Hartware.de heeft een preview van de Videologic Neon 250 32Mb AGP online gezet. De tekst is in het duits en zal voor de meeste mensen daarom waarschijnlijk niet al relaxed lezen. Wel hebben ze een paar benchmark grafiekjes, helaas zonder vergelijking met andere videokaarten. In onderstaand tabelletje heb ik de Neon 250 Q2 Crusher scores gecombineerd met de TNT2 en V3 benchmarks van Tom (allemaal getest op PIII-500). Geen wetenschappelijk verantwoorde vergelijking, maar beter dan niets...

Als dit klopt dan is de Neon bij 16-bit OpenGL rendering zeker geen slechte kaart. Meer benchmarks (Q3Test, Kingpin, Expendable en 3DMark99) vind je in de preview van Hartware.

Volgens speurwerk van The Register heeft Compaq al werkende samples van koper-interconnect Alpha's:

Conversations in newsgroup COMP.ARCH have revealed that samples of Alphas using copper and SOI technology are already up and running.

The reports come after yesterday's claim by Intel's John Miner that its IA-64 processors will outperform any Risc platform.

According to a user on the thread, he is already achieving clock speeds of 833MHz using Tru64 Unix on a UP2000 motherboard.

He said: "So what would you expect from a sped-up 0.18 um copper SOI part with much faster cache bandwidth than today (a 667 MHz EV-67 used in recent XP1000 SPEC results runs a 1/3 speed 222 MHz cache; what do you get when you increase the clock rat 2.2 times, and cache throughput 4.5 times - and double the cache size (and improve main mem bus speed?"

API demoed a 1GHz Alpha at PC Expo in summer running at 1GHz at room temperature. Compaq has also recently received SOI .18 micron parts from IBM.

Rambus aandelen zijn 11% gedaald nadat Intel gisteren bekend maakte dat het voorlopig toch PC133 SDRAM zal ondersteunen. Hier heb je een hap uit een nieuwsbericht van Techweb:

Shares of chip technology company Rambus fell 11 percent in morning trading Thursday, extending losses sparked Wednesday by Intel's decision to slow its planned adoption of a Rambus technology, analysts said.

[...] Analysts say the drop is a direct response to Intel's announcement that it will roll out another generation of an existing memory chipset, known as PC-133 DRAM , before shifting to a new family based on Rambus technology. Intel, the world's leading maker of semiconductors, is Rambus' main customer.

Intel zal in de eerste helft van 2000 een nieuw Xeon model met geïntegreerde L2 cache introduceren. Dankzij lagere produktiekosten zal deze processor zich ook op het low-end marktsegment kunnen richten:

Intel Corp. plans to push its Pentium III Xeon processor into the low end next year.

In the first half of 2000, the chip maker plan to introduce two 700+MHz Xeon procesor, one with 2MB and the other wit 1MB of integrated Level 2 cache.

"The integrated cache will allow us to extend Xeon into the more affordable low end", said John Miner, vice president and general manager of the comapany's communications products group, during his keynote speech here Wednesday at the Intel Developers Forum.

Fiets richting ZDNet voor het complete verhaaltje.

Ik lees bij AGN Hardware dat de nieuwe PowerMac G4 support zal krijgen van 3dfx. Het ziet er dus naar uit dat de Mac een T-Buffer zal gaan krijgen:

3dfx Interactive(R) Inc. (Nasdaq: TDFX) today announced that its 3dfx graphics technology will be available for Apple's new Power Mac G4. The addition of 3dfx's award-winning Voodoo graphics technology to the new Power Mac G4 creates a new level of graphics excellence and a truly immersive graphics experience for the user. 3dfx will also bring its recently announced T-buffer technology to the new Power Mac G4. T-buffer technology enables cinematic effects never before available on consumer level computers at real-time frame rates essential to attaining and experiencing true-to-life imagery. New Mac products from 3dfx that further capitalize on the power of the Power Mac G4 are planned in the near future.

"The new Power Mac G4 brings supercomputer speeds to personal computing," said Phil Schiller, Apple's vice president of worldwide product marketing. "The commitment by 3dfx to the new Power Mac G4 is a testament to its strength as a graphical powerhouse, particularly for processor-intensive multimedia applications."

"The Power Mac G4 delivers awesome speed and power that the gaming community and others who demand high performance are looking for," said Bryan Speece, director of Macintosh business development at 3dfx Interactive Inc. "The new Power Mac G4 coupled with our graphics technology brings us closer to delivering complete immersion on the computer. The efforts of Apple and 3dfx to elevate the level of graphics performance on the platform will be of great benefit to Mac users everywhere."

3dfx's T-Buffer technology will dramatically improve the rendering capabilities on the Macintosh, bridging the "realism" gap that exists today between motion picture and photographic imagery and computer-generated images. With its T-Buffer technology, 3dfx comes one step closer to fulfilling its mission to bring advanced cinematic effects and realistic imagery to consumers.

The T-Buffer technology enables cinematic effects never before available on consumer level computers at real-time frame rates essential to attaining and experiencing true-to-life imagery. Chief among these are real-time Full-Scene Spatial Anti-Aliasing, which removes the appearance of jagged edges in a rendered scene, and real-time Motion Blur, which dramatically improves the rendered realism for fast-moving objects. The T-Buffer technology also allows for the first time, acceleration of Depth of Field effects commonly used in professional photography and cinematography.

Het persbericht is ook hier te vinden.

Ja luitjes, hier nog wat meer GeForce 256 voor jullie. Thresh's FiringSquad heeft namelijk een verslag geschreven van de press conference van nVidia, waar zij natuurlijk aanwezig waren. Het meeste dat erin staat is niet echt nieuw, maar wat wel interessant is, is het Q&A met nVidia zelf. Zie hier een paar vragen:

[Q] When is the next part coming out? Will it be a new architecture or a clocked up GeForce 256?

[A] NVIDIA is on a 6 month product cycle, so the next big product will be announced at that time. None of the NVIDIA representatives would confirm or deny that the next product (NV15) would be a new architecture or just a clocked up GeForce 256.

[Q] The board supports up to 128MB but what will a typical board have?

[A] Dan Vivoli informed me that most of the first run of GeForce 256 boards will mostly have 32MB, and that 32MB is all you're really going to need. Probably a couple of the OEMs will create some hardcore boards with 64MB or more, but that will drive up costs to an even greater degree.

[Q] What price range are we talking about for the boards?

[A] This is another issue that NVIDIA danced around but understandably so - it's the OEMs who sell the boards and set the prices so NVIDIA is in no position to say how much the boards will cost. Huang did mention that graphics cards consistently increase in price with each generation, so based on that and the fact that TNT2 Ultras retail for $200-$230, and G400 Max's range from $200-$250, we'd say it's safe to assume that GeForce 256s will launch around the $300 mark or even a bit above.

Lees hier verder!

Sharky Extreme heeft z'n Weekly CPU Prices column geupdate met info over de Athlon-700, die naar verwachting begin oktober wordt gereleased!

The Athlon train just continues to roll down the track at the moment as Sharky Extreme has heard from several sources close to AMD this week that October is the target launch month for the next version of the Athlon CPU, a 700MHz model running on the current version's PC-100 SDRAM-based system bus.

The shocking part to us was not necessarily the mention of October itself, but the verification by multiple individuals that the actual ship date resides in "early" October.

AMD has been relentless in their pursuit of raw MHz speed. They seem to be thoroughly enjoying their current ownership of the "fastest x86 CPU in the world" title. How long will it take Intel to even match AMD's fastest parts, let alone exceed them, is unknown.

Currently, sources close to Intel indicate that a P3-733/133MHz Coppermine CPU could appear as early as mid-November which would cut short AMD's speed-King reign, at least in MHz.

Haha! ATi (fabrikant van semi-brakke videokaarten) heeft een belangrijke deal met Dell verloren. Als gevolg daarvan daalde de waarde van ATi's aandelen met 12%. C|Net heeft net nieuws:

ATI Technologies shares fell 12 percent after the No. 1 maker of graphics cards for personal computers lost the bidding for a contract to supply Dell Computer.

ATI shares fell by almost 11 percent to $12 a share after a midday halt in trading. The loss of the contract will cut sales by $10 million a quarter, the company said. Sales in ATI's third quarter ended May 31 totaled $302 million. Despite the lost business, ATI said fourth-quarter sales are "on target" and the outlook for the first quarter of fiscal 2000 is "solid."

"Dell remains a major customer of ATI, particularly in the notebook market," the company said in a statement.

[...]Chang declined to say how much of an impact the lost revenue will have on future earnings. She also wouldn't say whether the company lost the contract to rival Nvidia, as some analysts speculated. Officials at Santa Clara, California-based Nvidia weren't available to comment.

Check voor meer info ook dit artikel van GA-Source.

Een interessante posting bij Ace's Hardware (die chinees verstaan en dan ben je een held ):

According to this Digitimes Taiwan report , VIA aims to increase its production by 300% in year 2001 and intends to overtake Intel in terms of chipset production then. Also, VIA will be doing some sort of joint-venture with Motorola: Motorola will provide VIA with Alpha and Power PC technologies, and help VIA to release chipsets for servers by H2 2000. In exchange, VIA will provide Motorola with AGP 4X, AC97, USB and IEEE 1394 (Firewire) technologies. In the future VIA will further help Motorola with PC133 in exchange for copper technologies. The joint-venture will be limited to only chipsets and no cpus.

It is also reported that VIA is in close cooperation with AMD to come out with future versions of K7 chipsets. AMD's own K7 chipset, Irongate, will begin to cease production by the end of the year and VIA will take over the responsibility of supply K7 chipsets by next year.

ComChip heeft een review gepost van de Samsung SyncMaster 900SL, een 19" shadow mask monitor met 0,25mm dot pitch:

The thing that caught my eye with the 900SL is its boast that it is the shortest depth CDT in the world, at 407mm. Wow, that’s practically the size of my old 14-inch! Going from a desktop w/14-inch monitor to a full tower and the 900SL, I’ll actually be SAVING desk space in my room. For those of you caught in the dilemma of having to give up quality to save space and vice versa, the SyncMaster 900SL is your solution!

3D Spotlight heeft een review van de Tennmax TNT Detonator cooler. Zoals de naam al doet vermoeden moet je deze koeler op je TNT planten. Hoewel je niet veel snelheidswinst zult behalen houdt deze Tennmax de temperatuur wel redelijk naar beneden.

After I installed the Detonator, I promptly went to my overclocking utility to see if I could push the card to its limits. With the standard cooler I could set the Creative Labs TNT2 Ultra board I have to a core speed of 170MHz. The memory was able to go to 210MHz without visual distortion, but the cooler on the core won’t help you very much in getting better speed from the memory.

With the TNT Detonator, I was only able to get a core speed of 173MHz. That’s only 3MHz higher than 170MHz. I felt that this was odd, since I knew I could get more than 3MHz out of it. I then removed the unit from my Creative Labs board and carefully scraped the thermal pad off of the heatsink. After that was done, I took very fine grit sandpaper and smoothed the surface of the heatsink to remove any cuts caused by scrapping off the thermal pad.

After that was done, I put a small layer of thermal grease on the TNT2 core casing and pushed the TennMax Detonator against it to make sure an optimal connection between the heatsink and the TNT2 was established. I then pushed the pins back in and placed the Creative Labs board back in my system. I was surprised to see that this combination worked very well. I was able to get the core of the TNT2 up to 176MHz. This is 6MHz from the 170MHz I had with Creative Labs default unit and 3MHz higher than with the thermal pad.

Storage Review heeft er weer een review bij, ditmaal van de 27.2Gb grote Maxtor DiamondMax Plus 6800. Deze 7200rpm HD heeft 6.8gigs per platter en is voorzien van de ATA-66 interface. De review vind je hier.

Maxtor's entry in this latest battle is the DiamondMax Plus 6800. As its name indicates, this newest Maxtor is a 7200rpm unit featuring 6.8 gigs per platter. Its seek time is specified as "less than nine milliseconds." Unlike its predecessor, the Plus 5120, the Plus 6800 was designed from the ground up to go toe-to-toe with Western Digital's Expert and IBM's Deskstar GXP series. The drive features an ATA-66 interface and a two megabyte buffer. Judging from Maxtor's press releases and white papers issued on the drive, the manufacturer takes great pride in the electronics of the Plus 6800. This newest Maxtor incorporates the company's "DualWave" processor, theoretically increasing performance and reliability through the use of fewer parts. A look at the effects of a simple cache increase (Maxtor DiamondMax Plus 5120, OEM vs. Retail) proved that increasing cache size alone provides little if any increase in performance. Maxtor takes great pains to emphasize, however, that the DualWave controller was designed with two megabytes of buffer in mind. A three-year warranty protects the drive.

Over bij The Tech Zone is een vrij grote review te lezen van de Matrox G400 Max. Hier heb je wat over de 3D performance van deze kaart, hier staat de rest.

The Millennium G400 MAX can only be described as one hell of a Direct3D performer. Matrox have continued to build upon the excellent 3D image quality seen in the G200's but this time with a huge amount of horsepower behind them in both 16 and 32-bit color depths. The benchmark results from the Millennium G400 MAX on a PII 450 powered test system clearly show that it is being held up by the system up until 1280x1024 when the fill rate limitations start to kick in. From these results it looks like the G400 MAX will be very scaleable and will need a PIII or K7 to really make the most of it.

...

No matter how good the Direct3D performance of the G400 MAX is a large proportion of the gaming community will judge it solely on it's OpenGL performance. During testing the G400 OpenGL ICD has proved to be extremely stable. It has also produced some extremely impressive benchmarks, maybe somewhat disappointing a low resolutions in the light of results from other cards but it's a real contender at high resolutions and 32-bit color. The G400 MAX does give the feeling (especially when looking at the Direct3D benchmarks) that there is much more speed to come the OpenGL ICD, remember it is still a beta at this point, I saw an 11fps increase at 1600x1200 from the initial driver build I was using.

Sharky Extreme heeft een interview gehad met Bitboys Oy, het Finse bedrijf achter de Glaze 3D accelerator. Hier heb je alvast 2 vragen, klik hier voor de rest.

SE: 3dfx has hyped up their T-Buffer technology to the roof, now you are also saying that you will support the same features. Is this just a counter move to show 3dfx s new technology to be less than what people expect or were these features planned from the start?

Mika Toumi: 3dfx has now specified that T-Buffer is just an ordinary Accumulation buffer with some modifications. We wanted to point out that we support accumulation buffer. Accumulation buffer has been part of the OpenGL spec for years.

SE: What is the target price once the Glaze 3D retails? When do you expect it to be available on the retail market?

Mika Toumi: We expect Glaze3D to be in retail market Q1/2000. And the final board price is hard to estimate exactly at this point, but I can tell you that it will compete with 3dfx & nVidia in pricing also.

Bij The Adrenaline Vault is een preview te vinden van de strategy game Earth 2150. Deze game dreigt erg goed te worden, de graphics zien er in 3D in ieder geval erg strak uit.

Perhaps the game's most notable feature is that, like Warzone 2100, it abandons the RTS tradition of a top-down view in favor of a hardware-accelerated 3D environment. The landscape in Earth 2150 is satisfyingly three-dimensional with hills, valleys and lakes faithfully modeled and rendered in perspective -- this minor difference gives the whole game a much less stiff feel and appearance than other RTS romps that rely on orthogonal projections (think classic SimCity). Camera placement is accomplished through mouse movement, and panning, zooming and rotating the camera is so fluid and intuitive that it is an enjoyable activity in and of itself. When panning the camera, it swoops and dives cruise missile-like over undulating terrain. A three-window mode is even available, giving you one large and two small independent viewports, much like a CAD system. The visuals themselves are simply stunning. Weather and time of day add dramatic touches, such as lengthening shadows toward twilight, sudden rainsqualls, and snow softly accumulating on structures and trees. The first time a squad of my ED tanks switched on their work lights after sunset, I was struck by an intense feeling of déjà vu -- the resemblance to vehicles as seen from an airplane at night is uncanny.

Om jezelf alvast helemaal in de stemming te brengen heeft Tweak3D een drietal wallpapers online gegooid. Hier heb je ze in het klein, klik hier om de zips te downloaden. (Jpg's zijn in 1024 x 768)

De japanners zijn weer bezig geweest, lees ik bij Ace's Hardware. Door een ander klokkristal op een Athlon mobo te prikken hebben ze de Athlon middels een verhoogde busspeed over weten te klokken.

Nu is het alleen nog wachten op een Athlon plank waarmee dit met behulp van jumpers of vanuit het BIOS mogelijk is.

Mirabilis heef alwéér een nieuwe beta versie van ICQ 99B gepost. Versie 3.19 Build 2569 kun je hier naar beneden halen.

Asus heeft een nieuwe BIOSsen voor z'n P2B moederplanken gereleased. De nieuwe versies vind je op Asus' FTP site.

Intel is nu toch eindelijk overstag en zal support voor PC133 SDRAM aan z'n (i820?) chipsets toevoegen. EBN Online heeft de info:

Intel Corp. late today made it official: the company will add support of PC133 SDRAM to its chipset line, citing a near-term lack of availability of the Direct Rambus DRAM architecture it has been promoting for nearly three years.

[...] "It will be widely available in the market and the higher speed essentially comes at no extra premium, so why not use it?" MacWilliams said. "Basically it is lack of Direct RDRAM availability. We see PC133 as an interim memory co-existing with Direct Rambus through 2001," he added [break] De plannen van Intel om een eigen PC133 alternatief te ontwikkelen gaan dus waarschijnlijk niet door: [/break] Intel also confirmed that it will adopt the current PC133 industry standard, scotching earlier reports by EBN that the company might develop a slightly different 133-MHz specification. The company cautioned, however, that it is studying whether any interface changes may be needed to enable it to bring chipsets supporting the interface to market. [break] En om het nog leuker te maken zijn ze misschien ook nog van plan om DDR SDRAM te ondersteunen: [/break] However, Intel is seriously considering adding DDR support in its server chipsets, "based on what our OEMs are telling us," he added. Intel's DDR server interest may also be spurred by the disclosure earlier today by Reliance Computer Corp., San Jose, that most major server OEMs have endorsed its third-party DDR SDRAM-enabled chipset.

Hercules is bankroet. Vaarwel ...

A Greek tragedy it isn't, but Hercules Computer Technology's last chapter will be liquidation of all assets to pay creditors, customers, and back taxes. The graphics board pioneer that once held a herculean grip on the video graphics card market filed for Chapter 7 bankruptcy protection last week, ending a 17-year run for the firm.

[...] As of Tuesday, the company's Web site was operational and accepting orders. But those orders will not be honored, credit card accounts will not be debited, and invoices will be canceled, according to Joe Lau, former Hercules marketing manager. Lau says that "thousands" more orders for Hercules graphics boards received in the weeks leading up to the company's shutdown will likely be voided.

Lau attributed cutthroat competition, price erosion, and reduced product life cycles as contributing to Hercules' demise. Other sources at the Fremont, California, firm pointed to a management change that included abrupt resignations of the company's chief executive officer and vice president of engineering.

Check PC World voor het volledige overlijdens bericht.

Wazige Japanse site is terug met een aantal foto's van een Coppermine...

...en een Merced sample:

Meer pics vind je hier en daar.

Ace's Hardware heeft een aantal rauwe PowerPC G4, Pentium III en Athlon SPECint95 en SPECfp95 benchmarks van het net geraapt:

One should look at the PPC architecture page first to see what kind philosophy PPC architects have in mind. Then it time to jump to the G4 main page at motorola, where one get CPU die porn and some SPEC95 numbers(21.4 specINT, 20.4, specFP at 450MHz, probabily with 2MB 1/2 speed cache). From the Spec scores G4 is about 0.04756 SPECint95 per MHz, and 0.04533 SPECfp95 per MHz, while K7/Athlon (at 550MHz thus longer memory latency) rates at 0.04564 SPECint95 per MHz, and 0.03745 SPECfp95 per MHz, and PIII 500 XEON with 2MB L2 clocks in at 0.045 SPECint95 per MHz, and with 512KB L2 0.0318 SPECfp95 per MHz. This shows that G4 is about on par with top x86 CPUs in integer per clock while beating them in floating point, however due to short pipeline it suffers from a lower clock rate, much like K6 series. However consider the price, without L2 the 450MHz G4 cost 355 USD, $100 over the K7 500MHz version which does include L2, thus from $$$$ per performance point without enough Altivec softwares, the G4 is still not as attractive as the x86 offerings.

In een tabelletje ziet dat er zo uit:

Over de G4 kan ik weinig zeggen, maar de floating point scores van de Athlon lijken me t.o.v. de Xeon nog vrij laag. In de praktijk is de Athlon 50% sneller in pure floating point performance. Een Xeon zal daar met z'n 2Mb L2 cache weinig aan kunnen veranderen.