PC Paradox heeft een artikeltje in elkaar gedraaid over nieuwe technologiën waarmee het mogelijk moet worden om te netwerken via het bestaande stroomnet, tegen snelheden van ver boven de 1GBit per seconde. Jammie!

One company, MediaFusion, is hoping to take advantage of those powerlines that run into all of our house to provide a very high speed connection using existing infrastructure, very much like DSL uses our POTL (plain old telephone lines) and cable. The advantage they offer is that instead of talking about megabits MediaFusion is thinking gigabits and even exobits. This blows away cable at 30 Mb/s or even T1's and DS3's (depending on whether they will offer both a high speed download and upload). Thus the Internet 2 with their fancy fiberoptics could be replaced by those huge powerlines we always hear humming after it rains. Another advantage that beats out cable and DSL is that you do not have to a certain distance from the router or have a special type of cable. The one problem lies around how to deal with noise, those copper lines go through various routers along the way and keeping the packetloss at a minimum has proven a definite challenge.

Lam Research Corp - een of andere bedrijf dat handelt in chip-fabricage spullen - heeft vandaag bekend gemaakt dat AMD een order heeft geplaatst voor 0,18micron "Teres(TM) integrated chemical mechanical planarization (CMP) and post-CMP cleaning system". Whatever dat ook mag zijn, hopelijk gaat het voor ons betekenen dat AMD meer, snellere en goedkopere Athlons kan gaan bouwen:

Lam Research Corporation (Nasdaq:LRCX), a leading supplier of wafer fabrication equipment to the worldwide semiconductor industry, today announced a multi-system order from AMD for its Teres(TM) integrated chemical mechanical planarization (CMP) and post-CMP cleaning system.

AMD's leading-edge Fab 25 in Austin, Texas, is qualifying the initial system for 0.18 micron shallow trench isolation (STI) for AMD Athlon(TM) processor production. Subsequent systems will add capacity at AMD's Fab 25 and will be used for advanced CMP applications in AMD's Submicron Development Center in Sunnyvale, California.

AMD selected Lam's Teres for 0.18 micron STI direct-polish production and next-generation development because of its superior process performance, flexibility, and control as well as its industry-leading manufacturability. Fab 25's initial Teres system achieved a mean time between failure (MTBF) exceeding 200 hours during its evaluation and after its production release. AMD achieved excellent within-die and across-wafer uniformity as well as very low defect densities.

HardOCP heeft een review gepost van de Abit's BP6 dual socket370 wonderplank. HardOCP heeft geen benchmarks, maar wel een verzameling toffe pics en handige praktische info. Bijv. dit stukje over de stabiliteit & overklokbaarheid in een dual CPU config: twee C300A's die in hun uppie 558MHz doen, draaien samen niet stabiel op die snelheid:

This is where we encountered the problem with the FSB speeds. While 500MHz is solid as a rock, our CPU's when mounted together, would NOT do 558MHz as they would do by themselves. Bummer. This is why I made the statement on page 1 about the FSB's being being a little too narrow for us. [break]Als het bord echter eenmaal aangekachelt is, dan draait-ie ook meteen rock-solid:[/break] What we did do was hammer the ABIT BP6 as hard as we could under both NT and 98. We ran multiple programs that stressed the CPUs including games and such and also ran a battery of the Prime95 Torture Test for what amounted to a combined period of over 10 days!

To be quite honest with you, I was surprised. I did not think that the most inexpensive dual mainboard on the planet would deliver such solid performance.

AGN Hardware heeft een korte review van de FX Racing Wheel. Dit is een "force feedback hand held controller" van Interact. Je kent ze wel:

Now that the FX Racing Wheel is hooked up, it was time to play a few games using the controller. I have to admit that I was quite excited, thinking that I would not longer have to put a big wheel on my desktop. Need for Speed High Stakes is in the CD-Rom and I am ready to give the new controller a try. Picked the course an away I go, armed with the new FX Racing Wheel. The first thing I noticed is something is wrong, since I constantly remained in last place. The responsiveness from the controller just did not seem to be working right. I would turn but it did not seem to turn the way that I wanted it, resulting it in Jeremy being in last place.

The throttle on the controller is a bit difficult to use, since you have to pull to go forward and push to go back. Although this is not a problem for short gaming periods, longer ones create cramps in your finger. The buttons on the controller are also quite difficult to use, causing you to dig all over trying to find the right one to use.

Hm, ik hou het toch bij een traditionele gamepad.

...dat is de naam van de nieuwe column van Alex Ross (aka "Sharky" - van Sharky Extreme) de laatste ontwikkelingen in de hardware en games business bespreekt (zoiets dus als de Monday blurb van Tom, die hij inmiddels al weer afgeschaft lijkt te hebben). In de eerste aflevering van "Sharky's Private Eye" vind je o.a. info over T&L (Transformations & Lighting, oftewel onboard geometry acceleratie). nVidia zal dit met de NV10 als eerste introduceren:

This makes it rather interesting for lower-end system users. With hardware T & L, you might well be able to see similar quality visuals at a similar speed (in terms of fps) on a 'lowly' Pentium II 300Mhz. Having said that, when we see a REAL game with T&L (such as Black & White) we'll let you know. Don't expect to see many anytime soon though, after having spoken with game developers on campus it seems that games featuring T&L won't be seeing the light of day until late Q1 by our own guess. With the launch of NV10 (NVIDIA's T&L part) scheduled for the later half of this year, there will still probably be a couple of 'sponsored' titles by Christmas but it won't be 'the norm' as such. In order for game developers to hard-code in transformation and lighting, they would need to start from DirectX 7.0 ground-up. The trouble is that most games being developed for this Xmas have already been in the development cycle for a year and thus been coded in DirectX6.0. NVIDIA can thank Microsoft's DirectX7 schedule for this one… [break]Dit is dus errug leuk voor mensen met een 'trage' PC. Intel is minder gelukkig met deze ontwikkelingen. Het kopen van een een veel-te-dure PIII zou nu wel eens helemaal onzinnig kunnen worden:[/break] The one company that might threaten all of this T&L shenanigans could well be Intel who, in the next 18 months, do have some high-end graphics aspirations. How much longer will they continue to allow 3D chip manufacturers to do the geometry for their CPUs? We doubt they'll take kindly to an NV10 making their latest and greatest CPUs that bit more redundant with T&L being handled by the graphics sub-system instead of the CPU itself. Could this be the reason why 3dfx, PowerVR, S3, Matrox et .al are not yet opting or talking about a future hardware transformation & lighting part? Or are they just lagging behind NVIDIA?

Verder in Sharky's column: AMD's Athlon, Aureal vs EAX/DX8, ADSL en meer. Lees 't hier.

AMD Zone heeft een nieuwsposting geplaatst over het overklokken van de Athlon. AMD Zone bevestigd wat JC eerder zei: je zult een apart apparaatje nodig hebben om de multiplier van de Athlon te herprogrammeren:

I did some digging around about the whole no overclocking rumor that started here. I didn't want to post anything else until after I had better info. Looks like what JC had posted was basically correct. At present you will most likely have to have an external device to hook up to the processor to adjust its speed. I hope to get my hands on one of these, which could be manufactured by a third party. If anyone knows how to get such a device, or would like to let me borrow one for a while, I would be most appreciative.

Of het ook mogelijk is om de busspeeds omhoog te tillen wordt niet gezegd. 200Mhz = vast?

Andy van BXBoards heeft (in navolging van Upgrade Center, zie deze review) een
review van de Gigabyte GA-660 neer geknalt. De overklokbaarheid van deze tamelijk onbekende kaart blijkt verassend goed te zijn:

I really like this card! Firstly due to the cooling solutions employed it is very overclocker friendly! 180Mhz core clock is easily obtainable, and with a fan blowing onto the onboard "sandwich" cooler, speeds of 185Mhz was possible with total stability. This is Hercules Dynamite Ultra levels folks!

The use of 7ns SDRAM unfortunately prevents the GA-660 from reaching that cards 220Mhz memory clock - imagine what this design would do with 5.5ns memory! Nevertheless speeds of 70fps at 1024 x 768 x 16 were possible in Quake 2 with the core clocked to 180 - its hard to dismiss that sort of performance!

3DGN heeft een review gepost van de Elsa Erazor III, een TNT2 met 32Mb SDRAM en een magere 125/140Mhz kloksnelheid. Wel leuk zijn de (optionele) 3D shutterglasses:

Something Erazor III has that no other can claim is, the Amazing 3D glasses. The 3D Revelator. An interesting set of Stereo 3D glasses that lets you view anything that is Direct3D accelerated in actual 3D by using the trick of depth perception. I Have a review on the glasses, so find more here.

Ik lees op Voodoo Extreme dat er nieuwe drivers zijn voor de Guillemot Maxi Gamer Xentor TNT2's. Ze zijn gebaseerd op versie 2.03.

Xentor32 - win9x
Xentor32 - Win NT

Xentor16 - Win9x
Xentor16 - Win NT

De gasten van The Adrenaline Vault hebben wat info over Tomb Raider 4. Het schijnt dat deze game midden deze maand officieel wordt aangekondigd.

Eidos Interactive has uncharacteristically remained in the shadows regarding the next iteration of the Lara Croft series, Tomb Raider 4: Last Revelation. That is about to change. Earlier this year, the sequel was alluded to in the company’s financial report, and although the game was not publicly shown at E3, it was there, behind closed doors, being shown to a select group of insiders. The game’s official title and fall release date for a PlayStation version appeared on retailer’s release date lists in June, but still nothing from Eidos. As it turns out, the publisher was bidding its time and oiling the PR machine for a summer announcement.

The Adrenaline Vault has learned that an announcement regarding the next title will be made in mid-July. At that time, Eidos will announce the PC and PlayStation versions and reveal the location for the game. According to our source, the designers are returning to the roots of what made the series popular. No, not Madonna-esque pointed breasts, but the essence of exploration, adventure and puzzle solving, as opposed to the extremely challenging gameplay and combat of the third title. Although the game is being created for both platforms, numerous engine enhancements are reportedly being implemented.

Maar ja, zitten we er eigenlijk wel op te wachten?

Hm, ik lees op Voodoo Extreme dat mensen die deze kaart hebben besteld nog iets langer moeten wachten...

We have begun producing G400 MAX units and will start shipping them next week. Unfortunately, we are only getting a limited supply. We will therefore be shipping pre-booked units in order of order placement. Hence the customers who were first in placing their pre-orders will be first in receiving them.

We are doing our utmost to fulfill our commitment made to customers who placed their orders prior to June 11. We will be receiving more MAX units week by week during the month of July and will continue to ship these units to our customers as we receive them.

For our International orders, we will be sending batches of G400 graphic cards to our Dublin distribution warehouse in Ireland. They will then expedite your order via DHL Express. To account for the overseas shipping required, there will be an additional one week delay. An email will be sent to you, when the product will be available to ship, advising you of the shipping and handling charges and the VAT.

Please note that for all our customers who ordered after June 11, your ship date of 5 to 6 weeks is still in effect.

Woohoo!! Er is weer een nieuw filmpie uit van Command and Conquer 2: Tiberian Sun. Hij is geschoten tijdens de E3 beurs afgelopen mei. Het filmpje bevat 2 minuten aan in-game gameplay, erg cool kan ik je verzekeren. Ga naar deze site, daar vind je de mirrors. De eerste mirror heb ik overigens twee keer geprobeerd en twee maal was de zipfile corrupt, dus als het kan, probeer een andere. Er is zowel een high-quality versie van 16MB als een low-quality versie van 5MB.

Oh ja, bij de weg, de commentator tijdens dit filmpje spreekt Duits. De vertaling (indien nodig) staat hier.

Overigens was er deze week ook een kort 20 seconde (2,7MB) durend filmpje uit van Tiberian Sun. Die kun je hier lokaal downloaden.

Hier heb je wat screens van het nieuwe filmpje.

Stereovision.net heeft een koele artikeltje gepost waarin ze beschrijven hoe je de Elsa 3D Revelator aan de praat krijgt op een TNT kaart van andere merken dan Elsa. De oplossing is in feite errug simpel:

Rule: The chipset your video accelerator sports is the key. If Elsa offers 1) a board with the same chipset and 2) offers Revelator-compatible drivers for that board, then you may be lucky. How: Flash the BIOS of your board - use the utility that came with your board but (and that's the important point) don't use the BIOS you got: Instead, use a accordingly renamed Elsa BIOS appropriate for that particular chipset. That done, your board will be recognized as an Elsa board, you'll be able to install the Elsa- and Revelator-drivers. We can not guarantee that this works in all cases - but it did in all we heard about. Before you close that window now, read on, you may be happy you did - there are a few serious points to consider.

Zou tof zijn als dit werkt. Zo'n Revelator heb je al voor 109 piek.

Van Gigabyte's TNT2 kaart hebben we nog weinig gehoord. Upgrade Center heeft een review van de kaart, die geleverd wordt in een errug mooie (en zinloze) blauw gekleurde PCB. Plaatje:

De GA-660 draait default op 125/150Mhz, maar kan op een nogal achterlijke wijze middels een 'turbo jumper' naar 156/156MHz übergetakt worden:

The TNT2 chipset has an impressive list of features including: 128-bit TwiN-Texel (TNT) architecture, 32/16MB 166MHz frame buffer with 128-bit interface, AGP 4X/2X/1X interface with sideband addressing, 300MHz RAMDAC, single-pass multi-texturing, 32-bit true color rendering and per-pixel MIP-mapping. Regrading to the clock speed of the Core/Memory, Gigabyte provide's an option for the user to either use the default (125MHz core /150MHz memory) or a turbo mode (156MHz core /156MHz memory). There is a jumper marked TURBO on the GA-660. When the jumper is set to open, the chip will go to the turbo mode(156/156). Gigabyte claims to have put a lot of effort in this feature and it will run quite stable. Their next driver release will also provide the software option for overclocking the chip, so that the user can easily change the speed without opening the PC. The third party utility, Powerstrip, seemed to work fine in this regard. There are both 16MB and 32MB versions being made by Gigabyte as well as a flat panel display output option.

Overclockin.com heeft wat tips gepost over het koelen van je PII/PIII cache chips:

Well, Lazerus has come up with a great idea of how to do it. First, run down to your local hardware store (or maybe even just out to your garage) and pick up a set of feeler gauges. If you don't know what feeler gauges are, they are thin strips of metal that are marked with various thicknesses and are used to measure gaps between components (sparkplugs, valves, L2 cache, etc.). Here are a couple of pics of the gauges that Lazerus used (with a 486 included for scale):

Next, find the appropriate thickness feeler gauge for the gap between your processor's cache chips and the heatsink you are using. Just insert each strip until you find the one that fits the best. Once you have done that, cut the feeler gauge down to fit the cache chip and then put it on the cache chip with a little heatsink compound on each side. This should help conduct the heat away from those sensitive little cache chips. It is also a much easier solution than machining a spacer. Hope you like this tip and thanks again for the great info Lazerus.

Op een vergelijkbare wijze (trucje gejat van Tom) heb ik ooit twee gulden geïnvesteerd in m'n ouwe PII (de munten als warmtegeleider tussen cache chips en backplate).

ZDNet heeft een artikeltje gepost over Neptune, de codenaam voor de op NT gebaseerde consumer versie van windows. Het is de bedoeling dat Microsoft ons in april 2001 hiermee eindelijk gaat verlossen van de brakke 9x kernel:

Microsoft will target Neptune at consumers, positioning it as a follow-on to Millennium, the next Windows 98 upgrade that is due next year. But because Neptune will be based on the NT kernel, Microsoft is encouraging its Consumer Windows division and Business and Enterprise divisions teams to share components, such as the user interface. As a result, the new usability features developed for Neptune will likely show up in future releases of Business Windows, as Microsoft is calling post-Windows 2000 versions of the operating systems. [break]En net nu NT een beetje stabiel loopt is Bill van plan om er allerlei helende Jomanda krachten in te verstoppen:[/break] Neptune will be Microsoft's first operating system to make use of the WinTone service that CEO Bill Gates outlined last fall. Much like Sun Microsystems Inc.'s WebTone cenept, WinTone is aimed at making PCs self-healing and self-updating.

Microsoft's slogan achter dit nieuwe OS is "It just works". Dream on .

Joshua van AGN Hardware heeft de nieuwe Iomega Zip 250 gereviewed. Naast de toegenomen capaciteit (je kunt er nu 250Meg op kwijt) heeft Iomega de vormgeving licht getweakt:

Een hap uit de review:

The ZIP 250 is the latest incarnation in the ZIP line from Iomega. It boasts 2 1/2 times the storage capacity of traditional Zip drives without compromising the speed of the drive. The unit we received from Iomega was a SCSI-2 device with a MAC 50pin connector. Don't let the MAC connector scare you though, all SCSI Zip drives come this way and you can easily convert the drive to a 50pin centronics connector with the aid of a 3rd party connector. We picked up a Belkin connector from a local computer shop for about $45.

GA-Source bericht dat de Diamond FireGL1 vanaf vandaag in de (Amerikaanse) winkel zal liggen. Met een dikke 256-bits engine is dit wellicht een leuke kaart voor de professionele OpenGL'er. De prijs is trouwens navenant 'dik':

Diamond Multimedia Systems today announced the shipment of the retail version of the Fire GL1 professional graphics accelerator. The Diamond Fire GL1 accelerator, featuring IBM's first graphics chipset and the first 256-bit engine for the Windows NT platform. Priced at a U.S. estimated retail price (ERP) of $995, the Fire GL1 graphics accelerator is shipping now to Diamond's worldwide sales channel of partners and distributors.

Providing advanced performance in 2D and 3D NT workstation graphics, the Diamond Fire GL1 accelerator is ideal for CAD/CAM and digital content creation applications. With its 256-bit graphics engine, the Diamond Fire GL1 delivers up to 15 million anti-aliased vectors per second sustained and processes 200 million pixels per second fill rate. The new Diamond Fire GL1 accelerator outperforms the competition in both 2D and 3D testing. In 3D testing, the Fire GL1 posted an impressive ProCDRS-02 score of 22.3. Under the High-End Graphics segment of WinBench 99, the Fire GL1 yielded a 2D score of 414.

Mocht je dankzij deze posting een meer dan normale interesse voor deze kaart hebben ontwikkeld, dan zou'k ook ff in de laatste c't gaan kijken, want daar hebben ze de FireGL1 getest (samen met nog een stel andere OpenGL kaarten).

Riva Extreme weet ons te vertellen dat Asus vers gebakken TNT/TNT2 drivers naar z'n ftp geupt heeft. Hier heb je de download locaties:

V3400 / V3800 Beta 5 for Win9x
V3400 / V3800 Beta 2 for NT4

Voodoo3.net heeft een interview met Brian Burke van 3dfx op het net gekwakt. In het interview vind je o.a. een aantal vragen over Alienware's PGC technology ("nep-SLI"):

Something that has been on everyone's mind is the Voodoo3 PGC. Why did 3dfx turn that down? With 3dfx's competitors(Nvidia) turning to PGC, how does 3dfx plan on keeping up?

When we first created ScanLine Interleaving (SLI) we looked all the different ways to harness the power of two graphics boards together. We eliminated the technique of splitting the scene in half like PGC does for a lot of reasons. It is a technically inferior way to do it. We are confident that SLI is the best way to do it.

Regardless of the name on the front of the box, if an end user has a bad experience with a product based on the Voodoo3, it will reflect badly on 3dfx. It doesn?t matter if our name is on the front of the box or not. If 3dfx is not in control of the product, we can not control the quality of the hardware and supporting software, the marketing, the cost, the support policies, and all the many other factors that encompass the total end user experience. If we give that control to someone else, we put ourselves in a bad position.

Jaja, dat vinnik wel een errug lame smoes...

Okee, ik heb hier al vaker een posting over geplaats, maar meer info kan nooit kwaad. The New York Times heeft een berichtje over deze kwestie op z'n site gekieperd. Check it:

"The yields people are seeing for copper are not fantastic," said Dan Hutcheson, president of VLSI Researchin San Jose, Calif., "although for some people they aren't that bad." He added the learning curve from aluminum to copper is long, steep and expensive. Companies hoping to put off buying the pricey tools until the technology seems fairly stable might encounter a nasty surprise. "Copper is not a plain-vanilla process, and you can't just buy the tools and run them out of the box," Hutcheson said.

While Hutcheson noted that chip companies with the most advanced copper programs may be seeing yields of 40 to 50 percent, that's significantly below the 60 to 70 percent range that today is considered desirable for leading-edge technology. Other chip companies that are just beginning their copper development efforts may have yields close to zero. "Using copper is a multi-damascene integration nightmare, and that's what's catching all these companies unawares," said Hutcheson.

We mogen hopen dat AMD meer succes heeft met z'n produktie plant in Dresden, wanneer deze volgend jaar online gaat (1GHz Athlon zal niet slecht smaken). Intel blijft voorlopig nog evenop z'n luie bips zitten en slaat de eerste 'koper ronde' over.

Damage van Ars Technica heeft een vet artikel op z'n site geplant, waarin hij aan de hand van een stevige stapel benchmarks (variërend van Winstone99 tot een RH6 kernel compilation) de performance van de "God box" (dual PIII) vergelijkt met de "Nerd box" (dual Celeron systeempje van ComputerNerd).

Verassend genoeg blijken de Celerons het in vrijwel alle benchmarks te winnen van de PIII's. Het verschil tussen single en dual processor ligt over het algemeen op 25-40% bij multi-threaded proggies. Bij single-threaded apps zijn de dual systemen meestal een fractie trager ivm overhead van de tweede processor. Uiteraard hebben we het dan over de performance in een cleane benchmark omgeving. In de praktijk ligt de situatie anders, zo benadrukt Damage in z'n conclusie:

As a counterbalance to all of these benchmark numbers, I should probably say a word about the user experience on these SMP machines. Although many of these tests were not multithreaded and showed little performance gain when run one at a time, the applications being tested would almost certainly run quicker in real-world use. If your PC using habits are anything like mine, you'll have way too many windows or virtual screens open at once, and your PC will have lots of different irons in the fire. As I type this review, my PC is running a mail program, a spreadsheet, a telnet session, a paint program, about six browser windows, ICQ, a system monitoring program, my web page editor, and an MP3 player. This is standard operating procedure for many hard-core PC enthusiasts, and it's the kind of environment in which even a single-user workstation can benefit from SMP. There's nothing quite like kicking off a single-threaded instance of the SETI@Home CLI client or some other huge batch processing task, watching it occupy one CPU entirely, and still having another 500MHz processor in the system to take care of business. On a single-CPU system, I'd be left to twiddle my thumbs while processing Ars's server log file stats; not so with SMP.

In short, SMP offers a kind of performance enhancement in multitasking scenarios that isn't easily captured by benchmark numbers, but its benefits are no less tangible to the end user. Assuming you plan to use an SMP-capable OS like Windows NT/2000, Linux, or the BeOS, getting into a dual-processor rig may be worth the bucks--especially if motherboard and CPU prices keep dropping.

Mocht je nu zitten te wachten op je BP6, dan moet je dit artikel van Ars Technica zeker lezen!

EETimes heeft een artikeltje neer geknalt over de huidige vorderingen op het gebied van quantum computing (nogal mistige en heftige materie - c't heeft er regelmatig info over). Voor ons simpele zieltjes moet het allemaal gaan betekenen dat computers véél sneller gaan worden:

The quantum computer was an interesting theoretical experiment that helped both physicists and information theorists adopt a new view of their subjects. One problem that remained was how to design useful algorithms using quantum states.

A turning point came when Peter Shor, a scientist at Bell Laboratories, published a quantum algorithm showing that a fairly modest quantum computer could factor very large numbers in seconds. Since computer security systems depend on the difficulty of factoring large numbers, this result implied that a small quantum computer could potentially crack any security system. "It takes one trillion years to factorize a 200-digit number with present supercomputers, but it would take only one hour or less with a quantum computer," said Sone. The surprising result galvanized experimental scientists into the search for practical quantum-computing devices.

Voordat ze zover zijn moeten er nog veel obstakels overwonnen worden. Meer daarover lees je in het artikel van EETimes.

Ace's Hardware heeft een zooitje info gepost over de Sun UltraSparc III. Klinkt allemaal errug lekker:

Its most impressive point is the cache subsystem, unlike HP's PA-8500 which has 1.5MB L1 ondie, UltraSparcIII has 32KB instruction L1, 64KB Data L1, and 2KB each of Write and Prefetch cache, 4MB off die data L2, 8MB offdie instruction L2, with all ondie ones being 4way set associative, and all offdie ones being direct mapped. The L1 latency are 2 cycles, Prefetch 3 cycles, and L2 being 12 cycles. But its the bandwidth of these caches thats the most impressive, L1 data have 9.6GB/s, Prefetch have 18.4GB/s, L1 write have 13.6GB/s, L2 being 6.4GB/s, and memory interface which controls 4 banks of SDRAMs being 3.2GB/s. Also it allows 15 outstanding transactions per cycle from each processor which can complete out of order. So by doing this Sun assures that when the instructions observe no data dependency it can crank out lot of numbers by giving the processors lots of stuff to work on. The 16384 2bit branch predictor using Gshare algorithm is also one of the largest we have seem. L2 bus being 288bit and can fetch a whole cache line in one cycle. The ondie L2 tag also aloows quick miss detection. The instruction issue unit is 4 issue, with a 20 entry instruction queue. The integer excution unit have 4 pipelines and excutes most instructions in 1 cycle. The 4 pipelines are A0/A1 arithmetic, logical, shift unit (i assume one does add/sub, the other does mul/div, etc), 1 load/store, and 1 branch unit. The integer unit is 4 issue. The 2 issue floating point/graphics unit have 3 data pathes. Divide/multiply, add/subtract/compare, and an independent division data path, which lets a nonpipelined divide go through without blocking the mul pipeline. The latency on ADD and MUL are 4 cycles, divide being 20 cycles, and square root 24 cycles. the throughput are latency minus 3 cycles. (note these are all double percision 64bit stuff)

Hmm ja, leuk om een webservertje op te draaien!? (wij zijn nog op zoek naar sponsors ).

Ace's Hardware heeft een interessante posting over de huidige problemen bij Micron: een gigantische voorraad en

First here is an EBnews report on how much SDRAM chips Micron have stockpiled over the last year. :p

"Confronted with nearly six to seven weeks' worth of inventory, DRAM maker Micron Technology Inc. has been forced to offer special price-cutting deals with several major PC OEMs." "Micron created its DRAM surplus earlier this year when it began to aggressively ramp production. In fact, analysts estimate that the company can be pumping out as much as 47 million 64-Mbit DRAMs a month." "Micron previously said its total DRAM output will be doubled later this year when the acquired TI fabs fully come online with upgraded 0.21-micron processes."

Now the REALLY bad news from Taiwan and Japan, thanks to Daiki and KH. So according to these news, the 0.21um process 64Mb SDRAM chips Micron made at ALL its plants, especially the Boise, Idaho fab are going to be a total waste. Micron's new 0.21um 64Mb chips are already rejected by Dell, Compaq, and HP for quality reasons. Although Dell and Micron would not confirm or comment on this the Taiwan source said. The problem are said to be the optical Mask used in the lithgraphy, said the Japanese source and also hinted by the Taiwan report. This result would render 3 months of productions of 64Mb SDRAMs by Micron to waste, for both PC100 and PC133 class products. Also it would take 3 monthes for the new masks to come on line and let Micron reopen its production lines. The estimated loss is 450M for Micron for the lost chips alone. The market hiatus of Micron chips could result in higher SDRAM prices in the near future.

Again these are conflicting reports. So maybe Micron reduced its price as a means of trying to sale those low quality RAM chips.

Dat laatste klinkt minder prettig (RAM prijzen omhoog!?).

Mobo bakkers verkeren in staat van pissigheid ivm Intel\'s voortdurende gestuntel met de i810 chipset, zo bericht The Register:

A report in Asian magazine Eurotrade is suggesting that the number of errata (bugs) in the 810 chipset is causing many Taiwanese motherboard manufacturers to complain vociferously.

The magazine says that follows another fix for the 810, for the real time clock (RTC), which Intel confirmed to us yesterday was shipping.

An earlier problem with the 810 chipset caused incompatibilities with microprocessors which caused machines to hang.

The magazine reports that SiS (Silicon Integrated Systems) is picking up business from Taiwanese motherboard manufacturers, angry with the problems.

Ik kreeg Alex / Hardware Junkie een ICQ message dat-ie een review van de Abit BP6 online heeft gezet. Zijn info over de busspeeds:

66, 72, 75, 78, 80, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 104, 106, 108, 110, 124, 133 Mhz. [break]Hmm, leuk. Dus toch wél busspeeds tussen 100 en 124Mhz... Verder wat info over de onboard UltraATA/66 controller:[/break] The BP6 has a special U/DMA 66 controller on-board, sitting next to the standard BX IDE controller. It has it's own BIOS so you can boot from your fresh new U/DMA 66 HD, It uses 2 extra IRQ's (twice IRQ 11 on my system, don't ask me how, kinda funny tough, seems to work with both channels used) besided IRQ 14 and 15 from the 'standard' BX PIIXIDE IDE controller.

Ook leuk: een extra controller, maar het ding gebruikt maar 1 IRQ voor beide channels. Check de review voor meer info.