The Register heeft weer eens een paar mistige geruchten van de straat geplukt: nu denken ze dat 3dfx van plan is om S3 over te nemen. Dat zou voor beide bedrijven een errug relaxte marktpositie kunnen opleveren:

Register sources reckon 3dfx's subtle plan to take control of the 3D graphics industry and more is still a runner. The plan, which emerged back in June, has 3dfx launching a bid to buy rival 3D graphics specialist S3.

[...] That seemed to be that, as far as 3dfx's plan went, but according to our source, talks between the two companies have continued despite -- or, more likely, because of -- S3's Diamond acquisition.

According to our source, the two companies have agreed in principle to merge -- the haggling centres on the financial details and the make up of the merged operation's management structure.

The financial aspect is probably the most important sticking point. 3dfx's results haven't been two hot of late, whereas S3 is clearly moving on and up. That suggests that if the plan goes ahead, it will have to be a true combination of interests rather than a takeover by any other name.

Successful negotiations would lead to a business that combined two key 3D graphics players with Diamond's broader consumer product portfolio generating a powerful force in the PC add-in market. Both S3 and 3dfx are desperate to wrest leadership of the graphics market from ATI -- and together they might well be able to do so.

S3 brings some powerful technologies to the table -- it's texture compression facility, for one -- which 3dfx brings the best-known graphics brandname: Voodoo.

Intel gaat er nu toch redelijk hard tegenaan: op het Intel Developers Forum in Palm Springs heeft het een 800MHz Coppermine modelletje gedemonstreerd:

The 800MHz chip, likely a Pentium III based on Intel's Coppermine design, was based on Intel's 0.18 micron manufacturing process, he said. Coppermine is the code-name for a Pentium III chip build on the 0.18 micron process.

Barrett said the Pentium III Coppermine chip will debut in October at speeds as high as 700MHz. A 667MHz version is also expected, Gelsinger said.

Gelsinger also said Rambus Direct RAM, a new high-bandwidth memory technology that will work with the 820 chip set, is on schedule for delivery this quarter. He said 14 products from five suppliers had already been validated.

Check ZDNet voor meer info.

The Register heeft een nieuwsposting over de release van de PowerMac G4. In dit artikel claimt ome Steve dat de G4 bijna drie keer zo snel is als een PIII-600:

Jobs claimed the machine is 2.94 times faster than a 600MHz Pentium III according to data found on an Intel Web site, but Apple's product spec. more modestly credits it with being 100-200 per cent faster than the fastest PIIIs in CPU and Adobe Photoshop tests. [break] Verder nog wat specs van de G4 doos: [/break] The Power Mac G4 has 1MB of backside L2 cache running at half the speed of the processor, with a 100MHz system bus that allows a 800MBps throughput.

The built-in network is 10/100Base-T Ethernet. All three G4s have 128-bit internal memory data paths. Sensibly, the hard disks are either 10GB, 20GB or 27GB, and up to a total of 100GB is possible. The optical drives may be a 32x CD-ROM, DVD-ROM or DVD-RAM (both with video playback). There is a 100MB Zip drive, and support for two internal Ultra ATA-66 drives or three external SCSI drives (with an optional SCSI card). The graphics support consists of the ATI Rage 128 card with 16MB of SDRAM. There is support for 1600 x 1200 pixels at 32 bits per pixel and an 85Hz refresh rate.

Wazige Japanse site heeft een aantal foto's van de nieuwe PowerMac G4 online gezet:

Nog een blurp van Betanews:

Some rumors have been floating whether Windows 2000 will drop support for dual Celeron processors, so here's the official word. Windows 2000 will not officially support SMP Celeron machines and no technical support will be given, however using such hardware will not be blocked. Current builds of Windows 2000 read the dual Celerons as Pentium IIs and run them perfectly. If you do not mind losing technical support for such hardware, you should not have a problem once Windows 2000 reaches mainstream use. It is possible this will change before RTM in November, se keep checking back to stay updated.

Zou natuurlijk wel een extreem smerige klotestreek van Bill zijn als-ie Celeron SMP support uit Win2k ript.

Hier hebben we al eerder geruchten over gehoord bij The Register. Betanews legt het nogmaals uit...

After a piece of software hits release to manufacturing (RTM), a lot has to be done before shipments are made to retail outlets. CDs must be pressed, manuals printed, packaging prepared, etc. There is always dead time after a build goes Gold and a hefty one is expected with Windows 2000, due to the importance of this new OS.

First moved from October 6th to the 28th, Windows 2000 will now RTM on November 15th - the first day of COMDEX.

However, Windows 2000 will not hit store shelves until late January at the earliest, giving Microsoft time to deal with Y2K problems in the latter part of the month. It is possible we will see a huge Super Bowl Sunday shebang if Microsoft can get everything in gear by that point. Although, an early February release is a possibility as well.

Microsoft now has an excuse give an important keynote speech about the milestone and the COMDEX hoopla is bound increase ten-fold. Maybe these delays haven't turned out so bad for the software giant after all

Ik lees bij Slashdot.org dat er een nieuwe versie van XFree86 is gereleased. Ik ben niet zo into Linux, maar ik geloof dat dit wel redelijk belangrijk nieuws is :

- Support for S3 Savage4 and Savage3D. Limited to Linux/x86 at this point.
- Support for S3 Trio3D/2X.
- Support for DGux.
- Support for QNX.
- Fix bug in Mach64 server on Rage LT and Rage LT Pro.
- Fix SiS driver for 530 and 620.
- Fix the spurious underline problem on NVidia Riva TNT cards.
- Fix the PS/2 mouse problem with later Linux kernels.
- Misc updates and bugfixes in Rendition driver.
- Updates from SuSE and Red Hat, including more keyboards, PAM support, ARM and AXP fixes, security fixes.

De release notes vind je hier, download locaties lijken me wel bekend bij de Linux gebruikers...

Creative Labs heeft een nieuwe serie DVD-RAM en CD-RW drives aangekondigd. GA-Source heeft de press release:

Creative Labs today announced the latest additions to its family of multimedia storage solutions. The new Blaster CD-RW 6424 is an easy-to-use solution for storage and distribution of content with faster recording speeds. The new PC-DVD RAM 5.2GB provides a cost-effective solution for storing and archiving massive amounts of data.

The follow-up to the Blaster CD-RW 4224, the Blaster CD-RW 6424 now writes CD-Rs at 6x and CD-RWs at 4x, and reads CD-ROMs at up to 24X (a maximum read of 3,600KB/sec). It stores up to 650MB of data or over 74 minutes of CD audio on a single CD-R or CD-RW disc, offering customers a solution for archiving files, storing large databases or compiling MP3 files. By using inexpensive CD-R and CD-RW discs that are easily read by standard CD-ROM drives, the Blaster CD-RW delivers a low-cost alternative to sharing files at a fraction of the cost of Zip disks. Blaster CD-RW includes Ahead Software's Nero CD burning software and Prassi Software's abCD packet writing software, plus one blank CD-R.

Based on the DVD Forum's re-writeable DVD format, Creative PC-DVD RAM offers users 5.2GB of storage capacity on one double-sided DVD-RAM cartridge. In addition, PC-DVD RAM can also read DVD-ROM, DVD-Video, CD Audio, CD-ROM, CD-R and CD-RW and Video CD discs. With its large storage capacity, PC-DVD RAM provides a solution for backing up a complete hard drive or archiving enormous multimedia files. PC-DVD RAM connects to a PC via a SCSI interface ensuring reliable transfer of data. One DVD-RAM cartridge is included, plus Creative offers a 5-pack separately for US$199.99.

The Blaster CD-RW 6424 and PC-DVD RAM will be available at the end of July through Creative's network of distributors and resellers at an estimated street price of US$249.99 and US$599.99 respectively.

GA-Source heeft een drietal screenshots van de Dagoth Moor Zoological Gardens demo online gezet. Deze demo is door The Whole Experience (WXP) ontwikkeld om de supergeile features van de NVIDIA GeForce 256 te demonstreren:

Development of DMZG started right after E3. Our prototype hardware came soon thereafter, and we began work on the GeForce256 features. NVIDIA even sent several programmers up to our offices in Seattle to help out with optimizations and special GeForce256 features. It has been crazy to say the least: working with early hardware, early drivers, and early APIs is potentially the worst possible development scenario for any developer.

Surprisingly the process went quite smoothly, thanks in large part to the dedication and expertise of both NVIDIA and Microsoft, as well as the whole WXP team. I would like to give special thanks to Brian Harvey, Chris Donahue, and Mike Hara at NVIDIA for supporting this project, and having the vision to bring it to GeForce256 as a showcase of their amazing hardware.

De 1:1 versies van deze tumbnails vind je bij GA-Source. Voor meer info over de WXP demo kun je terecht op www.wxp3d.com.

Iomega heeft een nieuwe Zip drive gereleased. Het ding heeft een capaciteit van 250Meg en wordt middels de USB poort aansluiting. USB Workshop heeft de press release en voor zover ik het begrepen heb heeft dit model vergeleken met oudere Zip drives een compactere behuizing:

Iomega Corporation today announced a new Zip® 250MB Universal Serial Bus (USB) drive for both PC and Mac users.At one inch thick and weighing just under a pound, the new slim and light Zip® 250MB USB drive, which can be fit into a pocket eliminates the need for larger parallel port or SCSI conversion cables. The new Zip drive is compatable with the company's 100MB disks. The new Zip® 250MB USB drive is compatible with computers running Windows 95, Windows 98, Windows NT 4.0, or MacOs 8.1 or above, and provides hot ``plug and play'' (except when running Windows NT 4.0). The cable which allows the Zip® 250MB USB drive to be connected via a PCMCIA slot in notebooks running Windows 95 or Windows 98, is sold separately. Also with the PCMCIA connection, users do not need a power supply.

The Zip® 250MB USB drive ships today and is expected to be available by the end of Q3 1999 in U.S. and Japan. Availability in Europe is expected in Q4 1999. The estimated street price of the Zip® 250MB USB drive is at $179.95 (U.S.), with Zip® 250MB disks available at an estimated street price of $16.65 (U.S.) each for a package of six. The PCMCIA cable, which is sold separately, is expected to be available in Q4 1999 for about $39.95 (U.S.). The Zip® 250MB USB drive package includes one Zip® 250MB disk, a USB cable power supply, a USB guideline sheet and a USB ``Quick Start'' guide.

Ars Technica heeft een Windows NT RAM & page file optimization guide in elkaar gedraaid.

Unfortunately, most users never question their virtual memory, or "pagefile", settings. That's too bad because there are numerous ways to improve upon the default settings NT will establish on install, many of which can really affect the speed of your machine. It's not uncommon to see someone who, upon experiencing a good amount of disk thrashing, heads out to buy more RAM when RAM isn't the problem. You have to be able to analyze your needs intelligently. You shouldn't even consider buying RAM without first taking into account the state of your pagefile settings.

So, without further adieu, here's the Ars Guide to RAM and Pagefile mojo under Windows NT 4.0.

...en die info lees je hier.

Intel is begonnen met de levering van Merced test samples aan computer bouwers. De uiteindelijke release staat gepland voor midden 2000. MSNBC brengt het nieuws:

“This is the very first Merced coming out of the manufacturing line,” said Gadi Singer, vice president and general manager of Intel’s IA-64 processor division, as he conducted the first public demonstration of the chip at the Intel Developer Forum here.

The demonstration represents a key manufacturing step, or “first silicon” for the chip. Singer, who joined Intel president and chief executive Craig Barrett on the stage, demonstrated Merced running on an early version of Microsoft Corp.’s Windows 2000 and on a next-generation version of the free Linux operating system, both of which are currently in development.

[...] Barrett said that the company was on track for volume production for Merced in mid-2000. Company executives declined to say how many prototypes of the chip they are shipping to computer makers, or to identify which computer makers are receiving the samples. The companies are helping Intel with its testing and validation of the chip, which will take about nine to 12 months.

“It’s good news,” said Rick Doherty, director of Envisioneering, a consulting firm in Seaford, N.Y. “The updated timetable will ease a lot of developers’s concerns.”

Jaja, eindelijk zijn de langverwachte PowerDesk 5.21 drivers voor de Matrox G100, G200 en G400 kaarten uitgekomen. Voor Windows 9x dan. Er zijn een hoop bugfixes en snelheidsverbeteringen beloofd, en volges de Matrox Forums is er gemengd succes. De berichten varieren van 'langzamer' tot 'nix veranderd' tot 'meer dan 10% sneller'. Maargoed, hier kan je ze vinden, of direct neerladen door hier te klikken (4328 kB).

Tweak 3D heeft een interessante chatlog gepost. Er waren namelijk wat mensjes van nVidia op een IRC kanaal, en die hebben heel wat vragen beantwoord over de GeForce 256 GPU. Om de log door te spitten, klik hier.

Leviat4n: Will the NV10 (or GeForce) be CPU independent? That is, will I get the same kind of performance on a P2-400 as I would on a P3-550?
Chris Donahue: You will always get better performance with faster CPU's

Sang: What is the minimum CPU recommended for the GeForce256?
Chris Donahue: Minimum CPU = 266mhz

Dit is maar een klein stukje, lees hier verder!

Meer demo nieuws: vandaag is de demo gereleased van Rally Championship. Plaatjes:

De demo weegt wel 64MB (met cool filmpie), maar is het downloaden wel waard denk ik. Voor download locaties, fiets naar 3D Files.

In co-operation with 3DRacing.net and the French Rally Championship League, 3DFiles.com has posted the 64MB 'Joystick' playable demo of Rally Championship (2000).

"As rally enthusiasts eagerly await the launch, it has been revealed that Rally Championship will shock the market with a number of new features that have never been seen before:

Including multiple driving positions, fully operational 3D dashboard instruments, a superb selection of top rally cars, new crash, collision and damage routines and comprehensive multi-player options - adding up to the most exciting driving experience you can have without getting into a car and doing the real thing."

Ik lees bij Voodoo Extreme dan Mirage een nieuwe versie van de Mortyr demo uit heeft. Deze versie 1.06 bevat wat bugfixes, schijnbaar iets met een intro movie. Klik hier om te downloaden.

Computer Games Online heeft een artikel online gezet van de nieuwe Savage 2000 van S3, die net voor de GeForce 256 officieel was aangekondigd. Net als de GeForce heeft deze nieuwe Savage ook een hardware T&L engine:

Of course, most games still don't support hardware T&L, and even when they do, game speed will be limited by how fast the video card can texture those polygons and draw them. The Savage 2000 appears to shine in this area as well, with four textures in a single rendering pass. Current cards with multitexturing ability apply two textures at once, often each in their own texture pipeline. The Savage 2000 has two pipelines that can each apply two textures at once. The result is that, instead of drawing two pixels with two textures on each of them in two clock cycles of the chip, it can do it in one.

The clock rate should be nothing to scoff at, either. The Savage 2000 will come in two speed flavors. The Savage 2000 should use 166MHz memory and have a core chip clock speed of 143-150MHz. This would translate into a fill rate of about 600 million texels per second. Compare this with the Voodoo3 3500TV, which has a fill rate of 366 million texels per second. The Savage 2000+ will probably use 200MHz memory and have a core chip clock speed of 166-200MHz. This would produce a fill rate of 700-800 million texels per second. The higher the fill rate, the faster a 3D card can texture polygons and draw pixels. Both speeds will support up to 64MB of RAM.

Bij AnandTech is een review te vinden te vinden van de GA-6BXU Slot-1 i440BX mobo met onplank SCSI van Gigabyte. Deze high-end plank is erg stabiel, maar voor overklokkers niet geschikt:

A single jumper provides for bus speed selection, either 66 MHz, 100 MHz, or Auto. Auto, of course, uses the state of B21 to set the bus speed and the other settings force the specified FSB. A second jumper enables a 133 MHz FSB mode with a PCI divider of 1/4 so that the PCI bus remains within spec. As with all other i440BX solutions, AGP is still way out of spec when the 133 MHz FSB setting is in use. It would be nice to see some overclocked bus speeds here, but Gigabyte apparently does not think those are necessary for a "high end" onboard SCSI motherboard. Four DIP switches control the clock multiplier settings, which is of course not necessary to set on today's multiplier locked CPU's. Stability is excellent as we've come to expect from Gigabyte boards.

Gigabyte has also taken the standard Award BIOS and improved upon it just a bit, starting off with a feature well known on AOpen boards - a set of BIOS options that can be loaded and are already tweaked for performance. These options pretty much are optimal for the majority of users, but if you're a real hardcore tweaker, you'll want to go through and double check that everything set the way you like it. Upon boot, the BIOS reports the core voltage of each CPU and any voltages that may be out of spec. The onboard SCSI chip can be enabled or disabled from the BIOS as well.

Kijk eens naar dit schatje: zo klein en toch al 333MHz!

- 333 MHz Pentium II
- Micro-motherboard with PCI
- expansion, up to 256 MB
- DRAM, NeoMagic graphics,
-Phoenix BIOS, Super I/O, and more.
- Dimensions: 3 x 5 x 0.7"

AMD heeft besloten de busspeed van de nieuwe 500MHz K6-2's naar 133MHz te verhogen! Ace's Hardware heeft de info:

To make sure that the K6-x can run at higher speeds than 500 MHz and still perform competitive, AMD decided to increase the FSB speed and hence the speed of the L2-cache and memory to 133 MHz!

New chipsets from VIA, Ali, and SIS (SIS 540) will make this possible on NEW super 7 boards in Y2K. The K6-2 CPU's can run fine at 133 Mhz FSB right now, although there is no stable motherboard aviliable. AMD sure puts up a good fight for the low end with Celeron after they attacked the high end with the Athlon.

Nog een review van iXBT-Labs, deze keer hebben ze de Castlewood ORB drive aan een test onderworpen. Met een prijs van 200 dollar voor de drive en $30 voor de media (2,2Gb) is deze drive - zeker vergeleken met een Iomega Jaz - redelijk sympathiek geprijst. De performance van het interne IDE model is behoorlijk goed, ongeveer 1,5x zo traag als een Fireball CR harddisk:

Well, what can we say now? Frankly speaking, we were struck dumb with the results achieved. It's clear that the data density on Quantum is higher, and besides, it is a "real" hard disk drive, which simply implies that it should have better results. But we expected them to be at least 10 times better! And to our great surprise and delight the difference made only 3 times and only in High-End Disk WinMark, where a lot depends on the heads positioning speed, which should be lower by the drive with removable media. So, ORB made us a very pleasant surprise. The data transfer rate graph and namely its unusual shape is also worth mentioning: not a descending stepping graph but a V-shaped one. That is why ORB drive shows even better results than Quantum in the last sectors. Unfortunately, we can't explain why it happens so, but we will try to find out the truth, which appears quite interesting.

Onze russische vrienden van iXBT-Labs hebben een review van de Vantec P3D-5030 op hun site gemikt. Vergeleken met een ordinair PIII heatsinkje daalde de temperatuur van een PIII-600 dankzij deze Vantec cooler met 20C!

When we installed an ordinary noname cooler onto Intel Pentium III 600, the processor temperature quickly went up and having reached 71 degree the processor successfully crashed. Well, well, the ordinary coolers seem to be absolutely unfit for 600MHz CPU cooling.

When we installed Vantec P3D-5030 cooler onto Intel Pentium III 600, the processor temperature remained somewhere close to 52.2 degrees. All the tests were successfully finished.

Of course, this victory, and especially such complicated cooling system, inspired us to immediately overclock the processor. However, Intel Pentium III proved extremely badly overclockable: all the tests were carried out with the FSB frequency increased only to 103MHz and processor frequency - to 618MHz. The CPU temperature rose to 54 degrees. And with the external frequency set to 105MHz the processor crashed during the tests. We decided not to increase the core voltage remembering about the gift of Pentium III L2 cache to get burned in conditions like that, especially if we bear in mind that the CPU was working at 2.05V already.

Ik lees bij Betanews dat Be Inc. een tweede update voor BeOS Release 4.5 heeft gepost:

Be, Inc. has released a second system update to R4.5. It includes some fixes to NetPositive version 2.1.1, as well as a driver for the 3Com 3c905C network card, a driver for the EMU APS/10K1 audio processor chip, support for Adaptec's 7890 Ultra2 SCSI controller, and numerous bug fixes. This is a must-have update for all you BeOS users out there.

De update kun je hier downloaden.

Mercury Center heeft interview gepost met Linus Torvalds, de man achter Linux. In het interview zien we onze grote vriend Bill ook nog even voorbij fietsen:

It occurs to me that Linus might find it interesting to spend some time with Bill Gates. At first, he says he is not the least bit interested in meeting everybody's favorite nemesis. "There wouldn't be much of a connection point. I'm completely uninterested in the thing that he's the best in the world at. And he's not interested in the thing that maybe I'm the best in the world at. I couldn't give him advice in business and he couldn't give me advice in technology,'' he says. Then he adds, "Even if you're the best technology person at Microsoft, your goal isn't to make the best product possible.

"His whole mission in life is to see that anything that threatens Microsoft goes away,'' continues Linus. "He probably is the best businessman alive today. And he happened to be in the right place at the right time and he was technical enough to take advantage of it.''

But later, Linus admits that he feels sorry for the world's richest man. "To have such a huge ego and have everybody hate you must really be difficult for him,'' he says, "And he probably has trouble trusting anyone other than his oldest friends.'' Yes, Linus relents that he would actually like to meet with Gates, although he "wouldn't know how to start the conversation.''

En dit is 'm dus:

Bizznizzy.com heeft een comparison van de Alpha FH60 en Tennmax Mega 7 coolers in elkaar geknutseld. Wat betreft de temperatuur benchmarks is er nauwelijks sprake van enige concurrentie: de Alpha wint met grote voorsprong:

Well now. Where to start, what to say..... It's pretty clear, the Alpha is a higher performing cooler then the Mega7, in its current configuration. But, that doesn't mean you should go out and buy an Alpha instead of the Mega7. The Mega7 can be made to perform nearly as well as the Alpha, with minimal modifications. The TennMax also doesn't require hand tools to assemble. As a matter of fact, it comes pre-assembled, unlike the Alpha, which comes very disassembled, and in several bags. The Mega7's fan is also extremely quiet, and won't keep the neighbors awake at night. TennMax has also equipped the Mega7 with what is probably the most well thought out retaining device of any heatsink I've ever had the pleasure of using. TennMax also has some of the best customer support you'll ever find. Just check out Reseller Ratings. I've never seen a company answer e-mails as fast as TennMax does. I'm thinkin' those guys must live at their computers, cuz Its never taken much more then an hour to get a response outta 'em.

3D Gaming heeft een review gepost van de Sennheiser Surrounder, zonder enige twijfel de meest wazig vormgegeven speaker/koptelefoon op the entire aardkloot:

I found that the Surrounder's most ideal situation is DVD playback and gaming, since many users may find it inconvenient for everyday Windows use. One great benefit the Surrounder offers in DVD playback is the placement of the speakers. Usually when you buy a surround system, placement is quite an issue, since incorrect placement takes away from the whole 4 point idea. Sennheiser has already taken care of this problem and with perfect symmetry places the 4 speakers around your head. The same holds true for gaming, as any hardcore gamer knows that correctly placed sound offer quite an advantage in any game. You can't quite go wrong with symmetrical placement can you? Didn't think so.

Meer over dit 300 dollar kostende ding lees je in de review van 3DG.

AMD heeft bevestigd dat de nieuwe K6-2's van hetzelfde 0,25micron produktieproces gebruik maken als de Athlon. De K6-2 zal hiermee snelheden van minimaal 550MHz kunnen bereiken. Hier heb je de info, afkomstig van Ace's Hardware:

It has been confirmed! The new K6-2s were not build with the older CS44E process, but with the newer and better CS44E7. The Athlon is produced using the same 0.25um CS44E7 process! CS44E7 uses tungsten and Cu dopped Al for interconnects, and have a smaller effective gate length ( 0.16um) and oxide thickness. The result is faster switching transistors at lower power.

The new K6-2 (AFX) anmd 500+ should be able to reach at least 550 MHz and more on CS44E7 and should compete with Intel's lowend siblings throughout the year 2000.

The use of a newer process enables the K6-x to reach higher speeds at lower power dissipation. Sources inside AMD indicate that this new process is nothing but awesome for both the K6 series and the Athlon!

Kbench heeft een fotootje van een i820 moeberplank online gezet:

Tja, weinig bijzonders te zien...

Ace's Hardware heeft een artikeltje over webserver performance in elkaar geklust. Hier heb je een stukje over server parsed HTML:

Server-parsed HTML files consist of standard HTML markup code along with server-specific code that is interpreted by a 3rd-party handler program or the webserver itself before being transferred to the client. This is the next evolution from CGI, as it is much more efficient, and inline server-side extensions such as ASP, PHP3, or even generic server supported extensions have become incredibly popular. The intention of such extensions is to make website content more dynamic and modular, and thus, easier to maintain. Additionally, a server-parsed document promotes the thin client model, minimizing client-side load and bandwidth requirements. All of this comes at a price, however, as server-parsed documents must be interpreted by the server before being sent to the client, adding additional CPU overhead. [break] ...en database connectivity: [/break] Of course, the killer app on the Internet these days has become online databases and e-commerce. Web-enabled databases and application servers have seen the most growth lately and show the greatest promise for the future. From a performance standpoint, however, online databases (using Oracle, SQL Server, Sybase, etc.) and applications call for serious performance considerations to be made. For large sites, high volume HTTP traffic and database transactions can work against each other to either crash a server or slow it to a crawl in a very short time. In these instances, it is almost always preferable to use a dedicated back-end database server (also for security reasons) with HTTP servers acting as front-end processors.

Review Zone heeft een artikel in elkaar geparkeerd waarin ze de performance van de TNT2 Ultra, G400 Max en Voodoo3-3500 op een Athlon-650 en Pentium III-650 systeem vergelijken:

OK folks, hope you enjoyed that. It's an early look at what 650MHz possibly could mean, and to do that we had to overclock an engineering sample of the Pentium III to 650MHz to go on par with the Athlon 650. Based on that, as you've already seen, the Athlon does more than hold its own. In fact, if you're a serious gamer you'd do well to seriously think about getting your hands on an Athlon. The Pentium III performance is all right, but the point is simply that when considering the higher-end CPUs you can get an equally clocked Athlon for a lower price.

Meanwhile, about the graphic cards. The Voodoo3 3500 is an excellent card and scales very well indeed with higher-end CPUs, though in terms of Quake III performance the TNT2 Ultra card performs much better. With 650MHz CPUs, though, you'll need to push the TNT2 Ultra beyond 175/200MHZ to get the kind of performance that the Voodoo3 3500 does at its default settings.

Check Review Zone voor de benchmarks.

Apple heeft gisteren de nieuwe serie PowerMac G4's aangekondigd (dit ondanks geruchten bij The Register dat het ding gedelayed zou zijn tot 2000). De G4's draaien op snelheden van 400 tot 500MHz. De hip gekleurde turqoise behuizing heeft plaats moeten maken voor een deprimerend grauw blauw:

C|Net News.com heeft een artikel over de PPC G4:

As previously reported by CNET News.com, the Apple chief introduced the new G4 PowerPC chip, running at speeds of 400, 450, and 500 MHz. Designed to speed up tasks--like encoding digital media and running Internet applications like security software and Apple's own QuickTime streaming media software--the new G4 runs more than two times as fast as comparable Pentium III chips from Intel, Jobs said, adding that Apple used Intel's own benchmark tests. This claim will almost certainly be disputed by Intel and others. [break] De Apple propagande machine draait ook weer op volle toeren. Steve is bezig met de 'collectieve herprogrammering' van zijn volgelingen, die moeten allemaal massaal gaan zeggen dat de G3 een 'supercomputer on a chip' is: [/break]

Calling the new G4 a "supercomputer on a chip," Jobs said desktops running the 400-MHz version will be available starting today, featuring a new, faster system architecture to take advantage of the new chip. The 400-MHz G4 desktop computer will be priced at $1,599, he said. A 500-MHz system with DVD-RAM--the industry's first to ship with re-recordable DVD--will ship in October and is priced at $3,499.

"These tasks are bringing today's computers to their knees," Jobs said. "This is the kind of stuff we're already doing, but we're doing it too slowly." Apple said that its computer is so fast, it is classified as a supercomputer by the U.S. government and can't be sold to certain countries. Jobs played a new TV commercial based on this issue, as well as commercials for the previously announced iBook notebook.

The new G4s, Jobs quipped, "Will be available in all freedom-loving countries." [break] Allemaal bullshit natuurlijk (net zoals de [G3=2xPII performance] leugen, die Apple bij de release van de G3 de wereld in stuurde): [/break] Reaction to Apple's chip speed claims were swift from the Intel Developer Forum in Palm Springs, California. "There are lies, damn lies, and demos," said David Wu, an analyst at ABN AMRO. [break] Verder zegt Apple dat de release van MacOS 9 gepland staat voor midden-oktober: [/break] Just as he had at Macworld New York in July, Jobs touted Mac OS 9, as the new operating system will be called when it is released in mid-October. One of the new features is the "key chain" function of the software, which acts as a master password and sets in motion all other passwords for accessing corporate networks, applications, and the Internet. The new OS will be available for $99, he said.

Of particular interest to Seybold attendees was Jobs's demonstration of a new function of AppleScript in Mac OS 9, which allows users to automate remote production of documents over the Internet. "This gives new meaning to the term 'Net publishing,'" he said.

Betanews weet ons te vertellen dat Abit een nieuw BIOS voor de BE6 heeft gereleased, die problemen met Western Digital UltraATA/66 schijven zou moeten fixen:

;Received word that there is a new beta BIOS for the ABIT BE6 motherboard. A severe timing error has been found when you attach a Western Digital ATA66 CaviarAA to your BE6 motherboard. OEMs have found problems with this motherboard and the CaviarAA drives only. All other ATA66 drives seem to work fine. A new beta BIOS should be posted to the ABIT website within the next few days.

Check de BE6 FTP directory voor updates.

Matrox zal deze maand beginnen met de levering van z'n flat panel add-on, waarmee een OEM G400 in een hand omdraai van een digitaal flat panel poepgat kan worden voorzien:

The Matrox G400 Flat Panel add-on card features Digital Display Working Group’s (DDWG) Digital Visual Interface (DVI) specification. This allows the upgrade card to provide premium image quality, speed and color fidelity by bypassing the graphics card’s digital-to-analog converter and outputting data directly to the digital Flat Panel display. In addition, the Matrox G400 Flat Panel add-on card’s Silicon Image SiI 154 transmitter supports resolutions as high as 1280 x 1024. Priced at $59 ESP, this snap-on card connects easily to the Matrox Feature Connector Port on any Matrox G400 Series graphics card. This add-on does not require a special driver, or BIOS. Its unified drivers incorporate DVI support, meaning that once the DVI monitor is detected, digital output is automatic.

The Matrox G400 Flat Panel add-on card will be sold exclusively at Matrox’s online store as of September 1999.

Koen van Hardwareinfo.net heeft een review van de Terratec Soundsystem DMX online gezet. De DMX is gebaseerd op de Canyon3D chipset van ESS en ondersteund zowel A3D(1.0) als EAX:

De ondersteuning van zowel A3D en EAX is zeer uniek en brengt zo een mooie combinatie. Helaas is A3D alleen 1.0, zodat we het beroemde wavetracing systeem moeten missen. Ook EAX klinkt op de SoundSystem DMX niet zo mooi als op een Sound Blaster Live!. Toch blijft het feit staan dat de SoundSystem DMX een unieke combinatie brengt en het vrijwel de enige kaart is waarmee tegelijkertijd zowel positionele als ruimtelijke 3D audio effecten mogelijk zijn.

Check de review hier.

Koen heeft een artikeltje geschreven over de afgelopen Dealer Info Dag in Utrecht, waar hij met een aantal hardware distributeurs sprak:

Op de stand van Tekelec kwam ik tot mijn verbazing Patrick Brouckaert van AMD weer tegen. Met hem heb ik nog een tijde staan praten over de vorderingen van de AMD Athlon in Nederland. Volgens Patrick begint Nederland ondertussen gelukkig ook wel een beetje last te krijgen van de Athlon-hype. Dat het AMD Athlon artikel het tot nu toe meest gelezen artikel van Hardwareinfo.net is kon hij natuurlijk ook zeker waarderen. Hij stelde mij trouwens gerust dat niemand bang hoefde te zijn voor illegale opnieuw gemarkeerde AMD Athlon processoren. Volgens hem was het voor AMD niet moeilijk om te zien of processor echt is of door de Toms Hardware geïntroduceerde manier overgeklokt is. Verder zou bij het openmaken van de AMD Athlon CPU-cartridge deze helemaal beschadigd worden. Het namaken van de cartridges zou ook nog een extra moeilijkheidsgraad zijn voor vervalsers. Uiteraard raadde hij iedereen af om zelf de AMD Athlon processoren op de Toms Hardware manier over te klokken.

Verder vind je in het verslag nog info over nieuwe produkten van Diamond, Creative, Terratec, Elsa en Guillemot.

Andy van BXB heeft een review van de AOpen MX3W i810 moederplank op z'n site gezet. Zoals het een ware tweaker betaamd heeft Andy getracht deze plank over te klokken, wat met enige moeite mogelijk bleek te zijn:

Operation at 100Mhz and beyond is supported. For running 66Mhz Celerons at 100Mhz a jumper is required to correctly derive the AGP clock. This is rather fiddly - to run a Celeron 300a at 450 it is necessary to enter the BIOS, select a 4.5x ratio, and 100Mhz bus speed, then power down and set the jumper. Speeds at 83Mhz and below do not require the help of the jumper, although the memory schema of the i810 chipset means that at 83Mhz you will need to be running decent memory, capable of stable operation at 124Mhz.

Alex aka 'Sharky' heeft op de nieuwspage van Sharky Extreme een posting geplaatst met info over de trieste neergang van Hercules:

Sharky Extreme recently spoke with the former CEO of Hercules Graphics, Peggy Fong, who resigned from the company on August 16th. Ms. Fong confirmed that Hercules is not actively shipping any products out currently, and that the company's management team and employees have parted company and gone their separate ways.

Hercules' financial problems have been well documented over the last three years, and although their newer products had been selling well, it wasn't enough to eliminate the substantial debt load that had accumulated and would not have allowed both the company and its employees to prosper in the future.

Ms Fong expressed to us her heartfelt belief that Hercules' employees performed above and beyond the call of duty in 1999, particularly over the past two months in which they endured several weeks of not being paid due to the lack of funds, all the while attempting to ship out every last customer order possible.

As one of the earliest pioneers in the area of PC graphics, Hercules helped define the visual era of the 1980s. Along with producing some of the best selling video cards of all time, Hercules also generated fantastic levels of company income. Unfortunately the 1990s saw Hercules fall from the mountaintop almost as fast as they had ascended it, due to both increased competition within the video card market from Diamond and S3, as well as a fundamental shift in the way products are developed at the chipset level.

Intel heeft bekend gemaakt dat het al in oktober met een 700MHz (geen 133Mhz FSB?) versie van de Coppermine zal komen. News.com brengt het nieuws:

PALM SPRINGS, California--Intel will push to release the "Coppermine" Pentium III processor, an enhanced version of its high-end chip, in October and bring it out at a speed of at least 700 MHz, Intel chief executive Craig Barrett said here today.

The October surprise advances the Intel road map, because it means the new chip will arrive sooner and work faster than expected. The chip is an important part of the firm's plan to stave off the threat posed by AMD's Athlon processor, which performs better than Intel's current Pentium III chips, according to many tests.

Before now, most analysts predicted that Coppermine would come out at a speed of 600 MHz or 667 MHz. Originally due in September, Intel recently pushed it back to November, because it was having difficulty making large volumes of the chip at 600 MHz and faster.

Ik kwam bij Voodoo Extreme deze erg handige post tegen waarin het verschil tussen Megatexels en Megapixels (altijd verwarrend) wordt uitgelegd. Check it:

Megatexels is how many millions of texels can be fed down to the card's pipeline per second (peak). Megapixels is how many millions of pixels can be put out to the frame buffer (peak).

Now, for as a point of comparison the Voodoo1 was 50 megatexels and 50 megapixels. The Voodoo2 was about 90 megapixels and 180 megatexels. This meant that if multitexture was on, you could send down two textures combined at 90 megapixels. If you had separated it into two single-texture passes, it would go as two 90 megapixel passes, or 45 megapixels total. Now, the TNT (which supports multitexture) had (I *think*) about 125 megapixels and 125 megatexels. The reason for this is because in single-texture mode, it would fetch two texels and write two pixels at the same time (peak). In multitexture mode, it would fetch both texels from both textures and write one pixel as a result. So that's why the numbers are equal.

So when S3 claims that they have a 700 megatexel framerate, it's *feasible* that per texture it's 175 million texels, so a peak fill rate of 175 megapixels. Or it could be that their megapixel number is somewhere higher than 175 but lower than 700 and they would rather just announce the higher megatexel number. I really doubt that they're hitting 700 megapixels, as that would have been more impressive for them to say and they didn't say that. It's probably more on the order of 350 megapixels or so.

Either way, barely anything currently uses 4 texture units, though Unreal and Quake 3 could probably make use of them. It is nice that they have four on there for future stuff though.

In conclusion, the megatexel number and megapixel numbers are two very different numbers, and should be posted next to each other to gauge the performance of one card against another.

Laat ik de overgebleven GeForce nieuwsberichten maar ff snel samenvatten:

Leadtek GeForce 256 press release (3D GPU)
Creative Labs GeForce 256 press release (Business Wire)
Riva Extreme GeForce 256 preview
PC Rave GeForce 256 preview
CT News GeForce 256 preview.

Ik kwam bij Voodoo Extreme een aantal interessante postings tegen over developer support voor GeForce 256. Hier bijv. een stukje uit de Experience Developer's Diary:

Well, NVIDIA has finally dropped the bomb and announced their new card, the GeForce256 (previously known as NV10). This "Graphics Processing Unit" has a ton of exciting new features that will allow for enormous advances in realtime 3D graphics processing speed, and signifies the beginning of a new era in 3D gaming on the PC.

I'm proud to announce that WXP has been working very closely with NVIDIA for the past few months developing a special demo for NV10 called "Dagoth Moor Zoological Gardens". This demo is designed to showcase the power of the GeForce256, including full support for hardware transform and lighting.

Development of DMZG started right after E3. Our prototype hardware came soon thereafter, and we began work on the GeForce256 features. NVIDIA even sent several programmers up to our offices in Seattle to help out with optimizations and special GeForce256 features. It has been crazy to say the least: working with early hardware, early drivers, and early APIs is potentially the worst possible development scenario for any developer.

Surprisingly the process went quite smoothly, thanks in large part to the dedication and expertise of both NVIDIA and Microsoft, as well as the whole WXP team. I would like to give special thanks to Brian Harvey, Chris Donahue, and Mike Hara at NVIDIA for supporting this project, and having the vision to bring it to GeForce256 as a showcase of their amazing hardware.

It is important to note that, while DMZG runs under the Experience engine, it is NOT Experience. One of the nice side-effects of working with NVIDIA on this project has been that we were able to implement a lot of new features into the Experience engine while creating DMZG, and these features WILL roll over into Experience. Some of the new additions to the engine are:

- GPU accelerated geometry transformation
- GPU accelerated vertex lighting
- GPU accelerated cubic environment maps
- GPU accelerated character skinning
- Fading detail maps [break] Verder zal Tribes2 support hebben voor de T&L features van de GeForce. Deze info komt uit een Q&A'tje van Datumplane: [/break] James: Will there be T&L and all the alphabet soup in the GeForce 256 in Tribes2? What kind of performance should we expect on a GeForce in Tribes2? What will be the minimum graphics card requirement for Tribes2?

Mark: Yes, T&L will be supported in Tribes 2 (we're using OpenGL's transform pipeline, so it'll just work) Since we don't have a sample board, I can't comment on performance, but I expect it will be very good. Tribes 2 with the details cranked WAY down should play on pretty much any card with a good OpenGL driver... that said, expect play on low end cards to be suboptimal. [break] Planet Siege besloot daarop nog twee vragen richting Dynamix af te vuren... [/break] PSS:Will Tribes 2 support T&L exclusively with the GeForce 256, or will it work on all the new cards that support T&L? (I have seen that the Voodoo 4 and Savage 2000 will also both likely have T&L support).

Mark: Tribes 2 uses OpenGL to do all its transform and lighting - so any card that has T&L support and has an OpenGL driver that supports T&L will work with T2.

Tussen al het GeForce geweld door ook nog een berichtje van Videologic: shipping van de Neon 250 is begonnen en als het meezit moet de kaart binnenkort leverbaar zijn. Tegelijkertijd maakt Videologic bekend dat het de standaard hoeveelheid geheugen van de kaart van 16Meg naar 32Mb verhoogd heeft:

VideoLogic – the international company that develops, designs and supplies market-leading 2D/3D graphics, video and audio technologies and products – has today announced the shipping of its Neon 250 graphics card. Neon 250 is available from VideoLogic’s network of distributors, resellers and retailers for just £106 (UK RRP, excluding VAT) / $169 (USA).

Originally announced with 16MB of memory the Neon 250 will now ship with 32MB of high performance SDRAM memory at no additional cost to the end user.

Kevin Dale, marketing and sales director, VideoLogic Systems, says: "VideoLogic always strives to provide the very best value possible. Neon 250’s PowerVR 250 processor is the same ground-breaking PowerVR Series2 technology found in the latest arcade machines and Sega’s Dreamcast console. Now with 32MB of memory the Neon 250 is the essential card for PC gamers."

Leading UK PC magazine Personal Computer World presented a 16MB version of the Neon 250 with its Recommended Award in the July 1999 issue, saying: "Its advanced features, high performance and superb image quality make it a winner."

The VideoLogic Neon 250 features the NEC PowerVR 250 graphics processor with full 2X AGP support and integrated 250MHz RAMDAC and now comes with 32MB of high performance SDRAM memory running at 125MHz.

Neon 250 is based on the PowerVR 250 chip, part of the PowerVR Series2 family designed by VideoLogic Technology and NEC. Other PowerVR Series2 chips are used in Sega’s Dreamcast™ console, the Sega Naomi™ arcade system and on the R-Cade Vision 250™ from VideoLogic for the ArcadePC platform.

Misschien een interessante low-budget kaart. Eerdere previews hebben aangetoond dat de prestaties vergelijkbaar zijn met de TNT2 en dat beeldkwaliteit ook uitstekend is.

Elsa heeft gisteren de nieuwe Erazor X - gebaseerd op de nVidia GeForce 256 - aangekondigd. Hier heb je wat spul uit de press release bij GA-Source:

A complement to ELSA's high-end CAD solutions, the ERAZOR X offers an solution to serve the worldwide market for high-end entertainment PCs. The ERAZOR X features the newly announced GeForce 256 graphics chip from NVIDIA - the first to offer four independent rendering pipelines plus an additional transform and lighting engine. In action, this extra reserve of computing power crunches four-times the number of polygons than the previous generation TNT chips can handle, resulting in enhanced 3D images.

The ERAZOR X offers 32 MB of rapid synchronous DRAMs and a 300-MHz RAMDAC. Display refresh rates of up to 200 Hz can be freely defined in steps of 1 Hz with ELSA Settings, even with ultra-high resolutions. The ERAZOR X includes ELSA drivers for Windows 98 and Windows 95, with Windows 2000 and Windows-NT drivers in development. The ELSA drivers offer the easy and convenient adjustment of graphics settings to meet individual needs. Changes to resolution, resolution color depth and refresh rates can be made without having to restart the computer. A new integrated gamma correction function enables the user to adjust settings within any game via hotkey. Dark scenes can be conveniently adjusted within seconds at the press of a hotkey - eliminating the need to make changes to the monitor brightness.

Along with full OEM versions of the graphics programs Corel Draw 7 and Corel PhotoPaint 7, ELSA customers will receive the latest version of the racing game Need For Speed version 4, included with the ERAZOR X.

3D Concept heeft z'n super handige 3D chip table ge-update met de specs van de GeForce 256 en Savage2000. De belangrijkste gegevens van deze twee kaarten heb ik samen met de Glaze3D ff in een tabelletje gedouwd:

nVidia GeForce S3 Savage2000 Bitboys Glaze3D Availability Q4 '99 Q4 '99 Q1 '00 Fillrate 480MTex/sec 600-800MTex/sec 1200-4800MTex/sec Triangles/sec 15Milj/sec ? 15Milj/sec Clock speed 120MHz 150-200MHz 150MHz Buswidth 256-bits 128-bits 512-bits

Check 3DConcepts voor het complete overzicht.

Ook Riva Extreme heeft een preview van de nVidia GeForce 256 online gezet. Eén van de unieke features van de GeForce is Cube Environment Mapping (CEM), waarover je hier een stukje hebt:

This is one of the most compelling features of CEM - dynamic environments. Imagine walking through a huge field, with trees on each side. The terrain in front, behind, above and below you doesn't change that much, but the trees may change into a town, or mountains... the possibilities are endless. Now, instead of uploading an entirely new 'environment' texture, you can just change the textures on the two sides - making world transitions much more accurate and up to date. The sun? Well, just have that part of the top texture still, and move it across the +y (top) texture or have a large texture that your program 'pans' across (to send to the card). There are so many more possibilities for developers that you'll undoubtedly see a huge jump in gaming technology. Another benefit from CEM that I think is an impressive realism-enhancing feature is Fresnel Refraction. A perfect example of this is water. Here's a picture from the 'wave demo': You can see that as the viewing angle gets steeper, the more reflection you get and the less bottom you can see. This is a real-life situation that is performed extremely well by cube mapping, and will make the gaming experience better without you really noticing it.

Thresh heeft een preview van de nVidia GeForce 256 gepost. Firingsquad bevestigd dat de kaart op slechts 120Mhz loopt. Reden daarvoor is de complexiteit van de chip (23 miljoen transistors) en het daarmee gepaard gaande stroomverbruik (krijg je weer problemen met moederborden):

Yes, you heard correct the first time. The GeForce 256 can push out a sustained fill rate of 480 million pixels per second. That's over double the fill rate of the original TNT2. NVIDIA accomplishes this through GeForce's four-pixel pipeline rendering engine. We can derive the core speed of the GeForce 256 by dividing the pixel fill rate by the 4 pixel pipeline (480/4 = 120). Apparently, the GeForce 256 runs at a graphics core speed of 120MHz. By implementing massively parallel pipelines, NVIDIA is able to greatly increase the pixel fill rate without increasing clock speed. In fact, this speed is 5MHz lower than the TNT2 vanilla's core speed of 125MHz.

If we have a beef with the GeForce 256, it's certainly the clock speed. With over 23 million transistors, the GeForce exceeds Pentium III or Athlon in complexity, and the limitations of the AGP slot and power supply are already being felt. Being able to scale in parallel is a great alternative, but the fact remains that every title benefits from faster clock rate, and in many situations, there is no substitute.

De vraag is nu natuurlijk hoe de GeForce zich op hogere resoluties gaat verhouden tov de Savage2000. De Savage2000 heeft een hogere fillrate (700~800Mpix/sec), maar de GeForce heeft een bredere geheugen bus (256 bits).