Sharky Extreme brengt ons de info over ATi's geheime Aurora project. Zoals al hier en daar was gezegd gaat het om een nieuwe videokaart, de ATi Rage Fury MAXX (goh, creativiteit ten top). Het is de bedoeling dat op de kaart twee ATi Rage128 Pro chips in SMP mode gaan samenwerken. Verder moet-ie op zo'n 500 tot 600MPix/sec gaan uitkomen. Hieronder de specificaties:

- Twin (2) ATI RAGE 128 PRO graphics engines
- Dual (2) Triangle Setup Engines
- Dual (2) Texture Caches 500 M/Pixels per second total fill rate
- 4GB/sec memory bandwidth
- 2 x 32MB Banks of dedicated memory, totaling 64MB
- 32-bit true color 3D acceleration (16.7 million colors) up to 1920x1200
- 2D graphics up to 1920x1200@32bpp
- Full OpenGL ICD, Direct3D and DirectX acceleration
- Hardware DVD for full-frame rate, full-screen DVD
- AGP 4X (AGP 2X compatible)
- Bilinear/Trilinear Filtering
- Line & Edge Anti-Aliasing
- Texture Compositing
- Texture Decompression
- Specular Lighting
- Perspectively Correct Texture Mapping
- Mip-Mapping
- Z-buffering and Double-buffering
- Bump mapping
- Fog effects, texture lighting, video textures, reflections, shadows, spotlights, LOD biasing and texture morphing

The speed of the early Aurora board we tested had each of its Rage128 Pro chips running at 125MHz, while the ram speed was set to approximately 150MHz. This produces a maximum fill rate of 500M/Pixels per second, although it's likely this number will climb before the product's launch.

The reason why we expect a fill rate increase is the fact that we just received the single chip version of Aurora, which is basically the ATi Rage Fury Pro 32MB AGP card, and it's clocked to run out of the box at a 143MHz core speed with ram that operates at 155MHz.

If we take the high volume $149 Rage Fury Pro 32MB card's core speed and extrapolate the fill rate numbers if a dual-core Aurora card was set to run at the same speed (143MHz core x 2) we get an approximate fill rate estimate of 572M/Pixels per second.

That figure is a hair away from being double what a TNT2 Ultra's stock fill rate is, and it's also almost 100M/Pixels per second higher than NVIDIA's newest graphics chip, the GeForce256.

Hm, ik ben errug benieuwd naar de eerste benchmarks scores van dit kaartje. Fiets even naar Sharky Extreme voor de rest van de info!

Over bij Beta Bites hebben ze een kleine review in elkaar gezet van de nieuwe Aureal Vortex2 SQ2500 soundcard. Hieronder een gedeelte over zijn performance, de rest van de review staat hier.

For games, I played demos of Slave Zero, System Shock II, and the new Unreal Tournament, all of which support A3D technology. One of the most noticeable features of the games I played is the .wav files sound as if someone is in the room with you speaking to you. SSII and UT were just great when it comes to listening pleasure! Now, put on a pair headphones, and it is like being there. I found the SQ2500 to have great headphone support.

Aureal optimized the Vortex2 chip in SQ2500 for better performance in games. I did not do any comparisons on framerates between cards, so I can't confirm this optimization However, Aureal claims, you are supposed see some improvement in game performance with the SQ2500. I can tell you that other reviews on the SQ2500 have supported this claim with a slight performance improvement.

I also listened to several MP3 tracks on WinAmp. I found the sound outstanding as far as quality goes. The pureness of the sound is what I hear when I listen to A3D technology.

The MIDI sounds to me seem improved over the past Vortex2 cards I have listened too. The SQ2500 also provides a DVD output which is nice for those of you having one.

Bij Review Finder kun je een review vinden van ViewSonic's P817 21'" monitor. Hieronder wat info over z'n beeldkwaliteit.

Let me start with this: ViewSonic says this monitor is "For the imaging, CAD/CAM and graphics professionals who require the highest resolutions". Well, that must be true, because at a resolution of 2048x1536, the windows desktop is far too small. To get an idea of what a desktop of that size looks like (resized to 1024x768), click here. I couldn't imagine working in this resolution (especially using Small Fonts mode in windows). With this monitor, my preferred resolution is 1280x1024. At that screen size, text is about the same size that it was on my 17" monitor at 1024x768, and the refresh rate is rock-steady.

The picture quality is every bit as good as my ViewSonic 17PS, which is excellent. Since I'm not a CRT engineer, I'm not really qualified to comment on things like convergence, but I can tell you what I see. At almost licking-distance of the screen (Note: Please do not lick your screen), text is clear and sharp, even at the maximum resolution.

Check de hele review hier.

Eric van Fast Graphics heeft wat info gepost over nVidia's 3.34 reference drivers. Deze zouden namelijk de volgende chips van nVidia ondersteunen:

Installing the new 3.34 drivers for my GeForce test, I noticed that these new drivers support the following new chipsets:

- nVidia GeForce 256
- nVidia Riva TNT2 Pro
- nVidia Quadro
- nVidia Crush

What's all that mean? I don't know, but it seems like nVidia has quite a few chipsets coming up.. I'll keep you posted..

Die eerste twee komen me bekend voor, maar die andere?

Creative Labs heeft nieuwe win9x en NT drivers gereleased voor haar TNT en TNT2 kaarten. Ze zijn gebaseerd op de 2.08 reference drivers van nVidia.

Creative TNT/2/Ultra drivers voor Win9X
Creative TNT/2/Ultra drivers voor WinNT

Bij Sharky Extreme is net een review gepost van de ATI Rage Fury Pro videokaart. Dit beestje krijgt de Rage128 Pro chip mee die geklokt is op 143MHz. Het 32MB geheugen draait op een acceptabele 155MHz. Hieronder een hap uit de test:

Right away the advantages of the Rage128 Pro's 32bpp engine become clear as the card produced its best results when set to that color depth, regardless of resolution.

ATI stresses their product's visual quality and as such they're big believers in 32bpp applicATIons. Their drivers are optimized well for that color depth, and the degree of software engineering becomes evident in test results like the 1024x768x32bpp scores for the Expendable Timedemo. In that test, the Rage Fury Pro bested the TNT2 card by approximately 30% and this lead was retained as the resolution went up to 1600x1200x16bpp.

The Rage Fury seemed to have trouble scoring well in the older DirectX5.0 compATIble Forsaken benchmark, as the TNT2 card trounced it in that test's 16bpp tests.

But where it really counts, in DirectX6.1+ and OpenGL engines at 32bbp color depths, the Rage Fury delivers the performance gamers demand when playing their favorite entertainment titles.

Lees hier meer over deze kaart.

Als je echt geen genoeg kan krijgen van Lara Croft, GameSpot heeft een buttload (meer dan 100) aan pics van Tomb Raider 4 online gekieperd. Begin hier.

Was je al bezorgd over de relatief lage 480Mpix/sec van de nVidia GeForce 256, Voodoo Extreme laat nu weten dat je zelfs dit alleen haalt als je een GeForce kaart hebt met DDR RAM. Een 'gewone' GeForce kaart met SDRAM zou maar de helft, dus 240Mpix/sec trekken. Onderstaande is wat Mika Tuomi wat BitBoys (Glaze3D) aan VE e-mailde:

I wanted to mention one thing to you that I've noticed nobody seems to want to talk about. The GeForce memory types.

To my understanding, the fillrate number of 480Mpixel/sec nVidia announced for GeForce is only achieved with DoubleDataRate memory (DDR) which is not very well available at the moment. And SDRAM memory gives only half of the bandwidth. So, the first cards I've seen announced like Creatives etc, all use SDRAM. Which means the fillrate is 240Mpix/sec.

480Mpix/sec with 32bit color and 32bit Z requires one read (Z) and two writes (Color, Z) which is 5.76Gb/sec bandwidth. For example 128bits of SDRAM at 150MHz gives 2.4Gb/sec bandwidth, which equals to maximum of 200Mpix/sec (32bit pixels). For the comparison, Glaze3D has 12Gb/sec bandwidth.

So, people should really wait for the DDR version of the geforce boards if they want the announced 480Mpix/sec speed.

Alhoewel Billy van VE afgelopen vrijdag hierover al bij nVidia heeft nagevraagd heeft hij nog geen reactie ontvangen. Het zou in ieder geval zwaar zuigend zijn als dit verhaaltje zou kloppen. De DDR versie zal zo'n 200 piek extra gaan kosten namelijk. Hopelijk snel duidelijk van nVidia zelf!

In het tweede deel van het Next Generation Motherboard Platforms artikel kijkt Anand naar een drietal i820 moederborden, namelijk Intel's eigen Vancouver VC820, de AOpen AX6C en de Gigabyte GA-6CX. Hieronder een gedeelte uit het artikel.

You would think that it would be very easy to predict the future of the i820 based motherboard market. It should theoretically follow in the footsteps of the BX platform. But then again, when the BX chipset was released, there wasn't a true competitor to the chipset around. The i820 is a step in a different direction, the RDRAM platform has yet to be embraced by the industry and unfortunately for Intel, there are viable alternatives.

Micron has already launched a "we told you so" campaign praising their decision to pursue VIA's i820 alternative, the quietly released Apollo Pro 133A, over Intel's delayed i820. If we also take into account the fact that the i810 chipset was a huge disappointment in the minds of many motherboard manufacturers (it did begin to catch on after awhile but it was a very slow start), the future of the i820 chipset is a bit sketchy.

With no native PC133 support without the addition of a Memory Translator Hub (MTH) that will most likely not come without a performance penalty, the RDRAM prerequisite of the i820 platform will keep it out of the hands of many. At the same time, Intel has already made it clear that the upcoming Pentium III E (Coppermine) will be available in a Slot-1/100MHz FSB version which further extends the life of the BX chipset. It is unlikely that motherboard manufacturers will completely ignore the i820 chipset and continue to produce newer BX boards, but don't be surprised to see your favorite BX boards stick around for a little while longer while i820 becomes established.

Once the motherboard manufacturers, most of which are still recovering from the horrible earthquakes in Taiwan, do jump on the i820 bandwagon, you can expect quite a fewinnovations brought to motherboards based on this chipset.

De rest van het verhaaltje vind je hier.

Inprise (vroeger Borland) stort zich helemaal op Linux.

Inprise Corporation (Nasdaq: INPR), today announced that it is developing a high performance Linux application development environment that will support C, C++, and Delphi development. The project, code named "Kylix", is set for release next year and will be one of the first high performance Rapid Application Development (RAD) development tools for the Linux platform.

Volgend jaar dus al!
Daarnaast hebben ze al een Java JIT (Just-In-Time-compiler) voor Linux gemaakt en hebben ze een dealtje gesloten met Corel over een gezamelijke Linux aanpak.

Ik kwam bij Betanews wat info tegen over WindowBlinds 1.1:

Stardock is working hard developing the next version of WindowBlinds, the top user interface enhancement tool for Windows. Version 1.05 beta will be made available to Object Desktop users this week and other registered WindowBlinds users sometime next week. Version 1.1, slated for release next month will add a bunch of new features, some of which can be seen in the screenshot below. Notice what they're doing with the taskbar menu and with scrollbars. If you haven't tried WindowBlinds, you are missing out. Version 1.0 can be downloaded at the right and gives you complete control over the look of your desktop. Registering costs a mere $20 USD and gives you free access to upgrades. [break]
Zelf draai ik al ruim een week met WindowBlinds 1.0 en op m'n Win2000 doos draait dat stabiel, zonder problemen:

En nu we het toch over window managers hebben: er is ook een nieuwe release candidate van Enlightenment 0.16 gereleased: [/break] Release candidate 3 of Enlightenment 0.16 has been officially released. Enlightenment is a window manager for X. It's design goal is to be as configurable as possible - in look AND in feel. Currently in the look department it pretty much has that done. You can design your own window borders, menus, and other elements of your screen to look any way you like. You don't have to touch one line of code and you don't have to recompile anything. This release includes a few bug fixes over Monday's release candidate 2.

De binary kun je hier naar beneden halen.

Ik kreeg een mailtje van Jason Hurley van Savage Daily News, over een nieuwe MiniGL driver voor de Savage3D (dat is groot nieuws in S3 land, daar waar de drivers brak zijn ):

It's rough in spots, but still, this will let people play Quake3 with the MiniGL with better speed than before.

With that said, here are some obvious limitations of the MiniGL on Quake III:

1. Occasionally, you may see some objects connected by a thin line or triangle that shouldn't be there.

2. The animated textures on the bounce pads are slightly wrong.

3. Dynamic lighting, especially from things like the rocket launcher, are not working correctly. You'll have to change Q3 to use vertex lighting to avoid this problem.

4. Fog color is incorrect, so blue and red are reversed.

5. In the Q3Test2 map, sometimes you'll see odd triangles in the distance that appear to be part of the skybox.

Instructions for use (for Savage3D owners only):
------------------------------------------------
Put both files (metal.dll and opengl32.dll) in your Q3Test directory. If there is already an opengl32.dll file, you may want to make a backup of that file, just in case.

Once these files are in your Q3Test directory, you should be able to run Q3Test. Note that only very limited testing has been done on this MiniGL, so there may be some other problems. [break] De drivers kun je hier downloaden. Ik lees inmiddels bij Savage Daily News dat de driver ook bruikbaar is voor de Savage4 (framerate toename van 59,8 naar 67,3fps in Quake2). Ennuh... [/break] I don't speak Dutch, but I still visit your site several times a day. It's one of the best news sites I've seen! :-)

Hehe, cool

Bill heeft (als onderdeel van z'n NT Server site) een page in elkaar gerammeld waarin het de zogenaamde 'Linux Myths' weerlegt. Volgens Microsoft is Linux niet sneller dan NT, is het niet geschikt voor de desktop, is de beveiliging klote en is het besturingssysteem niet gratis (rekening houdend met Total Cost of Ownership).

Uiteraard vind de Linux community het niet leuk als Billie dat soort dingen gaat zeggen, met als gevolg een 65 kilometer lange thread bij Slashdot en een instant reaktie van Mandrake (die gozer van Enlightenment):

So naturally, surfing the net, I came across the official microsoft linux FUD page. I'm not sure I even know how to respond to this - instead of just attacking microsoft in general I'll stick to the bits I have personal knowledge of.

1. Reality: Linux Makes no Sense at the Desktop.
   We'll start with this point, since I find it the most amusing one, personally. I don't think they've been paying attention to the world, do you? look at all the various works that are being done by various parties, be it the GNOME camp, the KDE camp, enlightenment, windowmaker, etc. Just because it doesn't act like windows, and it doesn't look like windows, that doesn't mean it's not usable.
   Maybe they should spend some time investigating how difficult it is for people to use windows before they start harshing on other systems.
2. Linux does not provide support for the broad range of hardware in use today
   Well, maybe we don't support all the hardware out there, but because of companies like microsoft it becomes increasingly harder to support some hardware because you make people feel they have to hide how everything works. The REALITY of the situation is that linux really DOES support the bulk of hardware out there - exception maybe being modems, since winmodem is about the poorest idea that anyone has come up with.
3. Application support is limited
   Okay, this is just total crap. Sure, the same applicatiosn aren't found on linux as are found under windwos, but the reverse is also true. I can't use a lot of the familiar X applications under windows, either. So what's your point? You talk about ISV's supplying software for windows (93%) but then you forget about all the rest of us who don't write software that you have to pay money for. Maybe you should investigate this further, too
4. cumbersome nature of existing GUIs
   Right, you must be talking about windows here. makes sense, it's difficult to maintain windows boxes, so the inverse must be true, eh? ick.
5. Linux security model is weak
   because it's based on the unix security model? that's your argument?
6. a user who needs any administrative capability must be made a full administrator
   that's crap and you know it! haven't you ever heard of SUDO? oh, I forgot, you windows people don't really understand permissions delegation if it's not in a point-and-click gui.
7. windows nt security is easy to set up and administer with tools such as the security configuration editor
   right - this tool is going to keep your box actually safe from stuff. GIVE ME A BREAK! you can't configure all the services that users add onto their system from it, keeping THEM from having buffer overflow problems, etc. PLEASE do some research people!
8. total cost of ownership
   You never ONCE explain where all these bizarre linux costs came from. oh, wait, you made them up. you didn't bother taking into account all the free help out there for support, including web sites, community, newsgroups, etc. I hate it when people make up numbers like that to make themselves look good.
9. Linux needs real world proof rather than anecdotal stories
   there are LOADS of linux realworld success stories, not just anecdotes. Feel free to contact any of the customers of VA (and now companies like dell, penguin computing, etc) and talk to them for yourselves. don't take MY word for it.
10. Windows NT 4.0 outperforms linux on common customer workloads
    if you don't bother configuring your linux system you're asking for just as much trouble as if you don't bother configuring your windows NT workstation for testing and benchmarks. I would show you benchmarks to prove things the other way, but microsoft has lots of clauses where they don't allow you to publish benchmarks against them, unlike free software. we're not perfect, and we'll admit to it, but we don't try to hide behind a wall of beaurocracy until we have some tailored benchmarks biased against our competitors before we let some be published.
11. largest file size is 2gb
    oh come on - that's a system dependant limit - on a 64 bit system you don't have this problem (and iirc ufs doesn't display this problem als, even if only read-only ) NOTE: I clarify this due to some emails I received.
12. swap size limited to 128 mb
    that's not true with the systems you even tested with (2.1 removed this limitation iirc) even before that you could have multiple swap partitions

I'm going to stop here, because I just don't have time to go muhc further than this. I hope that microsoft does some more research before they put up more pages like this one. If anyone from microsoft is reading this, and wants to explicitly give me the rights to put up benchmarks of their OS, fileserving, webserving, SQL server, etc... I'll beat you at your own game. If not, then you're just lying chicken behind the same wall of crap you always do.

JC heeft SPEC benchmarks van een 700MHz Athlon Ultra te pakken weten te krijgen:

Wow, thanks to idiot at Ace's, we have now found Fred Weber's uPF presentation. Credit to this SI poster for being the first to show me the spec scores from that presentation:

CPU SPECint SPECfp K7-700, 266MHz bus 37 42

WOW! This is astonishing! Mind you, these are not official results (they're projected values with optimizations), but it's notable that the Athlon Ultra 700 seems to be well over twice the speed of the fastest current Xeon in specfp and quite a bit faster in specint (around 50% higher). Idiot is mentioning that the specint is "#3 in the world", matching the 21264-667. [break] Dit is inderdaad heel erg wow! Vergelijk dat met de c't SPEC benchmarks:[/break]

CPU SPECint SPECfp Pentium III 600MHz 18,9 12,4 AMD Athlon 600MHz 20,7 17,5 Alpha 21264 500MHz 18,1 22,8 Alpha 21264 500MHz Compaq Fortran compiler 39,7 AMD K6-III 450MHz 14,7 7,43 Intel Celeron 500MHz 14,3 9,75 Motorola PowerPC 750 (G3) 400MHz 14,8 11,1

AMD heeft z'n doelstelling wat betreft de oktober verkopen van de Athlon moeten bijstellen van 4 miljoen naar 2 miljoen units. Oorzaak is de slechte leverbaarheid van Slot-A moederborden. Hier de info van Ace's Hardware:

CTech Taiwan is reporting that AMD will sell 2 million K7s short of its targeted sales this year, reason being that there is a shortage of motherboards despite of all the K7s AMD has stocked up. AMD Far-East Region Production Manager is quoted as saying that AMD currently has more than 3 or 4 million K7 in stocks. The original target shipment for Oct is 4 million K7s has been reduced to 2 million because of the earthquake in Taiwan hampering production of K7 motherboards. The Production Manager is also quoted as saying that AMD is hoping to sell more than 10 millions of K7s by the end of the year. Now the numbers are a little far-fetched if you ask me, but we shall all get an accurate answer to that hopefully after tomorrow's Conference Call.

3D Soundsurge heeft een artikeltje gepost over DirectSound3D / EAX en A3D2.0 performance in de Unreal Tournament Demo. Uit de benchmarks blijkt dat de impact van A3D2.0 wavetracing op de framerate behoorlijk heftig is, rond de 30%:

The other thing the Vortex 2 results on both Computer A and B brings to light can be viewed as either disturbing or reason to be optimistic. On one side, the performance impact of wavetracing in Heretic II, which sounds great, is now almost non-existent after originally starting out in the order of 25 percent slower than direct sound (with no 3d). On the other side we have the older hybrid wavetracing implementation of Half-Life (hybrid because Half-Life has its own software reverb engine). It started out at around a 31 percent performance impact vs direct sound and on original Vortex 2 boards has been improved by a few percent to around 28 percent. Half-Life as noted is where the SQ2500 showed the biggest performance gains to date with wavetracing having only around a 17 percent hit on performance compared to direct sound.

Right now the Unreal Tournament impact is in the order of Half-Life on the SuperQuad for both original and revised Vortex 2 boards. What we must hope for is that Unreal Tournament’s A3D 2.0 performance impact, which like Half-Life has its own software reverb engine, can be optimized to similar impacts that we see in Heretic II.

The Tech Zone heeft een artikeltje neer geplant waarin ze beschrijven hoe je een draagbare rack-mounting kast kunt maken:

I found my rack-mount case on Ebay (just FYI, I do not work for Ebay). I also picked up a 4 space (just enough for the computer) SKB rotomolded music equipment carrying case to install the rack-mount case into. This isn’t necessary, but the SKB cases are lightweight, very durable, easily carried, have rubber feet on 2 sides, are stackable, and provide some moisture resistance. You can’t put this case in the bathtub, but if you want to carry the case to your friends in the rain, you should have no worries.

Errug handig voor LAN & demo parties!

Voor wie nog niet genoeg kan krijgen voor het altijd terugkerende debat over welke game platform beter is (als daar al een eenduidend antwoord op gegeven kan worden, wat volgens mij niet het geval is), heeft 3D Alpha een consoles vs PC's artikeltje geschreven. Hier een stukkie:

A good number of factors seem to be the criteria for determining a love for consoles or PCs. Chances are, if you've played games on a console since Nintendo, you're probably a console man who had very little or no love for a PC at all. Or, if you've been messing around with computers since the 286, you probably find more productivity in PCs than you would in consoles. Those are quite obvious statements, but being as obvious as they are, they do say something about how the gaming community is divided between PC games and Console games.

Hierna volgt een opsomming van nadelen en voordelen van de PC en console (gelukkig geen fundamentalistisch geschreeuw enzo). Check 't hier.

Onze vrienden van Hardware One hebben hun Creative Labs Annihilator review geupdate met een aantal nieuwe benchmarks (Quake III met bilinear en trilinear filtering, nVidia Tree demo en Exercizer). Hier de Q3Test resultaten:

Opvallend om te zien hoe de framerate lineair toeneemt met de memory clock (die DDR SDRAM versies van de GeForce gaan errug interessant worden). Anyway, check Hardware One voor de overige benchmarks.

The Register heeft meer info gepost over de optische storage technologie van C3D, waarover ik gisteren ook al een nieuwsposting neerplantte. Gisteren heeft het bedrijf werkende prototypes van een 140Gb ROM schijfje gedemonstreerd:

C3D yesterday showed off new storage technology which it says can offer over 100GB of storage on one disk.

[...] C3D technology works by using the properties of fluorescent incoherent light -- or light in which the waves are out of step. This affords much better resolution than current optical disk technology which is based on the properties of ordinary light.

Effectively, only two reflective layers of information can be packed on existing CD/DVD technology -- any more and the signal to noise ratio ( and hence interference) becomes too great.

With C3D's prototype technology, up to 10 layers of information can be read at once. In future, the sky's the limit as far as FMD and multi-layering are concerned.

Intel lijkt z'n strategie wat betreft de Merced / Itanium gewijzigd te hebben. Tot voor kort richtte Intel deze processor louter op het high-end segment, maar nu lijkt het ding al wat verder naar beneden te komen, op Xeon niveau. The Register serveert de soep:

At last year's Microprocessor Forum, Intel spokesfolks poo-pooed claims that Merced was anything more than a high-end architecture aimed at big league, 64-bit applications, specifically databases and operating systems.

Not anymore, it isn't. This time round, Intel's principal engineer and IA-64 microarchitecture manager, Harsh Sharangpani, clearly positioned Itanium very much in the mainstream "commercial" server and workstation markets, stressing the chip's Epic (Explicitly Parallel Instruction set Computing) architecture's benefits for everything from digital content creation to encryption and security roles for e-commerce and other "Internet applications" (Web browsing?).

In essence, then, Itanium will push right down into the space currently occupied by Intel's Xeon line, possibly even to the extent of supplanting it. And for anyone who thinks that's as far as it goes, Itanium will offer full MMX and Screaming SIMD Extensions (SSE) support in addition to full IA-32 compatibility.

Hmm, deze gewijzigde strategie kan twee oorzaken hebben:

- Intel plast in z'n broekie van de IBM Power4 en Compaq/API Alpha 21364 en denkt dat de Itanium niet snel genoeg is om met deze krachtpatsers te kunnen concurreren (wat aanvankelijk wel de bedoeling was).

- Intel plast in z'n broekie van AMD's 64-bit SledgeHammer plannen en wil de Itanium zo snel mogelijk richting mainstream brengen, omdat de bejaarde P6 core en al zijn varianten (Coppermine, Willamette) het niet meer kunnen trekken in vergelijking met de Athlon.

VIA Technologies heeft voor 14 miljoen dollar aan S3 aandelen gekocht. Interessant om te zien wat ze hier nu weer mee van plan zijn...

Taiwan's VIA Technologies Inc said today it had bought 1.35 million shares or 2.5 percent of U.S. firm S3 Inc for US$14.07 million. VIA Tech bought the S3 stake at US$10.419 per share, a company spokeswoman told Reuters.

``We already have cooperation in product development with S3, and we hope to strengthen ties through the purchase,'' she said.

The spokeswoman declined to elaborate.

In mid-September, VIA Tech said it had signed a final agreement to buy a stake in a subsidiary of U.S. firm Integrated Device Technology Inc for US$36 million. The deal followed VIA's August move to buy the assets of U.S. chip maker National Semiconductor Corp's (NYSE:NSM - news) money-losing Cyrix PC processor business for US$167 million.

Steven heeft gisteren de nieuwe iMacs geïntroduceerd. Hier een rijtje specs, gejat van MacOS Rumors:

- $999, Baby!

- The new iMac, based on the UMA chipset, does indeed sport a 100MHz main bus and uses full-size SDRAM DIMMs, reducing both initial and upgrade costs. The low-end model -- available only in Blueberry -- sports a 350MHz G3 processor with 512k backside L2 cache at half the processor speed, and includes 64MB of RAM as well as 8MB of video SGRAM standard. All models include an ATI RAGE 128 graphics card (!) on a 133MHz AGP bus. With muscle like this on even the low-end model, look out, Wintel....

- The DV model, which includes Firewire, RGB video out, DVD-ROM, a 10GB hard disk, 400MHz G3 CPU, and bundled iMovie video editing softwarte, is $1299. The DV Special Edition, which includes 128MB of RAM and a 13GB disk, is $1499.

- All models have a simple upgrade bay, the slot-loading CD/DVD, and the improved enclosures we showed you last week -- which use no fan at all, simply employing convective cooling through the handle/vent cluster. Silent running, indeed....

- The new Harmon-Kardon sound system sounds great -- but the Subwoofer is an optional USB peripheral. One very interesting feature of the USB Subwoofer is that as soon as it is plugged into either of the C2's dual USB busses, the Mac OS' sound settings automatically reconfigure to transfer low frequencies (bass) to the subwoofer. Very slick!

Ook The Register rapporteerd over de AMD presentaties op het Microprocessor Forum. Duidelijk wordt dat AMD nu serieus mee wil gaan spelen in het high-end segment, met snellere, SMP compatible versies van de Athlon, een nieuwe LDB I/O bus en de 64-bit SledgeHammer processor:

Central to Athlon's extended multi-processor role will be what AMD is calling the Lightning Data Transport (LDT), which was designed to provide a single, unified connection mechanism linking processor and North Bridge to multiple bus technologies -- PCI, System I/O (the combination of the NG I/O and Future I/O initiatives), etc. -- but neatly also serves as a multi-processor communication channel.

AMD will offer the new Athlon in a dual-CPU module which connects both processors to the North Bridge chipset (via an Alpha EV6 bus) and from thence to the AGP graphics card and the system DRAM. Plug four of these together and -- bingo -- you have an eight-way MP system. LDT is a point-to-point interconnect providing a throughput of up to 6.4GBps each way and a channel width of 8, 16 or 32 bits.

The 'Athlon Xeon' and LDT should ship sometime in 2000, the company said. Later, AMD will introduce its x86-64 ISA which will involve a further modification of the processor -- assuming it's not being held back for K8, of course -- that is essentially a 64-bit version of its current x86 implementation but allows the processor to behave as a 32-bit chip, when it's running existing 32-bit apps.

x84-64 will also sport some additional "specialised operations" to the x86 set and add what AMD calls "technical floating point instructions", which appear extend 3D Now! to something closer to Motorola's AltiVec technology than Intel's SSE.

Zie The Register voor de complete info.

Motorola heeft op het Microprocessor Forum plannen gepresenteerd die de G4 processor beter geschikt moet maken voor hogere kloksnelheden. De G4 dobbert op dit moment namelijk nog op 450MHz, terwijl AMD (en binnenkort ook Intel) al vrolijk op 700MHz rondfietsen. Hier de info, grof geleend van The Register:

To date, the PowerPC's clock speed has been limited by the size of its processing pipeline. The current G4 has a four-stage pipeline -- the path of an instruction through the chip -- and that's not enough to keep a 700MHz CPU fed with instructions and data. Motorola could simply up the frequency, but the processor would have so much idle time that the speed advantage would be lost.

The second-generation G4 increases the pipeline to seven stages, and to counter the reduction in the number of instructions a processor can handle per second inherent in long pipeline, the company has increased the number of instruction processing units in the chip.

According to Naras Iyengar, one of the chip's design team leaders, the new G4 will feature two extra integer units, taking the total to four, in addition to the existing floating-point unit and four AltiVec units. The AltiVec system has been enhanced to handle two instructions simultaneously, each being automatically passed to the relevant unit according to the type of data involved.

Following a clear industry trend, the new G4 brings the L2 cache into the chip itself to allow it to operate at the same speed as the core. The L1 caches remain the same size -- 32K instruction, 32K data -- while the on-die L2 will be 256K, connected to the L1 via a fast 256-bit wide datapath (up from the 7400's 64-bit path). Like AMD's K6-III chip, the new G4 will also support a third layer of cache between the CPU and the main memory bank, in backside configuration. It will support up to 2MB of this L3 cache.

IBM heeft z'n Power4 processor en de daaraan hangende architectuur op het Microprocessor Forum gepresenteerd. IBM gaat heel erg duidelijk op de brute toer (en dan bedoel ik niet zomaar 'bruut', maar 'wreed bruut'): wat denk je van een 170 miljoen transistors tellende dual-core CPU met tri-level cache architectuur, een L2 cache bandwidth van 100GB/s en kloksnelheden van boven de 1GHz? *kwijl*

Bandwidth is clearly all in IBM's eyes, and Kahle reeled off a stack of statistics such as the die-L3 line has a data throughput rate of over 40GBps, while the core-L2 line can handle over 100GBps. Each Power4 contains a chip-to-chip communications module to enhance multi-processing systems, and these modules operate at over 35GBps.

The point here is that the chip and its architecture have been designed from the ground up with server roles in mind -- as Power development head Charles Moore said when he introduced the chip at last year's Microprocessor Forum -- and servers are primarily about moving information from one place to another and that, in turn is primarily about bandwidth, especially in Internet roles where usage tends to fluctuate rapidly, with frequent high bandwidth demand peaks followed by periods of relatively low usage.

As Kahle put it: "Our design philosophy has been to get the right data to the right place at the right time."

Hmm, dat is me wel duidelijk ja . Dit is overduidelijk geen huis-tuin-en-internet CPU'tje die je in iMac-achtige doosjes aan zult treffen... Check The Register voor het complete verhaaltje.

Volgens Intel (en waarom moeten we die geloven ) zal de Coppermine 25% sneller zijn dan de huidige PIII's. Hier de info van The Register:

Coppermine will feature 256K of on-board L2 cache and despite retaining the same P6 core that Intel has been using for the last five-odd years, operate at around 25 per cent faster than the current, Deschutes Pentium III operating on the same 133MHz front-side bus that Coppermine uses.

According to Wilson, the improvement is due to the speed gains of bringing the L2 cache onto the die and upping the cache bandwidth, and increasing the chip's buffers to accelerate the flow of data through the processor.

Gaat er bij mij toch niet helemaal in dat een busspeed van 133MHz en 256Kb full-speed L2 cache opeens een prestatie verbetering van 25% teweeg kan brengen...

AMD heeft een press release gepost met meer details over de 'Lightning Data Bus', die gebruikt zal worden voor toekomstige Athlon (en SledgeHammer?) systemen. De maximale bandwidth van de LDT I/O bus is een brute 6,4Gig per sec, 24 keer zo veel als de (64-bit) PCI bus:

AMD's system bus of the future, called Lightning Data Transport (LDT) was also discussed at the forum. LDT provides a data link that is internal to the PC, providing up to 20x increase in bandwidth for I/O, co-processing and multi-processing functions. This increased I/O performance and bandwidth will improve overall system performance for AMD Athlon™ processor-based workstations, servers and personal computers.

"API plans to offer a two-way chipset and bridge chips during the second half of 2000 that incorporate a LDT interconnect, enabling industry leading I/O bandwidth and performance," stated Gerry Talbot, executive vice president, chief operating officer and chief technical officer, for API. "Pushing the envelope of I/O bandwidth will greatly improve router and web server capabilities, allowing technology to keep pace with the demands of markets like the Internet and telecommunications."

"LDT will compliment HotRail's smart switched fabric SMP core logic for servers based on AMD Athlon processors by providing increased scalability in the server I/O arena," said Rick Shriner, president and chief executive officer, HotRail Inc. "Together HotRail and AMD will drive servers to new levels of performance."

[...] About Lightning Data Transport™

LDT is an internal chip-to-chip interconnect that provides much greater bandwidth for I/O. It can achieve a bandwidth of up to 6.4 GB/sec per connection. Compared to current system interconnects that run up to 266 MB/sec, LDT provides more than a 20x increase in bandwidth. LDT complements externally visible bus standards such as PCI or SIO, and provides a very fast connection to both. LDT is the connection that can provide the bandwidth the new SIO standard needs to communicate with a server.

Lost Circuits heeft een review gepost van de Asus K7M, een Athlon moederplank gebouwd rondom de AMD Irongate North Bridge en de VIA Super I/O South Bridge. Het bord heeft busspeeds van 90, 95, 100 - 125 (in 1MHz stappen), 133, 140 en 150MHz en zou daarom ideaal kunnen zijn voor overclocking. Desondanks bleken de resultaten enigzins teleurstellend, want het plankje bleek onwillig te zijn bij busspeeds hoger dan 114MHz:

In our hands, overclocking was possible to a very limited degree, that is up to 114 MHz, beyond which the system would simply refuse to boot. This happened independent of the CPU used (Athlon 500 / 2 CPUs and Athlon 550 / 1 CPU), however, at least we do have the chance to compare the performance at identical clock speed using the default 100 MHz bus as opposed to the 110 MHz bus with a 5.5 x and 5 x multiplier, respectively.

Techweb heeft een erg lekker artikeltje over de plannen die AMD gisteren op het Microprocessor Forum heeft gepresenteerd, waaronder de 'Lightning Data Transport' (alweer een koele naam ) architectuur die voor de SledgeHammer ontwikkelt wordt:

AMD will debut its X86-64 architecture Tuesday at the Microprocessor Forum in San Jose, complemented by its Lightning Data Transport (LDT) I/O architecture.

[...] Following years of catastrophic manufacturing flubs, AMD has set out to ensure that manufacturing the new chip will be the easiest piece of the puzzle. Adding 64-bit capabilities increases the die size only 5 percent over the approximately 104 sq. mm that a 32-bit Athlon requires, using a 0.18-micron process, Lapinski said. While AMD's timetable still calls for 0.18-micron manufacturing to begin at Fab 30 in Dresden, Germany, later this quarter, the company on Monday showcased an 800-MHz Athlon running on the new process.

More importantly, AMD said that the combination of a small die size and its 0.18-micron process will allow the company to pack more than one 64-bit X86 microprocessor on a single die [wow, kick ass! ]. That's important, given that X86 integer instruction performance is closing in on RISC chips, Lapinski said. Through IEEE-compliant, triple-operand, double-precision floating-point instructions that AMD is designing for the new architecture, the company hopes to eliminate the floating-point advantage of RISC chips as well, he said.

[...] When pairing more than one microprocessor on-chip, AMD will use undisclosed custom logic to manage the infrastructure. Off-chip, however, AMD has designed the custom LDT bus for I/O and coprocessor chips. The LDT is a bidirectional bus, either 8, 16, or 32 bits wide; the bit width is negotiable at the device's initialization. Data passes through multiple logical channels in up to eight links or bridges, which can be connected to several daisy-chained devices.

Meer info in dit artikel:
- K6-2 Plus, 0,18micron, 128 of 256Kb on-die L2 cache, volgend jaar
- Geen Slot-B versies van de Athlon (Ultra)
- Alpha Processor / HotRail en Reliance Computing chipset suppliers voor SMP Athlons?
- Athlon Ultra volgend jaar, 1 of 2MB full-speed L2 cache, 266MHz bus
- Nieuwe K6 versies zonder L2 cache?

EETimes bericht dat Micron Technologies op het Microprocessor Forum werkende Double Data Rate SDRAM reepjes heeft gedemonstreerd. Naar eigen zeggen zou men al begin volgend jaar in staat zijn om in volume te kunnen leveren:

The DDR chips are currently available in 64 M-bit densities, although the first 128 M-bit samples are currently being produced at Micron's fab. Klein said the most likely early applications for the technology will be in the high-performance workstation and server markets.

Although the company expects to hit volume production within the next few months, the current designs are still showing bugs. Klein said the DDR chips still suffer from reference voltage leakage and asymmetrical data eyes, which are lowering operating speeds. The chips run at 133 MHz, but those speeds are not yet stable.

"We expect to have new silicon any day which should eliminate most of those problems," Klein said. "These chips are still first silicon, and this is excellent performance for first silicon." [break] Meer info over het DDR SDRAM gebeuren vond ik bij MaximumPC. Lijkt erop dat de benaming van de DDR SDRAMs is verandert: die heten nu PC2100 ipv PC266 (voor 133MHz DIMMetjes), wat verwijst naar de theoretische bandwidth (133 * 2 * (64/8) = 2,1Gbyte/sec).

Marketing technisch is dit uiteraard erg slim, aangezien PC266 nogal karig staat vergeleken met PC800 Rambus. Helemaal eerlijk is het niet, want PC800 RDRAM heeft een bandwidth van 1,6GByte/sec (maar heeft hierbij wel een hogere latency dan DDR SDRAM). Anyway, hier wat pruttel uit dat nieuwsartikel van MaximumPC: [/break] Micron Technology, not to be confused with PC maker Micron Electronics, showed off the first working demonstration of Double Data Rate memory on Monday.

Dubbed PC2100 for the 2.1GB of bandwidth it offers, the SDRAM-based technology is poised to go head-to-head with Direct RDRAM or Rambus memory in 2000.

The technology demonstrations, made privately to press and analysts, showed that DDR is indeed alive and kicking. Micron Technology is promoting DDR as a practical and cost-efficient alternative to Direct RDRAM.

MaximumPC heeft ook wat info over de Samurai dual Slot-1 North Bridge, die Micron speciaal voor dit DDR SDRAM gebeuren heeft ontwikkeld. Check 't hier.