NVIDIA/Intel Settlement Could Boost MacBook and MacBook Air Processor Speeds

I was beginning to wonder why the Core 2 Duo was still there...

if this means core i3/i5s in macbooks w/ the core 2 duo currently, then WOOT! took them long enough!

and that still means you won't get nVidia branded GPUs anytime soon...

I thought nVidia's CEO was quite clear that nVidia is out of the chipset business and the chipset team was transfered to work on Tegra, which supposedly nVidia's future. Even if they had some preliminary designs for Nehalem, Westmere, and now Sandy Bridge and they rushed the team back, I don't see how they can have anything production worthy in time for say a spring refresh of the 13" MacBook Pro. That also ignores the technical bandwidth limitations of handing a IGP of the DMI bus regardless of whether they get the license from Intel. They could hang the entire chipset of the PCIe bus, but that'd be kind of unorthodox and presents it's own risks including compatibility and probably more complicated drivers. Or they could stick with DMI and give the IGP it's own memory controller and dedicate VRAM, which increases power consumption and space usage partially defeating Apple's point of using a single chip IGP. None of these alternatives are optimal.

Interestingly, it has been confirmed that Sandy Bridge's IGP is OpenCL compatible, and graphics performance does seem decent admittedly not quite 320M class. The question is if Apple/Intel will provide decent OpenGL/OpenCL drivers, which hasn't been the case to date. Sandy Bridge's IGP is actually ideal for OpenCL since it shares the L3 cache with the processor so they can share data eliminating many of the bandwidth concerns for GPGPU through PCIe.

EDIT: It'd actually be latency concerns between standard IGPs or GPUs talking to the CPU, which Sandy Bridge's IGP addresses by sharing a L3 cache.

sounds good.

Not sure if Apple was in the loop on this settlement or not. If it was a surprise perhaps Apple is already too far in production of the next MBP refresh and this bit of news is coming too late?

I hope Steve had enough time to cancel the last truckload of C2D chips. :D

YES! Can't wait to get my hands on a 13 MBP with a Core i3/5 and Nvidia graphics! Add flash storage to that and Apple has just taken more of my money! :D :apple:

Charlie mentioned something about a cash settlement some time ago. Otherwise nVidia appears to have killed off and reallocated the personal for chipsets.

I'm not expecting anything.

I thought nVidia's CEO was quite clear that nVidia is out of the chipset business and the chipset team was transfered to work on Tegra, which supposedly nVidia's future.

Money works magic... watch. Also, Tegra hasn't really lifted up as it should have. More importantly, it was Tesla that nVidia said was their future. Well, I've got news, GTX480 or GTX580 based Teslas might be fast, but they suck so much power people shun them down.

Even if they had some preliminary designs for Nehalem, Westmere, and now Sandy Bridge and they rushed the team back, I don't see how they can have anything production worthy in time for say a spring refresh of the 13" MacBook Pro.

Which is why even if they are in talks, it doesn't mean any new chipsets until at least mid 2011.

That also ignores the technical bandwidth limitations of handing a IGP of the DMI bus regardless of whether they get the license from Intel.They could hang the entire chipset of the PCIe bus, but that'd be kind of unorthodox and presents it's own risks including compatibility and probably more complicated drivers.

You do realize DMI is just a x4 PCIe bus @ 10Gb/s? In our case, PCIe x2 @ 5Gb/s for mobile chipsets.

Or they could stick with DMI and give the IGP it's own memory controller and dedicate VRAM, which increases power consumption and space usage partially defeating Apple's point of using a single chip IGP. None of these alternatives are optimal.

No need.

Interestingly, it has been confirmed that Sandy Bridge's IGP is OpenCL compatible, and graphics performance does seem decent admittedly not quite 320M class. The question is if Apple/Intel will provide decent OpenGL/OpenCL drivers, which hasn't been the case to date. Sandy Bridge's IGP is actually ideal for OpenCL since it shares the L3 cache with the processor so they can share data eliminating many of the bandwidth concerns for GPGPU through PCIe.

Sandy Bridge IGPs will still suck even if Intel puts in OpenCL/GL. It's putting lipstick on a pig.

Charlie mentioned something about a cash settlement some time ago. Otherwise nVidia appears to have killed off and reallocated the personal for chipsets.

I'm not expecting anything.

From what I've read, Tesla is where their chipset engineers are now...

You do realize DMI is just a x4 PCIe bus?

Sandy Bridge is going to offer PCIe 2.0 lanes. I do remember stumbling upon the block diagrams for a possible MCP99 for a Nehalem/Westmere based system. It was a nasty mess of PCIe lanes and some DMI connections. The big question was the shared video RAM.

Next generation ION for Atom is stuck in a similar bandwidth limited situation hanging off of the NM10. Intel is closing the ranks. Try getting SLI support on an AMD board too. :rolleyes:

I can image Intel learning a few things about GPGPU from nVidia and a cash settlement but I'm not expecting the chipset department to spring back to life. Throw AMD into the mix with a ~480 shader IGP on their Llano solution. IGP solutions are going to need to be MUCH more robust than the Sandy Bridge GMA, Llano, or the current 310M/5470 ones.

You're looking at Mobility HD 5600 level power from something that's supposed to be cheaper than a discrete solution.

From what I've read, Tesla is where their chipset engineers are now...

I heard Tegra at last glance.