GPU and battery life: new Intel HD x Nvidia (with Optimus) x ATI

Hi.

I was just wondering and couldn't find this kind of comparison.

I'm looking forward to get one of the new Sandy Bridges laptops and noticed that it already comes with a fairly good GPU.
I don't play much, but eventually, with a new pc like that, I may like to give it a try.

But the gaming capabilities are not my concern here. Every benchmark I've seen already goes over and over this. I would like to see how Intel's new HD 3000 would differ from better-gaming-performance gpus, in the battery life issue, during regular use (non-gaming).

The option I have in mind, for my new notebook, has NVidia GeForce GT540m which has an interesting technology called Optimus, that would optimize its battery needs. But then I think that it still could be considerably more than using just Intel's internal GPU.

I would like to know from you if there will be significant difference by using these (Intel Hd 3000 and Nvidia GT540m) or other GPUs on a same notebook (let's say a 15.6in, new Sandy Bridge chipset), like an hour of battery life or so.

Best regards!

 

Anyone?

Isn't it a relevant aspect to consider for the next work/leisure equipment?

Or is it a dumb question?

 

The problem is that the GPU is only one portion of battery life; the hard drive, the screen, and many other components in the system will also affect the battery life. 2 notebooks that have the same CPU and GPU combinations can have vastly different battery lives, depending on the size of the battery installed, and the other components in the system. So in effect, there's no way to generalize the way you're asking. It's like asking what's the difference in gas mileage between a 200 HP and a 300 HP engine; the 300 HP engine will probably get less gas mileage, but the actual gas mileage will depend on the kind of car it's mounted in.

For Optimus, the Intel IGP is always on, and it turns on and off the GT540M at need. The GT540M, pretty much by definition, will always draw more power than the Intel IGP. So a computer with Intel IGP will have a longer batter life than the same computer with Optimus, and the Optimus solution will have a longer batter life than the same computer with GT540M only (probably; if you continually stress the Optimus system so that the GT540M is always on, it may have less battery life than the GT540M only solution, which could leave the IGP off).

 

Judicator, thanks for the reply.

I know, there are many factors for the battery life but my hypothetical question was about varying only the GPU, the other characteristics fixed.


So you think that the Optimus technology from GT540 would not add an extra battery eating footprint, it will be the same as just Intel IGP alone meanwhile the NVidia GPU is asleep?

 

As long as the GT540M is not being used, then yes, I think the power drain from it will be fairly trivial. Note that the usage of the GT540M is _not_ up to you. Optimus decides for you whether or not a program should use the GT540M or the Intel IGP, so basically unless you turn on your computer and never use any programs, or use only the most basic of programs, the Optimus solution _will_ end up using more power than the IGP only version, because the GT540M will get turned on at some point and draw more power.

 

I believe I can disable NVidia GPU on the bios, right?

Also, Clevo W150HNQ seems to have a button to switch between both GPUs while using Optimus. It doesn't work on Linux, though.

And do you have a clue on the difference on using just IGP or just GT540m (leaving Optimus variations aside) in the battery life on a regular notebook?

 

This feature is system dependent.
For instance, Alienware's M17x-r3 (with Optimus) does not have an option to disable the GPU in BIOS.

You can downclock the card via software though.

 

Depends on your BIOS. Lenovo T500/T400 can manually disable GPU in BIOS.

 

No, because it depends on the system, and what's being done. GPUs don't have a fixed draw, they'll draw more or less power depending on how hard they're being worked, so the numbers will vary.

 

ok. I though that it could be estimated somehow.

Today we see notebook that are expected to last 8 hrs (as manufacturers say) and others that are expected to last 2 hrs, both having similar configuration and sizes. I was just thinking that the GPU could be one of the greatest causes for such differences.

I have looked for many benchmarks between GPUs, any of them cover this matter. They are just concerned about gaming, gaming,...

 

When you're looking at battery life, you need to compare the battery life of an entire notebook.

Notebook_Model_A gets 8 hours of battery life.
Notebook_Model_B gets 2 hours of battery life.
Notebook_Model_C gets 4 hours of battery life

That is the information that is important. Knowing the exact GPU in those Notebook_Model_A/B/C is irrelevant, once you know the overall battery life that the entire notebook gets.

 

I see.
I'm starting to accept it as a very chaotic problem and that cannot be described in modular contributions. Maybe there is more things involved and I just neglect then.

Although I still believe it could be benchmarked and therefore predicted, within some control of the other variables (I believe the most important ones would be the notebook size, the CPU and the kind of hard drive and of course, its battery).

 

Which notebook? Was it the Acer Aspire 5750G? I was very tempted, but the screen resolution kept me away. (luckily!)

 

I'm actually looking forward to a Clevo W150HNQ (like the Sager NP5160). It has a 15.6in 16:9 screen with a Full HD resolution

And, just to explain those questions I'm making, I'm not concerned on the battery lifetime by its absolute value for this very model. I'm concerned on the impact it will have to use one or another GPU on a same hypothetical notebook model.

 

What's the battery life on that one?

edit: argh, don't tempt me!

 

I don't know, it hasn't been released yet (as far as I know).

It is a "gaming" notebook which uses the new Intel Sandy Bridge platform. Very good value though.

 

Unfortunately, where I live, the P151HM1 with the 460m (no Optimus) only costs $100 more. Same for the upgrade to the 1920x1080 screen.

That said, the Acer 5750G also costs about the same.

 

Well, here, just as an example, then, this is XoticPC's page of Sager notebooks. The Sager NP5160 (Clevo W150HN) is first on the page, and it's listed with 3-5 hours of battery life. You will note every system on there is listed with a range for the battery life, whether or not it has Optimus. This is fairly normal, because, as stated, battery life varies wildly depending on what you're doing. If you read the reviews that NBR does, they do a very specific battery test; they set the brightness at 75%, wireless on, and they have it periodically refresh a web page. That's where their numbers come from. If you do something more intensive, like play a game or run a "real" program, the battery life will go down as a consequence, because the battery will drain faster (as the GPU and CPU use more power to run the programs, the sound card/speakers will draw power to work, etc, etc.).

 

Yeah, that entirely true.

But those variations on the battery life are mostly due to cpu usage/scalling and hard disk accesses, except for the gaming or maybe video playback... so, it is possible to quantify what is consuming power from the battery, based on what the system is doing.

Besides that, one GPU will consume more than the other, both on trivial tasks or when gaming and is that difference that I'm trying to get an idea of. How much does one take advantage over the other on battery spending issue. Maybe it is very significant, or maybe Optimus does a great job and they even out...

To do so in a more exact way, it will be needed a very defined test, as the example you gave. I do have tried to find benchmarks like that but couldnt find. Then, I though that maybe people that are more into this kind of benchmarking stuff would have a more general idea on integrated or discrete GPUs and could help on this question, but it seems that it is not that trivial.

 

The idea for the benchmark is to test two identical systems (except by the GPU) and run several of these tests.
It would be easy to do on a case like integrated GP x discrete GPU, but to compare among different systems could be trickier to do, as even the internal structure of the notebook should impact on cooling and therefore fan powering.

Maybe I'll try that, if I get such notebook