Atomic I/O letters column #103Originally published 2009, in Atomic: Maximum Power Computing
Reprinted here February 3, 2010 Last modified 16-Jan-2015.
I've got a year-old media rig equipped with a 2.something-GHz Intel Core 2 Duo CPU, that's been working flawlessly for all my media needs since I bought it. Now, however, I'm contemplating getting a bigger monitor. I've got an old 19" 1280x1024 LCD, and I'm thinking about getting a 32" HDTV to use as a screen. (I have of course drooled over your monster screen several times, but I can't justify all those pixels, as I mainly use it for video.)
My worry is that the CPU won't be able to deliver enough pixels (1920x1080 or 1920x1200) in real-time. My question is simply whether a faster graphics card would help.
I'm now using the built-in GMA X3500 video of my Intel G35-chipset motherboard, that claims to support "1080p HD video" if using Vista. Due to Hollywood lobbying they won't let me use some of the driver acceleration in XP, though.
I'm mainly using VLC for video, and I'm unsure how much hardware acceleration it can use. Not much, I'm afraid.
Will I be able to view HD content with such a set-up, or should I upgrade the CPU? Would a serious graphics card help at all?
(Of course, most of the video won't be full HD, just up-scaled DVD, or 720p content "found somewhere on the net". But it would be fun to watch the occasional Blu-Ray, now that PC Blu-Ray drives are becoming reasonably priced.)
Note, I'm not interested in running Crysis, or doing any kind of real-time ray-tracing, which is why I've so far kept to my on-board graphics.
Yes, your onboard video will probably choke on some content, and a better graphics card - which is to say, pretty much any actual current-model separate graphics card - is likely to help with that.
There are so many combinations of data codec, video hardware capabilities, driver versions and playback programs, though, that pinning down exactly what will and won't play back smoothly can be a pain. If I were you, I'd just buy my new monitor and start playing stuff, and see if it works. If and when something doesn't (and you're unlucky enough to notice) is the time to buy more gear.
(Note that you also need to make sure that the HDTV you buy will actually work from ordinary computer input, without the magical pixie dust of HDCP Copy-Control Crap sprinkled everywhere. More on this here.)
Assuming that some HD stuff doesn't play smoothly on the new monitor - which is probable - then if the only problem is Blu-Ray, all you'll need is a graphics card with hardware-and-driver support for the melodiously-named "H.264/MPEG-4 AVC" and "SMPTE VC-1" codecs, plus Blu-Ray's dialect of MPEG-2, for older discs. Your built-in GMA X3500 graphics adapter actually does have VC-1 decoding built in, but for HD H.264 from integrated video you'd need at least a G45-chipset motherboard, or some other unusually-muscular integrated graphics adapter, like the one on Nvidia ION boards.
Nvidia and ATI's GPUs and drivers have been able to handle Blu-Ray for some time (well, the Windows versions have, at least). The drivers also help with some other codecs, plus deinterlacing and colour-correction and so on. Nvidia call their version of this "PureVideo"; ATI call theirs "Avivo".
You don't need a high-end graphics card. GPU-assisted 1080p decoding was working nicely on GeForce 8800s at the end of 2006, and all sorts of unassuming fanless cards can do it now. Nvidia have a PDF comparison chart on this page that tells you just how cheap you can get without losing Blu-Ray codec support.
Be prepared, though, to have to use more than one player program. If people were sensible then every player would work with every kind of content, but Copy-Control Crap and weird codecs mean that you're almost certainly going to need something in addition to VLC, like the Home Cinema mutation of Media Player Classic, and/or an actual licensed DVD/Blu-Ray player program like CyberLink PowerDVD. (You ought to get a licensed player bundled with your Blu-Ray drive.) VLC is really good, but it's not perfect.
There are also some particularly pathological video files that're badly undercompressed, giving them such a monstrous data rate that CPU and storage bottlenecks that normally don't matter at all suddenly give you stuttering playback. The only files I've found that're like this, though, are game-demo video clips, where whoever made the clip jammed every slider to maximum and so made a 125-second clip that's 500 megabytes in size.
I recently bought a Compaq ProLiant DL360 G2 on eBay, with the intention of using it as my SmoothWall machine. It has a stonkingly LOUD fan that I am unable to slow down with software (because I don’t know how to in SmoothWall). I was hoping I could solder on a few resistors and let them do the work, but I am not sure what I'm doing.
The fan is a 24-volt 0.65-amp behemoth, with 4 wires coming from it: Yellow, Red and 2x Black (though I suspect the black are spliced together on their way back from the fan, as only 1 black goes into it). My goal is to make the fan vaguely controllable so I can dial it right down to maybe 1/3 or 1/4 of its stock speed. Failing that, I'd settle for hardwiring it to 1/4.
Can you tell me what resistors I need and what wires I should be soldering them onto? I have a multimeter handy but am not quite sure what I should do with it.
The DL360 G2 is a one-unit rackmount server, and those are often loud, if they've got high-powered hardware in them. There's no room for big fans in a case only 44mm high, so they instead have small fans and/or blowers, turning really fast.
It's possible that the fan's current-draw specification is the peak current when the fan's spinning up, not when it's running at a constant speed. But I wouldn't be surprised if it really was a 24 * 0.65 = 15.6-watt fan. Which will indeed be very noisy, if it's not the size of a desk fan.
You can use Ohm's Law to figure out power-dropping resistors in a situation like this - but you don't necessarily actually want to.
Ohm's Law is V=IR, voltage equals current times resistance. If V is 24 volts and I is 0.65 amps, the equation tells us that the effective resistance of the fan is 36.9 ohms. If you double the resistance of the circuit by putting a 36.9-ohm resistor in series with the fan then current, and power, halve, greatly quietening the fan. Another 36.9 ohms in series with the fan would drop current to 0.325A and power to 7.8-watts, split evenly between the fan and the resistor; that means you'd need at least a four-watt resistor, to prevent it from burning up.
In the real world, there's no such thing as an off-the-shelf 36.9-ohm resistor, and 4W ratings aren't common either. You'd probably end up getting a 5W 33-ohm resistor, which'd be near enough.
(You can also wire resistors in series and parallel to get the resistance and power-handling values you want. If you can only get 2.5W power resistors, for instance, you'd take two 68-ohm 2.5-watters and wire them in parallel, and that'd give you a 34-ohm 5-watt result, which would also do this job fine.)
OK, now the reasons why you may not actually want to try this.
1: The fan may not even turn from a lower voltage. There's a decent chance that it'll turn from roughly half voltage, but there's no guarantee. It's also possible that the bearings may get stickier over time, so a fan that starts fine from lower voltage today will fail to start in a couple of years.
(Many, many people have for one reason or another ended up with fans that turn OK once they're going, but need to be poked with a pencil to get them started. It is important to remember to do this poking whenever powering up the device in question.)
2: The fan may be turning at this alarming speed for a very good reason. Cooling is a big problem for 1U servers, and if this one's using one fan for almost all of its cooling, turning that fan down may cause overheating in the short term, or early death of components in the long term.
Note also that this server seems to have some other small fans in the power supply, if this specs page can be believed. They're either very thin fans of reasonable diameter, or little 40mm units. Either way, they won't make a lot of noise by themselves, but may also not last very long. Small server fans shouldn't crap out after 300 hours like the similar-shaped cheaper fans in consumer hard-drive boxes, but they may.
Fortunately, there are ways past all of these problems, provided you don't have to install this server in its intended habitat, a cheek-by-jowl rack with other hot gear above and below it. If you can give the server plenty of breathing space - which you probably can - then you can just replace the standard fan with a much quieter outboard cooling arrangement.
Figure out where the air's going in and where it's coming out - including the power-supply-fan air-flow, so you don't end up trying to suck air in through PSU exhaust fans - then take the case off the server and hack holes in it to install normal PC fans in. You should only need to do this to the top panel. You won't necessarily even need to remove the stock fan - just unplug it.
With any luck, the server has a standard 12V fan socket or three on its mainboard. If not, no big deal; you could power 12V fan(s) from a 12V wire from the PSU, a separate hacked plugpack, or an old PSU.
You could even use the standard 24V fan rail to run 12V fans.
(Note that I'm assuming, here, that the standard fan runs from 24 volts DC. I'd be very surprised if it didn't, but I've found myself very surprised on numerous occasions.)
Two 12V DC fans in series will work fine from 24V DC, and the stock fan in your server has such a high rating that you could easily use SIX low-power 120mm fans (with 0.1-to-0.2-amp current ratings), in three parallel strings each containing two series-wired fans, presented in NetHack graphics below:
| + - + - |
| + - + - |
+ - + -
This many fans is unlikely to be necessary, mind you. Only two parallel-wired-series-strings-of-2-fans would probably give you more than enough cooling. Or if you wanted to get clever, you could use a single string of three series-wired 12V fans from 24V, so they'd each see 8V and run at roughly 44% power. Most 12V computer fans will work A-OK at reduced speed from 8V; 6V is pushing it, but 8V is a good bet.
You could hang a mains-powered 120mm fan off the top of the case too, if you liked. This is not, of course, a wise project for beginners. With low-voltage stuff, the worst you can do is cook your server. Mishaps with home-made mains wiring can be a lot more exciting.
Back in the DC world, you could try one of those huge 20cm-and-higher fans that are now quite commonly seen on the side, or even the front, of PC cases. They're usually quiet even from a full 12V. (As a general rule, increasing the fan size while keeping the power the same gives you more air flow and less noise, the bigger you go. The giant fans usually don't draw a lot of power, either.)
Anyway, that's the basic idea. Top off, hack a hole, install a fan - with gaff, if you like - add a finger-guard if you're a big wussy girl, and find some way to power it. Server quietened, warranty voided, job done.
(About ten minutes after I put this page up, a reader pointed out that the HP ProLiant Support Pack includes fan-speed control for Red Hat and SUSE Linux; anybody with some Linux experience ought to be able to get it working in SmoothWall as well. SmoothWall is specifically made so you can set it up via a Web-browser interface without needing to know anything about Linux, though, so the Support Pack won't be much help for many users.)
Are you aware of any secure way to buy MP3s from overseas services like Amazon without getting hit by the "You live in the backward Australian IP address range, and we don't allow sales of MP3s to that part of the world" roadblocks?
I figure the most likely solution is to use some form of proxy to defeat the region checking. My big problem is finding one that, one, is secure (free proxy lists don't fall into this category), two, presents an IP from an acceptable geographical region, and three, can be had for much cheapness (as otherwise I'm better off with sneakernet at the local JB).
(I've been chasing this for a couple of years, but haven't really found a good solution yet.)
I think, but am not certain, that it's as simple as going to https://www.amazon.com/ via Tor, and using a Tor exit node in the USA (which is where most of them are anyway). Amazon lets you start out like that, by going straight to https://www.amazon.com/ for an SSL connection, and https-via-Tor is I think as secure as https generally.
(I think the pitfalls for SSL over Tor apply to people doing ordinary browsing, where you often start out at an unencrypted http://www.whatever site and then click something to switch to https, and also just do general browsing and Googling and such with Tor turned on. This leaves you at least theoretically vulnerable to certificate fraud and other weirdness from the exit node; as I've written before, you shouldn't trust Tor exit nodes. Nobody without strong reasons for anonymity and/or Great-Firewall-avoidance wants to use Tor all the time anyway, though - it's far too slow. Which means it'll probably also be pretty painful if you're downloading 50 hours of 320kbps MP3s, but them's the breaks.)
Alternatively, there are companies that'll sell you a proxy or SSH-tunnel service from wherever you are to where they are, like Cotse.Net (who are in the States). This is overkill for just buying MP3s, but a cotse.net account with all mod cons - and honestly, the feature list goes on and on and on, and includes other stuff you might like - only costs $US5.95 a month. So what the hey. It should be fast, too.
Note that your credit card could be a problem too, if it's from an Australian bank.
(Shortly after I replied to Larry, TechCrunch published an article on this subject.)