Diary sections: current 2004 2003 2002 2001 older
Ahh, Christmas! It was a very white Christmas. This picture was taken in the late morning of 12/25/02.
I went out into this mess for a few 90-120 minute sessions, and helped clear a path to each house on our street, just in case of emergency. My wife used the zoom lens on the camera to find me in the snow. Of course, my hood was down because I was always looking around to make sure I wasn't dumping snow in the face of a Boy Scout as he was shoveling out an elderly lady's sidewalk.
And, by the next morning... Wow. That dark spot in the lower left corner is the front of a car that was abandoned by its owner after becoming stuck in the snow. (Hey, perfectionists: The colors are mismatched in this picture because I manually corrected them from their very-blue state. It was very early in the morning when this photo was taken, and the camera didn't handle the low-light condition. Sorry.)
Now here's why the DiamondMax Plus 8 40GB drive is a winner...
66 MB/second write, 104 MB/second read, on a cheap US$23 IDE controller! The above benchmark was on the following: Windows 2000 SP2, two Maxtor DiamondMax Plus 8 Series 40GB drives, and a Silicon Image SIL0680 RAID IDE controller.
Grr. Microsoft strikes again. I failed to receive a large email message, which was my fault because I'd previously set IIS 5.0's SMTP Virtual Server to a 2048 KB limit per message. When I discovered that this large email message had been lost, I deactivated the per-message and per-session size limits in IIS. From then on, all email was refused with error 452 -- "Not enough storage". I had 300MB of free RAM and 34GB of free disk space, so I concluded that this error resulted from my change to IIS. I re-activated the limits in IIS and set them to higher values, and the problem disappeared.
If you're wondering why I would be so crazy as to allow unlimited email in IIS, there's a simple answer. My email antivirus software also limits the size of the messages. By the time IIS receives an email message on my server, the message has already been inspected by the antivirus software.
Beta 15 of the 0.20 PCI patch is posted, although slightly hidden. Most people are able to discover the link, although at least one person posted a note at the Sudhian.com forums saying, "I give up -- where did you hide it?"
I responded to a problem with Windows Small Business Server's "Add New User Wizard" last night. My customer couldn't add new users; whenever he tried, a crash in ADDUSRW would occur. I checked with Microsoft, and they had two suggestions. Either the MFC42.DLL had been replaced, or else he failed to type in a description for the user. Huh?
The secret to VIA PCI bus parking is 75=08, 76=50. I obtained that answer by comparing WPCREDIT dumps of my P4PB motherboard ,well before I was under VIA's NDA. More recently, I'd been hearing that motherboard BIOSes are showing up with 75=09 76=50 instead. Well, a user in the StorageReview.com forums found that his system was miserable and unpatchable with those settings, but changing 75=08 solved his problems. Wow, I was going to test that this weekend, but now I might not need to.
The second browser was Opera until recently. I saw that Mozilla 1.2.1 had been released, and had a fix for my favorite Mozilla pet peeve. (I couldn't post messages into web forums easily, because word-wrapping was broken in edit boxes under Mozilla 1.0.) So, I'm trying Mozilla right now. I can't customize the cache as much as Opera permits, and I am unhappy with the one-to-four seconds needed to minimize or restore a Mozilla window. But, it's really usable otherwise, and it's free. Only time will tell whether I switch back to Opera.
I did a little work on the VIA PCI patch this week. Build 15 contains support for enabling PCI Bus Parking on the KT400 and P4X400 chipset, and allows enabling PCI Master Read Caching (including enabling the T1 optimization on Athlon chipsets). Build 15 isn't ready to post, because the Bus Parking support isn't tested yet. Unfortunately, it's been almost a month since I first discovered how to enable it, and people could really use that feature.
It's 1:10 AM. I have solved the "white bar" video capture issue with the eVGA Geforce4MX video card. Instead of using NVidia's WDM capture driver, use the adapted Microsoft driver here from Gainward. Thanks to the tek-tips forum for the answer.
Outlook 2002 (Outlook XP from Office XP) has been frustrating me. I use the Rules Wizard to sort my incoming email at home and at work. But, the sorting would not happen automatically. I found out that the problem is fixed in Office XP Service Pack 1.
Office XP Service Pack 2 is available, and Office 2000 Service Pack 3 is brand new.
What a week.
I'm trying to separate the topics. How does this look? See below.
|I met with my largest network customer, and asked whether my work for them would exceed their budget. Looks good, they said. Whew. That customer represents about 20-25% of my gross income.|
|I practiced upgrading servers from NT SBS 4.5 to SBS 2000. User accounts and proxy server settings seem good, but MS asked me for a domain name and then didn't use it when creating email aliases. Fixing the email issues involve several ugly steps too. But, overall, the process is not too difficult.|
|I ordered a lot of toys from newegg.com. Here's a quick summary.
|I heard about a
posting from VIA's Fiona Gatt. She
was looking for the "WPCREDIT guy", meaning
H.Oda, but he wouldn't reply to her.
Fiona: "A lot of you guys and gals use WPCREDIT for tweaking and system info etc. A lot of our latest chipsets aren't supported by the utility as yet. I want to provide what ever technical documentation the developer might need to include support but I have emailed the only email address on the http://www.h-oda.com/ web site, which is a web masters address, and havn't had a response. I guess the poor developer gets a whole heap of emails so it's not suprising. I was wondering if any of you happen to know the guy or know of someone who does etc. Any connections you have would be appreciated."
People were suggesting me as a suitable substitute. A friend emailed me to let me know this was going on. Three days after her posting, there were four replies recommending me to her. I posted a note to say that I was willing to help her. Five days went by and I didn't hear anything, so I posted a polite good-bye. Last Friday, Fiona reached me and offered me a chance to obtain VIA's datasheets in return for creating PCR files and articles about WPCREDIT. I've accepted.
So, eventually, I should have datasheets for all VIA chipsets, and we will have PCR files for them, and my VIA patches and skills can only get better and better...
I believe I owe Maxtor an apology. I received a sample of their new DiamondMax Plus 8 series hard disk drive. I had read the specifications and said, "This is going to be a big loser. How good could a single-platter disk drive be?" Well, it delivered a whopping 38000 rating in SiSoft Sandra benchmark. One reason for the high rating was explained by my HDSPEED utility, which is reporting that the drive delivers data sequentially at 59 MB/second. Now that I've seen these incredible performance numbers, I want to try the Plus 9 series!
In PCI patch news, I successfully created a WDM (.SYS file) version of my patch that runs on Win9X. This version has minimal modifications from the previous NT(2K+XP)-only version, so I have high hopes that it will still be fully functional on all the NT-based OSes. I haven't tested it on those OSes yet, so we shall see...
Creating a single patch driver for all OSes will halve the work I need to do when interfacing to the driver from my Control Panel applet.
Halloween is over. My whole family spent the week of October 28-November 3 being ill, including me. My daughter's surgery (adenoidectomy) on November 4 went well too. So, I'm back!
I was just informed that my patch has become part of a manufacturer's official solution for VIA chipset problems. This must mean I "made the big time"... :) Well, I knew that other manufacturers were emailing the patch to irate customers as a support solution, but this puts it in writing!
It's Sunday, and indeed a day for rest.
>> I found a stable, yet high-performance, PCI configuration for my A7V133 motherboard. This configuration will be in the next build of the PCI Latency patch. I reached 106 MB/second with a Promise Ultra133Tx2 controller and a Maxtor D740X hard disk. That sort of performance number was basically unheard-of up to now.
>> I completed and debugged the registry code for the 0.20 PCI patch. It might have a typo somewhere, but it works.
>> I found out the final reason why my Microtek 3800 flatbed scanner wasn't focusing well.
I'd already found the alignment of its CCD "eyeball" was wrong. Once I'd fixed that, the focus was still bad, so I tried realigning it again, but with no improvement. Well, today I opened the small lightproof compartment that holds the lens. The frame, which normally holds the lens rigidly in place, was cracked. Grrr! I hand-focused the lens and sealed the lens and frame in place with Super Glue. (Yes, genuine cyanoacrylate Super Glue from Super Glue Corporation. Accept no substitutes. This stuff is better than the four-for-$1.50 junk at the local store.)
Here's the proof. I scanned this image with the 3800 earlier today, after cleaning the flatbed glass.
|Click on the image to see the full-size 300 DPI scan. (800Kbytes)|
I am in the middle of debugging the registry support for the new 0.20 VIA PCI Latency patch. The registry support is the first step toward a totally-configurable patch. The build I'm testing will let users override these items in the registry:
Plus, there's a registry setting to control the overall patch behavior. It currently has four modes -- no patch, George's only, George's patch plus registry overrides, and registry-specified patches only.
How can I be expected to work in the IT industry when there are job descriptions like this -- J.Lo's personal tweaker?
(This was a reply I composed to an email today. I thought I'd share it.)
I know a few things about the Asus A7V333. I’ve spent a little time with it this weekend, as part of researching why my latest beta patch worked so badly on the A7V133.
The BIOSes on Asus’s various Athlon motherboards are configured to prevent the CPU from being powered down when idle. This will keep the CPU hot, but it has the side-effect of preventing most of the PCI problems that occur on other manufacturer’s VIA-based motherboards. When users complain about PCI problems on Asus boards, it is usually because the user has installed CPU-cooling software. The chipset’s PCI configuration settings become critical once CPU cooling has been enabled, and Asus’s settings are not good enough. So, under normal circumstances, I have to strongly suggest that CPU cooling software not be used with Asus motherboards.
There is one significant problem with the A7V333 right out of the box. Two jumpers control the voltage supplied to the RAM, and these jumpers are set to deliver too much voltage. Worse, these two jumpers are not discussed in the motherboard’s manual. The jumpers are JP1 and JP2, and they are located between the AGP slot and the FireWire connector at the “rear” of the board. Move both jumper caps to the “1-2” position, i.e. facing away from the FireWire connector. You can see the jumpers at this page:
That author suggests removing the jumpers. In that picture, you could instead move both jumpers to the right.
I reset those jumpers before I worked with the A7V333. I have not noticed a stability problem so far. If you reset those jumpers and you still have problems, please consider replacing the RAM. The RAM might be of poor quality, or it might be damaged from exposure to too much voltage. I suggest using RAM that’s made of Samsung chips for greatest stability.
NOTE: My PCI Latency Patch, version 0.20 beta 10, ran perfectly with the A7V333 today. But, there’s more. In order to test Asus boards with cooling, I have added the necessary settings to the patch so that the CPU is cooled automatically by Windows 2000’s and XP’s built-in ACPI support. This enabled me to tune the patch so it will correct the settings and thus allow PCI stability when commercial cooling software, e.g. CPUCool, is installed.
>> Since the DMCA has taken effect, hundreds of technologies we took for granted are now illegal because they might permit copyrighted material to be copied. Xerox copiers, CD burners, and tape recorders are all verboten, in theory. As of today, there's a new entry on the list of forbidden items: the PrintScreen key. That's right, criminals everywhere are using the PrtSc key to capture snapshots of screens containing material they do not own.
I expect the PrintScreen link to die rather quickly, because it's linking to the news section of someone's home page, so I'll summarize the link. The UK's W. H. Smith store chain management challenged this non-hacker fellow to break the copy-protection surrounding e-books and Microsoft Reader. The gent simply pressed PrtSc repeatedly to capture each page of a sample e-book, and shared the captured images.
>> In other copying news, Microsoft's licensing diskette for Small Business Server 2000 is easily copied. Since I have a customer who paid for ten license kits, yet CDW shipped only one diskette, and CDW's representative insists the customer does not need the other nine diskettes, I copied the diskette and tested CDW's claim. CDW is wrong! The diskette allowed a test installation of Small Business Server to increase its client access count from five to ten users. Then, the diskette's installer demanded a separate disk in order to increase from ten to fifteen users. My diskette had been permanently stamped with the number ten as the diskette's maximum client count.
I won't be divulging how I copied the diskette, and I have reformatted the diskette that I'd used for the test.
It's been a tiring week. I survived a round of layoffs at my day job. So, to keep you occupied, I'll post this link to a quiz about movies with character's faces removed.
And they say justice is blind...
My customer finally received his copy of Windows 2000 Small Business Server from CDW. The fourth CD is supposed to be Outlook 2000 software, but this package contains a second copy of the first CD instead! Well, Microsoft is to blame this time, but CDW was still involved...
Wow, some information we can really use... The distinctive sound of an IBM hard disk failing.
Someone sent an ad to every computer on the Time Warner network… It wasn’t an email message, or a web-server invasion. It was literally a message that popped up on the screen!
But you wouldn’t see it unless you were running Windows 95/98/ME’s WINPOPUP, or Windows 2000’s MESSENGER service. You’d only see it in Windows XP if you forgot to enable XP’s firewall.
My own PC runs Windows 2000, and I use Windows 2000’s built-in TCP/IP filtering to block unwanted incoming TCP traffic. But apparently WINPOPUP uses UDP/IP, which Windows 2000 doesn’t block properly!
I’ve installed Tiny Personal Firewall 2.0 in response.
I previously mentioned a customer who received 10GB tapes when he asked for 15GB DLT III tapes. Well, CDW-G listened to his complaint, and they responded... by sending him more 10GB tapes!
Google reveals secrets of PC technology used in their search engine. http://www.google.com/technology/pigeonrank.html
Here's one for the history buffs. MAD's answer to the World Trade Center?
I told ViaHardware.com that I won't be managing their Tweaks forum from now on. They've asked me to stay on as a contributing writer, and I've tentatively agreed.
One of the batches of Walmart "16X" CD-R's is giving errors in the LiteOn 40125S burner at 40X, although the batch seems to burn at 24-32X. Their Durabrand 24X CD-R's are doing well at 40X.
Windows XP Service Pack 1 is confirmed as shipping in a few days from today. The unofficial news is that the current beta build 1105 will be modified to contain the new anti-piracy licensing code, and will ship as 1105.1. Out of curiosity, I'm trying to obtain a copy of build 1105 to compare to the final product.
Since my son has been keeping us up at night, I've been playing American McGee's Alice to pass the time. I just finished the game in "god mode" last night. If he continues to keep us awake, I'll get through the game the normal way.
Wow. A 17-day gap in the diary. Yep, I've been busy.
CDW-G's sales rep apologized twice for mis-ordering thousands of dollars in software for my customer. He promised to make things right. But would he put the promise in writing? It's been eight weeks and even the CD media for the software hasn't arrived. Another customer showed me the crate full of new DLT tapes he got from CDW-G -- they were only 10GB tapes! Those will be going back -- the customer's smallest server has a 15/30 DLT drive and 13GB of data.
I installed a server upgrade for a customer last week. In one evening I installed a new tape drive, new hard disks, and new RAM. The whole process only took 3.5 hours. That may sound like a lot, but that time includes: backing up their old configuration, copying all their data to the new hard disks, running a full diagnostic on the new RAM, and searching for a driver for the new tape drive.
Here's a bonus from Wal-Mart. Their Durabrand CD-R discs are on sale right now. It appears as if they are clearing old stock out of their warehouse. But, among the many dusty 100-packs of 16X and 24X OEM discs carrying the Durabrand name, some are made by CDC Magnetics. These discs write perfectly at 40X in my LiteOn LTR-40125S burner, and they are almost flawless when burned in my Plextor PW-4012A drive at 40X! The discs can be identified by the computer-printed batch code in the center ring of the disc, which can be seen through the top of the package. The good ones have at least twelve letters or numbers in their code, begin with A, and end with LH. Be careful, because some others have A and LH in the middle of their code, and I haven't tested those.
On the down side of the Wal-Mart CD-R deal... Other Durabrand OEMs include GigaMagnetics, whose discs may or may not have imprinted codes. The GigaMagnetics discs are so bad... They show errors when burned at any speed higher than 16X in the LiteOn drive. Okay, so they're only labeled 16X, but where's the safety margin? The Plextor drive senses this specific media's ID, and slows down the drive to 12X, and burns the discs slowly-but-safely at that speed.
We are supposedly only a few days away from having the final Windows XP Service Pack 1 released. Apparently, build 1090 of SP1 would permanently label your XP's desktop as an "Evaluation" system, and build 1097 broke the support for USB 2.0 that was working in prior builds. We'll see just how good the final version is.
I am doing an emergency T1 move for CCS. I / They have been given just one week to move all their servers, the router, and all their DNS information to a UUNet T1. I'm almost done; once Verisign completes the changes to the old domain's DNS records, I can complete the move.
The truth about Fight Club revealed: It is really just about Calvin & Hobbes growing up.
I haven't posted in a few days. I've been too busy counting my money... Too bad that all the money was made by selling my old ham radio equipment at a loss.
I found a quote that belongs in the fortune file. Here 'tis.
Why he gave up a career in biochemistry: "I realized I could spend my life studying the pubic hair on a rat, and although I would be the world's foremost expert on the subject, I might still not be able to explain why it is curly." -- Eugene Jarvis, video game designer
Windows 2000 SP3 is ready for download. We'll see how long this preliminary link to it works... UPDATE: SP3 is for real! Windows Update now offers it.
My cable modem slowed down drastically in the last few weeks. It turns out that Time Warner is screwing new customers by throttling bandwidth severely. Their new "professional" service level is so bad (128K/1M), I'm left thinking that Time Warner deliberately chose these levels so that no one would want to buy them! Anyway, every time a technician reset my cable modem, they set it to one of the new levels. My old agreement calls for 768K upload and unlimited download, so I called and had them restore my original service level. I'm still seeing only about half of my original bandwidth, but that may be because it's daytime and their network is busy. I'll find out my real bandwidth by scheduling some overnight uploads and downloads.
The Utica hamfest is tomorrow, but it's already raining today. How am I going to set up a table to sell all my junk? I think it's time to begin selling on eBay...
I've been playing with CPU heatsinks and Linux lately. (Not at the same time, of course.)
In the world of heatsinks, I've been favorably impressed with the GlacialTech Igloo 2310 model. As usual, the secret to the heatsink's performance seems to be the fan; it's quiet yet has good airflow. I've also tried the Enlight Dr. Cool model, but I'm not as impressed. It has a much-louder fan with more airflow, yet it only reduced the heat of a busy Athlon by one degree Celsius. I have more sample CPU heatsinks and case fans arriving next Monday, so the playing will continue.
In the world of Linux, I've been staring at Mandrake and Red Hat versions of Linux. Mandrake 9.0 beta 1 is out, and I tried it. I found a few bugs right away. The installer's X-server setup screen doesn't show the available color depths for my ATI Rage Mobility AGP video card, for example. But, I went back to the released versions 8.1 and 8.2 of Mandrake, and they had their share of problems too. In Mandrake 8.1, with a USB mouse, I can't click on graphic buttons in a Java applet in either Netscape 4.7 or Mozilla 1.0 browsers. I can use Mandrake Control Center to map "mount points" for Samba shares successfully in 8.1, but I had varying trouble in 8.2 and 9.0.
The downloadable version of Mandrake 8.2 Linux ProSuite edition is basically empty! The purchasable version comes with a licensed copy of StarOffice 6.0, but the downloadable version omits it. My downloaded copy has no games and it installs no office tools, not even KOffice.
Red Hat Linux 7.3 is quite bloated. I watched it installing a ton of RPM packages, including four printing systems and Japanese font support.
A tip for Linux... Here's how to map a Linux client to a guest-accessible
share on a Windows 2000 PC. A correct line in /etc/fstab looks like this
//myserver/myshare //mnt/myshare smbfs guest,username="guest" 0
0Of course, you need to create the folder //mnt/share ahead of time.
I really must add this site to my portal page sometime. Adoption agency selling shaved apes as babies!
Which one of these "enterprise-class" virus-updating mechanisms is more goofy? You be the judge.
Company N: Their preferred method of updates was to make you subscribe to a "channel" using a third-party application, even though automatic FTP updates are possible through their software. This third-party app offers the choice of extracting all update packages to a folder whenever they arrive, but since the same "channel" sends both virus data and software upgrades, the app might fill your 'virus data updates' folder with patches and even whole antivirus applications. (They have finally ceased using this "channel" for virus updates.)
They have built an automatic FTP-downloading update system into their main antivirus software. It allows a server or client PC to download updates and save them to a folder, which can then be NT-shared or become part of one of your FTP servers. But the system is currently broken; the software for server PCs is one version behind the software for client PCs, so the server software does not know to download extra files needed by the clients. And, if you choose to NT-share the virus updates, your workstations cannot receive updates till they log on. Oh, and you can only get updates for virus signatures this way; you can't update your scanning engine or other software without creating a script to roll out the updates.
Company S: Their README file touts a new feature. Instead of manually downloading batches of updates using a separate application program, you can preconfigure the same program to remember the last batch you downloaded. Then, you can use a new command-line parameter for that program, so that it silently downloads the updates when it is run. To automate downloading of updates, you add this special command line as a task in the Windows Task Scheduler on an NT or Win2K server, but you must get all the path information correct of course.
Their update software uses HTTP or FTP internally, but the locations of updated files are "hidden" to prevent pirating. Of course, their update process starts with downloading a ZIP file containing a clear text list of the updates and their locations on the Internet... And this ZIP file's name never changes...
I received the ZS0K firmware update for the LiteOn 40X burner (LTR-40125S). ZS0G introduced P-CLV burning method, and ZS0K seems to be an update to provide improved P-CLV compatibility. I tested it and compared it to a Plextor PX-W4012TA 40x burner. The LiteOn drive is more sensitive to disc flaws than the Plextor, so I used the LiteOn drive and Nero CD Speed to verify the discs.
|Media type||LiteOn ZS0K||Plextor|
|TDK 40x CDR||40x: Flawless||40x: Flawless|
|WalMart Durabrand 16x CDR||40x: Flawless||40x: 99% Flawless|
|MediaPro "32x" CDR||40x: Many ECC errors
32x: Fewer ECC errors
24x: 96% okay
|40x: All of 2nd zone is ECC errors
32x: All of 2nd zone is ECC errors
Google gets a makeover...
I had a semi-relaxing day at home. I worked on the HDSPEED tool and the PCI latency patch. As a result, I believe I've set a new record for VIA PCI performance as a result! Check this out.
I've deliberately shown the 0D register and the 70-76 registers of WPCREDIT so that people can try to match this.
The configuration was as follows. A Biostar M7VIF (VIA KT333), Samsung PC2700 RAM, Geforce2MX video. A Promise TX2-133 non-RAID IDE controller. Windows XP with ACPI disabled, VIA 4-in-1's (4.38 I think). My new 0.20b4 patch, which sets register 0D to value F0 in the Promise controller. A single Maxtor D740X 40GB disk drive. HDSPEED configured as shown.
Sorry, no updates lately. Until I have a chance to catch up, here's some advice:
When on vacation in Hawaii, never joke about Spam.
On Windows XP, After installing the Q317277 update from Microsoft's Windows Update site, SoftICE can no longer perform its process of matching symbol tables to loaded modules. The Q317277 update contains new versions of NTOSKRNL.EXE and NTKRNLPA.EXE. If these files are installed, SoftICE will report the following errors in its logging window at startup:
*** API Hook Failure - SegLoad
*** API Hook Failure - PageIn
*** API Hook Failure - CopyOnWrite
I've asked NuMega (now CompuWare) for an updated core file, NTICE.SYS, to address the problem. Wish me luck! CompuWare's site has been structured in such a way that all old links to helpful information now point to pages that offer to sell "enhanced" support. I guess that, if you already paid for support, that isn't good enough!
I went to the Cortland hamfest anyway. Cold and wet. No major commercial vendors showed up.
Happy Flag Day (US).
I received the Kenwood TS-570D rig that I'd won on eBay. Its face was dirty, it had no microphone, the owner removed its crystal filter before sending it (but included the filter's documentation)... But, I paid less than half of list price for a used current-generation radio, and the description of the rig on eBay was technically accurate, so I can't complain... can I?
It looks as though I must scrap my plans to set up a table at the Cortland hamfest tomorrow. My assistant has car trouble and the weather forecast calls for rain.
Lots of stuff going on... All of the following items involve new toys...
I finally received evaluation copies of Symantec's latest enterprise-level antivirus solutions. Will they be better than NAI/McAfee's solutions? I have three customers whose network security needs to improve, and now is a good time to find out. (Psst! Hey, Symantec now offers single-server pricing on Norton Antivirus! But the resellers (SoftMart, Dell.com) don't know how to sell it yet! If they figure out how to sell me a copy for my web server, I'll let you know.)
UPDATE: Norton Antivirus Corporate Edition 7.6 will be tough to sell to one of these customers. The same version was loaded on their new Dell server, but the server rebooted whenever Veritas Backup Exec 8.5 began a scheduled tape backup!
I have been upset for two years with my Seagate DDS3 DAT tape drive in my server. If I perform a backup to a freshly-unwrapped tape, then reuse the tape a month later, the tape drive chokes on the tape! The tape drive continuously recalibrates, and it won't eject the tape. So, I went onto eBay and bought a Compaq DLT 35/70 tape drive. THEN I went to Seagate's web site and found a firmware update that solves the problem with my DDS3 drive! Oh well, I needed the extra storage and speed anyway. The 35/70 backed up 20.5 gigabytes of pre-compressed data in 1.3 hours -- not bad!
I'm also replacing other hardware around the house. I bought two of Western Digital's special-edition hard disks. These drives feature 80 GB of storage, 8 MB of cache RAM onboard, and are virtually silent. I've already done some basic benchmarks. My HDSPEED tool reports that it can read sequentially from this model at a 49 MB-per-second rate, which beats the best Maxtor drives (41 MB/sec) handily. Sisoft Sandra's benchmark reports a 31000+ drive speed, compared to 27000 on the Maxtor D740X. But, when I used a Promise FastTrak100 RAID controller to test the drives in a RAID 0 configuration, they scored about 42000 in Sandra, or about 1000 points less than the same configuration with Maxtor DiamondMax 60+ drives.
I received a $12 generic optical mouse, which I'll be using on my server. Actually, the mouse is priced at $6 on newegg.com's web site, but the site charges $6 in shipping for each one! That's right, I tried to add five of these mice to a large order, and the site tried to charge me $30 more in shipping. Perhaps this sort of scam is why pricewatch.com now compares the cost of items with shipping included.
Ham rigs have begun to arrive. I was using an ADI AR-146 for local 2-meter (144-148 MHz) operation, and an Alinco DX-70T for HF operation and shortwave listening (0.5-30 MHz). I've received a Yaesu FT-2600 for 2-meters, and a Kenwood TS-570D should arrive tomorrow for HF use. I've also bought new 10-meter and 2-meter antennas for the house, but I don't think I'll keep the 10-meter one because my handmade dipole is better. (The new 10-meter commercially-made dipole doesn't have a balun, which is probably a big factor in its poor performance.)
Aopen motherboard has vacuum-tube audio?
Here’s a funny one. I recently bought a Plextor (expensive) and Liteon (cheap) 40X burner. Because the Liteon was eating my Walmart 16X CDRs (fail even at 16X), I bought 40X TDK CDRs on sale at CompUSA. But I still had problems, so I grabbed a spare (heavier) power supply and I loaded Windows XP. Now the Liteon drive burns Walmart 16X CDRs at 40X without even any “soft” (correctable) errors! I did five in a row to be sure, and I used the “ScanDisc” (notice the ‘c’?) feature in Nero CD Speed diagnostics to prove they all were 100% perfect! Woohoo! Want to buy a 50-pack of TDK 40X CDRs? Going cheap…
I'm trying to hold back the laughter as I share the news that Joe Bob Briggs is a syndicated columnist now! This is a guy who presented himself as a "good ol' boy", which is a concept that welcomes illiteracy. Well, the funny thing is, I just read his column and it's not bad. You just have to add the Southern drawl as you read it... :)
Neato. Someone on the Internet is trying to hack into my web site using OWSSVR.DLL. I've renamed the file, and I can still edit my diary without it, so all that's left is to figure out who I report this to... This file is a component of SharePoint, so maybe this hack is already known to others.
A Tale of Two CDRW Drives...
The Plextor PW-4012TA is a well-engineered 40x12x40x drive, but it has one flaw. As of its 1.01 firmware version, it took 37 seconds to begin burning a disc. The disc is normally supposed to burn in three minutes or less, so this delay adds 20% or more to the required time. Thankfully, the new 1.02 firmware reduces the delay somewhat. Now, if Plextor would remove their three-minute limit on overburning...
The Liteon LTR-40125S is an inexpensive 40x12x48x CDRW drive that received great reviews. But, when I test it with generic Wal-Mart 16x CDRs (burning at 16x) and its ZS0A firmware, I get error-filled CDs every time. This problem seems to be greatly reduced by upgrading to their latest beta firmware, ZS0G. This new firmware is beta because it changes the way the drive burns. Instead of dividing the disc into zones and recalibrating between each zone ("Z-CLV"), this new firmware causes it to calibrate once at the beginning of the disc and use feedback from the lens to make small adjustments later ("P-CLV"). Since upgrading to the ZS0G firmware, it burns faster than ever on the 16x discs, and my last test disc only had one soft error according to Nero's CD Speed tool ("ScanDisc" with surface scan).
Yes, when people can download a CD's worth of data in thirteen seconds, a three-minute CD burner is just too slow.
To paraphrase Jim Allchin: Microsoft's bugs are so severe that, if hackers knew of them, it would destroy civilization as we know it. [Stifles a giggle] So that, your honor, is why we can never reveal all our source code. [Looks like he is choking on something -- or is he stifling a guffaw?] Yeah, I know, I'm being kind of hard on the guy...
I found a deal on blank CDs at newegg.com and I ordered a batch to see how good they are. ($23.00 for 100 32x blank CDRs, with free shipping.) They arrive today. I'll try them ASAP.
In honor of the release of Star Wars II in theaters, I offer this link to New Star Wars Characters.
It's scary to walk up to a server that's in use by hundreds of users, sign on and hit the Enter key to perform some routine action, and see the server freeze. This happened to my best friend last month, and it happened to me early this evening. I spent the rest of the evening recovering the server. Grr.
I heard from one of the original testers of my PCI Latency Patch. His name is Franck and he lives in Belgium. He wanted me to know that he's included me in the credits for his new album. I'm flattered. I already included him in the credits for my first versions of my PCI patches. He contributed a great deal of input to those early versions.
I've tentatively agreed to write a column for a new web site. I haven't submitted my first column yet, but the deadline is approaching.
Taiwan's Taipei Times has an opinion piece about using Microsoft proprietary formats to store government data. Coincidentally, this article is similarly aligned to the legislation in Peru.
Peru's government has a bill in their Congress that would require open-source software to be used for projects involving government data. Microsoft, blasted them for it, of course. The bill's sponsor has replied to Microsoft quite lucidly. Here's the site containing both Microsoft's and the sponsor's statements. I like his initial summary:
|"The basic principles which inspire the Bill are linked to the basic guarantees of a state of law, such as:
To guarantee the free access of citizens to public information, it is indispensable that the encoding of data is not tied to a single provider. The use of standard and open formats gives a guarantee of this free access, if necessary through the creation of compatible free software.
To guarantee national security or the security of the State, it is indispensable to be able to rely on systems without elements which allow control from a distance or the undesired transmission of information to third parties. Systems with source code freely accessible to the public are required to allow their inspection by the State itself, by the citizens, and by a large number of independent experts throughout the world. Our proposal brings further security, since the knowledge of the source code will eliminate the growing number of programs with *spy code*."
-- DR. EDGAR DAVID VILLANUEVA NUÑEZ
For those who are waiting for me to reveal something technically exciting, here's something you can read while you wait. It's science fiction from 1909.
I've been busy. Macintosh OS 9 has gross bugs when it comes to simple file sharing over TCP/IP. In order to stabilize it to the point where it will reconnect to servers after restarting the Mac, I have to do the following:
Even with those steps, MacOS 9.0.4 will freeze when copying a large file from one folder on a server to another folder on the same server. I'm hoping that Apple fixed this in an OS update, but their friendly web-update Control Panel applet freezes too! I have to manually download and apply patches in this order to bring a 9.0.4 OS up to date:
If that isn't exciting enough, Internet Explorer 5.1 for Macintosh can crash a proxy server! Its badly-formatted authentication packets are seen by the proxy server as an anonymous user with a real user's password. If the proxy server tries (and fails) to authenticate the anonymous user, the NT operating system will disable the anonymous user's account. The NT account, created by Internet Information Server for anonymous web browsing, is usually "IUSR_servername".
Here's a neat concept. In order to beat the criminal software publishers who try to make you accept an "End User License Agreement" after you break the shrink-wrap on the box, here's an app that will erase their EULA file before you can see it!
Wow. In 1983, while I was honing my assembly-language programming skills at roughly age 17, I was missing out on Castle Smurfenstein for the Apple ][!
I was relatively quiet when the sport of Extreme Ironing (German link too) premiered. But I can't say silent any longer, because we now have Extreme Croquet! I especially like the 2001 Monster Classic.
Apparently 3Com has a new line of 802.11b products with twice the bandwidth (22Mbits/sec) and longer range, but which are compatible with existing equipment.
I wrote this up for a person asking for help with an article. I didn't want it to go to waste, so I'll put it here for now. The following is a talk about the multiple banks inside SDRAM.
Banks can exist at multiple levels. It is acceptable to think of each SDRAM DIMM, or each side of a double-sided DIMM, or each 16-megabit section of a 64-megabit SDRAM chip as a bank.
The memory controller can control up to four DIMMs. The memory on a 168-pin DIMM has at the same number of bits of data ("width") as the current Athlon or Pentium II/III/IV CPU's bus, so one DIMM is sufficient to run a computer. It wasn't always so.
When the 80386 was at its peak, it needed 32 bits of RAM width, but the then-popular 30-pin "SIMM module" only supplied 8 bits. It was not unusual to see motherboards with 16 SIMM sockets. Since the CPU used four SIMMs at a time, each set of four SIMMs was referred to as a "bank". The 16-SIMM motherboard supported four "banks" of four SIMMs.
When the 80486 was released, the industry moved gradually from 30-pin modules to 72-pin modules. These new modules had enough data bits "width" to satisfy an 80486, but only half of the bits needed for a Pentium. But, the pinout of these modules allowed for a second set of memory to be mounted on the rear of the module's circuit board. The two sets of memory chips shared a common data bus, so they could not be used to provide all the data width needed by a Pentium, but they were able to be selected and deselected separately. Unfortunately, not all motherboard chipsets could handle these "double-sided" modules. These modules weren't commonly treated as having two "banks", because the chipset hid the details of handling the two sets of memory.
With the current (DDR) SDRAM arrangement, the data width of the DIMM allows one DIMM to be sufficient for current CPUs. So, the typical workstation design is:
A visual sample of an SDRAM chip and its banks is in this datasheet:
For people who are intimately familiar with EDO RAM, Micron offers a tech note on migrating to SDRAM:
I'd bought a Microtek Scanmaker 3800 scanner from Staples in February. It's a nice scanner, with plenty of big buttons on its front panel, and with support for 35mm slides. But it was their display model at the Staples store, and was slightly beaten, and Staples had lost both the original packaging and the power supply unit. The scanner lasted less than a day.
I received the scanner from the Microtek warranty repair center last night. Microtek was generous.
The original damaged top cover seems to have been replaced with a pristine one.
They sent a replacement power supply, and included a handwritten note warning me not to use my old one. I was worried that they would delay the RMA until I sent my original Microtek power supply... which, of course, I don't have.
They packed the unit in a genuine Microtek box. Normally, Microtek will charge $10 to replace an inferior box with a real Microtek one. But, I'd sent it in a nice box using plenty of Dell's new "airbags full of peanuts" packing material. I suspect that Microtek has an assembly-line method of repairing the units, and that the scanner was simply packed into the next available box on the line.
In other news, I installed the Biostar M7VIF into my main computer at home. It's quite fast, especially since I also upgraded to an Athlon XP1800+ processor at the same time.
I posted Alex's Easter picture (with Easter bunny) to www.georgebreese.com.
I finally made my peace with Newegg.com. I've reordered many of the new toys that I'd ordered, and canceled, previously. So, these toys aren't as hot and new as they were last month when I first wanted them, but they won't go to waste. :)
I included some Samsung DDR333 RAM in the order this time. My existing RAM isn't working on the KT333 when Trcd or Active-to-CMD timing is sped up from 3T to 2T. Perhaps the Samsung RAM will work. Without these extra timings, the Biostar KT333's Sandra benchmark numbers are roughly identical to the MSI KT266.
I don't have high hopes for seeing any performance increase with the current-generation Athlon with DDR333 RAM. The throughput of the Athlon's front-side bus is the same as DDR266 PC2100 RAM, right? Both today's Athlon and PC2100 are running at 133MHz, with Double Data Rate technology, and eight bytes wide, yes? It's time for me to double-check my datasheets on this subject...
What? My daughter's birthday celebration took two whole weeks? Why am I not surprised? :)
Actually, I worked hard on responding to all the incoming "feedback" email that has piled up since before the baby was born. As of last night, my feedback mailbox is finally empty. (My typical "feedback" mail message has a half-dozen questions in it!)
I will get back into the habit of posting to my diary, although I may move it to georgebreese.com sometime soon.
Boy, I hate to even provide this link, for fear that the target web server will be overwhelmed... Here's a Commodore 64 web server!
I added more photos to www.georgebreese.com. My life revolves around my wife and baby right now, and I'm trying to cope with the baby's night-feeding schedule, so I won't have much high-tech fun stuff to mention here for a while.
Here are a few notes that I can recall.
I bought an Olympus C2100 digital camera. Its sharpness is wonderful and its 10x/27x zoom is the best I could find anywhere for less than the price of a new car. This picture is a fair sample of its quality, although I reduced this picture so it would fit on a 640x480 web browser screen.
I've ordered my first motherboard with the VIA KT333 chipset. The weird thing is, the only board I could buy was the M7VIF made by Biostar. Biostar's boards were garbage, back in the days of the 80386SX. But the quality of my M7VKB was tolerable, and this board looks cleanly designed, so it might be a winner. We'll see...
Pictures of Alexander can be seen at www.georgebreese.com.
It's a boy! Alexander Richard Breese was born on 2/20/02 at 6:38PM EST. He weighs 11 pounds, 11 ounces (5.3kg), is 22 inches (56cm) long/tall, and is doing well.
After an problem with the fetal heart monitor this morning, an urgent ultrasound has just estimated the baby's weight at 13 pounds (6kg)! We've been ordered to go to the hospital immediately. I'm typing this as I finish packing the car. Wish us luck!
Still no baby yet. Testing on Saturday showed "progress". Huh?
This will probably be all over the news by tonight, but as long as my neghbor mailed it to me, I'll share it here while it's fresh. The next-generation DVD format was just revealed, with 27-50GB storage and a 13-hour capacity at 27GB.
No baby yet. Will spend the day at the hospital on Saturday for testing.
News from the world of networking
Gigabit-over-copper networks are halfway cheaper, like this:
The specialty at a new UK restaurant was rabbit stew. Business was brisk when the food inspector arrived. After examining the stew he said, "I can see that this stew has some horse meat in it. We allow horse meat in rabbit stew because rabbits can be expensive. But it can only be no more than fifty - fifty." With that, the inspector left.
The inspector returned to inspect again after a few months. After testing the stew he was upset. He said to the man, "I told you that you could use horse meat, but you had to keep the meat at fifty - fifty. This stew still has way too much horse meat."
The man answered, "I did make it fifty-fifty. One rabbit, one horse."
Gigabit copper cards are US$44 at buy.com today. Cheap, yes? But, a four-port switch is $700.
No baby yet.
It turns out that PCI Master Read Caching could work, as long as one related register is also set. So I posted version 0.19d of the PCI Latency Patch last night. One user still claims his A7V133 is trashed by it. Almost all others are good, but there is one report of crackling sound in a specific configuration.
If you are sick of www.amihotornot.com, you can try www.amibiosornot.com and www.amiannoyingornot.com. I'd like to learn how to redirect amihotornot.com to one of the others, using SuperScout.
Got the sample RAM from OCZ. Sexy, red DIMM in a copper heat-spreader. Time to play... Unless the baby comes...
No baby yet.
I finally got an answer from an Asus A7V133 (VIA KT133A chipset) owner about my 0.19c patch problems. PCI Master Read Caching is screwing up his PC. Based on my earlier KT133 research, my only other option for fixing the SBLive is to disable "STPGNT# CPU Idle", which has the unfortunate side-effect of heating Athlon CPUs 5-10 degrees Celsius. But, if it's necessary, then I suppose I had better do it.
I waited a week for OCZ to send my RAM, then I called. "Lost in FedEx shipping".
MSI released a BIOS update for their K7T266Pro2RU board. This is the fourth update since the fall, and they still have more to fix, In My Humble Opinion. They fixed the problem where the PC will restart when asked to shut down (PCI PME Wakeup is now disabled by default). But, I discovered that their 3.3 BIOS won't set the PCI Latency Timer; in fact, when I set that value in the BIOS, nothing changes!
The Register has reported that Valentine's Day viruses are on their way. What a way to say "I love you"... I know, we already saw a virus with that name...
I can't believe this was front-page news. Scientists Find Jurassic Age Dinosaur Vomit...
I think I've "battened down the hatches", settling almost all customer issues so that I can be ready for the baby to arrive.
This news story caught my eye. A "movie-viewing" (movie-pirating?!) site has opened up, in a country whose copyright laws are much more friendly than ours.
I posted a 0.19c version of my PCI Latency patch, but A7V133 owners are getting fed up with the problems. I don't blame them. I've heard from owners that have gross problems on Windows XP even when no patch is installed. Some testing is going on and one person posted his findings to the viahardware.com forums, but I think a lot more needs to be done.
I was offered a free sample of OCZ memory back in October. They sent other hardware, but never sent any RAM. I finally decided to order some from a web site. No such luck. The site forwarded the order to OCZ, who called me at 10:30PM on Saturday night to ask why I thought I had to pay for RAM. Well, I'd ordered that RAM to be shipped FedEx, and I still haven't seen it. I'm going to have to ask a friend to order it for me.
My 1-user, US$20 copy of McAfee WebShield SMTP arrived, and I received my grant letter. This deal is for real! Now if I could just get a reasonably-priced version 1-user license of McAfee NetShield for my server...
After the ice storm and wind storm of the last 48 hours, this year's weather report from Punxsutawney Phil should be no surprise.
Wow! Get McAfee WebShield SMTP for $20! Once I picked my jaw up off the floor, I ordered a copy! I am being peppered with infected emails, and so are my clients. This is a great time to get a copy of WebShield SMTP for your SMTP servers.
I was honored to be the recipient of a new strain of Klez email virus, which nothing was able to detect. I dutifully forwarded a copy to McAfee. Eight emails later, I'm hoping a human being has finally understood my message. I've dealt repeatedly with their auto-reply system, and the one human being who replied said "Duh, it's Klez" without understanding that this sample of Klez even survived the auto-reply system's disinfector!
Fun reading, although sacrilegious to Republicans: www.whitehouse.org and www.psychedelicrepublicans.com.
Customer issues and housecleaning have taken precedence over everything else. So, I have almost no technology news worth mentioning.
My best friend called to say that he'd just seen a gigabit-over-copper network card for sale at CompUSA, with a price tag of US$79. Yikes, here we go again, my network is obsolete...
My Pentium 4 computer with VIA P4X266 chipset just spent the last week as an Internet mail-and-database-and-DNS-and-Web server for 300 users. It performed flawlessly, and users couldn't tell the difference between it and their original server with multiple Xeons and SCSI disks.
I recently made the mistake of assuming that Compaq's new 5300-series RAID controllers had good enough sector translation to support an 8GB C: drive on NT 4.0. Um, no, only 4GB is possible. (When a BIOS handles SCSI or LBA IDE disks, it must translate old-style requests to new. DOS, or NT's loader, will request data by asking for the sectors located at a specific cylinder and head of the disk. NT 4.0's loader can only create requests for up to 1024 cylinders of a disk. If the BIOS emulates a disk having 255 heads and 63 sectors per track, then NT's loader can reach 1024*255*63*512 bytes, or 8422686720 bytes, or 8032.5 megabytes of data. The Compaq 5300's BIOS handles 1024*255*32*512 instead.)
Our baby is due in thirteen days. But, since a baby can typically arrive plus-or-minus two weeks from the estimated due date, the baby could come now!
Don't forget to send a get-well card to the Dalai Lama.
A lot of slow news days lately. When the real news doesn't work for you, make something up!
new iMac a hit with spendthrift public educators
Xerox wins GUI case against Apple, Microsoft, IBM, Red Hat, Be, SuSE, Palm, Amiga
Adobe defends Pirateshop 5.5 software
This web server was down for most of Wednesday, because an updated version of disk-defragmenting software chewed up the hard disk and spit it out.
No patch updates for another day or two. Gotta clean house before Grandma comes to visit!
Sick today. Something I ate? Nervous over my soon-to-be-fatherhood? I dunno.
Web changes have begun: Try http://www.georgebreese.com/ . When you're done, say hi at "gb AT georgebreese.com".
A customer was hit with an attack of BadTrans virus emails tonight. Somebody had the goal of swamping this customer's server. The server is configured to quarantine the infected mail files, but if this keeps up then I'll change it to delete them instead.
I am making some Web changes here. Stay tuned...
The UK version of buy.com is going bye-bye. The US version will go on...
Hey, amazon.com just showed a profit for the first time. So maybe venture capitalists should invest in dot-coms... Oops, they already tried that, sorry. :)
My latest patches (0.19 through 0.19b) are not consistently stable on KT133 and KT133A boards. I feel silly buying KT133A's to test with, too.
More fun: IBM's researchers use Sun workstations?!
McAfee Antivirus has an easy chance to be the antivirus of choice in the coming year. Here's why.
Microsoft's subscription-based software is going over like a lead balloon in businesses? Why am I not surprised? Is it because the small-business subscription price levels for software with a three-year lifespan are 50% of the retail purchase price per year? (In case you haven't had your caffeine yet: What's 50% times 3?)
Is it stupid for technology companies like Microsoft to have publicly-traded stock? The market forecasters and investors are looking for Microsoft to have a steady stream of income, but the software industry just doesn't work that way. Microsoft's software sales chart, in theory, should look like a sawtooth wave. Peaks occur when the new software ships, and valleys occur when the market saturates. Microsoft is working too hard to overcome the sawtooth-shaped revenue picture, and it's painful for them and us both.
Philips is my hero today. They are making two moves that I like. They are pressuring makers of copy-protected music CDs to stop, and they plan to produce CD copiers that can copy the protected CDs. Step #1 is for Philips to exercise their ownership of CD-related patents and trademarks, and they've begun by claiming that CDs containing nonstandard sectors are not CD's.
I was sick for a couple of days with a stomach bug. I tried to work on Monday, but I only lasted a couple of hours.
I picked up a "new" science fiction novel last weekend, titled Retief!. I'd read Keith Laumer's "Retief" novels as I was growing up, so I thought Mr. Laumer had returned to writing. Um, nope, it turns out that he's been dead since 1993. Another author is grabbing old Retief stories and novels, adding his own "CDT History" preface to each, and bundling them. The funny thing is, nobody told the reviewers... One review listed on the back cover says something like, "it's a pleasure to see Retief in new adventures". Oh well, at least it's one way to get hold of copies of the old out-of-print stories...
I wrote and posted a hard disk speed test program on this site last Friday. It should provide a cheaper way to discover the bandwidth of an IDE or SCSI connection.
I must create a patch that maximizes the throughput of the PCI bus. Since the TecChannel article came out, it seems to be the only thing anyone can discuss.
At least I have an initial set of registers that I know I should patch. I came up with the list by using my knowledge of the VIA chipsets, and testing a Promise controller using my hard disk test program and WPCREDIT. But I only tested on a KT266A motherboard. I need to do the same for the KT133 and older boards, in case they need different settings.
I'm almost done writing my first hardware review. But after seeing a note from Tuan at ViaHardware saying that the Serious Sam game needs to be added to the list of mandatory benchmarks, I panicked. Well, I've received word from ViaHardware this morning that I can skip it. (Why panic? Well, SysMark2001 was a finicky benchmark that required a whole evening just to test one motherboard configuration.)
Out till 11:00PM last night with a customer. (Note: WORKING with a customer! What were YOU thinking?)
Warning: new Euro currency is poisonous. So don't eat more than 400 at a time.
VIA released a new patch yesterday. I dissected it while waiting for my haircut this morning. The new patch looks for a Promise RAID controller, then patches the RAID so it stays on the PCI bus longer, and lengthens the VIA PCI arbitration timer so it will leave the RAID controller alone.
say space is turquoise
Astronomers in Maryland have worked out the color of the universe.
overload crashes civil service computers
A civil servant brought his work's computer system to a halt by downloading too much porn.
given alcohol during filming of anti-drinking advert
Swedish teenagers were given alcohol to make them happier while filming a state-funded anti-drinking commercial.
I didn't get a chance to test the Corsair RAM's abilities. Cassie got sick instead.
It's the end of an era. The 9-track tape drive is finally going the way of the dodo (and the IBM PCJr, and the 5.25" diskette, and ...). I must admit that I never did manage to add a 9-track tape drive to my computer hardware collection, even back when I was single and had room for such things.
FedEx just dropped off my sample of Corsair PC2700 RAM, made using "hand-picked" chips. For $175, it had better be good...
I'm busy with customers and with baby-prep. The crib is assembled, as is the recliner for my grandmother-in-law. The baby's room now looks vaguely like it should.
Reviews of the new "Northwood" edition of the Intel Pentium 4 processor are out. This new chip was compared to the new Athlon XP 2000+, with embarrassing results for Intel. The XP2000+ is well ahead of the 2.0GHz Northwood.
I'm starting to believe the claims of a problem with the VIA PCI bus, as documented by TecChannel. I believe I know which timer is interrupting the bus, and I know how to disable the timer, but I believe that SoundBlaster audio will suffer if I do. If I'm right, then the problem is as follows. VIA's PCI controller contains an "arbitration" timer that will detect when a PCI device has used the bus too long. This timer should be triggered when another device needs to use the PCI bus. But VIA's timer is being triggered when no other device needs the bus. As a result, no device can continuously use the bus for longer than the length of the timer. I wish I could prove whether another device needed the bus and triggered this timer, but I don't have the tools to do so.
I'm not having fun with the new 333MHz DDR SDRAM. Here's a quick summary of the situation.
I tried to review the Iwill XP333R. What a farce. I used it with my "unlocked" older Athlon CPU to figure out how best to overclock it, which is what OCZtech.com wanted. I did some speed benchmarking, which viahardware wanted, and it was pretty good. Then I put an Athlon XP1800+ on it for viahardware's review, and it wouldn't boot. I downloaded a new BIOS from Iwill's site, reinstalled the old CPU and flashed the BIOS, then reinserted the new CPU and began testing. All benchmark scores went down significantly! The Geforce 3 video card now runs like a Geforce 2. Oh, and even though this motherboard supports DDR333 RAM, my sample of KingMax is only detected as DDR266 in the BIOS.