ECS and SiS
By Allen Crider
Date: March 12, 2002
I was astounded to read about ECS's 20% greater Athlon vs. Intel-P4 mix of motherboards and it really shows the success of their K7S5A and their all-in-one K7SEM.
But the actions of both ECS and SiS indicate their success was more Forrest Gump-like than execution of thoughtful excellence.
First off, take a look at the ECS front page http://www.ecs.com.tw and you will see them pushing SIX new P4 motherboards and only two new Athlon motherboards. Is THAT the best way to handle success? There is no mention of any new KT333, their KT266A board is not played up on the front page and it seems they have dropped (or greatly delayed) their SiS740-based motherboard. I'm very puzzled why they would drop their winning hand and go fishing for something else.
Secondly, I've followed the new K7S6A, and ECS has done some pretty strange things. They added the ability to change the multiplier of the Athlon, but AMD's new organic packaging in the XP series has put quite a damper on the overclocking market. People aren't as interested in overclocking anymore, and since the K7S5A didn't overclock AND it was a huge success, why not go with WHAT THEY HAD? Or maybe make two models, one without OC and one with (and higher price/profits). Another issue with overclockers is that no one can get a 166/333 FSB on these new boards. This eliminates the REAL advantage of supporting PC2700 RAM. ECS simply didn't bother to engineer their board to support a 333FSB. Also, the new K7S6A doesn't have onboard LAN, which more than 85% of the K7S5As had. The issue is mostly SiS's fault and I'll address them below.
The very, very minor performance boost of the K7S6A makes it a no-brainer to get the older K7S5A with onboard LAN at $50 compared to about $65 (now) for the K7S6A. SiS's very high-speed connection between the Northbridge and the Southbridge (Mutiol) make this board the very best 32-bit low-end server solution. I've seen one running SCSI RAID with two gigabit ethernet cards. We ping-flooded both gb ethernet ports and at the same time ran disk-io benchmarks. The results were excellent--especially when compared to a VIA KT266A-based board. This Mutiol technology does breathe new life into 32-bit PCI.
I believe the K7S6A is going to be an ECS dud until AMD releases real 166mhz CPUs and the engineers revise the board to support this FSB.
Thirdly, SiS is doing some amazingly stupid things with their chipsets. They upgrade their successful 735 single-chip line to the 745 and replace the on-chip LAN with Firewire? Firewire!? What in the world were they thinking? Didn't any of their clients try and talk sense into them? Boggles the mind! I wouldn't be surprised if more than 65% of all 735 chipset motherboards are sitting in offices. Maybe firewire would be nice for a mobile chipset, but this dropping of LAN is not rational.
Also, their replacement for the 730 series, the SiS740, was delayed while they ramped up P4 chipset production. I'll let them have that call as they may have had many orders up front for the Pentium chipset, but now they are facing the imminent appearance of VIA's KM266. The 730 series was a very successful, reliable (if boring) product.
Lastly, SiS has closed their doors to documentation of their chips. It used to be that one could browse their website and download nice PDF databook files for all their products. SiS used to be active in the Linux community, and even assigned one of their best software engineers to the LinuxBIOS project. But I remember hearing rumors about VIA stealing some of their technology that made such a big difference in the performance between the KT266 and KT266A. Could that be the reason SiS has clamped down so hard on documentation?
The result is SiS-based products have no use running Linux. There are no drivers for the newer SiS AGP bridges, and no chance of getting X-Window drivers for either the SiS315 or the 740-650 all-in-one chipsets. That could be a big hit on low-end X-Terminals and diskless workstations which are gaining a huge following in 3rd world countries. I got a reply from them last weekend about documentation of the SiS315. They want an NDA. Nobody gets to write open-source code under an NDA.
So I predict that both SiS and ECS could be back to their usual selves by this time next year. They seem to be unable to run with their success, unable perhaps to recognize what it was they did right in the first place!
What a rant... :-)