Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: Tony Viola who wrote (84590)	6/28/1999 11:13:00 AM
From: John Koligman	Read Replies (1) \| Respond to of 186894

Overclocking article in today's WSJ. Celeron 300A mentioned. John When Fast Isn't Fast Enough, There's Always Overclocking By MIKE ELEK THE WALL STREET JOURNAL INTERACTIVE EDITION WHEN MY OLDEST BROTHER, Pat, was 16 years old, he bought a bright orange 1965 Chevy Malibu. Within a couple of months, he replaced the chrome grill with a blackout version, added a tachometer and hood pins, replaced the stock two-barrel carburetor with a Holley four banger and jacked up the rear end and fitted it with mag wheels and wide tires. My mother used to say it was a nice car before he got ahold of it. She'd probably say the same thing about my computer. Speed is relative, and within the month it became obvious that the 333 megahertz (MHz) my machine was giving me wasn't going to be enough. Sure, Internet Explorer and Netscape Communicator snapped open and Word 97 no longer annoyed me -- at least after I exorcised the animated Office Assistant. But I yearned for power ... computing power, that is. The solution: overclocking. In case you haven't heard, overclocking forces your processor to run faster than its specifications, and it's the latest rage among hobbyists. For some, it delivers a modest power boost; for others, a sizable power increase. One computer magazine even devotes a monthly column toward this pursuit. If you decide to go this route, there are a few warnings: You immediately void your warranty. Of course, if you've built your own computer, that's not a problem. But think twice before cracking open the case of your IBM, Gateway, Dell or Compaq. Forget about trying to overclock a laptop. Even if you could -- and you probably can't -- it would be perilous for the computer because of the extra heat that is generated. Obviously, you have no business altering a computer you use at work, unless you own the company. This also applies to leased computers, as well. You don't want to have to explain overclocking to tech support or the boss -- particularly if you made a mistake. I wouldn't try this with a computer that has valuable data on it. At the very least, back up all of your information. At worst, be prepared to buy a new computer. Not everyone will be able to boost their processor. Some processors simply don't budge from their intended speed. You get what you paid for -- nothing more, nothing less. If you're careless or reckless, it is possible to fry your processor or some of your other costly peripherals. Your hard drive, memory or video card might begin to act strangely. Windows itself might not run correctly -- or at all. The worst-kept secret among hobbyists is that Intel Corp.'s Celeron processor, specifically the inexpensive Celeron 300A, is a hardy processor that almost begs to be overclocked. This 300-MHz processor gets an "A" designation because it adds Level 2 cache -- high-speed memory that stores recently used information. The favorite combination is the Celeron 300A, affectionately known as "the Celery" in overclocking circles, attached to an ABit Computer Corp. motherboard. So in the interest of knowledge (and also because of an unfortunate accident -- unrelated to this experiment -- involving my AsusTek Computer Inc. motherboard), I picked up both of these items via mail order. Intel has stopped making the Celeron 300A, but its popularity persists, and like many things now gone, it has gained a cultlike following. It's probably the first time that interest in a slower CPU has actually increased after it was discontinued. The ABit motherboards are a good choice for hobbyists, because they allow nearly all changes to be made through software, instead of having to move jumpers or flip tiny switches on the motherboard. Other motherboards also offer this software-controlled BIOS, although ABit boards seem to be the most popular. Overclocking usually involves making two changes: altering the motherboard's front-side-bus speed or the processor multiplier, or both. The front side bus, usually known simply as the bus, is the electronic bridge that allows the processor to communicate with the rest of the computer's components, such as memory. The first Pentiums ran with a bus speed of 60 MHz and later 66 MHz -- the speed of the original Pentium processors. As new processors were introduced, the chip speed became a multiple of the bus speed. A 90 MHz processor was 1.5 times 60; a 133 MHz processor was 2 times 66 MHz; 150 was 2.5 times 60; 166 was 2.5 times 66 and so on. Back in those days, you could change the multiplier and get a small boost in power -- if it worked. With Intel's release of its Pentium II 350 MHz processors, the bus speed moved to 100 MHz, and suddenly overclockers had a new weapon in their quest for processor power. The multiplier on Intel's newer processors is fixed -- it can't be changed. The Celeron 300A's multiplier is locked at 4.5 and was designed to run in a computer with a 66 MHz bus speed -- 4.5 X 66 = 300 MHz (rounded slightly). Overclockers got around that problem by increasing the bus speed. Many third-party motherboard makers permit increases in small increments from 66 MHz to 75 MHz to 83 MHz to 100 MHz and beyond. The one thing that you must consider is how it affects the other components in your system. Increasing the bus speed also increases the speed of the PCI slots, which normally run at 33 MHz, affecting hard drives and video, sound and other peripherals. Some components are very sensitive to PCI bus speed, while others seem oblivious. Getting Along Here is the relationship between a computer's bus speed, PCI components and chip speed: When bus speed is: PCI slots run at: Celeron 300A speed is: (multiplier: 4.5) Celeron 333 speed is: (multiplier: 5) 66 MHz 33 MHz 300 MHz 333 MHz 75 MHz 37.5 MHz 338 MHz 375 MHz 83.3 MHz 41.65 MHz 375 MHz 415 MHz 100.2 MHz 33.3 MHz 450 MHz 500 MHz* 103 MHz 33.4 MHz 464 MHz did not boot 112 MHz 37.3 MHz 504 MHz* did not boot * Booted to DOS but wouldn't run Windows. As the table shows, you're much better off overclocking to a 100 MHz bus speed -- or even 103 MHz -- because the PCI interface returns to 33 MHz. The 333 MHz processor was unable to reach that golden 500 MHz speed, but my 300A ran problem-free at 450 MHz -- that's a 50% increase in processing power. That's an impressive punch for a $75 processor. That in a nutshell explains the Celeron 300A's popularity. Others report getting as much as 504 MHz from their Celeron 300A. But it usually involves disabling many settings, which degrades system performance. That offsets the advantage of overclocking. 'Juicing the Celery' Benchmark* Pentium 200 MHz Celeron 300 MHz ... ... overclocked to 450 MHz CPU Integer 500.4976 MIPS 878.6106 MIPS 1319.047 MIPS CPU Floating Point 231.5613 MFLOPS 349.5986 MFLOPS 522.4636 MFLOPS Memory 183.692 MB/s 522.5173 MB/s 802.8892 MB/s Cached Disk 33.72919 MB/s 49.15382 MB/s 75.65965 MB/s MIPS -- million instructions per second MFLOPS -- million floating-point operations per second MB/s -- megabytes per second * Results are from the WinTune 98 benchmark utility from CMP Media Inc.'s Windows Magazine. Keeping the processor and motherboard cool is extremely vital: Heat is the enemy of electronic components. The best way to keep your system cool is to make sure your power-supply fan is running, tie down the loose drive and power cables so they don't block air flow, and add extra cooling fans to your computer. To cool the processor, there are several Web-based companies that serve the overclocking market. Some use the tried-and-true combination of heat sink and fan, while others get more exotic with Peltier and refrigeration. I'm using a heat-sink-dual-fan combination from Computer Nerd (www.dudecomputer.com), which also serves up the fearsome-looking Wopper Celery Sandwich (two heat sinks and four fans that wrap around the Celeron processor). While the computer is running, I can hold my fingers on the heat sink and the rear of the processor, which means it's running cool. My older Pentium 200 MMX got so hot running at its rated speed -- even with a heat sink and fan -- that I couldn't touch the heat sink for more than a second or two. Why the difference? Several reasons. One is that the newer processors use a much thinner slice of silicon, which means they produce less heat. The other is lower voltage requirements. The older 486 and Pentium processors used from 5 volts to a split 2.8 or 3.3 volts/5 volts. My Celeron uses just 2 volts, which I bumped up to 2.1 volts when running at 450 MHz. Detractors say that overclocking your processor may shorten its life, as well as affect the surrounding components in your computer. There's no way of knowing how much time is lopped off a computer's life expectancy. But all computers' usable lives seem to be shrinking because of the frequency of newer and faster chips flooding the market and software that demands those faster chips. (What became of my Asus P2B and Celeron 333A? I accidentally erased the BIOS was forced to purchase a replacement BIOS from Asus. Oops.) I can't emphasize strongly enough the risks involved in overclocking any computer. It's a decision -- based on risk vs. performance -- that each person must make. As for me, I like this newfound speed. I can only hope that it satisfies my thirst for power ... computing power, of course. Very soon, the overclocking argument likely will disappear as Intel and other chip makers introduce processors that are blazingly fast and inexpensive, essentially rendering overclocking a hobby of the past. Still, the Celeron 300A will probably go down in history as the one of the most popular processors of all time among hobbyists. Write to Mike Elek at melek@interactive.wsj.com. Return to top of page

To: Tony Viola who wrote (84590)	6/28/1999 12:47:00 PM
From: Tenchusatsu	Read Replies (1) \| Respond to of 186894

<Have any idea what the Samsung 1 Mb DDR SDRAM chip described below (my post on Rambus) might mean to Rambus?> A 1 Mb DDR SDRAM made on 0.13 micron technology, running at 350 MHz, ready to sample by year's end? Seems rather hard to believe, but true or not, I'll bet this gives the anti-Rambus coalition new ammunition. Tenchusatsu

To: Tony Viola who wrote (84590)	6/28/1999 1:29:00 PM
From: kash johal	Respond to of 186894

Tony, Re:"Samsung 1Gb DDRSDRAM" I suspect it will be about a year before these are really available. My suspicion is the memory market will fragment into several options and OEMS will use such memory options to differentiate their systems. I know of at least one CPU that will be able to take advantage of that PC300 DDR memory. Here's another aticle on tester issues and varying forecasts for RDRAM. Rambus delay is starting to hurt test gear makers By Jack Robertson SAN JOSE -- It's deja vu all over again for chip production equipment makers. Just as the stalled 300-mm wafer effort put a stopper last year on next-generation equipment purchases, the new delay in moving to Direct Rambus DRAMs is having a deadly effect on orders for the new gear needed to test the new wideband memory. Earlier this year, chip makers had been hustling to get ready to move into volume production with the new DRAM that Intel Corp. had designated as the standard memory for its next-generation microprocessors. It was due to arrive this spring. But when the MPU giant suddenly delayed its big Direct Rambus launch until later this year, most DRAM vendors put their volume production plans on hold. And major new orders for the 1-gigahertz testers that chip makers had been pressing their gear suppliers to deliver as quickly as possible were pulled back. The delay in the ramp up of the Direct Rambus has left suppliers like Hewlett-Packard Co., a leading vendor of testers for the new chip, in the lurch. The holdup also will allow competitors not as far along in developing their Direct DRAM testers enough time to catch up with the leaders, points out a worried Pascal Ronde, HP's European general manager. That's okay with device makers, who aren't in as much of a hurry now to speed up their Direct Rambus production lines. By the time they get a better handle on how quickly the Direct RDRAM market develops, chip makers are counting on additional test vendors coming to market with production testers. More competition, they hope, will help to drive down prices of the expensive systems. Existing testers can't do the job of high-volume testing. Direct Rambus DRAMs operate at 600-MHz, 711-MHz, and 800-MHz speeds, far outstripping today's test gear that can't handle anything higher than 250-MHz. Major tester vendors do offer 1-GHz systems now that chip makers have been using for development work and low-rate initial production, but they don't have the throughput to handle Direct RDRAM chips economically in volume production. What chip makers are looking for are Direct Rambus testers that can handle 64 devices in parallel, similar to that of state-of-the-art SDDRAM testers. But the 1-GHz testers from HP and Teradyne Inc. have only 16 parallel stations and models from Advantest Corp. and Schlumberger ATE are limited to eight chips. As a result, test system vendors are going all out to develop a 64-station 1-Ghz testers as quickly as possible. Some suppliers are now talking about offering an interim 32-station tester. But their customers are taking a "wait and see" approach and aren't in any rush to invest heavily in Direct RDRAM testers. NEC Corp. is typical of where most DRAM makers are now regarding Rambus. "We have enough capability to make limited production of Direct Rambus now," says Shigeki Matsue, associate senior vice president. "We will wait to see how fast the Direct Rambus market develops before investing more heavily in new production equipment." That kind of attitude hits HP hard. The Palo Alto, Calif., company had counted heavily on jumping out in front of its rivals early in delivering Direct RDRAM testers. "We had developed our new high speed tester aimed at the Direct Rambus market," says HP's Ronde. "We will continue to build machines even if the Direct Rambus market is delayed. We will just inventory them," he says, "and wait for the market to take off." Other tester suppliers aren't all that unhappy about what's happening. One of them is Munich-based Gerald Kessler, senior managing director of Advantest Europe, which has a solid position in the market for SDRAM testers. He suspects that as memory vendors closely monitor the progress of the Direct RDRAM they may take an interim step and shift to PC133 and Double Data Rate SDRAMs. But the tester market for synchronous memory also will face the need shortly for 64-station, 1-GHz testers just as Direct RDRAM has. DDR versions that are due to come on the market shortly will need to be tested at speeds of 300-MHz and higher, points out a spokesman for Teradyne in Boston. "Conventional SDRAM testers begin to peak out at 250-MHz," he says. "We can probably push this slightly higher." But vendors moving to the next-generation high-speed DRAMs, he adds, will have to move to 1-GHz systems sooner or later. With minor modifications, however, the same tester can handle both DDR and Direct RDRAM chips. Doing that could increase demand and reduce the impact of the Direct Rambus bottleneck on tester orders. That's because chip makers are expected next year to ramp up their output of SDRAMs running at 300-MHz and higher, regardless of the status of the Direct RDRAM. Estimates vary widely on how markets will shape up for the next-generation memory chips. Dataquest has just revised its Direct RDRAM forecast down slightly to a 5% DRAM market share this year, 20% next year, and 50% in 2001. Semico Research Inc. is more conservative in how it sees the Direct RDRAM penetrating the memory market. Analyst Sherry Garber puts the new chip's share at 1.7% this year, 8.6% next year, and 13.6% in 2001. She sees DDR-SDRAM parts growing even faster, gaining a 8.6% share this year, 16.8% next year, and 26.4% in 2001.