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

Discrete GT 330M vs Iris Pro Graphics

FrankySavvy

FrankySavvy

macrumors 68000
Original poster
Mar 4, 2010
1,691
872
Long Island, NY
So I currently own a 2010 15" Macbook Pro. It has the NVIDIA GeForce GT 330M with 512mb Discrete VRAM.

How does that compare with current generation non-discrete "Iris Pro" graphics in the new Macbook Pros?

Not sure I want to shell out over $2000 to get a discrete card in my next Macbook Pro. I am a graphic designer that primarily uses InDesign, Photoshop and Illustrator and I do play some games on occasion like Starcraft 2 etc.

If Iris Pro graphics is leaps and bounds better then Intel integrated graphics, I don't see the need for the discrete VRAM.

P.S I am waiting for Broadwell :)

Help!
 
The HD 4600 is generally a bit faster than the 330M. Iris Pro has the extra eDRAM and 40 vs 20 eu. Is is a lot faster than a 330M.
If the software code supports it porperly it is in professional apps often no worse and in some cases like opencl even better than the 750M.

Starcraft 2 is pretty bad in OSX regardless. Compared to Windows but the 750M has just a 30% lead at best. The game is in 4v4 multiplayer often CPU bound anyway. It runs just fine with Iris Pro on reasonable settings (medium-high 1080p). I personally don't see much of a difference to ultra settings except for physic effects. 750M doesn't allow for higher settings really, Windows offers more performance on either GPU than the 750M over Iris Pro on OSX.

For professional apps the difference depends on the developers more than the GPUs. Well programmed the difference should be minor except inspecial cases. In the future Cuda will likely see less support in favor of OpenCL and Intel is okay there until you compare workstation GPUs which don't exist on mobile Macs. Almost the entire lineup is Intel now so developers will pay more an more attention to it. In the past they could rely more frequently on certain users to have fast GPUs in iMacs or the big MBP.
 
FrankySavvy said:
How does Iris Pro designate Video Memory from System Memory? Are discrete video cards a thing of the past?
Click to expand...

I honestly couldn't tell you how the memory is allocated between the two. I don't believe discrete graphics are a thing of the past by any means. The discrete graphics in even the higher end 15" Retina outperforms the Iris Pro by quite a bit. It's quite obvious when doing things graphically intense and while playing games.

The Broadwell chips will definitely make integrated even better but by then Nvidia and AMD will have improved their discrete. Intel is playing catchup in regards to performance and are generally about a generation behind in performance.

Edit: in the OP "if Iris Pro are leaps and bounds above Intel integrate".......Iris Pro is produced by Intel and is part of the "Crystalwell" package.

Edit Edit: It's not all about the VRAM as it is the discrete being more powerful in terms of processing power.
 
Last edited:
FrankySavvy said:
How does Iris Pro designate Video Memory from System Memory? Are discrete video cards a thing of the past?
Click to expand...

Iris Pro reserves up to 1.5GB of system memory for graphics. It also has 128MB of L4 cache that is basically used as a buffer of dedicated ram.

750M is still faster, but Iris Pro isn't bad.
 
I will be mostly using the Adobe Creative Suite. On occasion I want to play a game, but nothing crazy. Will see what the next update brings. I just don't want to spend over 2000+

What would you suggest I sell my 2010 15" Macbook Pro for (specs below)?

----------
Freyqq said:
Iris Pro reserves up to 1.5GB of system memory for graphics. It also has 128MB of L4 cache that is basically used as a buffer of dedicated ram.

750M is still faster, but Iris Pro isn't bad.
Click to expand...

Where did you get this infer from? That is pretty sweet.
 
FrankySavvy said:
any idea what the next gen Iris Pro graphics will be pushing?
Click to expand...

There is no official information yet, but unofficially, it should have 48 execution units (vs. the current 40), so I'd expect it to be around 25-30% faster on paper.
 
Thanks for the replies everyone, and starting this thread OP. I have been curious about this exact topic for a while now.
 
leman said:
Why is this even important? The Iris has direct access to the system RAM, so it can use as much of it as it wants anyway.
Click to expand...

I guess my understanding of video cards is very minimal. These new cards that can take system ram and access as much as they please compared to discrete video cards. I just don't understand how they work vs a video card with its own discrete ram.
 
FrankySavvy said:
I guess my understanding of video cards is very minimal. These new cards that can take system ram and access as much as they please compared to discrete video cards. I just don't understand how they work vs a video card with its own discrete ram.
Click to expand...

There is nothing new about an integrated GPU. They simply don't have any own RAM. Its a part of the CPU, it uses the same cache and the same memory controller as the CPU. As well as sharing the memory bandwidth of the CPU. It makes an integrated GPU more flexible (because the driver can eliminate the data copy in many cases), but also obviously much slower in bandwidth-limited situations (e.g. when one uses many high-quality textures etc.).

The reason why Iris Pro is faster is simply because the CPU has a large L4 cache — its own 128Mb RAM module which is very fast compared to the system RAM. Because GPU operations often use coherent RAM chunks, frame buffers and textures can be often mirrored in that fast RAM block, and this allows the GPU to get the data without needing to hit the slow system RAM. This is a very flexible approach but its obviously not perfect. It will fail if the assets are very large or you have lots of them. But it works fairly well in practice.

This is why there were some reports that AMD is looking into a different solution, which is integrating GDDR5 memory controllers into their CPUs, and then offering hybrid RAM, consisting of both DDR3 and GDDR5. No idea how far they are with this, though. The recently released Kaveri does not seem to have this included.
 
leman said:
There is nothing new about an integrated GPU. They simply don't have any own RAM. Its a part of the CPU, it uses the same cache and the same memory controller as the CPU. As well as sharing the memory bandwidth of the CPU. It makes an integrated GPU more flexible (because the driver can eliminate the data copy in many cases), but also obviously much slower in bandwidth-limited situations (e.g. when one uses many high-quality textures etc.).

The reason why Iris Pro is faster is simply because the CPU has a large L4 cache — its own 128Mb RAM module which is very fast compared to the system RAM. Because GPU operations often use coherent RAM chunks, frame buffers and textures can be often mirrored in that fast RAM block, and this allows the GPU to get the data without needing to hit the slow system RAM. This is a very flexible approach but its obviously not perfect. It will fail if the assets are very large or you have lots of them. But it works fairly well in practice.

This is why there were some reports that AMD is looking into a different solution, which is integrating GDDR5 memory controllers into their CPUs, and then offering hybrid RAM, consisting of both DDR3 and GDDR5. No idea how far they are with this, though. The recently released Kaveri does not seem to have this included.
Click to expand...

So with the current Iris Pro, it is ideal in most situations. The discrete GPU such as the GT 750M w/ 2GB would only really be needed for the Hardcore gamer or 3D rendering.

I won't see a difference in everyday tasks and Adobe Creative Suite. Correct?
 
FrankySavvy said:
So with the current Iris Pro, it is ideal in most situations. The discrete GPU such as the GT 750M w/ 2GB would only really be needed for the Hardcore gamer or 3D rendering.
Click to expand...

I don't think that a hardcore gamer will be happy with the 750M ;) But you are essentially correct, except that GPU is rarely utilised for 3D rendering. Some newer versions of 3D software support OpenCL acceleration for rendering, and Iris Pro should fare very well there. I am not aware of any benchmarks though.

FrankySavvy said:
I won't see a difference in everyday tasks and Adobe Creative Suite. Correct?
Click to expand...

If you are not working with very large images or videos, no. Benchmarks do show that 750M is significantly faster when working with 4K video or insanely big photos — these are situations where its high memory bandwidth gives it the decisive advantage. But for a 'normal' user, there will be no difference.

The bottomline is that is depends on the kind of data and algorithms you are working with. If you have lots of data and fairly simple algorithms (e.g. simple filter/effect application), then 750M will shine, because these workflows are more sensible to the data transfer speed. If you have smaller data and complex algorithms (e.g. ray tracing, which requires lots of random memory accesses), the Iris Pro will win by a large margin.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back