MXM?
Also 1 GB 650M in the classic MBP is fine it's just the retina where it's too weak for the price.
Even the non-retina has a 650M, which is confusing as I'd think they'd have more cooling to work with having all that extra space. Here's hoping for MXM in the next generation.
MX, like in the iMacs, not MXM which is a different technology and would be horrible for the retina.
Apple graphics cards have always been kind of weak though.
"Fast enough" has nothing to do with it. Lots of features like anisotropic filtering, high resolution textures, and buffering have only a minimal effect on processor usage, but a large effect on memory usage.
Fast enough has plenty to do with it. But I see there is no getting you to understand it.
Okay, so try explaining it to me why the arbitrary cutoff of the GT 650M could utilize no more than 1GB of memory.
It's different to argue that the 650m is fast enough to utilize 2gb of memory than to argue that a video card can't be slow enough to utilize no more than 1gb. In other words, you're changing the argument.
Sure, the speed of the 650m is debateable and it comes down to opinion. What I find as acceptable performance may be different than your acceptable performance. I don't doubt you could come up with scenarios where 2gb would be usable, but the overall performance just does not justify the bump in memory.
It's different to argue that the 650m is fast enough to utilize 2gb of memory than to argue that a video card can't be slow enough to utilize no more than 1gb. In other words, you're changing the argument.
Sure, the speed of the 650m is debateable and it comes down to opinion. What I find as acceptable performance may be different than your acceptable performance. I don't doubt you could come up with scenarios where 2gb would be usable, but the overall performance just does not justify the bump in memory. I'd rather take a 680m with 1gb memory than a 650m with 2 or even 3gb of memory.
I'm not changing the argument. It's always been about the 650M.
Rakehellion said:"Fast enough" has nothing to do with it. Lots of features like anisotropic filtering, high resolution textures, and buffering have only a minimal effect on processor usage, but a large effect on memory usage.
Okay, so basically you don't know what you're talking about. I named specific instances where processing speed is less important, whereas you're being intentionally vague to conceal your lack of information.
And we're not debating acceptable performance. I gave an example that you can play X game on Y settings and add anisotropic filtering to use 33% more memory with virtually no drop in performance. Memory and processor speed are two completely different things.
Are you saying that no one with a Core 2 Duo can ever use 16GB of system memory because their processor is too slow?
Please enlighten us as to how nvidia drivers handle low-level memory management and why 1GB is more than enough for a 650m.
Sure, I get your argument, like an analogy of using Y-rated tires on a 100HP car, but that point is that you have no idea what goes on in the background. There are lots of variables that can be attributed to increased memory usage without much extra GPU load, now and perhaps even more in the future. Anisotropic filtering alone can increase memory usage by ~30%.
So please, don't say that 1GB is enough to fully utilize a 650m like it's one of the commandments, because it makes you look foolish.
You have to apply the 650m in use with the RETINA screen, as this entire thread was based on the rMBP. Adding another 1gb may be useful on the non retina model, but it really is a waste on such a high resolution. Yes adding AF increases memory usage, great point! But if the 650m can't even handle the game without AF, what's the point? You already have to drop the resolution in modern games just to get it to play reasonably.
I can get increased performance out of any video card by dropping resolution.
Pretty sure I can run at native 2880*1800 with the IGPU, if that's what you were referring to?
I'm not being intentionally vague. I can also point out scenarios where more memory is NOT beneficial, but what's the point?
The basis for the discussion should be NATIVE RESOLUTION.
Quote where you gave X game on Y settings and 33% more memory utilization. Possibly in your head, but nowhere in this thread.
But in the end, we can't prove it for the rMBP.
None. Needing 1GB of memory and having 2GB is an entirely different scenario from needing 2GB and only having 1GB.
And a "what's the point" argument when I asked you a direct question is exactly what I meant by intentional vagueness.
"Should be," meaning a qualifier you just made up as you were losing the argument. I know that the Retina MacBook Pro can play Diablo III and Starcraft 2 at native resolution, but really, no one's forcing anyone to play at a resolution they don't want.
I suppose you could google "anisotropic filtering" if you don't know what it is.
I just did, whereas your stance is "I don't have to explain that" as you shifted goalposts.
If you don't know that texture accesses are not handled by shader units, then we don't have much to discuss.
You have to apply the 650m in use with the RETINA screen, as this entire thread was based on the rMBP. Adding another 1gb may be useful on the non retina model, but it really is a waste on such a high resolution. Yes adding AF increases memory usage, great point! But if the 650m can't even handle the game without AF, what's the point? You already have to drop the resolution in modern games just to get it to play reasonably.
I can get increased performance out of any video card by dropping resolution.
Maybe not everybody has the same expectations, but as far as my expectations go in regards to gaming, I prefer to run the highest resolution possible together with lower game settings.
What I just laid out here for you, is a VRAM limited scenario.
I could run many lightweight games at 2880x1800 (and basically remove the need for AA), but unfortunately the frame rate falls flat on it's face past 1200p because memory size and memory bandwidth limitations. The bottleneck is not the GPU fill rate.
So here you go, with my expectations higher VRAM and VRAM bandwidth would be welcome. If your expectations are to play at 800x600 on HIGH and have your eyes bleed, then obviously your reasoning is valid. But I would rather play at 2880x1800 on LOW, and I could use more VRAM. I could even use more VRAM (or rather faster dedicated VRAM) on an HD4000, nevermind a 650m.
Yes this is a valid scenario and what I too had in mind (although not my preference - I like 35fps + with high settings at native res). I don't think in your suggested scenario the game would go from unplayable to playable. Would you notice a difference? Possibly. Do you think it would go from say, 15-20fps to all of a sudden pushing 30+? I'm not convinced the VRam is your sole solution. Do you have an example of a game that shows this?
Yes, here's a stupid example for you. Fifa 13 on very low, field view (not close ups). It goes from 60+ fps at 1920x1200 to unplayable at 2560x1600 on intel HD4000.
Yes this is a valid scenario and what I too had in mind (although not my preference - I like 35fps + with high settings at native res). I don't think in your suggested scenario the game would go from unplayable to playable. Would you notice a difference? Possibly. Do you think it would go from say, 15-20fps to all of a sudden pushing 30+? I'm not convinced the VRam is your sole solution. Do you have an example of a game that shows this?
The 650M would run out of power before the VRAM limitation is hit.