I'm not arguing that it isn't possible. Just pointing out that it's not as cheap as some people think it is.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Haswell RMBP wont get a 750m... and heres why
- Thread starter yabot
- Start date
- Sort by reaction score
I'm not arguing that it isn't possible. Just pointing out that it's not as cheap as some people think it is.
1. The 2012 iMacs had the NVIDIA 650M, so did the rMBPs. Your argument is invalid.
2. Again, same specs in the 2012.
3. Broken record... If the iMacs didn't change in price from 2012 to 2013, why would the rMBPs.
4. This i can agree on. I think they will only do a max of 1GB VRAM.
Note though: avoid using words like "won't" and "never" in your titles unless you are certain. Clearly, research was not done.
No--not necessarily. The graphics switching prowess has come a long way with OSx...with Mavericks, even more power saving protocols have been added. If you watch the keynote---it's kinda cool, the example they give. You slide your calendar over the top of the website you're working with (Or browser with dozens of sites open)---it will intelligently shut the power down to the websites while using the calendar and until you re-unite with your browser...and it 'looked' like this was essentially ubiquitous with all apps and software in 10.9. The graphic switching capability is pretty damn good---and my current rMBP is good for 7 hours without power. The 750 is essentially an updated 650 with maybe a 10-15% increase in performance. Essentially the same chip. So if you're using whatever Apple decides to use as the iGPU from Intel (which are MUCH improved over this past year's 4000 series)---say the 5000, or even perhaps the new Iris Pro...it'll be less than often you'll need to utilize the dGPU for normal tasks. Video editing, for some reason the entire Adobe suite
....Flash sites, et al will turn your 7xx card on---but again, with gfxcardstatus you'll be able to control that. Along with Mavericks---I'm anticipating a 10 hour 15" rMBP this time around (in a controlled, 50% brightness environment) which is incredible for a laptop as nice as the rMPBUse Click 4 Flash---or any of the dozens of anti-flash measures and plug ins for your browsers. YOU can control this. As well, with the aforementioned gfxcardstatus app, it's free and can easily control when you do/don't want your dGPU fired up (Although it's free---if you have paypal, throw the dude a few bucks, he's maintained a KILLER app for a long time and we shouldn't take it for granted IMO).
As long as the logic board is set up for the dGPU there isn't a whole lot of 'extra' monetary concern. It's EFI on Apple's boards, not BIOS---regardless, if there's an option for the dGPU in certain CTO configs, that work will already have been done in the 'base' config so on the assembly line the extra GPU can be added. I understand and don't disagree with your thought process---but I think Apple has this down now (they're manufacturing). It's my HOPE they maintain the dGPU in the 15" for at least another round or too. While the Iris Pro does well in computation---it lacks in gaming and certain video titling and graphic programs, mainly Adobe's fascination with OpenGL (or do I have that backwards?)---as it is now, Adobe's After Effects works incredibly well on my 15" rMBP. it'll eat all the RAM you give it---but with the GPU offload for some of the work, it STILL makes sense to keep it---at least in the CTO options. IMHO of course---ymmv
Any chances you could share apps or software (other than benchmarking tools) that will throttle the rMBP? Generally curious---we've been using them solely for video editing and effects now over the past year with AK State Troopers, Deadliest Catch and a couple of History channel Docs before we send off to LA (Discovery and Nat Geo) and I have YET to see one of them 'throttle' or slow down---in fact, they easily keep up with my decked out 2009 MacPro with 32GB of RAM (with the exception of After Effects which will utilize every bit of RAM if you let it per frame). I'm not a gamer---other than XPlane, so maybe I'm missing something.
True---I agree with your responses. Unfortunately I HOPE they don't go this direction this year...the shared 640/650 of last year's iMac (low end) has now been replaced by the Iris Pro. Granted, I have ZERO experience with Intel's new bad boy iGPU---but the pricing of that Haswell chip with the Iris is almost in parity with last year's Ivy Bridge proc plus the 650m (I think both right around the $375 mark last check?). Hence adding the dGPU will increase the pricing if Apple uses the I/P iGPU in the rMBPs this year (and adds the dGPU as well).
Time will tell.
For those calling the OP stupid....has it ever occurred to you folks aren't as experienced with Apple's releases? Ages of posters on this board? I agree is wasn't worded perfectly with absolutes---BUT come on! Give the dude/dudette a break
---- I think OP---maybe next time instead of making absolute assertions, a post with a question mark would be better and not bring out the divas, and the know it alls. I saw a professional video guy/gal on here the other day arguing about the MacPro being a joke---not for pros, because he's a pro and his system NEEDS---never mind, I need not go on. When I looked at his profile, he joined this board almost three years after me....however, he has almost 5,000 posts in just over two years. I've NEVER met a professional video editor, colorist, finalist, sound dude----no one EVER in Alaska, Seattle, Portland or LA that has THAT kind of time to spend on Macrumors posting this gibberish Take it for what it is, but I understand the OPs concerns....and to the OP---if it's gaming and a bad ass video card you're after, build your own Windows rig and stack it up! Use an Apple machine to get your 'work' doneOr---maybe, better yet, for the same price as a 'gaming rig' --- you could buy BOTH the new XBox (1?) and PS4 for less cash and have access to most of the AAA titles you'd have on your computer. With Microsoft's transition to incorporating 'Glass' and Sony's expectations to allow iOS devices as controllers could be pretty nifty and a good way to incorporate your Apple purchases with the next gaming consoles.
Not sure what the answer is---but if you're a gamer, you can look elsewhere---Alienware, Razer, ibuypower, cyberpower....the boutique computer industry is active and VERY expensive if you want the monsters. I'm an Apple fan but still subscribe (through Newsstand
) to Maximum PC---and inevitably the review section monthly involves one of the latest, GREATEST Falcon Northwest $12,600 systems with Tri-SLI/CrossFire and 32GB of RAM with SSDs and XTreme Procs with damn near 2,000 watt power supplies that will soon need their own dedicated circuit to power----JUST the tower! I think those days are limited....we'll see what these new consoles are all about.J
Last edited:
The rmbp can handle full CPU load OR full GPU load but if you load up both to the max it can throttle, mainly due to power requirements (adapter cannot supply enough power). The other reason why its not as noticeable is that apple deliberately limits boost.
You are losing 200mhz speed right there. Apple charges extra for the cpu upgrade to run the cpu at 2.4ghz + boost speeds.
----------
750m is about 15% better. The only other option is 8870m. 760m is possible but consumes a lot of power and the rmbp is already near its thermal and power supply rating limits.
760m+ is NOT an option because the die size is larger. There is no spare room on the rMBP logic board to put in a larger dGPU.
Base MBP will have Iris Pro, BTO 15" will have 4600 + 750m.
Base MBP will have Iris Pro, BTO 15" will have 4600 + 750m.
Something twice as good?
That's here.The more pragmatic discussion can be found here and in my thread, along with a few other places.
Something twice as good?
The more pragmatic discussion can be found here and in my thread, along with a few other places.
Twice is good is still just a decent mid-range part for 2013. Not GREAT, but...
On a $2000 notebook really twice is good isn't close to being good enough.
Is there any actual practical shipping external GPU solution?
Obviously the holy grail would be something you plug in, which fits ANY video card, fits a Titan if you wanted to, and you just plug it in, OS X and/or Windows just recognize it, and...that's it.
Sounds like drivers or technical issues prevent that though? Not sure anything's shipping anyway.
Plus Thunderbolt's a bottleneck, only offering 4 PCIe lanes for the entire connector, when really you'd want 16 JUST for the video card + more for other stuff, but still it would sure as heck beat having no GPU or a terrible one!
EDIT: Years ago Sony did have a dock for a notebook that added a GPU through Thunderbolt, BUT it was only a low end GPU, just that one system, and Sony's always locking down drivers so you can't just install Nvidia's normal drivers on their systems.
A 750m is based on the GK107 die and is 118 mm^2.
The 760m is based on GK106 and is 221 mm^2.
750m and 760m are both 128 bit GDDR5 and so the tracing and soldered vram will take up a similar amount of space.
You are talking about 1 square cm difference in die size or about an increase of probably less than 0.5%. Die size has nothing to do with it.
No. Too much power (650m is already causing the battery to be used when gaming) and not enough cooling. Possibly 750m or 755m. A 760m would be possible with the 35 watt i7-4702mq.
I expect Iris Pro and 750/755m as BTO.
Wow dude. You have some relatively unrealistic expectations. Why do you think that magical Sony dock isn't around anymore?
Are you new to Macs? Just wondering because I've never heard such fantastic demands for a slim premium notebook before.
1 sq cm is a HUGE area on the already supremely cramped rMBP MB.
Please point out the 'extra' space.
http://www.ifixit.com/Teardown/MacBook+Pro+15-Inch+Retina+Display+Mid+2012+Teardown/9462/2
Yeah that's what I was thinking as well, just wanted to make sure my assumptions weren't off for some reason
. Really sucks that mostly cooling is the issue. Even though it's "Apple-like" to only put a 750M in there for battery alone, I don't think that's justifiable, and don't think they'd do that. It'll be even more interesting to see their decisions next year, when they have IGZO screens to cover almost all the "need" for battery. These machines are going to be super lame if they no longer offer dGPU next year. Broadwell's are going to be pretty good in the graphics department if we have Maxwell dGPU's in the upper end.Its a small area that can easily be accommodated. Mobo size could be slightly different given the change to haswell. Its not a huge area. Proportionally its a couple of percent max (consider that both sides of the mobo can be used) and should not be that hard to accommodate. Apple is a major company that can afford (and do) spend a couple extra dollars designing their computers. With their engineering prowess I don't think it would even be particularly challenging for them.
And don't forget the savings they make on the smaller chipset.
http://ark.intel.com/products/64339/Intel-BD82HM77-PCH
HM77 (used in rmbp) 25 x 25 mm.
HM87 (most or all of other variants are the same size) 20 x 20 mm
http://ark.intel.com/products/75528
So there is their die savings, going from 25 x 25 to 20 x 20 results in 625mm^2- 400 mm^2 = 225mm^2.
Edit: Probably should mention that its the package size that's important, NOT the die size. i3 + GT1 is quite a bit smaller than i7+ GT2 haswell but the package is the same size and so takes up the same space on the mobo (so for example in desktops they can be placed in any LGA 1150 socket). GPUs also follow this pattern; as can be seen the GPU package as a whole is several times larger than the actual die. We could probably increase the die significantly without increasing the package as much.
A GK106 package would be quite a bit larger but I think apple could fit it if they wanted to. However due to heating and power requirements I doubt it was ever really on the table in the first place.
Note that the CPU package is 37.5 x 37.5 = 1406 mm^2 (i7 quad and i5/i3 dual).
ULV (960mm^2 : 40 x 24 for ULV GT3).
So in this case the difference between die sizes quad and a non ULV dual are eliminated by the package. I do not expect this to happen for GPUs.
Last edited:
So...it appears after actually reading/researching you have reached the same conclusion as other members months ago...
The 760+ won't fit. It's at best a 755M, most likely a 750M.
I'll reiterate what I said in my original post. 1cm sq is a HUGE area. If you didn't understand that meant a physical as well as thermal limit, then...
It is still wrong. A 760M is actually easier to cool. To run it at similar performance as a 750/755M you need only about half the clocks. Lower clocks also come along with lower voltage which reduces total power consumption by more than just the relative clock rate. At similar power consumption you can probably run about 2/3 the clock rate.
Then you also have the heat more spread out on an area which makes it easier to transmit it to the heat pipe and the cooling block.
The bulk of PCB space used by a dGPU doesn't go to the chip itself, it is the wiring (which stays the same with 128bit MC, same PCIe width) and the memory chips (GDDR5 which should be 2GB now). The size of the chip actually helps more than it hurts.
An as cirus says it is package size that would matter first.
The only reason why they won't use a 760M is cost. Nvidia just wants more money for it than for the tiny 750M.
Then you also have the heat more spread out on an area which makes it easier to transmit it to the heat pipe and the cooling block.
The bulk of PCB space used by a dGPU doesn't go to the chip itself, it is the wiring (which stays the same with 128bit MC, same PCIe width) and the memory chips (GDDR5 which should be 2GB now). The size of the chip actually helps more than it hurts.
An as cirus says it is package size that would matter first.
The only reason why they won't use a 760M is cost. Nvidia just wants more money for it than for the tiny 750M.
Well, it would be great if they would put a 760 in there.
I'm not terribly hopeful.
Thanks for the info.
Now that you mention it, I do remember another thread where someone said the 760 would be easier to cool/more efficient than a 750/755.
I'm not terribly hopeful.
Thanks for the info.
Now that you mention it, I do remember another thread where someone said the 760 would be easier to cool/more efficient than a 750/755.
Its a small area is conjunction with the total size of the motherboard. The PCH has shrunk a bit. Don't forget that haswell has integrated VRMs on the CPU die freeing up even more mobo space.
Again, Its unknown how much the gpu package size will increase on a 760m vs a 750m. It could be a little, It could be a lot. The wiring is similar (Vram + traces). Either way gains in package size will be offset by reductions elsewhere.
760m consumes a lot of power and requires more cooling. I have no doubt that it would be harder to fit on the mobo but definitely possible (there are clevo notebooks out there with 45 watt quad core cpus and a 760/765m in a 13" chassis) but because of cooling and thermals the 760m was probably never seriously on the table.
Wow dude. You have some relatively unrealistic expectations. Why do you think that magical Sony dock isn't around anymore?
Are you new to Macs? Just wondering because I've never heard such fantastic demands for a slim premium notebook before.
I don't think he has unrealistic expectations. I think he's not expressed himself clearly that's all.
What he wants has been widely available on eGPUs for the past half decade (expresscard).
The sony dock is no longer around because it EOLed along with the previous generation of Vaio Z. And that was a proprietary and not extensive product.
If Intel (and Apple) would stop limiting thunderbolt development (through pulling products off shelves and preventing production), we'd be seeing cheap thunderbolt solutions and more consumers' needs could be met.
http://forum.techinferno.com/diy-e-gpu-projects/
It'll also stop thunderbolt from becoming totally useless.
#1: if the eGPU was worth the development, it would still be around today. Not that the idea isn't good, mind you.
2#: I'm fairly sure that Apple and Intel haven't decided to intentionally cripple their own profits with Thunderbolt.
I heard there are quite a few bugs with GPU over Thunderbolt and that is why no big brand seems confident to really commit to that. Sony also only released one single driver that you had to use. No big brand seems willing to commit to proper support for an eGPU creation. They probably assume it will be more trouble than it is worth.
I think it doesn't show, is because companies think it is too expensive for the expected volume of buyers. You need a GPU, a power supply, a PCB with Thunderbolt to PCIe and a case. It makes for an expensive package that too few people would care to pay for. I believe companies just gathered that those that game on notebooks are rather content with mobile GPUs and the rest won't be pulled away from consoles or desktop PCs anyway, especially not at the cost of basically a cheaper notebook for the eGPU.
The whole Thunderbolt fun is just too expensive. That has always been its problem. The tech is too complicate and requires too expensive chips and Intel is too greedy to allow it becoming a mass market product.
Most tech that currently is mostly worked on goes around streaming to/from a PC or any sort of wireless tech. eGPU is probably never going to take off.
I think it doesn't show, is because companies think it is too expensive for the expected volume of buyers. You need a GPU, a power supply, a PCB with Thunderbolt to PCIe and a case. It makes for an expensive package that too few people would care to pay for. I believe companies just gathered that those that game on notebooks are rather content with mobile GPUs and the rest won't be pulled away from consoles or desktop PCs anyway, especially not at the cost of basically a cheaper notebook for the eGPU.
The whole Thunderbolt fun is just too expensive. That has always been its problem. The tech is too complicate and requires too expensive chips and Intel is too greedy to allow it becoming a mass market product.
Most tech that currently is mostly worked on goes around streaming to/from a PC or any sort of wireless tech. eGPU is probably never going to take off.
eGPU development and driver writing still exists. And an eGPU chassis isn't just for graphics cards, the PCIE slot can take on many other expansion cards which are beneficial for professionals. Costly examples (often out of the reach of consumers) of this include PCIE chassis by Sonnet and OWC.
--
They won't cripple their profits (or let's talk revenue) by stymieing the development of thunderbolt by third parties. It really isn't very complex, Intel's fairly exclusive behavior is due to:
1) Capturing revenue that would otherwise go to thunderbolt developers
2) Controlling the user experience of thunderbolt
3) Lessening the complementary effects their technologies may have on direct competitors.*
*This is up to speculation.
This exclusive behavior by Intel has led to thunderbolt devices/expansion slots having artificially high prices due to a lack of competition and innovation. A case to prove this would be ViDock being denied certification of being a thunderbolt developer and hence being unable to license any thunderbolt tech.
The main thing about reading my reply is that don't take what I say to be absolute and extreme, Apple and Intel and great companies and masters at marketing, but their specific handling of thunderbolt is bound to reduce the overall value consumers can get out of machines with this port. And we may very well see history echo itself as with firewire.