Check with your local library or electricity company. The often have Kil-a-watts or other power meters available for loan.

I did another round of power testing on my machine -- here is some info that others may find useful. All numbers are with system idel:

1. 108W: No tuners in system.
2. 117W: Hauppauge PVR-150 tuner in system
3. 122W: PVR-150 + kworld-115 in system
4. 128W: PVR-150 + Kworld-115 + HD-5000 in system.

So the 3 tuners are consuming a total of 20 watts:
* 9 watts for the analog tuner
* 5-6 watts for each digital tuner.

Some other factoids:
* 1 watt: Power savings by dropping memory from 2GB (2x1GB) to 1GB (1x1GB) DDR-400 (PC3200) memory
* 7 watts: Additional power used to record 3 shows simultaneously.
* 9 watts: Power saved by using XVMC for playback of hidef recording.

A friend of mine built out a file server based on:
* AMD 45watt processor
* micro-atx board with integrated graphics
* 2 disk drives
* Antec NSK1380 box with 80% efficient power supply.

His system idles at 67 watts.

What this means to me is that if I rebuild my system using the same components he has in his file server, which probably represent "as good as it gets" on power consumption, my total power draw will go from 128 watts idle to 87 watts idle, for a saving of 41 watts. This will reduce my power bill by $118/year. I am less concerned about power consumption under load, as "load" mostly comes from watching TV, and we don't watch that much TV.

All of this brings me to the conclusion that the real "win" is to find a way to use "s3" sleep mode, which would enable me to throw the system into a state where it draws ~5-watts for an average of 12 hours/day. I wonder if anyone has ever gotten "s3" sleep mode to work with Knoppmyth...

Marc

Marc

I was recently thinking the same thing about s3 since for a long time I have wanted to use mythwelcome. My basic issue is I wanted to be sure it works correctly hardware wise. This post is old but it's informative


http://www.mythtv.org/pipermail/mythtv-users/2006-June/139210.html

for reference

my frontend

Gigabit GA-K8N51PVMT-9 (M-Atx)
Intel Gigabit PCI Desktop Nic
2GB memory (two sticks of Patriot DDR-400)
Seagate 750gb (Sata)
Nvidia 6150 (Integrated Graphics with Svideo)
Athlon 64 3700+ (Socket 939)
OCZ 400W Power Supply

64W Idle with Kill-A-Watt

Hi,



I can really recommend using mythwelcome for energy saving. I've been using it for almost a year now (actually from day 1 of using Knoppmyth) and it works almost perfectly:
1) I do not use S3. Instead, the machine shuts down completely.
2) Starting up takes a moment, but does not bother us (WAF is excellent)
3) Every 36 reboots or so the system performs an fsck over /myth volume, which leads to missed beginnings of recordings.
4) My BIOS cannot schedule a wake-up over more than 24 hours. Solved by scheduling one recording every day ("Sandmännchen" . I live in Germany), which I would record anyway.

The only *real* issue is number 3. Often, the fsck happens when we want to watch something, but occassionally (today, for example...) it bites.

Apart from that, this is a perfect setup (for me). The machine is off most of the day, powers up to record something, shuts down again and uses very little power.

My wife doesn't mind it taking some time to boot when we want to watch something. Neither do the kids or I.

I tried getting S3 to work for some time, but since shutting down worked like a charm, I never put enough energy into it. Maybe R6 will make using S3 easier.

jens

After doing some research, I've decided to go the shutdown / restart route. It looks like there are a lot of issues with S3-suspend. I have a "fussy" BIOS, but I've gotten it working. The one thing that seems strange is that the restart time must be expressed in terms of GMT, not my local time zone. Perhaps this is normal -- I don't know for sure.

Question -- does the "$time" variable in the "Set wakeup time" field return in local time or GMT? I'm trying to figure out if I will need to convert it. Thanks!

Marc

You may also want to use tune2fs to avoid the fsck problem mentioned above. Of course that may also mean doing a scheduled wake up once a week or so when nothing else is happening specifically to do the fsck.

The default for Linux is to use the BIOS to reflect GMT time and have the operating system apply the time zone shift for local time. Windows, on the other hand, uses the BIOS to reflect local time. Linux can be configured to do the same, but it is not the default configuration.

I sure would be nice if the setup let you choose which behaviour you wanted.

I know it is sacrilege to suggest that some of us dual-boot. It has bit me on more than one occasion. Boot to Windows -- fix the time. Boot to Knoppmyth -- fix the time. Boot to Windows -- fix the time. Boot to Knoppmyth -- miss a recording because I forgot to fix the time.

I know that setting the hardware clock to local time is just not the Linux way. But sometimes we don't live in a Linux all the time world.

Maybe I should add this to the R6 Feature Request thread.

I was playing with my "Tom's Hardware CPU Comparison" link from earlier in the thread.

This time I added price points (mentally - it's not in the charts).

Marc.aronson was originally thinking of an Intel E4500 -- price point about $125. Throwing that same number of dollars at an AMD processor (actually a little less because there wasn't an exact match), yielded faster encoding, and better HD playback results -- at least on those charts.

And the Intel processor was 65W vs the 45W AMD. Throw in the usual higher price point of boards for Intel processors and it seems like AMD might be the winner in the price/performance/power consumption race.

I know many a flame war has been fought difference between the two companies. I'm just saying that if I had the money to go shopping . . .

Before someone comments on the pricing, I used New Egg's April 21, 2008 pricing for this comparison.

Looking at the "Building a 'Greener' PC" thread, I see a note about the Black Edition processors that had slipped my mind.

While their unlocked nature does allow for Overclocking, it will also allow Underclocking, running things cooler, quieter, and with less power. Of course that would toss the CPU benchmark scores, used for comparison, out the window.

Also, because the Black Editions are targeted at the Overclocking, enthusiast, less price sensitive market, they tend to have a premium price.

fscking on reboot...

You could something tricky with the /fastboot and /forcefsk files/flags.

I am not sure how you know the next boot is going to be a fsck boot, but you could set it to /forcefsck every week at a known time and then go to sleep after that boot.

--or--
schedule a shutdown -rF now once a week based on idle or something?

_________________
--
stinga

as of 17-May-2018
R8.2 - sort of working.
MB: gigabyte GA-P43-ES3G | RAM: 2GB
VGA: PNY Nvidia GT240 1GB
4 x Technisat skystar 2 dvb | 1 x TBS6981 dvb-s2
pata: 1x300Gb | sata: 2x1Tb 2x2Tb
Problems:
TBS6981 Does not work.

Chapman1, I am looking to optimize both the cost of power AND the cost of acquiring the parts. Which 45W AMD processor did you see in the charts that is faster than the E4500? The only AMD processors I saw that were faster were the 65W AMD processors -- I must be missing something...

Thanks for the help!

Marc

Oops. Guess I didn't have my eye on the ball.

I was paying so much attention to price, I missed seeing that I had crossed the 45W/65W boarder. In fact I was looking so hard at the price column that I passed up lower priced, higher processing power chips. It was the performance of the 5200 I was looking at, though I now see that 5400's were priced less.

As Emily Litella, an old Gilda Radner character on "Saturday Night Live" would say, when confronted with facts that completely negated her editorial diatribe, "Never Mind!" (Said in a high squeaky voice, with that sheepish "opps I goofed" tone)

Looks like the top end of the 45W line for the moment is the 4050E. But still, that's only a $70 processor. The penalty over the Core 2 E4300 (extrapolated from the Tom's charts) is about 7 minutes more for h.264 encoding, 14 for DivX and 16 for Xvid, for a one hour video.

There are the 4450e and 4850e coming -- both 45W processors (reference - MSI CPU support chart for K9NGM3 - Athlon 64(Brisbane, FSB200, L2 Cache 1M) 4050, 4450 and 4850 listed as "Under Testing") I'd bet someone better informed than I, could tell you when they might be seen in the wild.

Might be worth waiting a while to see what shakes out in the two big IGP chipsets coming on stream. AnandTech has an early board based on the GF8200. In their article "IGP Power Consumption - 780G, GF8200, and G35 shows a slight lead for 8200 - about $12 a year in electricity, but a significant lead by both over Intel's G35.

But of course we don't yet know prices of 8200 based boards. And then there's drivers.

If AMD/ATI ever get their act together on Linux drivers, and gets us access to hardware decoding via the "Unified Video Decoder," it will be big player on the HTPC market. I don't care whether it is the closed source "fglrx," the company sponsored open source "RadeonHD," or the independent open source "Radeon" driver.

And who knows how long the 8200 Linux drivers will take. Though Nvidia drivers do appear more quickly than others.

Personally, I'm half tempted to go the "available now" 780G route, with a discrete Nvidia graphics card, while hoping and praying for IGP Linux drivers. I might even consider booting to that MicroFlacid thing for HD playback with the IGP, as the Windows drivers already take advantage of the "Unified Video Decoder" for MPEG-2, VC-1, and H.264 codecs. If all else fails, it could become my desktop machine.

All in all, exciting times to come in the next few months. While it's good to think green, might be worth burning a few extra watts by sanding pat, while we wait for things to shake out.

And hopefully the second part of this post doesn't contain another "Gilda" moment.

ChapmanI, thank you for the great write-up and the link to the article on power consumption!!! If I understood it properly, getting a motherboard with the ge8200 integrated GPU and the AMD 4850E CPU, the power draw when idle is only 50 watts. That is amazing. I suspect that the 4450E, while less powerful, probably has the same great low power consumption.

From what you wrote it sounds like the gefore8200 drivers aren't out yet. Do you now if other integrated geforce GPU's are equally power efficient?

I agree with your point about not making any moves at this point, but instead, waiting to see how things shake out. I've implemented the automatic shutdown stuff and my estimate is that the machine will be off for 12 hours a day, so I'm already cutting my power draw by 50%. A subsequent re-build to a system based on the above configuration should save another 50-60 watts per hour while active.

So to sum it up:

1. 3,240W - Power consumed/day by current system operating 24x7
2. 1,680W - Power consumed/day by current sys with shutdown mode
3. 990W - Estimated power/day for system described in article w/ shutdown mode.

Using shutdown mode, I am saving $180/year in power. When I move to the more power-efficient system, I will save an additional $79/year in power...

Marc

Not only are the 8200 drivers not out yet, but the boads containing that IGP are not even for sale. They're just starting to trickle onto the market. Biostar has a model, with the Windows drivers dated April 18, 2008, same with Asus. New-Egg lists the Elite-Group version, but has no stock. So that said, both the boards, and especially linux drivers are probably months away. I'm not savy enough in this area to know if the April 11 Nvidia drivers support it, though they do support higher end GPUs.

The 780G based boards are available now from several different manufacturers. The power consumption difference between it and the 8200 are not very great, as shown in the AnandTech article. Depending on their Windows Vista power profile (Balanced vs Performance) there is a 5-7 watts difference at idle, 8 watts during Mpeg2 playback, 7 watts during h.264 playback, and zero difference during game play. At $0.13/kWh, on 24x7 works out the less than $9 a year difference. Basically negligible, unless you're being extra ultra green, and gonna squeeze every last kWh -- but then you wouldn't be running a PVR, would you.

There are Linux drivers for the 780G from all three streams (fglrx, RadeonHD, and Radeon). None of them as of yet use the hardware decoding for Mpeg2, h.264, or VC-1 codecs. The Phoronix.com discussions are a little too deep geek for me. They've done gaming (Doom 3, Quake 4, Enemy Territory: Quake Wars), graphics (GtkPerf), LAME encoding, timed disk reads (hdparm), Gzip compression, and RAMspeed benchmarks. They tested the ECS version and declaired:
I can't find much about video playback under Linux on on the 780G yet. People seem to be able to play DVDs, but I'm not sure how well. It's definitely throwing more processor at it, due to the inability to offload the task to the IGP. But then again, the CPUs we've been discussing are more then up to the task.

It looks like AMD released all the secret sauce needed to make drivers on March 19. Phoronix has a premonition that the May proprietary drivers may be a big leap forward.

Can you tell I REALLY want a board with this chipset? If I can unload the GV-NX88T512HP Video Card I won from Gigabyte, I'm going to grab one. It seems to hold so much promise.

But the 8200 looks like it has a very similar feature set. And with the history of better Nvidia Linux support -- well there's a TON of promise there too. And for all I know, it might work right out of the box under Linux. Though there does seem to be some Windows driver teething problems.

My excitement about both of them is so palpable, I'm hopping up and down like I have to pee, with no washroom nearby. I can't wait to see "head to head" comparisons between the two chipsets.

Things I've read (and God, who can remember where) implied that the Intel equivalent is a long way off -- maybe more than an year. It could be more efficient because Intel can produce chips with a smaller process (the nm spec you always see quoted). But will they make it with the smallest process? And until someone actually has them, we can't even really guess what their power consumption would be like. If they're paying attention, it'll be in this same range.

Doing all this research, has be primed to go shopping!
[/Whimper on] I want some new toys so badly I can taste it! [/Whimper off]
(Note to self: Must refrain from licking processors and motherboards!) I just wish the bank account would let me pull the trigger.