3DMarks per Watt?

[LeStoffer]

Noise is another important issue.

... and even almost crucial for some of us. I invested in a watercooled rig because it was superior to solve the heat/noise problem. Cost is a concern of course, but beyond that you can have a state of the art game computer and still having it running at very low noise levels.

Seeing that the trend of high power consumption apparently is here to stay just made me convinced that I took the right path.

 

[Xmas]

My PSU is 400W, that doesn't mean it uses 400W/h, does it? :?

No. Watt is the unit for energy expressed over one second. If you want the watt-hour figure for your PSU, you need to multiply output by 3600.

400W is 400Wh/h, or 0,4 kWh per hour, that's probably about 0.06? per hour if you live in the EU.



At least I can be sure my current computer consumes less than 65W, even when playing 3D games And it's extremely quiet.

 

[Entropy]

Seeing that the trend of high power consumption apparently is here to stay just made me convinced that I took the right path.

"The trend of high power consumption apparently is here to stay"
Hmm.
Just how far do you think the trend will go?
400W? 800W? 2kW?

I think the trend is about to be broken.
It already has in a sense - a growing number of people (particuarly for private use) choose mobile computers, paying a premium for lower performance, but with the benefit of portability and mobile use, which requires low power draw. (The ones that draw the least power are also the quietest, since they do not need whiny fans.) The marketshare of portables is growing continously, and has already broken the 50% mark for some manufacturers. Industry-wide, its heading towards 30% which is remarkable since almost all private and public administrations prefer stationary systems.

Even there though, times they are a-changing. A recent report at digitimes mentioned that all the top manufacturers of thin clients had plans to start using VIAs series of very energy efficient systems. Administrative computing just don't need 100W CPUs, and their power draw/noise is only a problem with no corresponding benefit. That they are also high cost doesn't help either obviously.

As has already been mentioned, Intel is scrapping their NetBurst roadmap, and instead will focus on processors evolving from their current notebook design (Dothan). Where AMD will go is less clear, although their current path hasn't been quite as problematic recently from a thermal point as that of Intel, they are up against the same fundamental problems.

Does anyone here really care how much power his computer uses?

You could turn the question around - just who wants high power draw computers?
A: Only the ones who need the highest possible desktop performance with no consideration for power draw and its corresponding consequences in terms of cost and noise.
I think this group is getting smaller and smaller. Generally speaking, the industry is not likely to continue to follow the priorities of a progressively smaller group, even though it is the path that has traditionally driven sales.

 

[fallguy]

The Intel P4 produces much more heat, then a similar performing Athlon. Haven't heared anyone complaining about it...

You must have not want to hear it then, because many people have got a P4E, and complained about high heat.

 

[Cryect]

Hmmm, I know several people to complain about the heat of the P4's or actually Xeons to be specific. Have a couple of dual Xeon 3.06GHz systems and they really heat up your desk area that it can be unbearable kinda when for some reason they haven't got around to turning the AC on yet for the summer (no matter what cooling system there is in your computer it still will be spitting the same amount of heat out and unfortunately it gets stuck under the desk which doesn't have lots of breathing room).

True, this is a dual processor system meaning twice the heat. But, still at current rates won't be long before a single processor core reach those levels and you will see normal people to start to complain (on the otherhand they make nice feet warmers during the winter).

 

[ET]

It's not that hard to find power requirements in reviews (especially when they are excessive), and I don't think there's need for much more than this.

There is need for more than this - for everyone who wants to build silent PCs at least.

I have no problem with the additional argument of noise, but it's just a nitpick with my post. My argument was that what matters is the absolute power requirements, since they're a threshold for usability. You've given a further example of that, but you haven't given a reason for correlating power with performance. If you believe that's useful in some way, please explain how a benchmark per watt result will help in building a silent PC.

Also, if you're talking a silent PC, even there power usage has gone up, as cooling solutions improved. The dustbuster is a good example -- it was just a bad cooling design. The 6800 takes more power than the 5800, yet is quieter. So discussing the cooling solution and its noise in reviews will likely also be good to do as a standard. But performance per watt still doesn't make sense to me.

 

[Entropy]

I have no problem with the additional argument of noise, but it's just a nitpick with my post. My argument was that what matters is the absolute power requirements, since they're a threshold for usability. You've given a further example of that, but you haven't given a reason for correlating power with performance. If you believe that's useful in some way, please explain how a benchmark per watt result will help in building a silent PC.

Also, if you're talking a silent PC, even there power usage has gone up, as cooling solutions improved. The dustbuster is a good example -- it was just a bad cooling design. The 6800 takes more power than the 5800, yet is quieter. So discussing the cooling solution and its noise in reviews will likely also be good to do as a standard. But performance per watt still doesn't make sense to me.

Performance per watt has never been more than a ballpark figure even in the areas where the term has been used a lot. As you say, absolute power requirements is also necessary data.
Performance per watt is important though - embedded, cell phones, portables and these days it's even a critical parameter for desktops, since you're power limited there as well. To a certain limited degree, performance per watt is a measure of architectural goodness, completely useless however unless the target application area is taken into consideration.

For PC purposes, it's interesting to note that in a certain interval, dropping drive voltages and frequencies will give you large improvements in power draw at relatively modest decreases in operating frequency. (Or you can go the other direction as ATI did with the 9700 -> 9800 -> 9800XT and increase frequencies at the cost of power draw.) People interested in really quiet computing often engage in a bit of this "reverse overclocking". It tends to pay large dividends at this point in time, as manufacturers of CPUs and GPUs apparently push towards the upper spectrum of frequencies (and voltages/power draw). That seems to be where they need to position their parts to get marketing advantages as price/performance is the only metric a lot of hardware oriented people seem able to grasp.

This really is backwards though - instead of having mainstream solutions that are biased in favour of the interests of the performance nerds, it would make more sense to position the parts to fit more typical users, leaving the tweaking for maximum performance to those who are interested and able. IMHO.

 

[Pete]

well so much to say but better to just let you guys live the myth...
for a silent PC, the best investment is a AC. For all the work and parts to get a PC to shut up, you can just buy a AC and run your ambient lower.

I'm not sure how an AC contributes to lower noise? Have they developed really, really quiet ACs recently? Seems to me that a few low-speed 80+mm fans would be preferable to running an AC with your PC, in terms of both noise and cost.

 

[madshi]

I have no problem with the additional argument of noise, but it's just a nitpick with my post.

You claimed that the only important piece of information would be the power requirements. And I said that I don't agree. Why is that nitpicking?

performance per watt still doesn't make sense to me.

Well, I can agree that currently it doesn't really make much sense to do 3dMark per Watt benchmarks. That was more or less a teaser, nothing more. What I'm mainly after is to read in the reviews how much power each graphics card consumes (idle/load). The only review of 6800/x800 which contained this information was from techreport. All other reviews more or less ignored it.

Also, if you're talking a silent PC, even there power usage has gone up

Heh! Actually thanks to low-k the x800 has a lower power consumption than the previous generation's high end cards (9800xt/5950U). See what I mean? A lot of people don't know that. E.g. lots of websites say the cooler "whatever" can cool a 9800, but not a x800 because of increased power consumption. *Wrong*.

as cooling solutions improved. The dustbuster is a good example -- it was just a bad cooling design. The 6800 takes more power than the 5800, yet is quieter.

The noise of the graphics card cooler is only one of two pieces of the low-noise puzzle when talking about graphics cards. The other one is the power consumption. Look: If you put a passive Zalman Heatpipe on the 6800, the card itself is totally silent. But the air in the case gets very hot and thus the case/PSU fans have to work harder (= make more noise). So even with the best graphics card cooler it's still quite important to look for low power consumption, if you want to build a silent PC.

 

[ET]

You claimed that the only important piece of information would be the power requirements. And I said that I don't agree. Why is that nitpicking?

It just seemed to me that you were trying to argue the main point (per watt benchmarks) by picking at my short sighted assertion. That's why I felt it was a nit pick. I might have mis-read what you were trying to say.

As is, I quite agree with you.

 

[madshi]

Interesting new article about power consumption:

http://www.xbitlabs.com/articles/video/display/ati-powercons.html

 

[Pete]

Kudos to ATi for offering what appears to be very low idle power draw. I'm looking forward to the nV cards, in particular the 5700U and 6800GT.

 

[madshi]

Kudos to ATi for offering what appears to be very low idle power draw. I'm looking forward to the nV cards, in particular the 5700U and 6800GT.

Agreed and agreed. I'm curious to see the 6800GT results, too, since that card seems to have the best price/speed relation in the moment...

 

[Dave Baumann]

Mmmmm, now, I wonder if there is any correllation with R420's reduction in idle pwer draw in relation of 9800 and this!

 

[madshi]

You mean they tried to save engineering time by more or less fusing the mobile and desktop parts?

 

[Pete]

Mmmmm, now, I wonder if there is any correllation with R420's reduction in idle pwer draw in relation of 9800 and this!

Poppycock!

If that MR9800 is a unique die (which it almost must be), its power consumption must be impressively low when idle--I'm guessing along the lines of a 9600XT. Does ATi put out power figures for its cards?

madshi, that's one way to look at it, considering how all PCs are just getting smaller (DTR notebooks, SFF desktops, 1U server blades), and thus heat tolerances are getting tighter. I hope we see desktop Dothans soon to combat A64's Cool n Quiet features. The sooner PC parts makers focus on increased power efficiency and thus cooling and noise reduction, the better for everyone, IMO. I don't care if they initially market it as a luxury feature, along with variable-speed HSFs on the high-end cards, as it'll eventually work its way down to the whole product line. It would apear that GPU makers have already made the transition from the relatively power-ignorant R200/R300 and NV25 to the forcedly power-conscious RV360/R420 and NV30.

Enabling idle power saving features may serve to extend a GPU's life, too, no? The difference between 20W and 65W, or 15W and 55W, is huge.

 

[Dave Baumann]

If that MR9800 is a unique die (which it almost must be), its power consumption must be impressively low when idle--I'm guessing along the lines of a 9600XT. Does ATi put out power figures for its cards?

Uhh, Pete, did you read the rest of that thread?

 

[Pete]

I did and was going to post an apology for not doing so before replying, Dave. I should've guessed that they could use the same die, given how much less power the X800P uses compared to the X800XT via what appears to be a simple quad disabling.

Double check about you trying to finagle power figures out of them.

This is one more reason to believe that GDDR3, not the R420, is limiting supplies of X800Ps and XTPEs, right?

 

[Bjorn]

This is one more reason to believe that GDDR3, not the R420, is limiting supplies of X800Ps and XTPEs, right?

The mobile 9800 uses 8 pipes so wouldn't it rather be the opposite then ? (= Ati has a lot of 8 pipe R420's)

 

[Pete]

I was thinking more along the lines of they had R420s to spare to be used as 8-pipe MR9800s, but it's certainly possible they're seeing more imperfect R420s than I thought (due to the 100% X800P VIVO -> X800XT soft-mod).

Ugh, no more prognosticating for me.