World's smallest video server?

[RJM62]

A client gave me an old Acer One Aspire something-or-other. The one with the 8 GB SSD. Basically a pretty limited-use machine even if it was working -- which it wasn't. It was one of those situations where I think he would have paid me to take it away. He really hated it that much.

It had about 70MB of free hard drive space, despite almost nothing having been installed on it, just from temp files, caches, windows update uninstallers, and so forth. I deleted all that nonsense and got the free space up to about 1.8 GB.

The user profile was also corrupt, which also was an easy fix. But once I got Windows to work, I had no idea what to actually do with the machine.

Then I looked at my electric bill and realized how much money I'd saved once I switched over to a laptop for most of my Web design work, and decided I could save even more by moving the server for the live turtle tank video feed to the little Acer laptop, as the video server runs 24/7 and was drawing about 160 - 175 watts. It doesn't really have much of a load because all it does is run the Flash encoder and serve one video stream. I use a reverse proxy on the Apache server upstream to actually serve the video over the Web.

So I moved all the software on to the laptop, changed the port forwarding on the router, installed the webcam software, and voila: Turtles.

I figure this should save me about $15.00 - $20.00 a month in electric costs, not to mention reducing my "carbon footprint."

-Rich

 

[Steve Foley]

the video server runs 24/7 and was drawing about 160 - 175 watts.

I really find it hard to believe that any modern system can draw that much power.

Back in 1987, I was building XT Clones with gigantic motherboards (albeit with no heatsink needed) and spinning heavy 5-1/2" Full height (size of two CD drives stacked) hard drive. All this on a 150 watt power supply.

Does generating and dissipating the heat of the new systems really consume that much power?

 

[RJM62]

I really find it hard to believe that any modern system can draw that much power.

Back in 1987, I was building XT Clones with gigantic motherboards (albeit with no heatsink needed) and spinning heavy 5-1/2" Full height (size of two CD drives stacked) hard drive. All this on a 150 watt power supply.

Does generating and dissipating the heat of the new systems really consume that much power?

I'm not sure how it breaks down, but I do know that's how much it was drawing. I had the ConEd guy test it when he was here trying to figger out why my blasted electric bills were so high.

My now-retired desktop used to draw about 280 watts, but it had four hard drives and half a dozen fans.

I suppose they'll always draw considerably more on the input side than is being drawn on the output side, anyway, since the rectification process is not 100 percent efficient. Some of the power is lost to heat even before it does anything useful.

-Rich

 

[fgcason]

Back in 1987, I was building XT Clones with gigantic motherboards (albeit with no heatsink needed) and spinning heavy 5-1/2" Full height (size of two CD drives stacked) hard drive. All this on a 150 watt power supply.

The 150W was aftermarket stuff for more capability. At the time, 70-120 watts was a normal power supply. The 70W power supply would run an XT with a 5.25" hard drive and 5.25" floppy drive along with a lot of other stuff added on that most people didn't have.

When I was putting together a tower computer several years ago, it came with a 400W power supply which was considered absolute minimal power capability which didn't seem right if everything is now more efficient. I was told I need at least 600-800W minimum to run the computer without problems - it still works to this day and I'm convinced the 400W is excessive.

 

[mantakos]

I used to use one Dell Inspiron 9150 as a firewall/router/server, another as a desktop, and a Mac mini for doing iPhone development.

Then one day I bought one of those power meter thingies and started hooking everything up to it to see how much power it drew. I found that my two Dells were using 275 watts combined, while my Mac mini only drew 25 watts.

So I got rid of the Dells, and now I use the Mac mini as a desktop, and I picked up a used Linksys WRT54GS, which I installed OpenWRT (a stripped down version of Linux) on, which now serves as a firewall/router and wifi access point and draws only 7 watts.

It's amazing how much power can be saved just by beginning to pay a little bit of attention to it.

In the winter it doesn't make much difference, but in the summer these boxes are like little space heaters, and require additional power to run the AC to counter the heat they generate.
-harry

 

[mantakos]

... Back in 1987, I was building XT Clones with gigantic motherboards (albeit with no heatsink needed) and spinning heavy 5-1/2" Full height (size of two CD drives stacked) hard drive. All this on a 150 watt power supply....

And back then the CPUs didn't need heat sinks, but as transistor count and clock speed have climbed, power consumption has gone way up. And now we have one little super-computer as the CPU, and a second little super-computer just to render bad guys with machine guns for us to shoot at.
-harry

 

[flyingcheesehead]

Back in 1987, I was building XT Clones with gigantic motherboards (albeit with no heatsink needed) and spinning heavy 5-1/2" Full height (size of two CD drives stacked) hard drive. All this on a 150 watt power supply.

Does generating and dissipating the heat of the new systems really consume that much power?

Granted, your systems back then had certain parts that consumed much more power: Big hard drives compared to today's small ones (or SSD's which don't need a motor at all), for example.

However, today's processors consume a LOT more power than they did back then. What uses power in a processor is when something is switched from 0 to 1 or switched back. While the voltages used in processors are somewhat lower now than they were then (5 volts on the original Intel 80386 you were probably using then, and 1.2 volts on a typical new chip today), the clock speeds are MUCH higher (12MHz on the 386, up to 3.33GHz today). All else being equal, the CPU's power consumption alone would be 16 *times* what it was then, and since most CPU's today actually have several processor cores on a single chip, you can see that the CPU's of today are consuming a lot more power than they were then.

Because of the increased power consumption of the CPU, there is also a lot more heat that needs to be dissipated. That means fans - Usually multiple fans.

Add to that the extra power used by dedicated GPU's that are pretty much standard today, and the power consumed cooling them... Finally, wireless networks are common now and use more power than the equivalent wired networks of yesteryear.

So, power consumption on your average desktop computer has gone up quite a bit.

 

[Steve Foley]

5 volts on the original Intel 80386 you were probably using then, and 1.2 volts on a typical new chip today), the clock speeds are MUCH higher (12MHz on the 386, up to 3.33GHz today)

Actually, the XT clones used an 8088 processor, with a 'turbo' switch between 4.77 and 8 Mhz. The 'Super-Turbo' models were 10Mhz.

The first 80386 chip I saw was 16Mhz. That was a screamer.

 

[flyingcheesehead]

Actually, the XT clones used an 8088 processor, with a 'turbo' switch between 4.77 and 8 Mhz. The 'Super-Turbo' models were 10Mhz.

I remember the turbo switches... Didn't they even have them all the way into the 486 era on some machines?

 

[fgcason]

I remember the turbo switches... Didn't they even have them all the way into the 486 era on some machines?

A few 486's did though they started going away with the later 386's. I always liked the early geeky types with 386-16 that couldn't keep up with my 20mhz 286.

Anyone remember that a few of the old 8088 games that were processor speed dependent like Jet and Falcon? Nothing was more entertaining than having a smart mouth better than you type A personality flying one of those at 4.77mhz and tossing the switch to 10mhz on them mid flight. Approach speed instantly went from 150kts to Mach 1.3 before they knew what the unlabeled switch did.

I put together a ROM board for my 286 that loaded DOS directly into RAM on startup without even touching the HD until it was running. Boot time went from seconds to near instantly. Flip the power switch on and before you can get your finger off the switch, it's been waiting for you for a while.
Of course windows today is a totally different convoluted mess. 2.5 billion instructions per second and it still takes a couple minutes to boot up.

 

[Steve Foley]

Boot time went from seconds to near instantly.

I've got an Alphasmart Dana Wireless that I use to take notes in class. It's got a full size keyboard but runs Palm OS, so powers on instantly. Sometimes it takes 3 or 4 seconds to scan the SD card for new files. It reminds me of an old CP/M laptop I had a long time ago. A small LCD display, but still held more than I could type in a month. This one saves in MS-Word format and will hotsync through a USB cable, or wirelessly, when my router is in a good mood.

 

[nyoung]

I have a Kill-a-Watt meter which allows realtime monitoring of power consumption. A really cool product, by the way...
(http://www.p3international.com/products/special/P4400/P4400-CE.html)

Thought I'd share a few data points:

• Older PC (~6 yrs old, single core Pentium 2.4GHz) is about 90W
• New PC (dual core 3GHz, RAID drives, dual 24" monitors) is about 160W and the monitors are ~30W each.
• Both PCs, UPS, Cable modem, Wifi Router, VoIP ATA, cordless phone, and computer sound system draw about 300W combined.

 

[jesse]

I have a Kill-a-Watt meter which allows realtime monitoring of power consumption. A really cool product, by the way...
(http://www.p3international.com/products/special/P4400/P4400-CE.html)

Thought I'd share a few data points:

• Older PC (~6 yrs old, single core Pentium 2.4GHz) is about 90W
• New PC (dual core 3GHz, RAID drives, dual 24" monitors) is about 160W and the monitors are ~30W each.
• Both PCs, UPS, Cable modem, Wifi Router, VoIP ATA, cordless phone, and computer sound system draw about 300W combined.

Was all that equipment being stress tested? The power consumption levels vary with load.

 

[Arbiter419]

I'd like to see what my rig draws sometime...Opteron 1354, 2x2GB of DDR2 800, 3 SATA drives, one IDE, and an 8800GT. The BFG 650w power supply I have in it is overkill, but it's a nice buffer should I ever decide to upgrade my 8800GT. 400w at load sound about right?

 

[RJM62]

I'd like to see what my rig draws sometime...Opteron 1354, 2x2GB of DDR2 800, 3 SATA drives, one IDE, and an 8800GT. The BFG 650w power supply I have in it is overkill, but it's a nice buffer should I ever decide to upgrade my 8800GT. 400w at load sound about right?

Who knows. When the ConEd guy came with his meter, I was amazed at how wrong I was in my estimates of power consumption -- and I know how to do the math. So much seems to depend on the efficiency of whatever power conversion mechanism is in place that it's hard to make even ballpark estimates.

Battery backups, for example, can draw a lot of power even when there's nothing plugged into them. I saved a lot of watts by unplugging unused battery backups, and replacing battery backups for devices that were non-essential with surge protectors or plain old power strips.

Chargers for phones, MP3 players, cordless tools, etc. also draw some power even when they're not connected to their devices. The draw is small for individual chargers, but when you consider how many of them we have, they can add up.

Halogen and fluorescent lights can draw considerably more than their rated wattages because the transformers are not 100 percent efficient. They also draw some current even when turned off, if the switch is on the downstream side of the transformer (as is the case with most halogen and some fluorescent desk lamps, where the charger plugs directly into the wall). Placing these devices on a switched power strip solves that problem.

One thing I will say is that the guy from the electric company knew his stuff when it came to ways to save electricity. Most utilities offer this service for free, especially if you call to complain that your bills are too high. I think it's a very valuable service.

-Rich

 

[nyoung]

I wouldn't say stress tested, but the machines were being used. Some web-surfing on the old PC, code development and a RAID rebuild (grumble) were active on the new PC. As the PCs were doing various tasks, the power fluctuated a bit, but not more than +/- 20%, and typically less than 5%... the numbers I posted were average readings.

I do not know how accurate the Kill-a-watt is for capturing transient power surges. The display updates pretty frequently, maybe once every second or two. If the power spiked a ton for only a fraction of a second, I don't know if it would catch that. That said, it does have an integration mode where it captures kw-hrs used since plugged in. In this mode, you can capture total usage and and divide by the hrs to get a good estimate of average power consumption.

Was all that equipment being stress tested? The power consumption levels vary with load.