Atomic I/O letters column #56Originally published in Atomic: Maximum Power Computing Reprinted here April 2006.
Last modified 16-Jan-2015.
I've been an avid benchmarker of my PCs for a very long time now. As soon as I get a new PC, I load Windows, the relevant Service Packs and the latest drivers for all of the hardware. Then I sit down to a warm night of benching.
One day, after running through 3DMark05 (and watching my new PC crawl through the tests), a question started to bug me: If I buy the latest and greatest hardware, and it still performs like crap in these benchmarks, then what sort of 'off-world' hardware are these Futuremark guys using when they want to test their new benchmarks? If the new benchmark is designed to test the next generation of PC hardware, where are they getting their "next generation" hardware from? How do they know if their benchmark is going to turn out right?
I'll be able to sleep a few hours more each night if you know the answer to this.
Hardware companies may give pre-release gear to some software makers - mainly game companies, but benchmarking outfits as well - but lead times are pretty short in PC hardware. No game coders are signing NDAs and collecting their GeForce 15000s yet.
Instead, people writing code for hardware that doesn't exist yet use emulation. Sometimes that emulation's running on a more powerful (or at least much more expensive) system, like a workstation or the custom development boxes that console coders use, but it doesn't have to be. You can emulate any level of 3D acceleration on pretty much any other hardware. If necessary, you can do the whole darn thing in software, through any video adapter at all.
That's achingly slow, of course, but it's not necessarily a big deal for most game development. There's a lot to a game besides the graphics; most of the development can be done with low-polygon basic-shader placeholders, and indeed may have to be, since final art for a modern game can take a very long time to finish.
Futuremark are an extreme case, because 3DMark can be thought of as a severely unfinished game. If you're writing a real game, you have to be careful not to add features that you're later going to have to take out to get the thing to run tolerably well on normal computers. Futuremark don't have to worry about that - the whole idea of each new 3DMark version is that it should bring current hardware to its knees.
Emulating future hardware doesn't work perfectly, because the specifications the emulator's written to match - and against which the final code is probably checked for official correctness - don't necessarily match what the graphics chip companies end up implementing. But it beats building a time machine.
Every so often a CD I have will fail (usually a backup CD) and the error message I get is a "Cyclic Redundancy Check". Most of the data on the CD can be read, however some cannot. I'd like to know exactly what a Cyclic Redundancy Error is. Some sort of failure within the CD? Is there any way to fix the problem and recover the data?
Cyclic Redundancy Check, or CRC, is a very simple error-detecting system. It's computationally cheap and it doesn't need much space for the "hash" it produces from the input data, so you'll find CRC error detection all over the computing world. On CD-ROMs, the CRC data is in one of the "subcode" channels that are evenly distributed through the disc along with the data you can actually access.
When you get a CRC error, that just means the data on the disc is corrupted for some reason (or, less probably, that the data's fine but the CRC's corrupt). Almost always, this is because of disc damage. There's nothing you can do about it (besides checking the disc for muck or scratches you can possibly repair - see this column), because CRC doesn't provide any facility for error correction. CRC tells you when something's wrong, but it can't fix it.
To recover as much data as possible from a damaged CD or DVD, try IsoBuster.
I'm heading out on a road trip out west soon, and as such my only source of power will be a 650W generator. I was wondering whether or not it was a good idea to charge my TPG widescreen laptop off the generator, or whether the irregular spikes would be too much for the power supply to handle?
It ought to be A-OK. Switchmode power supplies of all kinds, including PC and laptop PSUs, are completely insensitive to power waveform, so they'll run fine from the output of any inverter or generator, even if it's an ancient squarewave unit that'll pretty much set a power drill on fire.
There's always the possibility of disaster, of course, and the management assumes no responsibility blah blah blah. If the generator's old and crusty (or just cheap and nasty) then it might be outputting too high a voltage, or have really spiky output that might cause problems. But, realistically, you're probably going to be fine.
LCDs don't get screen burn? Yes they do. But how?
My brother bought a very expensive flat panel LCD monitor, a nice 19 inch one that rotates on its stand for portrait and landscape view.
The other day I plugged it into my laptop, and at the bottom of the screen, while the light blue Windows booting screen was on, the Windows XP Taskbar was clearly visible. With my brother's system tray icons and everything. I was booting Win2K, so I know it wasn't mine!
The screen's not more than a year or so old, and I've never seen this on any LCD before, and indeed thought it was impossible. I just wondered if you had ever heard of this before?
Yes, LCD screens can suffer from "image persistence". It's not burn-in, though, in the CRT or plasma sense. The subpixels aren't actually getting any dimmer anywhere on the screen, and the "burn-in" shouldn't be visible when the monitor's displaying a white screen. But some sort of capacitive build-up does happen, giving the subpixels a bias depending on what they've been displaying for a long time.
The cure is supposed to be just turning the monitor off for a while (power-saving screen blanking ought to do the job, too), but it might be an unfeasibly long while. Some people report the image persistence just doesn't go away, and so might as well be CRT-style burn-in.
And, heck, maybe sometimes it is. It's not ridiculous to suppose that the characteristics of the thin film transistors in the LCD panel sandwich might change depending on how long and how often they're turned on. If this is the case, though, I don't know which kinds of panel are particularly susceptible.
My PSU had been vibrating a lot lately so I decided to take it apart and oil the fan. One thing I forgot was the orientation of the fan, but I figured it would be better if the fan sucked cool air from the outside and blew it over the heat sink.
My question, though, is: Was there any real danger in taking apart the PSU, or am I obscenely lucky to be able to type up this email?
I didn't take any real precautions; just unplugged the computer and pressed its power button before opening it up.
Your guess about the fan direction is logical, and perfectly matches the way PSU fans were meant to work in the original ATX spec. And it's wrong.
The standard airflow direction for PSU fans, enshrined in the revised spec that codified what everyone was doing anyway, is from the inside of the computer case to the outside. Yes, that means the PSU's breathing air that's been warmed by the computer, but it also means it's contributing to throughflow ventilation of the case. Front fans are, by convention, intake fans.
If you want to set your case up so the air flows back-to-front or top-to-bottom or whatever, that's fine, but if you've got a PSU fan and a front fan that're both trying to suck air into the case, they'll be fighting each other. You have to make sure that all of the fans are turned to match your brilliant ventilation scheme.
It is, however, pretty safe to open up a PSU that's not still plugged in.
The big electrolytic capacitors in PSUs can hold a charge after the PSU's turned off, but probably won't. Well-designed PSUs put bleeder resistors across their caps when they're turned off, draining them long before you could finish unscrewing the casing.
There's always the possibility that you've got a badly designed or faulty PSU, though. In that case, caps can hold a charge for ages (as in, weeks, easily). For that reason, treat all caps as if they have barbed rusty spikes sticking out of their terminals, until you've put a meter across them or just discharged them yourself (with, by preference, a resistor across the terminals; it's fun to discharge a cap the bang-and-weld way by sticking a screwdriver across it, but that's not good for the component).
If a cap is charged to some risky voltage, touching its contacts (or things connected to those contacts) will only give you a brief scary jolt at worst, unless you manage to touch one contact with one hand and the other contact with the other, in which case the jolt will be across your chest, which can be Bad.
Shocks from hand to foot can be lethal too (hand charged, foot earthed), but actual "primary" injuries from electric shocks, where the shock itself is what hurts you, are actually surprisingly rare. People who're injured by electricity are usually, in fact, injured because muscle spasms make them fall down and hit their head, jump off their ladder, stab themselves with their screwdriver, or whatever. If you're sitting at the kitchen table fiddling with an unplugged PSU, there probably aren't many such hazards waiting for you.
All usual disclaimers apply, of course. If anyone out there lights themselves up as a result of following my advice, then I never said a thing.