The race was on and last March Advanced Micro Devices (AMD) Inc. inched its way past the 1GHz mark just ahead of its arch rival Intel Corp.
Not to be outdone for long, Intel quickly followed suit, and on March 8, just two days after Sunnyvale, Calif.-based AMD began shipping its 1GHz processor, Santa Clara, Calif.-based Intel followed suit.
But some fear that the stiff competition between the two rivals means products sometimes aren’t being tested thoroughly before they ship.
On July 31, Intel shipped its 1.13GHz Pentium III processors, this time seemingly beating out AMD’s 1.1GHz Athlon processor. But within a month the 1.13GHz Pentium III was being recalled. A glitch in the processor caused it to freeze up under certain conditions. The news of the recall came the same day that AMD began shipping the 1.1GHz Athlon.
But this wasn’t the first or last time in 1999 that Intel had to do a quick turn around and recall a product. In May, Intel had to replace PC motherboards designed around its 820 chip when a faulty component sometimes caused system failures. In November, Intel had to correct a glitch in the software for its new Pentium 4 motherboards. Most recently, at the beginning of this year, some Linux OS vendors announced their operating systems don’t install cleanly on CPUs with Pentium 4 processors.
rush to market
“I think there are at least some cases where there has been a rush to market that’s contributed to some of these problems. In particular, the one that was most clear was when Intel had to recall the 1.13GHz Pentium III. They were under a lot of pressure from AMD at the time to get a faster chip out, and I think that contributed largely to the fact that they wanted to get that chip out before it was ready, and had to recall,” said Linley Gwennap, a senior analyst for the Linley Group in Mountain View, Calif.
Kevin Krewell, a senior analyst at Micro Design Resources in Sunnyvale, Calif., agrees that Intel’s desire to beat AMD may have contributed to the 1.13GHz Pentium III glitch.
“Certainly there’s an indication that that’s happened. I think the Intel 1.13GHz-recall was symptomatic of Intel trying to rush a product to market without extensive enough testing, to keep competitive with AMD,” Krewell said.
Not surprisingly, Intel has another take on the situation.
“The 1.13 hardly involved any volume at all. It was so small. It’s just one of these things that get really overblown. These chips have years of engineering built into them. They’re incredibly complex,” said Toronto-based Brett Udashkin, a regional manager at Intel Canada.
“Last year, I think, was an anomaly, clearly, if you go back and look at our history,” Udashkin said.
“We had a lot of product introductions last year.”
faster not always better
Despite the stumbles along the way, the race for speed continues. And Intel is claiming the lead with its Pentium 4 processor which runs at 1.5GHz. Yet some benchmarks suggest that with some applications at least, users may be better off with slower Pentium IIIs and Athlons.
The Pentium 4 is a new architecture, explained Micro Design Resources’ Krewell, and only applications that have been recompiled for the 144 new instructions in the Streaming SIMD extension 2 (SSE2) will benefit from the extra speed offered by the new processor, he said.
“Those applications that are recompiled for SSE2 show a big benefit. But…in general productivity applications like spreadsheets and PowerPoint and word processors and such show very little benefit from all that extra clock speed. So in those productivity applications, a 1GHz Pentium III is pretty competitive, and certainly the AMD Athlon is superior, even though the Athlon is 300MHz slower,” Krewell said.
According to Intel, the Pentium 4 was not designed with existing apps in mind, but with those still to come. E-commerce- and Internet-related apps, data mining, digital photography and gaming programs are the kinds of applications that will benefit from the new architecture of the Pentium 4, Udashkin said.
“I don’t know that you see a benefit with the Pentium 4 running a word processor any faster. I don’t really think that that’s the point. The point of this is really, what are those applications up coming? We architected the processor around those apps.”
et tu Transmeta?
But Intel was not the only chip maker to make recalls last year or face questions about the benefits of its chips. Santa Clara, Calif.-based Transmeta Corp.’s Crusoe chip faced a blow last year when IBM Corp. decided to hold off on including the processor in its ThinkPads.
The IBM decision to bypass the Crusoe was unrelated to recalls made by NEC Corp. due to problems with memory inside the processor in a faulty batch of chips. The recall was unrelated to a fundamental problem with the processor itself.
“The Transmeta problem, as I understand it, is more of a straight-forward issue of they were making some changes and didn’t get all the changes tested in time. Some of the chips slipped out the door. And it didn’t really have anything to do with them rushing to market,” Gwennap said.
But IBM’s decision was likely a result of the corporation not seeing any significant power savings from the Crusoe chip vs. what they could get from Intel, Gwennap said.
Krewell agreed. “I believe that Transmeta has made claims they can’t backup in terms of battery life and performance. Certainly they’ve been very reticent to put their part under the spotlight of benchmark performance. And every review I’ve seen so far of the Transmeta part has indicated that standard benchmarking performance of the Transmeta part is on the very, very lowest side. From an IBM perspective, I think the Transmeta part was too high a risk with too little return or reward for IBM.”
Word count: 202
Intel chip trips up Linux install
By Ed Scannell
A handful of top Linux operating systems distributors said their products do not install cleanly on Santa Clara, Calif.-based Intel Corp.’s Pentium 4 because they are unable to identify the chip upon installation.
The installation problem centres around Intel changing its CPUID (CPU identification) model numbering, catching many Linux distributors unaware. The result of this change is that shortly after users begin installation the process is suspended.
Officials at Intel admit the problem and say the affected Linux distributors must upgrade their versions of the operating system to fix the glitch. An Intel spokesperson said that the company told Linux distributors about the update to the chip in nondisclosure briefings before Nov. 20, but some distributors chose not to participate in those briefings.
By most accounts the only version of Linux that properly installs out of the box is Red Hat Inc.’s Linux.
Over the past few weeks several of Red Hat’s competitors, including Caldera Systems Inc. and TurboLinux Inc., have posted multistep work-arounds to avoid the problem.
Other top-tier Linux distributors such as SuSE Linux AG, Corel Corp. and MandrakeSoft Inc. also do not include information in their CPUID databases to identify the chip, although all are aware of the problem.