Which AMD CPU system best for lowest electricty usage?

ED · Post by ED » Fri Oct 15, 2004 3:49 pm

Am building an HTPC and want to find the system with the lowest heat generation, lowest noise generation and uses up the least electricity. Have been looking at getting an mATX motherboard that's undervoltable and underclockable but not sure which CPU type to go for - an AMD XP-M or an AMD 64 (not sure if you can get them in mobile version?) - as the AMD-64 has Cool'n'Quiet while the XP-M doesn't - would the AMD XP-M use less electricity than the AMD-64 or would the 64 bit processor be better due to Cool'n'Quiet? I would use the Zalmann 7000-Cu heatsink with the XP-M and the Thermaltake XP-120 heatsink with the 64bit processor, and aim to run either processor at around 1GHz.

Thanks very much for your help in advance

Cheers - Piers

roo · Post by **roo** » Fri Oct 15, 2004 4:24 pm

i would like to know as well

btw AMD makes the A64 mobile

mc2wheels · Post by **mc2wheels** » Fri Oct 15, 2004 5:05 pm

Actually, I am running a 35w mobile 2400+ running at 5x200 @ 1.1v. I could go lower by dropping the FSB (5 is the lowest multiplier my motherboard does), but I wanted to avoid sync issues with my memory and my video card. I am also using the Zalman 7000 alcu with the fan all the way down (uses fanmate). I also use a Coolmax Taurus CX power supply with one 12cm fan on low. Both of those are not really audible. I also use one 80mm Zalman fan running a 7v. My case temp and cpu stay at about 42 degrees. The only thing I can hear is the zalman front fan -- think I'm going to replace it with panaflo or nexus. Anyway, I was looking for a mATX board that had voltage, FSB, and multiplier control but I couldn't find one. I ended up going with a DFI NFII Ultra Infinity.

I never tried looking for Athlon 64 mATX boards (wanted to save money). But if you can find the board, and you have the cash, you might want to go with that.

sthayashi · Post by **sthayashi** » Sat Oct 16, 2004 2:00 pm

I REALLY want to know about the mobile Athlon 64 low power draw. So far, I haven't had the balls to give it a try.

What I DO have currently is an Athlon XP-M 2600+ matched with Mushkin PC2700 RAM, on an Aopen AK79D-400VN. There are 2 120mm Globe Fans and 1 Panaflo L1A on here. The hard drive is a Samsung 40GB. There is also a Seagate 7200.7 attached via USB (but power seperately).

I have it clocked stable at 1.2GHz at 1.1V. With a Seasonic Tornado 300W PSU, I measure about 49-50W power draw at idle. I underclock using 8rdavcore. At boot, it runs 2GHz @ 1.35V, and draws approximately 60-70W.

If anyone has any information on system power draw of an undervolted A64 system, I'd love to hear it.

Tobias · Post by **Tobias** » Sat Oct 16, 2004 4:48 pm

Well, my "kill-a-watt" isn't very trustworthy (right now it reports 18w, which is a tad on the low:) however it also report 0.3A which at 220V would be about 66w idle ([email protected]) and running prime95 it reports 0.39A ([email protected], A64 2800+) which would give about 86w. I also have 512MB RAM, 9600PRO and a seagate 7200.7 and an Abit mobo. CPU Power suggests that the CPU draws 19w at idle (counting with 89w max powerdraw, although the difference between idle and load suggest a maximum of about 75w), which leaves with about 47w for the rest of the system. I have to admit, it baffles me a bit.

My PSU is rubbish though, and it would be a happy surprise if it is even 60% efficient (it's old and not even close to brandname). I am pretty sure a Seasonic would push it down close to the numbers you have sthayashi

sthayashi · Post by **sthayashi** » Sun Oct 17, 2004 11:10 am

Tobias wrote:Well, my "kill-a-watt" isn't very trustworthy (right now it reports 18w, which is a tad on the low:) however it also report 0.3A which at 220V would be about 66w idle ([email protected]) and running prime95 it reports 0.39A ([email protected], A64 2800+) which would give about 86w. I also have 512MB RAM, 9600PRO and a seagate 7200.7 and an Abit mobo. CPU Power suggests that the CPU draws 19w at idle (counting with 89w max powerdraw, although the difference between idle and load suggest a maximum of about 75w), which leaves with about 47w for the rest of the system. I have to admit, it baffles me a bit.

My PSU is rubbish though, and it would be a happy surprise if it is even 60% efficient (it's old and not even close to brandname). I am pretty sure a Seasonic would push it down close to the numbers you have sthayashi

Keep in mind that that 220V @ 0.3A = 66VA, not 66W. VA * PF = W. Though at the same time, I've never heard of a PSU that has a PF of 0.27.

Thank you for the information though. Now I have to debate whether I should build a low powered A64 system. I just wanted to know the total system power draw because after a certain point, reducing the CPU results in diminishing returns, i.e. your chipset/memory/harddrive/videocard draws more power than your CPU. Until today though, I had not seen even a ballpark figure for low powered A64 system.

BTW, I don't fully believe that your CPU is drawing 19W if it's based on the 89W max, since almost no one knows how much an A64 2800+ draws at max (to put it into perspective, AMD gives the same max power draw for the A64 2800+, 3000+, and 3200+). Just my opinion though...

Edit: I should clarify this a bit. VA = W / PF. Between different PSUs, the wattage is the independent variable, not VA.

mczak · Post by **mczak** » Tue Oct 19, 2004 5:53 am

The 90nm A64 are probably the chips to get for low power consumption. There still aren't that many numbers out there regarding power consumption of these chips, but the numbers so far are truly impressive.
For instance http://www.anandtech.com/cpuchipsets/sh ... =2249&p=13 a 2.2Ghz 90nm 3500+ uses 30W less under full load than a 1.8Ghz 130nm 3200+! Translate the 30W from AC-outlet to cpu power, and that's still 20W less. Idle consumption is lower too.
It is not yet known though how well those chips will undervolt (the 130nm 2Ghz Clawhammers are known to undervolt without much trouble to around 1.3V/2Ghz and 0.9V/1Ghz), but it seems likely they won't disappoint.
The 90nm A64 imho look like THE chip to beat wrt performance/power (save the pentium-m).

pod03 · Post by **pod03** » Tue Oct 19, 2004 9:35 am

It appears that AthlonXPs will idle lower than A64s. Search for pod03 as the author and you will find the reasosns. Basically XPs can go down to 300MHz (5x60) at 1.1v (4.7W) whereas it appears that A64s dont go as low MHz with current BIOSs. However at 1000MHz an A64 machine would probably be more useful and efficient than a AXP at 1000MHz. So it depends how much idling the processor will do.

Mark

Bitter Jitter · Post by **Bitter Jitter** » Tue Oct 19, 2004 10:37 am

You can run the Athlon 64 at 800Mhz, with a 4 multiplier. To go lower you can lower the FSB to get a lower frequency. Clockgen should allow you to do this in windows with nforce3 boards.
Please correct me if i'm wrong.

Tobias · Post by **Tobias** » Tue Oct 19, 2004 11:08 am

Depends on the stepping of the CPU. my CG A64 will freeze up as soon as I even think about dropping the mulitplier below 5.

So does 66VA*PF (powerfactor, right?) mean that using a deacent (for a passive PFC-unit) that wattage draw is below 50?

mczak · Post by **mczak** » Tue Oct 19, 2004 12:42 pm

pod03 wrote:It appears that AthlonXPs will idle lower than A64s. Search for pod03 as the author and you will find the reasosns. Basically XPs can go down to 300MHz (5x60) at 1.1v (4.7W) whereas it appears that A64s dont go as low MHz with current BIOSs.

True, XP will go to a lower freqency more easily than A64. Though I consider the 4.7W number you quote to be just flat out wrong. AMD does unfortunately not publish enough information to figure out everything needed to calculate power draw at a given voltage/frequency - in particular you don't know how much it draws even if the clock would be 0 (and no that won't be 0W, not even close, due to current leakage which is independant of clock frequency). And this is exactly where the A64 should have quite a big advantage due to SOI which should reduce current leakage.

However at 1000MHz an A64 machine would probably be more useful and efficient than a AXP at 1000MHz.

Definitely. Also, just look at the published power draw numbers (which are, as mentioned, not available for xp-m, but are quite interesting for A64). A desktop A64 (cg-step) has a max power draw of 22W at 1.1V/1Ghz. The mobile versions are quoted as being capable of .95V/800Mhz which brings this down to 13W (actually the raw calculation would equal 14.5W). There is very little to zero doubt that any A64 (cg-step) can reach 0.95V/800Mhz (I know mine does 0.85V/1Ghz, and this is not extraordinary).
But at 15W there is already no problem at all cooling these cpus passively (tried that without problem). So why do you even want to go lower? Unless you're running on battery there doesn't seem that much of a point.

So it depends how much idling the processor will do.

If it's truely idling the power consumption will of course still be a lot lower - remember those 15W are the maximum power draw for a given voltage/frequency setting (i.e. running cpuburn). Stop Grant is below 5W including I/O power at 0.95V/800Mhz.
So, yes, the xp-m may have an advantage due to its very low achievable frequency - but as said, if it's really idling it doesn't matter at all, and if you actually need the cpu, are you sure 300Mhz are enough? And I believe wrt performance/power the A64 will blow it out of the water.

sthayashi · Post by **sthayashi** » Tue Oct 19, 2004 8:17 pm

Tobias wrote:So does 66VA*PF (powerfactor, right?) mean that using a deacent (for a passive PFC-unit) that wattage draw is below 50?

Yes. Assuming that portion of your Kill-A-Watt is working properly. BTW, I've edited my post above to hopefully make it a little clearer.

dago · Post by **dago** » Tue Oct 19, 2004 10:14 pm

For the VA/...
So does 66VA*PF (powerfactor, right?) mean that using a deacent (for a passive PFC-unit) that wattage draw is below 50?

Well, normally the 'optimal' powerfactor is 1. From what I know, Active PFC PSU are around 0.95, and passive PFC between 0.8 and 0.9.

It's strange that your kill-a-watt doesn't give you all of these (active power, total power, power factor).

Moreover, for your CPU and to measure its consumption, it's the active power that is important (in VA). Even when ppl talk about "W", the real electrical power used is the active part of it, but as the PF is supposed to be near 1. This means your whole computer is using 66 "W". (from my understanding).

pod03 · Post by **pod03** » Wed Oct 20, 2004 12:43 pm

The original question included which 'uses up the least electricity'. According to CPUpower at 300MHz the CPU is using 4.7W in my setup http://www.silentpcreview.com/article164-page1.html a figure which MikeC seems happy with. This compares favourably to 13W.

Mark

barleyguy · Post by **barleyguy** » Wed Oct 20, 2004 1:05 pm

The Athlon 64 Mobile will run at a multiplier of 4, at 0.8 volts. It only uses 12 Watts of power at that setting. It works just fine at minimum speed with a Zalman 7000. It even feels pretty fast at that low speed.

Clockgen can be used to set both the multiplier and bus speed. The 405 version of clockgen works for most KT800 boards, and there is a generic version of clockgen for all NForce 3 boards.

I'm using a Athlon 64 Mobile 2700, set at 225x8 (1800 Mhz) for normal use, and 200x4 when I want it to be extra silent.

By the way, the mobiles seem to default to a 4 multiplier on most desktop boards, so you have to use clockgen to get the normal multiplier if you want it.

Peace,
Harley.

sthayashi · Post by **sthayashi** » Wed Oct 20, 2004 8:41 pm

pod03 wrote:The original question included which 'uses up the least electricity'. According to CPUpower at 300MHz the CPU is using 4.7W in my setup http://www.silentpcreview.com/article164-page1.html a figure which MikeC seems happy with. This compares favourably to 13W.

Mark, you know I really respect you on this achievement. I can honestly say that your efforts in this area have directly inspired me to examine low powered systems. However, I would consider 4.7W to be a bit misleading for the original question.

The CPU may draw 4.7W, but I'd wager that the rest of the system pulls a lot more power than that. Really, even at the 13W level, Graphics cards and even just chipsets can be a significant source of power consumption in comparison. This is why at such low levels, system power draw becomes more interesting.

rperezlo · Post by **rperezlo** » Thu Oct 21, 2004 12:58 am

Regarding Clockgen and Cool'n'quiet I have found C'n'Q to be the more stable solution, almost as efficient as Clockgen and it is automatic.

Details: My MB is a Gigabyte K8NS Pro. I have an A64 3000+ normally running at 245x10 = 2450 MHz @ 1.55 V. Temperature at idle is 43 C.

- With Cool'n'quiet the low load mode is 5X (1225 MHz) @ 1.1 V. T = 36C.
- With Clockgen I can get 4X 185 (740 MHz) @ 0.8 V. T = 35 C. Occasionally manual changes of configuration crash the computer, and not only in extreme settings. I've noticed this doesn't happen at all if I let a couple of seconds between changes of different parameters.
- INI file configuration changes in Clockgen crash the computer almost always. It seems as it needs a time to settle down after a change. Probably a newer version with some real Nforce3 testing will do the trick.

For me the surprise is that the difference from 1.55 V to 1.1 V is a lot more than from 1.1 V to 0.8 V. I think this means that C'n'Q achieves it's power reduction not only by lowering Vcore and multiplier but also by activating some microprocessor 'save power' pin.

Frequency has very little influence in temperature compared to Vcore.

I switched from Intel to AMD a couple of weeks ago and I'm extremely happy with the power saving features. I can be working in my computer with only the CPU fan working at 15% rpm. Unfortunately the AMD stock cooler is incredibly noisy (I can hear it when it starts at 4%), so that I'll be changing it's fan quite soon, or maybe looking for a new HSF. Not sure if it is worthed anymore with this level of sound, shame of crap stock fan.

meglamaniac · Post by **meglamaniac** » Thu Oct 21, 2004 1:28 am

There is something important to remember when comparing draws between AthlonXP's and Athlon64's.
The majority of AthlonXP boards, especially the earlier revisions, have the HALT instruction disabled in the BIOS, so the processor never truely idles. It just sits there running NOOP instead, which keeps it significantly hotter (and uses more power).

This is because the voltage regulators on many boards couldn't cope with with the massive power swings the XPs were asking for when they went from HALT to full load in a few milliseconds. Voltages would spike out of range and the system would fall over. In some cases, this eventually killed the voltage regulators.
The "safe" solution was to disable HALT - as on my Asus A7V333, with the result that the processor idles a mere 7C below it's load temp.

Technology and quality has been improved now so new XP boards and all 64 boards should support HALT.

Just some info for you to consider.

rperezlo · Post by **rperezlo** » Thu Oct 21, 2004 1:47 am

Very good feedback, meglamaniac. My load temperature is about 58C, more than 20 degrees above the idle temperature. And this could also explain why several Clockgen changes without time to settle down crash my computer (power circuits not able to cope with changes). In fact sometimes it crashes just going back to 200x10 and I couldn't find an explanation for that.

Why C'n'Q doesn't crash in similar conditions is probably in the knowledge of AMD.

Does CLockgen also use the HALT line? I gues it doesn't because if it did the power compsumption at 0.8 V should be roughly half of that at 1.1 V, and in my case there is almost no temperature difference.

meglamaniac · Post by **meglamaniac** » Thu Oct 21, 2004 2:18 am

HALT is a privileged processor instruction.
That means only processes that have "special state" control over the processor can run it. I think, but I'm not sure, the processor has to be in Non-Interruptable Special State before it can execute a HALT (this would make sense - how would an interrupt execute if the processor isn't "listening"?).

Windows keeps Interruptable Special State for windows services and drivers, and Non-Interruptable Special State for the kernel.
Everything else runs in User State.
So, as far as I am aware, the only process that can call HALT is the kernel, which it will presumably do when nothing else requires processor time.

As you note, sudden speed changes by Clockgen will produce large changes in power draw. However, I would have thought the regulators should be up to the task by now - it's not like enough fuss hasn't been made about how much 12v power Athlon 64's need, and it's well known the 12v line is the least likely to be on target.

My guess would be that Clockgen switches the processor speed instantly, whereas C'n'Q maybe has preset stages and ramps through them on it's way up or down to smooth things out.
I have no evidence to back this up, it's just a shot in the dark.

pod03 · Post by **pod03** » Thu Oct 21, 2004 2:21 am

sthayashi, it was a bit misleading, sorry. However, it is very difficult to get a good idea of system draw without building and testing systems with a meter at various system setups. I only have one system and no meter. So if anyone has good low speed A64 and AXP set ups with PSUs with the same efficiency (same psu?) and a reliable meter I would be interested in a comparison at lowest possible speed, voltage, etc. Also at higher settings where there is similar performance.

I checked DVD playback at 300MHz last night and it was fine on my system into stereo. I suspect that with soundstorm directly sending DirectSound3D signal as a Dolby Digital encoded stream to an external receiver this might be efficient and high quality. So for a HTPC this might be the best set up for good quality, low noise, energy and low heat. I cannot be sure.

Mark

rperezlo · Post by **rperezlo** » Thu Oct 21, 2004 2:24 am

You may be right. Changes with Clockgen are instant (i.e. from 10X to 5X) but CnQ has steps in the middle. It is not easy to track them with CPUz because it's readouts don't refresh fast enough but I have seen 9X at least once.

Also, with CrystalCPUID you can set some timings and delays. I'll try them to see if they avoid the PC to crash.

mczak · Post by **mczak** » Thu Oct 21, 2004 4:52 am

pod03 wrote:The original question included which 'uses up the least electricity'. According to CPUpower at 300MHz the CPU is using 4.7W in my setup http://www.silentpcreview.com/article164-page1.html a figure which MikeC seems happy with. This compares favourably to 13W.

Sorry, but just because CPUpower says that doesn't mean it's true. I believe this program just uses the known power consumption numbers and extrapolates from that to underclock/undervolted cpus. You are _guaranteed_ to get completely bogus numbers at the very low frequency side of things that way.
And even if it would only be 5W, considering the performance difference to the 800Mhz A64 (whose 13W is a number with a lot more confidence), the A64 still comes out on top performance/power regardless - the xp-m might lead "least amount of electricity" but that's imho not that interesting - my pocket calculator easily undercuts that

.
The xp-m ought to be quite a bit cheaper though.

Jan Kivar · Post by **Jan Kivar** » Thu Oct 21, 2004 6:12 am

meglamaniac wrote:There is something important to remember when comparing draws between AthlonXP's and Athlon64's.
The majority of AthlonXP boards, especially the earlier revisions, have the HALT instruction disabled in the BIOS, so the processor never truely idles. It just sits there running NOOP instead, which keeps it significantly hotter (and uses more power).

This is because the voltage regulators on many boards couldn't cope with with the massive power swings the XPs were asking for when they went from HALT to full load in a few milliseconds. Voltages would spike out of range and the system would fall over. In some cases, this eventually killed the voltage regulators.
The "safe" solution was to disable HALT - as on my Asus A7V333, with the result that the processor idles a mere 7C below it's load temp.

I've read that the reason not to have HLT in early VIA boards is that the PCI implementation in the (early? read the fine print) VIA chipsets is so poor. Searching for "kt133 686B SBLive" will give lots of reading.

From my personal experience I can tell that with my old KT333-based computer, sound tearing occured (SBLive) when there was a read/write operation on my RAID array. Using VCool (known program to put those XPs into HLT with VIA boards) made it even worse. After fiddling with latency patches it started working ok.

In the latest BIOS there is a separate control for enabling the HLT command. I haven't tested (=reinstalled) after flashing the BIOS so I can't tell You whether it would work fine without playing with the patches.

I've also heard that because old XP boards took their CPU power from the 5V line that would have caused some PSU/VR damage. Again, I can tell from personal experience that whenever the CPU goes to full load, all fans in the computer speed up. This is because the board takes up much more current in the 5V rail that it messes the 12V regulation a bit (this is with an 350W Enermax, BTW). After two years, both the motherboard and the PSU are still alive.

And no, I'm not picking a fight.

Cheers,

Jan

sthayashi · Post by **sthayashi** » Thu Oct 21, 2004 7:21 am

Jan Kivar wrote:I've read that the reason not to have HLT in early VIA boards is that the PCI implementation in the (early? read the fine print) VIA chipsets is so poor. Searching for "kt133 686B SBLive" will give lots of reading.

From my personal experience I can tell that with my old KT333-based computer, sound tearing occured (SBLive) when there was a read/write operation on my RAID array. Using VCool (known program to put those XPs into HLT with VIA boards) made it even worse. After fiddling with latency patches it started working ok.

In the latest BIOS there is a separate control for enabling the HLT command. I haven't tested (=reinstalled) after flashing the BIOS so I can't tell You whether it would work fine without playing with the patches.

I've also heard that because old XP boards took their CPU power from the 5V line that would have caused some PSU/VR damage. Again, I can tell from personal experience that whenever the CPU goes to full load, all fans in the computer speed up. This is because the board takes up much more current in the 5V rail that it messes the 12V regulation a bit (this is with an 350W Enermax, BTW). After two years, both the motherboard and the PSU are still alive.

Holy S**t, you may have just given me the explanation I've been looking for these past couple years. I own a Stereo-Link, and even though they claimed high audio fidelity, I couldn't understand why it made a lot of noise when I accessed my Seagate V Raid-0 array (on a dual Athlon system no less.

). I use it now on my low powered system, and there hasn't been a problem. All this time, I've looked for drivers or what not, but it never occured to me that because Athlons suck down so much 5V in such varied manners that the balance could be thrown off somehow.

Sooty · Post by **Sooty** » Thu Oct 21, 2004 8:12 am

Which of the desktop Athlon 64's are 90nm? I had a quick look on the AMD site but couldn't find specs.

mczak · Post by **mczak** » Thu Oct 21, 2004 11:34 am

Sooty wrote:Which of the desktop Athlon 64's are 90nm? I had a quick look on the AMD site but couldn't find specs.

AMD hasn't updated the tech docs yet for the 90nm parts. The 3000+, 3200+ for socket 939 are 90nm parts, and a 3500+ for socket 939 exists both as 90nm and 130nm part.

mczak

edit: actually they have upated the docs (must have been minutes ago...) especially have a look at "AMD Athlon™ 64 Processor Power and Thermal Data Sheet" all 90nm parts have 67W! TDP in contrast to the (bogus) 89W of the 130nm parts. Min P-state is still 21W though.
Linky: http://www.amd.com/us-en/Processors/Tec ... 03,00.html
(argh they removed the power consumption of the mobile cpus!!!)

Bitter Jitter · Post by **Bitter Jitter** » Thu Oct 21, 2004 12:12 pm

The motherboard voltage regulators should be able to handle the change in voltage, after all they boot from 0v to 1.55v just fine.
Maybe its the FSB changing instantly that it doesn't like?

sgtpokey · Post by **sgtpokey** » Thu Oct 21, 2004 1:18 pm

I believe by specs you can get both the XP-M or AMD 64-M to comparable levels of power consumption/heat consumption, by fiddling with your motherboard bios.

but I'd go with an XP-M for a couple of reasons (I run an HTPC with a Mobile Xp running at 1.2ghz)
* Been around longer and has more variety of Micro ATX and ATX boards [not sure how many AMD64 boards have bios support for undervolting or even support the mobile 64]
* Overall board cost is cheaper and more variety in boards
* Been around longer so more literature to help in troubleshooting (kind of important when building an HTPC, there are a lot of component combos that you may end up trying in both hardware and software and the user community is smaller than other computing audiences)
* Not sure if there's a good way to force an AMD 64 to be at a specfic speed and power consumption level, but the XP has several well-documented (Do a google search for "AMD XP Pin Mod") and easy ways to do this EVEN if your motherboard doesnt' support these things: which means you can use any old xp board lying around (or purchasable for cheap)
* The extra power of the 64 is nice if you're gaming on this machine too, but if you've already decided power consumtion and heat are factors, you'll want to be prepared to live without cutting-edge gaming (if you put a gaming-class video card & a gaming class processor speed in there, you defeat your primary goals)
* lots of replacement parts for cheap available if something goes wrong (Might as well use mature parts instead of cutting-edge parts for a project such as this)

mczak · Post by **mczak** » Thu Oct 21, 2004 2:41 pm

sgtpokey wrote:I believe by specs you can get both the XP-M or AMD 64-M to comparable levels of power consumption/heat consumption, by fiddling with your motherboard bios.

Ok, fair enough.

but I'd go with an XP-M for a couple of reasons (I run an HTPC with a Mobile Xp running at 1.2ghz)
* Been around longer and has more variety of Micro ATX and ATX boards [not sure how many AMD64 boards have bios support for undervolting or even support the mobile 64]

I agree only partly. Undervolting with Athlon-XP boards can be tricky, asus for instance generally don't support it at all (interstingly, the K8V board supports it, even though it spits out a "overvoltage error" if you undervolt too much...). However, pretty much all A64 boards support Cool 'n' quiet which is all you need (and in contrast to undervolting capability it's easy to figure out if a board supports it before you buy it). There are not many micro-atx boards out for A64 though probably. And A64 boards might not support A64 mobile, but the point is you don't really need one (in contrast for the XP-M which you need for unlocked multiplier unless you want to fiddle around with bridge modding).

* Overall board cost is cheaper and more variety in boards

Definetely true. Both solutions don't really cost a fortune though.

* Been around longer so more literature to help in troubleshooting (kind of important when building an HTPC, there are a lot of component combos that you may end up trying in both hardware and software and the user community is smaller than other computing audiences)

Someone needs to pioneer new stuff, no?

* Not sure if there's a good way to force an AMD 64 to be at a specfic speed and power consumption level, but the XP has several well-documented (Do a google search for "AMD XP Pin Mod") and easy ways to do this EVEN if your motherboard doesnt' support these things: which means you can use any old xp board lying around (or purchasable for cheap)

This is actually the biggest advantage of the A64. There are still some socket A boards which will not allow undervolting nor will they work with the tools for dynamically changing the multiplier (I believe this affects all nforce2 boards). And not everyone likes pin mods...
But you should have no trouble at all with any A64 board which supports Cool 'n' Quiet. Those are _designed_ for changing voltage/multiplier on the fly, it doesn't matter if you change to a setting which is officially supported or not.

* The extra power of the 64 is nice if you're gaming on this machine too, but if you've already decided power consumtion and heat are factors, you'll want to be prepared to live without cutting-edge gaming (if you put a gaming-class video card & a gaming class processor speed in there, you defeat your primary goals)

If it doesn't use more power, why not go for it if it is faster?

* lots of replacement parts for cheap available if something goes wrong (Might as well use mature parts instead of cutting-edge parts for a project such as this)

I'm not sure what you're planning to destroy hardware

but A64 are much harder to destroy than XP-M... you shouldn't need that many replacement parts (but yes they are cheaper for a XP-M system)

Which AMD CPU system best for lowest electricty usage?

Which AMD CPU system best for lowest electricty usage?

I am running a 1 Ghz mobile