03/30/2001

A big letdown. A lot of people (myself included) miss the days when you could build an SMP box on the cheap by getting an Abit BP6 motherboard and a pair of low-end Celerons. I read on Ace’s Hardware this week that Via’s C3 processor is SMP-capable. The C3 is a 733 MHz chip, derived from the Centaur WinChip 3 design, that plugs into Socket 370 and costs $54 in quantities of 1,000. Since a lot of places sell CPUs at or slightly below that cost and make their money on shipping, you can expect to buy it for under $60. The drawback with the C3 is weak floating point performance. For applications use that doesn’t make much difference, so for people like me who just want to multitask a bunch of productivity apps wicked fast, it would be nice.

But I was skeptical. Earlier Cyrix and Centaur CPUs (VIA now owns both design houses) didn’t support Intel’s APIC protocol for SMP due to patent problems. So Cyrix and AMD invented their own protocol, called OpenPIC, and prototype SMP chipsets existed but were never commercially released, probably due to lack of demand. AMD attempted to solve this problem by licensing the Alpha processor bus (and therefore its SMP architecture) for the Athlon/Duron and getting into the chipset business.

It was a Herculean labor for VIA to get the rights to use Intel’s P6 bus. I’d be shocked if they managed to wrestle SMP out of Intel as well. But if these rumors were true, it would have ushered in a whole new era of inexpensive SMP, albeit with a slightly limited audience due to the C3’s poor (but better than the AMD K6-2) gaming performance.

But VIA’s site made no mention of SMP. None of the reviews of the C3 or its predecessors mentioned SMP. Finally, I found confirmation of the truth on www.viahardware.com . The C3, in spite of photos of dual configurations originating at Cebit, doesn’t support APIC and therefore won’t do SMP. Bummer.

Want another letdown? OK. The C3 doesn’t do out-of-order execution like every other modern CPU (including even the Cyrix 6×86) does. That’s part of the reason why the C3 struggles to keep up with an equivalently-clocked Celeron, even if the Celeron is running on a 66 MHz FSB while the C3 runs on a 133 MHz FSB. For OOO, you’ll have to wait for the next revision of the processor, due later this year.

The lone drawing point, besides price, for the C3 is its cool operation and low power consumption. It can operate with just a heatsink, no fan. You could team it up with a fanless 135W power supply, a 4400 RPM hard drive (or a very quiet 5400 rpm drive), and an integrated motherboard to have a silent PC. You can’t do that with anything from AMD or Intel. So for quiet PCs, the C3 has an audience.

Hey, someone could take that chip, put it on a microATX board, and put it in a tiny squarish Lucite case with the CD-ROM drive up top, so you put the CDs in like toast in a toaster, and sell the computer on size, quietness, and looks alone. Oh, wait a minute. Someone already tried something like that.

You’ll also notice VIA is scrapping the Cyrix brand name, which is probably a good move. Cyrix chips weren’t bad; they weren’t ideal for 3D gaming but for everything else they were a fine chip. Cheap and fast. Unfortunately they were usually paired up with very cheap and very low-quality hardware (particularly cheap power supplies) and when the systems had problems, everyone blamed Cyrix. But my friends and I, pairing Cyrix CPUs up with Abit, Asus, and AOpen motherboards and Diamond video cards and Creative or Ensoniq sound cards, never had any problems whatsoever with the CPUs.

Discussion groups. I’ve often longed for the days of the old-style BBS. I never ran a BBS myself–in the golden age of BBSing, I was just a teenager, and a good BBS required a US Robotics dual standard modem, a 386, and a gigabyte hard drive, all of which could easily set you back $2,000.

The Internet has so many advantages to those BBSs. When you dialed in, it was very easy to spend an hour online. In the meantime, no one else could use the BBS. With 24 hours in a day, even with an average call length of 15 minutes, fewer than 100 people would get in, and that makes it hard to facilitate meaningful discussion. It happened, but unless the BBS was part of a network, the communities stayed small. The Internet doesn’t have those disadvantages. The line’s never busy (if you’ve got a decent ISP at least), so the community can be much larger.

The discussion groups facility on this site have always been very under-utilized. I think a grand total of four people have posted messages here. That’s largely my fault; I never configured the discussion area, nor did I ever get rid of that stupid skull and replace it with something intuitive (like, say, the word “Discuss…”). I started looking into configuring it, and lo and behold, it’s possible to create a nice discussion board with Manilla. The interface is a little different from UBBS, which seems to be what most of the popular discussion groups of today use, but it’s not bad.

Like most other online bulletin boards, you have to be a member to log in and post. There is no charge to be a member. Let me emphasize that. There is no charge to be a member! Understood? Excellent. There’s also no validation process, none of that other stuff. Manilla does maintain a database of members that I can look at. I’ve looked at it once. I just don’t have time to go snooping around there. I’m too busy to invade your privacy.

Non-members can read messages. Messages posted are indexed by this site’s search engine. It’s really nice.

To become a member, click Join Now to the left. It will ask for an e-mail address. That address is used for two things. If you forget your password, your password hint is mailed to that address. And optionally, you can get your daily (or more, if this board gets popular) dose of the Silicon Underground e-mailed to you. Probably most people will turn that option off. If you’re concerned about spam, or concerned about privacy, feed it a bogus e-mail address. Tell my site you’re billg@microsoft.com or something. I really don’t care. Honest. (A lot of Web robots seem to have problems navigating Manilla sites, so spam harvesters may find this site more trouble than it’s worth, but I can’t make any guarantees.) And if you want to use a handle, that’s fine too.

Discussion groups get their own calendar. When you click on March 29, 2001 in the calendar, you get that day’s messages, plus the rest from the previous week. If you just want to see just that day’s topics instead, click the link that says Chronological View, and it’ll switch. Sorry, I don’t know how to make that a preference that gets saved for you.

The advantages of a discussion group are many. First, this becomes more of a community and less something that’s all about me. When you want to have your say, you can just log in and respond and it’s instantly there. When you e-mail me, I won’t see it until I get home, and then I may or may not post it, depending on a number of factors. When you post, if someone else sees it first, they can respond. So if you’re having a problem and need a quick response, someone else may see it and respond before I get to it.

You’re still free to e-mail me of course, but I had this resource here and it’s really a shame I haven’t been using it. I’ll continue to respond to mail and have it posted, for those who prefer a more moderated discussion (a small few, if page reads are any indication).

You can get to the discussion groups at least two ways. You can click on the Discussion Groups link to the left. Or you can click the Discuss link at the bottom of a message.

Here’s hoping this will become a valuable resource.

03/19/2001

Another useful hidden utility. If you’ve never used Sysmon.exe, remember it. With Windows 98 and newer, you can use it to track CPU usage, memory usage, and disk throughput (usunted information, I always searched the Web. When I wanted useful information, I hit DejaNews. Sure, there was a lot of junk out there, but 50% of it was good stuff, and most of that never made it onto the Web. I never did find any useful information on the Asus SP97V motherboard on the Web, because the hardware sites weren’t into it. I found out what I wanted to know about it from DejaNews. When I wanted to know how to get Windows NT Workstation machines to authenticate against an OS/2 domain, I found out how on DejaNews. When I needed information about XTs and ATs for some insane reason, I hit DejaNews.

I spent a little time in the comp.sys.ibm.pc.hardware hierarchy for old time’s sake yesterday. I’m sure I’ll get a ton of spam now because I probably didn’t spam-filter all of my e-mail addresses, but that’s OK. It was pretty fun. I’ll have to do it again someday soon. It’s the closest thing I can find to an old-style BBS that still exists and has a sizable community. The scary thing is, some of the old WWIVnet message boards had a bigger community than comp.sys.ibm.pc.hardware seems to have. The questions I answered were hardly difficult ones, and some of them had been sitting for a couple of days, which never would have happened on WWIVnet. And I know WWIVnet wasn’t even the biggest of the BBS networks, it just happened to have a lock on the St. Louis market in the late 1980s and early 1990s. I wonder where all those wizards went?

I ought to install a text-based newsreader on my Linux box to give myself a more authentic atmosphere though. This stuff just doesn’t look right when it’s running in a GUI. Not to me at least–back when I was dialing up BBSs, nobody ever ran Windows. At the very least, it should be running in a terminal window. Hmm. Maybe next time…

Even if the community is smaller, Usenet does have one big advantage over the old-school BBS though: No busy signals.

Monitor buying tips

Monitor time. Looks like I’m in the market for a new monitor now. I was talking about monitors the other day with a Web developer, who observed that when you buy a good monitor, it tends to stick around forever–the CPU just changes. He’s right.

Read more

Wednesday mailbag

Short stuff today. There’s mail on nostalgia and my infamous books today, so let’s get right to them.
———-

From: “Jeff Hurchalla”

Subject: optimizing windows – ramdisk

Hello Dave,

First I wanted to thank you for the great book on optimizing windows. I’ve read it all now and picked up quite a bit. I think the book nicely fills a niche for windows tweakers- I know I was looking for it for a while. One thing I noticed, if you do a second ed. you might include information on hard disk setups -how IDE and SCSI work, RAID, and what cable and role(master/slave) to use for that new second IDE hard drive.

One tip to add could be the format /z command when using fat32. This undocumented /z switch sets the cluster size in number of sectors.. so “format /z:64” would create 64 sectors/cluster or 32k clusters. I got this from prorec.com a while back. You might want to check out their site just for the music recording discussions, never mind that they do a great job describing how to optimize for digital audio(I think the advice is generally applicable for any computer).

I really want to ask you about a problem I’m having when I run windows95(osr2) from ramdisk. I have it working, mostly as described, but user.dat and system.dat are pesky. I had to put these both into the ramdisk hard drive directory(the one with himem.sys. setver.exe, etc) to get it to boot. Io.sys looks at the registry early in the boot process way before it gets to autoexec.bat(where the ramdisk is created). So it’s not trivial to find a way to create the ramdisk and get the registry in there, where I think it belongs, in time to satisfy io.sys. The solution to put the registry in the ramdisk directory isn’t very good because windows constantly updates the registry(and usually for no apparent reason). Do you have any ideas for how to get the registry in the ram early, or how to get windows to switch to using the registry in the ramdisk later on? It would also be ok, though maybe not as satisfying, to somehow set windows to stop automatically updating the registry just because it’s apparently noticed 15 seconds of time has passed. I also thought to use lilo (possibly modified) to do some work with a ramdisk and then load a bootsect.dos, but I don’t know enough to tell if it could work. I used to have linux set up, but everything’s standard right now.

I read in your views that the outlook for your linux/win2k book is not very good. That’s a shame. I’d really like to read it in the future. I hope something works out.

Take care, Jeff

———-

Well, first off, thanks for the compliments. It’s things like that that make the time and effort that go into writing a book worthwhile.

I had forgotten the FAT32 cluster trick; that is a good one but you might as well just use FAT16 if you’re going to make big clusters–unless the 2GB limit bites you. So, yes, I can see it being useful.

I believe I did talk about hard drive setups, but I may have neglected it. Rather than look it up, I’ll be lazy and just state it: Hard drives should always be masters. If you must put two drives on a channel, make the newer, more modern drive the master. RAID would certainly be a good topic, and one that’s sorely missed from the book. I think I did a decent job of predicting some of the up-and-coming stuff, but two things that I missed were IDE RAID and USB networking. They didn’t exist in the summer of 1999 when I wrote the bulk of the book, and I didn’t anticipate them.

The chances of a second edition appear to be pretty slim, unfortunately. I could write another book for another publisher, provided I’m very careful not to violate the copyright I sold to O’Reilly, but I would almost certainly burn some bridges by doing that and that’s not exactly something I want to do, whether I agree with how O’Reilly marketed the book or not. I’m almost hesitant to say anything about it or even mention that possibility.

As for the Linux/W2K book, it’s tough to say what to do with it. There are a lot of things I’d love to say about it, but again, I’d burn some bridges that I probably shouldn’t. A year ago when it looked like both W2K and Linux 2.4 would be released within six months, it looked like a blockbuster book. With 2.4 delayed and W2K failing to take over the world, it’s becoming a tough sell, and frankly I’m getting really sick of the topic. I also know I tend to take some unpopular stances on the issues involved, which probably doesn’t help. If I thought the book would cover the expenses I’d have to outlay to write it, I’d be much more inclined to do it. So far for me, writing for O’Reilly has proven to be a very expensive hobby. (And if anyone from O’Reilly reads my site and doesn’t like hearing me say that, tough.)

I’ll probably re-evaluate after the 2.4 kernel starts firming up some more, and once my wrists start looking like they’ll hold up or I get the hang of NaturallySpeaking. As for my next book, if there is another one, I don’t want it to involve Linux, and I want to write for a small indie publisher who only releases a half-dozen or so books a year.

As for the registry problem in your ramdisk, did you make the modifications to msdos.sys? Specifically, the line WinBootDir parameter must point to the ramdisk, while WinBootDir points to your directory containing the boot files.

The only other thing that I can think of that might cause those symptoms would be if when you installed Windows to the surrogate partition, Setup may have found your old registry and grabbed some data from it, forcing you to keep the registry on your HD rather than in ram.

I messed around with the ramdisk trick for an entire weekend (literally–I didn’t do anything else for two and a half days) getting it to work right, and for all I remember, I may have spent a few evenings preceeding that weekend working on it. I do know this: Microsoft never intended for anyone to do that, which of course just made me all the more determined to do it.

Let me know if that doesn’t fix the problem, because I’m really curious now what may have caused it. I’ll think on it some more.

———-

From: “Gary M. Berg”

Subject: Pascal on the Mac

You can always pick up Turbo Pascal 5.5 from Borland for free: http://community.borland.com/article/0,1410,20803,00.html

———-

Thanks. For that matter, Borland made older versions of Turbo C and Pascal available for DOS too, which isn’t bad for quick-and-dirty stuff. I found the older versions of Turbo Pascal for DOS produced smaller, tighter executables than the later versions (I mostly used version 7, back in 1992-93). And I believe at least one of their Windows C compilers is available for free now too.
———-

From: “Brent Dickerson”

Subject: voice recognition stuff

In case your interested in reading another’s short eval. The link may not work after today. brentfd

http://www.gazette.com/weekly/ibiz/biz9.html

———-

Thanks. That’s the first positive piece I’ve read on VoiceXPress, but their findings on ViaVoice are certainly consistent with my experience and with what I’ve read elsewhere–though ViaVoice improves considerably with a good mic. I still prefer NaturallySpeaking though, and I get better results than he did, but again, with a highly recommended mic and sound card combo.
———-

From: Robert Bruce Thompson

Subject: RE:

Hmm. When I was in high school, the first integrated circuits were still in R&D labs. That was 1971 or thereabouts.
———-

By the time I was in high school (1989-1993), “build your own computer” meant driving up to Gateway Electronics, buying a 386sx motherboard, I/O card and video card, some memory and a case, then assembling it all yourself. This was slightly more exotic than today when you can go down the street to get your parts.

I was an anomaly in that I owned an Amiga, whose system boards didn’t change much and were very well documented, so there were literally dozens of hacks available on BBSs to add features to them–most involved at least a little soldering and some involved discrete components. So that’s where most of my experience with discrete components comes from.

Most twentysomethings like me never had to build our own circuit boards and can’t imagine doing so. I was surprised to see modifications requiring soldering to start popping up on the hardware sites, because I know an awful lot of overclockers are my age and have never soldered delicate electronics before.

I dated a girl a few years ago who vaguely remembered her dad building a computer from a kit in the late 1970s. He’d tell her what component he needed, and she’d find it from the pile and give it to him. With me being a computer professional, she was pretty proud of that. She knew not many people in our generation had even that experience.