• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Graphics for 15" rMBP Haswell

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
badb0y said:
The 650M would run out of power before the VRAM limitation is hit.

A number of 7850s ship with 1 GB and game fine even at high resolutions. The VRAM is not limiting performance in this particular case. If the GPU was something like the GTX 675M then the 1 GB VRAM would limit the card.
Click to expand...

The vram argument here is not really that the gpu will support the high settings (it will not) but if 1 GB is insufficient for the highest res textures (difference in fps between high and low textures is virtually 0 as long as you have sufficient bandwidth and vram). 1GB might require high or medium vs very high in some games.
 
dagamer34 said:
I expect the MacBook Pro in 2014 will have 2GB GDDR5 memory as 4Gbit chips will be available (though a good chunk of them will be going to the PS4).
Click to expand...

Another year? : (

The memory standards need to be updated across the board anyway. The iMacs for last year only having 512 MB on three out of the four models was sad. If anything at the very least, there should have been 1 GB in the base 27" for the GTX 660M.
 
Ice_Dragon said:
Another year? : (

The memory standards need to be updated across the board anyway. The iMacs for last year only having 512 MB on three out of the four models was sad. If anything at the very least, there should have been 1 GB in the base 27" for the GTX 660M.
Click to expand...

Apple is always stingy with RAM for some reason. Don't know why they are more optimistic than reality has shown...

4Gb GDDR5 chips won't be in mass production in time for a fall update this year (and again, I'd bet most of those chips are going to the PS4 since it requires them to hit 8GB total).
 
dagamer34 said:
Apple is always stingy with RAM for some reason. Don't know why they are more optimistic than reality has shown...

4Gb GDDR5 chips won't be in mass production in time for a fall update this year (and again, I'd bet most of those chips are going to the PS4 since it requires them to hit 8GB total).
Click to expand...

To me they're selling themselves short. People who need the ultimate iMac for example will buy it and get the GTX 680MX as a BTO.

Those who don't will buy maybe the base model 21.5".
 
Rakehellion said:
Will Apple really put a dent in their magnificent battery life to add a gaming GPU with more memory?
Click to expand...
The current 15" rMBP already has a dGPU, so there's no reason to believe the next version couldn't also host one. The question is whether Apple wants the better performance of a dGPU or stick with GT3e and get better battery life than the current generation rMBP.
 
ViRGE said:
The current 15" rMBP already has a dGPU, so there's no reason to believe the next version couldn't also host one. The question is whether Apple wants the better performance of a dGPU or stick with GT3e and get better battery life than the current generation rMBP.
Click to expand...

The GT3e should only go into the 13" or go into a less expensive 15" if they are going to do what they did in 2009 with a 15" with just the 9400M.
 
ViRGE said:
The current 15" rMBP already has a dGPU, so there's no reason to believe the next version couldn't also host one. The question is whether Apple wants the better performance of a dGPU or stick with GT3e and get better battery life than the current generation rMBP.
Click to expand...

But jumping up to a 660M or 2GB of memory would suck up more battery. Could that be better spent on an upgraded CPU or other features?
 
Rakehellion said:
But jumping up to a 660M or 2GB of memory would suck up more battery. Could that be better spent on an upgraded CPU or other features?
Click to expand...

The GT650M in the current Retina MacBook Pro already runs at 660M speeds, and I don't think VRAM is as big a power draw as the display. The major issue is chip density (amount of space 2GB of VRAM would take on the logic board) and cost (can we get it for cheap).

The calculation really isn't that hard, if it's this year, it'll be a high-end SKU, next year, maybe the baseline version will have it, but most definitely by 2014 when 4Gbit chips should be more plentiful.
 
Rakehellion said:
But jumping up to a 660M or 2GB of memory would suck up more battery. Could that be better spent on an upgraded CPU or other features?
Click to expand...
We're up to the 700 series now, which have the advantage of a new boost mechanism and slightly lower power usage due to node improvements. A 750M or whatever would not necessarily consume power more power than the 650M did. It won't be terribly faster, but it doesn't mean power consumption has to go up.
 
dagamer34 said:
The GT650M in the current Retina MacBook Pro already runs at 660M speeds, and I don't think VRAM is as big a power draw as the display. The major issue is chip density (amount of space 2GB of VRAM would take on the logic board) and cost (can we get it for cheap).

The calculation really isn't that hard, if it's this year, it'll be a high-end SKU, next year, maybe the baseline version will have it, but most definitely by 2014 when 4Gbit chips should be more plentiful.
Click to expand...

It's faster than 660M speeds, I believe the 660M has a core clock of 835MHz while the 650M in the rPro runs at 900MHz.
 
joshhedge said:
It's faster than 660M speeds, I believe the 660M has a core clock of 835MHz while the 650M in the rPro runs at 900MHz.
Click to expand...

650m-735mhz base, 850 boost. (the ddr3 versions run at 850 base no boost).
retina 650m-900mhz no boost
660m-835 base, 950 boost
750m--967 base + boost 2.0 (often 1058-1100 depending on game)

Boost is almost always running unless the computer is overheating.

ViRGE said:
The current 15" rMBP already has a dGPU, so there's no reason to believe the next version couldn't also host one. The question is whether Apple wants the better performance of a dGPU or stick with GT3e and get better battery life than the current generation rMBP.
Click to expand...

They probably won't get better battery life. OSX supports automatic graphics switching (optimus) and so the vast majority the user is on the computer's battery they are using the igp and not the dgpu (which only kicks in when really required which isn't often when on battery). Bootcamp does not support grpahics switching and always runs on the dgpu (this is why windows battery life is poor). Switching the gt3e would not bring about any performance gains (probably a net loss) but would allow apple to use a slimmer case or help to eliminate thermal problems. The gte3 versions also have a lower clockspeed than the gt2 quads.
 
Enigmoid said:
650m-735mhz base, 850 boost. (the ddr3 versions run at 850 base no boost).
retina 650m-900mhz no boost
660m-835 base, 950 boost
750m--967 base + boost 2.0 (often 1058-1100 depending on game)

Boost is almost always running unless the computer is overheating.
Click to expand...

I see. No where on the Nvidia website did it state anything about boost being enabled. The advantage the rPro has is that due to the new thermal system inside, the GPU and CPU rarely throttle at all as shown by Anand, I mean I've never experienced slow downs. I don't expect a 660m to run at maximum boost all of the time, especially in sub-par laptops.

ViRGE said:
The current 15" rMBP already has a dGPU, so there's no reason to believe the next version couldn't also host one. The question is whether Apple wants the better performance of a dGPU or stick with GT3e and get better battery life than the current generation rMBP.
Click to expand...

Sitting in the library studying for the summer exams I'm getting 9:30 battery life with word, iTunes and Webkit open. I think Apple has hit the sweet spot here for battery life personally and should focus on increasing performance at this level of battery life through the inclusion of a 750m or a 760m.

Apple should have gone with the GTX 660M to begin with, sounds more premium and fitting for the rPro than the GT 650M.

Rakehellion said:
So why is it a 650M? Does the 660M have significantly more cores?
Click to expand...

Nope, both have the same 384 Kepler CUDA cores. It was probably Apple being cheap, binning the GPUs that don't clock highly into the cMBP and the ones which do into the rMBP.
 
Last edited:
joshhedge said:
The advantage the rPro has is that due to the new thermal system inside, the GPU and CPU rarely throttle at all as shown by Anand, I mean I've never experienced slow downs. I don't expect a 660m to run at maximum boost all of the time, especially in sub-par laptops.
Click to expand...

If the rmbp 15 which is definitely on the far end of the 'laptops with heat problems' scale doesn't have a problem with the 650m overclocked then very few laptops are going to have a problem maintaining boost.

Generally with a few exceptions you can run most i7 +650/660m laptops on furmark and prime (real stress test not anand's wimpy HL2 test) and not throttle below stock speeds (turbo on the cpu may not be active but generally full gpu clocks are possible). If you try to do that on a rmbp the computer will throttle like crazy (and the 85 watt power supply doesn't seem to be large enough either). Pretty much all but the real duds of the computing world will keep the the turbo.

Notebookcheck

Prime 95 and Furmark can extract the maximum performance out of the hardware. When both programs are run simultaneously, the core frequency drops (caused by Prime 95) to 1.2 Ghz. Running Prime 95 alone, we measured temperatures between 92 and 104 degrees Celsius (197.6 and 219.2 degrees Fahrenheit), and the word "Throttling" flashes repeatedly in HWiNFO. Furmark on its own cycles through all available Turbo Boost steps. After about one hour, the temperature settled at a fairly consistent 85 degrees Celsius (185 degrees Fahrenheit) with the fan system running. We never measured above 100 degrees Celsius (212 degrees Fahrenheit) with Furmark. This limitation might be due to the power adapter: in the above scenario, the system would require 86.6 watts - but the power adapter can only supply a maximum of 85 watts. Immediately after running our different tests, we subjected the system to 3DMark 06 and CineBench R11.5. The results were identical to after a cold boot.
Click to expand...

47675.png


That cpu score is lower than expected. The 3630qm gets that score but is a weaker cpu.

Notebookcheck get 5.52 r11.5 multi points for the 2.3ghz version when most of the other computers with a similar cpu (same clockspeed but the igp speed is different are slightly over 6).
 
Last edited:
Enigmoid said:
If the rmbp 15 which is definitely on the far end of the 'laptops with heat problems' scale doesn't have a problem with the 650m overclocked then very few laptops are going to have a problem maintaining boost.
Click to expand...

I'm not sure I fully understand what you mean, are you saying that the rPro has a heat problem? Because I've never experienced throttling, nor has the laptop got uncomfortably hot in any of the video encoding, gaming or 3D CAD modelling software that I use.

Enigmoid said:
Generally with a few exceptions you can run most i7 +650/660m laptops on furmark and prime (real stress test not anand's wimpy HL2 test) and not throttle below stock speeds (turbo on the cpu may not be active but generally full gpu clocks are possible). If you try to do that on a rmbp the computer will throttle like crazy (and the 85 watt power supply doesn't seem to be large enough either). Pretty much all but the real duds of the computing world will keep the the turbo.

Notebookcheck



47675.png


That cpu score is lower than expected. The 3630qm gets that score but is a weaker cpu.

Notebookcheck get 5.52 r11.5 multi points for the 2.3ghz version when most of the other computers with a similar cpu (same clockspeed but the igp speed is different are slightly over 6).
Click to expand...

Again, I have the 2.6GHz rPro and I haven't experience throttling.

Edit: I should probably make myself clear that I don't deny that the rPro probably does throttle at points, but the new and improved cooling system in the laptop is good enough to make that only occur at rare occurrences.
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top