Lag in HiDPI resolutions will never be fixed on current rMBP

ppone · May 17, 2015

Apple will never fix the lag on the rMBP because it is impossible to do it on the current hardware.

I have base 15' 2012 rMBP with a 650M GPU. I have also briefly tested on newer versions of the 15 rMBP and none of them solve the lag issue.

HiDPI resolutions with the upscaling and downscaling are really intensive. The animations will never be as smooth as iOS. It is amazing that Apple was able to create HiDPI resolutions with current generation GPUs. However most of it is through deep optimizations through the software side.

When people say the current lag can be resolved through software, I disagree. Apple might have made lag worse between Mavericks and Yosemite, but even Mavericks had lag. Apple is going against the limitations of the hardware and facing a massive computation problem has increasing HiDPI resolutions does not scale.

Increasing the HiDPI resolutions creates an exponential drop in UI performance. Hence why iPad and Iphone 6 can be buttery smooth 99% of the time and rMBP only 90%; due to HiDPI resolutions which are picked based on the device screen size.. The higher the resolution you go the more intense computing/gpu power you need.

The only workaround is to "Reduce transparency" and use a non-HiDPI resolution high enough that it still looks good on the retina screen. Using rdm you can use a resolution of non HiDPI 1900x1200 or 2048x1280.

Resetting the p-ram or any other workarounds will not work. Excited to what WWDC brings on the rMBP front. Apple had 3 years where they did not improve the rMBP besides minor updates. I hope they do some custom silicon to handle all the intensive upscaling/downscaling or create their own GPU.

mattopotamus · May 17, 2015

I would think it could be fixed for reasons you mentioned. Things like turning on reduced transparency helps greatly. I can only imagine that further tweaks to the software can be made to eliminate the lag altogether. Not to forgert, this lag was introduced with yosemite.

ppone · May 17, 2015

No it can't be fixed. Lag has been with the rMBP since it was introduced in 2012.
Increase or decrease in lag between OS versions are going to be there.

You don't realize how much software optimizations Apple has already done. There is not much room to do more.

The issue is it a very hard to problem to solve. Software can help lag but it won't eliminate it. IMO only custom hardware or GPU specifically designed for HiDPI is the only resolution. Remember GPU were not designed for HiDPI scaling in mind.

mattopotamus said:
I would think it could be fixed for reasons you mentioned. Things like turning on reduced transparency helps greatly. I can only imagine that further tweaks to the software can be made to eliminate the lag altogether. Not to forgert, this lag was introduced with yosemite.

leman · May 17, 2015

Nonsense. It has been explained a number of times here that contemporary GPUs have more then enough performance to render HiDPI desktops. My HD4000 has no problems with a 3360x2100 backing buffer. Nor do I have any lag (just like many other rMBP users).

The lag exists only in isolated, algorithmically flawed applications (such as App Store) or specific conditions, such as activating Mission Control when certain applications have visible windows. And there seem to be some occasional bugs in the Macs power management.

----------

ppone said:
Remember GPU were not designed for HiDPI scaling in mind.

GPUs are designed to filter multiple gigatexels worth of texture data every second. There is nothing special about HiDPI in that regard.

ppone · May 17, 2015

Sorry please stop spreading you can fix this with software nonsense.

It pretty well know there is a lag on all rMBP models. Do you even know how the rMBP handles HiDPI resolutions.

Maybe you should read this review from Anandtech[1]. It not the direct handling of 3360x2100 resolutions. It is in fact doing much more with the upscaling than downscaling.

Quote from Anandtech[1]

"To be quite honest, the hardware in the rMBP isn’t enough to deliver a consistently smooth experience across all applications. At 2880 x 1800 most interactions are smooth but things like zooming windows or scrolling on certain web pages is clearly sub-30fps. At the higher scaled resolutions, since the GPU has to render as much as 9.2MP, even UI performance can be sluggish. There’s simply nothing that can be done at this point - Apple is pushing the limits of the hardware we have available today, far beyond what any other OEM has done. Future iterations of the Retina Display MacBook Pro will have faster hardware with embedded DRAM that will help mitigate this problem. But there are other limitations: many elements of screen drawing are still done on the CPU, and as largely serial architectures their ability to scale performance with dramatically higher resolutions is limited."

1- http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/8

leman said:
Nonsense. It has been explained a number of times here that contemporary GPUs have more then enough performance to render HiDPI desktops. My HD4000 has no problems with a 3360x2100 backing buffer. Nor do I have any lag (just like many other rMBP users).

The lag exists only in isolated, algorithmically flawed applications (such as App Store) or specific conditions, such as activating Mission Control when certain applications have visible windows. And there seem to be some occasional bugs in the Macs power management.

----------

GPUs are designed to filter multiple gigatexels worth of texture data every second. There is nothing special about HiDPI in that regard.

newellj · May 17, 2015

ppone said:
Sorry please stop spreading you can fix this with software nonsense.

It pretty well know there is a lag on all rMBP models. Do you even know how the rMBP handles HiDPI resolutions.

Maybe you should read this review from Anandtech[1]. It not the direct handling of 3360x2100 resolutions. It is in fact doing much more with the upscaling than downscaling.

Quote from Anandtech[1]

"To be quite honest, the hardware in the rMBP isn’t enough to deliver a consistently smooth experience across all applications. At 2880 x 1800 most interactions are smooth but things like zooming windows or scrolling on certain web pages is clearly sub-30fps. At the higher scaled resolutions, since the GPU has to render as much as 9.2MP, even UI performance can be sluggish. There’s simply nothing that can be done at this point - Apple is pushing the limits of the hardware we have available today, far beyond what any other OEM has done. Future iterations of the Retina Display MacBook Pro will have faster hardware with embedded DRAM that will help mitigate this problem. But there are other limitations: many elements of screen drawing are still done on the CPU, and as largely serial architectures their ability to scale performance with dramatically higher resolutions is limited."

1- http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/8

To be fair, that article is three years old, when Intel's HD 4000 integrated graphics was the best Apple could offer.

sjinsjca · May 17, 2015

What is this lag of which you speak? My late-2014, fully-stuffed rMBP exhibits none of it except when the CPU is at full-throttle, such as when switching between Spaces desktops while loading and booting multiple virtual machines. Then and only then do I see anything that might be called lag. I call it a busy machine.

It suggests to me that the lag OP sees (and I don't) is indeed a software/algorithmic issue in poorly optimized applications, or perhaps a consequence of having the screen resolution set at some value other than default for the display. I don't deny that the GPU is running hard in this machine, but it really does appear to be adequate, and I work it hard.

Michael Goff · May 17, 2015

It can be fixed and I don't know why you're saying otherwise. As you have noted, the lag has gotten WORSE from OS to OS. This implies a SOFTWARE issue. But let's put it into perspective. My rMBP has a resolution of 2560x1600 in an effective of 1440x900. My phone has 2560x1440 at... I'm not completely sure, to be honest. Anyway, my phone has less lag on the UI than my laptop at times.

My laptop has a faster GPU than my phone.

GFXBench 3.0 - Manhattan Offscreen OGL off screen
min: 17.7
GFXBench 3.0 - Manhattan Offscreen OGL off screen
min: 44.7

The HD 6000 should easily power the UI at 60fps.

----------

sjinsjca said:
What is this lag of which you speak? My late-2014, fully-stuffed rMBP exhibits none of it except when the CPU is at full-throttle, such as when switching between Spaces desktops while loading and booting multiple virtual machines. Then and only then do I see anything that might be called lag. I call it a busy machine.

It suggests to me that the lag OP sees (and I don't) is indeed a software/algorithmic issue in poorly optimized applications, or perhaps a consequence of having the screen resolution set at some value other than default for the display. I don't deny that the GPU is running hard in this machine, but it really does appear to be adequate, and I work it hard.

Your Mission Control never drops frames? Your text input on Safari never lags behind a little and then speeds up to magically get an entire word done in a second?

ppone · May 17, 2015

Well Nvidia 650M is that same machine with Intel 4000 as well.

newellj said:
To be fair, that article is three years old, when Intel's HD 4000 integrated graphics was the best Apple could offer.

827538 · May 17, 2015

Never had any lag issues and I've run both Mavericks and Yosemite as well as boot camping Windows 7, I have no idea what this troll is talking about.
Both the dGPU and iGPU can handle HiDPI scaling fine without issues.
Hell I can play CSS at native 2880x1800 without issue.

Michael Goff · May 17, 2015

SCOLANATOR said:
Never had any lag issues and I've run both Mavericks and Yosemite as well as boot camping Windows 7, I have no idea what this troll is talking about.
Both the dGPU and iGPU can handle HiDPI scaling fine without issues.
Hell I can play CSS at native 2880x1800 without issue.

I guess you're calling a lot of people trolls, then. This is an issue that more than a couple people are complaining about.

leman · May 17, 2015

ppone said:
Sorry please stop spreading you can fix this with software nonsense.

...

Maybe you should read this review from Anandtech[1]. It not the direct handling of 3360x2100 resolutions. It is in fact doing much more with the upscaling than downscaling.

I am happy that you finally woke up after three years, but this had been discussed in great depth on these forums and elsewhere. I alone have probably written several dozens of posts, providing benchmarks, speculations and performance estimations. I am not going to repeat all of this again.

Just food for though for you, as you are clearly too lazy to look for material on the topic: A HiDPI buffer at 1920x1200 takes around 35MB. An Ivy Bridge with HD4000's total memory bandwidth is around 25GB/s. This means that an Ivy Bridge has enough bandwidth to copy the frame buffer around more then 700 times per second. The HD4000 was shown to have over 1.5Gpixel/s fillrate in actually benchmarks, which allows it to draw a 3480x2400 farmebuffer at over 170 FPS; as well as over 2Gtexel/s texturing performance, which would allow it to rescale such a buffer at around 200FPS. And this is only the HD4000, current-gem MacBooks have IGPs that double and triple these specs! So its quite easy to see that even the lowly HD4000 has more then enough horsepower to perform the task (I am not even talking about stuff like caching and frame buffer compression which can improve the performance of these operations dramatically). Of course, the reality is much more complex as the GPU needs to share the memory bandwidth with the GPU and there is some overdraw (translucent effects). But also, in reality, you'd be stupid to redraw everything at 60fps (and Apple engineers are far from being stupid). OS X only redraws as things change, and only the areas that have changed. So if the hardware is getting to its limits, then it would be only in situations where you have Mission Control with dozens of windows with animated contents, because Mission Control actually reflects all the changes in real time.

No, the issues that some people are having (as said before, my 2012 machine is lag-free) can be (and have been) explained by algorithmic issues. Any slight browser lag that I might have had experienced was finally fixed in Yosemite. But of course, you are welcome to ignore everything else and just rely on your 'facts' and quotes from a 2012 Anandtech article which is talking about performance of an OS X that became obsolete two years ago.

ppone · May 17, 2015

Well Apple thought the founder of AnandTech was good enough, so they hired him. When is the last time you saw any of these tech journalist get hired by Apple. So clearly, he knew what he was writing about.

I have no clue what you wrote about onhow Intel 4000 is powerful enough to smoothly power a retina display. It does not even make sense. I would trust Anand from AnandTech who was hired by Apple over what you wrote.

Please stop saying there is not lag and it is all algorithmic.

A recent MacRumors article https://www.macrumors.com/2015/04/06/2015-macbook-pro-ui-lag-os-x-yosemite/ the issue is alive and well.

leman said:
I am happy that you finally woke up after three years, but this had been discussed in great depth on these forums and elsewhere. I alone have probably written several dozens of posts, providing benchmarks, speculations and performance estimations. I am not going to repeat all of this again.

Just food for though for you, as you are clearly too lazy to look for material on the topic: A HiDPI buffer at 1920x1200 takes around 35MB. An Ivy Bridge with HD4000's total memory bandwidth is around 25GB/s. This means that an Ivy Bridge has enough bandwidth to copy the frame buffer around more then 700 times per second. The HD4000 was shown to have over 1.5Gpixel/s fillrate in actually benchmarks, which allows it to draw a 3480x2400 farmebuffer at over 170 FPS; as well as over 2Gtexel/s texturing performance, which would allow it to rescale such a buffer at around 200FPS. And this is only the HD4000, current-gem MacBooks have IGPs that double and triple these specs! So its quite easy to see that even the lowly HD4000 has more then enough horsepower to perform the task (I am not even talking about stuff like caching and frame buffer compression which can improve the performance of these operations dramatically). Of course, the reality is much more complex as the GPU needs to share the memory bandwidth with the GPU and there is some overdraw (translucent effects). But also, in reality, you'd be stupid to redraw everything at 60fps (and Apple engineers are far from being stupid). OS X only redraws as things change, and only the areas that have changed. So if the hardware is getting to its limits, then it would be only in situations where you have Mission Control with dozens of windows with animated contents, because Mission Control actually reflects all the changes in real time.

No, the issues that some people are having (as said before, my 2012 machine is lag-free) can be (and have been) explained by algorithmic issues. Any slight browser lag that I might have had experienced was finally fixed in Yosemite. But of course, you are welcome to ignore everything else and just rely on your 'facts' and quotes from a 2012 Anandtech article which is talking about performance of an OS X that became obsolete two years ago.

hberthier · May 17, 2015

So... it's a hardware issue and Microsoft created some sort of weird voodoo, enabling the same hardware to deal normally, without a single bit of lag, the same resolutions with all the upscaling etc?
Nice work microsoft.

windywalks · May 17, 2015

ppone said:
Well Apple thought the founder of AnandTech was good enough, so they hired him. When is the last time you saw any of these tech journalist get hired by Apple. So clearly, he knew what he was writing about.

I have no clue what you wrote about onhow Intel 4000 is powerful enough to smoothly power a retina display. It does not even make sense. I would trust Anand from AnandTech who was hired by Apple over what you wrote.

Please stop saying there is not lag and it is all algorithmic.

A recent MacRumors article https://www.macrumors.com/2015/04/06/2015-macbook-pro-ui-lag-os-x-yosemite/ the issue is alive and well.

Anand was a great journalist, while he wrote for Anandtech, but his role at Apple has never been disclosed so please don't overrate his authority as an all-knowing guru, he has his limitations like anybody else and he's definitely not overseeing every single department at Apple for inconsistencies and making a list and checking it twice if software is playing naughty or nice.

Moreover, please stop calling everyone on what they know and don't, since all you seem to know is what little you have read and some of your experiences. Nothing you write has any technical implications as to why three years after the introduction of the original Retina Macs the tech wouldn't be powerful enough to handle HiDPI. You're post title is nothing more than click-bait/rant.

ppone · May 17, 2015

First of all my post is a rant that lag on the rMBP exists when using HiDPI resolution. The current rMBP are power enough to handle HiDPI but with lag.

If you can't understand that than maybe you can't perceive lag.

As for proof, just run a xbench with the same resolution using HiDPI and not using HiDPI. The non HiDPI will perform at least 40% faster on graphics test.

Attached xbench graphic results running in HiDPI and non HiDPI on 10.10.3.

windywalks said:
Anand was a great journalist, while he wrote for Anandtech, but his role at Apple has never been disclosed so please don't overrate his authority as an all-knowing guru, he has his limitations like anybody else and he's definitely not overseeing every single department at Apple for inconsistencies and making a list and checking it twice if software is playing naughty or nice.

Moreover, please stop calling everyone on what they know and don't, since all you seem to know is what little you have read and some of your experiences. Nothing you write has any technical implications as to why three years after the introduction of the original Retina Macs the tech wouldn't be powerful enough to handle HiDPI. You're post title is nothing more than click-bait/rant.

Freyqq · May 17, 2015

I refuse to believe that my computer can run Starcraft II on high at 60fps, but it can't render a transparency effect and some windows being rearranged at 60fps. Totally a software problem.

----------

ppone said:
Sorry please stop spreading you can fix this with software nonsense.

It pretty well know there is a lag on all rMBP models. Do you even know how the rMBP handles HiDPI resolutions.

Maybe you should read this review from Anandtech[1]. It not the direct handling of 3360x2100 resolutions. It is in fact doing much more with the upscaling than downscaling.

Quote from Anandtech[1]

"To be quite honest, the hardware in the rMBP isnt enough to deliver a consistently smooth experience across all applications. At 2880 x 1800 most interactions are smooth but things like zooming windows or scrolling on certain web pages is clearly sub-30fps. At the higher scaled resolutions, since the GPU has to render as much as 9.2MP, even UI performance can be sluggish. Theres simply nothing that can be done at this point - Apple is pushing the limits of the hardware we have available today, far beyond what any other OEM has done. Future iterations of the Retina Display MacBook Pro will have faster hardware with embedded DRAM that will help mitigate this problem. But there are other limitations: many elements of screen drawing are still done on the CPU, and as largely serial architectures their ability to scale performance with dramatically higher resolutions is limited."

1- http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/8

Indeed. Apple's current solution is fundamentally inefficient. It's a software problem.

ppone · May 17, 2015

The power required for running games vs rendering OS UI elements at 60 fps are two totally separate problems.
You can see in the benchmark I posted above the OpenGL scores are almost the same. Games mostly just use OpenGL. However the Quartz and UI test are where the differences are.

If it was such an easy problem to fix why hasn't Apple released a software fix already.

When the xbench for HiDPI and non HiDPI modes give the same results under OS X, then I would say the issue is resolved.

Freyqq said:
I refuse to believe that my computer can run Starcraft II on high at 60fps, but it can't render a transparency effect and some windows being rearranged at 60fps. Totally a software problem.

Michael Goff · May 17, 2015

ppone said:
The power required for running games vs rendering OS UI elements at 60 fps are two totally separate problems.
You can see in the benchmark I posted above the OpenGL scores are almost the same. Games mostly just use OpenGL. However the Quartz and UI test are where the differences are.

If it was such an easy problem to fix why hasn't Apple released a software fix already.

When the xbench for HiDPI and non HiDPI modes give the same results under OS X, then I would say the issue is resolved.

Yeah, the games are more demanding. There's a thread around here where people disable the v-sync and they're able to get lag-free on their 4k monitors. So something tells me it's not a hardware issue.

Edit: Just checked, and the weirdest things are killing my frame rates. Typing? Moving the cursor a lot? Having the cursor be over a link or near links? Using mission control? And this is supposedly a hardware issue?

Freyqq · May 17, 2015

ppone said:
The power required for running games vs rendering OS UI elements at 60 fps are two totally separate problems.
You can see in the benchmark I posted above the OpenGL scores are almost the same. Games mostly just use OpenGL. However the Quartz and UI test are where the differences are.

If it was such an easy problem to fix why hasn't Apple released a software fix already.

When the xbench for HiDPI and non HiDPI modes give the same results under OS X, then I would say the issue is resolved.

So, it's fundamentally a software problem. I never said it would be an easy fix.

Michael Goff · May 17, 2015

I just noticed sometimes my framerate drops just sitting on a page for too long. Today is a very informative day.

leman · May 18, 2015

ppone said:
Well Apple thought the founder of AnandTech was good enough, so they hired him. When is the last time you saw any of these tech journalist get hired by Apple. So clearly, he knew what he was writing about.

You are hilarious. If Anand is such an authority for you, then be a doll and look up his subsequent articles where he notes that all the lag issues he used to have were fixed by subsequent software releases. But you don't want to do that, right? That would ruin your story.

ppone said:
I have no clue what you wrote about onhow Intel 4000 is powerful enough to smoothly power a retina display. It does not even make sense. I would trust Anand from AnandTech who was hired by Apple over what you wrote.

I gave you hard facts about hardware. If it does not make sense to you, refute my numbers and/or my arguments. If you can't, you are clearly not competent to make any statement whatsoever on the matter. Your appeal to authority is completely worthless.

ppone said:
A recent MacRumors article https://www.macrumors.com/2015/04/06/2015-macbook-pro-ui-lag-os-x-yosemite/ the issue is alive and well.

A minor issue with a specific model. I am not aware whether it has been resolved since, but it looks to me like another instance of the SMC bug. I used to have heavy lag in my 2012 rMBP until Apple has released a firmware fix.

dusk007 · May 18, 2015

leman said:
You are hilarious.

That is what I thought. This thread is funny.

Not a great deal of technical knowledge but he knows because Anand was hired by Apple. Possibly for Qualityassurance and not software engineering but if you are hired it means everything you say is gold in every subject matter relating to Apple.

sjinsjca · May 18, 2015

Michael Goff said:
Your Mission Control never drops frames? Your text input on Safari never lags behind a little and then speeds up to magically get an entire word done in a second?

Well, maybe that's it: I rarely use Mission Control. And yes, when I try it I can see it's a weensy bit stuttery. Shrug. I rarely use Mission Control. And the stutter strikes me as CPU rather than GPU related: the machine is fetching, shrinking and arranging live screens from all open apps and Spaces, without interrupting anything else that might be going on.

Interestingly, I see the same very slight stutter even if there are two or more live streaming videos underway. The videos continue to play. So again, CPU rather than GPU exhaustion seems likely.

As to the Safari issue... no, can't say I see that behavior. One might even characterize it as "snappy" on this machine. Does the issue occur in Chrome or Firefox too? If not, seems like it's not a GPU issue.

Michael Goff · May 18, 2015

sjinsjca said:
Well, maybe that's it: I rarely use Mission Control. And yes, when I try it I can see it's a weensy bit stuttery. Shrug. I rarely use Mission Control. And the stutter strikes me as CPU rather than GPU related: the machine is fetching, shrinking and arranging live screens from all open apps and Spaces, without interrupting anything else that might be going on.

Interestingly, I see the same very slight stutter even if there are two or more live streaming videos underway. The videos continue to play. So again, CPU rather than GPU exhaustion seems likely.

As to the Safari issue... no, can't say I see that behavior. One might even characterize it as "snappy" on this machine. Does the issue occur in Chrome or Firefox too? If not, seems like it's not a GPU issue.

Never tried it with Firefox, but Chrome just plain sucks in general.

Lag in HiDPI resolutions will never be fixed on current rMBP

macrumors regular

macrumors G5

macrumors regular

macrumors Core

macrumors regular

macrumors G3

macrumors 68020

Suspended

macrumors regular

Cancelled

Suspended

macrumors Core

macrumors regular

macrumors newbie

macrumors 6502a

macrumors regular

Attachments

macrumors 601

macrumors regular

Suspended

macrumors 601

Suspended

macrumors Core

macrumors 68040

macrumors 68020

Suspended

Our Staff