Atomic I/O letters column #88Originally published 2008, in Atomic: Maximum Power Computing Last modified 16-Jan-2015.
I am trying to cut some corners with some leftover pieces I have, to put together something to sell. I would like to set this computer up with a RAID 0 setup, because I only have two 10Gb hard drives, left over from a couple of Xboxes I have modified.
I've got a Foxconn 6100K8MA-RS socket 939 motherboard, Opteron 165, 2 x 512Mb Corsair RAM, Lite-On DVD burner and Biostar GeForce 8500GT graphics card, as well as the two 10Gb Western Digital hard drives. I am planning on installing Windows XP Pro on this.
Could you please help me get this started? It is something I have never tried and I am not really sure where to begin.
I presume RAID 0 appeals to you because those old 10Gb drives just aren't unreliable enough for you. Making a stripe-set out of them will, of course, add their unreliabilities together!
OK, point one: If those drives haven't been "unlocked", you're not going to be able to use them in a PC at all.
Point two: Your motherboard only supports RAID on its SATA ports, not the PATA ones. And XP Pro can make software stripe-sets, but it can't boot from them. So I guess you could get adapters to convert the drives to SATA and do it that way, but this is still a pretty pointless exercise.
I mean, a sixteen-gigabyte CompactFlash card now costs well under a hundred bucks delivered. That fact is the universe telling you to boot this computer from some other drive.
If you really want to use the Xbox hard drives in this new computer, go right ahead, if they're unlocked. You could put swap files on them, and thereby reduce flogging of the main system drive. But that main system drive should be a cheap new SATA unit. Reliability and capacity aside, a new drive will be much faster than the old Xbox drives, even if you RAID 'em.
That's partly because a new drive will almost certainly be a 7200RPM unit, while the Xbox drives are 5400RPM. But it's mainly because any new drive will have much higher data density than the old drives, so far more data will pass under the heads per platter revolution.
Personally, I'd crack the old drives open and harvest the platters for wall decorations, and the voice-coil magnets for toys.
Can you advise what is a reasonable way to clean an LCD screen?
Much conflicting advice on the Web. I don't want to buy proprietary cleaning kits/stuff if not necessary.
Get a cheap microfibre "glass-cleaning" cloth from the supermarket. Dampen it a bit and it should be good enough for most monitor gunk. Press only gently, because the sandwich structure of LCD monitors is a lot more fragile than the thick glass pressure vessel of a CRT.
Add a dash of dishwashing liquid, or even ammonia, to your water if there's greasier stuff on the monitor. Alcohol (isopropyl or just meths) will also work well, but anything more than soap and water may damage coatings on the monitor over time.
Just a dash of alcohol in a bowl of water shouldn't hurt anything unless you really make a habit of using it, though, and that should be more than enough to clean even a very grody screen.
Also, don't go nuts desperately trying to find every tiny speck on the monitor. Yes, you can see more dirt if you turn the screen off and examine it closely - but why bother cleaning off dirt that you can't see when the monitor's on?
(I don't know if you actually have a tendency toward this sort of obsessive-compulsive over-cleaning, but I mention it anyway, because I do.)
Something weird just caught my eye in the ThinkPad X300 review here:
"I noticed that the solid-state drive on my review machine was far from silent — it makes some noise when electrons are being shuffled around. It's quieter than a normal hard drive, but you can still hear it."
What's going on here? I doubt it's the sound of thousands of electrons moving in unison.
A current through a conductor creates a magnetic field. That field will apply a force to anything nearby that's affected by magnetic fields, which includes other conductors that have current flowing through them. The result is vibration in time with the current flow.
(In the days before strong permanent magnets, all of the magnetic fields in normal electric generators and motors were created by current through conductors. This extended even to specialised "motors", like loudspeakers; the modern permanent-magnet loudspeaker, which doesn't need a field coil, was a great advance. Early permanent-magnet motors used alnico magnets, which needed to be periodically remagnetised.)
This is why transformers hum, and it's why all sorts of solid-state hardware also makes noise, though lower-powered devices may make too little of it for a human to hear, or operate at too low or (more commonly) too high a frequency. Any AC in the audio range, from the 50/60Hz of mains current to the several kilohertz of the inverters that drive cold-cathode backlights in LCD monitors, can create a matching audio-frequency vibration of some component or other.
In this case, the solid-state drive itself may or may not actually be the source of the audible noise. It could be power supply components on the mainboard buzzing when the drive's drawing more power, for instance.
I have an Abit AB9 QuadGT with 2 x 2Gb of DDR2 RAM. My video card is a GeForce 7950 GX2 that was a hand-me-down from a friend (free!), and it looks like I'm experiencing what you are saying is a "3Gb barrier".
This board does not have a memory remap feature, and I'm running WinXP x64. In BIOS it shows 3.1Gb of memory, and it does the same in Windows. Is there any way to somehow jiggle this memory free, at least in the OS, or am I just screwed?
Also, do you think this video card is the culprit? What if I upgraded it?
Once you get past the BIOS stage of startup, nothing can do anything about the "memory holes" created for Memory-Mapped I/O (MMIO) for video card and other RAM.
If your BIOS itself is reporting 3.1Gb, that probably means that there's an option to "reserve memory" above 3Gb, and it's turned on. But if you've got a big ol' slab of graphics-card video memory as well, which you do, then it doesn't really matter if the reserve-memory option is turned on or not, since that big ol' slab is going to slap its huge MMIO footprint down into your fourth gigabyte of RAM anyway. So it doesn't matter if the BIOS was already told to not let the system use it.
If MMIO-induced RAM-disappearance problems can be cured, they have has to be cured at the very early boot stage. After that, there's no way at all to get the "shadowed" RAM back.
Actually, with the gigabyte of video memory on your GeForce 7950 GX2, you ought to have significantly less than 3Gb of real accessible system RAM. Dual-GPU cards, like multiple-graphics-card SLI or CrossFire setups, are particularly bad memory-hole offenders, because each graphics unit has its own slab of RAM, containing almost the exact same data as the other slab.
I don't know whether there's a memory-"hoist" feature in your motherboard's BIOS setup program. If there isn't - and I wouldn't be at all surprised if there weren't - then there is indeed nothing more you can do to make your RAM visible.
You could, however, add more RAM. Since you're running XP x64, memory above the 4Gb line will be perfectly accessible. And your motherboard uses DDR2, which is dirt cheap these days. Your mobo can also accept up to 8Gb of RAM.
If you already had all four memory slots filled, you'd have to scrap some or all of your existing RAM. Even if you bought four all-new 2Gb modules, though, you'd only be paying $AU200 to $AU300, depending on the speed rating.
As of now (early December 2008), there are lots of graphics card options that'll give you significantly better 3D performance your memory-hole-hog 7950. Any 512Mb card will give you back 512Mb of system RAM compared with what you've got (oddball BIOS memory-reserving options aside), and your options are pretty much wide open; even a humble GeForce 9600GT for $AU170 or so will be a little faster than what you've got.
The bang-per-buck winner for you, assuming you don't feel compelled to get a cutting-edge card, would be something like a GeForce 9800GT (actually a little slower than the 8800GT, but also cheaper) or Radeon HD3870.
I'm going to upgrade my rig this year and am considering a modular power supply.
PC Power and Cooling insist that corded power supplies are superior because "modular" power supplies, the kind with plug-in output cables, increase resistance in the lines. I've looked through many forums and how-tos but haven't seen this issue addressed.
What's the real deal?
Yes, the extra connectors in a modular PSU's plug-in leads do increase resistance. And PC PSUs need to deliver low voltage at high current, which is exactly the situation where connector resistance can really make a difference.
In the world of radio-controlled cars, there's a particular kind of cheap white-nylon pin-and-socket battery connector that's commonly supplied as standard equipment. When people upgrade the motors in their cars and start moving really impressive amounts of current when accelerating and braking, those cheap connectors often get rather hot, even to the point of melting the plastic housings and becoming impossible to unplug.
That's partly because the connectors get dirty and loose, but when they're new they still cannot help but work no better, and quite likely considerably worse, than the output-cable connectors on modular PSUs.
I just grabbed one such male-and-female R/C connector pair from my bits box, and measured the resistance from the negative side, through the connection, through a paper clip stabbed into the snipped-off wires on the other end, and back through the plug again to the positive side.
Just connecting my test leads to each other, the reading was 0.0 ohms. Adding the crappy plug and paper clip to the circuit gave me... 0.0 ohms.
So I switched to a possibly-more-precise analogue-digital hybrid multimeter. That one reckoned the resistance of the test leads (which included a couple of dodgy alligator-clip leads, themselves each adding two small-area metal-to-metal contacts to the circuit) was 0.3 ohms.
When I added the lousy nylon plug and paper clip, the reading rocketed up to... 0.3 ohms.
So I changed tactic. I replaced the paper clip with another dodgy skinny alligator-clip test lead, put five amps from my regulated bench supply through the contraption, and measured the voltage drop across it. 0.25 volts.
Then I measured the drop across the skinny alligator-clip lead by itself: 0.232V.
That meant the voltage drop across the two sides of the connector was 0.018 volts at five amps, which indicates a resistance of 0.0036 ohms. Let's be pessimistic and say 0.004.
Let's also say that you insert that much resistance into a 100%-loaded output from a thousand-watt PC PSU.
That sort of PSU will be able to deliver something like 25A at 3.3 volts, which makes the 3.3-volt rail the most resistance-sensitive output - even if you combine the multiple 12V rails, they won't quite equal it. If 3.3 volts causes 25 amps to flow, there must be about 0.132 ohms of resistance in the circuit. Add 0.004 ohms to that, and the resistance of the circuit has only risen by three per cent.
But the 3.3V output from the PSU doesn't come out on just one wire. There are eight +3.3V wires on a standard ATX connector. Split the current between all of those, and adding 0.004 ohms to each of them is the equivalent of adding 0.0005 ohms to a single-wire output. Which brings the resistance increase down to less than 0.4%.
All of those outputs from a PSU are also 16AWG wire, at the thickest - actually, everything but the ATX plug on many PSUs is fed by thinner 18AWG wire. 16AWG copper wire all by itself has a resistance of about 0.013 ohms per metre. About a foot of the wire all by itself, therefore, has the same resistance as both sides of a cheap nylon R/C battery connector put together.
Since all of this wiring is feeding PC hardware which either has its own quite heavy-duty regulating and smoothing circuitry (PC CPUs and graphics cards), or will just inherently tolerate a 10% sag with no problems (pretty much everything else), the fraction-of-1% sag that one extra connector may possibly add to the equation seems very unlikely to ever matter.
(I later tried the plug-resistance experiment with a chain of five cheap Molex-plug Y-adapters - considerably cheesier than the connectors on modular PSU cables. All five of them put together gave a five-amp voltage drop of 0.55V. That means they managed only 0.11 ohms between them, or 0.022 ohms each.)