Power consumption of 12 games: guess which is most close to Furmark

[voidshatter]

Since nVidia "cheats" in Furmark, I'm interested in finding an alternative test to determine the real power consumption.

Test platform:

Intel Core i7 980X [6C6T] @ 4GHz 1.25v
Corsair H70
EVGA X58 Classified E760
6 x 2 GB Kingston DDR3 1600 C7 1.65v
Intel X25-M G2 160G SSD
2 x MSI 6950 2GB Twin Frozr II 810/1250 1.1v
Corsair HX1000
Lian Li 7FNWX

Microsoft Windows 7 64bit SP1
Catalyst 11.5 WHQL "Power Control settings" = + 0%

All games/programs use max settings at 1920x1200 (with vsync off, and PhysX disabled where appropriate) unless otherwise stated.

Readings are from a power meter between the PSU and the wall. Each number is the maximum value I can observe / manage to tweak from each game/program. (However not guaranteed to include the most power hungry scenes.)

[voidshatter]

It is known that the only working driver to enable full crossfire performance in Crysis 2 is Catalyst 11.4preview (Mar 23) + 11.3 cap5. Therefore the power consumption of Crysis 2 measured above (with Catalyst 11.5 WHQL) may not represent the max power consumption from the game.

Supplementary data:

CPU usage of Furmark:



CPU usage of Metro 2033:

[Zhriver]

No Bad Company 2? I believe I've managed 730 watts on my 2600k along with 2x 580's.
Of my portfolio of games that was the hardest hitter.
Also I imagine Arma 2 with every single config tweaked for maximum fidelity, along with a massive scenario could be a contender.

[voidshatter]

Originally Posted by Zhriver

No Bad Company 2? I believe I've managed 730 watts on my 2600k along with 2x 580's.
Of my portfolio of games that was the hardest hitter.
Also I imagine Arma 2 with every single config tweaked for maximum fidelity, along with a massive scenario could be a contender.

Oh too bad I just uninstalled that game yesterday... :S

By the way, it's been reported that BFBC2 gets a huge fps boost by upgrading dual core CPU to quad-core CPU, thus I guess much part of your 730w may come from your CPU?

[W1zzard]

there is no "real" power consumption, find a benchmark that you think puts a realistic load on the card.

also consider different cpu usage = different power when measured at the wall socket.

i like the power consumption profile of crysis 2 very much, which is why i'm using it for typical gaming power consumption testing in my reviews

[voidshatter]

Originally Posted by W1zzard

there is no "real" power consumption, find a benchmark that you think puts a realistic load on the card.

also consider different cpu usage = different power when measured at the wall socket.

i like the power consumption profile of crysis 2 very much, which is why i'm using it for typical gaming power consumption testing in my reviews

But lots of newbies keep asking questions like "which PSU to choose for my graphics cards / which graphics cards can my PSU support".. It would be safer to suggest according to the maximum power consumption numbers, and I believe it's why you bother to offer a function in GPU-Z to disable the power throttling of GTX580/570.

[RejZoR]

Try this:
- Unreal Tournament 3
- Morbias Space Station map
- use Excessive Overkill Mutator
- use Spectator mutator
- set game speed to max
- set bots to hardest difficulty
- set max number of bots
- set graphics to max
- set match time or score to some very high number

And leave the game running. If you want to put more load on CPU, give them Flak Cannons and use a mutator that makes shrapnels ricochet more than once and that it fires more shrapnels.
This is how i used to test my overclocks. Observing this insane botmatch should do the trick.

[Dippyskoodlez]

I'm curious to see what the min, max, and average FPS obtained [in each game], is, in comparison to power consumption.

Just an interesting thought

[W1zzard]

Originally Posted by voidshatter

But lots of newbies keep asking questions like "which PSU to choose for my graphics cards / which graphics cards can my PSU support"..

look at the graph in the first post, the number is 600 W. 500 w realistic use + some margin. it's not that the psu will explode, burn your house and get your gf pregnant the instant the system uses 601 w

[W1zzard]

that's card only power consumption for my last benchmark suite with some random high performance card. dont know the benchmark names though. bottom line is that power varies wildly between tests and resolutions

[RejZoR]

Originally Posted by W1zzard

look at the graph in the first post, the number is 600 W. 500 w realistic use + some margin. it's not that the psu will explode, burn your house and get your gf pregnant the instant the system uses 601 w

Most good PSU's can pump up even more power than they are rated at. I remember 1kW PSU's to deliver 1,2kW in some tests without any problems.

[voidshatter]

Originally Posted by W1zzard

look at the graph in the first post, the number is 600 W. 500 w realistic use + some margin. it's not that the psu will explode, burn your house and get your gf pregnant the instant the system uses 601 w

How would you recommend PSU for a GTX590 then? (Without knowing the real furmark number)

In the US/EU most PSU are built with high quality, but in China it's not always true. If you don't leave enough margin you might end up with BSOD or even damaged graphics cards. (Lots of posts about this already)

Originally Posted by W1zzard

http://img.techpowerup.org/110419/Capture1924.jpg
that's card only power consumption for my last benchmark suite with some random high performance card. dont know the benchmark names though. bottom line is that power varies wildly between tests and resolutions

I don't know about the unit but I can see the maximum value doesn't diverge much.

[W1zzard]

Originally Posted by voidshatter

In the US/EU most PSU are built with high quality, but in China it's not always true.

as far as i know no PSUs are built in the EU/US. some companies are designing them there, or putting on the sticker, but that's it. basically there are only a handful factories on the planet that make pc power supplies

How would you recommend PSU for a GTX590 then?

350-400w (for the card only), so 750-850w for the system

[voidshatter]

Originally Posted by W1zzard

as far as i know no PSUs are built in the EU/US. some companies are designing them there, or putting on the sticker, but that's it. basically there are only a handful factories on the planet that make pc power supplies


350-400w (for the card only), so 750-850w for the system

I may not be stating this very clearly.. The PSU you can buy in the US/EU are major brands. However there are many "minor" brands in China, which are "castrated" or "shrinked" or whatever the word is, down from the standard quality. Also if taken yield and quality into consideration, it's not safe to push a PSU near its limit. That's why we used Furmark numbers as a guideline when possible.

[AsRock]

First off you should never buy a no brand as the result if you do it's totally your fault that ya pc exploded.

If i was building another system i would do what i did last time. Find the pc parts that i want and find out how much power would take while gaming then buy a psu double the power of what the pc going to be running at while gaming.

While gaming on my system it's around 300-370w ( whole system). Although over the last 3 years i have had it the GPU( 2900XT\4890\6970 has been changed 3 times and the cpu once from a E8400 to 9550 and it's always been around 300-370w.

Although the 2900XT and 4890 was heavy overclocked were as my 6970 is stock which i am sure could push it 400w but because i got a quality psu with qa well known name with twice the power that i needed i have had 0 issue's.

And as far as furmark goes it's unrealistic power usage.

[voidshatter]

Originally Posted by AsRock

First off you should never buy a no brand as the result if you do it's totally your fault that ya pc exploded.

If a certain brand is made in China, they do quality control (similar as binning CPUs/GPUs when the yield rate is low) and sell the top class to the US/EU, and sell the lowest class in China.

Also keep in mind that £1000 for a pair of graphics cards might be cheap for people in the UK, but such price is a heavy burden to most people in China. Many people have to consider brands, e.g. "Yeston" etc making graphics cards selling at lower price than ASUS/Giga/EVGA/MSI etc. The same for PSU - you may have never heard about the brands before, yet these are popular there.

[silkstone]

If you want to know the maximum possible power your computer would use why not use the PSU stress test in OCCT? i think that would put a pretty large load on your PSU.

[AsRock]

Originally Posted by voidshatter

If a certain brand is made in China, they do quality control (similar as binning CPUs/GPUs when the yield rate is low) and sell the top class to the US/EU, and sell the lowest class in China.

Also keep in mind that £1000 for a pair of graphics cards might be cheap for people in the UK, but such price is a heavy burden to most people in China. Many people have to consider brands, e.g. "Yeston" etc making graphics cards selling at lower price than ASUS/Giga/EVGA/MSI etc. The same for PSU - you may have never heard about the brands before, yet these are popular there.

IF yeston is a good brand in there country buy it if it's not don't.. After all it's why i used the word brands and did not name companys as i do know more than most that one brand is not available in another due to moving to another country and experiencing it first hand.

Not much for your example either as £1000 is not cheap for 2 video cards a a hell load of people in the UK even more so when i can get the same thing reasonably cheaper.

[voidshatter]

Originally Posted by silkstone

If you want to know the maximum possible power your computer would use why not use the PSU stress test in OCCT? i think that would put a pretty large load on your PSU.

Because we want to give recommendations of PSU regarding graphics cards (and also try to find out the performance per watt of graphics cards). In such case, stressing CPU is not very helpful unless a systematical method is used to fully stress the CPU with low priority process in background while benchmarking games to make a fair comparison. Furmark used to be good until nVidia started cheating.

[yogurt_21]

Originally Posted by voidshatter

Because we want to give recommendations of PSU regarding graphics cards (and also try to find out the performance per watt of graphics cards). In such case, stressing CPU is not very helpful unless a systematical method is used to fully stress the CPU with low priority process in background while benchmarking games to make a fair comparison. Furmark used to be good until nVidia started cheating.

W1z essentially does this with each review

[voidshatter]

Originally Posted by yogurt_21

W1z essentially does this with each review

These are stated in w1z's reviews:

Originally Posted by w1zzard

Maximum: Furmark Stability Test at 1280x1024, 0xAA. This results in a very high non-game power consumption that can typically be reached only with stress testing applications. Card left running stress test until power draw converged to a stable value. On cards with power limiting systems we will disable the power limiting system or configure it to the highest available setting - if possible. We will also use the highest single reading from a Furmark run which is obtained by measuring faster than when the power limit can kick in.

When it is no longer possible to avoid cheating in Furmark, how are we supposed to get an alternative?

[W1zzard]

Originally Posted by voidshatter

When it is no longer possible to avoid cheating in Furmark, how are we supposed to get an alternative?

call nvidia, request to speak to jen-hsun and ask your question

[voidshatter]

Originally Posted by W1zzard

call nvidia, request to speak to jen-hsun and ask your question

Then I'm sure he would respond: "don't worry, we have optimized GTX580 and it now consumes a lot less power"

Originally Posted by Advertisements of GTX580

- Lower power consumption compared to GTX 480, only 244 watts