CyberJunk86 schrijft: "Ja, ja je leest ut goed: de oude Pentium FDIV bug is terug, maar nu in de P3....hmzzz, hoe bedoel je ze raken gestresst door AMD! Neen Intel, dit is ook niet de manier om klanten te trekken....":

Lang geleden, toen Intel met de Pentium op de proppen kwam, bleek de processor een fout te bevatten. Dat bleek een kleine financiële ramp te zijn te zijn voor Intel, dat verplicht was om iedereen die toen een Pentium kocht terug te betalen. Het betrof een floating point divide error (FDIV), waardoor sommige berekeningen een foutief resultaat opleverden. Meer bepaald: het resultaat had een negatief in plaats van een positief teken. Volgens sommige bronnen op het Internet is hij terug, ditmaal bij de Pentium III. Op de lijst van fouten die tot nu toe gevonden werden bij de P3, vind je onder nummer 56 een fout die veel weg heeft van de originele FDIV-error. Het is niet duidelijk of het probleem opduikt bij de Katmai- of Coppermine-versies van de Pentium III.

Gerard Groenewegen schrijft: "Beste tweakers,

Er is weer een storm opgelaaid over Superweb. Superweb heeft namelijk vandaag voor de maand december abbonnements geld afgeschreven. Dit is volgens vele gebruikers wederom een afspraak die ze niet nakomen.

Hiervoor worden twee redenen aangedragen. Ten eerste zou het abbonnement pas ingaan als Superweb zou hebben gemeld dat de testfase is afgesloten (zie www.superweb.nl - service - ledenhoek) en de gebruikers een brief met uitleg hebben ontvangen. Zoals op de pagina van Superweb is te zien is dit nog niet het geval, noch heeft iemand een brief met uitleg ontvangen. Ten tweede staat hier dat het abbonnement dan pas ingaat wanneer Superweb vindt dat de testfase voorbij is.

Het afschrijven van abbonementsgeld duidt erop dat Superweb blijkbaar vindt dat men vanaf nu voldoet. Echter dan zouden ze pas op 9 januari weer geld mogen afschrijven, aangezien iedereen al een keer heeft betaald bij wie nu geld voor december is afgeschreven. Wanneer men belt met Superweb krijgt men te horen dat zij van mening zijn dat ze al vanaf 5 november naar behoren functioneren. Er zullen weinig gebruikers zijn die dit beaamen".

Oh ooh, /me ziet een heftige wolk reakties boven deze nieuwsposting open barsten...

Betanews bericht dat de Microserfs hun Mobile Explorer software hebben aangekondigd. Met deze zak vol internet applicaties laat je je mobieltje gezellig tegen het internet babbelen:

Microsoft aims to enable the use of the Internet, e-mail, location services, and e-commerce over mobile phones with hopes of Mobile Explorer becoming the stepping-stone for developing applications for new wireless technologies. Current services such as Microsoft Exchange Server, the BackOffice family, and MSN Mobile Services, will also be integrated into Mobile Explorer. In a press release issued today, Harel Kodesh, vice president of the Productivity Appliance Division at Microsoft said, "Microsoft's goal is to enable secure information access any time, anywhere, from the device of the users' choice."

Mobile Explorer will come in two versions, one for "smart phones" and the other for Web-ready "feature phones." Mobile Explorer for feature phones will include an operating-independent, browser based system. It will be the first dual-mode microbrowser that can display both HTML- and WAP (Wireless Application Protocol) 1.1-authored Internet content. Mobile Explorer for smart phones will include a full suite of personal management and Internet tools in addition to features offered on feature phones.

McMiGHtY schrijft: "Lopen die Duitsers dus bijna op ons voor... Het wordt tijd dat iedereen in Nederland aan het internet gaat, al was het alleen maar om de bandbreedte van de duitsers in te korten".

Nounou. Hoe meer vriendjes op het net, hoe leuker het wordt :

Duitsers waren in de maand oktober gemiddeld vijf uur online. Dat is slechts dertig minuten minder dan Amerikanen, de traditionele voorlopers qua internetgebruik.

Dat blijkt uit een onderzoek van Media Metrix. Het is de eerste maandelijkse studie naar internetgebruik in Europa van dit onderzoeksbedrijf, vergelijkbaar met het onderzoek dat het in Amerika doet.

Semi-anonieme Fred Internet schrijft: "Nederland is weer een URL redirect service rijker. http://glow.nl is lekker kort en uiteraard geheel gratis! Mijn homepageadres bij WISH ben ik al vergeten. Ik glow, jij ook?":

Nadat u uw domeinnaam heeft aangevraagd wordt er een zogenaamd subdomein aangemaakt. Dit subdomein komt altijd vóór glow.nl te staan. Uw nieuwe Internet-adres ziet er dan bijvoorbeeld uit als: http://flower.glow.nl, of http://www.flower.glow.nl. Op de adresregel van de Internet browser blijft ook netjes uw nieuwe adres staan.

Eén van de grote voordelen is o.a. dat uwnaam.glow.nl gemakkelijker in te tikken is dan een lang homepageadres.

VHware schrijft: "In dit artikel van CPU Review wordt een vrij grote lijst met KX133 mainboarden opgesomd. Verder wordt gezegd dat nieuwe revisies van de chipset mogelijk SMP support meebrengen:":

I am looking forward to reviewing motherboards based on the KX133 chipset; by supporting PC133 it will increase the memory bandwidth from 800Mb/sec to 1Gb/sec.

Future revision of the chipset will support DDR SDRAM; which would allow 1.6Gb/sec with PC200 and 2.1Gb/sec with PC266 SDRAM - equaling and surpassing 800Mhz RDRAM's 1.6Gb/sec bandwidth.

Early next year will be even more interesting; AMD is reportedly working on an SMP chipset, and while the VIA FAQ states that the KX133 is not SMP capable, the feature list makes a point of mentioning that an IO/APIC is included. I would not be at all surprised if it turns out that a future KX133 revision will support SMP.

VHware kwam aanzetten met dit artikel van AboutLinux.com, over Athlon performance tweaking in Linux:

Recently I wrote a review (at CPUReview) of Linux running on an Athlon 600 based system. While the benchmark results I obtained for that review were excellent; I could not help but feel that the scores could be higher... and egcs has all those nice optimization flags...

I decided to run a series of experiments; covering the spectrum from no optimization ("how to turn your K7-600 into a K6-266") to the equivalent of a K7-720 for memory and integer oriented code - and by following the optimization settings shown in the article, you can get up to 73% improvement on some types of code.

I was not the only one to think of trying different optimization settings; I got an e-mail yesterday from a reader suggesting that I play with compiler settings...

EGCS developers should take note of these results; the best (overall) performing configuration did not turn in the best results for all tests; therefore examining the optimization settings show here and the code for the benchmarks can tell them what sections of the code generator and optimizer need to be tweaked to produce even better Athlon results!

biwan@multiweb.nl" rel="external">Obiwan schrijft: "Nope, dit is geen japans plastic prut setje":

The first time you hear the Quake railgun humming through 160watts or pure sass, you'll be heading out to the local store to pick up your favorite brand of adult diapers. The sheer bass that emits from the Promedia's was enough to send the Midiland 4100's and the DTT2500's scurrying away begging for mercy. I'm being overly descriptive here to make a point. Keep in mind that no review can "prove" to you how well something sounds, or how loud they might be.

Het frustrerende is dus dat dit super geile setje niet in ons kleine klotelandje verkrijgbaar is .

DOS stuurde deze link naar een nieuwsbericht van ZDNet, die meldt dat Nortel een technologie heeft ontwikkelt waarmee kabelgebruikers beschermd kunnen worden tegen de woeste inbraken van op hol geslagen scriptkiddies:

Sommige kabel-abonnees beschermen zichzelf tegen infiltratie met speciale software of 'firewall'-hardware. De oplossing van Nortel Networks, Secure Cable, is een firewall die aanvallen van hackers kan blokkeren voordat zij de pc van een kabel-abonnee kunnen bereiken. Secure Cable is onderdeel van een nieuwe netwerk-connectiebox.

Nortel heeft soortgelijke oplossingen voor telefoon-verbindingen en DSL. De laatste is een high-speed telefoonverbinding die ook permanent open kan staan, want de gebruiker betaalt geen telefoontikken. Om die reden is DSL net zo hack-gevoelig als een kabelverbinding.

Gewoon een vuurmuur dus. Hopelijk brengt het providers niet op akelige ideeën, zoals het dichtgooien van (downstream) FTP poorten.

Fastgraphics heeft een roundup van 7 GeForce kaarten in elkaar getimmerd. Onder de deelnemers bevinden zich de ASUS V6600, ASUS V6600 Deluxe, Creative Labs 3D Blaster Annihilator, Creative Labs 3D Blaster Annihilator Pro, Elsa Erazor X, Guillemot 3D Prophet en Hercules 3D Prophet DDR / DVI. De laatst genoemde blijkt dankzij DDR SDRAM en een iets hogere core clock als snelste uit de test te komen:

"Hercules 3D Prophet? ... But I thought that 3D Prophet was a trademark from Guillemot ?" Well, as mentioned in the intro of this review, Hercules has been acquired by Guillemot, and starting with this card all new Guillemot graphic cards will be sold under the Hercules brandname. So basically this is a Guillemot card with Hercules's name on it.. The card is exactly the same as the board from Creative (nVidia reference) but it does have some nice extra's. Like the normal 3D Prophet this card as an S-Video output and in addition we also find a DVI port on this card. DVI stands for Digital Visual Interface. A DVI output delivers digital image output (so the RamDac is bypassed) for new high-end CRT monitors or digital flatpanels. DVI supports resolutions up to 1600x1200 with 32 bit color. Unfortunately I don't have a DVI ready CRT or flatpanel monitor, so I wasn't able to test this feature.

Guillemot also did something which is officially not allowed by nVidia: They're shipping this board with a default speed of 130/150, even though nVidia forbids vendors to ship cards running above the default speed, which is 120/150 for DDR cards. Due to the higher core speed of the 3D Prophet DDR-DVI it performs a bit better than the Annihilator DDR from Creative Labs.

hjs schrijft dat bij nV News een interview te vinden is met de enige echte Steve Mosher van Creative Labs. In het interview wordt wat geblaat over 3dfx, transform and lighting en Kerstmis . Hieronder een gedeelte:

nV News: S3 seems to have dropped the ball on hardware transform and lighting (T&L) for the time being. Do you feel that T&L from NVIDIA lost a competitor or an ally?

Steve: I don't think the Savage part competes with the GeForce part. I can't speak to the functionality of the T&L in their part, since I personally haven't evaluated it. I believe that T&L must be in every graphics chip going forward. The sooner you get it in the sooner you get all the issues worked out. Anybody who has ever programmed complex 3D graphics welcomes hardware T&L.

This sounds a lot like the 32bit color debate to me. When guys don't have 32bit color, they say its not necessary. Then they catch up and try to thump their chest about it. People remember.

nV News: How much weight do you place on full-scene anti-aliasing? Can higher fillrate eliminate edge "jaggies" through higher resolutions in conjunction with increased triangles sufficiently? If so, what resolution will be the new sweet spot, now that the goal of 1024x768x32bpp @60fps has been achieved?

Steve: Boy that is a lot of different questions. How much weight on AA? AA is nice if it comes for free, if it costs too much fill, you can get the same effect by boosting resolution. Also, you have ask what kind of AA filter you are using - 2x2 or 4x4.

The benifits of AA are also content dependent. All content has edges, but contrast at the edge also contributes to the effect. So the only way to tell is to run some content and compare. You really have to do this side by side in real time. Screen captures miss the temporal effects and over emphasize the spatial effects.

The new sweet spot? As far as your eye and brain are concerned, we are a long way from saturating the human visual system.

hjs kwam vertellen dat er een patch beschikbaar is voor het kontschoppende spel Unreal Tournament. Hieronder de updates die Unreal Tournament v402 Update ons schenkt:

Bug fixes:

- D3D detection works properly.
- removed log warnings for players ghosting out of the level
- fixed saving last start spot for players
- timemessage sounds play correctly if viewing another player
- don't allow spaces in names when using console name or setname commands
- suicide can now end overtime of tied game
- fixed stuck animation if killed while coming out of feign death
- no connect time on team scoreboards in single player
- attach decals in post begin play when zone is properly set
- fixed player sounds appearing to change location when using 3d sound
- turn fog off by default for D3D - was causing texture thrashing w/ D3D drivers
- fixed CTF-coret alternate path problems
- fixed guided warhead blowing up if no client response with 300 msec after spawning
- players can no longer switch to illegal team if server is full and force team balance set
- fixed bots not always playing take hit sounds
- don't kill spectators in the pressure chamber!!!

Server performance improvements:

- Don't replicate actor properties of GameReplicationInfo
- Don't replicate actor properties of ZoneInfo
- Less frequent visibility checks for currently relevant actors
- don't keep checking visibility of optional effects
- less frequent checks for temporary actors after 0.2 seconds
- reduced playerreplicationinfo update frequency from 5/sec to 4/sec
- inventory class-specific scripted replication only done if bNetOwner
- native replication for weapon, pickup, levelinfo, and ammo classes
- don't spawn decals on dedicated server
- don't spawn non-replicated effects actors on dedicated server
- sort static actors so can ignore non-replicated static actors when testing for replication


New features:

- added flesh hit sound to bullets
- automatic path builder (documented in updated UT_AI.htm)
- postnetbeginplay() called for actors spawned on network clients
- Pawns use PostNetBeginPlay() to make sure they have a valid skin and mesh
- improved D3D texture pool allocation when world and actor texture detail is medium
- Pawns add momentum to velocity after effects of damage mutator
- Optional package downloading implemented.


Download locations

Download from Unreal Universe Mirror 1 (3.5 MB) FTP (fast) 350 users

Download from FilePlanet

GamePC heeft z'n site voorzien van een review van de Abit BE6-II moederplaat. Zoals bekend istie uitgerust met SoftMenu III en een ATA/66 controller en behoort tot de beste BX planken van het moment.

Comparing the new BE6-II to the old BE6, we notice that both boards are very similar yet differ in areas, which is more than likely why the BE6-II didn’t get a different name. The BE6-II uses SOFT-MENU III, which is a newer BIOS that is derived from SOFT-MENU II found on the original BE6 mainboard. The differences in the two BIOS’s are significant enough to take note of for all the controlling features offered to the user. One example is the tremendous amount of FSB settings this board allows, 66, 75, 83, and 84 – 200 (that’s right 84 MHz to 200 MHz in increments of 1 MHz). A real shocker to me with this board was that the BE6-II board had DIP switches (same thing found on ASUS’s P3B-F) build on the board that could also control CPU settings. Though not a very common thing to be found on ABIT boards, but it is a helpful feature to get a system operable again when new BIOS settings that you may have changed to are incorrect. Using a new board layout the BE6-II is 10mm shorter in width then the BE6. Hardware monitoring and cooling features is nothing new with the exception of differently placed fan headers and a slightly smaller heatsink that cools the 440BX chipset. There is no answer why ABIT went with a smaller heatsink, but now height won’t be an issue when buying a Pentium III heatsink/fan. Previously, large coolers like the Vantec P3D-5030 couldn’t fit correctly on the BE6, thanks to the larger BX chipset heatsink that it included, now it’s not an issue.

Other than SoftMenu III, there is very little difference between the two boards. They both have a single AGP 2x slot, 5 PCI slots, 3 DIMM slots, two Ultra DMA/66 and two Ultra DMA/33 channels. The BE6-II did strip off a single ISA slot, a good start, but Abit decided to not put on a 6th PCI slot, a change that would have been welcomed by everyone. Abit also did some minor modifications to component placing, like slightly modifying the position of the power connector, fan connectors, BIOS chip, and other miscellaneous items, nothing major, but all are positive moves if you work with computers all day long.

The Super 7 Hardware Guide heeft een kleine en beknopte review online gepleurd van de Gigabyte GA-7IX Athlon mobo. We gaan maar meteen FFWD naar de conclusie:

Gigabyte's GA-7IX is a workhorse of a mainboard offering good stability and quality construction. A bit on the pricey side, as are all Slot A mainboards, the board would be a great choice for the neophyte upgrader or system builder. It's lack of access to overclocking settings will probably raise a ruckus with the overclocking community but if you're looking for a build it forget it system that has some teeth and upgradability, I don't see any reason to shy away from the GA-7IX. In fact, I am so impressed with its solid performance that it has become the heart of my personal workstation.

Bij AnandTech Webnews liep ik tegen het volgende bericht:

I read at Overclockers' Workbench that Microstar will be releasing another version of MSI 6167 Athlon board with multiple FSB support like the ASUS K7M in mid December with a model number of MSI 619x.

Bij Hardware Abyss is een preview te lezen van Bill's Windows Millennium, nog steeds gebaseerd op de 9x 16/32-bit hybrid. De preview geeft je veel info over het OS (nu in Beta2) en je vindt er ook een zooi screenshots. Hieronder wat over de performance en stabiliteit:

Windows Millennium is suprisingly stable, compared to the previous versions of 9x. Although it is not as stable as Windows 2000, I believe that this is the best they can do, under a 16/32-bit enviornment. The performance has increased in general. It is a noticable increase as applications load up faster, and there is a smoother transaction between multitasking applications. Memory management has also improved, and it can be seen through the amount of resources which is taken up when a program is loaded. Beta 2 has truely impressed me.

Het sprookje over de zaligmakende Camino chipset krijgt wederom een vervolg. Volgens The Register gaan VC820 planken lastig doen wanneer 2 RIMMetjes van verschillende fabrikanten gemixt worden. CC820 borden worden leip wanneer meerdere geheugenbanken gevuld worden met 'onrein' SDRAM:

Although Intel engineers solved the problem by shutting off one of the sockets with a terminator, there is an additional factor which you will need to take into account if you want to populate the beastie.

If you use two RIMMs and they are from different manufacturers, you can expect problems, according to very reliable sources. That will mean that if you are buying or specifying a system that uses the VC820, be very careful that you use the right kind of memory.

And in further i820 news, dealers are reporting what they describe as "serious problems" with CC820 i820 systems. One dealer said: "One piece PC-100 128MB memory seemed to run file, while two pieces easily crashed the system. Intel engineers vehemently claim that there is something wrong with most manufacturers' memory -- they told me 'most of them are more interested in making a buck than in making PC-100 memory according to specs'".

Intel engineers told him that "there are some serious timing issues that will be resolved in later revisions of the chipset".

Lijkt erop dat die Packard Hell en Compaq bakken eindelijk van een redelijke videokaart voorzien gaan worden, want nVidia is er in geslaagd om een OEM deal met beide bedrijven binnen te hengelen:

NVIDIA's products have been chosen to boost graphics performance on both the PACKARD-Bell Platinum line and Compaq Pesario platforms. The RIVA TNT2 Pro 3D on PACKARD-Bell's new Pentium III machines will deliver 286 million pixels of processing power per second with a 32MB frame buffer, 32-bit color pipeline and 32-bit Z-stencil buffer. The Riva card supports OpenGL and Direct3D acceleration.

NVIDIA's GeForce 256 is now available through Compaq's "Built For You" self-configuration program for the Presario. The GeForce 256 includes nearly 23 million transistors, more than twice the complexity of the Intel Pentium III microprocessor. With powerful floating-point calculation capabilities, the device delivers 15 million sustained polygons per second and 480 million pixels per second. GeForce 256 supports up to 128MB of frame buffer memory, full AGP with fast writes, and a 350MHz RAMDAC to drive resolutions and color depths up to 2048x1536 at 75Hz.

JC heeft wat info bijeen weten te harken over de performance van de Itanium systemen, die op dit moment aan hard- en software developers geleverd worden. De eerste samples draaien op 600MHz, performance is boven verwachting:

Merced is sampling between 500MHz and 600MHz right now. Clock frequencies at release are expected to be much higher. Performance is above expectations. Confirmation of L1, L2, and L3 cache (and another woohoo from me that they ceased calling it L0, L1, and L2). They'll have a few hundred samples out by the end of this month, and several thousand starting January. It's intended for systems that will cost over $100,000 -- Intel is finally making good on their promise to accelerate the internet.

- Claimed to be ten times faster at SSL than "its peers".
- Peak of 6 GFLOPS with two fpus (is this a single precision count? Using SIMD?).
- Programmers' reference manual in 2000 H1

Dean Kent van Silicon Investor heeft een nieuwe Industry Update gepost, met wederom een brede doos info over de bewegingen in de processor, chipset, moederborden en geheugenmarkt:

Rambus production began again in earnest last month, in conjunction with the i820 release. Samsung expects to grab the majority of the market, particularly since Micron is does not appear anxious to get in. In addition, sources say that Intel has disqualified all Rambus memory suppliers in their qualifications, with the exception of Samsung. Reportedly, the testing was system-level and suppliers were failing due to excessive noise. These sources indicate that Toshiba proposed a mask revision to fix the problem, but Intel rejected it as too time-consuming. Intel is reportedly working on a PCB fix.

VIA is now pushing the PC2000 standard, which is DDR SDRAM. DDR, which means Double Data Rate, allows the data to be transmitted on both the rising and falling edges of the clock. While the internal DRAM speeds (and thus the latency) are the same as PC133, relatively large internal buffers allow the data to be sent out more quickly, resulting in an effective transfer rate of 266MHz. Micron has some DDR SDRAM on their roadmap for the near future, as do several other manufacturers. Once chipset support is widely available, this memory will compete head-to-head with DRDRAM.

De DDR chipset van VIA staat gepland voor het tweede kwartaal van 2000. De release date van de KX133 chipset is vooruit geschoven naar december. Check het artikel voor meer info.

Jason van AnandTech heeft een review van de dikke Iiyama Vision Master Pro 510 neer gemikt. Deze 22" heeft erg goede specs en is met een prijskaartje van ca FL 1925 sympathiek geprijst:

Photo Tests: I used a resolution of 1600X1280 at 85Hz with the color depth at 32bit. The clarity was fantastic here. Colors were vibrant and sharp, and with the size of this monitor, it just makes it that much more enjoyable.

Text Test: Setting it at various resolutions, I loaded text documents into Microsoft Word, and Notepad. I examined the clarity of the text on different colored backgrounds. Text was clear and sharp, at all resolutions up to 1600X1280. I found that anything above that, text gets too small for proper use.

Resolution Test: Here, I simply tried every resolution up to the monitor's maximum, 2048X1536, and looked for defects and clarity issues in the display. In my opinion, anything above 1600X1200 for text starts to get unreadable for my eyes; although, the monitor performed extremely well right up to its maximum 2048X1536.

Unreality.org heeft een interview gepost met Tim Sweeney, wat o.a. een aantal verhelderende antwoorden over Unreal 2 en Epics' next-gen 3D engine opleverde:

Bren86: I'm a little confused as to what the code difference between "Unreal 2" and "next engine" are. With Unreal 2 just now in design stage for Legend, will it be benefitting from "next engine" or will it be an enhancement of the UT engine? If the latter is true, will Epic and/or Digital Extremes be collaborating in some way with Legend on Unreal 2?

Tim S: Unreal 2 = Sequel to Unreal 1, being developed by Legend Entertainment, using an enhanced version of the Unreal Tournament engine. We will be supplying all-new skeletal animation code and improved mesh handling. Legend will be adding a bunch of features. We aren't ready to announce any of the other cool stuff that will be added for Unreal 2.

Next-Generation Engine = The engine that will be used for Epic's next large-scale project. The timeframe for this is "late 2001 or later". The theme and style are still being invented, so don't ask -- we'll be talking about it when we're ready. Don't assume it will necessarily be another game in the style of our past games. All we can say is that it will be 3D and it will be revolutionary.

Digital Extremes = The Canadian co-developers of Unreal 1 and Unreal Tournament along with Epic. We've been working closely with James Schmalz and the Digital Extremes team since 1993 (remember Solar Winds, Epic Pinball, and Extreme Pinball?) Now that Unreal Tournament is done and both companies are in great financial shape, we're each doing our own project. Since UT shipped, they've been adding lots of cool new features to the Unreal Tournament engine and are starting on their new project independently using the engine.

VHware schrijft: "Samsung maakt een chipset waarmee het mogelijk is om de AMD Athlon en de Digital (Compaq) Alpha processor aan te sturen":

Samsung also told EBN that it plans to break into the Alpha EV6 mobo and chipset market with a single-CPU chipset codenamed Caspian in Q3 (sampling in March) and Tasman, a four to eight-way server chipset, sometime in 2001.

The clever part is that Samsung will use Caspian to get into the x86 market, by supporting AMD's Athlon, which uses the same bus as the Alpha EV6, in the same chipset.

Curiously, given Samsung's desire to drive the Rambus market, Caspian will not support that memory technology, only PC133 and DDR SDRAM.

Was het niet leuk geweest als ze daar meteen een SMP chipset van gemaakt hadden?