If the rMBP maxes out the Graphics Card is it "Futureproof"

[calderone]

Whatever ad hominems help you sleep at night.

You mean like this?

You read some stuff somewhere and now think that you know everything about how graphics hardware works.

 

[Slivortal]

You mean like this?

Even if the first part of his argument was arguably an ad hominem, over 80% of that post was completely fact-based. Responding to the 20% header and none of the 80% body just seems like you're asking for a trollfight.

 

[calderone]

Even if the first part of his argument was arguably an ad hominem, over 80% of that post was completely fact-based. Responding to the 20% header and none of the 80% body just seems like you're asking for a trollfight.

Facts that supported what I was saying (note he/she never admitted it).

 

[Slivortal]

Facts that supported what I was saying (note he/she never admitted it).

Facts supported what you were saying, and facts supported what they were saying. They admitted that there was overhead associated with the graphics changing operation, but argued that the single display type that required such overhead was mostly obsolete, and that the overhead generated in this specific case was inconsequential to the everyday running of the machine.

 

[calderone]

Facts supported what you were saying, and facts supported what they were saying. They admitted that there was overhead associated with the graphics changing operation, but argued that the single display type that required such overhead was mostly obsolete, and that the overhead generated in this specific case was inconsequential to the everyday running of the machine.

The topic of the thread is essentially asking "Inconsequential for how long?"

In other words, sure it may not be a problem TODAY (questionable thus far, given the reported lag issues), but how long can we reasonably expect the hardware and software to keep up with ever increasing graphical requirements?

Based on the idea that a machine such as this should perform in three years, much as it does today.

 

[stevelam]

Facts supported what you were saying, and facts supported what they were saying. They admitted that there was overhead associated with the graphics changing operation, but argued that the single display type that required such overhead was mostly obsolete, and that the overhead generated in this specific case was inconsequential to the everyday running of the machine.

i don't believe this for a second. i think its pretty obvious to anyone whos used the RMBP that the gpu can't handle normal UI operations efficiently at all. i have ML GM installed and i can easily see things slow down/lag here and there. the sad fact is the loss of FPS for simple things like scrolling/dragging makes the computer feel much slower and less responsive than it should be.

even when i have an external monitor plugged in, dragging a photoshop window from one screen to the other stutters horribly. little things like this add up to give an overall poor user experience.

 

[Slivortal]

The topic of the thread is essentially asking "Inconsequential for how long?"

In other words, sure it may not be a problem TODAY (questionable thus far, given the reported lag issues), but how long can we reasonably expect the hardware and software to keep up with ever increasing graphical requirements?

Based on the idea that a machine such as this should perform in three years, much as it does today.

Huh? I'm not sure exactly what you're saying... The lag issues AREN'T GENERATED by this overexaggerated issue because...

1. The majority of the UI, if any, doesn't run off of this specific software framework to begin with.

2. The costs associated with this software are EXCEEDINGLY small. The new Macbook Pro Retinas decreased their battery charge indicator power consumption by 100%, but that did nothing to the overall power consumption, because the cost reduced was so imperceptibly small. These are the kinds of changes we're talking about.

Most RAMDAC controllers use hardware overlays to begin with, so we're adding an inexpensive process to preexisting code. We're already paying up by spending time in a hardware overlay to begin with, so performing inexpensive processes is even more inexpensive than going out of our way to do them.

http://en.wikipedia.org/wiki/RAMDAC


The software lag issues are NOT hardware based, but rather SOFTWARE based, and all conventional evidence points to this.

Now I see what is going on. You read some stuff somewhere and now think that you know everything about how graphics hardware works.

Anand is probably talking about RAMDAC/resolution converters which is used to transform the digital video buffer into the signal fed into the monitor. This unit essentially supports free scaling in a sense. AFAIK, RAMDAC in modern consumer hardware is not built for the tasks HiDPI rendering ala-Apple has to perform, as the RAMDAC cannot downscale a buffer above the native resolution anyway (I might be mistaken about this point though). The shader-based downscaling Apple uses is still very cheap. Current IGPs have fill rates well over 1Gpixel/sec (the texture filtering performance is substantially higher). The full 2800x1800 frame buffer is 5 megapixels. This means that the IGP can fill the buffer 60 times per second (maximal required fps) without even breaking a sweat. The performance overhead is so low compared to the work required to actually draw the UI on the screen that any attempt to optimize that area is likely a waste of time.

P.S. Strictly speaking, RAMDAC is a more or less obsolete unit used for analog video output. They have been replaced by other hardware which outputs digital video signal instead. I still use the term RAMDAC here, for convenience reasons.

i don't believe this for a second. i think its pretty obvious to anyone whos used the RMBP that the gpu can't handle normal UI operations efficiently at all. i have ML GM installed and i can easily see things slow down/lag here and there. the sad fact is the loss of FPS for simple things like scrolling/dragging makes the computer feel much slower and less responsive than it should be.

even when i have an external monitor plugged in, dragging a photoshop window from one screen to the other stutters horribly. little things like this add up to give an overall poor user experience.

There are lags in the current RMBPs. But all evidence points to them being software-based.

 

[wordoflife]

The term future proof is ridiculous.

Yes and no.
In my opinion, it depends on what you're comparing it too.

For example, when you are building a PC and SLI multiple GTX 690's (mind you, they are quite expensive) in order to 'future proof' it, then yes, you will most likely succeed and use it for many years to come. Why? Because most people don't have $1000+ graphic cards (let alone SLI - link multiple graphics cards together for MOAR performance). Software (games) won't require cards like that for MANY years to come simply because the majority of it's users will have something less powerful. Newer cards and technology will catch up, but never be as powerful for only a fraction of the price immediately (as in, less than a year).

The issue with the Mac is that exactly the opposite happens. Newer cards and technology will replace old stuff (generally less than a year old) and still be the same price. They won't make a new Mac with poorer performance than the last version, so anything newer is basically better than what you had.

So my advice would be to just get what you need for now. Remember, I'm talking only in a matter of less than 12 months, and you're talking about 4 years. At that point, you'll already be so very behind and outdated anyways.

 

[leman]

Facts that supported what I was saying (note he/she never admitted it).

Oh come on. What you do here is typical non sequitur fallacy. You state some (correct) facts — namely that downscaling requires some work from the GPU, but then for some reason conclude that downscaling is the sole reason of all performance issues the rMBP may or may not be suffering from. What I did is to point out that the cost downscaling itself is practically zero compared to the cost of actually drawing the stuff on the screen — and this is where the actual performance loss is coming from. If 95% of your time is spent in the shop's queue and only 5% packing your bag, it won't save you any time if you optimize packing the bag, because even 10x speedup there would only result in a few percent of overall performance win. You have no reason to be so defensive. I have been optimizing software since I was a teenager and I used to do lot of work with 3D APIs. I do have some competence talking about these things.

 

[ixodes]

Futureproof is a myth from the past.

 

[JoeG4]

There's no such thing as futureproof.

The internals can handle 2880x1800 just fine. The issue is that the software hasn't yet been optimized...

Knowing Apple, if they haven't fixed drivers on something within the first few weeks of it being on the market, it'll probably never happen. AFAIK X3100 MacBooks still support fewer things than GMA 950 MacBooks lol.

 

[leman]

i don't believe this for a second. i think its pretty obvious to anyone whos used the RMBP that the gpu can't handle normal UI operations efficiently at all. i have ML GM installed and i can easily see things slow down/lag here and there. the sad fact is the loss of FPS for simple things like scrolling/dragging makes the computer feel much slower and less responsive than it should be.

Why do you people are so keen on claiming that rMBP performance issues are in the GPU? Complex UI rendering is still mostly done on the CPU. And scrolling (especially fast scrolling) usually involves complete redrawing of the window. Of course, if you have a complex website in a fullscreen window, the rendering task overhead is often 4x compared to the non-retina display. However, all these things can most likely be fixed by using better (optimized) rendering algorithms. I think ML GM already showed this.

 

[stevelam]

Why do you people are so keen on claiming that rMBP performance issues are in the GPU? Complex UI rendering is still mostly done on the CPU. And scrolling (especially fast scrolling) usually involves complete redrawing of the window. Of course, if you have a complex website in a fullscreen window, the rendering task overhead is often 4x compared to the non-retina display. However, all these things can most likely be fixed by using better (optimized) rendering algorithms. I think ML GM already showed this.

its pretty obvious its the gpu. why would the ipad 3 have basically the same clocked CPU as ipad 2 but 4x the GPU power? so you're saying all the retina rendering on ipad is just done by mostly the CPU?

 

[Centsy]

its pretty obvious its the gpu. why would the ipad 3 have basically the same clocked CPU as ipad 2 but 4x the GPU power? so you're saying all the retina rendering on ipad is just done by mostly the CPU?

ITT: A MacBook Pro is the same as an iPad

 

[leman]

its pretty obvious its the gpu. why would the ipad 3 have basically the same clocked CPU as ipad 2 but 4x the GPU power? so you're saying all the retina rendering on ipad is just done by mostly the CPU?

No idea how the iPad does it because PowerVR is a completely different pair of shoes. I am pretty sure that font rendering, for instance, is done entirely on the CPU on the OS X.