Slashdot Mirror
AMD Buys Pre-VIA Cyrix Media-GX Division
An anonymous reader writes "A long time ago, in what feels like a different universe, Cyrix created the first sub-$1000 PC based on a 2 chip solution called the Media-GX. Soon after National Semiconductor bought Cyrix, keeping the Media-GX team and selling the 686MX team to VIA.
In the meantime, the Media-GX team have created the a series of single chip PCs, and a totally new CPU, the GX2. Now National Semiconductor is
selling the division to AMD, which should give it a higher profile and better fab technology again."
Reader jlouderb reminds us of National Semiconductor's Device Girls promotion, "a lame take-off on the Spice Girls," and points to coverage at eWeek of the purchase.
Ugh, can't anyone appeal to common sense instead of sex drive? Make way for bad hardware related puns.
They certainly looked better than the Spice Girls. I bet they sounded better, too, though who cares about that. :)
Perhaps what AMD wants is not their CPU, but the stuff that they've integrated in with it to create a single-chip PC. In a year or so, we might see a single-chip system based on one of the AMD processors.
AMD will not use the 80x86 division of NSM to create a 80x86 embedded processor. 80x86 chips fare poorly in the embedded market, which is dominated by ARM.
This looks like it could directly compete with the EPIA Mini-ITX. I don't know what the fastest media gx core is, but it would be cool if AMD released a Mini-ITX board based around it. Especially if it could be done cheaper than the VIA board.
AMD seems to be finally starting to take chipset design somewhat seriously, plus they finally got a decent third party chipset manufacturer a couple years ago when nVidia signed on. I don't think that they need to make motherboards themselves, but I do believe that they should contract one of the Taiwanese companies to make AMD-branded motherboards.
As for AMD's naming scheme though, I rather like how they are doing things with the Opteron. It's VERY simple but also much more descriptive than using MHz/GHz to describe a chip. You have three numbers, first represents how many of these chips you can pack into a single system (1, 2 or 8), the second represents the core revision (currently only version '4', which I guess is a rather random number to start with), and the last represents relative performance within that core version (0, 2, 4 and now 6).
Personally I find that much easier and less confusing than just MHz. Take Intel's Pentium 4 for example. There were 3 different versions of the 2.0GHz P4 and now 3 different versions of the 2.4GHz chip. The 3.06GHz P4 is faster than one version of the 3.0GHz P4, but slower than the other version of the 3.0GHz P4. To differentiate all these different versions Intel just sticks letters like 'A', 'B' etc. at the end of the clock speed in a fairly random manner. For the 2.0GHz vs. 2.0A chip, the 'A' referred to having more cache and lower power consumption. For the 2.4GHz vs. 2.4B vs. 2.4C chips it's all bus speed (400MHz vs. 533MHz vs. 800MHz). Of course, Intel REALLY out did themselves back with the old 600MHz PIII, where they had no less than 6 different versions of the chip! What's worse, a socket PIII 600MHz chip was NOT the same chip as a Slot 1 PIII 600MHz, but rather the same as a PIII 600E.
Long story short, using MHz/GHz to name chips is VERY confusing in this day and age. Things like cache size, I/O bandwidth and memory bandwidth are of major importance, while clock speed plays only a limited role in many situations.