The new macbook pro besides the typical intel video card that it has use before, they have removed the nvidia card in the configuration for preference of an AMD video card.
AMD Radeon R9 M370X with 2GB of GDDR5 memory and automatic graphics switching
I wonder if anyone would like to opine about this move. Maybe I'm unfair to AMD. I have always have prefer NVIDIA. One of the reasons, which is not the best of the reason, is that if I want to run some cuda code, it can run there. Also, Nvidia seems to provide great video cards...
Anyways, I'm not trying to say that one is better than the other one because I just don't know. I wonder what is behind this? Just cost? better card?
Macbook pro moving to AMD Video Card
-
iblues1976
- Gnome
- Posts: 379
- Joined: Fri Sep 16, 2011 4:54 pm
- x 10
-
Kojack
- OGRE Moderator
- Posts: 7157
- Joined: Sun Jan 25, 2004 7:35 am
- Location: Brisbane, Australia
- x 538
Re: Macbook pro moving to AMD Video Card
Apple switches around a bit. imacs used to be amd, then they switched to nvidia around 2012.
The main thing I've always noticed with amd vs nvidia is that amd tended to be way more strict at things like shaders. For example amd didn't allow fixed function to mix with shaders (such as fixed function vertices with a pixel shader, or a vertex shader with fixed function rasterisation), while nvidia would allow it. Or mixing shader model 1/2 with shader model 3. This caused a LOT of problems in Ogre.
I don't know how it stands these days but I know years ago the basic rule was amd was faster at code with flow control and branching, while nvidia was faster at code with lots of sequential math (no branching or loops). This is at least partly because amd had a smaller warp size, it only worked on around 32 pixels at a time compared to much more (I think around 512) on nvidia. Lots of pixels in parallel is great when they are all the same flow, but when pixels differ in flow (some have a true if statement, some have a false, etc) the entire block is slowed down. Small blocks of pixels handle that better.
These days, who knowns. I haven't touched AMD since the 5780 card (which was damn cool for the time).
If I was writing gpgpu code for my own use and I had nvidia, I'd use Cuda. If I was writing it for others, I'd go with OpenCL so it's amd/nvidia/intel compatible.
I'm still annoyed at amd for dropping Render Monkey. Nvidia's fx composer is crap in comparison. No shader editor is as simple to use as Render Monkey, I still use it for SM3.0 stuff.
The main thing I've always noticed with amd vs nvidia is that amd tended to be way more strict at things like shaders. For example amd didn't allow fixed function to mix with shaders (such as fixed function vertices with a pixel shader, or a vertex shader with fixed function rasterisation), while nvidia would allow it. Or mixing shader model 1/2 with shader model 3. This caused a LOT of problems in Ogre.
I don't know how it stands these days but I know years ago the basic rule was amd was faster at code with flow control and branching, while nvidia was faster at code with lots of sequential math (no branching or loops). This is at least partly because amd had a smaller warp size, it only worked on around 32 pixels at a time compared to much more (I think around 512) on nvidia. Lots of pixels in parallel is great when they are all the same flow, but when pixels differ in flow (some have a true if statement, some have a false, etc) the entire block is slowed down. Small blocks of pixels handle that better.
These days, who knowns. I haven't touched AMD since the 5780 card (which was damn cool for the time).
If I was writing gpgpu code for my own use and I had nvidia, I'd use Cuda. If I was writing it for others, I'd go with OpenCL so it's amd/nvidia/intel compatible.
I'm still annoyed at amd for dropping Render Monkey. Nvidia's fx composer is crap in comparison. No shader editor is as simple to use as Render Monkey, I still use it for SM3.0 stuff.
-
EricB
- Bronze Sponsor
- Posts: 360
- Joined: Fri Apr 09, 2010 5:28 am
- Location: Florida
- x 213
- Contact:
Re: Macbook pro moving to AMD Video Card
With HBM GPUs right around the corner, and Zen coming in 2016. I wouldn't be surprised to see some Zen based APUs with HBM stacked GPU memory on the chip coming to Macs by 2017. Specially with the professional graphics marketshare Apple enjoys.
A good example of this is the speed improvements in Adobe software on Carrizo cpus. While I don't see Carrizo design wins for AMD with Apple. You do see them warming up to each other. As well as other OEMs landing Carrizo designs where they've largely ignored everything since Richland.
So interesting times are coming to AMD.
That being said, I'm the opposite. I haven't paid for a Nvidia product since the Geforce 440m. Bastards shipped a dx7 card with OpenGL 1.0 support up until 2008! Needless to say I was pretty pissed when my low spec Ogre game couldn't even run on it.
Disclaimer: I have short term financial interest in AMD shares.
A good example of this is the speed improvements in Adobe software on Carrizo cpus. While I don't see Carrizo design wins for AMD with Apple. You do see them warming up to each other. As well as other OEMs landing Carrizo designs where they've largely ignored everything since Richland.
So interesting times are coming to AMD.
That being said, I'm the opposite. I haven't paid for a Nvidia product since the Geforce 440m. Bastards shipped a dx7 card with OpenGL 1.0 support up until 2008! Needless to say I was pretty pissed when my low spec Ogre game couldn't even run on it.
Disclaimer: I have short term financial interest in AMD shares.
