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.
Comparing Iris Pro vs 750m
- Thread starter Atomic Walrus
- Start date
- Sort by reaction score
DDR3 is nothing like GDDR5 when it comes to gaming performance. Just look at the performance delta it causes between the Xbox One (DDR3 + EDRAM) and the PS4 (GDDR5).
It will make a big difference, it's much faster, all dedicated and designed with gaming in mind.
Yup, if gaming is what you do, the 750 would generally do better then the 650 (do take note you'll have to use the 660 as the rMBP GPU is closer to it and comes with 1gb GDDR5) and the 750 performance will be closer to the 755 (but there's no data for 755).
Things are a little different when it comes to computation softwares though (I'm looking at those CAD Softwares)
Iris Pro vs GeForce 660M
SolidWorks
15 fps vs 9.08 fps
Siemens NX (higher end then SolidWorks)
2.26 fps vs 4.3 fps
Its a little odd that Iris Pro is faster with Soldiworks but slower in NX considering both uses similar graphics engine to display the 3D models (ironically, SolidWorks is using the graphic engine owned by Siemens)
Any idea when is Anandtech review coming up? I'm still trying to decide to get the Iris Pro only rMBP or the Iris Pro + Nvidia GPU rMBP
Anandtech usually take their time but it's always worth it. No one I respect or trust more, some very intelligent and well educated reviewers.
I'm sure it won't be long but it will be thorough. No doubt they'll be as intrigued as we are to look at the GPU's and their respective performance.
Just be patient. For me I know I'm going for the higher end 750M, that was the plan all along. So I just need to wait for it to arrive.
You're right the 650M in the Ivy rMBP is more like a 660M as they are golden samples and are overclocked accordingly but I imagine the 750M to be the same, it's the same architecture, we will know soon enough. Getting tired of reading speculation and misleading benchmark results.
I'm sure it won't be long but it will be thorough. No doubt they'll be as intrigued as we are to look at the GPU's and their respective performance.
Just be patient. For me I know I'm going for the higher end 750M, that was the plan all along. So I just need to wait for it to arrive.
You're right the 650M in the Ivy rMBP is more like a 660M as they are golden samples and are overclocked accordingly but I imagine the 750M to be the same, it's the same architecture, we will know soon enough. Getting tired of reading speculation and misleading benchmark results.
So if you're just using CAD and video work (assuming extra battery life is not important), the Iris Pro would be better?
I thought it's the opposite? Since at Gaming the Nvidia chip mostly beats the Iris Chip while at OpenGL softwares, the Iris Pro usually beats the Nvidia chip?
----------
I guess good things take time after all
I thought of going for the 750M as well but since CAD, Photo, Video, 3D rendering is more important then gaming and earlier IrisPro benchmark show it's generally better at these applications and that Apple made the price of a similar spec rMBP (without the Nvidia GPU) the same as the one which has, makes the decision much harder
Since I'm aiming to keep this machine for the next 5 years, gotta be sure I made the right decision
DDR3 is nothing near GDDR5, and the whole siphoning of your systems resources (which is also actively being used by your system) has to have additional drawbacks.
You can't just "solve" the OpenGL problem. Iris Pro isn't designed to be good at OpenGL. If you want it to be good at OpenGL then you'd have to completely redesign it to basically act like a dGPU, in which case it wouldn't be nearly as god at OpenCL anymore. You can't really have both (unless of course, you have both chips) But we've yet to see how that works with the two side by side yet. Maybe the system chooses which chip is better for each circumstance, but that's doubtful.
Last edited:
If dGPUs are going to stick around (and I have to imagine after things came so close this year Nvidia will be fighting back seriously next year) Apple really needs to find a better way to handle the switch.
I thought things might have gotten better in MBPs released after the 2010 machine I've been using, but it's just not the case. Even loading Chrome still forces the dGPU on, which is absurd when we're talking about a maximum 20% performance gap between the chips with serious power consumption differences.
I'm fine with using gfxCardStatus to regulate the situation myself, but that shouldn't be necessary. Worse yet Apple is still unwilling to support use of the iGPU in Boot Camp, so using the machine in native Windows on battery is generally a bad idea (in fact this is one major reason someone might intentionally pick an iGPU-only model).
Another anecdotal data point: Borderlands 2 finally brings the 15-20% performance difference I was expecting to see from the start. The 5200 actually performs quite well, but the 750 leaves room for extra graphical features or higher resolutions (though I'm really liking the crisp no-blur pixel doubling mode that some games are able to deliver at 1440x900.. no idea why that doesn't work on the desktop if you explicitly set 1440x900).
It's worth noting that running the 750m results in lower per-chip temperatures and lower fan RPMs. Noise levels are similar since both fans ramp up but with the 5200 the CPU gets quite hot by comparison.
I thought things might have gotten better in MBPs released after the 2010 machine I've been using, but it's just not the case. Even loading Chrome still forces the dGPU on, which is absurd when we're talking about a maximum 20% performance gap between the chips with serious power consumption differences.
I'm fine with using gfxCardStatus to regulate the situation myself, but that shouldn't be necessary. Worse yet Apple is still unwilling to support use of the iGPU in Boot Camp, so using the machine in native Windows on battery is generally a bad idea (in fact this is one major reason someone might intentionally pick an iGPU-only model).
Another anecdotal data point: Borderlands 2 finally brings the 15-20% performance difference I was expecting to see from the start. The 5200 actually performs quite well, but the 750 leaves room for extra graphical features or higher resolutions (though I'm really liking the crisp no-blur pixel doubling mode that some games are able to deliver at 1440x900.. no idea why that doesn't work on the desktop if you explicitly set 1440x900).
It's worth noting that running the 750m results in lower per-chip temperatures and lower fan RPMs. Noise levels are similar since both fans ramp up but with the 5200 the CPU gets quite hot by comparison.
some quick tests on the rmbp:
15", 8GB, 256GB, 2.0GHZ, Iris Pro, Windows 8.1
2880x1800 all settings ultra (aa is only ever fxaa, and turning it off does nothing)
cs:go 47fps
dota2 35fps
tf2: 62fps
guild wars 2: 17fps (24fps with shaders & shadows med)
---
under osx: VRAM(total) 1024MB
Graphics Adapter Information Windows
Total Avaliable Graphics Memory 1792MB
Dedicated Video Memory: 32MB
System Video Memory: 0MB
Shared System Memory: 1760MB
so it seems windows is sharing more ram with the graphics, however i dont understand why the dedicated is listed as 32MB when crystalwell is 128MB
15", 8GB, 256GB, 2.0GHZ, Iris Pro, Windows 8.1
2880x1800 all settings ultra (aa is only ever fxaa, and turning it off does nothing)
cs:go 47fps
dota2 35fps
tf2: 62fps
guild wars 2: 17fps (24fps with shaders & shadows med)
---
under osx: VRAM(total) 1024MB
Graphics Adapter Information Windows
Total Avaliable Graphics Memory 1792MB
Dedicated Video Memory: 32MB
System Video Memory: 0MB
Shared System Memory: 1760MB
so it seems windows is sharing more ram with the graphics, however i dont understand why the dedicated is listed as 32MB when crystalwell is 128MB
Last edited:
It's the max. Mavericks can assign tot he Iris Pro what it needs at that moment, up to 1GB.
Just as I thought to be pretty sure that Iris Pro is what I want, I start all over again by reading this thread. Can't wait for some solid info, from people here or from anandtech.
I'm starting to think that in my case, both Iris Pro only or Iris Pro + 750M are bad. Iris Pro seems the way to go specifically for Solidworks (being even better then the GTX 780M in this case), it's a Windows only program and I would have to avoid the 750M (or else it's always on in bootcamp). On the other hand, the 750M seems to be much better at other things which I also intend to do/use. But if I get the IP + 750M, my Solidworks performance will be much worse.
So no rMBP for me? Damn this entire situation -_-
I'm starting to think that in my case, both Iris Pro only or Iris Pro + 750M are bad. Iris Pro seems the way to go specifically for Solidworks (being even better then the GTX 780M in this case), it's a Windows only program and I would have to avoid the 750M (or else it's always on in bootcamp). On the other hand, the 750M seems to be much better at other things which I also intend to do/use. But if I get the IP + 750M, my Solidworks performance will be much worse.
So no rMBP for me? Damn this entire situation -_-
Is this true? From the comments I have read, it's not so easy to simply turn of the 750M in Windows. Or is there really such an easy solution that nobody knows?
I've never heard of this actually working before (from what I've read disabling the dGPU simply disables some driver functionality, like removing your GPU drivers on Windows). Everything I can find suggests that Boot Camp does not provide any way to access the iGPU, which Windows sees as not "connected" to the system.
I'll likely end up testing this sometime this weekend either way, as soon as I figure out how a Windows 8.1 Boot Camp installation works, but I'm sure someone already knows the details.
This really seems like something that Apple should be working on fixing, doesn't it? I can't see any reason they would want to discourage full-featured use of Windows on their machines, as this is a major selling point: The only laptops that run both OS X and Windows.
what do you mean? i was running those at 2880x1800.
in the morning i'll do some more extensive testing, (native on max; what settings needed to reach 60fps on native; 1440x900 results) including screenshots for proof
Where are you getting this information from? OpenGL is a graphical API, whereas OpenCL is for computations. OpenCL is designed to allow multiple processor types to be involved in workloads - that includes discrete GPUs. That the Iris Pro supports OpenCL doesn't mean that its OpenGL performance is poor; it's a false dichotomy to say that a processor can only be good at one or the other.
Solidworks is not the only app which is apparently is faster with the Iris Pro; Maya and Lightwave too per notebook check website.......I would imagine Spaceclaim falls in that category too!
Remember though that if you're not trying to build an entire automobile assembly in SW, parallels may be adequate.
I'm running Spaceclaim in parallels 9 on my 2011 15" MBP..
Hello, this is my first post. I have been following this forum while waiting the release of the Haswell MBPr. It appears that there is a confusion regarding IrisPro vs 750m performance in OpenCL ang OpenGL. Assuming that both OpenCL and OpenGL drivers are well-written for both GPUs, the statement "this hardware is better in OpenCL and worse in OpenGL than another hardware" is wrong. Both APIs access the same hardware. In OpenCL you can control the parallelism a little bit better while in OpenGL the developer of the driver is responsible for panellizing the computations when for example you ask the GPU to draw a polygon on the screen. So, if the drivers are perfect the performance of the two APIs should be the same. Something that I have to mention here is that one of the main bottlenecks in GPU applications is the memory access. The more you access the GPU memory the worse it becomes. Now, why Iris seems to be stronger in OpenCL? I would guess that the OpenCL software used to benchmark the hardware can fit all its data in the on-chip memory Iris Pro has. As a result it doesn't suffer much when accessing the memory. 750m doesn't have this on-chip memory and this results poorer performance. Now, if you find an OpenCL application that uses lots of memory (and also accesses this memory frequently) and the on-chip memory of IrisPro is not enough to accommodate all the data, then IrisPro performance should drop. That's why 750m gets ahead in games, because games usually need lots of memory for textures, geometry, frame buffer objects and so on.
Having said that, I am not quite sure what kind of OpenCL applications were used to benchmark IrisPro. I just didn't have free time to investigate this.
Thank you.
Having said that, I am not quite sure what kind of OpenCL applications were used to benchmark IrisPro. I just didn't have free time to investigate this.
Thank you.
First High-end Macbook Pro Retina (late 2013, 15″) benchmark
Hi,
I was one of the lucky few who received their high-end macbook pro yesterday. Today I took the time to run a few benchmark to compare the Iris to the 750m. Overall on synthetic benchmark (heaven), it seems that the 750m is like 33% faster but on Starcraft 2 it is close to 45%.
Interestingly enough when I compared to the first generation (but on OSX 10.7) the 750m is less performant than the 650m so there might be a driver issue. My friend didn't have Starcraft 2 so I can't compare on SC2.
Overall, I feel that unless you need a ton of ram (e.g for photo editing) or you want to play there is not a whole lot of value to get the high-end.
I wrote a small blog post with all the benchmark results if you are interested: http://bit.ly/1ij77d5
Hope this help answer your questions.
Hi,
I was one of the lucky few who received their high-end macbook pro yesterday. Today I took the time to run a few benchmark to compare the Iris to the 750m. Overall on synthetic benchmark (heaven), it seems that the 750m is like 33% faster but on Starcraft 2 it is close to 45%.
Interestingly enough when I compared to the first generation (but on OSX 10.7) the 750m is less performant than the 650m so there might be a driver issue. My friend didn't have Starcraft 2 so I can't compare on SC2.
Overall, I feel that unless you need a ton of ram (e.g for photo editing) or you want to play there is not a whole lot of value to get the high-end.
I wrote a small blog post with all the benchmark results if you are interested: http://bit.ly/1ij77d5
Hope this help answer your questions.