Power efficient builds for gaming and its importance

Tags:
Last response: in Systems
Share
October 29, 2010 2:51:43 PM

I seem to get scoffed at when I mention how important this is. So I thought I'd do some math and get some actual figures on paper, then electronic papel to show everyone my points have validity.

Before I get into the real math and figures, I will say that in a year's time, the differences in costs are pretty substantial. Substantial like the fact that we critically argue how crucial initial costs of CPUs and GPUs per their performance marks, but when power efficiency comes into play we seem to forget operational cost per performance. If the video card you just bought for \$30 cheaper performs even on par with another, but is significantly inefficient with its watt usage and heat generation in comparison and you'll spend that \$30 you saved over the next 6 months - then more in the following months - wouldn't that be something you'd consider when originally purchasing it? Don't think energy costs that much? I dare you to continue reading.

Based off of reading an energy costs article from Anandtech from late 2008, I figured the math for a flat rate kWh consumption (http://www.anandtech.com/show/2668/2). Now, I will remind you that running a computer not only takes power, but also creates additional heat, which increases the demand on your A/C. I will not be exploring that aspect of cost factoring in this post, but I will provide you with a difference in my energy bill usage and costs for the last 4 months. This will provide you a good real-world example of what more heat and energy usage can do to my actual energy bill.

I live in Georgia, USA for those that are curious of rates. For those that don't know, the climate here is usually hot and humid (temperatures easily reaching 100+ with 90% humidity in summer months); a heavy demand on any A/C unit. Here, the pricing is tiered for the kWh you use and if it's during summer or winter months. For example: the first 650 kWh is charged at a rate of \$0.045991, the next 350 kWh is at the rate of \$0.03946, and if you go beyond 1000 kWh the rate is \$0.038737 for winter. Summer months the rates increase as you reach the next tier - up to \$0.078765. http://www.georgiapower.com/pricing/pdf/2.10_R-16.pdf Below are the costs exactly as it's coming out of my pocket. I was going to take out the fees and taxes, but then decided it wouldn't matter since every energy company charges this, and fees & taxes of the actual services provided is inherent to business.

June 2010 - 1,438 kWh - \$176.48
July 2010 - 1,004 kWh - \$126.58
August 2010 - 834 kWh - \$102.70
September 2010 - 722 kWh - \$80.44

Note that all four bills are in the summer charge rates. The only variances here are the heat outside and the fact that I got my brother to turn off his computer when he wasn't using it sometime in early-mid August (only reflecting in Septembers bill at this point). I'm certain my next month's bill will be even lower since it's cooler and my brother is no longer living with me. I always turn my computer off when I'm not using it. It's on for about 8 hours a day; 4-6 hours at full load (gaming) and 2-4 hours idle'ish.

Due to the way energy exchange works, 100% efficient PSUs can't possibly exist, and 82% is an industry standard. For availability, an example would be taken by looking at what Newegg has. Out of a total of 250 listed PSUs, 189 are rated below 80+, at 80+ or Bronze certified (82%+). Silver certified is where 85%+ is necessary. I know most custom builders probably use PSUs that are 500-800W and if building for initial cost effectiveness, I doubt the majority will spend more than \$75 on a PSU. That being said, there are only a handful of PSUs listed at or under \$75 that are Bronze certified, much less 80+ certified. So if you spent even less on an energy inefficient PSU to save your initial cost, chances are you can basically take all the numbers I'm providing here and multiply them by the % difference. Just like living in an area where energy costs are doubled (such as CA), this can make a HUGE difference.

Now onto the figures based on Anadtech's article. I based my initial figures off the charge rate of \$0.07/kWh from NC (which happens to be pretty close to what GA charges as well). To be fair to PSU efficiency and actual watt usage, an 82% energy efficient standard PSU watt amount is provided in paranthesis. The final costs are based on these amounts. You can modify the final amounts by difference of energy cost (\$0.07/kWh = 100%) or percentages away from a Bronze certified PSU (82%+ = 100%; scaled opposite).

Example #1 - Inefficient gaming rig
-at idle, will consume 310 watts (378)
-under full load, will consume 550 watts (671)

#1-This computer ran for 10 hours a day (8 hours full load, 2 hours at idle): \$167.65/year; \$13.97/month
#2-This computer ran for 24 hours a day (8 hours full load, 16 hours at idle): \$312.57/year; \$26.05/month

Example #2 - Efficient gaming rig
-at idle, will consume 160 watts (195)
-under full load, will consume 350 watts (427)

#3-This computer ran for 10 hours a day (8 hours full load, 2 hours at idle): \$104.19/year; \$8.68/month
#4-This computer ran for 24 hours a day (8 hours full load, 16 hours at idle): \$178.93/year; \$14.91/month

In both examples, you can see that by simply leaving your computer on increases costs dramatically. These costs do not reflect your monitor, only your case and what's inside.

As an extreme Example #3 for California costs and a less efficient PSU (75% instead of 82%) (example numbers used from above):

#1-\$366.48/year; \$30.54/month
#2-\$683.28/year; \$56.93/month
#3-\$227.76/year; \$18.99/month
#4-\$391.14/year; \$32.59/month

So what are the actual differences between an extremely inefficient computer and one that is more efficient as well as someone not utilizing simple efficient power practices? Humungous! A difference of \$34/month. This does not include the additional heat generated either. So operational costs can easily be \$40/month between extremes in California, and \$20/month between extremes in the southeast (NC and GA at least). I'll let you figure out the exact amounts, but in the 2-4 years you own that component of your computer, you could saved enough money in operational costs to upgrade again -or- purchase a higher quality product in 1/3rd to 1/8th the time.

Fortunately, not only for our wallets, but for Mother Earth as well, both AMD and Nvidia seem to have graciously adopted the importance of energy efficiency. The new 6800 series seems to only improve on the 5800's efficiency, and the GTX460 is a huge improvement over the 470 and 480. As dies get smaller, CPUs become more clustered and lower in frequency - less heat and watts used overall. Things are looking much better. However, ultimately, the choice is still yours.

Enlightened, you shall be.
a b B Homebuilt system
October 29, 2010 3:19:29 PM

Good read. Maybe you can do a study on putting a computer to sleep as opposed to completely shut down as well.
October 29, 2010 3:25:19 PM

Ouch, math makes my head hurt this early in the morning!! :-p

Well....non-GHZ math!!!

But it is very good info. One other area you might want to look into is the LCD and LED monitors. Some of those monitors put off a LOT of heat - and to check on how much energy and heat they put off when they are sleeping/stand-by.
Related resources
October 30, 2010 10:57:13 AM

kureme said:
Good read. Maybe you can do a study on putting a computer to sleep as opposed to completely shut down as well.

I'd love to get some actual readings from the wall, but alas, I'm a humble personal builder - not a professional or journalist with equipment. *cough* /wave Tom's *cough*

Actually, I just checked how much watt monitors ran for and to my surprise not that much at all. I'll have to look into getting one.
a b B Homebuilt system
October 30, 2010 1:48:22 PM

Wow, what does you inefficient gaming rig include?
I have never build a system with over 300 watts idle, Hell most of my systems do not break 300 at load.

I have listed several 80+ examples in the past as well. Low noise was always a nice side effect....

My game rig is off when its not gaming the rest of my computing is done mostly on my Media center(most of my less demanding games are played on there now).

high 30's low 40's at dead idle(with nothing running Intel can drop power consumption quite a bit). Surfing the net with music, 46-54ish

H55N USB3
I5 750(undervolted)
5770 @ stock
4 gigs of memory
1tb hard drive(thinking of adding one more[mod my case a bit]. media takes a lot of space)
SG05 case with 300 watt power supply.
a b B Homebuilt system
October 30, 2010 2:00:38 PM

kureme said:
Good read. Maybe you can do a study on putting a computer to sleep as opposed to completely shut down as well.

When I get home later today, I will try to get you those tests(on a few computers). They will just be with a basic power meter(http://tinyurl.com/yz3pvrf), but they should give you an idea.
October 30, 2010 4:06:44 PM

nukemaster said:
Wow, what does you inefficient gaming rig include?
I have never build a system with over 300 watts idle, Hell most of my systems do not break 300 at load.

Those sample rigs are taken from the article and I'm speculating are generalizations, not actual builds.

However, I can find GPUs that can almost suit those amounts alone (not to mention if they're overclocked) (http://www.tomshardware.com/reviews/geforce-gtx-480,2585-15.html]; http://www.tomshardware.com/reviews/radeon-hd-5970,2474-15.html):

Idle= 114-211 / Loaded= 280-620
a b B Homebuilt system
October 31, 2010 5:56:22 PM

@ kureme. Only had a chance to test on 2 systems, but gives you an idea.

I7 system(I7 920 undervolted @ 3.5,X58A-UD5,4 sticks of memory(normally has 6),5 hard drives, 5870, Sound Blaster Audigy2 zs, Corsair 850 watt psu)
Hibernate/Off - 10
Sleep - 12-13
Idle - 141-143
Winamp 144
Starcraft2 - 280(MAX)
Prime - 238
Just Cause 2 - 302(MAX)

Media Center(I5 750 undervolted,H55N USB3,2 sticks of memory,1 hard drive,5770,300watt SFX fsp psu)
Hibernate/Off - 4
Sleep - 4
Idle - 52
Winamp - 59(since the cpu can not enter ultra low power states, those are off on the I7 system so they do not make a difference)
Starcraft 2 - 157(MAX)
Just Cause 2 - 165.

Generally sleep works well(only a few watts more then off). The big difference from off on both systems is most likely due to the board/psu differences. The X58 board has more crap(including more usb ports with power on when the system is off or a sleep). Some power supplies have an effect on this "off" power as well.
a b B Homebuilt system
November 1, 2010 12:23:41 PM

Thanks for the numbers nukemaster. It really doesn't seem to be much significant difference between putting the system in in sleep mode as opposed to completely shutting down, maybe only a few cents a month.

This also brings to mind how much power you're still using even when shutting down the PC.
a b B Homebuilt system
November 1, 2010 12:42:13 PM

On some boards in the bios, there is an option called EUP support or similar. This lowers power off power(if i remember it was claimed at less then 1 wall) by not leaving power on usb and other ports. If I get a chance, I will test that too.
November 1, 2010 6:04:07 PM

Man....

I wish I was only paying \$0.08 per KWH.

I don't know how much watts my primary rig (see sig) and HTPC draw since I don't have a Kill-A-Watt.
November 5, 2010 8:33:11 AM

Nice testing, nukemaster.

One thing that I noticed while reading Hardwarecanuck's reviews, is that voltage changes will skew watt readings from a wall-tester. That's something to consider as well.

The negligible difference between hibernate and off is outstanding.

Quote:
I don't know how much watts my primary rig (see sig) and HTPC draw since I don't have a Kill-A-Watt.

I wouldn't worry too much. The higher demanding parts of the computer are older Athlons and the Fermi GTX 270 & 280 specifically. There are plenty of reviews out there for hardware within the last 4-5 years showing examples.
a b B Homebuilt system
November 5, 2010 11:30:44 AM

With my meter actually showing voltage as well, it may be smart enough to take that into consideration.

It does not line up with my UPS, but as long as I measure all devices with one or the other it should be fairly good.

Hibernate and off should always be the same load since hibernate saves all memory to the hard drive and shuts the power off. In hibernate you can even unplug the computer and it will be fine on next boot. I have used this on laptops for years now since it does not use any power.

Under Windows Vista/ 7 there is Hybrid Sleep(it is the default sleep setting). It combines both, so the computer saves memory to hard drive and uses normal suspend to ram together. The result is faster wake up from sleep while still being able to cope with a power out while the system is sleeping.
November 6, 2010 5:39:34 AM

Quote:
With my meter actually showing voltage as well, it may be smart enough to take that into consideration.

Maybe. I'm sure you can do a little research and find out if the model you're using does.

Sounds like hibernate has come a long way. I always stayed away from it since it reduced performance so much upon wake. Doesn't really affect me though - I'll just continue to turn my computer off when I'm not using it.

Either way, good info though. Thanks nuke.

Quote:
...Fermi GTX 270 & 280 specifically.

I actually meant 470 and 480. My bad.
March 16, 2011 1:03:22 PM

Reviews always do a great job at convincing us that card A is more powerful than card B, but isn't as practical for use. Then card C is quieter and costs less. So that leaves a huge grey area. Inquiring minds like to see less grey and more black and white. As a follow-up to this forum of the importance of having efficient hardware, I've taken actual numbers from reviews and determined, to some extent, how efficient the most popular GPUs are on the market today.

Disclaimers::
• This table, albeit missing some data that wasn't provided, will clearly show how each card stacks up in terms of initial cost/fps and total operational costs/fps. I've included using some CF and SLI setups as well. My next step will be to get averages across several games or benchmarks.
• Feel free to verify, however, before you say "but this isn't what you put" remember that prices change and FPS results change depending on the exact testing platform.
• Also, take note that I've based these numbers off of a PSU that is 100% efficient. So typically if you want more accurate "real world" numbers you'll need to scale the costs up to accomodate that correction (see first post if you're not sure what I'm talking about).