Want to game on battery? Forget a dedicated GPU.

[Finn]

This is something I have been meaning to do for a while with my vortex 3, and its not something that I have ever seen any review mention (which seems very surprising, as going by the forum posts its knowledge that's not fully widespread) But I decided to benchmark integrated vs dedicated, mains vs battery. I used 3dmark vantage just because thats what I had installed and I wasnt particularly worried about exact score, more how they compare.

I'm testing a GTX 675m, i7-3720qm in the vortex3 chassis

(click scores for their results page)

	Mains	Battery
iGPU	3234	3258
dGPU	14072	1478

So it wasnt particularly surprising that dedicated on mains is vastly better than integrated but what was surprising was that the dedicated on battery was a lot slower than integrated - less than half the speed. One thing that did immediately spring to mind was how pointless and complicated optimus is. It might aswell purely switch between integrated/dedicated based on mains/battery rather than keeping a hardcoded list of apps. Since I dont care about power draw on mains and activating the dedicated on battery actually costs me performance.

One thing I would love to see is other people repeat this test on other GPUs, I'm curious whether lower power draw GPUs perform better when on battery. It would seem counter-intuitive, but i'm curious if when gaming on battery you are actually better with a lower spec GPU which doesn't throttle down so much on battery.

Edit: Okay run the same test on my Mrs' GT 630M 3603 on mains vs 3515 on battery, so despite being an extremely low end GPU, its actually better on battery than my 675!

edit2:
Scores so far

	Mains	Battery
i7-2670qm	1220	1238
i7-3720qm	3234	3258
GT 630M	3603	3515
GTX 670M	11045	1169
GTX 675M	14072	1478

 

[tom6561]

Was wondering about this because I'd noticed the horrific performance in games when on battery power, sometimes I hadn't realised the charger had come out and was struggling to understand why my performance had suddenly dropped so much. Anyway I went ahead and did the same as you, using 3dmark vantage. I wasn't sure how to do the table like you did, so I borrowed your formatting, hopefully it's worked... This is on my Skyfire II by the way, so it's HD3000 graphics as well.

	Mains	Battery
iGPU	1220	1238
dGPU	11045	1169

It is very weird that the dedicated graphics card is essentially completely redundant when running on battery power, there must be a way to turn off the GPU throttling at the cost of much battery life, you'd hope so anyway.

 

[marwoo578]

I believe the dedicated GPU throttles back or off on battery as the GPU needs more wattage to run than the battery can give it
Hence the need to plug the charger in

Higher the GPU wattage means will throttle back more which is why Finns Mrs' GT 630M performed better on battery as lower wattage ( throttled back less ) than his 675

 

[tom6561]

That does make sense, but it seems strange that the 675M performs worse than a 630M on battery. Presumably the 630M is more efficient at that wattage than the 675M so can stay at a higher clock, but you'd think that a gaming laptop would have a battery capable of supplying high power.

 

[marwoo578]

The 630m runs at 35 watts while the 675m runs at 100 watts
Be some battery to put out 100 watts

 

[tom6561]

But still, take my 670M, runs at 70 or 75 watts, can't remember which, say the battery can supply half of that which is roughly 35. My core clock drops from 600Mhz down to 71Mhz, surely that's not right?

 

[marwoo578]

Yeah does seem a huge drop esp when the VortexIII battery is supposed to be 76.96WH
I guess though with powering the CPU/Screen/HDD/Fans etc the amount left free will be a lot less

 

[tom6561]

Well that means it should be able to supply about 70 watts of power for an hour or so, I'd understand if performance was cut in half, but it's no better than integrated graphics.

 

[Finn]

I dont think its a particularly linear relationship, eg cutting the power in half doesn't mean half the speed, these figures show its most more than 50%

I understand WHY it happens, marwoo, what I am trying to do is quantify by how much and for differtent GPUS. For instance I have never seen a single laptop review benchmark performance on battery, until trying it myself I has assumed that performance on battery was proportional to performance on mains, eg double speed on mains = double speed on battery. This is evidently very much not the case. What I find surprising is that no-one has done this test and documented the results. You can find many sites comparing mains performance but not a single one comparing battery performance.

Say for instance you do want to game on battery, which GPU do you buy? I dont know and I dont know where to find that information. I was kinda hoping that more people would respond to this thread with figures but then it seems like most people dont actually care? Something else that I find surprising. Seems most people dont actually want to run their battery powered computer on battery, I guess most just want something they can cart between mains points?

 

[tom6561]

One of my friends has a GT555M, I'll try and get him to do a benchmark comparison and post it here when I have done.