Cost Justifying the Move to Clarkdale CPUs

Ecological issues around computing. This is an experimental forum.

Moderators: Ralf Hutter, Lawrence Lee

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Cost Justifying the Move to Clarkdale CPUs

Post by ces » Fri Apr 02, 2010 9:11 pm

Does anyone have this information in their head?

Clarkdales lower their energy usage at idle. Their motherboards use less energy.

If you are replacing an existing computer, how much energy costs can you save. How long does it take to pay for the Clarkdale setup with energy savings?

Does it make a difference what motherboard you use? Which Clarkdale chip you use?

What about replacing an Intel Q9550 with an overclocked Clarkdale? This is not that crazy. An overclocked Clarkdale still downshifts into very low energy saving speeds when idle (which is almost all the time).

If overclocked with a supercooler, it can probably hit about 4.5Ghz, which makes it about equivalent to a non-overclocked Intel Q9550. And the Q9550 doesn't downshift when idle.

What do you think. What can you say?

swivelguy2
Posts: 404
Joined: Sun Jul 04, 2004 9:18 pm
Location: Illinois, USA

Post by swivelguy2 » Sat Apr 03, 2010 5:59 am

I'm not gonna run the actual numbers, but the payoff times for this sort of thing, when related to computers, are generally very long, unless you get most of the cost back by selling the old components. By very long, I mean 5+ years, and possibly never at all if you apply discounting.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 03, 2010 12:08 pm

swivelguy2 wrote:I'm not gonna run the actual numbers, but the payoff times for this sort of thing, when related to computers, are generally very long, unless you get most of the cost back by selling the old components. By very long, I mean 5+ years, and possibly never at all if you apply discounting.
Approximately say you go from 120 watts to 50 watts (and maybe a lot less when the clarkdale is idling) How much savings per year do you think?

$10, $50 $100?

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Sat Apr 03, 2010 2:02 pm

As for me, I don't see the point in going for a Clarkdale sitting on the Wolfdale. It won't be faster (a bit -- doesn't count), it will need a lot of cash to change the whole system (including RAM type). My current system is 22W DC idle rated, and I don't know if a Clarkdale setup will be better in this mean.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 03, 2010 2:47 pm

Ksanderash wrote:As for me, I don't see the point in going for a Clarkdale sitting on the Wolfdale. It won't be faster (a bit -- doesn't count), it will need a lot of cash to change the whole system (including RAM type). My current system is 22W DC idle rated, and I don't know if a Clarkdale setup will be better in this mean.
That is one option I am considering. A g31 chip set, probably running a downvolted Q8400. The downside is that everything is moving away from the 775 socket. All things equal, in the long term it is easier not to struggle against the direction of change.

Even with the platform you are talking about, how do you cost justify it. If you are comparing an E3300 with a G31 chipset against a standard Q9550 with say a 5450 Radeon.

To what extent does the savings pay for the cost of moving to the greener system. What is the delta in electrical usage. How much do you save over 3 years?

Does it at least pay for the cost of replacing the older system with the greener system. It assuredly does, but is the payback in 12 months or 12 years? Or most likely, somewhere in between. That is a not unimportant question.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Sat Apr 03, 2010 3:22 pm

ces wrote: That is one option I am considering. A g31 chip set, probably running a downvolted Q8400. The downside is that everything is moving away from the 775 socket. All things equal, in the long term it is easier not to struggle against the direction of change.

Even with the platform you are talking about, how do you cost justify it. If you are comparing an E3300 with a G31 chipset against a standard Q9550 with say a 5450 Radeon.

To what extent does the savings pay for the cost of moving to the greener system. What is the delta in electrical usage. How much do you save over 3 years?

Does it at least pay for the cost of replacing the older system with the greener system. It assuredly does, but is the payback in 12 months or 12 years? Or most likely, somewhere in between. That is a not unimportant question.
There will not be to much saving from a G31 + Qxxxx vs H55 + core i3 or i5 less than 10W for sure and if you want the same performance then there will be no difference.

But I will calculate for you as an example what a 10W difference will save you on electric bill.
I will assume you use the computer 12h a day 7 day a week for one year.
So you have 10W x 12h x 365days = 43800wh this is 43.8kWh at 10 cent /kWh is 4.38$ so less than 5$ :) I do not think you can make any upgrade for 5 or 10 $
I live in Canada and the price here is about 10 cent/kWh but is not completely true because if you use as I do about 300kWh/month the I pay about 60$ for this since there is a fix monthly surcharge + some municipal surcharge and some taxes and in the end I pay 20 cent /kWh

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Sat Apr 03, 2010 6:20 pm

Quite frankly, I think you'll be looking at longer than 12 years payback unless you sell your current system. Even longer if you don't leave your PC on 24/7. If your goal is to save money, then it's not worth it, imho. If it's to help the environment, then consider that majority is spent during the manufacturing process so that's even more reason to keep your Q9550.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Mon Apr 05, 2010 6:09 am

ces wrote:To what extent does the savings pay for the cost of moving to the greener system.
Dear ces, it won't be that much greener :? That's a illusion force-feeded by vendors as they must sell their goods by any means. Green is another pseudo-goodish consume reassuring trend. You buy a new hardware that is <5% greener than previous and the factories consume energy and material resources to give you that. And you have a pain in the ass how to get a maximum late hitech toy for a minimum bucks, how to run it perfect as your previous one was, and where to throw outdated stuff for some pay back money.
What is the delta in electrical usage. How much do you save over 3 years?
A saving? I'm just wasting so far! ;)

Well, I'm not using my compie too much. Maybe 5 hours a day. And not every day I swith it on.

electrodacus
Wow, I got 1kWh for 1USD! :) But still don't think I should upgrade to Clarkdale :P

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Mon Apr 05, 2010 7:42 am

electrodacus wrote:But I will calculate for you as an example what a 10W difference will save you on electric bill. I will assume you use the computer 12h a day 7 day a week for one year. So you have 10W x 12h x 365days = 43800wh this is 43.8kWh at 10 cent /kWh is 4.38$ so less than 5$ :)

I do not think you can make any upgrade for 5 or 10 $ I live in Canada and the price here is about 10 cent/kWh but is not completely true because if you use as I do about 300kWh/month the I pay about 60$ for this since there is a fix monthly surcharge + some municipal surcharge and some taxes and in the end I pay 20 cent /kWh
1. Let's assume 10 cents per kWh. You were going to have to pay those fixed costs no matter what (until you go to wind and sun - and even then you have to pay another form of fixed costs).

2. Saving 10 watts will produce a savings of 87600 watt hours if you calculate it at 10W x 24 hrs x 365 days (electrodacus why did you use 12 hours in your calculation?)

3. So Saving 10 Watts = Saving 87.6 kWh per year
At 10 cent /kWh this produces a savings of:
8.76 over one year
$26 over three years
$44 over five years
$88 over 10 years
(not calculating in the time value of money which reduces the savings)

Other Factors:
(a) Manufacturing Costs. As Ksanderash points out, if the purchase of a new system or new components results in non-use of the old system or the manufacture of a new system that otherwise would not be manufactured, you actually are increasing total energy usage in the world by the amount of energy it costs to manufacture the new system.
(b) Capital Cost of Producing the Extra Watts. If you plan to go off grid, generating your own energy by wind or solar, the reduction of the need for 10 watts, might save you a lot more money in reducing the capital cost of generating that 10 watts (and I guess your time and/or labor costs). I don't know what that costs. My guess would be maybe $50 or $100 for an additional 10 watts of peak air or sun generated energy - what with batteries and all.

Am I missing anything?

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Mon Apr 05, 2010 2:25 pm

No point to calculate the savings over ten years, because in five years or so you will replace the system anyway to be able to run some new unforeseen killer application.

Also you are forgetting the cost of your time needed to asssemble the new system, install OS, drivers, applications and updates and doing stability testing, risk of replacing some doa parts, migrating data and settings from the old system, wiping the old system and selling it.

For these reasons, you should only buy a new system when you actually need it to do things you couldn't do at all with the old system, with reasonable speed at least.

However, there might be some parts you could replace in your current system, that would both give you a noticeable performance improvement and lower your power usage.

For example, if you have many old small hard drives in your system, consolidating to a single large hard hard drive could make your machine cooler, quieter and more performant. If you do this, you should sell the old drives.

If your current system uses ddr3, then replacing only the motherboard and CPU would be an option, but even that only makes sense if you get a large performance boost in addition. Nothing you can buy at the moment is that much faster than q9550 that it would warrant an upgrade, especially as you don't mention that you were unhappy with your current performance in any way.

Your plans of 4.5 gigahertz especially don't sound like saving. The supercooler you mention needs power too and isn't cheap. Also you need higher voltage which means more power.

The q9550 is a good CPU with a low idle power. Do you really have 120w idle or is that just a guess? My signature system idles at about 110w and half of that is the gpu.

My advice is to forget this idea and upgrade only when your system won't run your stuff anymore. Even then, strategic single part upgrades should help you get far. For example an Intel ssd would make your system fly. If they seem expensive now, wait for the 25nm versions due later this year.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Mon Apr 05, 2010 5:32 pm

ces wrote:
1. Let's assume 10 cents per kWh. You were going to have to pay those fixed costs no matter what (until you go to wind and sun - and even then you have to pay another form of fixed costs).

2. Saving 10 watts will produce a savings of 87600 watt hours if you calculate it at 10W x 24 hrs x 365 days (electrodacus why did you use 12 hours in your calculation?)

3. So Saving 10 Watts = Saving 87.6 kWh per year
At 10 cent /kWh this produces a savings of:
8.76 over one year
$26 over three years
$44 over five years
$88 over 10 years
(not calculating in the time value of money which reduces the savings)

Other Factors:
(a) Manufacturing Costs. As Ksanderash points out, if the purchase of a new system or new components results in non-use of the old system or the manufacture of a new system that otherwise would not be manufactured, you actually are increasing total energy usage in the world by the amount of energy it costs to manufacture the new system.
(b) Capital Cost of Producing the Extra Watts. If you plan to go off grid, generating your own energy by wind or solar, the reduction of the need for 10 watts, might save you a lot more money in reducing the capital cost of generating that 10 watts (and I guess your time and/or labor costs). I don't know what that costs. My guess would be maybe $50 or $100 for an additional 10 watts of peak air or sun generated energy - what with batteries and all.

Am I missing anything?

Totally agree with lm.

1. Yes you need to pay the fixed cost but this are adding to your electric bill so I actually pay 20cent/kWh right now if I will use more than 300kWh/month the price/kWh will go down but not by much.
I will soon move off grid and I will use about 80% wind and 20% solar.

2.I used 12h/day since this is how much I use in average my computer and I did not understood that you want to use your computer 24h/day.

Other factors.
Electronic devices are quite efficient so if you want to save electricity you need to look at other appliance (illumination can be reduced 5x or more by using LED to replace the incandescent bulbs;washing machine and especially dryer;electric stove;air conditioning and heating can be reduced 10x by using better house insulation.)
I'm originally from Europe and I see huge waste of energy in North America. Out of my 300kWh/month 70kWh/month I pay for common illumination in may building is a two story 8 apartments and there are 8 x 100W incandescent light bulbs that are working 24h a day no daylight sensor at least for the one that are outside there are 2 on the front door one at the back door and the rest are inside on the stairs and hallway.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Mon Apr 05, 2010 9:08 pm

How many watts difference in energy usage is there between:
(a) a standard H55 1156 socket mother board with a Clarkdale chip, using the Clarkdale video capability
vs.
(b) that same motherboard using a Radeon 5450 video card

Bobert
Posts: 13
Joined: Sat Mar 20, 2010 5:13 pm
Location: NC

Post by Bobert » Fri Apr 09, 2010 7:15 pm

How many watts difference in energy usage is there between:
(a) a standard H55 1156 socket mother board with a Clarkdale chip, using the Clarkdale video capability
vs.
(b) that same motherboard using a Radeon 5450 video card
In round numbers probably 10w difference AC.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Fri Apr 09, 2010 8:01 pm

Bobert wrote:
How many watts difference in energy usage is there between:
(a) a standard H55 1156 socket mother board with a Clarkdale chip, using the Clarkdale video capability
vs.
(b) that same motherboard using a Radeon 5450 video card
In round numbers probably 10w difference AC.
That means then, adding the low power Radeon 5450 video card to a motherboard with integrated video generates an additional cost of approximately $10 per year.

Presumably one of those super dual video card setups might cost $100 extra per year to run. Maybe more.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Fri Apr 09, 2010 9:11 pm

ces wrote: Presumably one of those super dual video card setups might cost $100 extra per year to run. Maybe more.
No because they will not be able to run those computers 24h a day :)
... and if they are able to do that the 100$ a year will be worth :))

PS: not a gamer but I can understand.

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Fri Apr 09, 2010 9:35 pm

" the Q9550 doesn't downshift when idle"

I thought that the Q9550 down shifts from 2.83 GHz (8.5 x 333MHZ) and 1.25 volts to 2 GHZ (6 x 333 MHz) and 1.1 volts when at idle.

Generally you will get a bigger energy savings, and therefore noise reduction opportunity, by using IGP instead of an add-in video adapter.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 10, 2010 7:23 am

Wayne Redpath wrote:" the Q9550 doesn't downshift when idle"

I thought that the Q9550 down shifts from 2.83 GHz (8.5 x 333MHZ) and 1.25 volts to 2 GHZ (6 x 333 MHz) and 1.1 volts when at idle.
Can anyone say if this is so, or not?

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Sat Apr 10, 2010 10:21 am

ces wrote:
Wayne Redpath wrote:" the Q9550 doesn't downshift when idle"

I thought that the Q9550 down shifts from 2.83 GHz (8.5 x 333MHZ) and 1.25 volts to 2 GHZ (6 x 333 MHz) and 1.1 volts when at idle.
Can anyone say if this is so, or not?
I'm sure the Q9550 at idle is 2Ghz (6x333) I have the Q8400s and mine works when all setting are default but if I underclock and undervolt the speed step will be deactivated and this is normal I usually use mine at 2GHz (6x333) and 0.925V
You can see the spec for Q9550 here Link as you can see Q9550 uses Enhanced Intel® Speedstep Technology as probably all modern Intel CPU.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 10, 2010 11:29 am

electrodacus wrote:I'm sure the Q9550 at idle is 2Ghz (6x333) I have the Q8400s and mine works when all setting are default but if I underclock and undervolt the speed step will be deactivated and this is normal I usually use mine at 2GHz (6x333) and 0.925V
You can see the spec for Q9550 here Link as you can see Q9550 uses Enhanced Intel® Speedstep Technology as probably all modern Intel CPU.
Can you confirm that it also downvolts from 1.25 volts to 1.1 volts?

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Sat Apr 10, 2010 12:17 pm

ces wrote:Can you confirm that it also downvolts from 1.25 volts to 1.1 volts?
That's dependent on a specific sample. Not every Q9550 runs on the same voltage.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 10, 2010 12:22 pm

ilovejedd wrote:
ces wrote:Can you confirm that it also downvolts from 1.25 volts to 1.1 volts?
That's dependent on a specific sample. Not every Q9550 runs on the same voltage.
Well that's interesting. Do you know what the typically variance is, or an intelligent or educated guess at what a typical varience will likely be?

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Sat Apr 10, 2010 2:22 pm

ces wrote:Well that's interesting. Do you know what the typically variance is, or an intelligent or educated guess at what a typical varience will likely be?
Unless you've got a whole bunch of these things, you can't really say but Intel specifications state VID range of 0.85V – 1.3625V.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 10, 2010 2:43 pm

ilovejedd wrote:
ces wrote:Well that's interesting. Do you know what the typically variance is, or an intelligent or educated guess at what a typical varience will likely be?
Unless you've got a whole bunch of these things, you can't really say but Intel specifications state VID range of 0.85V – 1.3625V.
Wow. That must mean when I buy a low cost cpu from Microcenter, especially if they are selling under their house brand, I could be getting something that was left after getting picked over by others.

Do you know if that voltage varies from core to core? I have a CPU with a core that just plain runs too hot. Could that mean it is running on higher voltage than the other core?

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Sat Apr 10, 2010 10:37 pm

"Not every Q9550 runs on the same voltage."

During the manufacturing process the 'chips' are tested and then built for different markets. It goes something like this:

==>server market - high speed, low voltage, low heat, highest reliability

==>notebook market - various speeds, low voltage, low heat

==>desktop market - various speeds, middle voltages, middle heat

==>enthusiast market - various speeds, highest voltage, highest heat

The enthusiast market gets the poorest quality product.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sat Apr 10, 2010 11:13 pm

Wayne Redpath wrote:"Not every Q9550 runs on the same voltage."
During the manufacturing process the 'chips' are tested and then built for different markets. It goes something like this:
==>server market - high speed, low voltage, low heat, highest reliability
==>notebook market - various speeds, low voltage, low heat
==>desktop market - various speeds, middle voltages, middle heat
==>enthusiast market - various speeds, highest voltage, highest heat
The enthusiast market gets the poorest quality product.
How do you buy the good stuff, or should you even care?

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Sun Apr 11, 2010 8:20 am

"should you even care?"

Many do care and spcr is great for assisting us to make better informed decisions. Our situations are unique:
- geographic location
- bank account balance
- personal preferences and social influences (wife...)

Some battles we fight today and some tomorrow...maybe...

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Sun Apr 11, 2010 8:45 am

ces wrote:Wow. That must mean when I buy a low cost cpu from Microcenter, especially if they are selling under their house brand, I could be getting something that was left after getting picked over by others.
Doubtful, unless it's open-box. They buy their processors in bulk so I somehow doubt that they do any cherry picking on their end. As far as I'm aware, Intel doesn't list VID in the packaging anymore so it's hard to say whether a chip is particularly low-power or a heavy overclocker until you have it installed on your motherboard. Batch numbers aren't as magical as what oc'ing forums might suggest.

On the Intel side, the "cherry-picked" chips get sold under the Xeon brand where they're able to charge businesses much, much more. :P

Most consumers wouldn't be particularly concerned if their processor is consuming a few watts more than another. However, when you've got a room full of rackmount computers or blades, those few watts matter, not because of the power consumption of the computers themselves, but rather lower cooling costs.

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Sun Apr 11, 2010 8:54 am

Wayne Redpath wrote:==>enthusiast market - various speeds, highest voltage, highest heat

The enthusiast market gets the poorest quality product.
Not really as bad as that. I'd say the enthusiast market gets highest speed at the expense of heat and voltage.

Wayne Redpath
Posts: 104
Joined: Wed Jul 29, 2009 7:42 am
Location: Calgary, Alberta, Canada

Post by Wayne Redpath » Sun Apr 11, 2010 9:26 am

"a room full of rackmount computers or blades...lower cooling costs"

I've heard that the general 'rule-of-thumb' is that every 1 KW of heat generated incurs 4 KW of air cooling cost. This is not a problem in a residence in Calgary, Alberta, Canada. However in in Miami, Florida, USA...

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Sun Apr 11, 2010 12:10 pm

electrodacus wrote:But I will calculate for you as an example what a 10W difference will save you on electric bill. I will assume you use the computer 12h a day 7 day a week for one year. So you have 10W x 12h x 365days = 43800wh this is 43.8kWh at 10 cent /kWh is 4.38$ so less than 5$ :)

I do not think you can make any upgrade for 5 or 10 $ I live in Canada and the price here is about 10 cent/kWh but is not completely true because if you use as I do about 300kWh/month the I pay about 60$ for this since there is a fix monthly surcharge + some municipal surcharge and some taxes and in the end I pay 20 cent /kWh
1. Let's assume 10 cents per kWh. You were going to have to pay those fixed costs no matter what (until you go to wind and sun - and even then you have to pay another form of fixed costs).

2. Saving 10 watts will produce a savings of 87600 watt hours if you calculate it at 10W x 24 hrs x 365 days (electrodacus why did you use 12 hours in your calculation?)

3. So Saving 10 Watts = Saving 87.6 kWh per year
At 10 cent /kWh this produces a savings of:
8.76 over one year
$26 over three years
$44 over five years
$88 over 10 years
(not calculating in the time value of money which reduces the savings)

Other Factors:
(a) Manufacturing Costs. As Ksanderash points out, if the purchase of a new system or new components results in non-use of the old system or the manufacture of a new system that otherwise would not be manufactured, you actually are increasing total energy usage in the world by the amount of energy it costs to manufacture the new system.
(b) Capital Cost of Producing the Extra Watts. If you plan to go off grid, generating your own energy by wind or solar, the reduction of the need for 10 watts, might save you a lot more money in reducing the capital cost of generating that 10 watts (and I guess your time and/or labor costs). I don't know what that costs. My guess would be maybe $50 or $100 for an additional 10 watts of peak air or sun generated energy - what with batteries and all.

One major correction to the above

So the one thing I am missing is that 1kw of energy generates 4kw of air conditioning load, if you are in a climate where you are running air conditioning you need to multiply the above numbers. Let's say you are running air conditioning for 6 months of the year you need to multiply the above numbers by 2. And you get:

Saving 10 Watts = Saving 87.6 kWh per year x 2
At 10 cent /kWh this produces a savings of:
$18 over one year
$52 over three years
$88 over five years
$176 over 10 years
(not calculating in the time value of money which reduces the savings)

Post Reply