November 97 News

December 97 News

Monday, december 22:

Santa Claus was a little early this year for me: I received this afternoon a Federal Express box. I was slightly puzzled until I saw the name of the sender: Cyrix Corporation. Opening the box, I found inside a neatly packed Media GX motherboard with a 180MHz processor. Well, thanks guys. :-)

I had asked Cyrix for a sample MediaGX motherboard so that I could test its compatibility and performance under GNU/Linux. My first impression on this motherboard: it's small, but looks very complete. The video, sound and game circuitry are included on-board. The MediaGX processor has a 486-sized heatsink/fan on top, and the motherboard has a switching voltage regulator for the CPU. Just one thing is missing: an L2 cache! I'll begin testing ASAP.

Just a little question for you all to think about during these holidays: if you were going to integrate a Java-capable network-ready sophisticated operating system and a < $500 "appliance PC" (with gross margins of < 10%), would you choose a $60 license of Windows CE or a free, customized version of Linux (with all source code available)?

I doubt I'll update these pages in the next few days, so: Merry Christmas to all :)

Wednesday, december 17:

Some short news:

1. I got an email from Koen in his usual terse style. Quoting:
ADB> Here is set6x86 version 1.5.
...
Got it, tested it, approved it :-)
Executive summary: Works fine for me.
2. AMD is reported to have increased the yield on their K6-233 from 15% in the first quarter to better than 50% in the third quarter 97. Expect much lower prices on the K6-233 soon.
3. I am having a hard time maintaining a dialogue with Cyrix representatives. I sent them an email on November 29 and I am still waiting for an answer. :-(
4. Likewise, I sent two emails to IDT/Centaur technologies about a month ago, to see if they could send me a sample C6 for testing under Linux. No answer yet. Right now my feeling about the C6 is: too little, too late, too expensive and too badly marketed. Note that they are going after exactly the same x86 CPU market segment as NS/Cyrix with their 6x86MX: the low end. But NS/Cyrix is already firmly entrenched in this segment. The 6x86MX PR-200 is 10-15% cheaper than the C6-200 and performs much better, as well as being distributed all over the world (with the distribution network of National Semiconductor). The IDT/Centaur people had better get their act together before they become just an x86 CPU history oddity, like so many others.
5. I already discussed a little bit the market positioning of Intel, AMD and NS/Cyrix. Intel is the leader with an estimated 85% x86 CPU market share and occupies the top part of the price/performance curve with the PII-266 and PII-300, AMD comes second in market share and stands in the middle, NS/Cyrix comes third and is firmly entrenched in the low end. IDT/Centaur is trying to fight NS/Cyrix in the low end but hasn't even scratched them yet. AMD is having a tough time because "being in the middle" is always the worst positioning one can have, but they are going after Intel with their new high-performance CPUs (the K6 3D and K6+ 3D). Brave but foolish? A year from now we should be fixed.
6. Meanwhile NS/Cyrix is missing its deadline for the release of the 6x86MX PR-266, but this does not matter much. The important fact is that NS/Cyrix is finding itself at the right spot at the right moment, namely making cheap CPUs for the < $not_much (it was $1.000 a month ago; now it's $800) PC market. Coincidence? Luck? No. This successful marketing strategy (outsmarting the guys at Intel, who are no slouches) is the work of one man at Cyrix: Steve Tobak, now promoted to Vice-President of Marketing and Public Relations at National Semiconductor. Just a bit of history: initially Cyrix was trying to sell the 6x86 for higher prices than the corresponding Intel Pentium processors. But then Microsoft "detected" a bug in the L1 cache in the early revisions of the chip (2.6 and earlier) and decided that NT would disable the cache on 6x86 processors. Cyrix was locked out of the high-margin, high-performance x86 CPU market and had to completely rethink its strategy. At about that time Steve joined Cyrix (1995), and soon 6x86 chips were being marked with their PR rating instead of their true core clock rate (whereas Intel raised clock rates beyond the capabilities of 6x86 chips). In retrospect I doubt Cyrix would have survived these last two years were it not for its bold marketing strategy. And of course the merger with National Semiconductor is a stroke of genius.
7. I received some results for the Pentium II 300MHz CPU as tested by the latest version of nbench-byte (see the Linux Benchmarking Project site). Both the K6 and the 6x86MX do better than the Pentium II in terms of performance/MHz on all but the floating-point intensive tests.
8. Aleksandr Konosevich sent me an email with some bad news: Computer Week - Moscow used his article in c't magazine and information from my site without even a mention of our names. Perhaps they don't know the meaning of the word "copyright"? :-(
9. The Korean economy is taking a deep plunge these days. Expect much lower DRAM prices in January. But soon after SDRAM DIMMs may be in short supply: it appears there is a shortage of 100 MHz SDRAM testers, with shipping delays > 6 months (sources: Les Echos and Electronique International).

Sunday, december 14:

Set6x86 version 1.5 released, with various small improvements. It is immediately available for download here (26Kb).

The IBM 6x86MX PR166 rev. 1.4 chip I received a week ago has been running at 187.5MHz (2.5 x 75MHz) without any problems. According to Cyrix this is a PR233 rating, but I paid just $85 for this little overclocking-friendly jewel marked as a 2 x 66MHz part.

The money saved over an equivalent Intel part would be almost enough to buy the new 6.4Gb IBM UDMA drive I also mentionned below. But did I say this is the first EIDE drive I have ever seen with a 512Kb cache? It is extremely fast, and runs cool and silent.

Saturday, december 6:

Two important news (three in fact):

1. Koen Gadeyne has passed on the role of maintainer for his set6x86 package to me. Thanks, Koen!
2. Mike Jagdis has agreed to have his experimental patch (which he was not maintaining anymore) merged with mine. Thanks, Mike!
3. Consequently, I am releasing a new version of set6x86, which is in fact a combination patch/utility package for Linux 6x86 users. This combination 6x86 kernel patch/utility combo supercedes all previous patches and utilities.

Set6x86 version 1.4 includes everything one needs for correct 6x86 identification/configuration under Linux:

1. my latest Linux kernel patch,
2. the set6x86 utility,
3. sample set6x86 scripts,
4. a new utility to examine all 6x86 relevant registers,
5. the source for a demo "coma" bug program and
6. two different documentation files.

It is immediately available for download here (24Kb). I will be uploading it to sunsite.unc.edu Real Soon Now.

Also, I have received a new ASUS TX-based motherboard and a Rev. 1.4 6x86MX chip, as well as a new IBM UDMA 6.4Gb EIDE drive. I am using them right now to write these pages. Stay tuned for test results/comments on this hot setup.

Last updated on December 22, 1997.

Copyright 1997 Andrew D. Balsa