Atomic I/O letters column #19Originally published in Atomic: Maximum Power Computing Reprinted here 26-Mar-2003.
Last modified 03-Dec-2011.
Cheap flat panel?
I just bought a Triplex GeForce 4 Ti4200 and was wondering how I could go about connecting a laptop display to the secondary display plug on my new card. I've looked online for the answer but cant seem to find it. I know that if this project was possible then it would generate quite a bit of interest, as laptop monitors are cheap and easy to come by.
This is a variation on the "how do I connect a laptop display to plain old VGA" question (as seen in this column), and the answer's much the same: You don't, unless you want to pay more than you would for a basic boxed LCD monitor that's made to work with PCs.
The Digital Visual Interface (DVI) output from the second connector on your Ti4200, is, at least, closer to the digital data that the laptop panel wants, so you wouldn't need an analogue-to-digital conversion stage in the driver hardware. But, still, no such driver hardware exists, at least not as a sensibly priced off-the-shelf product. You can't use the hardware from a toasted laptop - the source of many surplus laptop panels - because laptops don't bother producing a VGA or DVI signal to drive their panels. There's no need for them to bother with that, when they only have to drive the one flavour of panel, with no cable to speak of between the video hardware and the display. Sure, there's a VGA connector on the back of the laptop, but that won't help you.
Decent 15 inch flat panel monitors are down below $AU600 now, and if you want to go bargain-hunting you can find older, smaller and/or dodgier LCDs for quite a bit less. Unless you want to spend money on nothing, though, all a bare laptop panel is good for is replacing a busted laptop screen.
When changing my video card, I noticed that my old card had a 26 pin header on it, with no labels other than "JP1". I'm pretty sure that its called a "Feature Connector". What features does it connect to?
The VGA Feature Connector is an old Video Electronics Standards Association (VESA) standard expansion connector for video cards. It was used for things like video overlay cards, TV tuner cards, head-mounted displays and so on; you hooked up the Feature Connector to another card in the computer with a little ribbon cable.
The Feature Connector can only manage 256 colour video at lousy resolutions, though, and it doesn't deliver a full video signal, just raw pixel values that have to be marshalled into an image with the help of data delivered by other means.
Today, you'll find it filed under "goodbye, and good riddance".
M56 Smart Drive
I consider myself to be fairly computer literate, but am intrigued by one setting on all motherboard BIOSes. What on earth is "hard drive S.M.A.R.T. capability"? All the different mobos I've used have this setting, but none give any clues as to what it is.
"SMART" stands for Self-Monitoring, Analysis and Reporting Technology; it's a standard for monitoring disk drive health and reporting problems. When the drive motors, media, or other components start behaving badly, SMART can warn you of an impending disk failure, so you can transfer your data over to a new drive before the old one carks it. SMART can't detect everything that can go wrong with a drive - it can't tell that you're about to kick your computer over on a tiled floor by accident - but it can notice a lot of failures before they happen. That's the theory, anyway.
Windows has supported SMART natively since Win98, but that doesn't mean it actually makes use of it. You have to run monitoring software that looks at the drive and tells you if anything seems to be amiss, or else the drive will be doing its SMART checking, but nothing will be asking it what it's found.
Various all-in-one utility packages do SMART monitoring, and there are stand-alone monitors as well - SANtools' SMART Disk Monitor, for instance.
To swap or not to swap
If you are running Windows XP in a system with 500Mb+ RAM, would your system perform better if you removed the swap file altogether?
When you've got lots of physical RAM (ignoring hard drive virtual memory), and you're not actually using any more memory than the physical RAM you have, you'd think that the swap file would just be sitting there doing nothing.
What actually happens, though, is that applications tend to allocate more memory than they really need, just in case they suddenly do need it for some operation that you might perform. Applications are likely to actually use more than two thirds of the memory that they allocate, but the rest lies fallow.
Operating systems that're somewhat sensible about virtual memory - including WinXP - deal with this by assigning memory allocations to virtual memory at first, and assigning to physical RAM - assuming there's enough of it - only the memory that the application actually uses. This process doesn't cause disk-flogging, because it's not really "swapping" at all. There's nothing in the virtual memory when it's re-assigned to physical memory. So if you've got enough physical memory to handle all of the memory that's actually being used for something, you get to stay out of Swap File Hell, and the swap file just soaks up over-allocations and leaves more physical RAM for actual use.
The other reason not to disable or limiting the size of your swap file is that doing so just gives you a point at which you'll get out-of-memory errors, if the system does at some point need to use more memory than it has physical RAM.
I just recently built my first PC and was wondering if you could tell me what is the maximum temperature for the P4 1.8A Ghz (Northwood) CPU? I read articles and messages posted up on the Web, and they say that P4 should be around 50 degrees C max... however when I played some 3D games, it went up to 68 degrees C. Is this normal?
Well, it's not worryingly ABnormal, though it does suggest that your case ventilation might be lousy, or your CPU cooler might not be attached quite right. Don't have a cow over it, though.
Your processor's got a 67 degree Celsius "maximum junction temperature". That's not the point where the CPU will emit a little pop and die, but it is the temperature above which the CPU can reasonably be expected to get flaky.
The temperature your motherboard reports for the CPU, though, is not necessarily the processor's actual core temperature. The P4 does, at least, have an internal temperature sensor whose calibration ought to be pretty good, and which is reasonably thermally close to the core. But the calibration of the motherboard hardware that connects to that sensor is unknown. Motherboard thermal readings are often inaccurate.
My standard hot-CPUs rule is that if the computer ain't crashing, don't worry about it. Since the P4 has a built-in "Thermal Monitor Feature" throttling mechanism that slows it down when it's hot, though, there's good reason to keep its temperature below the maximum junction temperature, because that's presumably more or less the point where the throttle cuts in.
It's easy enough to tell if your CPU's throttled down - a utility like WCPUID should tell you what the CPU frequency is at any moment. If you're still suspicious, you can also run the CPU benchmark of your choice right after booting the machine, and again when the CPU's warmed up, and see if there's a difference.
I have been using an old Hitachi 22" monitor, and lately it hasn't been going so well. After booting up XP, it starts flickering between full colour and purple and then remains purple.
I know how to do basic adjustments to monitors (i.e. focus) so at least I know how to take the case off. But I'm wondering if there is anything that can be done about my sick Hitachi? I don't want to part with it just yet.
The "purple" colour you're seeing is, I'll bet, a magenta tint, caused by the green channel dropping out. Many monitors develop problems like this caused by plug or cable damage, or by a dud solder joint or failed component inside the monitor itself. Your problem sounds as if it's thermally triggered; something's probably expanding and losing contact as the monitor warms up.
Lose green and your screen will be tinted magenta; lose red and it'll be tinted cyan (blue-green), lose blue and it'll be tinted yellow.
Often, these problems can be solved with nothing but a bit of resoldering, but finding the spot that needs to be resoldered can be tricky. If it's a mere dud cable, and your monitor has a plug-in cable, then you can just change it.
PSU plug plethora
I've noticed that some PSUs have the words "Pentium 4 power connector" in their description. Do Pentium 4 motherboards require a special power connector?
Yes, they do, but it's not just a P4 thing. The extra connector is the "ATX12V" plug and socket, a four pin connector that provides two more 12V supply wires to the motherboard, along with a couple of grounds.
A motherboard with an ATX12V receptacle on it may work OK without the ATX12V connector plugged in, just as a server motherboard may work OK without the "AUX" power plug connected. All these plugs do is provide more wires connecting the motherboard to the power rails. If you don't connect anything to the extra power inputs, though, then there'll be a more resistive path from the PSU to some of the components on the motherboard that need power, and the voltage those components see may be too low.
Some motherboards have an ATX12V socket and also a regular four pin "Molex" power receptacle on the motherboard; you can plug a drive power connector into them instead of the ATX12V plug and they'll be happy. There are also cheap Molex-to-ATX12V adapter cables, which let you do the same trick with normal ATX12V mobos.
Hitting the wall
Why is it that newer hard disks beyond 32Gb don't work on some older motherboards?
My friend's Gigabyte 2000 Dual BIOS P3 mobo has this same problem with his 40Gb disk; he needs to set the jumper to 32Gb, so there goes 8Gb wasted...
I run a K6-2 300MHz with a PC Chips mobo and have run into the same problem.
Because 32 gigabytes should be enough for anyone.
Or, more helpfully, because the 32Gb limit is just one of many hard drive size limits forced upon users by, among other things, address space restrictions. Go back further through the annals of PC history and you'll find limit lines at 32Gb, 16Gb, 8Gb, 2Gb and 540Mb; the limit after 32Gb was 137Gb. Or wasn't, actually, as I explain in more detail in this letters column.
Some PCs that can't handle drives over 32Gb can be cured with a BIOS update, some can be bludgeoned into compliance by using special software from the drive manufacturer, and some can't be fixed at all. Fortunately, the Gigabyte GA-BX2000 and GA-BX2000+ seem to both be in the first category; the latest BIOS from here will, apparently, solve the problem.
Your old PC Chips motherboard, on the other hand, probably can't be fixed.
Pick a number...
When I bought a computer with an Asus P4B533 motherboard and Pentium 4 2.00 GHz processor, I wanted to overclock it, but during the process I faced one problem - I am getting reports of 3 different CPU core voltages, depending on where I look.
The BIOS and Asus' PC Probe utility are showing different CPU Voltage. (I have tried all available BIOS versions.)
For example, if I set Vcore to 1.625V in the BIOS setup, after saving settings and rebooting, the BIOS setup Power -> Hardware Monitor display says 1.68V. But if I boot into Windows XP and run PC Probe, than it says Vcore is as high as 1.712V!
Who should I trust? Is it 1.625, 1.68 or 1.712? That's a big difference. If you don't know the exact Vcore, it's easy to damage the processor. I've read some articles on the Web where people said that Northwood core can not be taken further than 1.7V.
You should probably trust the BIOS Hardware Monitor display. If you're nervous about it, though, nothing short of hunting around the mobo with a thin and pointy multimeter probe will do.
The reason why the numbers vary is that different monitoring programs use different fudge factors.
All hardware monitoring chips of a given model ought to act pretty much the same on the output side, assuming something isn't severely fouled up (their power supply, for instance). So, presented with the same input from a motherboard, a given hardware monitor chip should always give about the same voltage and temperature reading output.
Unfortunately, though, the input side of the hardware monitoring chip has components between it and the actual things it's measuring. The power rails are probably fed to the chip through a voltage-dropping resistor bridge, for instance, and the negative rails have an inverter between them and the chip.
These intervening components don't necessarily have the same values across all motherboards, and the mobo manufacturers take that into account when they set the compensation values for their BIOS setup software. Assuming they get it right - and they sometimes don't - the BIOS should therefore report accurate temperatures and voltages. You won't necessarily get the exact voltage you asked for, but the voltage the BIOS reports ought to be pretty much right. Other mobo monitoring software, though, may or may not get the fudge factors right for a given motherboard, and may therefore misreport both voltage and temperature.
Fortunately, the thermal difference between Vcore values of 1.625V and 1.712V is only about 11% (CPU heat output increases linearly with core speed, but also with the square of the core voltage), and that's a pretty good measure of the amount of extra strain you're putting on your CPU if it is running at the higher voltage. So you're not likely to beat your processor to death as a result of these inaccuracies.
Yes, there have indeed been reports of Northwood cores working for a while at elevated Vcore and then dying, but a core voltage difference of a few per cent shouldn't be enough to cause this problem. It might make it happen faster, but if that's the case, then it probably would have happened anyway.
I hope this cheers you up.