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

[Eastcoasthandle]

See, that is the fault with your analogy. Truck companies TEST their trucks for maximum payload. Who cares if 98% of the people never reach that payload, or if the "regular" payload is alot less. The fact is the truck companies give you the maximum payload and the truck is 100% capable of that. Graphics card companies, according to this article, seem to give you a maximum wattage but not when the card is being pushed at 100%.

Actually no it's not the fault of my analogy or the other poster's analogy. The car analogy is just fine when explaining the actual load used between that program and daily programs and games used. After reading this post it's clear to me (at least it's my opinion) you simply don't understand what's being conveyed. Regardless, the difference between the load from daily programs and games (IE a car with 4 occupants) vs a stress test/torture test program (IE: a car pulling a boat) is quiet clear for some of us.

[STEvil]

Actually no it's not the fault of my analogy or the other poster's analogy. The car analogy is just fine when explaining the actual load used between that program and daily programs and games used. After reading this post it's clear to me (at least it's my opinion) you simply don't understand what's being conveyed. Regardless, the difference between the load from daily programs and games (IE a car with 4 occupants) vs a stress test/torture test program (IE: a car pulling a boat) is quiet clear for some of us.

Except the load being implied is far greater than what the car should ever attempt to handle where the truck can pull a load but the load being given is larger than average.

There are games and programs out now that can create as much or nearly as much as furmark.

[Eastcoasthandle]

Except the load being implied is far greater than what the car should ever attempt to handle where the truck can pull a load but the load being given is larger than average.

That is the essence of the analogy IMO. Or at least that's how I interpreted it.

There are games and programs out now that can create as much or nearly as much as furmark.

I can't speak for everyone but I haven't come across such a program and game that does.

[Beefy22]

Based on the example given it's no longer a broad field but a very specific on. However, to each his own, what's implied remains the same.

If by "what's implied remains the same" means you're satisfied with a product that can not operate as advertised 100% of the time ok, you got me.

That is the essence of the analogy IMO. Or at least that's how I interpreted it.

Except cars are not designed to tow boats nor are they advertised in such ways... Graphics cards are designed to run games and fur rendering algorithms are part of some games.

A better analogy would be... Two trucks attempt towing the same 3 ton 28' Donzi up a steep hill and one of them can't make it over the hill. The hill with a steep incline might not be a normal everyday occurrence but you would expect your truck to handle it.

One more time, our load-results are that high because we used FurMark as load. Have look at this comparison of tools. As you can see FurMark is much more demanding than 3DMark06. I think these 3DMark06-results should be pretty much what you expected, right? But you can also see, that there are many games that are more demanding than 3DMark06, not only FurMark.

We have one :P.
http://ht4u.net/reviews/2008/messtechnik/index8.php
And we are currently working on an separat article that will deal with the temperatures of the 4870 (and the high voltage regulator temperature). And yes, we´ll do some pictures with our thermal camera .

Greetings,
Leander [HT4U]

I look forward to that!

[Eastcoasthandle]

If by "what's implied remains the same" means you're satisfied with a product that can not operate as advertised 100% of the time ok, you got me.

It's obvious we won't agree here. Therefore, there is no need to resort to "prodding" the issue.

Except cars are not designed to tow boats

Neither was my Olds Firenza (back in the day) when I towed that Explorer but it was done. What is considered "not designed for" is what is expressed by some when a stress test/torture test program is used to determine power consumption.

nor are they advertised in such ways... Graphics cards are designed to run games and fur rendering algorithms are part of some games.

Isn't the article implying/stating in part that the TDP of those video cards are not accurate? If so, then it's only natural to inquire about the test methods used. And, we can only disagree on that the power consumption of "fur rendering algorithms are part of some games" as being the same.

[STEvil]

I imagine grass rendering algorithms would look very similar, and possibly even leaf ones to an extent ones..

edit, could have sworn I replied earlier with a couple examples. Anyways, Folding@Home and Flatout Ultimate Carnage tend to push GPU's very well.

[Brianwulf]

Yeah! Why not? Hummer stopped making H1's in 2006 because we decided that they slurped way too much gas.

Furthermore, Hummer sold 54,052 H2's in 2006 and only 6,095 H2's in 2008. A huge difference, eh?

I'm sorry to say the biggest reason for that is that in 2008, even people with very good credit were having trouble getting car loans. Also there was the very high gas prices.

[STEvil]

Due to release of H2 and H3 as previously stated in this thread.

H1's were never intended to be civilian vehicles.

[Manicdan]

This page is the best part i think:
http://ht4u.net/reviews/2009/power_c...ics/index9.php

a change of 20 watts just by having better cooling. id love to see a review where they see the power coming out of the wall using stock cooling on a quadcore and high end video card, then replace it with a water cooling setup and see the difference. id bet people would buy a water cooling kit when someone says it saves 10-20% of their computers power. then i bet it could do even more if they undervolt it at stock clocks.

[cegras]

This page is the best part i think:
http://ht4u.net/reviews/2009/power_c...ics/index9.php

a change of 20 watts just by having better cooling. id love to see a review where they see the power coming out of the wall using stock cooling on a quadcore and high end video card, then replace it with a water cooling setup and see the difference. id bet people would buy a water cooling kit when someone says it saves 10-20% of their computers power. then i bet it could do even more if they undervolt it at stock clocks.

While at first that also seemed utterly ridiculous to me, you can't forget that silicon will behave very differently at different temperatures ..

[saaya]

One more time, our load-results are that high because we used FurMark as load. Have look at this comparison of tools. As you can see FurMark is much more demanding than 3DMark06. I think these 3DMark06-results should be pretty much what you expected, right? But you can also see, that there are many games that are more demanding than 3DMark06, not only FurMark.

hey kylian, nice to have you on xs

i remember that furmark heats up gpus more than anything else, but im quite surprised that its supposed to heat up high end cards almost 100W more than games?

[STEvil]

saaya - check the two replies previous to yours. Maybe the issue is temperature PLUS utilization

[SKYMTL]

The issue is that FurMark is still quite CPU bound, even at the higher settings.

[Chruschef]

i don't get why this is such a big deal...

GPU's have to power an entire circuit board, RAM, GPU. GPU cores also designed to handle more heatload, than fragile CPU's so it isn't the same as a "high-TDP" processor... though, these recent GPU's are quite power hungry.

[AbelJemka]

And we are currently working on an separat article that will deal with the temperatures of the 4870 (and the high voltage regulator temperature). And yes, we´ll do some pictures with our thermal camera .

Greetings,
Leander [HT4U]

http://translate.google.fr/translate...ml&sl=fr&tl=en

[largon]

[IMG]http://ht4u.net/images/reviews/2009/power_consumption_graphics//4870_wandler_3.png[/IMG]

Now it is no supprise anymore, why AMD lowers the load synthetically. The card simply can not not bear the load FurMark creates. While the GPU-Temperature is absolutely uncritical, the voltage regulator temperature is not. Meanwhile the datasheet of the voltage regulator is available and Vitec specifies the temperature to 125 degrees.

Via FurMark we reached within a few minutes temperatures from about 125 degrees. Thus we can say: The Radeon HD 4870 definitly has a problem with its voltage regulators when using FurMark. Surely FurMark is a worst-case-scenario, but it points out that the voltage regulators on the Radeon HD 4870 work near its maximum specification.

http://ht4u.net/reviews/2009/power_consumption_graphics/index8.php

That "max 125°C specification" applies only to the Vitec 59PR9853 multiphase inductor. The actual voltage regulators which the "VRM phase [number] temperature" -figures in Rivatuner come from are the Volterra VT1165SF chips - and their max specified operating temp is unknown (datasheet is not public). My point being, those Vitec inductors should run cooler than the Volterra chips since the coil surely consumes only some minute amount of energy (~5W), and as such their temps must be lower than the 120-125°C of the Volterra's which generate a considerably higher amount of heat.

[Kylian]

hey kylian, nice to have you on xs

i remember that furmark heats up gpus more than anything else, but im quite surprised that its supposed to heat up high end cards almost 100W more than games?

Have a look on it in a procentual way. Difference between 3DMark06 and FurMark are about 20-30%. If you keep in mind, that even many games are causing a notably higher power consumption than 3DMark06, those values aren´t to high anymore IMHO. But you still need to handle FurMark as an worst-case-scenario, though. If you don´t believe it, just check it your self, get FurMark and start some little testing .

The issue is that FurMark is still quite CPU bound, even at the higher settings.

Sorry, but this is one more time simply not true. We had look at this, and tested Power Consumption on 4870 X2 with different CPU-clock-Speeds (1,8 vs 2,6 GHz). The differences in Power Consumption had been in the range of meassuring tolerance.

That "max 125°C specification" applies only to the Vitec 59PR9853 multiphase inductor. The actual voltage regulators which the "VRM phase [number] temperature" -figures in Rivatuner come from are the Volterra VT1165SF chips - and their max specified operating temp is unknown (datasheet is not public). My point being, those Vitec inductors should run cooler than the Volterra chips since the coil surely consumes only some minute amount of energy (~5W), and as such their temps must be lower than the 120-125°C of the Volterra's which generate a considerably higher amount of heat.

But this one is really true. We updated the page accordingly. Thanks a lot for your hint, largon.

Greetings,
Leander - HT4U

[Manicdan]

If you don´t believe it, just check it your self, get FurMark and start some little testing .

did you imply that i should fry my computer? i hope not

[Kylian]

Nah, you won´t kill your card just with little testing with FurMark. There´s no need to run FurMark for long time, just to see that it causes a much higher Power Consumption than 3DMark06.

[SKYMTL]

Sorry, but this is one more time simply not true. We had look at this, and tested Power Consumption on 4870 X2 with different CPU-clock-Speeds (1,8 vs 2,6 GHz). The differences in Power Consumption had been in the range of meassuring tolerance.

Try a HD 4870. At high res you shoud see between 20-25% CPU utilization on a quad core. If not, please post your FurMark settings since I have yet to see a single GPU card not post high CPU utilization in FurMark.

[Kylian]

Sorry, but i don´t get your point. You reffer to a CPU-Utilization from about 20-25%. So where is the problem? As long as CPU-Utilization is as low as you stated, CPU-Perfomance shouldn´t have any influence to our measurements since it is simply no bottleneck. And we additionally confirmed this with our meassurments at different cpu-clocks.

Greetings,
Leander - HT4U

[Shintai]

25% CPU utilization on a quad usually means singlethreaded and 100% load on 1 core.

[Kylian]

No, usually this is not the case. Just if the Programm is perfectly multi-threaded you´ll get from 25% CPU-Utilisation on a quad to a 100% CPU-usage on a single-core. But nevertheless, we used dual-core-processor in our test. CPU-Performance didn´t have any influence on Radeon HD 4870 X2 as we meassured, so why should it suddenly influnece consumption on slower cards like the Radeon HD 4870?

[SKYMTL]

No, usually this is not the case. Just if the Programm is perfectly multi-threaded you´ll get from 25% CPU-Utilisation on a quad to a 100% CPU-usage on a single-core. But nevertheless, we used dual-core-processor in our test. CPU-Performance didn´t have any influence on Radeon HD 4870 X2 as we meassured, so why should it suddenly influnece consumption on slower cards like the Radeon HD 4870?

High CPU utilization will affect any card's power consumption numbers since the card may not be being used to its full potential. Basically, in some instances it may be waiting on the CPU for information to be processed; call it a bottleneck if you will. This will in turn affect power consumption since the GPU would not be working to its fullest potential.

I know you didn't test like this but CPU utilization also becomes an issue when measuring overall system power consumption. The CPU can highly influence the overall consumption numbers of higher end cards; some of which may be bottlenecked in certain applications. Hope that makes sense.

[Parad0x]

Trislot coolers cant be far away.

what do we have here?