vice president for consumer and commerce business. "This fundamentally changes business models."

As part of its new approach to the Internet, Microsoft said it planned to sell part of Expedia in an initial public stock offering. Expedia will become a separate company but will be majority owned by Microsoft.

The Adrenaline Vault bericht dat er sinds de release van de Sega Dreamcast in de VS al meer dan 500.000 kastjes zijn afgezet, ver boven Sega's verwachtingen:

Thursday, September 23, 1999 - 11:50 CDT Dreamcast Reach 500,000 Units Sold Sega of America's Dreamcast videogame system has topped 500,000 units sold. The advanced videogame console has sold through a total of 514,000 units since it launched on Sept. 9. This is in contrast to the PlayStation reaching the same number of sales in four months in 1995 and Nintendo 64 reached the same number after two months at retail when it launched in 1996, according to the NPD Group of Port Washington, NY. Today's announced sales figures for Dreamcast well surpass the company's previous projections of hitting 400,000 in the system's first month of availability at retail.

MacOS Rumors heeft een interessante posting over de AMD / Motorola deal (kwam er kort op neer dat AMD produktiecapaciteit beschikbaar zal stellen in z'n 0,18micron koper-interconnect fab in Dresden):

New information from Motorola sources suggests that not only will Motorola be taking advantage of AMD's excess production capacity to build more and faster G4s, but that it may actually be able to reduce its costs by outsourcing production to AMD over time.

Motorola will also reportedly be licensing several technologies from AMD, including its on-chip, high-speed Level 2 cache design, high-speed memory controller systems, and potentially some of its floating-point processing technologies.

The biggest news of all has to be the persistent rumor at Motorola that management is looking for a new arm to the AIM (Apple, IBM, Motorola) PowerPC alliance....and that AMD could be it. What possible purpose could a new partner serve that couldn't be culled from the existing members? Could it have something to do with a process technology prototyped by AMD which could infringe on AIM patents, but in alliance with AIM could be used on both the Athlon and PowerPC? We may know soon enough....

Een alliantie tussen IBM, Motorola, Apple en AMD? Dat zou kewl kunnen worden .

Ik kwam bij JC wat interessante geruchten tegen over een K6-2+ met 128Kb L2 cache:

Freaky day, eh? First off, the famous old rumour of the "K6-2+" with 128K on-die L2 has resurfaced, as suggested on c't (English trans). The basic idea is probably to try to convince people that the 128K K6 is meant to compete with the 128K P6 (Celeron), and the 256K K6 is to compete with the 256K P6 (PIII Coppermine). Interesting gambit, sort of an expansion on that segmentation strategy so that people may associate PIII with K6-3, leaving the Athlon as being considered a step up from both. Of course, this could just be an idle rumour which would make any of this speculation dead. [break] Ook heeft-ie info over de gevolgen van de i820 delay in de aandelenmarkt. Je zult begrijpen dat die blunders niet echt vertrouwen wekken bij aandeelhouders ... [/break] Still a crappy stock market day for some. Intel got hit a bit by the above news, and on top of that by BancBoston Robertson Stephens (an analyst), who mention that demand for systems from Intel has gone down by as much as two million units recently. ASPs appear, according to Dan Niles, to have dropped from an anticipated $206 to a surprising $198 or perhaps lower. Another analyst, Gerard Klauer Mattison &smp; Co., mentions that many companies that have been trying to emulate Dell's "low inventory" strategy may alter their plans as the disaster in Taiwan has meant that such companies would not have sufficient inventory in a crisis where they're not getting a constant uninterruptible stream of parts. [break] Enneuh, het zal je niet meer verbazen, maar er blijkt ook nog een bugje in de PIII Xeon met 512Kb en 1Mb L2 cache te zitten... [/break] Uh, oh -- just before I sent this, I found some more Intel-related news. Not too great at that. My source is SI, the news article is here. The problem? A delay in two of Intel's 550MHz PIII Xeon models (the versions with 512K and 1MB L2 cache), due to an odd glitch, are unable to work on the new 8-way motherboards. The 2MB version (which, of course, offers the greatest profit margin) appears to not experience this problem. Hmmm, oddly I find myself somewhat drooling over the prospect of 8-way x86 processors with 2MB L2 cache.