Which in all probability is the same issue. Again folks :
30" ACD ran at 2560x1600 like 5-6 years ago. GPUs aren't the problem, it's something in the HiDPI software stack.
I personally wouldn't trade my rMBP 15" for the world.
I hope you're right.
Which in all probability is the same issue. Again folks :
30" ACD ran at 2560x1600 like 5-6 years ago. GPUs aren't the problem, it's something in the HiDPI software stack.
I personally wouldn't trade my rMBP 15" for the world.
You're saying the same thing I am except making a few wrong assumptions. The resolution isn't doubled then halved at all in HiDPI mode, it's only downscaled after upscaling if you use the scaled modes (1680x1050 or 1920x1200). Running at 1440x900 (2880x1800) only upscales non-Retina graphics to 2880x1800 and leaves it at that. Running 1920x1200 (which I run permenantly) runs a 3840x2400 framebuffer which it then downscales to 2880x1800 for displaying on the display.
And what I've been saying all along is quite correct, the problem is not the pixel pushing ability of the GPU. Pixel fill rates, look them up. Again, if you can't understand that a 4 year old GPU, the 9400m, could power a 30" ACD and that we're now in 2012, 4 years later, with GPUs 3 generations newer, I don't know how to explain it to you.
The 9400m could push 2.32 GP/s. Again, 4 years ago. 2880x1800 at 60 fps is roughly 311 MP/s. Are you getting this ? Even if what you say is true and HiDPI was so horridly optimized as you claim (which it isn't), that's still only half of the pixel fill rate of a now 4 year old Integrated GPU.
Anyway, I own a rMBP 15" and I've hardly met any "lag" whatsoever. I guess some people are just too sensitive. Thank god Apple knows it's a software issue and is fixing it for those people.
----------
The hardware is capable. You just don't understand GPU hardware or HiDPI to make such claims. Anand has always been full of it.
Apple didn't do anything, webkit did which is mostly people from KHTML which is what webkit was forked from.
The only thing I agree with you is about the "fullness" of Anand.
I understand perfectly well what I am taking about. You are right though that at native it's just rendered pixel doubled. But in any other scaled mode it's exactly as I decribed it. Let me remind you that the ACD at 30" has a 2560 × 1600, that's about 4 million pixels, in 1920x1200 (and to a lesser extent in 1650x) we are talking about (2x2=4 times) 9.2 million pixels for the video card to draw, and then operate the scaler down to retina resolution. That's two and 1/4 acds. If it's the "software" (and it's simple operations that prove that's not the case) then how come even after ml the ui lags and stutters have not been solved?
All of WebKit has nothing of KHTML in it. The QTWebKit port is being backported to KDE's structure and I would appreciate if people realized that for the past 5 years nothing from KDE has existed in any Apple WebKit code.
For the last 5 years, WebKit has not been an Apple only project. Heck, 2 years ago, Google surpassed Apple in terms of raw commits to the code base. Of course WebKit has evolved since its days as KHTML, it still is a fork of KHTML, and is maintained by quite a few folk outside of Apple. In fact, Apple aren't even the main contributors anymore.
Thats all fine, but google keeps a lot of its top improvements in their forked chrome project while everything apple does is available to benefit everyone.
So why did Apple release a product that clearly wasn't ready yet?
this was specific to bad Safari performance, its looks marginally acceptable in Chrome as it is.
Because the software scalers probably aren't yet fully optimized at all. There's probably still a ton of operations left up to the CPU, not the GPU, so it's not fully using the pixel fill rate capability of the GPU and it's getting bogged down by the process scheduler.
Anyway it's moot, most people are not running the scaled modes, using plain old 1440x900 points (2880x1800 pixels) and getting this lag, so it's obvious there's something wrong with Apple's scaler in Quartz Extreme. Let's face it, HiDPI mode is quite new and it's not like Apple has ever been quick at optimizing their graphics stack (look at how long it took them for OpenGL 3.2..).
The GPU itself can do, again, way more than 2.32 GP/s. That's Gigapixels, billions of pixels per seconds. It's quite enough to run the resolution and do proper updating at 60 fps. Something else is amiss in the software stack where we're not getting done with a frame in less than 15 ms (the maximum render time per frame to maintain 60 fps).
I maintain this isn't a hardware problem at all.
People here are missing the point. This only fixes the issue for some websites.
- The Verge is still choppy as are many other websites.
- Webkit crashes in double tap mode.
- Webkit slows down the UI frame rate (it's measureable). And no, it's not a memory leak. I did all the tests for that.
Bottom line: not ready for prime time.