Dr. Jerry Pournelle

Email Me


Why not subscribe now?

Chaos Manor Subscribe Now



Useful Link(s)...

JerryPournelle.com


Hosting by
  Bluehost


Powered by Apache

Computing At Chaos Manor:
December 5, 2006

The User's Column, December, 2006
Column 317, part 1
Jerry Pournelle jerryp@jerrypournelle.com
www.jerrypournelle.com
Copyright 2006 Jerry E. Pournelle, Ph.D.

This will probably be posted Tuesday, December 5, 2006 in keeping with the new schedule. The Mailbag will be up Monday. I will probably continue this schedule. It makes my weekends a lot less hectic. If anyone seriously objects, please let me know.

Subscriptions have been encouraging. I can use more. With enough subscriptions I can encourage others to make regular contributions. Note that a subscription to Chaos Manor Reviews is also a subscription to www.jerrypournelle.com where I have more to say on many issues, and where we have a large and expanded mail section. Of course this is a rotten time of year to be asking people to pay discretionary income.

Christmas Gift Suggestions

This is the time of the year - actually I am a bit late - when I solicit suggestions for Christmas Gifts. Please send your Christmas gift selections for next week's column. Merry Christmas.

Annual Chaos Manor Orchids and Onions

It is also time for you to send your nominations for the Annual Chaos Manor Orchid and Onion Parade. If you have several nominations, please send a separate email for each. Nominations may be short, or detailed; use your judgment as to how obvious something is, and by all means include as many details and opinions as you think relevant. Please put the word Orchids or the word Onions in the subject of each nomination.

You may also nominate products for Chaos Manor Users Choice Awards. Please put the words User's Choice in the subject, and one nomination per email. You may send as many nominations as you like, although several in a single category will probably cry for explanation.

The 2006 Users Choice Awards, and the 2006 Annual Chaos Manor Orchids and Onions Parade, will be given in January, 2007. I weigh rather than count nominations, and the final decision is mine (after discussion among my advisors). All nominations must be received before the end of December, 2006. I will guarantee to read them all, but I am not likely to acknowledge them.

Quest for Power

I have four AMD systems. Two are now working properly. A third may be, but it's a bit early to tell.

The first is Alexis, my main communications computer, an AMD dual processor on an ASUS A8N32-SLI Deluxe motherboard. That works splendidly. The second is Satine, the system I am writing this on. Satine is a single processor system built on an ASUS A8N-SLI Deluxe. I told you about her in the final November column.

The third is Dianne, also AMD dual processor, also ASUS A8N32-SLI Deluxe, and until yesterday utterly unreliable. She's working now, and a story goes with that.

The fourth ran the February build of Vista reasonably well but had unpredictable catastrophic failures which I ascribed to Vista, but now I am not at all sure. I'll get to that one later.

You will recall that last week when I tried to revive Satine, one of my attempts got the result that she'd start up, but I got error reports: memory test failures, and other problems. For details see the column. The point of the story was that taking Satine off Antec power supplies and installing a medium price Rosewill power supply caused her to work perfectly. No more memory errors, no more video errors, no more disk errors: she just works. Note also that this is the system that formerly ran a test version of Vista until the errors got so painful that I used Ghost to restore her to XP. She had an Antec power supply which subsequently failed catastrophically (again see last week for details). It's possible that power supply irregularities caused memory errors which caused the Vista problems. We'll never know.

I speculated that if Satine's startup memory errors could be cured by using a non-Antec power supply, perhaps Dianne could be cured also; so I got another Rosewill, installed it in place of the Antec Neo HE, and Lo! Dianne came up and began working. I scrubbed her down to bare metal and installed XP, and so far she has been working just fine. We'll see how long that lasts. Dianne was originally intended to become my main writing and games machine, but Satine is doing that so well I have no real incentive to replace her, and Satine has the great merit of being built into an Antec LANBOY Lan Party case, meaning it's very easy to fling her in the back of the Explorer and take her down to the beach house in San Diego, or off on a writing trip in the desert with Niven; since she's been the main writing machine, I don't need to do much before leaving.

Which leaves Dianne as the experimental AMD machine for Vista, assuming that changing power supplies has in fact cured all her problems. I'll know more on that after a week or two of hard use with XP.

Until I began experimenting with new power supplies, I had tentatively concluded that the problems were with ASUS motherboards and/or AMD CPU chips. This didn't seem reasonable, which is why I kept that conclusion to myself until I understood things better. It is now clear to me that the problem is power supplies. ASUS/AMD may be more sensitive to power quality than Intel systems, or it may be that the issue is specific to Antec power supplies. In any event, feed ASUS/AMD systems the proper power and they work.

I will probably replace the two Rosewill power supplies, most likely with PC Power and Cooling equipment of the proper wattage. That will be a precaution: I have no complaints about the Rosewill power supplies, but given their price I don't think they can be as well built as PC Cool systems, and either Satine or Dianne will be my main writing machine, and thus deserving of premium equipment.

Quest for Power, II

I have been making inquiries about Antec power supplies with AMD CPU and ASUS motherboards. There is no doubt that there are serious problems; serious enough that absent new data I will not again use an Antec power supply with an ASUS motherboard.

I want to be careful here. Antec builds great cases - it was their cases that got me trying their power supplies, because Antec cases are far better than PC Power and Cooling cases. By better I mean more convenient for construction and maintenance. I have no hesitation recommending their cases.

Nor do I know of any general problems with their power supplies; indeed I have good evidence that they build excellent power supplies with heavy duty components. Most power supply makers rate their wattage outputs with the power supply at 25° degrees C; PC Power and Cooling measures theirs at 40°; Bob Thompson tells me that Antec measures their output at 50° C. They use top quality components. The workmanship is good.

But they don't work with ASUS motherboards, and the word is getting around. Even computer gaming magazines, heavily dependent on advertising revenue, are beginning to hint at problems. One says "Few are as well engineered as Antec's NEO HE (which can be problematic in some configurations)". Careful web searches produce the same results: well engineered, good workmanship, but "problematic in some configurations."

In my own case, two of the Antec NEO (not NEO HE) simply failed. Antec no longer sells that model, so if that were all the data I had I would not bother with this story. Alas, two more NEO HE power supplies caused AMD systems on ASUS motherboards to report memory errors, and a third wouldn't let the system boot at all. The systems in question are now running just fine with Rosewill power supplies. Rosewill is an average quality power supply, not premium like Antec.

There's something weird going on here. I've got inquiries in to Antec, but apparently they're busy there. I am asking for a selection of their power supplies in the proper wattages to see if they work properly; I am willing to change power supplies on working systems to make these tests. In particular, I would like to see if a new Antec power supply cures the problem of my fourth AMD ASUS system: that one was used to test the first Vista release candidate, and I had far more problems than my colleagues reported.

Alexis the communications machine, exposed.
Alexis, my main communications system. ASUS A8N32-SLI Deluxe motherboard, AMD dual processor, Antec case and power supply. [View larger]
The rubber band mounting scheme from Antec.
Antec's new superquiet drive mount. One of the drives is still mounted with the big rubber bands. I had to remove the other because one of the bands broke. I love these Antec Performance One cases: they are very easy to work with. I do not recommend the rubber band drive mount system. [View larger]
Drive in tray, and the beloved Screwball ratchet screwdriver.
Hard drive installed in mounting tray ready to go back into the computer. Note the S/V Tools Screwball ratchet screwdriver. This ball handled screwdriver has been one of my most useful tools for nearly thirty years. If you ever have a chance to get on, do so immediately. [View larger]

Note that I have several Intel systems with Antec power supplies and I have had no problems with them; also, there is Alexis, the main communications system. Alexis has AMD on an ASUS motherboard. Clearly some Antec power supplies work with that combination.

This isn't the end of this story, but for now, if you are contemplating an AMD ASUS system, I cannot recommend that you use an Antec power supply. And as always, I do not hesitate to recommend PC Power and Cooling power supplies.

Antec Shock Mounting

Antec has recently been providing an "extra quiet" mounting system for hard drives. These look like big rubber bands. You mount the drive by twisting them into these rubber bands having first removed the drive trays that also come with the case.

When I opened Dianne to replace her power supply, I found that one of those rubber bands had broken, and one hard drive was dangling from a single band. It hadn't harmed anything. The drive still spun up and the data cable was attached. Still, needless to say, that wasn't a very satisfactory situation.

Dianne had not been mistreated. She had been at two work benches, and laid on her side a couple of times, but otherwise she'd never been transported. I have no idea what broke that big mounting band; it certainly wasn't abuse of the system.

It did not take long to put that drive in a standard mounting tray and reinstall it properly.

Given this failure, I do not recommend using the big rubber band system for mounting hard drives. Stay with mounting trays. They're not much noisier, and they're sure a lot more secure.

Live OneCare Settings

It has happened three times now. I have a system that works without problems. It is running Windows OneCare as the only firewall/virus/spyware protection other than the D-Link Gaming Router that stands between me and the Internet. Then I upgrade Live OneCare and the problems begin.

Nearly all of my systems have been protected by beta versions of Live OneCare with automatic updating, and I have no problems to report about that. When the beta period was ending, I bought a copy of Live OneCare by downloading it and paying the fee. There were no problems then, either; at least none I know of.

The difficulty came when Alexis crashed. I told that story last month. In essence Alexis had to be scrubbed to bare metal, hard disk reformatted, and XP reinstalled. As part of the reinstallation I downloaded the free 90-day copy of Live OneCare. Still no problems.

Then Satine crashed. After I had her restored, she was running the beta copy of Live OneCare. I decided to update; the license allows you to install it on three machines. I did the update installation. Live OneCare was happy. And now my problems began.

Satine and Alexis could see each other, but neither could access the other. Each machine could send and receive files from all the other machines on my local network; it was only when I tried to get them to access each other that communications failed, with the message that access was denied. Eventually that was fixed by going into the advanced settings - see last month for details.

Next I decided to update Live OneCare on Alexis. Up to then she'd been running on the 90-day free download. I upgraded by using my Passport connection so that now I was on paid status.

The instant I did that, I could no longer send mail. Outlook reported that it could not connect to my outgoing mail server. A quick check with my ISP revealed there was nothing wrong at that end, and indeed, Alexis could go to Internet Explorer, Web Mail, and send messages that went out just fine. She just couldn't send mail through Outlook. Mail and replies were piling up. The outbox was full, and Outlook couldn't access the outgoing mail server. This was driving me mad.

The remedy is Control Panel | Network | Local Area Connection | Properties | Advanced, and keep drilling until you find a list of things that can be allowed. These include the mail server. I set those and my mail service started again.

In other words, Live OneCare changes the firewall settings when you upgrade it, and it does not tell you it has done so. Worse, doing the proper settings is not simple. I doubt that Aunt Minnie will be able to figure it out: all she will know is that she can't send mail. I am waiting for this to happen to Larry Niven; then either Eric or I will have to go out and change his settings, because I doubt I can explain it over the telephone. It's that complex.

Microsoft, are you listening? This is serious.

Delayed Write Error

The crazy fight to tame the Windows firewall was bad enough. The response from the Microsoft Live OneCare rapid response team was, well, not very satisfying. The responses on the Delayed Write Error bug have been non-existent.

The Delayed Write Error can potentially destroy your files. It is as if every time you take your car out there is a one in one thousand chance that the wheels will all fall off. I have been bitten by the Delayed Write Error three times now (that I have logged) and while two of those "only" affected an external backup drive (requiring me to start over on backups) the third actually wiped out the registry on the C: drive on Alexis, and eventually caused me to nuke the system, reformat, and reinstall everything.

As the farmer said when he saw the two trains racing toward a head-on collision, this is a hell of a way to run a railroad. An operating system that can wipe you out sometimes - in my case, while I was trying to make a backup restoration point! - is dangerous.

I don't know if Vista has corrected the problem. I am fairly sure that Microsoft doesn't know what causes it in XP.

XP, Vista, and the Microsoft OS Product Cycle

I had an interesting thought the other day. Think of it as a pure speculation.

Have you noticed that when you install a new version of Microsoft Windows, everything is just rosy for a while? "Best Windows yet" is the usual comment, and except for the execrable Windows ME each new edition generally is the best Windows yet. It's clean, efficient, it just works, and we're all happy.

Then over the years problems develop. Programs that used to Just Work are a lot of trouble. Nero comes to mind; there was a time when it was child's play to use Nero to burn an ISO image onto a CD, or make a new data DVD; now it takes time and you have to figure out what it's doing, and if you make a mistake you may have to restart the machine. Feh. Perhaps that's not Microsoft's OS at fault. It may just be feature creep. But it sure is annoying, and I am sure you can think of other examples.

It begins to take forever for My Computer to open. Making network connections takes longer. Things still work, but it's just harder to get your work done.

About then a new edition of Windows comes to the rescue. Buy that and your troubles are over.

Now I would never accuse Microsoft of sending out glitchmaking routines with their automatic updates. I am sure that most of the hassles we have with new security settings are required, even if they do make things harder to use. And of course all kinds of new stuff comes out during a Windows product cycle. New drivers are needed, and they have to be compatible with legacy equipment, but the best programmers have been taken off maintaining the old OS and put on new product development; and therefore keeping up with Driver Hell gets very difficult.

We need not ascribe this deterioration of performance over time to malice. Indeed, it may not actually happen: few of us remember what it was like when we were running an older version of Windows.

But it sure seems that way. And now that Vista is coming out, I predict that Windows XP will get harder and harder to use...