HD4000 performance with (upscaled) retina?

[0x000000]

Hi everyone,

I and many others are very curious about the performance when forcing the rMBP to use the HD4000 with gfxCardStatus. And very interesting would be how the mac performs when using the HD4000 and upscaled resolutions, as they should theoretically be more taxing.

Has anyone got experiences in this? Some claim it's no problem, some that it's sluggish. I just don't know what to believe and it's of course very relevant for many, with the 13" rMBP on the horizon.

Thanks a lot, it'll be an enormous help.

.

 

[philipfreire]

I'm not sure if a driver update to the HD4000 would fix this or not but I am experiencing sluggish results on any scaled retina resolution.

I have a 2.7/16GB/768 (Maxed out specs)
I also have a week left on my 14 day return, and will be shipping it back on tuesday after the new iMacs are released. I'll just wait for the Haswell update. It will be designed to support 4k resolution screens without any issues and the iGPU will be capable of running this retina display smoothly in any normal operations.

 

[0x000000]

I'm not sure if a driver update to the HD4000 would fix this or not but I am experiencing sluggish results on any scaled retina resolution.

I have a 2.7/16GB/768 (Maxed out specs)
I also have a week left on my 14 day return, and will be shipping it back on tuesday after the new iMacs are released. I'll just wait for the Haswell update. It will be designed to support 4k resolution screens without any issues and the iGPU will be capable of running this retina display smoothly in any normal operations.

With some further reading this seems to be quite the case. Or has anyone made any clearly different observations?

Some reported hiding the dock or putting it to the left or right side of the screen increases ui speed a bit, did you try that?

 

[bill-p]

Not seeing any sluggishness.

rMBP has issues with EFI that may cause more performance issues than the HD4000 GPU itself.

 

[0x000000]

The current results so far are more confusing than I thought with just as many people experiencing sluggishness and people who say it's fine...

I don't know what I expected.

 

[nontroppo]

No sluggishness, apart from very rare stuttering, and I run mostly in HiDPI1920 (the "worst" of the scaled options). Also of note, the memory bandwidth (how much pixel throughput the GPU handles) is about 16x greater than the demands of the worst scaled resolution, so there is no technical reason the HD4000 can't handle retina resolutions.

 

[Exana]

It's not just a GPU memory bandwidth question. What takes times is downsizing each oversized image (ie : 3840x2400) to 2880x1800. This seems to be handled by CPU not by the GPU.

BTW, HD Graphics 4000 and GeForce GT 650m run fine until 1680x1050. 1920x1200 is not great all the time.

 

[0x000000]

Went to the store today and tested a low end rMBP against a classic. Safari speed was identical, the verge scrolling wasnt super smooth on both, but still 'very' smooth.

Switching spaces was slightly less nice on the rMBP. The speed is the same: the rMBP doesn't take longer than the classic, it's just as if it uses an ever so slightly lower framerate. The difference is negligible though and only barely visible, like comparing 30 to 60 fps. I expected a more visible drop of performance when using scaled modes, but the performance stayed the same...

The difference doesnt feel great at all, it's visible only in direct comparism, but its so enormously cosmetic that I'd not sacrify retina resolution, silent fans or battery time for those few fps

I'll test it again on slower macs to better assess the speed.

 

[stevelam]

No sluggishness, apart from very rare stuttering, and I run mostly in HiDPI1920 (the "worst" of the scaled options). Also of note, the memory bandwidth (how much pixel throughput the GPU handles) is about 16x greater than the demands of the worst scaled resolution, so there is no technical reason the HD4000 can't handle retina resolutions.

because scaling is done via software. theres your reason.

 

[AirThis]

Went to the store today and tested a low end rMBP against a classic. Safari speed was identical, the verge scrolling wasnt super smooth on both, but still 'very' smooth.

Switching spaces was slightly less nice on the rMBP. The speed is the same: the rMBP doesn't take longer than the classic, it's just as if it uses an ever so slightly lower framerate. The difference is negligible though and only barely visible, like comparing 30 to 60 fps. I expected a more visible drop of performance when using scaled modes, but the performance stayed the same...

The difference doesnt feel great at all, it's visible only in direct comparism, but its so enormously cosmetic that I'd not sacrify retina resolution, silent fans or battery time for those few fps

I'll test it again on slower macs to better assess the speed.

+1. Sums up my experience perfectly.

 

[nontroppo]

because scaling is done via software. theres your reason.

Do you have a reference to an article or something that hardware acceleration is disabled on the HD4000? I don't see significant CPU hit at all when monitoring Window Server and moving or resizing windows. Apple has spent years increasing the use of the GPU for quartz window drawing and it would be odd if they wen't back to using a software drawing engine...

 

[Tsuchiya]

I was wondering about this on the 13" rMBP, but I think it should be fine tbh.

The resolution is the same as my 27" monitor hooked up to a 2011 i5 Mac Mini with Intel HD 3000. The setup runs flawlessly.

 

[0x000000]

Tested it against lower geared MacBook Airs and Pros, Mac Pros etc. the difference in smoothness is apparent. The rMBP was pretty much the only more recent mac I could find that showed those performance "problems" when switching spaces.

Again, this doesn't change productivity and may be by design... but it may still serve as an indicator for how much performance the retina display actually swallows - and that it does.

 

[ImperialX]

I use the HD4000 in my rMBP in 3840x2400 quite frequently and I've never noticed a problem with normal use. When I use anything intensive, GT650 kicks in and the performance is still good.