Dan's Data letters #8
Publication date: 5-Nov-2002.Last modified 03-Dec-2011.
Standby drain?
Having recently moved into a student house in my 2nd year of university, there is much concern over keeping our electricity bill down and thus saving money. Among the steps decided to reduce power consumption was the issue of leaving computer monitors on standby as against switching them off.
Is it better to switch a monitor off than leave it in standby mode? How much energy is a monitor getting through when in standby mode? Where is it putting all this energy if it's not displaying anything on the screen?
Peter
Answer:
In standby, monitors should only draw a few watts. The Energy Star
spec for monitors
requires that they step down to less than 15 watts, and then less than 8
watts, if they have the full two levels of standby. Modern screens can be
expected to drop to less than five watts in standby. If five watts makes
a significant difference to your power bill, I don't know how you managed
to afford computers in the first place.
Note that a monitor that's just displaying a black screen, but isn't in standby mode (standby mode should be indicated by the monitor's power light flashing or turning orange or something) will still have all of its electronics powered up, including the power-sucking tube assembly. It will still, therefore, be drawing pretty much full power.
When a monitor's in standby with its tube powered down, it's still it's running the circuitry that allows it to power up again when the computer tells it to. And, as the nearest physics student should be able to tell you, all of the energy that goes into the monitor ends up as heat.
(This, by the way, means that computers make perfectly good electric heaters. A single PC doesn't have the heat output of even a small fan heater, but two or three PCs kept on all the time will be enough to keep most rooms comfortably warm, cost no more to do it than an electric heater of the same power would, and be able to do much more interesting things with the electricity on the way.)
Jumper hunting
I just bought a second hand PC to upgrade my mother's old computer. Problem is, the BIOS is password protected and I can't find how to reset it. The motherboard is a Gigabyte GA-6BXC Rev 1.7. I found a manual that says it can be reset using JP11, but I cant find JP11 on my motherboard. I have tried removing the battery, but it still boots with the BIOS password protected.
Kiran or Trish
Answer:
There's no mention of CMOS clearing in the manual for the v1.0 revision
of this board, but the manual for the v1.9 revision turns out to have v1.7
info in it as well (both manuals are available in PDF format from
here), and mentions
JP11. But, as you say, doesn't tell you where it is. I, however, happen
to know where it is, thanks to a reader who volunteered the information
after I first put this page up; it's apparently between the second and third
ISA slots from the edge of the board, next to the case connector pins.
Remember to turn off the hard power switch on the back of the computer, or unplug the computer from the wall, if the PSU has no hard power switch, before you use the jumper. If you can't find it, you can achieve the same result by turning off/unplugging the PC and then removing the lithium coin cell backup battery from its holder next to the PCI slots. Leave it out for several seconds, put it back in again, and the CMOS should be cleared. If doing that didn't work before, it's probably because you didn't remove the battery for long enough, or because you left the computer plugged in; the motherboard uses the standby supply line from the PSU to maintain its CMOS contents.
In other news, the Table of Contents entry for section 1.2 of each of the PDF manuals mentioned above lists that section's title as "KEY FEATHERS".
Thank you, and goodnight.
SCSI salvage
I have a an old HP Kayak XA workstation. It has a SCSI card already in it, and a 4.3Gb IBM drive. I was planning to go down to an auction and pick up three more 4.3Gb SCSI drives, to make it a server for the home LAN. I was going to RAID 0 them, so I was wondering exactly how reliable these drives would be. I mean, they're far from new, and they have a limited lifetime. Since the primary purpose of the server is to back up my main system when I go LANing, the reliability is a big-ish issue, but I don't want the loss of space associated with RAID 1 or 0+1.
Also I was curious, can you install Win2000 Server on a software RAID array created in Win2000?
Byron
Answer:
Don't put important data on second hand drives. Ever. You don't know where
they've been. RAID 0 sets of second hand hard drives, where one failed drive
will take down the whole stripe-set, are obviously an even worse idea. I
don't know much about Kayak XAs, but it's my understanding that they come
with either a SCSI controller or an ATA one. So you can't just plug
in a new commodity IDE drive, which is what I'd recommend if it were possible.
You're not going to find a cheap new SCSI drive, of course, so the best option I suggest is that you find a single used SCSI drive that's big enough for the data you want to store, and use that. Leave the original drive as your boot device, for simplicity.
Regarding the installing-on-a-RAID-array issue - Win2000 can't have its system or boot partition on a software striped volume, but I think mirrored volumes are OK.
Cheapie case
I've recently purchased a "Ritmo" case with two fans, front USB ports and a 400w PSU. It was only $AU90... cheap, I know.
Will I be able to trust it to run an Athlon XP 2200+, an Asus A7V8X motherboard, 512Mb PC2700 RAM, a GeForce4 Ti4200, a 60Gb Western Digital HD and a 20Gb Seagate HD? Oh, and a Sound Blaster Audigy? Is there any cheap, practical way to test the PSU to see what it can produce, before frying all my shiny new components?
Brad
Answer:
If that's a real 400 watt PSU in that case, it'll be fine. Heck, if it's
a real 300 watt PSU, it still ought to be perfectly OK.
If it's a cheap piece of crap with a "400 watts!" sticker on it, though, all bets are off.
Many PSUs can't deliver what they say they can - see here for amusing examples. Now, even a stacked PC isn't likely to actually have a more-than-300-watt continuous power draw, but if any PSU's not going to live up to its label by a large margin, then an allegedly-400-watt one in a 90-Australian-dollar case is that PSU.
There's no easy way to test a PSU's capacity. You can make like Tom's Hardware and load up the rails with power resistors, but you probably don't have a bunch of low-value power resistors sitting around. Gimcrack power resistors can be made out of fencing wire hanging in buckets of water, to name just one of a variety of mad-scientist options; all you need to make it all work is a multimeter to help you guesstimate your resistor values. But that isn't a five minute job either.
In any case, a dodgy PSU is unlikely to barbecue your computer, even if it dies. It's much more likely to never actually die, but instead just give you a really annoyingly flaky computer, until you swap the PSU for something better.
Network mystery #7623
I have four PCs running Windows 2000 Server. Connect them to a cheap 5-port palm hub, and they work well. Swap the hub for an eight port mini switch (I've tried CNET and LanTech switches), and the four PCs are no longer networked together. Can't even ping each other.
Swap back the hub, and network connectivity is back.
If I use the switch with two PCs running WinXP, they work just fine.
Network settings/installations on the four PCs are default, and I only gave each a fixed IP and subnet mask.
There is nothing to configure on the switches themselves. I'm thinking that there must be some service or setting in Win2000 that's confusing the switch.
Norman
Answer:
I'm assuming you're using the same network cables for the computers that
work as for the computers that don't. If not, then a dud cable or two could
be the problem, especially if the hub is 10BaseT only and the switch is
10/100BaseT.
I'm also assuming you're not plugging any PCs into an uplink port on the switch by mistake. That ought not to shut everything down, I think, but it sure wouldn't help.
I'm further assuming that you haven't just been unlucky enough to get two dodgy switches. The fact that the XP boxes works suggests that that isn't the problem.
Things you can try:
- Update everything's network card drivers. The drivers can, conceivably, stop a network card from correctly negotiating with the switch. A 10BaseT hub doesn't require speed or duplex negotiation; a 10/100 switch does.
- Lock all of the network cards down to 10-megabit-per-second, half-duplex mode in their Properties. Again, if it's something to do with the higher speed or duplex negotiation that's causing the problem, this ought to solve it.
- Turn everything off, with the network all plugged together. Turn on the switch. Now turn on the PCs. If the network adapters aren't negotiating properly when "hot-plugged" (unlikely, but possible), this will solve the problem.
- Try different network cards. Swap the ones from the XP boxes into a couple of the 2000 boxes and see if it helps. If the 2000 boxes have built-in NICs, you can disable them in BIOS setup and try a PCI-card NIC.
Thing you can try that probably won't work:
- Shuffle cables around. You've probably already done this, but if anything can see anything when you plug everything into different ports on the switch from the ones they had before, then that indicates eldritch cable problems; replace (and/or shorten) cables until the problem goes away.
4+4+4=5?
I have a Microsoft developer CD which contains the following:
Windows 2000 Professional
Windows 2000 Server
Windows 2000 Advanced Server
Windows 2000 Security Rollup Package 1(All on one CD-ROM, yes CD-ROM, not DVD-ROM)
When I look at the properties of the CD from Windows Explorer, it says the disc's "Used space" is 510,263,296 bytes (486 MB).
However, when I copy this CD to a hard drive - simply select all files and drag to a folder on C: - the copied files take up 1,239,522,616 bytes (1.15Gb). The "Size on disk" is 1,289,445,376 bytes (1.20Gb)
These are full working versions of all 3 Windows 2000 Operating Systems straight from Microsoft. On the CD, each OS folder looks exactly like the single CD with that operating system, and the number of files are exactly the same (eg. Microsoft Windows 2000 Server CD - 7817 files, Win2000_EN\ENGLISH\WIN2000\SERVER folder on above CD - 7817 files).
My question is, how is it possible to fit 1.15GB of data on a CD that then only shows up as only 486Mb? I have tested duplicating this CD, both using a proper CD duplicator and also using software with a burner, and the duplicator can copy the CD no problems, but using software and a burner, there is just no way it can fit all the files back onto a CD-R.
I have noticed that there are a large number of files on the CD, most of which are small (less than 1Mb); for instance, in the folder Win2000_EN\ENGLISH\WIN2000\SERVER\I386 there are 4646 files and only 18 are greater than 1Mb. Also, I have tested copying the files to NTFS and FAT32 partitions, with roughly the same results (the above size information is from an NTFS partition).
I have been stumped with this one for a long time and have mystified a lot of people with the logistics of this problem. Any insight you could provide that would help this mystery would be greatly appreciated.
David
Answer:
At first, I thought this might be related to slack
space, but that doesn't
actually make sense.
Slack space is the space that's left over in the last disk cluster taken up by a file. Disk space is allocated by the cluster; the minimum amount of space a file can take up is one cluster. Your NTFS hard drive probably has four kilobyte clusters; make a one byte file, and it'll take up four kilobytes on the disk. Make a 4097 byte file, and it'll take up eight kilobytes on the disk. Make 1024 one byte files and they'll take up four megabytes.
The average amount of space wasted by files on a 4kb-cluster drive is, therefore, two kilobytes per file. A collection of files with a given total amount of actual data will waste more and more space as the file size falls, unless many of the files are whole multiples of the cluster size in size, or not far short of a whole multiple.
Windows' Properties display for a bunch of files shows you the "Size", and also the "Size on disk" for the files. "Size" is the actual amount of data; "Size on disk" is the total disk space taken up, including slack space.
This explains the 49,922,760 byte discrepancy between the "Size" and the "Size on disk" of the files once you've copied them to your hard drive. Three directories each with something like 8000 files in them gives a nice neat two-odd kilobytes wasted per file, which is what you ought to see from four kilobyte clusters. So there's nothing weird going on here.
The weird thing is, as you say, the discrepancy between the amount of CD taken up - the absolute bleeding-edge maximum possible capacity of weird super-over-burned exotic CDs that many drives can't read is 900Mb, but this CD's a perfectly normal one that isn't even close to the old "74 minute" 650Mb ceiling - and the amount of hard disk taken up by the data after you copy it. CD-ROMs only have about one kilobyte slack per file, but you'd need a rather large negative amount of slack per file to make up the difference in this case, and I doubt Microsoft have developed negative-slack technology yet.
What is actually happening is, basically, a neat little hack, which may be familiar to anybody who's used to dealing with Real Operating Systems in which hard and soft links are everyday tools.
The three operating systems on your CD share lots and lots of files. There is not a big difference between different versions (Home, Server, Advanced Server...) of the various NT-series OSes; the more expensive versions come with more extras and less deliberate crippling, but they're basically the same thing.
So what Microsoft have done is jimmy the CD file system so that various duplicate files in the three directories are actually made up of the exact same data on the disc - the disc's table of contents points to the same thing for different files. The total amount of actual data on the disc can, therefore, be 486Mb, but not much of that data is different. Most of it is duplicated stuff, the same in all three OSes. When you copy the files to the hard drive, you copy the same CD data three times, to three different directories (maybe only two times, for some files), where it ends up as sets of separate but identical files, all taking up hard drive space.
The utility Microsoft use to make these discs, automatically linking duplicate files, is called CDIMAGE.EXE, and as it turns out it's not hard to find.
To duplicate these sorts of CDs without using dedicated copying hardware, just do an ISO copy, either with a CD burning program that can read and write the ISO-9660 filesystem on the fly, or by reading the CD to an ISO file on your hard drive and then burning a new CD from that file.
