Real Power Consumption - 4870 X2 & GTX295 out of Spec!

[Manicdan]

and AMD did make it so FurMark would run using much less power (and less performance) with one of their drivers. FurMark is definitly the worse case scenario, and i think it does show us that the card companies should build their cards to cut off the super FPS if the power goes too high. while this may not be a real case, imagine if a game was able to pull that much power out of your card, it wont take long for you to smell something going wrong, hoping that the company knew about it soon enough to limit its output.

and btw if AMD made a CPU that was 2x 940s using 8 cores and generating 300w and costing about 500$ i would buy it. to some of us, the power draw isnt that crazy, we dont all keep out systems at load 24/7, but when we use them, we must enjoy them. i have a 2900xt that runs at 1.35v in 3D applications and overclocks to 950mhz, im scared to know how much power it uses then, well more like curious.

[Tuvok-LuR-]

so this means 33% of your 4870 - 4870X2 is idling while playing games :O

[XS Janus]

Err.. its like twice as fast?

Thats what I meant... where's the progress

[vinister]

All the specs are tested under worst-case scenario, too. All those connectors can supply their rated current with ambient temperatures at their max rated (usually 80-100 degrees C), and for 100% duty... which means 24/7/365. Pushing their limit at regular room temperature for a few minutes is child's play.

Does the article mention any of the video cards crashing during the test?
Does it mention anything melting?
Does it mention any problems whatsoever?

not that I saw. So, what's the big deal? "Look how far past spec we can push this stuff" Cool. That's what this industry's all about!

I think instead of adding more and more power connectors to these cards, they definitely should be working on a high-current standard for power supplies and connectors.

Its just stupid to connect an 8 pin power connector that is just full of small-guage powers and grounds, a single 8 AWG pair will supply video cards doubling today's power draws. It would look alot cooler, too I say just give me a power supply with a terminal strip, say 10 terminals of +12V and ground, and let me wire it myself, so I can spec whatever size cable fits my needs.

[Manicdan]

Its just stupid to connect an 8 pin power connector that is just full of small-guage powers and grounds, a single 8 AWG pair will supply video cards doubling today's power draws. It would look alot cooler, too I say just give me a power supply with a terminal strip, say 10 terminals of +12V and ground, and let me wire it myself, so I can spec whatever size cable fits my needs.

couldnt agree with you more. i had a 265W TEC that i wired to a PSU by using a molex, and it turned from white to yellow and was pretty hot. im sure the wire was doing ok, but it was still enough to damage the plastic connector. they should offer PSUs with real wire connectors and different gauge wires paired with black/red and yellow/black, or any color and black. truth is people would need to use alot of wire as they change things around, and companies love nothing more than repeat customers due to disposable parts.

while it wouldnt be for normal people, many of us who build custom computers can wire the parts together like childs play.

[STEvil]

FurMark isn't a prefered benchmarking tool since it pushes GPUs above and beyond what they are specified to run at. You can't say that just because a FurMark produces abnormally high power consumption and that the GPUs are running out of spec. It is the PROGRAM that causes the GPUs to run above and beyond, NOT the GPUs themselves.

I've seen the way ATI tests for power consumption and they run every game you can think of at multiple resolutions to determine their max board power consumption and I am sure Nvidia does the same thing. However, I would go so far as to say that FurMark is buggy in the fact that it loads the GPU in ways it was not meant to be.

So should we stop using Linpack and other tools which utilize CPU's to their full potential as well?

The GPU's were rated at a "working" power usage level which is fine by me, but to say that we cannot measure their full peak power draw because they never attain that doesnt sit right with me. Who's to say that some GPGPU application(s) might not use more of the GPU? What if a game comes out that uses the GPU more, are they going to have to write "idle" routines into the next driver release so you dont go over the allotted power draw (ie: Furmark renaming)?

I think two things should happen.

1: They should be rated at their peak.
2: More reviews should be done like this.

[vinister]

I can't help but disagree. Its abusing the product.

You can buy a car and tow a boat with it if you like, but its abusing the product, and it will break it. Its not the company's fault, because although the car was able to pull the boat, thats not what the car was designed for.

This is the exact same situation.

While your hardware will run a silly test like furmark, it was simply not designed to. It might break it. Lucky for us, they are built well enough that they don't break, they just don't get their advertised fuel economy

[clayton]

GTX 260 looks very efficient at full blast.

[SNiiPE_DoGG]

I can't help but disagree. Its abusing the product.

You can buy a car and tow a boat with it if you like, but its abusing the product, and it will break it. Its not the company's fault, because although the car was able to pull the boat, thats not what the car was designed for.

This is the exact same situation.

While your hardware will run a silly test like furmark, it was simply not designed to. It might break it. Lucky for us, they are built well enough that they don't break, they just don't get their advertised fuel economy

+1 best analogy and exactly the case with fur mark

[mibo]

So should we stop using Linpack and other tools which utilize CPU's to their full potential as well?

The GPU's were rated at a "working" power usage level which is fine by me, but to say that we cannot measure their full peak power draw because they never attain that doesnt sit right with me. Who's to say that some GPGPU application(s) might not use more of the GPU? What if a game comes out that uses the GPU more, are they going to have to write "idle" routines into the next driver release so you dont go over the allotted power draw (ie: Furmark renaming)?

I think two things should happen.

1: They should be rated at their peak.
2: More reviews should be done like this.

+1

The car manufacturer would only guarantee that his product works for "normal use".

That "normal use" is what we discuss about.

"normal use" of a cpu: run every software you want on it - and still the cpu will not exceed specifications. (prime, occt, ...)

"normal use" of a gpu: for me: run every software you want on it - and still the gpu does not exceed specifications.

[Jamesrt2004]

+1

The car manufacturer would only guarantee that his product works for "normal use".

That "normal use" is what we discuss about.

"normal use" of a cpu: run every software you want on it - and still the cpu will not exceed specifications. (prime, occt, ...)

"normal use" of a gpu: for me: run every software you want on it - and still the gpu does not exceed specifications.

wouldn't call burning programs to make the processor/GFX as hot as it can be "normal"

[largon]

^Indeed.
If power viruses are "normal use", then what is "abnormal use"?

[Martijn]

+1

The car manufacturer would only guarantee that his product works for "normal use".

That "normal use" is what we discuss about.

"normal use" of a cpu: run every software you want on it - and still the cpu will not exceed specifications. (prime, occt, ...)

"normal use" of a gpu: for me: run every software you want on it - and still the gpu does not exceed specifications.

Exactly... If your car overheated everytime you pushed it to the max would you accept that as well? I don't think so.

[Slay0r]

Mainly windmills at night. And then there is hydropower from norway aswell at night. Both are "uncontrolled" and thereby having large excess capacities at night and with windmills also when its very windy.

Its calculated that 700 windmills can keep the entire danish carpark running on electricity. Even making gas into electricity is a large plant would be more efficient that out in the car

But again, we also burn biofuel for CO2 neutrality. So its a win/win.

I'm sorry but the electric/hybrid car market is a big joke. Mainly because the arabs won't let an actual substitute for gas on the market, because that means less money for them. This may sound very childish and/or unlikely but you just have to look at the amount of good ideas for "clean" energy that get thrown in the bin for unclear reasons.

Look at what we have now. electric cheapo cars that won't even do 100mph and that will run out of battery charge in less than 100miles. Or that failure of the hybrid prius, which gets its ass spanked by the newer 1.3/1.4l diesels in terms of mpg and performance. The market REALLY does want you to overlook the green cars as of now.

Back on topic, I'm sure the cards weren't designed to run a very heavy stress test. If you game/render/work on them and they overheat, then that's another matter. AND if they draw double than what they're advertised to draw, then you have the right to complain. I mean, just look at how big the stock gfx cooling solutions are getting.. Soon we'll need a case with its own airflow just for the card itself. The problem is that both NV and AMD will not back down on their specs/anything, since we still buy them no matter what they draw, because they're the best of their league..

[Shintai]

I'm sorry but the electric/hybrid car market is a big joke. Mainly because the arabs won't let an actual substitute for gas on the market, because that means less money for them. This may sound very childish and/or unlikely but you just have to look at the amount of good ideas for "clean" energy that get thrown in the bin for unclear reasons.

Look at what we have now. electric cheapo cars that won't even do 100mph and that will run out of battery charge in less than 100miles. Or that failure of the hybrid prius, which gets its ass spanked by the newer 1.3/1.4l diesels in terms of mpg and performance. The market REALLY does want you to overlook the green cars as of now.

Back on topic, I'm sure the cards weren't designed to run a very heavy stress test. If you game/render/work on them and they overheat, then that's another matter. AND if they draw double than what they're advertised to draw, then you have the right to complain. I mean, just look at how big the stock gfx cooling solutions are getting.. Soon we'll need a case with its own airflow just for the card itself. The problem is that both NV and AMD will not back down on their specs/anything, since we still buy them no matter what they draw, because they're the best of their league..

First fo all. Arabs have nothing to say. This is something the entire country is geared towards. Secondly I think you failed to understand the concept of the bettery changing stations. Its no different than putting gas on your car. Its just faster with the battery change.

And hybrids was always a joke. Same reason the Volt is an utter disaster (But thats just how GM works and why they can never be saved). Pure electric is the way. Plus a gas/diesel engine is so horrible in efficiency that you can get twice the power by putting the fuel into a powerplant, convert it to electricty and then charge a battery and use it in a car.

In terms of the power consumption. Furmark is not really a power virus. And even if, can the GPUs throttle when they overheat? The answer is no.

Its seen ebfore, atleast with older nVidia cards that drivers+game pushed the GPUs beyond their limits. And there is artificial limits in both nVidia and AMD drivers to keep games running within the power profile.

[gOJDO]

Exactly... If your car overheated everytime you pushed it to the max would you accept that as well? I don't think so.

It's like you buy a scooter motorbike, designed for a city ride and you complain becouse it failed climbing on a rocky mountain.

[Slay0r]

First fo all. Arabs have nothing to say. This is something the entire country is geared towards. Secondly I think you failed to understand the concept of the bettery changing stations. Its no different than putting gas on your car. Its just faster with the battery change.

Well.. My thought was generalized to the world as a whole, I wasn't referring to one country, not yours so to speak, since I frankly don't know what goes on over there in particular because I'm not documented enough! I can just take your word for it. As for the battery changing thing, I replied down below.

And hybrids was always a joke. Same reason the Volt is an utter disaster (But thats just how GM works and why they can never be saved). Pure electric is the way. Plus a gas/diesel engine is so horrible in efficiency that you can get twice the power by putting the fuel into a powerplant, convert it to electricty and then charge a battery and use it in a car.

While to an extent we may agree here, I'd rather have a not so efficient diesel that can do 1000miles with one tank of fuel rather than have to stop every tenth of that to swap batteries!

In terms of the power consumption. Furmark is not really a power virus. And even if, can the GPUs throttle when they overheat? The answer is no.

Its seen ebfore, atleast with older nVidia cards that drivers+game pushed the GPUs beyond their limits. And there is artificial limits in both nVidia and AMD drivers to keep games running within the power profile.

I don't know enough about power castration within the driver to elaborate a proper comment on this, but it doesn't sound completely true. If it was the case, overclocking would give gains until the power profile maximum is reached, then start being useless. Is this the case or I'm misunderstanding your reasoning?

[Humminn55]

Exactly... If your car overheated everytime you pushed it to the max would you accept that as well? I don't think so.

But that's not the problem here.......just pushing something to the max on occasion is not the issue.....the issue is how Furmark overworks the card(s) in an unrealistic scenario.

Take this analogy......you are looking at purchasing a used car and you have two Honda Civics to choose from.

The first is one owned by a middle-aged couple who used the car "sensibly," meaning while it was accelerated at WOT (wide open throttle) to merge onto interstates, etc., it spent the majority of its life "putting" around at under 75% of its max use.

The other Civic is owned by a "tuner." Wing, engine "massaged" internally with replaced pistons and head which raised compression ratio, Nitrous injection, and raced nightly for a year or so.

Which do you buy for longevity and hopefully the car won't need major repairs during your payments?

Naturally, you'd take the "sensible" Civic. Why? The "tuned" Civic has been thrashed within an inch of its life, has been run flat out so much that failure is inevitable. After all, the adage that the candle that burns twice as bright burns half as long applies to almost everything.....candles or computer components.

So, push a car to its limit ALL the time and a major failure will happen much sooner than one driven "easier."


What I really don't understand is the fascination a lot of "O'Cers" have in running Furmark or Linpack for hours and hours at a stretch, esp. in a heavily overclocked system.

True, overclocking can create instability, but to "test' for stability by running a system flat out for hours just to try to catch some sign of instability seems rather senseless to me.

I O'C for fun and to get a little "free" speed out of my system......don't we all here? On the other hand, I'm not running Linpack or Furmark for hours trying to get instability to show up......

If my O'C is unstable, it'll show up in a gaming situation....or it won't boot worth crap, or the system will freeze when opening programs....all signs the O'C is too high and I need to back off.

But, instead, you get those that obsess about running programs that run their components at the ragged edge for hours and then wonder why they get overheating problems, early component failures/death, etc.

What's that? You're depending on ultimate stability for something "mission critical," which is what I always hear as the reason for this obsessive testing..... To that I retort, if you're depending on an ultra stable system for things like CAD design or quality HD conversion, etc., run a fast but stock speed system. Seems to work for people who depend upon ultra stable systems for their work, like graphic artists at Industrial Light and Magic or Pixar. (You honestly think they'd take a chance using overclocked sytems that have the chance of data corruption? Not on your life..........)

Blaming the video card makers for having too much power, more than "advertised," being consumed running Furmark is akin, at least to me, to blaming a car manufacturer for your gas mileage sucking when you've been driving down the interstate for hours at 100mph.....


And if running Linpack and Furmark seems just dandy to you, try this test for stability on your own vehicle you've got parked outside your house.

Go start it up, leave it in neutral or park or whatever, block the wheels so it cannot move, and put a brick on the accelerator pedal so the engine is running wide open. Come back 6 hours later and see if your car is still running.......that's about what Furmark is doing for your video card.

Sorry for the ramble, but throwing around blame and astonishment that Furmark makes video cards draw more power than what the card makers say they do is silly.....Furmark, while a 3D program, is a wholly unrealistic test...it runs the card wide open without a single pause anywhere, something no game does, even Crysis. To not expect the video card to draw more than advertised power when running wide open without pause at all is an unrealistic expectation and should not surprise anyone.

[Shintai]

I don't know enough about power castration within the driver to elaborate a proper comment on this, but it doesn't sound completely true. If it was the case, overclocking would give gains until the power profile maximum is reached, then start being useless. Is this the case or I'm misunderstanding your reasoning?

Misunderstood it. Power profiling would not have any effect on overclocking.

To put it short. 9 work, 1 idle. If you does it 10 times per second or 1billion times per second doesnt change the power profile.

[Eastcoasthandle]

I can't help but disagree. Its abusing the product.

You can buy a car and tow a boat with it if you like, but its abusing the product, and it will break it. Its not the company's fault, because although the car was able to pull the boat, thats not what the car was designed for.

This is the exact same situation.

While your hardware will run a silly test like furmark, it was simply not designed to. It might break it. Lucky for us, they are built well enough that they don't break, they just don't get their advertised fuel economy

I have to agree with this car analogy here. a good while back I use to have a Olds Firenza and one of my friend's truck broke down (a 1980 something Ford Explorer). I towed him some 10 miles back home. When I got home the anti-freeze in the radiator boiled like a hot tea pot. Clearly, that vehicle was not design to tow a vehicle at twice it's size and weight.

So yes, even though you can take a CPU or GPU to it's extreme maximum doesn't mean you are suppose to. So I agree here.

Edit:
Forgot to mention that not only did I have to top off the antifreeze but, I also had to change the oil and replace the gas as well do to the towing. Having done this was I suppose to assume that was the average MPG I usually received had I not towed the car? Was 1200 miles the best I could get out of the antifreeze and motor oil because of that tow? Of course not, and it would be silly to think so. The same applies when you use programs like that as a means to determine power consumption and heat.

[SKYMTL]

I love the car anology but I think it is a little of both. First of all you can't run a program like FurMark or Linpack, run a product far outside of its normal parameters and then point to it and say: "it is consuming more power than the manufacturer stated!" Power consumption is a slippery slope that can vary from app to app or even from chip to chip within the same SKU.

What needs to be done is a broad range of tests from gaming to worst case scenarios like Furmark to normal gaming to Folding. That way we can have an idea of average power consumption since let's be honest; none of us will sit there all day while FurMark runs on our computers. What DOES matter to us is the efficiency or lack thereof in the tasks we use our computers for on a day to day basis.

There are many ways to generate an influx of interest to a site and one of those is to create a controversy which people talk about. This is exactly what is happening here since I feel that while this test is one of the more accurate when determining power consumption, so much more could have been done in order to get a better idea of overall power consumption.

[SNiiPE_DoGG]

I love the car anology but I think it is a little of both. First of all you can't run a program like FurMark or Linpack, run a product far outside of its normal parameters and then point to it and say: "it is consuming more power than the manufacturer stated!" Power consumption is a slippery slope that can vary from app to app or even from chip to chip within the same SKU.

What needs to be done is a broad range of tests from gaming to worst case scenarios like Furmark to normal gaming to Folding. That way we can have an idea of average power consumption since let's be honest; none of us will sit there all day while FurMark runs on our computers. What DOES matter to us is the efficiency or lack thereof in the tasks we use our computers for on a day to day basis.

There are many ways to generate an influx of interest to a site and one of those is to create a controversy which people talk about. This is exactly what is happening here since I feel that while this test is one of the more accurate when determining power consumption, so much more could have been done in order to get a better idea of overall power consumption.

on a side note, if someone were to do a full battery of tests like you suggest I would also like to see a test of the powerconsumption at full load using the stock cooler versus the power consumption at full load using a fullcover waterblock. Everyone says the cooler components run the more efficient they are, but the question is then how much more efficient

[ZOMGVTEK]

Is that why there is always an insane amount of scalding hot air coming from the back of these things?

[Eastcoasthandle]

Is that why there is always an insane amount of scalding hot air coming from the back of these things?

Any GPU that is forced to create a high load will produce what you suggest. So, I don't really understand your question. Any other time or set of circumstances would solely depend on your tolerance for heat.

[Barrok]

I think the car anology isn't all that great and here is why. Sure, you wouldn't ask a corvette to pull a boat but you could easily go find a vehicle (truck, suv) that CAN pull the boat. So, there is a way to take that massive load and pull it within specs of a vehicle. Can the same be said of furmark? Can you find a Graphics card (maybe the Quadro) that is capable of that load like you can find a car that is capable of the load as well?

I think its funny that we can find programs that stress CPU's, 8 threads 100% max but once you find a GPU program that 100% max stresses the GPU that all of a sudden thats a bad program and no one should use it because it could harm the card... weird.