Retina MBP UI performance - with video

[jcljk]

Oops, just got which figure you're talking about. I'd argue though that what activity monitor is reporting there is irrelevant because at 2880x1800 it's perfectly smooth and at the 1920x1200 "setting" it's extraordinarily choppy. So it's got to be a bottleneck somewhere else in the software stack than simply the core's utilization while scrolling. The core may fly up like that because of the high resolution of touch input the mac trackpads collect too ( which is why scrolling is so different and better on the mac trackpad than almost any other input device ), totally independent of the content on the screen itself scrolling, which is more likely to be handled/processed by the video card ( my guess, don't have data to back this up ).

----------




At 2880x1800 there's no scaling happening, that's the true/native/actual resolution of the retina screen. So there is no extra calculation happening to show things on the screen. Hence, it's faster. The video card inside is capable of handling most things at that resolution very quickly...

At any other resolution however, let's take the default scaled resolution ( 1440x900 ), what happens is that for each frame per second ( imaginary, it's not really measured that way ) the video card has to render everything like it normally does, meaning figure out what to show for each pixel, and THEN do another full calculation for each pixel to drop it down to simulate the 1440x900 from its first rendering at 2880x1800. I'm doing a terrible job of explaining this but in a nutshell: It's actually a lot less work to use the much higher resolution of 2880x1800 than any scaled resolution that OS X ships with ( the only options available ).

I'm not an Apple App Developer so I don't have access to Mountain Lion. Though I wants it

i see.
How's the performance on Adobe Photoshop for 1440x 900 and 2880x1800?
Opening many Pictures and dragging one of the big one around. or scrolling up and down ?

Here's the link where you can download ML DP4. http://imzdl.com
Make sure you choose Developer Preview 4

 

[jcljk]

Yeah, here you go. Anything else you want to see?

http://www.youtube.com/watch?v=bkVnUoNfu5Q

Give it a few minutes to process HD

hi Ken,

did you get a chance to test adobe photoshop ?
www.theverge.com is a heavy site and can be quite laggy in RMBP.
Since you have updated to Mountain Lion, Please help to test the performance on this site.

thank you

 

[bleomycin]

hi Ken,

did you get a chance to test adobe photoshop ?
www.theverge.com is a heavy site and can be quite laggy in RMBP.
Since you have updated to Mountain Lion, Please help to test the performance on this site.

thank you

Please try theverge, scrolling quickly on my rmbp on lion is just as smooth as in your video for engadget. The Verge really, really chugs. Thanks!

 

[Vicobe]

I got the CPU % up to 100% between two Safari processes, however, the CPU User% was only about 12%. Makes me think Safari is only using one core.

I can get my 2.6 GHz (16GB RAM) RMBP up to 100.7% CPU on the Safari Web Content process. The Safari process is only up around 15% at the same time.

Thank you all. My assumption is justified: Safari eats 100% of one core during scrolling on RMBP when using any of standard display modes (except native display resolution 2880x1800).

I'd argue though that what activity monitor is reporting there is irrelevant because at 2880x1800 it's perfectly smooth and at the 1920x1200 "setting" it's extraordinarily choppy. So it's got to be a bottleneck somewhere else in the software stack than simply the core's utilization while scrolling. The core may fly up like that because of the high resolution of touch input the mac trackpads collect too

We can only assume what happened there internally. I think 100% CPU (1 core) is used because it use more CPU for showing web page content and using less GPU for that (wrong coded by developers).
This RMBP is not first computer in history with such high resolution of display, because before at least apple users was used Cinema/Thunderbolt Display with also almost same high resolution. The difference is in scaling and rescaling - Lion in RMBP is working not in native resolution, simulating different modes (1440, 1920 and etc) all except native display resolution 2880x1800 in which as I understand no any choppiness during scrolling. So RMBP with shipped Lion is the first time in worlds history of apple with such kind of rescaling output to user. I hope that next OSX (Mountain Lion) will bring smooth scrolling not only for special apple apps like Safari, but any third-party app will work smooth and fine.
At least this puppy (RMBP) have quad core processor which can (should!) handle such things if properly used by software/drivers developers.

 

[seaw]

At least this puppy (RMBP) have quad core processor which can (should!) handle such things if properly used by software/drivers developers.

1. The 27 inch iMac now exists for some time and was first driven by an Intel Core2Duo!!!!!
2. It has just A FEW pixels less than the MacBook Pro Retina!!!!!
3. It could handle this high resolution just fine!!!!!

Why should this new SUPERMACHINE of a MacBook Pro Retina a couple of generations later not be able to handle its pixels with ease?????

Sorry, but Apple just programmed a crapy upscale downscale process!
That new Notebook just should have come out together with Mountain Lion next month.... that's it!!!!

 

[pmau]

Whatever the specs are ...

When 1920x1200 is used, the whole content is rendered into a 3840x2400 pixel texture and then scaled down to 2880x1800.

This is stupid, the specs are not relevant. It's a waste of resources just to show off a new product.

Apple has repeatedly claimed to work on resolution independence. On the iPad and iPhone they dictated integral scaling (2x).

On the MacBook they realized that they cannot get away with that because people need more screen estate to work properly.

This is really a half-baked approach to a problem that is simply papered over by massive use of hardware resources.

 

[kirky1234]

We can only assume what happened there internally. I think 100% CPU (1 core) is used because it use more CPU for showing web page content and using less GPU for that (wrong coded by developers).

At least this puppy (RMBP) have quad core processor which can (should!) handle such things if properly used by software/drivers developers.

Just for my understanding: The problems occure because the software developers wrote their programms (safari) in a way, that they may only use 1 core. If they would just allow them to use multiple cores, the cpu could handle the scrolling just fine? What really confuses me is that even with the nvidi graphics card forced on, it can't handle the scrolling. On the other hand, the same nvidia card can handle diabolo 3 with highest possible settings at 20 fps....

 

[Vicobe]

kirky1234, even when you switch in RMBP to dedicated nvidia GPU, its still not enough power because it uses same 1 core of CPU at same 100%. There is 2 solutions in future for developers: 1) use more CPU cores to draw items on screen 2) Use GPU where is possible.
As I understand Apple will try to use more GPU in Mountain Lion for this and Safari there work more smoothly than in Lion. Bot what about other apps? Lookin on user reports, it is seems like iTunes still crap on Mountain Lion. And this tells me that also other (and third party) apps also should be recoded by developers for retina display and this is sad, because using GPU or more CPUs(cores) to draw something on screen is not easy for developers, the using single CPU and not use GPU for simplification for years, but now "retina display" is some kind of bottleneck...

I hope that that rescaling mechanism used by Apple for HiDPI modes ("best for retina" and all scaled modes except native display resolution 2880x1800) may be rewritten by apple developers to use more GPU and less CPU and in this case all apps will take benefit of GPU and will work smoothly.

 

[davidmcguigan]

Thank you all. My assumption is justified: Safari eats 100% of one core during scrolling on RMBP when using any of standard display modes (except native display resolution 2880x1800).

No that's incorrect. Look at my 2880x1800 screenshot. It still uses 100%+ on 2880x1800. Identical resource utilization, different results.

 

[davidmcguigan]

I don't know about that

Whatever the specs are ...

When 1920x1200 is used, the whole content is rendered into a 3840x2400 pixel texture and then scaled down to 2880x1800.

This is stupid, the specs are not relevant. It's a waste of resources just to show off a new product.

Not necessarily. The reason they do that isn't to "show off". It's actually because when you render any non-native resolution on any flat panel monitor the text becomes blurry and semi-unreadable. What Apple has done here is actually quite brilliant ( the engineers, it would seem, just didn't figure out how to make it fast in time for the announcement / shipping date ). By rendering things at twice the "simulated" resolution and then down calculating they've actually made the text quite crisp ( maybe not perfect, but literally 20x better than any non-native resolution I've ever seen on any other flat panel, which is a lot ) and the resolutions usable. I imagine they're using a downscaling algorithm that's actually more involved than simply dividing and rounding, but it uses readability-enhancing techniques to detect and improve legibility for text on the screen. This is one time where I think Apple's is actually doing something extremely innovative, just not necessarily doing it well yet.

----------

Just for my understanding: The problems occure because the software developers wrote their programms (safari) in a way, that they may only use 1 core. If they would just allow them to use multiple cores, the cpu could handle the scrolling just fine? What really confuses me is that even with the nvidi graphics card forced on, it can't handle the scrolling. On the other hand, the same nvidia card can handle diabolo 3 with highest possible settings at 20 fps....

Kind of. My personal opinion/hypothesis is that scrolling web pages in Safari is offloaded ENTIRELY ( 100%, NO WORK HAPPENING ON THE CPU ) to the video card. The 100% process utilization you see in activity monitor is because during multitouch scrolling a CPU core focuses very heavily on very very granularly tracking your finger movements including acceleration, direction change, etc. at a super high resolution to attempt to realistically interpret your acceleration, momentum, give the page a feeling of a sense of inertia, etc. That hypothesis would seem to coincide with the scrolling being super smooth at 2880x1800 with the same reported CPU utilization in activity monitor as any of the scaled resolutions and previous models where scrolling is smooth at lower resolutions. But, it's just a hypothesis and doesn't really help our situation any ( unless an Apple engineer reads my initial suggestion and is like OMG WHY DON'T WE JUST HAND-OFF THE DOWNSCALING CALCULATION TO THE CPU INSTEAD OF THE GPU SO THEY CAN WORK IN TANDEM AND SOLVE THIS PROBLEM. IN FACE WHY DON'T WE HAVE IT PRECALCULATE THE SCROLLING FRAMES IN BOTH DIRECTIONS WELL IN ADVANCE SINCE WE HAVE 4 CORE CPUS JUST SITTING THERE NEVER USED ON THESE MACHINES? WE CAN MAKE SCROLLING EVEN BETTER THAN IT'S EVER BEEN BEFORE THANKS DAVID MCGUIGAN WE'RE GOING TO HIRE YOU FOR IDEAS AND GIVE YOU TIM COOK'S STOCK OPTIONS!!! )

 

[NoMoreSony]

Whatever the specs are ...

When 1920x1200 is used, the whole content is rendered into a 3840x2400 pixel texture and then scaled down to 2880x1800.

This is stupid, the specs are not relevant. It's a waste of resources just to show off a new product.

Apple has repeatedly claimed to work on resolution independence. On the iPad and iPhone they dictated integral scaling (2x).

On the MacBook they realized that they cannot get away with that because people need more screen estate to work properly.

This is really a half-baked approach to a problem that is simply papered over by massive use of hardware resources.

No way it is stupid. This is typical Apple trick to push developers for quick adoption of Retina resolution in their programs. Do you remember the story when iPhone Retina applications looked crappy on iPads screens in 2x mode? It was specially done just to push developers to creating new iPad version of the same apps. It was changed only in the new Ipad.

If Apple did direct conversion from 1 pixel to 2x2 all of the old non-Retina progs looked nearly the same, as new Retina adopted programs. Most of old programs then were never updated. Hopefully, some smart guys can program small application to avoid artificial bad scaling and switch on direct 1 ->2x2 conversion.

We don't need (yet?) jailbreaking our Macs to use third party applications. And most of us can then simply download this program and use our old programms in perfect scaling instead of investing a lot of bucks in the new updates. Just to rid off bad scaled icons, buttons and fonts.

 

[davidmcguigan]

Wow

Happy to report that peformance on Windows 7 Pro is absolutely insane. Feels at least 4x as fast as OS X to me, and is actually completely usable at the native 2880x1800.

I think I may have died and gone to heaven. What a gorgeous screen / OS combination.

 

[killmoms]

1. The 27 inch iMac now exists for some time and was first driven by an Intel Core2Duo!!!!!
2. It has just A FEW pixels less than the MacBook Pro Retina!!!!!
3. It could handle this high resolution just fine!!!!!

The 27” iMac has nearly 1/3 fewer pixels than the MBPR. Or, to put it another way, the MBPR has 40% more pixels to drive than the 27” iMac. It’s a considerably larger display. I think this is simply a side-effect of the machine being released slightly before Mountain Lion, where it performs much better. Might be why every person who buys an MBPR gets a free upgrade to ML.

Tempest in a teapot.

----------

…and is actually completely usable at the native 2880x1800.

Maybe if you’re Superman. For the rest of us with normal vision, 200% DPI scaling in Windows 7 is pretty laughable compared to OS X’s HiDPI mode.

 

[zeraien]

My story is slightly different since I went to an unofficial mac store in Stockholm, where they don't run the kiosk software that you'll find in real Apple stores. Just a stock Lion with some Apple content.

I checked activity monitor and stuff. But the main difference is that I didn't test it with 4 windows/apps open. I opened every app on the dock (the standard lion dock), including a bunch of Safari windows. Mission Control was laggy. It was not horrible, but it was laggy.

The 2012 non-retina MBP standing next to it had NO PROBLEM what so ever with the same number of apps and windows. It rendered them very smoothly!

I also tried changing the scaling on the retina MBP, and certainly the upscaled resolutions had more lag, but even "best for retina" resolution had some lag.

Another quick test is to press the new Launchpad button. Press it and you shall see noticeable lag. Especially when you press the same button on the 2012 MBP without retina and see how smooth it's there.

Don't just take my word for it, have a look at real benchmarks - in most graphics benchmarks in Cinebench that i've seen, the new retina MBP gets lower or comparable FPS performance that my early-2011 MBP did!

Note, this is about the GPU only, not memory or SSD performance - you get all those in the non-retina 2012 MBP as well.

retina gets 39fps (older mbp still wins, getting around 45fps): http://www.barefeats.com/mba12a.html

retina gets 34fps: http://www.youtube.com/watch?v=hVniAF9I99o&feature=player_detailpage#t=383s

early 2011 MBP gets 35fps: http://www.youtube.com/watch?v=EEkFNVNkgWM&feature=player_detailpage#t=224s

retina gets 33fps: http://www.youtube.com/watch?v=Zg8C7hIvdCc&feature=player_detailpage#t=96s

 

[chituan]

Here's a quick video to show the UI. iTunes window is still just a little laggy but that's always been the case on any computer.

http://www.youtube.com/watch?v=VLzRuxXZiK4

May I know the screen resolution that is used here ?
I just read Anandtech review and know that native retina (5 mil pixels) is less laggy than 1200p for example because the later has to render up to 9.2 mil pixels. I am buying the rMBP to use it solely at 1200p as I am a developer so if it is that smooth at 1200p in ML then I would definitely buy it.

 

[paulinbognor]

Just a query to those lucky enough to have their hand on a MBPR When running an external monitor 30" or 27" from the machine (with the lid closed) I preume the machine has no problems at all with lag etc and its all just from dealing with the Retina display.

The reason I ask is that I intend to use the machine as a mobile workstation which will spend 80% of its time attached to an external screen. thanks in advance

 

[davidmcguigan]

For the rest of us with normal vision, 200% DPI scaling in Windows 7 is pretty laughable compared to OS X’s HiDPI mode.

Apple actually sets it to 150%, but I think you're missing the point. You get 2880x1800 of real estate versus just 1440x900 with HiDPI mode. Unfortunately the scrolling performance and responsiveness of the machine overall in OS X is just too terrible to keep this machine, taking it back to the Apple store this week.

It's literally unusable if you're looking for more than a light web browser. Scrolling is choppy and just abysmal at any of the non 1440x900 settings. I love everything about this machine except how dog slow OS X is in almost every use case and how terrible the non retina-ready ( 98% of all apps ) apps are. I should also mention the new magsafe is SUPER SUPER LOOSE and falls off most of the time when I use it on my bed or couch. Sad. Maybe the next update's graphics card will be able to handle retina life better, the Anandtech review says that though Mountain Lion will help the graphics cards and OS X's poor choice of offloading a lot of graphics processing to the CPU instead of GPU just aren't ready for these resolutions, and even the Facebook news feed scrolling will still be slow as **** in M-Lion.

 

[seaw]

...and even the Facebook news feed scrolling will still be slow as **** in M-Lion.

So, what is going on??? Will Mountain Lion fix this issue or not?
I'm pretty tired of going back and fourth about thinking scrolling will be fine and scrolling will lag forever!!!

Why is FACEBOOK still sluggish under Mountain Lion?

 

[leman]

At any other resolution however, let's take the default scaled resolution ( 1440x900 ), what happens is that for each frame per second ( imaginary, it's not really measured that way ) the video card has to render everything like it normally does, meaning figure out what to show for each pixel, and THEN do another full calculation for each pixel to drop it down to simulate the 1440x900 from its first rendering at 2880x1800. I'm doing a terrible job of explaining this but in a nutshell: It's actually a lot less work to use the much higher resolution of 2880x1800 than any scaled resolution that OS X ships with ( the only options available ).

This is completely wrong. At HiDPI 1440x900, the system is working at the native 2880x1800 resolution, it simply draws everything four times bigger, so that the UI sizes are the same as for a native 1440x900 display. Other scaled modes take a performance hit because the rendering happens at much higher resolutions. E.g. 1680x1050 renders to a 3360×2100 offscreen buffer and then downscales it to 2880x1800. The downscaling itself is an extremely cheap operation compared to everything else. Basically, there is no reason at all for unscaled 2880x1800 to be faster than scaled 1440x900. Actually, I would expect the native res to be slower, as it has to render more UI elements.

 

[prfrma]

Tried one in store today.

At highest resolution in final cut pro, a 1080p movie takes up less than a quarter of the screen size, and I was just scrubbing through the timeline serenely without much care in the world. People really just don't know what there talking about and how mental this is for any computer to do right now.

I'll be upgrading fro a 2008 uMBP which was underbaked and unpowered for what I did with it for work - this thing, even as a first gen seems very impressive, lots of screen real estate with little compromise in terms of general performance, where as if your at native it chomps though whatever you throw at it. There is a little stuttering when at highest resolution setting switching between full screen apps but thats it really.

 

[Office Hours]

My observations.

The OS runs fluidly at native retina and 1680 scaled under the HD4000. 1920 runs slightly jittery (minor) but resolves with the dGPU.

Safari runs slow under anything. Chrome with GPU acceleration enabled, runs perfect... almost. Sites that use a lot of fancy CSS and have fixed components do lag a bit but I would say that this is 10% of the sites of visit. It is NOT images that cause the lag, it's complicated css effects. This is not isolated to RMBP, if you run Safari at full screen on a high res iMac, those sites will lag too. If you web dev you're probably familiar with this slow down on desktop/laptops. They figured it out with the iPad 3 so I have faith that they will with the RMBP.

 

[CUsurfer]

The 100% process utilization you see in activity monitor is because during multitouch scrolling a CPU core focuses very heavily on very very granularly tracking your finger movements including acceleration, direction change, etc. at a super high resolution to attempt to realistically interpret your acceleration, momentum, give the page a feeling of a sense of inertia, etc.

Could you test this by simply plugging in a wired mouse, perform a scrolling operation in Safari whilst looking at Activity Monitor? I doubt it's the trackpad. I'm guessing it's just hard for the scaling to keep up.

OMG WHY DON'T WE JUST HAND-OFF THE DOWNSCALING CALCULATION TO THE CPU INSTEAD OF THE GPU SO THEY CAN WORK IN TANDEM AND SOLVE THIS PROBLEM. IN FACE WHY DON'T WE HAVE IT PRECALCULATE THE SCROLLING FRAMES IN BOTH DIRECTIONS WELL IN ADVANCE SINCE WE HAVE 4 CORE CPUS JUST SITTING THERE NEVER USED ON THESE MACHINES? WE CAN MAKE SCROLLING EVEN BETTER THAN IT'S EVER BEEN BEFORE THANKS DAVID MCGUIGAN WE'RE GOING TO HIRE YOU FOR IDEAS AND GIVE YOU TIM COOK'S STOCK OPTIONS!!!

That doesn't sound like a good idea to me. I want my CPU free to perform real tasks like the millions of complex floating point calculations performed by my audio software. That type of stuff needs to be completely offloaded to the iGPU or dGPU, not the CPU.

 

[killmoms]

Apple actually sets it to 150%, but I think you're missing the point. You get 2880x1800 of real estate versus just 1440x900 with HiDPI mode. Unfortunately the scrolling performance and responsiveness of the machine overall in OS X is just too terrible to keep this machine, taking it back to the Apple store this week.

Using 200% DPI scaling (doubling the size of interface elements in Windows) is the same as Apple's 1440 x 900 HiDPI mode in OS X. 150% DPI scaling in Windows 7 is equivalent to to the 1920 x 1200 HiDPI mode in OS X.

 

[davidmcguigan]

So, what is going on??? Will Mountain Lion fix this issue or not?
I'm pretty tired of going back and fourth about thinking scrolling will be fine and scrolling will lag forever!!!

Why is FACEBOOK still sluggish under Mountain Lion?

According to Anandtech, no. In fact a lot of the sub par performance of the new Retina MBP will stick around indefinitely because the graphics card and drivers in OS X just can't handle it well enough yet. The reviewer predicted/hoped that subsequent retina macs will come with much beefier graphics processing, which judging from what Google announced today ( a 16 core graphics card inside their new Nexus tablet ) might not be that far off.

----------

This is completely wrong. At HiDPI 1440x900, the system is working at the native 2880x1800 resolution, it simply draws everything four times bigger, so that the UI sizes are the same as for a native 1440x900 display. Other scaled modes take a performance hit because the rendering happens at much higher resolutions. E.g. 1680x1050 renders to a 3360×2100 offscreen buffer and then downscales it to 2880x1800. The downscaling itself is an extremely cheap operation compared to everything else. Basically, there is no reason at all for unscaled 2880x1800 to be faster than scaled 1440x900. Actually, I would expect the native res to be slower, as it has to render more UI elements.

I think if it were done the way you're imagining the aliasing effects and appearance of text would be identical to those done when you set the screen's resolution to 1440x900, which literally just "draws everything four times bigger". Instead they render at the super high resolution and then intelligently choose what to smooth and colorize where to product a much sharper, higher definition image on the interface chrome. Make no mistake, even at the "best for retina" display performance is still terrible on anything with a lot of images, including the app store. If you open the app store, resize it to about a quarter of the screen and then click the zoom ( plus/maximize ) button, you'll see how terrible this machine performs. It resizes that window at around 10-15 frames per second. Resizes. A. Solitary. Window.

----------

Tried one in store today.

At highest resolution in final cut pro, a 1080p movie takes up less than a quarter of the screen size, and I was just scrubbing through the timeline serenely without much care in the world. People really just don't know what there talking about and how mental this is for any computer to do right now.

I'll be upgrading fro a 2008 uMBP which was underbaked and unpowered for what I did with it for work - this thing, even as a first gen seems very impressive, lots of screen real estate with little compromise in terms of general performance, where as if your at native it chomps though whatever you throw at it. There is a little stuttering when at highest resolution setting switching between full screen apps but thats it really.

Right, working with a dedicated app where there's only 1 image on screen at a time ( even if it's rendering 30 of them per second in the same spot ) will be much faster and more responsive than a giant slew of other tasks like web browsing for example. I've been using the machine every day now for almost 2 weeks and in practice the machine is just slow. It feels on par with a Core 2 Duo running OS X with laggier scrolling about 70% of the time. I wish they offered this exact machine with this exact display at 1920x1200 instead of 2880x1800, that would be a machine I could actually use for work on the go.

----------

My observations.

The OS runs fluidly at native retina and 1680 scaled under the HD4000. 1920 runs slightly jittery (minor) but resolves with the dGPU.

Safari runs slow under anything. Chrome with GPU acceleration enabled, runs perfect... almost. Sites that use a lot of fancy CSS and have fixed components do lag a bit but I would say that this is 10% of the sites of visit. It is NOT images that cause the lag, it's complicated css effects. This is not isolated to RMBP, if you run Safari at full screen on a high res iMac, those sites will lag too. If you web dev you're probably familiar with this slow down on desktop/laptops. They figured it out with the iPad 3 so I have faith that they will with the RMBP.

Unfortunately it does not run "fluidly". Maybe "closer to fluidly". But if you do any real grown up tasks you'll start to see it jitter up. Unfortunately CSS has absolutely nothing to do with it. You'll see the same issues in and out of a web browser.

----------

That doesn't sound like a good idea to me. I want my CPU free to perform real tasks like the millions of complex floating point calculations performed by my audio software. That type of stuff needs to be completely offloaded to the iGPU or dGPU, not the CPU.

If you're ever able to fully saturate all 8 virtual cores of the quad core ivy bridge they put in this thing then you're the kind of person that's also got access to a dual hexacore class desktop workstation and should be doing your rendering there instead. Consumers have MOST of their CPU free or underutilized approximately 100% of the time. Trust me on this hahaha.

 

[seaw]

According to Anandtech, no. In fact a lot of the sub par performance of the new Retina MBP will stick around indefinitely because the graphics card and drivers in OS X just can't handle it well enough yet. The reviewer predicted/hoped that subsequent retina macs will come with much beefier graphics processing, which judging from what Google announced today ( a 16 core graphics card inside their new Nexus tablet ) might not be that far off.

I watched the whole 36 min Anandtech review and was impressed by the depth. He clearly explains the scrolling issue under lion and mountain lion. BUT I THINK:
I'm not really into all the technical stuff. I do not know much about 16 core graphics, CPUs and what they can and could handle in the past...but I dont understand how one, that is so deeply reviewing a product, can just say: "The hardware can not handle the many pixels. People should wait a couple of years, then it should get better!"
I don't get it.

First:
the resolution of the iMac is not so much smaller than the res on the retina. And the iMac deals with this high resolution for SOME time now even at core2duo times. I've never experienced a lag.

Second:
If the many pixels would cause this problem....then....hook up the cinema display with a standard Macbook Pro. This machine is not lagging at all driving even MOOOORE pixels.

Corect me if I'm wrong.