Do you also long for the old days when a floating point coprocessor was packaged in a separate chip?
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.
Would you buy a rMBP without a dedicated GPU?
- Thread starter johngwheeler
- Start date
- Sort by reaction score
Do you also long for the old days when a floating point coprocessor was packaged in a separate chip?
I know I am posting here against this Iris Pro only system. But I am not against a future without Dedicated GPU's - All I'm saying is, right now the Iris Pro is not fast enough to warrant a $2,000+ Laptop without a dedicated GPU. The performance is not there yet. Maybe in 2014 or 2015 but not now.
I know I am posting here against this Iris Pro only system. But I am not against a future without Dedicated GPU's - All I'm saying is, right now the Iris Pro is not fast enough to warrant a $2,000+ Laptop without a dedicated GPU. The performance is not there yet. Maybe in 2014 or 2015 but not now.
Link
In that article they are talking about it outperforming a GT 640. But Apple used an overclocked 650m in the rMBP. It performs as high as a 660m. The 640 is a lot slower than a stock 660m or overclocked 650m.
I really don't think Iris Pro is ready for the big time. And this is still a comparison to last years GPU. We are dealing with the 750m and 760m this year and each of those offer 5 to 10% performance boosts over their predecessors.
I don't think you really understand the new CPU models. The MBA also got a 1.3Ghz CPU that replaced a 1.7/8 Ghz CPU and gets pretty much exactly the same benchmark numbers at 2W less TDP. The base clock is lower because the GPU is bigger and can take a bigger chunk of the total TDP. When the GPU isn't active which it is in most situations hat you need a fast CPU, the CPU ends up running due to turbo at pretty much the same clock speeds.
2.4Ghz is the new 2.8Ghz. Base clock is a misleading metric here.
A 2.8Ghz Haswell will NOT be any faster than the 2.4Ghz Iris Pro Haswell. Even the few Watts that the 128MB eDRAM needs will be worth it for some workloads where latency or really high bandwidth is useful.
Just compare the MBA benchmarks between 2013 and 2012 from anandtech.
http://www.anandtech.com/show/7085/the-2013-macbook-air-review-13inch
That should tell you how ridiculous it is to simply say one is slower based on base clock only.
The 750M is a joke. It is just the same GPU with a bit more clock speed because it has a more mature stepping. You can overclock any 650M of a newer stepping probably to about the same level (in Windows at least).
The Iris Pro can definitely compete with those GPUs. iGPUs used to be half or a quater as fast as anything dedicated. Now they are in striking distance even trading blows on some non gaming pro level stuff. That GPU is absolutely competing. Just because it doesn't beat the fastest dGPU yet doesn't mean much. It also doesn't add 40+W just for 20% more performance.
As far as Pro applications go drivers is all that matters and at least in Windows Iris Pro leaves a 650M in the dust in some of these. Geforce drivers are crippled intentionally but on OSX there is enough raw power on the Iris Pro to make drivers that you won't notice any difference.
http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested
Gamers can complain but anybody else really has no reason to. A 760M would be quite a bit faster in games but since when does Apple care.
From all the reviews of Quad Core Haswell notebooks I read the PGA models seem to show no signs of any better battery life (most of these were not terribly well optimized gaming notebooks though) than Ivy Bridge while the BGA HQ Model in the Razer Blade is almost ultrabook level. It also has a **** screen so the rMBP won't be as good but still. I have yet to see a PGA Haswell with any decent battery life.
The 4950HQ is too expensive to show up in anything other than the high end rMBP unless Apple miraculously changed to become a very generous company.
I understand perfectly. You are neglecting to include a GPU load in to any of your turbo estimations. It is very rare that the GPU will be just sitting idle. Even using YouTube utilises the GPU for decoding tasks now and even nominal loads cause these GPU's to clock up to 50 to 75%.
Doing Photoshop work for example that will use both OpenCL and the CPU. This will cause the CPU to remain at 2.4GHz under any kind of serious work. I'm sure that 3.8GHz boost clock will really come in handy while I'm writing a novel in textedit.
I'm actually surprised that people are still outraged by NVIDIA rebadging old GPU's and bumping the clocks. They did this with the 8000 to 9000 Series the 200m to short-lived 300m series the 400 to 500 and the 600 to 700. Everyone knows they do this, it is not new and the increased clock speed does make a difference.
And yes I'm aware that the current Haswell chips while clocked lower than Ivy Bridge perform the same as the old higher-clocked Ivy Bridge chips. My point is, can Apple give us a notebook that performs the same as last year? That isn't what I want and it isn't what I think Apple will do with their Pro notebook.
The Air is all about portability and working while mobile, it made sense there to freeze performance to the same as the previous model while drastically increasing battery life. The same does not hold true on a Pro machine. For starters the first rMBP still feels underpowered for its screen and the Mac Pro which is a Pro machine received two very high end graphics cards. It is clear Apple thinks fast graphics are important in their Pro machines, I remember a time when the Mac Pro shipped with a single low end NVIDIA 7300GT!
Problem is even if that dGPU is off the compromises in the chassis to accommodate it are still there. That wasn't a huge problem a few years ago when powerhouse laptops were roughly the size and weight of a small bungalow but today every millimeter counts. The cooling system has to be able to deal with the extra heat of the dGPU, the motherboard has to be that bit larger to provide a physical place for it to sit and, as a result, the flexibility of the chassis is limited a bit more than it would be if you didn't have it in the first place. Now for some that'll be a compromise they're willing to make but I suspect that number will shrink as iGPU's get more and more capable.
Personally I'd be more than happy with a rMBP without a dGPU. Heck, I just bought a 13" rMBP! Put simply I don't game or do video editing so even the current HD4000 is enough for what I use the machine for. Again though it's all personal preference, there's no right answer to this.
The 13" rMBP has never had a dedicated GPU so.. yes I would? There's no difference with the today's rMBP.
And if you watch a youtube video what you you need a CPU at full speed for?
Photoshop OpenCL usage is a couple of specialized routines which at the moment of running won't need all that much cpu. You won't get much from having the CPU at full speed at the same moment.
If you ran some really heavy background process that uses OpenCL or GL and still need the 15% extra clock speed on whatever else you do that would be slightly different. But a Photoshop filer that you run right at the moment is not really CPU limited if it uses OpenCL. I doubt there is a noticeable difference. Also Apple due to the strong enough cooling system might opt for running the CPU at 55W TDP which should put average CPU speed definitely ahead.
In many workloads the eDRAM will probably make up for quite few Mhz.
It is only really games that need the cpu too at a significant speed but even there the quad core is enough, it is a big problem on the 15W dual cores though (the HD 5000 is barely any better than the HD 4400 because there is so much CPU limitation in many modern games).
With a 760M with a 2.8Ghz is most definitely faster but the difference in everyday work (apart from games) should be quite minor. Also you wrote in one post about mulit monitor stuff. The Iris Pro is actually equipped to push a ridiculous amount of pixels. Intel made sure some UltraHD revolution is not going to be a problem. In 2D that chip can handle any amount of monitors as long as there is some way to transport the picture info to those. You don't get surround gaming afaik but just extended monitors for work are no issue at all.
Just because some app uses the GPU like Chrome for rendering a web page doesn't mean it needs it at right the same second as the cpu. Usually when stuff is offloaded to the GPU the CPU load is a constant and generally not a limiting factor. The chip can switch between powering up either part of chip in less than a millisecond. I very much doubt that in any real world usage you would see any noticeable difference because of that base clock difference. The limited GPU use we do have is largely not an issue in everyday stuff especially the stuff that most people need lots of CPU speed for. Pro applications that make heavy use of the GPU can only use so much CPU at the same time. The code is not optimized for right the perfect balance of cpu/gpu performance available in any possible machine (only consoles get that treatment). Games also only need a decent base line of CPU speed with everything above being pretty much a waste.
I am aware of rebranding GPUs being something common. It is still just rebranding and ergo not really appropriate to speak of different generations. 500M vs 600M was a different generation. 600M vs 700M is more like Ivy Bridge 3720QM vs 3760QM. It isn't a different generation.
-----------
Personally I think the best thing about a Iris Pro rMBP is that one wouldn't have to deal with that crappy GPU switching solution of Apple's. Finally being able to use external displays for normal stuff without having to deal with an unnecessarily hot notebook because the dGPU is active. Even a HD 3000 can easily handle youtube or VLC on a second monitor. And even a stone old MHD 4500X could handle a power point presentation.
For a mobile notebook a dGPU 750M isn't really worth it for the minor speed bump. A 760M with a more intelligent gpu switching like Optimus would be. On an OSX notebook that won't happen; Apple has just been buying time to get to this kind of iGPU. Otherwise they would have fixed there uselessly inefficient muxing GPU switching a year ago.
It makes sense Apple seems to have been aming for the same thing with the MBA. About similar performance but great battery life.
Gamers won't like it but the majority of the people are better of without a 760M. For Pro rendering workload the difference isn't big enough to really complain and if somebody really needs lot of performance a 700M won't be any decent either and a Mac Pro is the way to go.
i have no idea how you could guess the needs of the majority.
if it performs as well as an overclocked GT650 then i'll be perfectly fine with it, anything less is not acceptable.
we need a comparison ( no no, not those useless game benchmarks or benchmarking softwares ) like CAD or Photoshop, rendering those on very high DPI to actually simulate real life performance.
telling someone to get a Mac Pro because rMBP isn't good enough is like offering customer a pizza when you're running out of orange juice.
I don't have the patience to read your whole post but I was using youtube to illustrate how many common things are now using the GPU. I also said Photoshop but you left that out of your quote to make an invalid point.
I've been holding out for a proper rMBP re-fresh. Partially because I hope they resolve the burn-in issues & my 2011 17" is still doing me ok. I planned to upgrade on the next bump. If it doesn't have discreet graphics, then I'm absolutely ****ed and have no idea what product I'm going to upgrade to...
It's not just performance with an intel chip. For games it's reliability and optimisation as well. ATI & NVidia rule the world. Is a game ported from PC > Mac & then further to a graphics card that isn't going to be given a 2nd thought for most game developers going to be any good at all?
I'm sick of hearing the "Eheeh Macs aren't for games, you're using it wrong" excuse. It is for me. My Macbook Pro is my main machine, and I want to do everything on it to a reasonable level.
It's not just performance with an intel chip. For games it's reliability and optimisation as well. ATI & NVidia rule the world. Is a game ported from PC > Mac & then further to a graphics card that isn't going to be given a 2nd thought for most game developers going to be any good at all?
I'm sick of hearing the "Eheeh Macs aren't for games, you're using it wrong" excuse. It is for me. My Macbook Pro is my main machine, and I want to do everything on it to a reasonable level.
Its not an excuse - just plain fact actually. Despite Apple trying to sell the Mac for games, its never come close to decent for anyone that understands the true gaming potential on computers... and for that matter, all laptops even purpose built PC gaming laptops are still quite a compromise.
Using a Mac for gaming is a bad excuse for not having proper gaming hardware.
I will. 13"rMPB. Will do video editing with it. Will not be an issue. Will stay away from 4K.
Btw - I would like 16GB RAM and a large SSD.
Btw - I would like 16GB RAM and a large SSD.
The 15" Pro and/or base retina Pro should at minimum include Iris Pro, and definitely not go with the 28W Iris.
If Apple didn't go with the 765M, could they go to a 760M with 1 GB? Could that drive the retina screen? I'm only going off what Razer uses and remember even the Blade Pro is only 1920x1080.
If Apple didn't go with the 765M, could they go to a 760M with 1 GB? Could that drive the retina screen? I'm only going off what Razer uses and remember even the Blade Pro is only 1920x1080.
There's a lot, so I won't quote anyone specifically, but here's a continuation:
1) No dGPU in rMBP 15" Haswell: as explained, the benchmarked CPU is a top-end model. Its clock speed is affected because of the TDP headroom from Iris Pro, but it's still considered a top-end product in its category. And it costs an arm and a leg. Even if Apple works out a deal with Intel, it still costs them a lot more than a non-Iris Pro CPU. They can't put it into a computer at a lower price point. So it obviously isn't a "low-end" product by any standard.
That CPU aside, there really is no worthy upgrade to the overclocked 650M in last year MacBook Pro. nVidia is just rebadging its GPU. The gain in performance would be minimal. They can go ATI, and there are viable alternatives that would provide quite a boost to performance, but it's clear they decided against that route. Most likely because they wanted to reduce the number of third-parties they had to make deals with, so only having to deal with Intel is favorable. Also they can negotiate better pricing by making a commitment. Apple was pushing Intel pretty hard for better integrated graphics after all.
2) Why Iris Pro may not be the solution: for desktop and 2D applications, we likely wouldn't see a problem. And for OpenCL, we may even see a boost. But other than that, Iris Pro isn't quite that fast at everything else. Namely OpenGL. And for a GPU-accelerated interface on OSX, OpenGL is more important than anything else.
Mavericks does improve graphics performance, but that's a global thing. If Iris Pro gets faster, so does the 650M in the last generation. The gap is still there.
Disregarding the interface, it's also clear that if iGPU is the only solution in use, then a portion of system RAM would be permanently shaved off to the iGPU. This may not bode well for systems with 8GB of RAM because Iris Pro may ask for up to 1GB.
And last but not least, it's very obvious that Iris (Pro or not) suffers from memory bandwidth issues. That's why they suck at rendering at higher resolutions. Even with the combined bandwidth of L4 cache and DDR3 at 100% efficiency (which is impossible, by the way), the 650M from last generation still has more bandwidth with GDDR5. Problem is... the Retina Macbook Pro is a case where Iris will have to push graphics at higher resolutions. It's not just for gaming, but also for 3D rendering applications... like AutoCAD, Maya, etc... and it's those applications that will hog memory bandwidth like crazy depending on the size of the project. (guess why Quadro GPUs need massive amount of video RAM)
So no matter how we try to look at it, Iris Pro is not a breakthrough. It matches the performance of the last generation in the best case, and in general, it wouldn't be comparable depending on what's running. For instance, Photoshop actually ends up using OpenGL more than OpenCL. OpenCL is only used for very specific tasks, but OpenGL can be enabled globally to accelerate drawing over the whole canvas over multiple layers. If you work with illustrations, it speeds things up tremendously especially at higher DPI settings.
Side note: battery life differences between Mavericks and Mountain Lion running Photoshop CS6, with all GPU-accelerated features enabled:
Mountain Lion: 3:30
Mavericks: 5:45
My guess is that Iris Pro would be able to get 8 hours under Mavericks doing the same task. That's roughly >30% battery life improvement, though performance may suffer a little tiny bit.
So the tradeoff is there. You just have to figure out which is more important. Slightly lower performance for some extra hours of work (this does make a difference especially if you're on an airplane), or smoother performance but losing 2 hours of work.
1) No dGPU in rMBP 15" Haswell: as explained, the benchmarked CPU is a top-end model. Its clock speed is affected because of the TDP headroom from Iris Pro, but it's still considered a top-end product in its category. And it costs an arm and a leg. Even if Apple works out a deal with Intel, it still costs them a lot more than a non-Iris Pro CPU. They can't put it into a computer at a lower price point. So it obviously isn't a "low-end" product by any standard.
That CPU aside, there really is no worthy upgrade to the overclocked 650M in last year MacBook Pro. nVidia is just rebadging its GPU. The gain in performance would be minimal. They can go ATI, and there are viable alternatives that would provide quite a boost to performance, but it's clear they decided against that route. Most likely because they wanted to reduce the number of third-parties they had to make deals with, so only having to deal with Intel is favorable. Also they can negotiate better pricing by making a commitment. Apple was pushing Intel pretty hard for better integrated graphics after all.
2) Why Iris Pro may not be the solution: for desktop and 2D applications, we likely wouldn't see a problem. And for OpenCL, we may even see a boost. But other than that, Iris Pro isn't quite that fast at everything else. Namely OpenGL. And for a GPU-accelerated interface on OSX, OpenGL is more important than anything else.
Mavericks does improve graphics performance, but that's a global thing. If Iris Pro gets faster, so does the 650M in the last generation. The gap is still there.
Disregarding the interface, it's also clear that if iGPU is the only solution in use, then a portion of system RAM would be permanently shaved off to the iGPU. This may not bode well for systems with 8GB of RAM because Iris Pro may ask for up to 1GB.
And last but not least, it's very obvious that Iris (Pro or not) suffers from memory bandwidth issues. That's why they suck at rendering at higher resolutions. Even with the combined bandwidth of L4 cache and DDR3 at 100% efficiency (which is impossible, by the way), the 650M from last generation still has more bandwidth with GDDR5. Problem is... the Retina Macbook Pro is a case where Iris will have to push graphics at higher resolutions. It's not just for gaming, but also for 3D rendering applications... like AutoCAD, Maya, etc... and it's those applications that will hog memory bandwidth like crazy depending on the size of the project. (guess why Quadro GPUs need massive amount of video RAM)
So no matter how we try to look at it, Iris Pro is not a breakthrough. It matches the performance of the last generation in the best case, and in general, it wouldn't be comparable depending on what's running. For instance, Photoshop actually ends up using OpenGL more than OpenCL. OpenCL is only used for very specific tasks, but OpenGL can be enabled globally to accelerate drawing over the whole canvas over multiple layers. If you work with illustrations, it speeds things up tremendously especially at higher DPI settings.
Side note: battery life differences between Mavericks and Mountain Lion running Photoshop CS6, with all GPU-accelerated features enabled:
Mountain Lion: 3:30
Mavericks: 5:45
My guess is that Iris Pro would be able to get 8 hours under Mavericks doing the same task. That's roughly >30% battery life improvement, though performance may suffer a little tiny bit.
So the tradeoff is there. You just have to figure out which is more important. Slightly lower performance for some extra hours of work (this does make a difference especially if you're on an airplane), or smoother performance but losing 2 hours of work.
If you read my post, you'd see I didn't leave that out. You show no sign of having any real understanding of how Photoshop uses the GPU and which processing resources affect performance for when it does. Hint: If the GPU is full engaged in a 750M like GPU accelerated filter comparison, the left over CPU speed doesn't matter, unless you encode with handbrake in the background.
Only way I would buy a non dGPU rMBP is if it were $1599, came with/upgradable to 16Gigs RAM, came with 256Gigs SSD and was likewise upgradable.
I can live with pretty good, but I don't want to be stuck with the same config the entire lifetime of the machine. Bad enough to compromise on dGPU, don't want to walk away with a greatly depreciated anchor after a year - which is very far from the case in present Mac machines. Well, I expect the 8Gigs rMBP being sold today to take a bigger hit on resale a year from now, than anyone seems to want to admit.
I can live with pretty good, but I don't want to be stuck with the same config the entire lifetime of the machine. Bad enough to compromise on dGPU, don't want to walk away with a greatly depreciated anchor after a year - which is very far from the case in present Mac machines. Well, I expect the 8Gigs rMBP being sold today to take a bigger hit on resale a year from now, than anyone seems to want to admit.
Yes of course, even if dGPU where at same price as Iris Pro i would go more for a Iris Pro rMBP 15" as for my needs Iris Pro rMBP 15" would be a perfect laptop as Iris Pro is pretty damn good for 2D tasks as what i mostly do on my laptop is Coding and since i prefer 15" screen over 13" screen so it would be really awesome from apple to release 15" rMBP with IRIS PRO only
Yes and i will buy the next rmbp only if it does not have a dgpu.
If your waiting for apple to release another MacBook with dedicated graphics you are gonna be waiting a long time!
Apple has 2 choices.
Announce a 15 inch which is basically 10% faster then the current model on graphics and CPU. About the same battery life.
Or release an iris pro version which is only slower on gaming (Opencl may well be faster on iris) , still gains like 5% in CPU but then they can claim all their MacBooks have all day battery life. Because of the energy savings and the bigger battery.( smaller PCB = bigger battery)
Not hard to predict which way apple is going IMHO. Especially because we have seen the geekbench score for such a machine already.
And for people suggesting a 765m please stop. Those razers get extremely hot even with an energy efficient cheap tn panel. Which is maybe acceptable in a gamers laptop but not for one aimed at pro's.
If your waiting for apple to release another MacBook with dedicated graphics you are gonna be waiting a long time!
Apple has 2 choices.
Announce a 15 inch which is basically 10% faster then the current model on graphics and CPU. About the same battery life.
Or release an iris pro version which is only slower on gaming (Opencl may well be faster on iris) , still gains like 5% in CPU but then they can claim all their MacBooks have all day battery life. Because of the energy savings and the bigger battery.( smaller PCB = bigger battery)
Not hard to predict which way apple is going IMHO. Especially because we have seen the geekbench score for such a machine already.
And for people suggesting a 765m please stop. Those razers get extremely hot even with an energy efficient cheap tn panel. Which is maybe acceptable in a gamers laptop but not for one aimed at pro's.