De USB poort waarvan iedere pc nu al zo'n twee jaar vergezeld gaat lijkt eindelijk enig nut te krijgen. Na webcams, muisen en joysticks komt IOmega nu met een USB versie van z'n Zip drive. Een handig alternatief voor parallel (elke moderne PC heeft USB, dus je kunt 'm overal aan hangen, wat bij SCSI niet lukt), alleen nu nog hopen dat de snelheid ook beter is, want die is bij parallel dikke drek:

Performance was good. In real-world benchmark testing, the USB Zip drive was anywhere from 2.4 times to 5.4 times faster than the parallel Zip. The internal ATAPI Zip drive was slightly quicker, but note that it uses the much faster IDE interface (5 to 33 Mbyte/Sec) while USB is only capable of 1.2 Mbyte/Sec. The USB Zip drive also performed much better than the parallel Zip drive in Ziff-Davis' Business Disk WinMark 99, garnering a score of 381 Kbyte/Sec. The ATAPI drive was 20% faster in this benchmark. What's interesting, however, is that the CPU utilization for the USB Zip drive during the test was a mere 7%, soundly besting the other drives. This means that when you copy files to or from the USB Zip drive, your computer will still be free to perform other operations such as printing.

Weer een review bij Thresh's Firingsquad, ditmaal met de ATI Rage Fury in de spotlights. De performance van de ATI Rage 128 hangt dicht rond die van de TNT, maar bij 32-bits rendering neemt ATI de leiding. De performance en beeldkwaliteit bij 32-bits rendering is perfect, maar 16-bits schiet de ATI tekort:

That's why the noticeable lack of dropoff in 16 and 32 bit resolution scores for the Rage 128 is exciting news. Normally, the additional bandwidth required for 32-bit is prohibitive. After all, 32 bits is double the information for every single pixel, for every single frame! The Rage 128 pulls off 32-bit color with only a minor performance penalty in Q2 and DethKarz, whereas the TNT suffers a noticeable dropoff at 800x600 and higher when switching to 32bpp. This indicates a superior memory architecture for the Rage128, and finally provides us with the ability to run at high color depths without loss of framerate.

Unfortunately, the excellent 32-bit performance comes at a price-our beta drivers have the worst 16-bit rendering I've seen since the bad old days of the Riva 128! According to ATI, this is primarily due to the beta drivers for the Rage128, but it's quite a disconcerting sight. The dithering method used leaves visible artifacts with heavy graininess and posterization. More on that later.

Gamespot heeft een review van Logitech's Formula Force racestuur. Je hebt 'm vast wel eens in de winkel zien liggen. Het ding ziet er erg vet uit, veel koeler dan het stuur van Mikroschoft (wil er nou zo'n ultra lame stuur van M$ op z'n burea hebben). Hier wat info uit de review van Gamespot (waar overigens weinig spul in staat):

After resetting the wheel to a "blank" configuration (most Windows driving games let you customize the control setup anyway), we booted up Need for Speed III and Speed Busters. At first, the forces seemed a bit weaker than the Microsoft wheel. A trip to the game controller control panel let us tweak the forces up to 150 percent (!). At those maximums, the effects came through strongly, but weren't overbearing. The rubber grip of the wheel was quite comfortable, but the plastic foot pedals had very little resistance and were hard to use with any precision. The wheel itself seemed fairly precise, so the car would generally go where we wanted. Like all such devices, it will not work with most DOS games.

Diamond heeft een nieuw BIOS uitgebracht voor de Viper V550. Het dingetje van 201Kb kun je hier neerladen...

Een C400 op 731MHz, het klinkt erg onwaarschijnlijk en ik geloof er dan ook niet al te veel van. Iemand op BXBoards message board zegt het voor elkaar gekregen te hebben. Hier wat info uit die thread:

I just witnessed a Celeron 400 slot 1 clocked at 133Mhz x 5.5. It booted all the way into Win98 and were able to run CPUID. It got so hot the alarm started going off in a few seconds. Luckily no damage. This Celeron must be a "Pre-Production" version as the multiplier is not locked. We bought it at a local PC store, but it was the only one they had until Feb. The motherboard was a Gigabyte (with onboard SCSI and Lan). We were stupid enough to try it once, but probably won't again because the mb was around $300...who cares about the cpu. Here are the saved results:

CPUID : GenuineIntel, Typ=0, Family=6, Model=6, Step=0
Processor Core : Pentium II, > Deschutes (Katmai? Williamette?)
L1 Data Cache : 16 KB, 4way associative
L1 Code Cache : 16 KB, 4way associative
L2 Cache: : 128 KB, 4way associative
L2 Cacheable Area : 4 GByte
L2 ECC : present and enabled
Actual clock rate : 731.5 MHz
System clock rate : default: 133 MHz, measured: 133 MHz
Ratio (cpu/syst) : 5.5
ctP2info has detected : Pentium II 400.0 MHz with ECC

Yahoo! The Street.com heeft een interview met Anand van ... AnandTech met wat gelul over de leeftijd van Anand (hij is pas 16) en zijn review van de K6-3 (waarmee hij de eerste was):

When evening comes, Shimpi turns into Anandtech.com, a burgeoning online enterprise. The first time we attempted to talk to him, he had to call us back because of an unfolding crisis over his new $40,000 server, a piece of equipment he paid for with ad revenues. To ease up, he recently hired three employees -- two editors, who receive a small salary, and a database expert, who works for the gadgets computer companies send Shimpi to review.

[...] But his site has evolved quickly from something that less than two years ago started out as a self-designed Web page on GeoCities ( (Nasdaq:GCTY - news) ).

"It was mostly my thoughts, online," Shimpi says. "I got 36 visitors on the first day." Nothing near the audience he garnered when he released his notes on AMD's new chip: In the first 12 hours, the site received more than 300,000 hits. "It took my server down," he says. Over the next three weeks, his readership rose from an average of 150,000 to 260,000 page views a day.

"I really have no way of assessing him," says Nathan Brookwood, an analyst with Insight 64. "But the benchmark results for the Sharptooth were about what I had expected, so he must be a pretty sharp kid."

Ik kreeg vandaag de PU van februari in de bus, dus hier meteen maar even een korte inhoud:

Eerst maar even een lijstje met ge(p)reviewde games:

• Settlers 3 (PC)
• Sim City 3000 (PC)
• Tomb Raider 3 (PC)
• Grand Touring (PC)
• Buggy (PC)
• NBA Live 99 (PC

• Buggy (PSX)
• NBA Live (PSX)
• NHL 99 (PSX)

en verder:

• Dreamcast Special
• meer (p)reviews
• Powerplay special over Gameshardware

Ook bij Singapore Hardware hebben ze wat te melden oever Intel...

Intel will soon announce their .18-micron fabrication process. This size reduction will result in faster CPUs, with Pentium IIIs up to clock speeds of 700MHz. The article says Intel can get rid of Slot 1, like it was already obsolete. I remember a while back that another Intel rep had said that socket or slot didn't make a performance difference. Why did they go to Slot 1 in the first place? The extra space was needed for the L2 cache modules. How come Intel didn't just push for Pentium IIs to be more like the new Celerons??

We may never know the truth. Coppermine and Cascades, the first to chips to be made under the new process, will contain 256KB of integrated cache memory. Just think of a Celeron with a .18-micron core and double the L2 cache. Now that's impressive. Click here to read the full article.

http://home.hawaii.rr.com/biosfear/

Oeh, er staat veel te gebeuren, op Rambus gebied, ging eerder Acer al licentie aanvragen bij Rambus Techn., nu gonst het weer dat Samsung, Intel gaat helpen bij de productie...

Intel has invested $100 million in Samsung to help them speed up production of Rambus ram.

Following an investment in Micron Technology, the investment in Samsung marks the second time in less than a year that Intel, in Santa Clara, Calif., has taken an equity stake in a DRAM company to ensure a reliable supply of the high-speed Direct Rambus memory devices, which suppliers are having difficulty producing in large quantities this year.

The video gurus at SciTech have released GLDirect, a universal Direct3D wrappers of sorts. It enables Direct3D video cards to run OpenGL applications without a specific OpenGL driver. GLDirect currently supports a whole host of D3D accelerators including those from 3Dfx, nVidia, Intel 3DLabs and many, many more. A full list is available at www.scitechsoft.com/gld_hrdw.html. You can download a shareware version of GLDirect from www.scitechsoft.com/down_gldirect.html. Have fun.

Het gonst van de geruchten ook bij nVidia, sinds 3Dfx STB overnam, zijn speculaties niet van de lucht...en vooral nVidia met zijn aandelen de markt opging.

For some more news and a info on NVIDIA's IPO, our buddy Ken Feinstein has posted a small interview with NVIDIA's CEO, Jen-Hsun Huang talked with Ken about their goals for the future. You can read the full article here. Here is an interesting snip from the interview: When asked about the strength of archrival 3Dfx's brand in the retail space, and Nvidia's prospects for competing with the upcoming Voodoo 3, Huang had some tough comments. "Even costing $50 more, TNT boards are outselling Banshee-based boards in retail," he said. "The Voodoo brand appears to be worth negative $50. If I were 3Dfx, I'd run away from that brand as quickly as possible. The purpose of a brand is to create value, and value is measured only by profits. By buying Nvidia's products, a cardmaker gets an extra $50."

www.rivazone.com

Als je dacht dat Cyrix bij de pakken neer ging zitten, dan heb je het mis, MX wat de klok slaat...

Daiki tells me of this news story, which says that Cyrix has finally decided to make the jump to the inexpensive PGA370 platform with their MXs, which we previously knew as Jedi. The MXs, as we've known (for, well, a really long time), is an M2 with the Cayenne floating point enhancement. This means that it has an x87 unit (fpu) at about the level of the Pentium!!! (some parts of it better, some not as good, but mostly the same in latency and throughput, if the numbers I've seen are correct). The estimate of the article is that the MXs will exceed the performance of the PII by 30% at like clock speeds (which is probably using 3DNow!, implying a lesser implementation compared to AMD's). The story goes on to say that the MXs will start at 400MHz and eventually ramp to 600MHz. Personally, I think those are PR values, we won't see those go away until the Jalapeno in early 1999. So, the MXs will be very fetching at same-MHz comparisons, but we'll still likely see it lagging in x87 performance with the PR rating, but it will still be worth looking at -- finally! The MXs will be sampling in April and is expected to come out by July.

over het eventueel verdwijnen van het slot-1 stond iets bij BXBOARDS...

My, the times they are a changing. This News.com article reports that Intel is considering phasing out the Slot-1 architecture altogether to lower costs. The article alludes to the fact that Slot-1 is now unnecessary because of the obviously successful Celeron "A" technique of moving the L2 cache on to the CPU die instead of separating it as in the P2 design. Look for a new Intel CPU package sometime in that latter half of this year.

The shift to 0.18 microns will yield similar, but different results. With these chips, Intel is going to start integrating more functions onto the processor. Coppermine and Cascades, the first to chips to be made under the new process, will contain 256KB of integrated cache memory, said sources. Current Pentium IIs have 512KB of secondary cache memory, but it sits alongside the processor and is not part of the same piece of silicon.

Toon mij je chip en ik vertel je wie je bent...

Dat scenario wordt mogelijk door de plannen die Intel in het geheim heeft ontwikkeld om alle processors te voorzien van een ‘digitaal identiteitsbewijs’. Het gaat daarbij volgens ingewijden om een uniek nummer waaraan elke chip te herkennen is en een ingebouwd component dat registratienummers genereert die kunnen worden gebruikt door bedrijven die via het Internet diensten en goederen verkopen. Een online winkel zou dankzij het ingebouwde ‘paspoort’ altijd weten met wie er zaken worden gedaan. Het ID-nummer kan worden gebruikt om gestolen computers die door de politie worden opgespoord weer bij hun rechtmatige eigenaar terug te bezorgen. De plannen van Intel zijn uitgelekt via organisaties die zich inspannen voor behoud van de privacy, zoals de machtige burgerrechtenbeweging in de Verenigde Staten, de ACLU. De ACLU zegt dat het ID-bewijs voor de nieuwe generatie Pentiums een prachtig middel is om veilig producten op het Internet te kopen. Ook kan gevoelige informatie beter worden afgeschermd, omdat deze alleen door de bezitter van de chip met het juiste nummer kan worden ontvangen. Het ‘digitale paspoort’ maakt echter ook een einde aan de mogelijkheid om je anoniem over het Internet te begeven. Intel heeft al wel rekening gehouden met mogelijke protesten tegen zijn plan. De chipgigant wil gebruikers een programma bieden waarmee de nieuwe technologie wordt uitgeschakeld.

www.webwereld.nl

gelezen bij de AMDZone...

K6-3 According to this The Register article the K6-3 will be coming out sooner than February 24th. Here's the buzz:

AMD is now expected to introduce its SharpTooth K6-3D processor earlier than the 24th February.

Meanwhile, rumours are circulating in the industry that Gateway 2000 is likely to use the K6-3D processor in some of its machines.

And look, Gateway might just be in!

www.amdzone.com

Next Generation Chip Article-ruiner...

Source: c't

c't has translated this article on next generation processors so that us non German speakers can get some of the goods. Here's a portion:

Due to the other bus protocol as well as the recently licensed Direct Rambus interface the K7 needs special chip sets and boards. Both should be shipped at the same time as the processor chip. AMD hopes that at the end of 1999 the new fab in Dresden will be ready to shrink the processor to 0,18 µm allowing clock frequencies to increase up to 800 MHz.

www.amdzone.com

Ik vond een vrij uitgebreid stuk bij Tom over degeruchten van de K6 en K7 en Katmai introducties...wilde jullie dat niet onthouden.

I spoke to Dana Krelle, AMD's vice president of marketing, and he gave me the latest scoop on AMD's plans. K6-3 is supposed to ship very close to Intel's Pentium III launch. AMD is trying to release this new CPU with 256 kB on-die L2 cache already prior to Intel's Pentium III release.

Most of you know that K6-3 will run on the same Socket7-motherboards that currently host K6-2 CPUs, you may only require a BIOS update. The performance of K6-3 will be high enough to make it very competitive to Intel's Pentium III, the K6-3 version at 450 MHz will most likely outperform the new Pentium III 500 in many benchmarks. The highly official version says that K6-3 will definitely ship in Q1/1999, but we know that Q1 ends on March 31. Let's hope K6-3 will be available earlier than this.

K7

K7 is also still on track, planned release time is Q2/1999. This seems surprising, because the delay of K6-3, which was expected in January 1999. K7 will be running in a new unique platform using a slot called 'SlotA', which is mechanically identical to Slot1, but using a completely different bus protocol. Thus K7 will not run in Slot1-boards and no Intel CPU will run in SlotA-boards. K7 will require its own new chipset and AMD will be the first supplier of one.

The bus protocol used by K7 is called 'EV6', known from Digital's Alpha CPUs. The EV6-bus of K7 is supposed to run at 200 MHz, offering a significantly higher bandwidth than the 100 or later 133 MHz bus clock used by Intel's Pentium II and Pentium III CPUs. The performance of K7 is expected to be significantly above Pentium II or Pentium III level, in office, multimedia, 3D as well as floating point intensive applications. For more information about K7 please look at 'Why AMD’s K7 will be Intel’s toughest competitor ever'.

nog meer lekkers...
hier te vinden

Er waren nogal veel discussies over vloeibare vormen, omtrent de koeling, ik kon het niet nalaten wat ik bij Heat-Sinkguide.Com heb gelezen even door te geven.

In theory, it should be possible to reach sub-zero temperatures, which would allow extreme overclocking, with high performance peltier elements. The main problems with high-end peltier coolers (just like with any other supercooling method) is condensation. If the cold CPU is exposed to ambient air, water will condensation, and possibly cause a short circuit.

I've been thinking about a solution, a Peltier sandwich for Celeron CPUs, where no cool part is exposed to the ambient air. The entire CPU core is encapsulated in an insulating material, e.g. silicon, polyurethane or epoxy resin. Two peltier elements cool the CPU, one on each side, and heat up the two heatsinks. Note: Until now, nobody has ever built such a unit - all info presented here is completely theoretical. However, it would be relatively easy to build the cooler, and if you are interested to build one, please contact me to share your experiences. The problem is that because of the insulating material, the process is not reversible.

If you decide to give it a try, I would suggest that you fix the heatsinks and peltier elements first, and then apply the insulating material all around the CPU (silicone is most convenient, I guess). Since the front heatsink has fans and the rear heatsink has not, I recommand to use peltier coolers with different performance, e.g. a 40W element on the front, and a 20W element on the back, or (for those with a big power supply) a 80W model in the front and a 40W model in the back.

http://www.heatsink-guide.com

Dit is namelijk wel toepasbaar op de werk plek en maakt nauwelijks rommel of herrie!