I'm glad I found this forum, you guys really seem to know what you're talking about when it comes to low power systems.
Currently, my Windows home Server consists of a Celeron 430 + Foxconn G31MV-K + 640GB WD Blue + 1TB WD Green + 380W Antec Earthwatts 80plus PSU. With this system, I idle at around 45W and load at 60W. Im planning on taking the CPU and motherboard from this system to build a basic PC for my parents, so I'd like to replace my current WHS CPU + Mobo with something else.
So after reading probably every low power related thread on this board, I've decided that maybe Ill get a AMD LE-1250 cpu paired up with the popular Gigabyte GA-MA74GM-S2 motherboard. I am just wondering what you guys think before I pull the trigger. Is this is lowest power consuming system I can build (ignoring intel atom)? I know there is a thread that suggests that the Biostar A760G-M2+ will consume even less power, but I live in Canada and I cannot get that board anywhere online cheaply. I particularly like the built in 6 SATA ports of the Gigabyte. Any other suggestions?
Since I dont do much with my WHS other than to serve/archive files and backup PCs, I'll probably underclock/undervolt the processor and memory as much as possible. I'll also disable any unneeded onboard peripherals such as IDE, sound, ports (serial, coms, floppy). Ideally Im aiming for 30W, the lower the better.
WHS Build. The lowest power possible. LE-1250 + 740G?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
In theory, 740G was the shrunk process version of 690G. With that in mind, 740G should draw less power as it doesn't add transistors to its core, unlike 780G brethrens. But, in reality, you've to face some difficulties when choosing 740G boards:
1. lackluster driver support: AMD has kind of discontinued pre-DX10 GPUs's support. In contrast to 78xG, who likes to have 12~15 driver releases next year, 740G will have only 3~4 driver releases in 2010.
2. scarce info concerning motherboard: Many mobo vendors don't send cheap of cheapo to reviewers, thus no info available. Gigabyte 740G was the sole 740G board that has been verified by SPCR, a far cry to 780G's flooding.
3. VRM phase changing: it's the technology that adjusts VRM phases based on CPU load in order to improve conversion efficiency. In other words, to save a few watts. Lately, I've tried similar technology with MSI and Asrock on AMD side, both of them have successfully shaved 3~5 watts at idle. The less VRM phases, the more power you save at idle. There's no love for 740G board.
4. Slower speed compared to "True" K10 chipset: rooted bact to 690G, 740G was indeed a K8/K9 chipset who can house AM3 CPU with reduced HyperTransport 1.0 (2000MT/s). You may lose 5% performance with the same AM3 CPU running identical frequency in comparison to the 785G boards. Of course, it doesn't matter to AM2 CPU like Sempron 1250.
As for lowest power consumption, ECS 740G sort of gave me a surprise as it drew 1~2 watts less than Gigabyte 740G board. However, Gigabyte's OC bios and 2 controllable fans is superior to ECS. Just my 2 cents, anyway.
1. lackluster driver support: AMD has kind of discontinued pre-DX10 GPUs's support. In contrast to 78xG, who likes to have 12~15 driver releases next year, 740G will have only 3~4 driver releases in 2010.
2. scarce info concerning motherboard: Many mobo vendors don't send cheap of cheapo to reviewers, thus no info available. Gigabyte 740G was the sole 740G board that has been verified by SPCR, a far cry to 780G's flooding.
3. VRM phase changing: it's the technology that adjusts VRM phases based on CPU load in order to improve conversion efficiency. In other words, to save a few watts. Lately, I've tried similar technology with MSI and Asrock on AMD side, both of them have successfully shaved 3~5 watts at idle. The less VRM phases, the more power you save at idle. There's no love for 740G board.
4. Slower speed compared to "True" K10 chipset: rooted bact to 690G, 740G was indeed a K8/K9 chipset who can house AM3 CPU with reduced HyperTransport 1.0 (2000MT/s). You may lose 5% performance with the same AM3 CPU running identical frequency in comparison to the 785G boards. Of course, it doesn't matter to AM2 CPU like Sempron 1250.
As for lowest power consumption, ECS 740G sort of gave me a surprise as it drew 1~2 watts less than Gigabyte 740G board. However, Gigabyte's OC bios and 2 controllable fans is superior to ECS. Just my 2 cents, anyway.
Thanks for the details.
1. Is driver support really that important though? On my current WHS, I think the only drivers I installed were the VGA ones. As long as the board is supported by Windowns 7 (since the new WHS will be based on this), then I figure it should suffice.
As for your other points, I definitely have to agree in that from the online reviews, it's really hard to determine how a certain board will consume power. I suppose the only way is to hear from everyone's experiences. Even the number of VRM stages is usually hard to figure out.
The 1 - 2 watt less power consumed by the ECS board is practically negligible. Im guessing by controlling the fan's voltage and its better underclocking options, those 1 - 2 could easily be saved on the Gigabyte board. Also, my experience of ECS boards are that they are usually crappy compared to some of the mainstream boards like MSI, Gigabyte, Foxconn etc...
1. Is driver support really that important though? On my current WHS, I think the only drivers I installed were the VGA ones. As long as the board is supported by Windowns 7 (since the new WHS will be based on this), then I figure it should suffice.
As for your other points, I definitely have to agree in that from the online reviews, it's really hard to determine how a certain board will consume power. I suppose the only way is to hear from everyone's experiences. Even the number of VRM stages is usually hard to figure out.
The 1 - 2 watt less power consumed by the ECS board is practically negligible. Im guessing by controlling the fan's voltage and its better underclocking options, those 1 - 2 could easily be saved on the Gigabyte board. Also, my experience of ECS boards are that they are usually crappy compared to some of the mainstream boards like MSI, Gigabyte, Foxconn etc...
1. I just don't like out-of-box discontinued product support. But 740 ought to suffice basic windows usage.
As for undervolt, I used CrystalCPUID to control frequency/voltage on both two boards. When it comes to other choices, though I don't have experience, Foxconn 740G also has 6 SATAs. I've seen other people recommending it on SPCR forum, but you may need to find the related info on yourself.
Btw, I already found Gigabyte 740G rev 4.0 has only 4 SATAs, beware! Well, what's your power? Power is deadly important under low power computing circumstance. In the past, I've had reduced the whole system from 66~68W to 55~56W with a simple power swap. And I just installed a Seasonic 200SFD, a dated 200W power, to replace an out of date Seventeam 350W. To make matters worse, all of my two PSUs are phase-out models by today's efficiency standard. All in all, that's where Pico-like unit really shines.
As for undervolt, I used CrystalCPUID to control frequency/voltage on both two boards. When it comes to other choices, though I don't have experience, Foxconn 740G also has 6 SATAs. I've seen other people recommending it on SPCR forum, but you may need to find the related info on yourself.
Btw, I already found Gigabyte 740G rev 4.0 has only 4 SATAs, beware! Well, what's your power? Power is deadly important under low power computing circumstance. In the past, I've had reduced the whole system from 66~68W to 55~56W with a simple power swap. And I just installed a Seasonic 200SFD, a dated 200W power, to replace an out of date Seventeam 350W. To make matters worse, all of my two PSUs are phase-out models by today's efficiency standard. All in all, that's where Pico-like unit really shines.
Im running an Antec Earthwatts 380 which I would estimate to be 75 - 80% efficient when running a super light load of 30 watts. I suppose that could always be replaced. I'm not sure how well those Pico-PSUs are at powering 4-6 hard drives though. I could always get a CarPC power supply which has the advantage of a wider unregulated voltage input, but the problem is finding a high current 12V AC-DC brick.
Considering the significantly higher cost of a AC-DC brick + DC-DC psu as compared to a standard ATX psu, the time it takes for the extra power savings to pay itself out is very long. I wish they would make more 100 - 200 watt standard ATX power supplies that are 80+.
How efficient are the GeForce 6100, 7025 and 760G boards compared to the 740G? I know the 780G is less efficient than the 740G.
Considering the significantly higher cost of a AC-DC brick + DC-DC psu as compared to a standard ATX psu, the time it takes for the extra power savings to pay itself out is very long. I wish they would make more 100 - 200 watt standard ATX power supplies that are 80+.
How efficient are the GeForce 6100, 7025 and 760G boards compared to the 740G? I know the 780G is less efficient than the 740G.
I'm afraid that EA380 won't be that efficient at 30W loading. You need to know 30W loading for normal ATX PSU is ridiculously difficult to achieve good efficiency.
Take a look at Corsair CX400, a similar wattage rated 80Plus PSU, review for your reference.
http://www.silentpcreview.com/Corsair_CX400W
I suggest this DC-DC unit for you, SPCR has reviewed&proven anyway. I didn't experience it, but it looks good from paper. Also, the owner ships goods to canada.
http://www.silentpcreview.com/Winmate_DD-24AX
viewtopic.php?t=55153
As for efficiency, you cannot judge power consumption solely on chipset alone. The same chipset with different motherboard vendors' implementation will draw varying power figures across the boards despite identical chipset TDP rating. Please refer to SPCR's 780Gs review, you'll learn this fact. To complicate matters, I think mobo implementation is more important than chipset spec from my experience. For example, 760G is 780G minus HD Video offloading. In theory, 760G should consume more power than 740G. But Biostar A760G-M2+ seemed to prove that chipset TDP is not a clear indicator based SPCR forum discussion. That said, you need to specify the board you want instead of general chipset request. Were I you, I'd choose a boad that can meet my demand and then try to use all known techniques to lower power consumption.
There are two suggestions for you:
1. try to stick with proven components like DC-DC unit, Gigabyte 740G, MSI 785G, or Biostar 760G board, regardless of cost.
2. get the component that you can purchase at the lowest pricing. After that, use all known techniques to reduce power consumption. Btw, you may consider sharing your findings to SPCR members if you go this route. That's where proven parts coming from.
Take a look at Corsair CX400, a similar wattage rated 80Plus PSU, review for your reference.
http://www.silentpcreview.com/Corsair_CX400W
I suggest this DC-DC unit for you, SPCR has reviewed&proven anyway. I didn't experience it, but it looks good from paper. Also, the owner ships goods to canada.
http://www.silentpcreview.com/Winmate_DD-24AX
viewtopic.php?t=55153
As for efficiency, you cannot judge power consumption solely on chipset alone. The same chipset with different motherboard vendors' implementation will draw varying power figures across the boards despite identical chipset TDP rating. Please refer to SPCR's 780Gs review, you'll learn this fact. To complicate matters, I think mobo implementation is more important than chipset spec from my experience. For example, 760G is 780G minus HD Video offloading. In theory, 760G should consume more power than 740G. But Biostar A760G-M2+ seemed to prove that chipset TDP is not a clear indicator based SPCR forum discussion. That said, you need to specify the board you want instead of general chipset request. Were I you, I'd choose a boad that can meet my demand and then try to use all known techniques to lower power consumption.
There are two suggestions for you:
1. try to stick with proven components like DC-DC unit, Gigabyte 740G, MSI 785G, or Biostar 760G board, regardless of cost.
2. get the component that you can purchase at the lowest pricing. After that, use all known techniques to reduce power consumption. Btw, you may consider sharing your findings to SPCR members if you go this route. That's where proven parts coming from.
Thanks for all the suggestions so far. I'll definitely post my results in the end.
I ended up picking up a Gigabyte GA-MA74GM-S2 locally. I picked this board because of its proven low power consumption and 6 SATA ports. Although not as low as the Biostar A760G, but I couldnt find that board anywhere.
Now Im looking for a processor. I'm aiming for a cheap LE-1100, LE-1200, LE-1250 (ideally this one), LE-1300, LE-1640, or LE-1660. My most economic choice is to buy one used, so if anyone has one of these processors for sale, feel free to send me a PM. I've been scouring other forums for parts...it's somewhat surprisingly difficult to find specific parts (that aren't marked up ridiculously high) here in Canada sometimes.
As for the power supply, I dont intend on upgrading from my Earthwatts 380W unless I find another use for it. It wouldnt really make much economic sense. Let's say I could save maybe 3 - 5 watts by using a DC-DC. I'm guessing that would only work out to be around $2 - $4 in savings per year. The payout period for this incremental savings is much too long and not really worth it. However, if I was building from scratch or if I find use for the EA380W for another system, then it would be something to consider.
I ended up picking up a Gigabyte GA-MA74GM-S2 locally. I picked this board because of its proven low power consumption and 6 SATA ports. Although not as low as the Biostar A760G, but I couldnt find that board anywhere.
Now Im looking for a processor. I'm aiming for a cheap LE-1100, LE-1200, LE-1250 (ideally this one), LE-1300, LE-1640, or LE-1660. My most economic choice is to buy one used, so if anyone has one of these processors for sale, feel free to send me a PM. I've been scouring other forums for parts...it's somewhat surprisingly difficult to find specific parts (that aren't marked up ridiculously high) here in Canada sometimes.
As for the power supply, I dont intend on upgrading from my Earthwatts 380W unless I find another use for it. It wouldnt really make much economic sense. Let's say I could save maybe 3 - 5 watts by using a DC-DC. I'm guessing that would only work out to be around $2 - $4 in savings per year. The payout period for this incremental savings is much too long and not really worth it. However, if I was building from scratch or if I find use for the EA380W for another system, then it would be something to consider.
Very close to plugging the trigger on a cpu now.
What do you guys think about the Sempron LE-1200 vs Athlon LE-1640 more specifically? I can get the LE-1640 for only $6 more than the LE-1200. I probably wont need the extra processing power, but which one is more power efficient? The G1 stepping on the LE-1200 is less efficient than the G2 on the LE-1250/LE-1640, but Im assuming the difference is negligible? The LE-1640 has virtualization.
I'll be underclocking the processor probably to 1GHZ @ 0.9v regardless of the stock speed.
What do you guys think about the Sempron LE-1200 vs Athlon LE-1640 more specifically? I can get the LE-1640 for only $6 more than the LE-1200. I probably wont need the extra processing power, but which one is more power efficient? The G1 stepping on the LE-1200 is less efficient than the G2 on the LE-1250/LE-1640, but Im assuming the difference is negligible? The LE-1640 has virtualization.
I'll be underclocking the processor probably to 1GHZ @ 0.9v regardless of the stock speed.
-
MikeC
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
I have a system being used as a torrent & music server & HTPC. Close enough to your application to be relevant. My main comment is that you do not need to accept low speed/performanice to get super low power consumption.
Here are the details --
AC power:-
35-36W idle w/ screen off (I don't use a screensaver)
40-41W idle w/ screen on
50-55W playing 1080p video
55-60W extreme multitasking - playing 1080p video, running utorrent, winrar & Avira antivirus, streaming music via ethernet -- and amazingly, the video remains perfectly viewable >99% of the time.
Occasional peaks to around 65-70W are seen mostly when starting programs, but these are very short term. 50W is the typical in-use power when watching video.
Components:-
AMD Athlon2 X2 240e (dualcore 2.8 ghz, 2mb cache - 45W TDP)
MSI 785GN-E65 motherboard
4gb DDR3-1333 (2 sticks)
Seagate 500gb Momentus 5400rpm
Antec NSK1480 w/ its own PSU (80+ but not very efficient at these low power levels)
All BIOS settings are pretty much on auto or default; no voltage tweaking at all. Of course it is running CoolnQuiet, on Windows 7 ultimate - 64bit.
My guess is that w/ a more efficient PSU (the Winmate or picoPSU w/ efficient adapter or maybe an 80+ Gold 300W PSU -- if they every become available), the power could drop another 5W or more.
Here are the details --
AC power:-
35-36W idle w/ screen off (I don't use a screensaver)
40-41W idle w/ screen on
50-55W playing 1080p video
55-60W extreme multitasking - playing 1080p video, running utorrent, winrar & Avira antivirus, streaming music via ethernet -- and amazingly, the video remains perfectly viewable >99% of the time.
Occasional peaks to around 65-70W are seen mostly when starting programs, but these are very short term. 50W is the typical in-use power when watching video.
Components:-
AMD Athlon2 X2 240e (dualcore 2.8 ghz, 2mb cache - 45W TDP)
MSI 785GN-E65 motherboard
4gb DDR3-1333 (2 sticks)
Seagate 500gb Momentus 5400rpm
Antec NSK1480 w/ its own PSU (80+ but not very efficient at these low power levels)
All BIOS settings are pretty much on auto or default; no voltage tweaking at all. Of course it is running CoolnQuiet, on Windows 7 ultimate - 64bit.
My guess is that w/ a more efficient PSU (the Winmate or picoPSU w/ efficient adapter or maybe an 80+ Gold 300W PSU -- if they every become available), the power could drop another 5W or more.
Last edited by MikeC on Fri Jan 08, 2010 2:27 pm, edited 3 times in total.
-
MikeC
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
No way. Under 70%, maybe 65% at best. We got 71% at 43W load w/ an EA430 in the lab.speedboxx wrote:Im running an Antec Earthwatts 380 which I would estimate to be 75 - 80% efficient when running a super light load of 30 watts.
Last edited by MikeC on Thu Jan 07, 2010 2:17 pm, edited 1 time in total.
With my build I was thinking of picking up an Arctic Cooling 7 Pro Rev2 and running it fanless. But then I came across this article which seems to indicate a correlation between VRM cooling and power efficiency:
http://www.silentpcreview.com/article285-page4.html
The cooler the VRMs, the less power will be drawn. I couldnt find much information on this, maybe you guys could share some experiences? I'm almost certain that I can run an underclocked LE series processor off this heatsink fanless, but maybe it would be worth it to incorporate some sort of dedicated fan to cool the VRMs?
http://www.silentpcreview.com/article285-page4.html
The cooler the VRMs, the less power will be drawn. I couldnt find much information on this, maybe you guys could share some experiences? I'm almost certain that I can run an underclocked LE series processor off this heatsink fanless, but maybe it would be worth it to incorporate some sort of dedicated fan to cool the VRMs?
-
MikeC
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
I've mentioned this in the forums over the years but haven't done a serious article on the main site. The gist is simple: W/O active cooling, under extreme load (like in heatsink testing w/ ridiculously high, long loads on the CPU), the VRMs get really hot over time, and power demand increases significantly w/ temperature -- beyond some ideal range. Most MOSFETs (which are usually used for VRM circuits) have an ideal temp range where efficiency is maximized. This is much hotter than you might think -- too hot to touch, generally -- but go beyond that, and efficiency drops quite quickly. Similar to most machinery, including car engines, etc.speedboxx wrote:With my build I was thinking of picking up an Arctic Cooling 7 Pro Rev2 and running it fanless. But then I came across this article which seems to indicate a correlation between VRM cooling and power efficiency:
http://www.silentpcreview.com/article285-page4.html
The cooler the VRMs, the less power will be drawn. I couldnt find much information on this, maybe you guys could share some experiences? I'm almost certain that I can run an underclocked LE series processor off this heatsink fanless, but maybe it would be worth it to incorporate some sort of dedicated fan to cool the VRMs?
Is this significant in actual use as opposed to extreme lab test conditions? If you game for long periods, do lots of other CPU-intensive taks or have high ambient temps -- say 30C or higher -- w/o some forced airflow around the VRMs, your power draw will probably be higher.
How much higher? Hard to say. As little as a watt or two if you don't push the CPU that hard but as much 10W if you have a hot CPU and extended high loads on the CPU.
With a low power Sempron, it's probably not significant, esp. as you probably won't be gaming w/it.
Personally, I prefer low-speed fan cooling for most of my builds. A tiny bit of forced airflow is still better in most cases, than none. Cheap insurance against f-ups.
Thanks for the detailed reply. I was curious because SPCR did a review on the popular Gigabyte GA-MA74GM-S2 and I believe it was paired up with a fairly low power processor, yet the review stated that the VRMs get really hot. Im assuming that motherboard designers design the boards such that the VRMs get cooled from the air blown past the processor, so maybe it wouldnt be a good idea to go passive.
I was hoping that my power supply fan would be the only fan in the system, but I guess Im rethinking that strategy.
I was hoping that my power supply fan would be the only fan in the system, but I guess Im rethinking that strategy.
The only real disadvantage of this Gigabyte board is that its BIOS insists that the PCIe x16 slot must have a GPU in it. If you get to the point where you need to add additional disks and try to put a SATA controller in that slot then the board won't POST and you are dead-stop. The Biostar board is happy with any type of card in that slot. Other than that detail you've made a good choice.speedboxx wrote:Thanks for all the suggestions so far. I'll definitely post my results in the end.
I ended up picking up a Gigabyte GA-MA74GM-S2 locally. I picked this board because of its proven low power consumption and 6 SATA ports. Although not as low as the Biostar A760G, but I couldnt find that board anywhere.
I agree with MikeC here - you don't need to go to the bottom of the CPU barrel to get a good, quiet low power system. I know they are getting a bit old now, but the Athlon-II x2 series provided a really good balance of speed vs power. You do not really want to go with a single-core processor- pick just about any CPU from that series that you can find on sale and you'll be happy. Get out of the LE-xxxx lineup.speedboxx wrote:Now Im looking for a processor. I'm aiming for a cheap LE-1100, LE-1200, LE-1250 (ideally this one), LE-1300, LE-1640, or LE-1660. My most economic choice is to buy one used, so if anyone has one of these processors for sale, feel free to send me a PM. I've been scouring other forums for parts...it's somewhat surprisingly difficult to find specific parts (that aren't marked up ridiculously high) here in Canada sometimes.
-
MikeC
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
speedboxx --
It really should be noted: The VRMs on most boards always get really hot at high load.
Looking at all the details.
That gigabyte 740 drew less power at idle than the boards it was compared to. It was more efficient than the other boards at low power. But at extended full load, the giga740 drew a bit more power. Why is hard to tell, because the other boards didn't have the same chipset.
Again, if you don't push the CPU hard for long periods, the nearby fan in your PSU would probably be enough (with a big cpu heatsink optimized for low airflow) that there'd be no hit on efficiency. But if you push the CPU hard a lot or have a heatsink with tightly spaced fans, keeping a fan spinning slowly on the heatsink is advised.
Arctic Cooling 7 Pro Rev2 has fairly tight fin spacing, btw -- compared to a Ninja or similar, for example.
It really should be noted: The VRMs on most boards always get really hot at high load.
Looking at all the details.
That gigabyte 740 drew less power at idle than the boards it was compared to. It was more efficient than the other boards at low power. But at extended full load, the giga740 drew a bit more power. Why is hard to tell, because the other boards didn't have the same chipset.
Again, if you don't push the CPU hard for long periods, the nearby fan in your PSU would probably be enough (with a big cpu heatsink optimized for low airflow) that there'd be no hit on efficiency. But if you push the CPU hard a lot or have a heatsink with tightly spaced fans, keeping a fan spinning slowly on the heatsink is advised.
Arctic Cooling 7 Pro Rev2 has fairly tight fin spacing, btw -- compared to a Ninja or similar, for example.
Yea, I realize the Biostar A740G is better than the Gigabyte board in almost every way (ECC support, SATA card support in the PCIe-x16, lower power), but like I mentioned, I really couldnt find anywhere to get the Biostar for cheap here in Canada.
But I figure Ill probably never need more than 6 hard drives. Im having trouble even filling my existing 1TB drive. There's always the two PCI and single PCIe-x1 slot for additional SATA cards.
I picked up an LE-1250 today, will probably buy the Arctic Cooler 7 Pro heatsink after work and I'll let you guys know how things turn out =)
But I figure Ill probably never need more than 6 hard drives. Im having trouble even filling my existing 1TB drive. There's always the two PCI and single PCIe-x1 slot for additional SATA cards.
I picked up an LE-1250 today, will probably buy the Arctic Cooler 7 Pro heatsink after work and I'll let you guys know how things turn out =)
Update: Ok I got my new system installed. Summary of what I have:
AMD Sempron LE-1250 CPU + Arctic Cooling Freezer 7 Pro Rev.2
Gigabyte GA-MA74GM-S2 Motherboard
1.5GB Ram (Kingston ValueRam 512MB + 1GB)
640GB WD Blue OS drive
1TB WD Green data drive
Antec Earthwatts 380W PSU
APC BE450G UPS
Im surprised how easily the migration went. I just unplugged the drives from my old system then put it in the new board and it booted straight up! The only missing driver was an "SM Bus Controller" and of course I had to reactivate WHS. Very thankful that I didnt have to do a full reinstall.
Everything is fully underclocked and all uneccesarry peripherals turned off in the BIOS (IDE, serial port, audio, parallel). HT Link, Ram, VGA, FSB frequencies are all set to the lowest possible. CPU is running at 1ghz, 0.85v. Im idling at about 38w and full load never exceeds 45w measured at the UPs inlet. I was hoping for close to 30w idling, but oh well. In comparison my old Celeron 430 + Foxconn G31MV-K idled at 45W and load 60W with the same drives and psu.
Im going to let this run for a bit and see how it goes. First impression is that the CPU seems to be a bit more stressed from looking at the Task Manager. Also, Ive noticed my physical memory has about 300MB less available after a fresh boot as compared to before. I wonder if maybe clocking my CPU a bit higher would actually decrease over power consumed since the CPU would spend less time at a higher load.
AMD Sempron LE-1250 CPU + Arctic Cooling Freezer 7 Pro Rev.2
Gigabyte GA-MA74GM-S2 Motherboard
1.5GB Ram (Kingston ValueRam 512MB + 1GB)
640GB WD Blue OS drive
1TB WD Green data drive
Antec Earthwatts 380W PSU
APC BE450G UPS
Im surprised how easily the migration went. I just unplugged the drives from my old system then put it in the new board and it booted straight up! The only missing driver was an "SM Bus Controller" and of course I had to reactivate WHS. Very thankful that I didnt have to do a full reinstall.
Everything is fully underclocked and all uneccesarry peripherals turned off in the BIOS (IDE, serial port, audio, parallel). HT Link, Ram, VGA, FSB frequencies are all set to the lowest possible. CPU is running at 1ghz, 0.85v. Im idling at about 38w and full load never exceeds 45w measured at the UPs inlet. I was hoping for close to 30w idling, but oh well. In comparison my old Celeron 430 + Foxconn G31MV-K idled at 45W and load 60W with the same drives and psu.
Im going to let this run for a bit and see how it goes. First impression is that the CPU seems to be a bit more stressed from looking at the Task Manager. Also, Ive noticed my physical memory has about 300MB less available after a fresh boot as compared to before. I wonder if maybe clocking my CPU a bit higher would actually decrease over power consumed since the CPU would spend less time at a higher load.
Another update. It appears that the GA-MA74GM-S2 board + LE-1250 boots fine and WHS runs fine when the LE-1250 is clocked to 1ghz and 0.8v, but the board will hang when rebooting. I had to bring the voltage up to 0.9v for it to be stable. Unfortunately now, my idling power is 42W which isnt much of an improvement from my old Celeron 430 based system. I think the LE-1250 would actually perform worse since it's so heavily underclocked.
It's silly to pick out components specifically for power-saving and then ignore the potential of dynamic undervolting. But from your comments, it does not seem like you actually understand what programs like RMClock and CrystalCPUID can achieve.psiu wrote:Or just let Cool N Quiet handle things, I generally don't even worry about the extra 0.005V of undervolt I might get. Plus you get full speed when needed.
Using CrystalCPUID with custom p-states, you get full speed when needed, but can typically idle at signficantly less voltage than what Cool 'N Quiet will provide (as Cool 'N Quiet must be conservative).
No I've used both programs--I guess it depends on how much time one wants to spend experimenting for stability, for me, it's not that much. Picking out low power parts also seems to result in the processors being marked as low voltage anyway--CNQ will drop down to the minimum voltage and speed itself. My BE-2300 won't run at a 4x multiplier, and if I remember right it won't boot at .8V but auto-clocks down to .825 with CNQ.davidh44 wrote:It's silly to pick out components specifically for power-saving and then ignore the potential of dynamic undervolting. But from your comments, it does not seem like you actually understand what programs like RMClock and CrystalCPUID can achieve.psiu wrote:Or just let Cool N Quiet handle things, I generally don't even worry about the extra 0.005V of undervolt I might get. Plus you get full speed when needed.
Using CrystalCPUID with custom p-states, you get full speed when needed, but can typically idle at signficantly less voltage than what Cool 'N Quiet will provide (as Cool 'N Quiet must be conservative).
Ive never actually used an AMD CPU with CnQ before actually. With WHS, how do you ensure that it is working? Do you have to install a driver for QnC to work, or is it handled by the motherboard?
I already clocked my LE-1250 CPU down to 1ghz @ 0.9v so I doubt CnQ will do anything further past that. This is the lowest I can go without stability issues.
I already clocked my LE-1250 CPU down to 1ghz @ 0.9v so I doubt CnQ will do anything further past that. This is the lowest I can go without stability issues.