Columns Performance per what?

[Tim S]

http://www.bit-tech.net/columns/2007...nce_per_what/1

Many manufacturers have started to use the term "Performance per Watt" in an effort to draw attention away from increasing power requirements. Tim Smalley thinks it's time these manufacturers started focusing on lower power consumption before this obsession gets out of hand.

[DougEdey]

I've been moaning about this for ages, it'll take a good 18months before we see a significant decrease in GPU power consumption.

[The_Beast]

if you get a low power consuming computer you don't get performance, if you want performance you have to buy a power hungry monster, there is no in between

[DougEdey]

AMD have proven that to be wrong, remember the 35W AM2 X2 3800+ they bought out?

At the time it was top notch.

[LeMaltor]

I did mean to ask this the other day, just how high can PSU's go? I understand about having more than one maybe if it ever came to it, but are we ever going to see like 10K watt PSUs lol

[Firehed]

This is why I had a massive eye-roll when you guys did the ~750w PSU roundup, with plans for a 1kw roundup shortly after. I mean, PCIE provides something like 3x the power that AGP did through the slot, and from the beginning we've been using connectors that double that. Now we've got a pair of even bigger connectors?

And I remember back when a 500w PSU was massive overkill...

[LeMaltor]

I think we should be more worried about our electricity bills that saving the planet, if you want to save the planet, turn your computer off and dont use it at all, as Tim Westwood would say "Go hard or go home!"

[Nexxo]

I just upgraded my ATi Radeon 9800Pro. Because my watercooling setup is finely balanced at a ceiling of about 230Watts, I did not want to get anything that would contribute significantly more heat to the system. Imagine my surprise when I found that an ATi XT1650Pro runs about 2.5 times faster than the 9800Pro, but at fewer Watts: about 45W instead of 48W. Gotta love that die shrinkage...

Yet the heatsink on the XT1650Pro is decidedly meatier than on the 9800Pro. What gives?

Marketing gives. Nobody wants to buy a GPU with a weedy-looking heatsink, because they believe it won't perform. People want big-ass cards with big-ass sinks, promising power like steroid-crazed body builders, like people want big PSUs and big radiators and big bore tubing and big pumps with listed flowrates that they simply not achieve in reality.

Let's change the script, guys. Efficiency is a good modding criterion too.

[Kipman725]

meh more transistors = more computational power = more heat

in the long term the heat disapation will increase. In the very long term it may decrease due to new tech but for now we are stuck with it.

[Ramble]

In other words, manufacturers should stop using efficiency as the only performance meter and start looking at the overall picture?

[Nexxo]

Not necessarily. The XT1650Pro has about 50 million more transistors than the 9800Pro, but runs cooler. It is a matter of building hardware to be more efficient, as well as more powerful. Nobody wants a PC 5 years from now that uses 5000 Watts.

[Tim S]

Look at the Core 2 as an example - more performance than chips manufactured on the same process node with double the TDP. It's all in the design and that proves that it can be done.

[leexgx]

thay need to add power states into the GPUs like CPUs do

like when not in game mode auto Downclock and use less power

[Bindibadgi]

Originally Posted by Nexxo

Not necessarily. The XT1650Pro has about 50 million more transistors than the 9800Pro, but runs cooler. It is a matter of building hardware to be more efficient, as well as more powerful. Nobody wants a PC 5 years from now that uses 5000 Watts.

In respect though, the 9800pro was the fastest card of the 9xxx series. You'd have to compare it to the HD2900 XT

But then again I spose 320 Stream processors is a bit more than 8 pixel shaders

[LeMaltor]

Originally Posted by leexgx

thay need to add power states into the GPUs like CPUs do

like when not in game mode auto Downclock and use less power

Dont they have that already with 2D and 3D mode?

[Gordy]

I run my main machine on single 120w psu. It runs all the games I want on a 24in display yet does need to treble my electricity bill to do it.

So they can create systems that require very little power, its just they wouldn't be able to claim its the fastest ever blah blah blah.

"For what its worth its a intel core duo laptop chip with 2gb of dd2 and a sapphire x800 gto2 with full sized hard disc and a laptop dvd writer."

[Tim S]

Originally Posted by LeMaltor

Dont they have that already with 2D and 3D mode?

They do, but does a high-end chip really need to run at 500MHz in order to drive an Aeroglass desktop smoothly?

[Mankz]

I have to say that I think with the 8800 SLi and Quad core we've about reaced the top of wattage requriments. I'm pretty sure by the next gereration or at the second for sure, the companies will be looking to make high performance products with low power requirements.

For example, a 2Ghz Core 2 Duo, 2Gb of RAM, 15", 160Gb and an X1600 GPU can all be powered by under 65W (all be it they are mobile)......

[Krikkit]

I agree completely with the sentiment Tim - before I moved to uni (and had to start paying my own power bills), I didn't give a toss about how much power my computer used, but now it's a real concern.

Looking at the new-gen GPU's, power consumption's gone mad. Intel have managed a decent effort with C2D, because it's a powerful processor for a reasonable TDP, but the graphics boys have just gone completely mad these days.

[atanum141]

Originally Posted by HardOCP

Performance per What?

bit-tech’s own Tim Smalley has posted an editorial on the use of Performance pet Watt as a standard on PC components. Mr. Smalley feels that maybe we need to consider making some changes to the system at hand.

Let’s start with the CPU manufacturers, because they have been responsible for a large percentage of any system’s power draw since the day the megahertz race started. Out of the two, Intel has been the biggest culprit of poor energy efficiency in the past when its 130nm to 90nm die shrink actually resulted in a 25 percent increase in thermal design power (TDP).

Kyle's Note: Anyone that has not been seriously considering PPW in their hardware evaluations for a while now should be questioned about their opinions overall. You can read here, here, and here where power consumption and performance heavily weighed on our conclusions. As for his tree hugger stance on thinking about the planet where PPW is concerned, well, that made me laugh. Market demand will shape the CPU and GPU future, and I know most tech companies are already fully embracing higher PPW usage scenarios. Think R700 for starters.

Comments

Performance per What?

A second comment regarding what we put on this site? Whats Kyle at H trying to do? I swear he's starting a E-Beef.

Bit & H are my two faves, but taking little diggs at Bit are starting to anoy me.