Power consumption 'review' and chart for Video cards...

[Jas420221]

No 2900xt or 8800Ultra though...

We not only focus on the 3D performance of a new graphic card, but also power consumption. Today we got about 8 cards to do this comparison, but it is a pity that we don’t have NV & AMD’s latest flatship: 8800 Ultra & HD 2900XT.

Remarks:

Except AMD’s reference HD 2600XT, all other cards are retail product.They all running at their default frequency and using stock cooler. We using Seasonic’s Power Angle to get the actual power consumption of the whole platform. Because all the cards are using the same platform so we believe the data we collect is trustful.

We got the results from two way:

1.idle in vista for 15minute

2.using ATI Tool’s show 3D view to simulate graphic card’s fullload.

http://en.expreview.com/?p=72

[Ryoko]

jeez... the gtx would tank my old power supply just by itself

i guess being a gamer is definitely not an eco-friendly lifestyle...

[eatyour1337ies]

Impressive. Slightly less consumption than the GTS at load yet performance somewhere betwixt the GTX and the Ultra. So how is big N going to move all that G80 stock if they keep prices where they are?

[ChrisRay]

jeez... the gtx would tank my old power supply just by itself

i guess being a gamer is definitely not an eco-friendly lifestyle...

8800GTX is a very very power hungry card. Yet for a long time it was the best performance per watt ((kinda crazy as technology evolves)) But imagine running 2 of them in SLI.. or maybe an even higher configuration?

Thats the beauty of process shrinks., And its why I love the 8800GT so much. While the 8800GTX is more performant., The 8800GT comes within shouting distance at so much less power usage. I can get nearly 8800GTX SLI performance from 2 8800GTs in SLI with a 550 Watt Enermax PSU.

Chris

[acroig]

Thats the beauty of process shrinks., And its why I love the 8800GT so much. While the 8800GTX is more performant., The 8800GT comes within shouting distance at so much less power usage. I can get nearly 8800GTX SLI performance from 2 8800GTs in SLI with a 550 Watt Enermax PSU.

Chris

Hey, Chris, I just got my kid an 8800 GT and it's clocked to 700 core 1000 RAM. Should that card not beat my GTX in most games @ 1600x1200? I'm just curious since you've mentioned that the GTX would win in some scenarios. Can you point me out to some numbers?

Thanks!!

[Sound_Card]

2.using ATI Tool’s show 3D view to simulate graphic card’s fullload.

Is this some sort of joke?

When using ATI Tool’s 3D view, Geforce 8800GT and Radeon HD 3870 is very close, 8800GT even 2 watt lower than HD 3870. And comparing the load mode of the new G92-8800GTS with old G80-8800GTS also have 2 watt’s diference; but in idle mode, there is a 20 watt’s gap between them. I guess using the new 65nm processing to shrink GPU core only affect their idle power consumption.

lol? This is a joke.

[ChrisRay]

Hey, Chris, I just got my kid an 8800 GT and it's clocked to 700 core 1000 RAM. Should that card not beat my GTX in most games @ 1600x1200? I'm just curious since you've mentioned that the GTX would win in some scenarios. Can you point me out to some numbers?

Thanks!!

It really depends on your settings..

[Jas420221]

Is this some sort of joke?

ehh...dunno. I dont know what exactly its doing, but I do know it heats my core up just as high if not higher than any game I own. It should be stressing the core and memory. I think it is b/c I get artifacts when I push my core to high, other artifacts when I my memory goes too high as well. So its doing something. How though, I dont know.

Where is Ray Adams when you need him?

lol? This is a joke.

My theory...that is a silly statement (them), have to agree there.

[acroig]

It really depends on your settings..

OK, 4xAA / 16xAF @ 1600x1200. Which card would get the best fps? My oc'd GTX or the GT @ 700/1728/2000?

[ChrisRay]

Probably a little faster. But it's gonna depend the game. For instance FEAR seems to prefer the 8800GTX design due to higher bandwith/stencil shadowing performance. But Bioshock will probably do better on the 8800GT ((at those clocks))

Your mileage will vary. But I'd say they were close to equivalent. As far as linkage numbers. No I dont have any as that configuration isnt widely available or benchmarked for reference. I could probably attempt to make some if I can keep my 8800GT's stable at those settings for long enough to do some real testing. The settings where the 8800GTX are going to be undisputed will be high memory utilization scenerios such as 8xQ, 16xQ, 16xS, 32xS and high resolutions. Below is an example of Bioshock on my system using dual 8800GTX verses 8800GT where 16xQ causes stuttering and higher FPS deficit due to running out of memory.

Bioshock 1680x1050



Performance Thoughts: Bioshock does extremely well on the 8800GT with 16xAA enabled. You can definately say that this will be the preferred setting for DirectX 9.0 users. Unfortunately under 16xQ I ran into some framebuffer limitations on the 8800GT SLI configuration which caused some hitching and prevented performance from being as optimal.

From my 8800GT SLI preview.

http://forums.slizone.com/index.php?showtopic=9798

But as I said in another thread. If you use 4xAA/8xCSAA/16xCSAA its almost impossible to tell the difference between 8800GTX cards and my 8800GT ones. Unless your actually looking for something to point out. But as games progress and memory amount increases. You'll find the 8800GTX hardware will behave more consistently.

[ChrisRay]

Ok finished editing. Now your question should be answered better.

[acroig]

Ok finished editing. Now your question should be answered better.

As usual, awesome job. Thanks so much. The GT is the new Ti 4200. What an awesome card. The same performance as GTX for about half the price. Incredible.

[-n7-]

As usual, awesome job. Thanks so much. The GT is the new Ti 4200. What an awesome card. The same performance as GTX for about half the price. Incredible.

Huh?

GT is a severely bandwidth crippled card that often gets raped by the GTX at high resolutions w/ AA & AF...& that gap will only get bigger as newer more demanding games come out.

Is the GT a fantastic deal compared to the GTX?

Absolutely.

But saying it's the same performance as the GTX is nonsense, unless you run at uber low resolutions.

Start benching @ 2560x1600 w/ eye candy turned up, & that "same performance" becomes remarkably dissimilar.

[acroig]

Huh?

GT is a severely bandwidth crippled card that often gets raped by the GTX at high resolutions w/ AA & AF...& that gap will only get bigger as newer more demanding games come out.

Is the GT a fantastic deal compared to the GTX?

Absolutely.

But saying it's the same performance as the GTX is nonsense, unless you run at uber low resolutions.

Start benching @ 2560x1600 w/ eye candy turned up, & that "same performance" becomes remarkably dissimilar.

Please take a look at post #9. No one is talking about stupid high resolutions, just 1600x1200. That's not "uber low" is it now?

[-n7-]

Please take a look at post #9. No one is talking about stupid high resolutions, just 1600x1200. That's not "uber low" is it now?

Kinda yeah, but i'm used to a tad *cough* higher.

I don't know why you'd want a GT though since you have a GTX, unless you want SLI...which i of course cannot recommend, since taking two bandwidth limited cards results in even worse scaling...

But each to their own.

I'm personally waiting for real nV release...GT & GTS 512 MB only annoy me, since having nothing really better than the GTX for over a year doesn't make me happy.

[acroig]

I don't know why you'd want a GT though since you have a GTX....

Please look at post #5.

And yes, I too would like to see a new high end from nV.

[acroig]

Kinda yeah, but i'm used to a tad *cough* higher.

That's the blessing and the curse of owning huge displays. Awesome to look at but tough to get top performance at native res.

[Sound_Card]

ehh...dunno. I dont know what exactly its doing, but I do know it heats my core up just as high if not higher than any game I own. It should be stressing the core and memory. I think it is b/c I get artifacts when I push my core to high, other artifacts when I my memory goes too high as well. So its doing something. How though, I dont know.

Where is Ray Adams when you need him?



My theory...that is a silly statement (them), have to agree there.

ATi tool is made by w1zzard. You must be thinking of ATi tray tools. I hope they did not use tray tools, as that would be even worse. My x1800gto2 get's over 300 fps on ATi tools 3d view.

It's a 3D hairy cube spinning in random directions. It's going to increase power consumption just like any 3D application is, but obviously not every 3d app is the same and is going to vary from work load. Which is why I get a little frustrated when power consumption is measured with 3dmark 2006 and 3dmark 2006 only. Not a good indicator of how todays GPU's throttle across work loads.

[ChrisRay]

Huh?

GT is a severely bandwidth crippled card that often gets raped by the GTX at high resolutions w/ AA & AF...& that gap will only get bigger as newer more demanding games come out.

Is the GT a fantastic deal compared to the GTX?

Absolutely.

But saying it's the same performance as the GTX is nonsense, unless you run at uber low resolutions.

Start benching @ 2560x1600 w/ eye candy turned up, & that "same performance" becomes remarkably dissimilar.

I dont think the gap will "increase" greatly as newer games come out. ((Unless framebuffer limitations come to account)) you will probably see similar behavior patterns throughout the "useful" lifetime of either of these cards. The only exception being is when that 512 barrier gets hit. The 8800GT has some interesting bandwith saving technology and the shader domain is clocked high enough to where there is not a large difference in shader power.

Chris

[acroig]

The 8800GT has some interesting bandwith saving technology and the shader domain is clocked high enough to where there is not a large difference in shader power.

Chris

I noticed. 1728, impressive. The best I could get out of my GTX is 1620.

[Crag2804]

That power drain on the 8800 gtx is why I built a low power downloading machine. Running my gaming machine over night was a huge waste of power to download things. Now I've got a small box that pulls 42W total when it's downloading. Much less than the few hundred watts it used to cost me to see heroes.

[razor1]

ATi tool is made by w1zzard. You must be thinking of ATi tray tools. I hope they did not use tray tools, as that would be even worse. My x1800gto2 get's over 300 fps on ATi tools 3d view.

It's a 3D hairy cube spinning in random directions. It's going to increase power consumption just like any 3D application is, but obviously not every 3d app is the same and is going to vary from work load. Which is why I get a little frustrated when power consumption is measured with 3dmark 2006 and 3dmark 2006 only. Not a good indicator of how todays GPU's throttle across work loads.

That 3d hairy cube is very graphically intensive, anyone want to look into how many calculations and what parts of the GPU being stressed?

Just a hint, why we don't see fur shaders or grass shaders that fill up the screen in real time games yet.......

[Jas420221]

That power drain on the 8800 gtx is why I built a low power downloading machine. Running my gaming machine over night was a huge waste of power to download things. Now I've got a small box that pulls 42W total when it's downloading. Much less than the few hundred watts it used to cost me to see heroes.

Your gaming rig DL things at 400W? Impossible. Unless you have an SLI setup, a sheeeeeeeeee ton of HDDs and active USB perihperals, its not pulling 400W to DL.

That 3d hairy cube is very graphically intensive, anyone want to look into how many calculations and what parts of the GPU being stressed?

Just a hint, why we don't see fur shaders or grass shaders that fill up the screen in real time games yet.......

Thank you for your support that was much more intelligently put!

[Sound_Card]

That 3d hairy cube is very graphically intensive, anyone want to look into how many calculations and what parts of the GPU being stressed?

Just a hint, why we don't see fur shaders or grass shaders that fill up the screen in real time games yet.......

It may be graphically intensive from a artistic standpoint, not GPU intensive. I hope your not trying to argue it's a valid form to measure power consumption from. It is not. It's a 400x400 box with a even smaller cube and a solid background in which a x1600xt get's about 100 frames per second on.

hint: these GPU's are not going to run all out on ATi tool 3d view.

Here is what happens when you take power consumption from a handful of games and factor them together.

[razor1]

next time actually link to the article

http://www.hothardware.com/articles/..._RV670/?page=5

what motherboards are they using again?

We tested all of the graphics cards used in this article on either an EVGA nForce 680i SLI motherboard (NVIDIA GPUs) or an Asus P5E3 Deluxe (ATI GPUs) powered by a Core 2 Extreme QX6850 quad-core processor and 2GB of low-latency Corsair RAM. The first thing we did when configuring the test systems was enter their respective BIOSes and set all values to their "optimized" or "performance" default settings. Then we manually configured the memory timings and disabled any integrated peripherals that wouldn't be put to use. The hard drive was then formatted, and Windows Vista Ultimate was installed. When the installation was complete we fully updated the OS, and installed the latest DX10 redist and various hotfixes along with the necessary drivers and applications.

[Sound_Card]

Too bad the difference is not 50 watts between x38 and 680i.

http://techreport.com/articles.x/13351/13

[ChrisRay]

The amount of power consumed by a board will depend greatly on whats installed. 3 GPUS for instance will use more power ((at the motherboard level)) than 2. On the 680I boards.

[Jas420221]

Too bad the difference is not 50 watts between x38 and 680i.

http://techreport.com/articles.x/13351/13

Weird how that picture shows the 2900xt consuming less power than a GTX....was it always that way? I thought the 2900xt took that crown by about 20W not using less by almost 30W???

It may not be 50W but it is ~30W.

[jam2k]

Yeah I remember everyone screaming when 2900XT launched: "ZOMG I'll need a power plant to run R600 !"

[Jas420221]

Yeah I remember everyone screaming when 2900XT launched: "ZOMG I'll need a power plant to run R600 !"

Who can forget...

Thats why I personally dont like that picture that was posted.

[hazindu]

It may be graphically intensive from a artistic standpoint, not GPU intensive. I hope your not trying to argue it's a valid form to measure power consumption from. It is not. It's a 400x400 box with a even smaller cube and a solid background in which a x1600xt get's about 100 frames per second on.

hint: these GPU's are not going to run all out on ATi tool 3d view.

Here is what happens when you take power consumption from a handful of games and factor them together.

I don't think ati tool's hairy box thing per se is a great stress test as it doesn't use the latest shaders and other features, but that has nothing to do with it being a 400-400 box. Small windowed demos should work great assuming the vsync is off, and the card doesn't have separate 2d and 3d clock settings which to my knowledge, the geforce 8s don't, as there is no processor or memory bottlenecking the graphics card, so it can go full boar. I've seen some of my highest gpu temps while running Humus's demos in a 200 by 200 window. Unlike Oblivion or Crysis, these demos do not allow the card to take a breather while caching memory or waiting for something to load.

Also keep in mind that some of the greatest processor stress tests are simple small programs that do things like test prime numbers, or calculate PI.

[razor1]

Yeah I remember everyone screaming when 2900XT launched: "ZOMG I'll need a power plant to run R600 !"

For the performance that it gives, its alot.