Graphics State of the Union 148
Tom's Hardware has put out a nice recap of where computer graphics have been and where they are headed in the near future. While there are some definite shiny toys being displayed in new product releases and on the test beds, the overall problem of power consumption continues to rear its ugly head demanding attention. From the article: "while all of these things are interesting, exciting and new, the problem remains the same. Getting smaller and faster only makes sense if the design also is less demanding on the wall socket and cooling system. We all want different things when it comes to advancements, but first and foremost we need better power management. The bottom line is simple: graphics makers must take a step back from feature brainstorming until the power issue is resolved."
Power & Physics (Score:3, Interesting)
What AMD can bring to the ATI deal? (Score:5, Insightful)
Intel came out with a truly Power-Hungry CPU.
AMD came out with a cooler and better CPU.
Intel came out with an even cooler CPU that out performed the AMD one. (Core Duo/Core 2 Duo)
The ball is now in AMDs court.
In other words, the presure on Intel was that they had to compete in that area in order to be competitive.
Perhaps AMD, coming from their battle with Intel can help focus the ATI division on less power consumption/heat generation, and perhaps that is that AMD can help bring to the table.
If they even BEGIN to make inroads in this, while maintaining a competitive stance against Nvidia, it will force Nvidia to compete on this point also, which should move GPUs in a cooler direction
Re:What AMD can bring to the ATI deal? (Score:2, Interesting)
I stated this in the amd+ati deal, but somehow it didn't get modded :)
AMD had to go together with ATI to get _low power_ systems up. They won't make it on the processor line alone, they need to have a chipset+cpu solution (which is the most important anyway). ATI has just that, low powerconsumption (I believe their GPUs use less power too). Add to this that in the recent presentation they gave clue [dailytech.com] of what could be done to use less power. Namely the scaling of systems, this would optimise the compute
Re:What AMD can bring to the ATI deal? (Score:2)
Yes, I can see where for AMD adding ATI's chipset over Nvidia's makes sense, especially if ATI's is less power hungry.
Interesting article, and definately in line with what we were both thinking. Suddenly it looks like Nvidia and Intel might have some competition on their hands.
Older GFORCE does the trick (Score:2)
Re:Older GFORCE does the trick (Score:2, Interesting)
FUD, pure and simple.
For $30 (and less) you can buy a nearly noiseless heatsink that will cool your card better than most stock heatsinks (e.g. Zalman VF(7/9)00, Arctic Cooling Silencer line), and if you don't want to go the DIY route there are more and more cards coming out that have quality quiet heatsinks preinstalled (e.g. ICEQ3 by HIS).
Re:Older GFORCE does the trick (Score:2)
Re:Older GFORCE does the trick (Score:2)
The only thing that worries me is the amount of flex the heatsink puts on the PCB when you tighten it down all the way like it says in the manual. It flexed so much that it actually made inserting the card into the AGP slot a bit difficult...
Re:Older GFORCE does the trick (Score:2)
Re:Power & Physics (Score:2)
The games that people remember tend to be the good/original titles. ie.) Super Mario Bros. Zelda, Contra
Does anyone else have thoughts about this notion?
Wrong. (Score:5, Insightful)
Today this is irrelevant. If consumers continue to purchase ever more power hungry graphics cards, what is to stop the companies from making them? When the market actually changes and people start considering the power requirements of their cards, then I'll believe this statement about the bottom line. Because right now the only thing I hear from people building or buying new computers about the power requirements is "make sure you get a PFC PSU and get lots of watts", not "make sure you get a low-power GPU". For one thing, some people actually enjoy saying they have a 600+ watt PSU. I can imagine that with current power costs today this trend will continue. Do the math, it's not actually costing a person much more per month to go from 600 to a 1000 watt PSU, especially since most people don't use their GPU to full power most of the time.
Power requirements take a back seat to overall performance, and will continue to do so until electricity costs are driven up further. It's simple economics. People are willing to pay for the power-hungry cards. And until they're not, power consumption will continue to be less important to the producers than performance. This is analogous to today's vehicles, still being built and shipped with huge fuel sucking engines. For many people, and I'd wager to say enough to sustain the market for years to come, the cost of energy (either liquid or electrical) is still low enough that they aren't going to give up their cherished powers, be they piston driven or transistor.
TLF
Re:Wrong. (Score:5, Interesting)
And what choice do we currently have? If the companies made Watt-Friendly cards, I'm willing to bet people would buy them especially for laptops. But they don't. We don't have the choice BUT to buy these double bay, amp eating, juggernauts we have today.
Re:Wrong. (Score:5, Informative)
A lot of the newer mobile GPU (like GeForce Go) are capable of greatly reducing their overall consumption when their total demand is low. They ramp up when needed.
Of course this doesn't address the fact that, when needed, and when ramped up, they consume a lot of power. To which I say, yes, we need more power efficient cards.
This is unique to the mobile sector for now, but of course will eventually find its way into the entire realm of graphics computing.
Unless of course we find a way to produce power more cheaply and abundantly than with hydrocarbons. In which case the only thing we'll care about then is cooling
TLF
Re:Wrong. (Score:1, Insightful)
The other problem with this solution seems to be that with Vista coming, and making use of the graphics card for its user interface, there will be very little down time for the videocard.
abundant power not the answer (Score:2)
Currently a moderate cooling system that is capable of cooling a 600W gaming rig costs $150-200. With 50kW systems, it's going to be much more expensive, large, and cumbersome, ask people in the neares
Re:abundant power not the answer (Score:2)
New project idea: how to turn a Windows cluster into an on-demand water heater.
Re: (Score:2)
Tell that to SUV drivers.... (Score:2)
In the playstation generation, nobody cares about power consumption.
Re:abundant power not the answer (Score:2)
yeah, I went with the older socket amd (754), yeah the cpu wasn't that big and powerful.... and the integrated graphics first annoyed me, then i realized they weren't half bad.
i know, full eye candy games take a lot more processing power. but there is a trick if you can stand the bare bones ugly mode in most modern games. just turn off all the eye candy enhancers. yeah, it's true that doesn't sell
Re:Wrong. (Score:2)
Mobile? The desktop ATI cards, including the twin-slot monoliths, have done this for years.
Re:Wrong. (Score:3, Insightful)
Re:Wrong. (Score:2)
You haven't bothered looking much, have you? Fanless video cards are available for the taking. They're quite prevalent in Home Theater PC's (HTPC's) due to low noise levels, and the lack of a fan pretty much puts a very conservative upper limit on how much juice it can pull. Even given those limitations, you can find cards that'll give you fairly decent performance. Just look at AnandTech's recent HD-DVD/Blu-ray video card comparison.
The point is this: people do hav
Re:Wrong. (Score:2)
Depends on your viewpoint. Some of the graphics cards I own don't even need a bloody heat-sink, yet they do desktop things as well as cards made in the third millennium. (Or would have, if video RAM hadn't been so expensive back in 1996
Re:Wrong. (Score:3, Insightful)
"High energy consumption" is a very relative term. When compared to the video cards sporting multiple 5000rpm fans, heatpipes, and three pounds of copper-cored heatsink fins, a little one-inch-by-one-inch heatsink covering the GPU generally signals low energy consumption.
As for cards with no heatsinks at all, I think you'll agree that such animals are becoming very, very scarce these days, and they represent the barest fringe of
Re:Wrong. (Score:2)
Buy two when they come out [newsforge.com].
These soul-sucking proprietary vendors can piss off.
Re:Wrong. (Score:1)
Re:Wrong. (Score:2)
If you want the issue to be one of cost, wattage use isn't the problem. Consumers will only care when the up when it becomes too expensive to buy both a
Re:Wrong. (Score:1)
You almost need top-of-the-line stuff to get a constant 60 FPS in games that have been coming out recently.
Personally I would be nothing but happy if game developers slowed the rate at which they are improving graphics in games. Having to continuously purchase new hardware is really expensive, and the improved visuals rarely add anything of value to the game.
Re:Wrong. (Score:4, Insightful)
Oh, and Nethack.
Re:Wrong. (Score:2)
Re:Wrong. (Score:2)
Yeah, but note that using electricity for heating is expensive and wasteful, unless you have no other options for heating. If you had used district heating or something, the figures would have looked diffe
Re:Wrong. (Score:2)
Re:Wrong. (Score:2)
Re:Wrong. (Score:3, Insightful)
Eventually, the market will force GPUs down the sam
Re:Wrong. (Score:1)
Now to play the latest games you need a minum requirement for a card that requires a leaf blower attached to cool it. So lets see here, they've got power hungry computers, that are loud as hell to power PC games that are starting to struggle w
Re:Wrong. (Score:2)
Re:Wrong. (Score:2)
Re:Wrong. (Score:3, Informative)
No, you'll never hear "low-power GPU". You will, however, heard "fanless videocard" ALL THE TIME, and it's effectively a code for the same idea.
Regular people understand the issues far more than geeks give them credit for.
Re:Wrong. (Score:2)
Yeah, regular people who don't know why they need a video card since they don't play videos on their machine. Regular people who want to know why you want them to bring in their 'hard drive', when it is the computer that is the problem. Regular people who are using Office 2003 as their operating system. Regular people who don't know why their computers aren't running during a blackout.
These are the regular people.
The peop
Re:Wrong. (Score:2)
"Regular People" don't know the terminology, they don't know the difference between "RAM" and "Hard Drives", they don't know the difference between "Operating System" and "Programs", they wouldn't recognize a "hard drive" if it fell on them. Still, if you skip most of the terminology, average people are perfectly able to understand.
Re:Wrong. (Score:3, Informative)
A 1000 watt PSU does not use 10/6 times as much power as a 600 watt PSU. There are two reasons for this:
Re:Wrong. (Score:3, Informative)
Power just became more of an issue. (Score:5, Funny)
Re:Power just became more of an issue. (Score:1)
Re:Power just became more of an issue. (Score:2)
unless it is about 10,000 words, it should be 1-2 pages.
Re:Power just became more of an issue. (Score:2, Informative)
http://www.tomshardware.com/2006/07/21/the_graphi
Removes most of the ads, and puts the article on one page.
Great graphics... not so great games? (Score:4, Insightful)
Nowadays it's all about the graphics, and the gameplay tends to (but not always) suffer. Even the best of the best new games have these problems. FEAR? A pathetic 8-9 hours of gameplay, though it was pretty fun while it lasted. Oblivion? Tons of hours of gameplay, but completely SHALLOW in terms of the overall experience. Even Morrowind had this game beat IMO. Battlefield 2? Awesome graphics, and fun gameplay... oh, but don't try running more than a few bots on your machine unless you want to run at 2fps, and forget about coop play, and don't expect single player with more than 16 player maps (mods notwithstanding).
It seems to me that the more games focus on graphics, the more they lose in other areas. They either have cut features, performance issues, lack of content, or something... this isn't always the case (think Half Life 2), but unfortunately we're paying for the 'shiny factor' more often and losing out on the content that made the old games fun. Maybe I'm getting too old, or maybe I'm just jaded, but I still miss the old style games.
Re:Great graphics... not so great games? (Score:2)
Graphical output trumped text. VGA trumped CGA. Etc. Etc. The advance in graphical capabilites and the emphasis on graphics has always been with us. Why was Dragon Slayer wildly successful in 1984? Because it was the most extreme gameplay for graphics tradeoff ever.
What is new is that the vast majority of big p
... and when I said Dragon Slayer... (Score:3, Insightful)
Also, consider this an official request for a way to edit posts in the first minute after they're posted.
Video Game Nostalgia Effect (Score:3, Insightful)
Re:Great graphics... not so great games? (Score:2)
Re:Great graphics... not so great games? (Score:2)
It was a good game, but just that 'good', nothing more nothing less. It was pretty much the same as SuperMetroid and MetroidZero was also very much the same, it just felt like 'been there, done that'. When I look at the SNES with StreetFighter2, StarFox, MarioWorld, YoshisIsland, Zelda3, MarioKart, PrinceOfPersia, AnotherWorld, FinalFantasy, CronoTrigger, SuperMetroid, PilotWings, Actraiser, Simcity and a heck of a lot other great and original games, the
It is the noise, not the power that is killing me (Score:2, Interesting)
I wouldn't care a bit about power consumption if it wasn't so closely connected to noise levels.
Maybe it's just me... But is this really new news? (Score:2)
Re:Maybe it's just me... But is this really new ne (Score:2)
1100 watt power supply??? (Score:3, Insightful)
I'm sorry. No video game is worth that much power.
Re:1100 watt power supply??? (Score:4, Insightful)
Wait until you have to switch your PC from a regular 110V outlet to a round 220V outlet like the ones they use for electric ovens.
Maybe if you had a little meter next to you that rang up how much you were paying for electricity since you turned on your pc people would be more conservative. Right now it is a bit of a hidden cost since it all gets lumped together into a monthly bill, along with your AC, fridge, etc.
About The Subject On Power (Score:1)
Re:About The Subject On Power (Score:2)
I love it when people take RECOMENDATIONS out of context, and suddenly start calling them REQUIREMENTS.
"MINIMUM SYSTEM REQUIREMENTS" are nothing of the sort. They are not the minimum requirements for the device to work, they are simply a VERY general "minimum" value with lots of play built-in.
For
consider the consumer (Score:1)
Re:consider the consumer (Score:1)
NURBS (Score:2)
I wish it wasn't just workstation gpu's that got the good 2d and line drawing support (I guess it's not economic for consumer cards) I'm really interested in NPR (nonphotorealistic rendering) but all graphics cards are concerned
Misleading title and summary (Score:4, Informative)
No. It's an article more-or-less solely devoted to discussing the issue of power consumption in new and upcoming graphics cards. It doesn't describe the state of the union or even have much to say about any shiny new toys beyond their likely impact on power consumption.
It's an interesting article, but not the article that goes with its title nor the Slashdot summary.
Current buffer-swap implementations don't help (Score:5, Insightful)
When rendering in double-buffered mode with vsync on, the graphics card driver needs to wait for the display's vertical retrace before it swaps (or blits) the back buffer to the front. Today, all Windows drivers that I know of accomplish this with a spinlock. This means that an animated app grabs ALL available CPU cycles, even if the CPU actually needed to redraw each frame is trivial, and thus runs much hotter than it ought to for the amount of work being done.
For a high-end game that stresses the system anyway, this isn't a big deal. For more modest games or non-game applets, it's embarrassing to have a single rotating triangle forcing the machine to run all-out, particularly on battery power.
Application-level 'fixes' for this problem are very unsatisfactory - mostly trying to guess how long you've got until the next flip, Sleep()ing a bit and hoping you get woken up in time. It's clumsy, imprecise and the wrong place to be solving this. Why can't the driver wait on the flip - the flip it controls, for crying out loud - in some more efficient manner? (Can the new MWAIT instruction in EMT64 help with situations like this?)
Re:Current buffer-swap implementations don't help (Score:2)
Are you sure about that?
I have done a bit of hardware programming though I am no expert, and from what I know the only thing the driver will ever wait for is a full pipeline to empty. Modern GL/D3D implementations batch commands up into arbitrary sized buffers which are then DMAd to the hardware and processed internally .... this includes glSwapBuffers() commands .... calling glFlush or whatever the D3D equivalent is will cause the system to wait for the pipeline to empty but in any modern game so much ti
Re:Current buffer-swap implementations don't help (Score:2)
Well, graphics is only a hobby for me, but yes, I'm pretty sure. Profile a helloworld-type app, and it shows pretty much all your time in SwapBuffers.
calling glFlush or whatever the D3D equivalent is will cause the system to wait for the pipeline to empty but in any modern game so much time is spent actually calculating what commands to send that the overhead of synchronising with the card is quite low.
I'm not talking about modern games; with those, you expect the machine to sweat. I
Re:Current buffer-swap implementations don't help (Score:2)
Re:Current buffer-swap implementations don't help (Score:2)
Re:Current buffer-swap implementations don't help (Score:2)
Hey, some good news! Thanks for the info.
Re:Current buffer-swap implementations don't help (Score:2)
Actually, are you sure about that? I've just installed 84.21 for my ageing 4200 Go, and the 100% CPU problem is just as bad as ever.
Re:Current buffer-swap implementations don't help (Score:2)
Re:Current buffer-swap implementations don't help (Score:2)
Re:Current buffer-swap implementations don't help (Score:2)
Particularly some of the comments to the Old New Thing [msdn.com] blog.
Nice try, but power consumption isn't sexy enough (Score:2)
"Better power management" gets funny looks.
Funny look != profits
Step 2. is NEVER "generate funny looks"
Is the problem even tractable with current trends? (Score:2, Interesting)
1. GPUs have higher transistor counts than modern CPUs.
2. The development cycle for GPUs is much shorter than CPUs
3. The shelf life of GPU designs is much shorter than CPU designs (the C2D is a direct descendant of the P5, (pentium 4 arch is a dead end evolutionarily).
Given the preceding, it is unlikely that a reduction of power consumption will be the focus of GPU companies in the future, it would be suicide in a market which demands performance above all else. nVidia has shown that there ar
correction (Score:1)
Tom's Hardware is part of the power problem (Score:4, Insightful)
Why would a PC builder take power usage into consideration if the major review sites don't?
This is fucking stupid (Score:2)
Heat is the boundry of performance. The entire chip industry spends every day working out how to make things cooler per unit of performance so that they may increase total perform
Re:This is fucking stupid (Score:2)
See, here are a couple formulas off the top of my head:
1/heat * 1/wattage * speed = 1 (design tradeoffs)
efficiency coefficient * process coefficient * design tradeoffs = total goodness
Heat, wattage, and speed are linked. Non-mobile GPUs will ALWAYS be on the hot side because speed is king in that arena.
Going back to the ARM: efficiency is easy for a GPU because its computations are trivially parallelizable, which is why I
Question: (Score:2)
I don't want 400fps at max settings, I just want something relatively new that can run 3 year old games decently without turning the room into a sauna. Unfortunately with the things I've heard about Matrox and S3 it doesn't look like I've got much of a choice
Re:Question: (Score:2)
Power consumption in modern PCs is disgraceful (Score:3, Insightful)
Re:More important then the power problem (Score:4, Insightful)
A) A new interface (like PCI Express version 2 now with MORE POWER(tm))
B)Onboard power management and the ability to take power straight from the power supply and bypass the motherboard.
Re:More important then the power problem (Score:2)
D) With ads on it, with revenue paying your Power Company Bill.
Re:More important then the power problem (Score:1)
This seems like a good idea especially for notebooks.
They can be good to use for web use on your lap or at a coffee shop but when you put in on your docking station and hook it up to a 21 inch monitor you also plug in 4 video cards and have an awesome gaming machine from a laptop.
C) Use less power in the first place (Score:2)
Re:More important then the power problem (Score:2, Insightful)
Re:More important then the power problem (Score:2)
Price & performance will always be more import (Score:5, Insightful)
This makes all those "Green PC" claims a joke. I remember my first PC. It wasn't a "Green PC", but it had a 100W power supply, no heatsinks etc. My latest PC is a "Green PC" but has a 400W power supply. I'm not sure how a 400W-based system is greener than a 100W based system, but hey it says Green so it has got to be good right?
Re:Price & performance will always be more imp (Score:4, Insightful)
Re:Price & performance will always be more imp (Score:2)
Go ahead, take a look at the specs. I'll bet your old 300w powersupply couldn't top more than 10A on the +12v line, but most 300w powersupplies sold today hit the 15A mark for +12v. You could
Re:Price & performance will always be more imp (Score:3, Interesting)
Until prominently labeling approximated cost of power consumption is mandated by law, that is. When people are blatantly shown the cost to their pocket, they'll wisen up. It's working for appliances (well, at the mid and upper income ranges, anyway).
For people who don't pay for their electricity directly (like most college students) this won't be as big a factor, but
Re:Price & performance will always be more imp (Score:2)
And, out of curiousity, any idea of what the market share of these extremely high-powered cards are? I'm pretty sure it's relatively insignificant in terms of total power used.
Re:Price & performance will always be more imp (Score:2)
The Japanese care and I'm sure Google cares but, as a percentage of the market, just how many care? Power consumption these days is about making things work at all.
Re:Price & performance will always be more imp (Score:2)
Re:Price & performance will always be more imp (Score:2)
I am no specialist in this area, but I believe much of the "green" part of green PCs is what is used to build the computer. There are a lot of metals in computer boards that aren't good for the environment. Green pcs replace these with other metals or with plastic etc.
Re:Price & performance will always be more imp (Score:2)
Re:Price & performance will always be more imp (Score:2)
Re:A typical VACUUM draws MUCH more power than PC (Score:2)
Re:hmm (Score:2)