MBPR vs MBP w/SSD: which is faster

[derbothaus]

But what Anand implied, and people have repeated here is that the rMBP is close to its hardware limits; this is patently false.

It's only "patently false" if you run Windows. If you are running OS X the demands are what they are. You need more. Whether or not it SHOULD be able to do it. This may change. Hoping, wishing , and theory does not make any issues go away.

I think this is not a straight forward estimation. The rMBP has a higher clocked GPU, and a much better thermal system that stops the GPU/CPU from throttling. That means that it is certainly possible over the course of working on real-world tasks, the rMBP maintains performance as the cMBP slows substantially. That will certainly catapult the rMBP above the cMBP running on an external display, and will result in tit-for-tat performance on the internal display.

The thermal system does not even come into play until something like a video game is being played or major GPU based rendering or CUDA. Drawing the screen does not throttle anything below usable value (with or without an external display). We are not talking games. Macbook and "gaming" is kind of a ridiculous thing to get into anyway. This is just straight up user experience in browsers and OS elements and animations.

 

[nontroppo]

It's only "patently false" if you run Windows. If you are running OS X the demands are what they are.

I think we are saying the same thing, this is a software problem (the operating system being part of a software equation).

Part of my work involves building psychophysical stimuli for vision research. I use OS X Matlab via an OpenGL wrapper, and I can benchmark my drawing code by disabling vertical sync and measuring fps. If for example I generate 6 patches of sinusoidal modulated gratings using GLSL shader programs and variable cosine smoothing across edges, the fps do not vary substantially across the HiDPI modes, I get ~285fps @ 1440, ~275fps @ 1920 and ~310fps @ native2880. This is a fairly demanding task as I am controlling the GLSL programs via an interpreted language over a single thread, and for comparison my 2010 15"MBP does ~180fps. Even when driving a reference CRT at 120Hz we still have lots of headroom, much more than my 1.8year old MBP.

The thermal system does not even come into play until something like a video game is being played or major GPU based rendering or CUDA.

Try importing large batches of RAW photos into Lightroom, running a computational model in Matlab or edit video in Final Cut Pro X. There are many cases when professional machine will be working at maximum capacity and throttling may come into play, not just for those playing games!!! My rMBP handles these situations exceptionally well, and certainly flies through tasks cooly that turn my 2010 MBP into a fan-spinning lava ball. Thermal management is critically important to anyone who runs demanding tasks, not just gamers shooting monsters in dungeons...

 

[derbothaus]

Thermal management is critically important to anyone who runs demanding tasks, not just gamers shooting monsters in dungeons...

This is why I run a Mac Pro as my main rig.
..and yes, we are saying the same thing.
I also shoot monsters in dungeons.

 

[terraphantm]

Simple fact and common sense : Resampling consumes cpu and gpu resources, and in the case of dGPU, more data has to be moved from RAM to VRAM...

I have nothing against the retina display. I just don't like the sealed parts of current rMBP...

Just how much CPU do you think it consumes? We're talking maybe a .5% impact on overall performance

 

[stevelam]

Just how much CPU do you think it consumes? We're talking maybe a .5% impact on overall performance

thanks for your made up number. very helpful and not pointless at all.

the cmbp will definitely be faster especially since you can install a second hard drive in it. This allows you to use it as a scratch disk for various programs that utilize them (photoshop, audio DAWS, etc etc). you can't do that with the rmbp.

 

[nontroppo]

thanks for your made up number. very helpful and not pointless at all.

So I've just measured this. Run a rMBP at 1920x1200 HiDPI *and* a 27" Cinema display. Run Quartz debug to measure FPS and Activity Monitor and focus in on WindowServer, the process which does the work of managing the drawing and compositing. I set up a transparent terminal and drag the activity monitor window around as fast as possible, forcing maximum UI compositing. FPS on both screens are 60fps and CPU is:

Internal: ~16%
External ~13%

At 2880 and 1440hiDPI the CPU use is ~13.5%, identical to the external. Screenshots caught using timed mode with grab.

So ~3% of one core, or ~0.75% of total CPU resource is consumed driving the worst of the HiDPI modes during window compositing across transparent windows. What I haven't measured is GPU resource usage, but as the bandwidth of both the Intel and NVidia chips is so much higher than the display demand, that is unlikely to be substantially different.

... since you can install a second hard drive in it. This allows you to use it as a scratch disk for various programs that utilize them (photoshop, audio DAWS, etc etc). you can't do that with the rmbp.

Finally a useful statement! The cMBP certainly benefits on-the-road without access to external drives. My 2010MBP has an SSD + 750GB scorpio black, and it is great to carry so much storage without the need for any external drives. Such a benefit is negated if using a thunderbolt attached drive however, which many professional users would travel with anyway...

 

[chrslbrt]

Stating that the retina MacBook's hardware is incapable or even I'll suited towards running it's display is a dated notion, and a bit absurd.

Mine has absolutely no problems whatsoever and I run activity monitor from time to time to see if it's even breaking a swear, which it doesn't.

For all we know, that Anandtech reviewer could have had a defective model, not to mention he used an older operating system.

People will listen to one guy on the Internet despite dozens of people saying otherwise.

Just as stupid as the lg vs Samsung screen scare.

 

[chaseychasem]

Thank you for that enlightened appraisal of empirical evidence and your mastery of the possessive pronoun.