27th > June > 1999 Archive

The Register breaking news

Strange AMD litigation pops up from nowhere

AMD is sueing Singapore Airlines. The company filed the case at the end of last week in the US District Court in San Francisco. The case seems to circle round a negotiable instrument. Whether this is a blunt instrument or not is unclear. This one's bizarre, isn't it? We'll attempt to find out more... ®
Mike Magee, 27 Jun 1999
The Register breaking news

API publishes Alpha Slot B benchmarketing

Correction Alpha Processor Inc, a company partly owned by Samsung and Compaq, has released benchmarks for the Slot B technology we wrote about eight days back. The company's benchmarketing is here. *Blush* we published the wrong URL link before -- this is now corrected. Sorry API and readers.. On the face of it, the Alpha technology appears to trounce the Intel technology. Now API/Compaq just needs to get its marchitecture right too, if the benchmarks are to be believed. ®
Mike Magee, 27 Jun 1999
The Register breaking news

Compaq, Winbook sued for patent infringement

An individual has taken out a summons against Compaq, Winbook Computer and Chem USA for alleged patent infringement. The case, filed by Pao-Chin Chen on the 22nd of June, was filed in the US District Court for the Southern District of Texas, that is Houston. A preliminary date for the case was set by Judge John D. Rainey for the 7th of October this year. Chem USA and Winbook specialise in the supply of notebook computers, although their relationship with Compaq is unclear at this stage. ®
Mike Magee, 27 Jun 1999
The Register breaking news

Intel busts Via over PC133 – this time it's personal

Intel has gone legal again in its attempts to block shipments of Via Technologies' PC133 chipsets. Chipzilla, which filed against Via in San Jose in March owing to a clerical error, filed against it there again last week. Presumably experience of the March dummy run will help accelerate the legal process. Intel is generally held to be trying to slow the progress of PC133 while it either implements its own version or (unlikely one this) hauls its own favoured Direct Rambus out of the mire. Speaking in London last week Craig Barrett made it pretty clear that PC133, which is still officially ruled out, isn't really ruled out after all, but he's been making that pretty clear to anybody who'll listen for some time now. Here's one we prepared earlier. The latest Via lawsuit, which we're presuming won't get pulled again on Monday, claims breach of contract, false advertising, patent infringement and (choke) unfair competition. The row itself revolves around the licensing deal Intel did with Via last November. Since March Intel has been claiming that Via is in breach of this, and has warned off both Via and its customers. Intel contends that Via chipsets using a 133MHz front-side bus or 4X AGP aren't covered by the licence, and this is an obvious problem for Via's Apollo Pro133, which has a 133MHz bus and supports PC133 memory. There's clearly been some tough wrangling between Intel and Via since March, and the latest nuclear strike (which also includes the revocation of Via's P6 bus licence) means they've broken down completely - probably. Intel is not noted for the subtlety and sensitivity of its negotiating style, so there's still a slight chance the company is just dropping what it thinks of as a gentle hint. But by bizarre coincidence last week, Intel granted a licence to Acer Labs Inc (ALi) to sell chipsets compatible with its P6 bus architecture. As SiS has one of these as well, the move clearly puts pressure on, and potentially isolates, rival Via. So you might reckon that by taking pot-shots at Via while leveling the playing field between the other two, Intel is both encouraging Via to behave or lose business, and discouraging Via's two competitors from going chasing down the same road Via did. A sort of carrot and ICBM approach designed to keep the chipset boys in line while Intel re-tarmacs its roadmaps. By the way, we're tickled to note that Intel is of the view that Via manufacturing its chipsets at NatSemi, which has a broad-ranging cross-licensing deal with Intel, doesn't put Via in the clear. Humorously Intel recently went down in the Alabama courts against Intergraph, having claimed unsuccessfully that Intel's use of Clipper technology was OK because of, er, its broad-ranging cross-licensing deal with NatSemi. ®
John Lettice, 27 Jun 1999
The Register breaking news

Linux loses in NT tests – Mindcraft numbers still wrong?

PC Week Labs has unveiled the results of the latest round of the Linux versus NT benchmarks and, as generally expected, Linux lost. But the results of the five day rematch also tend to indicate that the initial Microsoft-funded Mindcraft benchmarks weren't quite all they should be. After the publication of the initial Mindcraft test results earlier this year various Linux luminaries criticised the way they'd been set up and the class of hardware they'd been run on. The Mindcraft tests, it was argued, unfairly hobbled Linux and ought to be re-run on a level playing field. That was the point of the PC Week Lab tests, which had Microsoft, Red Hat, Penguin Computing and Mindcraft representatives present to ensure fair play. Mindcraft's tests had claimed NT outperformed Linux as a file server by 2.5 times and as a Web server by 3.7 times. PC Week, bless 'em, doesn't seem to draw attention to the point, but its Labs results come up with NT being 2.3 times faster as a Web server (IIS 4.0 versus Apache) and around about the same as a file server. They also found that Linux performance didn't collapse when client load exceeded 16 computers, as Mindcraft's Apache benchmark had found, and the performance difference was a lot narrower on a single CPU box, although NT still won. The performance difference here was 41 per cent for Web server and 52 per cent for a file server. The PC Week Labs results also arguably understate their divergence from the Mindcraft ones in some areas. Earlier PC Week Labs tests (still however conducted after the Mindcraft test) had found Linux-Samba could outperform NT when NT Workstation clients were used. MS performance engineers turned loose on NT after that found performance improved if four NTFS partitions, each with their own transaction log, were used rather than one. So NT's performance in the latest tests benefited from this discovery. PC Week also reports that the lack of a multithreaded IP stack in the Linux networking subsystem causes a performance plateau in the OS, and the charts in the report are particularly interesting. They show fairly conclusively that Linux was barely able to take advantage of multiprocessor boxes at all, so this is a hill the OS needs to climb soon. PC Week Labs Report. ®
John Lettice, 27 Jun 1999
The Register breaking news

Corel compensated over unfair MS Office contract

Here are some interesting numbers. Corel has won C$9.9 million compensation over a Revenue Canada contract, the bidding for which Corel successfully argued in court was skewed in Microsoft's favour. But Microsoft Office got the contract via GE Capital anyway - if all of the options are taken up, then the deal will be worth C$55 million. So Microsoft must be crying all the way to the bank, right? Well, up to a point, Lord Copper. The contract in question, which Revenue Canada put out to tender, is an interesting example of where integration blurs into immolation. Revenue Canada was already an MS Office shop, and had asked companies bidding for the deal with software other than MS Office to include estimated conversion costs in their bids. A series of Canadian courts ruled that this was unfair, and that Corel hadn't been given enough information to be able to submit a bid for WordPerfect. Revenue Canada was told either to reopen the competition or pay compensation. The deal certainly remains a win for Microsoft, but Corel's legal victory represents a shot across the bows for the Canadian government - its departments will be less likely to be quite so cavalier in 'open' tenders in the future. And of course you could view it as an awful warning for business in general - when you're getting into this integration thing, remember to figure out your likely exit costs first. Of course, as the majority of businesses are now already in there with Office, and standing on the old upgrade ratchet escalator, that's not a lot of help to them, is it? ®
John Lettice, 27 Jun 1999
The Register breaking news

AMD posts K7 benchmarks against PIII

Chip company AMD has now posted a set of benchmarks comparing its Athlon K7 processor against Intel's offerings. AMD figures claim that a 550MHz K7-Athlon with 512K cache registers 146 per cent against the 100 per cent of a PIII-550 using the SPECfp_base95 FP benchmark. It registers 140 per cent against the 100 per cent of the same processor in 3D performance, claims AMD. And even on integer performance the 550MHz K7 beats Intel's part. Go here for AMD's figures. We await Intel's benchmarketing and competitive response with some interest. ®
Mike Magee, 27 Jun 1999
The Register breaking news

Games sites push 3DNow support

A coalition of Web hardware sites has put its weight behind the 3DNow! instruction set supported by AMD, IDT and Cyrix, and wants software developers to write to the standard. The group includes JC's PC News, Linux 3D Gaming, FullOn3D and AMD Insider so far. The ginger group wants to push development of applications and hardware that takes advantage of the instruction set. Go here for more information. On a wider point, The Register will look into the type of speed boost users can expect when SIMD instructions such as 3DNow! and Katmai New Instructions are used. ®
Mike Magee, 27 Jun 1999