Slashdot Mirror
AMD's Turion 64 on the Desktop
Toasty16 writes "SPCR has an overview of using an AMD Turion 64 mobile processor in a desktop system. There's a good bit of info about motherboard compatibility and power consumption as compared to a Pentium M processor. There's also links to articles from the Techreport and LaptopLogic on the same topic. If you've been thinking about building a low power HTPC or file server, mobile processor on desktop is an interesting option."
I have used fanless Via C3 chips for several years now. One is running a Linux Fileserver at home, the other a DVB multi-tuner PVR.
Last I saw they are on the C7 chip. Not so famous as AMD, but for certain tasks, get the job done nicely.
- Paul
I have been using a turion based PC for a little while. For what I do (no gaming) it has been brilliant. Its good to see that there are other people writing articles about using Mobile chips on their desktop.
I've been using Mobile Athlon XPs for a couple years now. Picked it up on the premise that they ran cooler, on lower voltages, and didn't have a multiplier cap. Worked wonders for hitting an 800MHz overclock on air (2.0 to 2.8). They also seem to work in all the mobos i tried, although some needed a BIOS flash.
Honesty may be the best policy, but by process of elimination, dishonesty is the second best policy.
I love the idea of low power, high-concept CPUs on desktop machines.
m l
I love the fact that these chips are 754 pin and thus compatible with an existing socket- motherboard manufacturers won't have to ramp up any new hardware to start selling boards for these in every color of the rainbow.
Turion Shuttles/insert small FF MB here/, anyone?
and there are loads of legacy boards available. Socket 754 boards are dirt cheap. here's a handy list of compatibles from the article:
http://angelfall.s39.xrea.com/area2ch/turion-e.ht
I would have jumped all over the P M, except there was no desktop gear for it; unless I bought a notebook PC and did some expensive hacking, which, ad publicae geekio, is a contradiction in terms.
score +one for AMD.
Speaking just to the power-savings benefit of using a mobile CPU in a home system, unless you are running a home server, the best way to conserve power in any PC would be to turn it off.
This link notes one person's cost of leaving systems powered up, a little more than $23.00 USD per month.
So boot it up each morning, make the coffee and toast, and by the time you sit back down in front of Unbuntu, Mandriva, BSD, SUSE or that friendly chair-throwing group from Redmond, you can feel better about not burning up more oil to play solitaire (YOU'RE FIRED).
"Let us raise a standard to which the wise and honest can repair" - George Washington
I've had a great experience with Turion 64 chips in a laptop. High frame rate on graphics-intensive applications and, in general, good responsiveness even when running a whole boat load of RAM and CPU-heavy apps like Eclipse and Server JVM. I've used these chips from the ML-26 to the ML-44, and the cost/benefit analysis of AMD Turions versus the alternative just makes more sense. For the dollar, it seems like I can get 30% more performance in the apps I care to run.
But, Turion 64 on a desktop, not quite so fast, if performance is important to you, why go to all the trouble to install a mobile CPU? Either turn the thing off at night or drive less.
------ Tim O'Brien
Any chance of getting some part numbers or manufacturers for that system?
The need for cooler running desktops has surpassed the need for faster desktops in the case of most of my customers. They like to browse the internet, write email, and play bridge online, but rarely render anything in 3D, encode much media, or play any games. They'd rather stuff the computer in a cabinet and not have to stare at it.
In the latter part of last year, we replaced three Dells that cooked themselves inside a cabinet, and have at least three more where the customer complains of frequent crashes. We're almost certain that heat is the issue. All Dell will do is send us new case fans.
We've tried cutting vent holes in the desks too. Short of an active fan based ventilation system, it appears that a fast system is not suited for life inside an enclosed cabinet. Enter the need for something not quite as fast, but cooler and quieter.
Modern video cards accelerate a decent portion of MPEG-2 playback, but you still need a decent amount of CPU.
I think with the most recent video cards, it's something like 50/50.
Note: I'm not counting hardware resolution scaling here. Output scaling is one aspect of video playback that is historically EXTREMELY CPU-hungry, but has been supported in hardware on any video card made in the past decade or more. Even with hardware scaling, you need a 2-3 GHz+ CPU to play back high def MPEG-2, and additional HW acceleration (IDCT, MoComp) offloads 20-30% at best. VIA's video chipsets offload much more of MPEG-2 playback than most other video cards, but until the CN400 series, they were only able to offload standard def content. (90% of hardware MPEG decoders on the market only support MPEG-2 MP@ML, i.e. standard def content. MP@HL decoders for high def content are rare and expensive.)
retrorocket.o not found, launch anyway?
How about a stack of twenty PIII 500s with CPUs from eBay for thirty bucks in shipping. File servers don't need Ghz plus CPUs. Slot 1 PIIIs are fanless too. If I recall the PIIIs are about thirty watts which is actually less than the PIIs.
l
Here's a CPU wattage comparison list.
http://www.silentpcreview.com/article31-page1.htm
I was running two very old machines for fileservering and routing/firewall duties. My estimate put these machines at about ~$30-40/mo to run depending on what I was doing. I was able to drop this down by about half moving to a mini ITX board with the via C3 processor, and it only cost about $150. I could have spent less, but I upgraded the power supplies as well.
You can easily measure how much power your computer draws with a multimeter from the hardware store - last time I was there I saw them for about $10. Put the meter on the AC amps scale, make sure the wires are plugged into the amp reading ports, and then wire it in series with your computer.
I guarantee you'll be suprised. I was.
..don't panic
Here is the problem as I understand it. Things should turn on and off based on various conditions.
This problem matches the fundamental function of digital electronics. Your contention that I should do something is silly. The task at hand is easy, yet tedious. You described a design flaw that has an obvious electronic solution. A machine should do it.
I have a Mobile AthlonXP 2600+ in my desktop machine and that think still overheats with a 40usd heatsink/fan on it.
I very much doubt that the Turion in a desktop is anywhere near as low power as the Pentium-M, certainly doesn't seem to be from the laptop reviews I've seen.
I'd say for a HTPC, you'd have to go VIA.
#include <sig.h>
I was wondering why SPCR wasn't working.
Bittorrent..
Most switching power supplies are very efficient and take power factor into consideration. There are some high-end units that approach ~0.99+. I would imagine most to be around ~0.85.
:)
My point being for $10 you can get an acceptable measure of your computer's power draw. Do be careful, though.
..don't panic
They placed a huge hunk of copper with a large noisy fan on that chip. That doesn't make for a Silent PC.
I would have liked to see some testing with a fanless heat sink.
I've got a 800Mhz via mini-itx board that i use for a mythbox. I didn't buy it for low-power, i bought it for silent operation. Nobody wants a noisy PC in their lounge when they're chatting with friends. And your friends certainly dont appreciate the noise the next morning sleeping in the lounge with a hangover.
The C3 chip is pleanty powerful at the moment. With DVD-B in the UK the TV is transmitted as an MPEG2 stream, so there's no encoding required. Playing back divx movies with no hardware acceleartion (unless you count basic video overlay and scaling (XVideo)) leaves the CPU 50% idle. I've yet to try using XvMC to utilise the north bridge/grahpics chips hardware decoding assitance for DVDs etc.
However something more powerfull will be needed one day for h.264 etc, and maybe high-def TV if that ever arrives in the UK or NZ.
A 1Ghz Ultra Low Voltage Pentium-M would be nice for fanless operation, but the CPU alone would cost more than my entire setup (TV included). And is 1Ghz pentiumM going to be enough of an improvement over an 800Mhz C3? Remember the regular pentium-Ms are 2Ghz+ now. An underclocked cheap Turion might be just the ticket if a suitable mini-itx MB exists.
At last I will be able to buy a powersaving chip that doesn't come from Intel's Israeli labs. Divest in Israel, the world's only terrorist state!
Deep Thought was wrong! The answer isn't 42! It's 52.1! (rounded off, of course, no need to waste signifigant digits on the answer to life, the universe, and everything, is there? By golly no!!)
Installing and running cpufreq is relatively easy and the savings are considerable. For newbie linux users I have an explanatory step-by-step post http://pkt3141592.blogspot.com/2005/07/fun-with-li nux-cpufreq-driver.html on the subject in my (almost abandoned) blog.
Running a Turion is a hard-core option, but PowerNow should be enabled in ALL Athlon64 desktops.
P.
What would be interesting is if someone would build into the next generation of graphics cards a power-saving mode, much like AMD's badly-named "Cool-n-Quiet," so it consumes much less power when not needed. This would be especially beneficial for machines that aren't used for 3D gaming (many home computers and most business machines).
I usually have my machine at home on while I'm at work, and while I sleep, as it's generally downloading something from somewhere; turning it off isn't usually an option.
I went out and bought my Athlon 64 3200+ when they first came out. It's the original Socket 754 variet. AMD has long since moved on from that socket, and have stopped upgrading those chips.
So, the Socket 754 mobile chips might be another upgrade path for those of us that don't want to swap out the whole motherboard.
They would have the added benefit of lower power/heat characteristics. This would be nice, since I'm using my current A64 in a MythTV HTPC.
Without having RTFA or the replies, I present the above question. Venice doesn't consume more than 30 watts from what I read. See this thread: http://forums.anandtech.com/messageview.aspx?catid =29&threadid=1780053&enterthread=y
This already exists. Modern Nvidia cards reduce their core clcok speed and voltage when not doing 3D, and ATI cards reduce both their core and memory clock & voltages when not doing 3D.
For example, my 6600 GT runs at 500 MHz during 3D games, but it clocks down to 300 MHz and a reduced voltage in 2D operation. The power consumption is reduced by half in 2D.
These are technologies adapted from their mobile chips, although they're not as aggressive as the power saving modes of those mobile chips. I do wonder, however, what we are going to do once Vista starts stressing the 3D engine full-time.
Man is the animal that laughs.
And occasionally whores for Karma.
I think it'd be interesting to see a really aggressive version of that. Does the computer know when the monitor is off? If it does, let the graphics card power down completely. If your computer is downloading something overnight or whatever, just shut the whole graphics bit down and let the rest of the computer do its thing. This may be possible with other devices in the machine, too - audio card off if speakers are off and no headphones plugged in, etc. We've definitely not seen a big push for energy efficient desktop machines, yet, so the possibilities are numerous.
The only downside is that your old motherboard is incapable of doing the powernow throttling for the turion chips, according to TFA.
The DXR3 is a fun card to play with. Hardware accelerated MPEG-2 playback.
and under Windows, I'm quite happy with crystalcpuid http://crystalmark.info/?lang=en
m l
It also happened that my athlon64 3000+ is still stable when going under the default lowest consumption state (1GHz at 1.1V). It works fine at 800MHz and 0.8V
a nice guide here: http://www.silentpcreview.com/article231-page1.ht
The reason these chips are pin-for-pin compatible is because they aren't any different. Turion is just a marketing name. You're getting a regular Venice (or whatever) core and so of course it comes in the regular package.
I do think making desktops from laptop processors is a good idea, but in this case, you're not actually doing anything different, just using different words for the same thing.
Intel currently does have separate desktop (P4) and laptop (P-M/Core Duo) processors, but they're about to switch to all laptop-derived processors with Merom or Conroe or whatever it is called.
http://lkml.org/lkml/2005/8/20/95
Yes, well, you're an idiot for being so sure, even though you're COMPLETELY WRONG. Sorry, but that's the way it is.
Many videocards now have MPEG-2 decoding built-in, but that's not computing-free at all... For instance, a 300MHz system wouldn't be able to playback 1080 MPEG-2 video, even with XVMC hardware acceleration.
Plus, as I've been saying a lot lately, there is a cross-over point, where hardware acceleration becomes counter-productive. With a system that is more than about a 3200+ or so (AMD=2GHz P4=3.2Ghz) you can playback 1080 MPEG-2 video faster in software, than you can with hardware acceleration thanks to overhead of XVMC, interrupts, and many other things like that.
Finally, I don't know of any cards that can decode WMV9 and H.264, which are most widely used for HD content. Those are the most CPU-intensive codecs, and there's no hardware that will handle them.
NO NO NO NO NO!!! The blue-screen is called an overlay. It's used because the uncompressed video (WHICH HAS BEEN DECODED BY SOFTWARE) is transferred directly to the videocard (via DMA) rather than being routed through X11's standard (slow) output methods.
The decoding is still done entirely on your CPU.
No such thing. Hardware acceleration has to be designed-in for EACH CODEC you wish to use, which is vastly impractical. So MPEG-2 is the only widely-supported codec.
Also not entirely true. The overhead with hardware decoding is not tiny. And, with a reasonably fast system, software decoding is FASTER.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
If you're looking for low power firewall machine, you should really look at Soekris: http://soekris.com/
They are fantastic small machines/boards that are perfect for that kind of job and they works great with *BSD and Linux.
They offload some of the processing, but they'll still take up a hefty chunk of CPU.
I'm talking about 100% hardware MPEG decoders that take an MPEG stream in and give video out, such as the decoder on MyHD MDP-1x0 series cards. (Unfortunately not supported under Linux.) The only CPU those will use is that required to shift a stream from your hard drive to the card.
retrorocket.o not found, launch anyway?
No such thing. Hardware acceleration has to be designed-in for EACH CODEC you wish to use, which is vastly impractical. So MPEG-2 is the only widely-supported codec.
i hate to break this to you, but MPEG-4 has Also been supported widely in better video cards for the past 4+ years.
i mean you were having such a fun time disecting gps post..
and even though you've made many assertions, the CPU load for mpeg-2 and mpeg-4 streams when being decoded with hardware acceleration is roughly a 50% reduction. a CPU that is running at 50% of load uses a lot less electricity and cranks out a lot less heat than one at 100% load.
Since the graphic card are so highly optimizied for processing video, the portion of the decoding process they are allocated to perform (reguardless of your assertions they Are performing computations on the video needed for final display. to a tune of reducing the load on the cpu by 50%) doesn't even cause them to break a sweat, which is why laptops with what passes for full mpeg-2/4 on-board graphic acceleration can add more than 45 minutes of 'battery' life, while playing back video.
now, could hardware acceleration for video be handled better? Yeah, It could. Considering the fact that GPUs come with 300+ million transistors Custom tuned for video processing, there is little reason why a modern vpu couldn't handle 100% of the video decoding computations except that that's not the way the specs were designed etc. why would i WANT to run those computations on my 30 million transistor CPU?
The GPUs in VIA systems for the past couple years have done MPEG-4 decoding, because they really, really had to (terribly under-powered CPU)... It was just recently that ATI began including MPEG-4 support, and I'm sure NVidia hasn't been doing it for very long either.
No, definately not widely supported. Sure, you can find a graphics card that will do it, but it's certainly not common at all.
Now that is just a moronic assertion. First of all, unless we're talking about a specific CPU, there's no way you can say it will use-up 50% less time... Is that 50% on a 300MHz system, or 50% of a 4GHz dual-core CPU?
Besides, as I've said before, the performance improvement is directly proportional to the speed of the CPU. With about a 2GHz Athlon, software decoding is FASTER, and takes less CPU time than XVMC (hardware decoding). You can do the benchmarks yourself, or you can ask someone else who knows this stuff even better than me, like Ivor Hewitt, the main man behind openchrome.
Also a terrible assertion to make. Read my former journal entry about S2K Bus Disconnect. There are many 32-bit AMD CPUs out there which barely change their power consumption and heat during idle vs. load. Newer motherboards support S2K now, but older ones do not, so that's not always the case.
A hardware decoding card will do much of the decoding, but that's not how "all videocards" work, as the parent claimed (I think it's safe to assume you're actually the parent, posting as an AC, but that doesn't really matter right now...). Even with videocard cards that support hardware acceleration, most people aren't using that functionality at all, for numerous reasons.
The overlay the parent was talking about isn't a matter of number-crunching being easier for the GPU rather than the CPU. It's just a case of an overlay being trivially easy to scale, and not going through the standard display calls saves time. You could have nearly as good performance without hardware scaling though, by just displaying it normal sized, and instead of zooming, just changing the videomode to get the video to fullscreen.
I can think of many, many reasons. First of all for future compatibility... An ASIC isn't a general purpose CPU, so it can't be upgraded when the next video codec comes out, or a current one gets an improvement. Second, because it's a tremendous waste of money to include all the major codecs in a videocard. You'd be better off spending that money on a larger battery, a much more effecient CPU, etc. With that better, faster CPU, you can also do things OTHER than video decoding, such as video ENCODING for instance... MPEG-2 acceleration is common just because it's so easy to do.
Also because your videocard can't do postprocessing, good deinterlacing or inverse telecine. Because interlacing and telecining methods change over the years, and software can keep-up with them, but hardware can't.
Plus, because software decoding can actually be faster... Even with full hardware offload, you are spending CPU time on interrupts transfering the video from the hard drive into RAM, then to the GPU. Interrupts are something PCs have a TERRIBLE time with, as opposed to number-crunching, which it has a very easy time with. At s
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant