Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Will the Haswell rMBP be announced in September with a dGPU option?


  • Total voters
    407
  • Poll closed .
150 votes for a dGPU. Glad I got my vote in before the iMacs came out. :D
 
though i still wonder if due to the binning rumor (link) if apple will get a 50pipeline version of the 5200, currently its 40pipelines @ 200 - 1300MHz

it also seems weird to be for apple to put the 5100 in the 13" and a 5000 + 7xx in the 15" and not use an iris pro at all in the macbooks (although they might use them in the 15" cmbp if they dont kill those)
 
though i still wonder if due to the binning rumor (link) if apple will get a 50pipeline version of the 5200, currently its 40pipelines @ 200 - 1300MHz

it also seems weird to be for apple to put the 5100 in the 13" and a 5000 + 7xx in the 15" and not use an iris pro at all in the macbooks (although they might use them in the 15" cmbp if they dont kill those)

Is a 5200+dGPU physically possible?
 
Is a 5200+dGPU physically possible?

size wise yes - they are from the same generation thus same socket thus same size physically

i thought i read somewhere that TDP was a concern however the current cpu i7-3635QM is 45 watt and the iris pro cpus are 47 watt - so i guess its just the cost possibly holding it back

though that kinda debunks the binning rumor (unless they have the two options, igpu with 50pipleines and igpu with 40pipelines + dgpu)
 
Last edited:
size wise yes - all the cpus are the same size, otherwise they would require different sockets

i thought i read somewhere that TDP was a concern however the current cpu i7-3635QM is 45 watt and the iris pro cpus are 47 watt - so i guess its just the cost possibly holding it back

though that kinda debunks the binning rumor (unless they have the two options, igpu with 50pipleines and igpu with 40pipelines + dgpu)

I think a 5200+750M would blow my mind.
 
670mx/675mx/760m/765m/770m they all have the same gk106 core

however their differences are in the clocks and the bus width, the 760m and 765m have a 128 bus, the 670mx and the 770m have 192, the 675mx has 256

simply put, gk106 isn't going to happen. thats very unfortunate, but they receive a very large discount if they buy the same parts for the most common pcs that they have, i.e. its either the 750m or 755m

basically, no performance gains whatsoever. it might actually be worse, given that the cooling doesn't handle that well the heat, with cripple boost 2.0 they might even perform worse.

That's extremely disappointing to hear. On the other hand, apple states up to 1.4x gain in performance for their iMacs. However I fail to see what model that correlates to. It might be the 755M in the low end 27 inch, or high end 21 inch (750M). Or do you think it's the 775M in the high end 27 inch? If so, it's irrelevant.
 
That would be redundant. A 4600 and 755M would make much more sense.
not really you forget that display with all those pixels to push.

That's extremely disappointing to hear. On the other hand, apple states up to 1.4x gain in performance for their iMacs. However I fail to see what model that correlates to. It might be the 755M in the low end 27 inch, or high end 21 inch (750M). Or do you think it's the 775M in the high end 27 inch? If so, it's irrelevant.

I don't know how they drew those comparisons. but its real,

775m looks like the 680m, which is indeed around 20% slower than the 680mx that apple used, which is coincidentally the 780m, yes the 680mx = 780m

however the 680m in turn is indeed around 30% more powerful than the 675mx

the 640m LE/640m/645m/650m/660m/730m/740m/750m/755m are the same gpu, they vary with the use of DDR3 or GDDR5 (yes all models here can use the DDR3 and be called any of that, except the 660m and I don't know if the 755m is going to be apple only), and they use the same core with different clocks.

anyway its the same damn thing from last year, because there is nothing new in the market.
 
not really you forget that display with all those pixels to push.

No, I didn't. If the 4600 can push "all those pixels" in Mavericks with no lag, than the 4600 can definitely do it too.
 
That's what I was wondering - physical constraints seem to be king.
Originally Posted by dusk007 said:
But a 760M chip running at lower clocks is more power efficient than a 755M.
Cooler?
Yes easier to cool too. Bigger chips that have the heat source more spread out make heat transfer easier than very cramped heat sources. That is among other things why the smaller and small chips get hotter.
I also meant more efficient. Apple always clocked GPUs any which way they liked. The 330M was under clocked the 650M over clocked.
If you pick the optimal clock on a 760M, you can get more performance for a given power consumption than you can with the smaller chip of a 750M/755M.
The 760M is significantly more expensive though and Apple is grasping.

I don't have numbers for it but I am guessing a 760M at default clocks might beat a 755M in power consumption. With the Turbo 755M runs at very high clocks that probably are quite a bit removed from the optimally efficient clocks.

From a consumer perspective a 760M is also much better because as the chips get better (cooler) a few months in you have the freedom to change clock speeds and have good overclocking. A 755M will have virtually no room for any worthwhile over clocking because it would be pushed into too high voltages long before you see any noticeable speed increase.
 
Yes easier to cool too. Bigger chips that have the heat source more spread out make heat transfer easier than very cramped heat sources. That is among other things why the smaller and small chips get hotter.
I also meant more efficient. Apple always clocked GPUs any which way they liked. The 330M was under clocked the 650M over clocked.
If you pick the optimal clock on a 760M, you can get more performance for a given power consumption than you can with the smaller chip of a 750M/755M.
The 760M is significantly more expensive though and Apple is grasping.

I don't have numbers for it but I am guessing a 760M at default clocks might beat a 755M in power consumption. With the Turbo 755M runs at very high clocks that probably are quite a bit removed from the optimally efficient clocks.

From a consumer perspective a 760M is also much better because as the chips get better (cooler) a few months in you have the freedom to change clock speeds and have good overclocking. A 755M will have virtually no room for any worthwhile over clocking because it would be pushed into too high voltages long before you see any noticeable speed increase.

It's all very interesting - I guess I have little faith that Apple would use a different chip in the MBPs than in the iMacs - at least they haven't in the past.
 
Yes easier to cool too. Bigger chips that have the heat source more spread out make heat transfer easier than very cramped heat sources. That is among other things why the smaller and small chips get hotter.
I also meant more efficient. Apple always clocked GPUs any which way they liked. The 330M was under clocked the 650M over clocked.
If you pick the optimal clock on a 760M, you can get more performance for a given power consumption than you can with the smaller chip of a 750M/755M.
The 760M is significantly more expensive though and Apple is grasping.

I don't have numbers for it but I am guessing a 760M at default clocks might beat a 755M in power consumption. With the Turbo 755M runs at very high clocks that probably are quite a bit removed from the optimally efficient clocks.

From a consumer perspective a 760M is also much better because as the chips get better (cooler) a few months in you have the freedom to change clock speeds and have good overclocking. A 755M will have virtually no room for any worthwhile over clocking because it would be pushed into too high voltages long before you see any noticeable speed increase.

Well the 755M is an overclocked 750M. I would assume the 755M is a separate product specifically for being a pre-made overclocked 750M. Since it is the same exact chip other than clock speed. With that said then yes the performance of the 755M could not likely be altered to achieve higher performance.

The 760M is clocked rather low to begin with, so that does seem like a better bet for higher clocking, which would create a significant boost of performance from the 650M of the current years. But you stated price as a restriction, so that sucks :(

What I'm intrigued by (specifically when speculating for the new rMBP's) is the fact that apple has spread 3 models out across the GPU's Iris Pro --> 750M --> 755M

I think it's basically impossible for the rMBP to get the 760M though because they only go up to 755M in the iMac line (ignoring the 775M)
 
What I'm intrigued by (specifically when speculating for the new rMBP's) is the fact that apple has spread 3 models out across the GPU's Iris Pro --> 750M --> 755M

I think it's basically impossible for the rMBP to get the 760M though because they only go up to 755M in the iMac line (ignoring the 775M)

The 2011 iMacs used 6750--> 6770--> 6970
The 2012 iMacs used 640M--> 650M--> 660M--> 675MX--> 680MX
The 2013 iMacs use.. 5200--> 750M--> 755M--> 775M --> 780M

I'm not trying to argue that the rMBP would get the 760M, but what intrigues you about the 3 model spread this year, vice previous years (with at least 3 models as well)?
 
The 2011 iMacs used 6750--> 6770--> 6970
The 2012 iMacs used 640M--> 650M--> 660M--> 675MX--> 680MX
The 2013 iMacs use.. 5200--> 750M--> 755M--> 775M --> 780M

I'm not trying to argue that the rMBP would get the 760M, but what intrigues you about the 3 model spread this year, vice previous years (with at least 3 models as well)?

Well, the baseline is a "regular" Iris Pro, not some magical beefed up version, based on the SKU.
 
The 2011 iMacs used 6750--> 6770--> 6970
The 2012 iMacs used 640M--> 650M--> 660M--> 675MX--> 680MX
The 2013 iMacs use.. 5200--> 750M--> 755M--> 775M --> 780M

I'm not trying to argue that the rMBP would get the 760M, but what intrigues you about the 3 model spread this year, vice previous years (with at least 3 models as well)?

I can't say anything regarding the 2011 because I'm not familiar with AMD, but regarding 2012 vs. 2013; they are the exact same. The 2013 line up are simply the rebranded 700 serious chips. The 650M is the 750M, the 660M (overclocked 650M) is a 755M (overclocked 750M), the 675MX is a 775M, and 680M is 780M. The 700 series has ~10-20% improvement in performance in general over the 600 series.

I don't mean to contradict Dusk, but I think the 760M is out of range. The 750M has 1300 million transistors, while the 760M has 2560 million. Also the 750M is rated for Medium sized laptops, while the 760M is rated for large laptops.

Comparing watts: The 750M is around 35-40W. The 755M (overclocked 750M) sits at 50W, and the 760M sits at 50W as well (and 765M at 60-65 watts). I think we can definitely throw the 760M out of the equation, simply for heat/TDP constraints. If it is indeed more expensive too, then it's out (which would make sense, given the iMac line up not having it).

That leaves the 750M and 755M remaining. Based on what they did in the current rMBP - placing a 650M (again, same as 750M) but clocking it at 660M (755M) speeds, I would assume they would do the very same for this years model. Because nothing else has changed, and there really aren't any other options this time around. That is, of course, ignoring Iris Pro, and or assuming that will be the low end model, and the dGPU the high end model.

My guess is they'll have a optimally overclocked 750M which will still be labeled 750M but acting closer to the (755M) in performance. That way they can probably get around 40-45W power consumption out of it instead of a constant 50W. Assuming that was the advantage to having a overclocked 650M instead of putting a 660M in the current rMBP. But I would have to let someone else explain another reason for why they did that, if there is one.

FYI I know nothing about chips other than what I'm able to see on Notebookcheck, and using my own analytical skills :p
 
I can't say anything regarding the 2011 because I'm not familiar with AMD, but regarding 2012 vs. 2013; they are the exact same. The 2013 line up are simply the rebranded 700 serious chips. The 650M is the 750M, the 660M (overclocked 650M) is a 755M (overclocked 750M), the 675MX is a 775M, and 680M is 780M. The 700 series has ~10-20% improvement in performance in general over the 600 series.

I don't mean to contradict Dusk, but I think the 760M is out of range. The 750M has 1300 million transistors, while the 760M has 2560 million. Also the 750M is rated for Medium sized laptops, while the 760M is rated for large laptops.

Comparing watts: The 750M is around 35-40W. The 755M (overclocked 750M) sits at 50W, and the 760M sits at 50W as well (and 765M at 60-65 watts). I think we can definitely throw the 760M out of the equation, simply for heat/TDP constraints. If it is indeed more expensive too, then it's out (which would make sense, given the iMac line up not having it).

That leaves the 750M and 755M remaining. Based on what they did in the current rMBP - placing a 650M (again, same as 750M) but clocking it at 660M (755M) speeds, I would assume they would do the very same for this years model. Because nothing else has changed, and there really aren't any other options this time around. That is, of course, ignoring Iris Pro, and or assuming that will be the low end model, and the dGPU the high end model.

My guess is they'll have a optimally overclocked 750M which will still be labeled 750M but acting closer to the (755M) in performance. That way they can probably get around 40-45W power consumption out of it instead of a constant 50W. Assuming that was the advantage to having a overclocked 650M instead of putting a 660M in the current rMBP. But I would have to let someone else explain another reason for why they did that, if there is one.

FYI I know nothing about chips other than what I'm able to see on Notebookcheck, and using my own analytical skills :p

lol so solid. This is about as good an explanation as any. :)
 
775m != 675mx

775m looks like a rebranded 680m, which is gk104 derivative (670 core), while the 675mx is a gk106 derivative, 660 core
 
That would be cool ! 5200 for compute intensive apps like CAD, 3d modeling, etc and 750M for games!.
Maybe Apple will in fact surprise!

I can't remember who posted it but it raises that concern that a low end rmbp with 5200 only would be better at some tasks than a high end rmbp with a 4600+dGPU. Not sure how that would be handled, but it will be interesting.
 
I can't remember who posted it but it raises that concern that a low end rmbp with 5200 only would be better at some tasks than a high end rmbp with a 4600+dGPU. Not sure how that would be handled, but it will be interesting.

I think it's pretty much been determined that the way they'd handle that is by putting Iris pro in the cheaper line with HD 4600 + dGPU in upper line. Which is weird, because for open cl that makes the cheaper model perform better. I'm curious as to what 21.5 Inch iMac users are thinking. Those who are choosing between the two models. We'll just wait to see with the MBPs though *sigh*
 
for open cl that makes the cheaper model perform better.

I'd want to see some benchmarks* before I doubled down on that statement. I'm sure there's been some overlapping performance in Apple's lineups in the past, but I can't see it being embarrassingly significant.

*of all the Haswell iMacs
 
I'd want to see some benchmarks* before I doubled down on that statement. I'm sure there's been some overlapping performance in Apple's lineups in the past, but I can't see it being embarrassingly significant.

*of all the Haswell iMacs

Idk all that talk in the Waiting for Haswell thread made it seem obviously significant that Iris Pro is good with most everything except 3D and rendering. The benchmarks of the iMac praise it too. So I don't see that changing. Unless the HD4600 is up there too, which it isn't haha
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.