I would get a s939 ATX motherboard, probably a non-NForce one if you want low heat/power/noice. Most of them will let you undervolt from the BIOS, no need for crystalcpuid.

Most matx boards don't allow undervolting, so you'll probably want to stick with full ATX.

But I think you'd be surprised how cool, low power, and quiet a normally volted Venice 3000+ would be. If you're building a machine for a home server, or even light gaming, I would get one of the new NVidia 6100 mATX boards. All passively cooled (at least the one I got is) and no need for a video card, it has decent onboard graphics. It'll be very low power, cool running, etc... and it'll be dirt cheap. with cool'n'quiet it'll clock down to like 1.1GHz most of the time.

unfortunately i am still using a agp card, x800xtpe. as far as i know, all the nf4 board are pci-e...

just out of curiosity, can it go lower than 1.1? i know my wifes laptop with a p-m 1.6 will go down to 600MHz...

jvrobert: Wrong, VIA should be avoided if any.

bob4432: Get this mobo. Low cost, good overclocker and both AGP and PCIE that works for real!

Bob, I use a Abit KD7 and AXP 1700+ JIUHB core which is rated for 1.6vcore. It can actually do 2ghz at 1.6vcore, and I run it at 1166mhz when it is at 1.1vcore. As I do mostly only browsing and such, I do not need additional speed most of the time so I do not change clock on the fly in windows, however I will reboot and go into the bios if I ever decide I need a little cpu boost once in a while for cpu intensive tasks. The cpu can actually do 1.2ghz at 1.1vcore, however due to FSB and also just to be a little more conservative I run 166x7 rather than 133x9, as I prefer not to run PCI/AGP out of spec and there is no lock option on my motherboard. My AXP is unlocked as you can see, and most are factory unlocked or can be unlocked with conductive laquer (spelling?). The KD7 is really a great motherboard for its time and has the best vcore range I've ever seen on a motherboard (not that I need it) ranging from 1.1vcore to 2.325vcore. If it had vcore lower than 1.1 I would try to use it.

Edit: I should also mention I have a AC power measuring tool and when my cpu is set to 1.1vcore at 1166mhz my entire tower consumes only 60w. However when I boost the clock (and more importantly, the vcore) consumption IIRC will go well over 100w of course, depending on the exact speed and vcore I set. For conserving power, undervolting the cpu can make a difference. At the same time using a high end graphics card or a CRT monitor is counter productive for low power, and I suggest to get a good quality psu like a seasonic for its efficiency. I do use a CRT because I like the image quality of a good sony CRT and I'm too cheap to fork over the cash for a 21 inch equilavent LCD with inferior image quality. I use a Radeon 9000pro for graphics which is easily passively cooled, I just unplug the fan and no other special cooling arrangement. I assume it is low power graphics.

-Ken

thanks for the info, i am very familiar with that board and have been eyeing since it came out. not too worried about o/c, although with a 3000Venice i would have too , but underclocking/undervolting. anybody have one that can verify that this one can u/c and use the cnq to throttle down?

appreciate the response

Pardon my lack of knowledge, but can the range of vcore and clock speeds that CnQ uses be set?

-Ken

I too have been doing a great deal of research in the Turion 64 for a low power, low heat HTPC. After an exhaustive search I've finally came across a Japanese site that lists compatible (either fully or partially) and incompatible motherboards.

http://angelfall.s39.xrea.com/area2ch/turion.html

Bookmark the page and check back from time to time for updates.

DFI has three motherboards that officially supports Turion 64, at least according to their own site.

http://us.dfi.com.tw/Support/mb_cpu_sup ... ll&SITE=NA

excellent find, wonder when they posted that because on 11/09/2005 i got this response:

Turions are mobile processors and not supported on desktop motherboards. Thank you.

Please use *Reply All* option when communicating with tech support.

Sincerely,
Robert AKA RGone...
robert@dfiweb.com
rgone@dfi-street.com "forum addy"
www.dfi-street.com

Very nice! There's an english language version:
http://angelfall.s39.xrea.com/area2ch/turion-e.html

** BUMP **

In case anyone is looking for a Turion compatible MB.

So my Asus K8V-MX should be okay, by the look of it as it has the same VIA K8M800 chipset as a few of the other mobos listed.

It's not about the chipset, but the BIOS.

I read recently a post (don't remember where, though) from a mobo vendor rep that AMD is now telling the mobo vendors that they can't do anything to officially or unofficially support turion 64s on desktop boards or AMD will prevent them from being able to develop, sell and manufacture mobos for AMD cpus. (ie revoke their IP license, stop giving them volume discounts or something of that nature).

I think AMD has been threatening that for a while, but maybe they're finally getting serious about it.

I'd assume this is a large part of why they're moving to a new socket (Socket S1) for mobile CPUs.

No I don't think so. I think it more of an issue related to "grey market" CPU where a lot of OEM CPUs have been sold to consumers by the "grey market middle man." Those OEM CPUs were intended to go to the likes of Dell, HP, etc. The reason for cleaning this mess is to insure that no vendor will sell altered CPUs as well as to level the pricing field, and if one of these OEM CPUs were to die the blame would inadertantly land on AMD, instead of the seller.

I haven't fully thought this out yet, but I think even though the Turion 64 is sold in retail boxes, the intended target is the mobile market which is why there is no HSF. If AMD were to officially offer support for the Turion 64 then a HSF has to be included, otherwise a neophyte would install an inappropriate HSF, or worse, assume that it is not necessary.

Personally, I would like a m-ATX MB to have official support for the Turion as it would make a great CPU for a HTPC. It may or may not produce more heat than the Pentium M, but it is definitely less expensive to build. Forget the price of the Pentium M. Have you seen the price for a Pentium M motherboard?

Asus CT-479 and a P4P800-VM is not really expensive.

The Asus CT-479 is a add-on card to use a Pentium M CPU in a Pentium 4 motherboard like the P4P800-VM. There is a catch though, IIRC, speed-step cannot be activated. Therefore, the CPU will run at full speed all the time.

Not quite energy efficient.

The cheapest "real" Pentium M motherboard the does support speed step is the AOpen i915Ga-HFS $220 (very mixed reviews) and the AOpen i915GMm-HFS which has better review but costs $285.

Still, it would be very low power when idle, 10 W?
Doesn't CrystalCPUID work either? I know that it's one of the few mobo's that you actually can change multiplier in BIOS, that's a good sign.

Don't forget about MSI 915GM, $200 here in Europe.

Have you seen any more reviews of Aopen 915GA besides the one VR-Zone made? I'm still curious about that strange PCIE slot.
The PCIE slot is placed in the same way on the MSI mobo, what's going on?

Only the customers' short review (complaints) about it at Newegg.com.

For now I'm not going to bother with upgrading/rebuilding my HTPC. My primary rig takes precedence now and will probably be quite expensive since I am considering the A64 X2 4600+, 4800+ or the soon to be released 5000+. Of course I will need to replace my motherboard; I'm pretty sure I can't fit an Athlon 64 in my Socket A mobo. Then again, isn't it the purpose of a hammer to make things fit?

By the time I get around to upgrading my HTPC next year, I will be looking for a m-ATX for the "new" socket 479 Memron, or the Socket 1 Turion replacement.

Not sure if people are still looking towards the Turion as an alternative CPU. But I came across some additional information that may help to ultimately decide whether or not the Turion is worth the trouble.

GamePC has an old review of the Turion MT-34 (1.8GHz). They used a DFI nForce3 250GB motherboard as a testbed, and the Seasonic Power Angel to measure wattage.

They compared the Turion to the desktop Athlon 64 S754 130nm, and the Pentium M. Forget about the Pentium M. The Turion MT-34 (1.8GHz) matches up against the Athlon 64 3000+ S754 130nm (1.8GHz) in terms of benchmark performance.

Based on the components used in the review total system power consumption is 97w when the Turion is idle and under load it consumes 128w. The S754 Athlon 64 3000+ consumes 98w when idle and 149w under load.

The Turion MT-34 only consumes 1w less than the Athlon 64 when idle, and 21w under load.

Okay, those numbers are based on the older Newcastle Athlon 64s which uses the 130nm process. The current generation of Athlon 64 uses the 90nm process. That is significant decrease in size of about 30%.

I would speculate that a 90nm S754 or S939 would cause total system power consumption to drop conservatively to about 143w under load (using the same components of course). Thus, the Turion MT-34 would only consume 1w less when idle and about 15w less under load than the 90nm Athlon 64 desktop equivalent.

In my opinion a hypethetical decrease of only 15w in a desktop environment is not worth the restrictions that a Turion 64 would impose in a desktop environment. A Turion 64 MT-34 sells for around $190, the 90nm S939 Athlon 3000+ is around $150, and the 90nm S754 Athlon 3000+ is around $125. The marginal increase in electricity expense can be partially offset by the lower price of a desktop Athlon 64 CPU.

Please note that in the game PC, they test the 25w Turion64 at 35w voltage. The CPU should be at 1.2V, not 1.35V. Also, some of their performance benchmarks just don't seem to add up. IMHO the GamePC just seems kind of fishy. I emailed them a time or two about the incorrect voltage. They did not reply, and do not appear to have added anything to have correct it in the review, which leads me to believe the worse.

We're working on an article that examines several Turions in a couple of different 754 desktop boards including one that supports it 100% (even CnQ, etc). These will be compared to a Pentium M 2.13GHz + AOpen 915 board.

A quick peek:

W/ one drive, integrated graphics and a 512mb stick of RAM, an ML44 (2.4GHz, 1mb cache, 35W part) system idles at 45W (total AC power draw) w/CnQ & ~90W at 100% CPU load.

The closest desktop A64 runs >15W more in idle and something like 30~40W higher at full load.

It's not always about "value for money" -- it's about a priority on low power consumption & low heat -- without taking a big hit on performance or price. (I.E.: Keep buying gas guzzlers now because they are cheaper... and more advanced efficient cars will never get cheaper or better or come soon enough to avert an escalation of energy depletion & its attendent ills.)

MikeC: I guess it's a bit hard to decide how to perform a test like that. Testing the components as they are out of the box is never wrong, however most of us are interested in what a little bit of tweaking can bring. Of course I'm talking about undervolting. The negative aspect of undervolting in a test is that the results are individual for each CPU. On the other hand it could be just as individual as the acoustics of HD's.

1. What's your opinion about underclocking desktop and mobile AMD CPU's? Have you noticed any difference? If comparing a desktop and a mobile CPU with the same stock Vcore (like T64ML and some A64) the mobile one probably wins the low power race. But for some reason I have this idea about desktop CPU's having more underclocking headroom than the mobile counterparts.

2. Does the individual TDP of newer AMD CPU's mean anything in real world usage? Will I see notice difference in heat output between a 50 W and a 80 W CPU? (same OPN) It feels like I'm the only one who wants to know this, I haven't seen any kind of discussion about it here at SPCR, surprisingly. See here, last posts. Just the fact that there are standard Opteron 175 (2.2 GHz, dual core, 2x1 MB L2 cache) available with a 35 W TDP (they are documented by AMD, not talking about low power "EE" or "HE" models) would be interesting enough, but so far I've been treated as a daydreamer when I've mentioned it here in the forums. I just don't know what to believe. The individual TDP could be something important when doing some tests with AMD CPU's, and maybe preferrable than the max TDP. The problem is that you need to use the CPU to find out the TDP, so there's no chance you can cherrypick one when buying. Here's the software if anyone is interested.

3. Do you think thermal probes on the CPU heatsink could be a good way to measure the heat output? The other two ways doen't seem reliable to me. Measuring the power input on different mobos gives different results, and the temps reported from the mobo seems pretty different on every model.

If these questions will be answered in the article then I'll wait for it.

Mats --

1) Can't really say on this one. It's not a focus of the article; we're looking at things that don't require any more hacking or tweaking than absolutely necessary. IE, all the processors are examined w/CnQ, full load at stock voltage, and/or CrystalCPUID - a bit of the last. But I suppose we could try one simple comparison, just for you.

2) It means something, but not to compare a 3500+ to a 3800+, per se. My take on their TDP is to take the highest clocked model in a range (where all are rated at the same power like 67W), then use that model as the baseline and estimate real max power for the slower models in proportion to clock speed. A 10% drop in clock speed = 10% drop in power.

3) No. Too many variables affect temp, not just pure power dissipation. The power measurement is much more accurate.

I agree, and that's the classical way TDP's are determined, and still is for the max TDP of the CPU's I mentioned. But the new individual TDP seems a bit more complicated. AMD's new system looks both good and bad to me. The good thing is that you can get an accurate TDP for your CPU. The bad thing is that you don't know it until you've bought it (not really worse than before though).

Like the Opteron 175 I mentioned, if you're lucky you'll get a TDP = 35 W, but in worst case it's 110 W (according to thermal profile I on page 11, which is used for the 175, page 37). I wonder if it's possible to see such differences in power draw too. Now I need two identical processors to try it out, which I never have...

Mats, sorry for not reading through all your links. I see what you're saying. Will have to check that little utility & read the rest of the AMD docs before I can give you any useful comment. Certainly looks interesting.

www.xtremesystems.org seems to be offline, btw.

Hey, isn't that a little misleading? I would want TDP to tell me the maximum power that the CPU could possibly dissipate under worst-case conditions, not the "well if you stand on one leg and undervolt the CPU you might get 35W" TDP.

I think the two chip companies should also introduce an ADP (Average Dissipation Power). This should represent "if you use our processor at stock voltage and at stock speed, it will dissipate X amount of power at an ambient temp of 20 degrees C when running [insert program that moderately loads the CPU here]". I know everyone's setup is different, so practically no-one will actually get the ADP, but it would make it much easier to choose a cool-running chip.