Apple dumps NVidia from the Macbook Pro

[exar333]

Apple can keep the MacBook Pro at the same price, while likely getting charged considerably less for the dGPU from AMD. Cheap is cheap. Apple likes to make more money and AMD is ok with lesser margins vs. NV. It is probably as simple as that...

 

[crisium]

Apple + All 3 consoles means AMD has a constant revenue stream to stay alive. Profits be damned, they just need revenue. They are surely undercutting Nvidia by a lot, but they really have to.

 

[dacostafilipe]

Apple really pushes OpenCL? Final Cut Pro is about the only major product from Apple that uses OpenCL from what I have seen. I'd guess the real reason is AMD gave Apple a deal they couldn't resist. You know, why AMD's margins will remain 50% lower than their competition.

There's a lot of OpenCL in OSX, like the new Photos app. The trend seem to indicate that you will not be able to use OSX in the future without OpenCL.

Not a problem for most "normal" OSX users, but for those (like myself) that use Hackintosh, this could be a problem (aka less choice in GPU)

 

[AnandThenMan]

Second is the... Freesync technology. Retina 4K and 5K Thunderbolt Displays...

I never thought of that you could be right.

 

[Glo.]

Well if we will put things together:

OpenCL - far superior performance on AMD GPUs than on Nvidia.
Freesync, technology that will be extremely useful for Apple 4 and 5K Displays.
Mantle(Vulkan) - works perfectly with AMD GPUs.
As of right now AMD GPUs are in Mac Pro, iMac, Macbook Pro.

Well, Its not that bad perspective from my point of view.

 

[exar333]

Apple + All 3 consoles means AMD has a constant revenue stream to stay alive. Profits be damned, they just need revenue. They are surely undercutting Nvidia by a lot, but they really have to.

This is very true. Its a good fit for AMD because Apple is VERY price conscious (not for the end-user, but from their suppliers) to keep their margins high. AMD needs revenue and Apple is looking to cut costs. Perfect match!

 

[kasakka]

There's a lot of OpenCL in OSX, like the new Photos app. The trend seem to indicate that you will not be able to use OSX in the future without OpenCL.

Not a problem for most "normal" OSX users, but for those (like myself) that use Hackintosh, this could be a problem (aka less choice in GPU)

Yet it's still Nvidia that offers the most trouble-free experience with a Hackintosh while AMD requires driver hacks to get 3D acceleration working properly and even then there seem to be problems like some display outputs not working. It's actually quite interesting that Nvidia has kept pushing out web drivers that support even the desktop GPUs. I really hope that doesn't change in the future because I love having a dual boot OSX/Windows system on my PC.

 

[Enigmoid]

You are comparing desktop chips to mobiles? The 750Ti is not a mobile chip, its performance does not represent a 750M.

Also nVidia's OpenCL performance does not compare to AMDs, and the fact that they JUST NOW started to support OpenCL 1.2 (Which AMD supported YEARS AGO) is another issue. Apple has been pushing for 2.0 support, which Intel and AMD have.

The equivalent mobile chips (860m/850m/960m/950m) all use the same die at different clocks. It does not represent a 750m but the next gen replacement of the 750m.

 

[jpiniero]

I wonder how much of this has to do with Nvidia's patent lawsuits.

This is the most likely answer.

 

[tential]

I think that it's odd that amd doesn't market itself well. In all major consoles, and Apple products and not one advert stating that. Amd just doesn't use the opportunities it gets to market itself. They deserve the situation they're in. Good hardware, horrible everything else. They don't know how to play this game at all.

If you put nvidia in this same position, including cash etc, they'd turn this around in under 2 years with that marketing team I'm not even exaggerating. That's the lack of confidence amds team shows me.

 

[destrekor]

I cant believe they are charging $500 to upgrade from intel iris to a lowly M370X. They are basically shaving $100 off the cpu cost and adding a $200 gpu for a total net cost of $100 and gouging the heck out of the consumer for $400.

You get both GPUs. The system will use the Iris Pro GPU for light tasks, and will switch to the discrete GPU when heavy processing is detected.

And as others have said, you get the other upgrades as well. $500 buys you a discrete GPU with a dedicated 2GB of RAM, another 256GB of PCIe flash storage, and a faster processor.

 

[destrekor]

The equivalent mobile chips (860m/850m/960m/950m) all use the same die at different clocks. It does not represent a 750m but the next gen replacement of the 750m.

They use the same die, but they aren't at all the same save for different clocks. The mobile parts are heavily cut down, either with purposefully disabled cores or those parts that didn't pass inspection for their higher-revenue parts.

 

[Enigmoid]

They use the same die, but they aren't at all the same save for different clocks. The mobile parts are heavily cut down, either with purposefully disabled cores or those parts that didn't pass inspection for their higher-revenue parts.

There is absolutely nothing cut down about the 860m and 960m. The 850m and 950m are also fully enabled GM107 parts but with DDR3.

Mobile is higher binned and higher revenue that desktop. Some may be cut down but even then the higher end cut down chips end up in mobile.

 

[JDG1980]

The new MacBook Pro is supposed to be on sale today, isn't it? I'm surprised we don't have any benchmarks yet. No one seems to know for sure what the R9 M370X is; all we have is a questionable benchmark chart from an old rumor.

I think it would be great if the R9 M370X is a new chip, but I suspect it's probably just Bonaire.

 

[destrekor]

There is absolutely nothing cut down about the 860m and 960m. The 850m and 950m are also fully enabled GM107 parts but with DDR3.

Mobile is higher binned and higher revenue that desktop. Some may be cut down but even then the higher end cut down chips end up in mobile.

Umm... the 960 has 8 SMMs, the 960M has 5.
The 980 has 16 SMMs. The 980M has 12.
The 970 has 13 SMMs. The 970M 10.

Looks like the 970, a lower-bin from the 980, is a higher bin than the 980M.

Every single GM204 GPU that does not bear the name "GTX 980", is a cut-down/binned product. GM107 doesn't even exist in the desktop, so there I guess you can say they are mobile-specific and at least one model is sure to be fully enabled.

Also, I misspoke - the 960M is not a cut-down GM204 - it is a GM107 with lesser performance than the GM206-based 960.

So the GM107 models may be just different clock variants of each other, but they do not have a desktop comparison. The higher-end mobile 900 series all share the same GPU with the desktop series, but albeit, are binned/cut-down.

 

[JDG1980]

GM107 doesn't even exist in the desktop, so there I guess you can say they are mobile-specific and at least one model is sure to be fully enabled.

Of course they exist on the desktop. The first GM107 releases were the GTX 750 and GTX 750 Ti.

 

[RussianSensation]

Very lackluster update. Same chassis, same CPUs, barely improved the battery life. SSD is the only bright spot of the updates. Otherwise, this looks like a gen to skip as the next one will be Skylake and maybe a chassis update with USB Type-C.

 

[destrekor]

Of course they exist on the desktop. The first GM107 releases were the GTX 750 and GTX 750 Ti.

Whoops, so they were. I didn't look that far back, I forgot Maxwell debuted in the Kepler number-family.

 

[Silverforce11]

With the great efficiency of Maxwell, I really dont understand why Apple went back to AMD, unless like others said it has to deal with open GL vs CUDA. Either that or there is something political going on behind the scenes.

Nothing hard to understand, Apple ships apps with OSX that has native OpenCL acceleration, they've decided to leverage the iGPU from Intel, particularly Iris Pro to enhance performance of apps. It just so happens AMD has superior OpenCL support & performance, so when a dGPU is used, the only route that enhance their software is AMD dGPU.

This is a major reason for NV to push for better OpenCL.

 

[Silverforce11]

The new MacBook Pro is supposed to be on sale today, isn't it? I'm surprised we don't have any benchmarks yet. No one seems to know for sure what the R9 M370X is; all we have is a questionable benchmark chart from an old rumor.

I think it would be great if the R9 M370X is a new chip, but I suspect it's probably just Bonaire.

5K support will rule that rebadge out. It's either a cut down Tonga (unlikely due to the TDP associated on that platform!!) or a new chip using GCN 1.2+

 

[wilds]

I'm kind of annoyed at Apple still using the gmux switch in 2015. Will Apple ever support GPU switching in Windows or Linux without 3rd party support?

 

[Abwx]

People are still recycling this old argument which doesn't pan out.

Recycling old arguments, hey, with a DT GPUs review..?.

Are you sure that those Radeon Mxxx are not well reworked chips.?.

http://www.anandtech.com/show/9236/amd-announces-radeon-m300-series-notebook-video-cards

 

[Phynaz]

Apple really pushes OpenCL?

Apple invented OpenCL.

 

[BFG10K]

Apple + All 3 consoles means AMD has a constant revenue stream to stay alive. Profits be damned, they just need revenue. They are surely undercutting Nvidia by a lot, but they really have to.

Agreed - even Intel/nVidia users should be hoping for a profitable AMD. Competition benefits all consumers.

 

[JDG1980]

5K support will rule that rebadge out. It's either a cut down Tonga (unlikely due to the TDP associated on that platform!!) or a new chip using GCN 1.2+

It wasn't really full 5K - they said 5120x2160. Full 5K would be 5120x2880. I can't find the link, but I think someone did the math and found out that 5120x2160 @ 60 Hz is just within the bandwidth limits of DisplayPort 1.2.

I just did the math and it seems that this is the case:
60 fps * 24 bits per pixel (8 each R,G,B) * 5120 * 2160 = approximately 15.9 gigabits per second. DP 1.2 is rated for up to 17.28 Gbps.