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 Prediction Poll: Sept? dGPU?
- Thread starter GSPice
- Start date
- Sort by reaction score
true, if we aren't talking about GK110 chip.
----------
for gaming - no. for GPGPU - yes, it is.
From the previous tests of desktop chips that are equal to 750m, 760m and 780m, it looks like 780m and Iris Pro are pretty similar in terms of GPGPU performance, 780m is probably faster a bit. But it's the 100w tdp chip alone, which will never be used in rMBPs, while Iris Pro is only a part of CPU that has 47w tdp total.
Yea.. I don't think so. The Iris Pro doesn't even match the 650M. Show us benchmarks of it consistently beating or being close to a 780M.
The only place the Iris Pro does consistently well is in synthetic benchmarks and even at it's best it only just hangs around with the 650M. For anything that requires massive memory bandwidth and pure brute force performance it falls flat.
The other part of the equation is that people who truly need a GPU for professional work don't take chances with Intel drivers. If you understand the needs of professionals as you say you do, I don't have to expand on that point any further.
Leaving the Intel drivers question aside, do you even know what GPGPU is? From what I see, it looks like you don't. And that's the reason, why I don't think we should discuss it any further.
He is talking about GPGPU not games which is all you seem to care about.
Kepler has always been rather poor in GPGPU at least the Geforce versions. Even Fermi outperformed it on that area while Fermi is much slower on games and stuff.
Update: Kaellar beat me to it.
Kepler has always been rather poor in GPGPU at least the Geforce versions. Even Fermi outperformed it on that area while Fermi is much slower on games and stuff.
Update: Kaellar beat me to it.
I've written fluid simulators which ran on the GPU and were used on various film projects so take your pretentious attitude elsewhere or stop using it as an excuse not to provide benchmarks to back up your claims.
----------
As I said, show me the benchmarks which show the Iris Pro consistently beating out even a 650M for GPGPU tasks that are more than simple synthetic tests. I've seen the Iris Pro doing well with synthetic benchmarks and on simple encoding tests but that's about it.
CLBenchmark fluid simulation is actually a great indicator of OpenCL performance difference between the GPUs, and I admit you've seen those benchmarks already. What real-life tests are you asking for? Those of more specific software most of which work is done on way more powerful desktop machines? Done on something almost not presented on the market yet? Is it me, or it's really trying to beat someones opinion simply by demanding something not present and not actually applicable to the situation?
Last edited:
there are no 5000 in the tdp range of the rmbp
though I still think its going to be 5200 + dgpu (I hope its amd, though there are people that need cuda) and the 5100 for the 13
I really hope that broadwell brings the best igpus to the 35w range and quads on top of that
basically it isn't worth to me to change from a SB mbp 13 to a haswell rmbp 13 if I don't get a leverage on cpu power. and surely I would opt for the cheaper mba that gives me the same power as the rmbp 13, albeit with the loss of the real state
Because
- the price is so high (in consumer minds),
- the Retina Macbook has a medium appeal (compared to old MBPs)
- the Intel 5XXX are barely more powerful than the weak HD 4000...
...I think they will still include a dGPU, even if it's still a GTM 650.
- the price is so high (in consumer minds),
- the Retina Macbook has a medium appeal (compared to old MBPs)
- the Intel 5XXX are barely more powerful than the weak HD 4000...
...I think they will still include a dGPU, even if it's still a GTM 650.
You need to amend your polling options to include a "year". Then change it to "Broadwell".
Doubt it'll be 5200+dGPU.
There are currently three CPUs with 5200. iirc, only two are released. Third is coming Q4. rMBPs have three different CPU configurations. Waiting for the third of the 5200, the high end version, seems likely why the rMBP isn't being released yet and will be next month or so.
That being said, it could mean that the 5200 is the only GPU. 5200 + dGPU would probably push the TDP too high.
I think what you are saying is that because the chip hasn't come out yet we don't have a lot of benchmarks? Just so we are on the same page lets define GPGPUs as a GPUs that can be utilised by software as processing units for any purpose other than 2D/3D rendering. So when you say Iris 5200 outperforms the 650m you are talking OpenCL performance yes ?
I am just wondering why apple can't put a workstation class dGPU like Nvidia Quadro K2100M on their rMBPs? I would consider it the ultimate beast of a laptop if it had that GPU. Solidworks 3D, Autodesk Maya, ProEngineer will run so fluently with rMBP with Iris Pro 5200 + NVidia Quadro K2100M. You can even play some serious games with K2100M as well. I'm very sure that the gaming performance should be as fast as a NVidia GeForce GT 750M based on the core and memory clocks.
Ok cool we have a baseline to work from.
The Iris 5200 is faster for openCL
The 650M is faster for games
So the most important question is :
Outside of very specialist computing (scientific/simulation etc) what commonly used software currently uses openCL and to how much of an improvement from pure CPU processing?
The chip's TDP is 55 + whatever CPU TDP is. A lot of battery drainage.
----------
Also, majority of the people who buy macs, don't buy macs for gaming.
I primarily game on a desktop where I can use my own GPU I picked. Or on a console.
If people game on a laptop, they're probably going to buy a gaming laptop.
Macs have never been known as gaming devices for hard core gaming. So why put a gaming class GPU in there?
You need to amend your polling options to include a "year". Then change it to "Broadwell".
We better not see anybody voting for September anymore!
I guess time will tell. The CLBenchmark fluid simulation does have good results for the Iris Pro but again it's a synthetic test. The reason why I'm not that happy with the Iris Pro is it seems whenever memory bandwidth becomes needed the performance falls sharply. That's why it does well with synthetic tests and not so much in "real world" tests.
there are actually 4 cpus with the 5200, 4750HQ, 4850HQ, 4950HQ and the 4960HQ
and the tdp is going to be raised by 2w from last year, not a game breaking thing, specially considering that the tdp for the dgpu actually lowered
Not everyone has the means/desire/ability to have a notebook plus a desktop and/or console(s). Not everyone who likes to game also likes gaming laptops, either, as they are usually large, heavy, and have worse battery life. There are those who prefer OS X, need a notebook, and would like to have the ability to game sometimes.
Personally, I do not like having more than one main computer, and building a desktop just for the few hours a week of gaming I get in is not practical. I need portability and something that looks professional so that leaves out a desktop and massive gaming notebooks with flashing lights. I prefer OS X to Windows but I also like to get in some gaming every now and then.
While likely a minority, I would surmise that there are others with similar requirements, which the 15" rMPB meets perfectly: thin, light, nicely designed, OS X, and decent casual gaming capability. I understand why Apple could be considering ditching the dGPU, but for me and others with similar requirements it would definitely be a disappointment.