MacBook Pro with a dedicated Nvidia GPU (15" or 17")

[kennyman]

Hello,

I have recently been tasked to tag and manage some old media files (original files are on Adobe Premiere and After Effects CS4 & CS5).

I require CUDA to run the mercury playback engine smoothly. I am not very familiar with MacBook Pros. I mostly use a Mac Pro and a MacBook Air on a daily basis.

Now that I am on the road, the small screen and low spec CPU on the MBA is slowing my work down. I would like seek for your advice, looking for a MBP 15/17 inch with a dedicated Nvidia card.

Any recommendations? Many thanks

 

[Glmnet1]

The 17" is discontinued since 2011.

The latest MBP with an Nvidia card is the Mid 2014 15". Apple doesn't sell it anymore so you would have to look for a used one.

Another possibility would be to buy a 2016/2017 and an external Nvidia GPU but for now they often require some tinkering before they work so I wouldn't recommend it unless you're comfortable with that.

 

[kennyman]

The 17" is discontinued since 2011.

The latest MBP with an Nvidia card is the Mid 2014 15". Apple doesn't sell it anymore so you would have to look for a used one.

Another possibility would be to buy a 2016/2017 and an external Nvidia GPU but for now they often require some tinkering before they work so I wouldn't recommend it unless you're comfortable with that.

Thanks, I guess I will have to go with the mid 2014 15" then. I would like to avoid the external GPU kit, it is not very practical while moving around (travelling a lot lately).

 

[Poki]

I'm not familiar with the Mercury playback engine, but have you done your research if there's an alternative that runs on AMD GPUs? AMD has done quite some work with software developers to increase compatibility of their render engine and reducing the reliance on CUDA.

 

[Glmnet1]

I just read about the Mercury playback engine and it's part of Adobe CS5 and one of its features is that it can use CUDA to optimize some operations.

So, I think you'd be better off buying a more recent MBP, I'm pretty sure it's going to run better than the 2014 one even without the CUDA optimization. Also, they are now optimizing for Metal and it appears to be as good or even better than for CUDA.

 

[leman]

The current MBPs with AMD GPUs are 3x or more faster than the NVIDIA cards on previous laptops. I seriously doubt that CUDA support can overcome this kind of advantage, and if it can, Adobe should probably be sued for fraud and dishonest customer treatment

 

[kennyman]

Just want to update this post, I have successfully managed to tag all my old video files. I bought a used MBP from eBay, CUDA is a must for the old CS4 and CS5. Adobe were kind enough to provide additional support for my old plug ins.

I actually like the MBP 2013 and I am going to keep it, my mid 2017 MBA struggles with big files. I always thought that the latest Intel chipset should be overall better in terms of bus speed, IO and peripherals. But in actual usage, the 2013 MBP is way faster than my 2017 top-of-the-line MBA. I am going to throw the MBA on eBay to recoup cost.

I was just comparing geekbench scores, comparing MBP 2013 to MBP 2017 (comparing Intel chipset and processor specs).

MBP 2013
Single-Core Score is 4308
Multi-Core Score is 15666
Processor is i7 4960HQ
Cores/Threads is 4/8
Cache is 7MB
Base Frequency is at 2.6 GHz
Turbo Frequency is at 3.8 GHz
TDP is 47W

MBP 2017
Single-Core Score is 4969
Multi-Core Score is 16999
Processor is i7 7920HQ
Cores/Threads is 4/8
Cache is 9MB
Base Frequency is at 3.1 GHz
Turbo Frequency is at 4.1 GHz
TDP is 45W

So Intel did what in 4 years - these are just incremental updates.

I have also tested the battery life and it is very strange, I get almost 60-75 minutes more out of the MBP 2013. I thought MBA should be more efficient than a 4-core processor.

I have the same workflow everyday, couple of apps and couple of taps opened, screen is at 65% brightness with Bluetooth and Wifi:

I get 10.5 hours max out of late MBP 2013 with Yosemite 10.10.1
I get 9 hours max out of my MBA 2017 with the latest High Sierra Build

It is very difficult to accept that an old 4-year old MBP gets better battery life than a MBA (with no dedicated GPU). This does not make any sense - any input?

 

[leman]

I actually like the MBP 2013 and I am going to keep it, my mid 2017 MBA struggles with big files. I always thought that the latest Intel chipset should be overall better in terms of bus speed, IO and peripherals. But in actual usage, the 2013 MBP is way faster than my 2017 top-of-the-line MBA. I am going to throw the MBA on eBay to recoup cost.

There is nothing surprising about that. The MBA has been all but discontinued and hasn't received any real technology updates since 2015. The Haswell in the 2013 and the Broadwell in the currently sold MBA has only minor architectural differences and you are pitting a 45W quad-care vs a 15W dual-core here.

I was just comparing geekbench scores, comparing MBP 2013 to MBP 2017 (comparing Intel chipset and processor specs).

[...]

So Intel did what in 4 years - these are just incremental updates.

Because Geekench is barely representative of real-world performance. I wish people would stop looking at this benchmark. Whether you will notice a difference between Haswell and Skylake depends on what to do with your laptop, but Skylake is around 20% more efficient clocl-per-clock under optimal circumstances.

I have also tested the battery life and it is very strange, I get almost 60-75 minutes more out of the MBP 2013. I thought MBA should be more efficient than a 4-core processor.

Yeah, that is super weird. Shouldn't be happening. My guess is that there is some problem with either your MBA battery or your software setup. Or you are super lucky and got an MBP with an exceptionally good battery

 

[cruisin]

It is very difficult to accept that an old 4-year old MBP gets better battery life than a MBA (with no dedicated GPU). This does not make any sense - any input?

The MBA has great battery life if it spends most of it idling. If it is struggling, then it is running at full power (can hear the fans) and not saving any battery. The MBA does not get the same battery life no matter what you do with it.

The MBP doe not use the dedicated GPU unless it needs to (to save power) and it has more CPU power available. So things that take a bit will always be faster and the CPU can go back to idling while the MBA is still struggling (and needing more power). Plus your MBA was designed to be small and efficient at the cost of power and so it will naturally be slower.

 

[mj_]

Because Geekench is barely representative of real-world performance. I wish people would stop looking at this benchmark. Whether you will notice a difference between Haswell and Skylake depends on what to do with your laptop, but Skylake is around 20% more efficient clocl-per-clock under optimal circumstances.

20% more efficient does not necessarily mean 20% faster. While it's true that Skylake is a "tock" in Intel's tick-tock model (a microarchitecture redesign using the same 14nm manufacturing process introduced with Broadwell) the actual performance differences are negligible. It is, however, much more power-efficient. See for example this article: https://www.pugetsystems.com/labs/articles/Haswell-vs-Skylake-S-i7-4790K-vs-i7-6700K-641/

Thus, OP isn't wrong when he says that the differences between Haswell and Skylake are rather small. You can measure the difference running various benchmarks but you will never really be able to tell the difference between the two (unless of course you're comparing a dual-core with a quad-core and even then it depends on applications used).

The rest of your post I agree with. The MBA is not only ancient technology but also a dual-core with lower clock speeds. It is thus quite obviously much slower than the MBP, even an older 2010 or 2011 MBP.

 

[leman]

20% more efficient does not necessarily mean 20% faster.

Its exactly what it means. If you have two equally clocked CPUs and code that can mot efficiently utilise the CPU's resources, the Skylake should complete around 10-20% faster than the Haswell. This has to do with additional execution unit as well as other architectural improvements. Of course, no applications will probably approach these ideal conditions since there is always some stalling here or there. Real-world difference between Haswell and Skylake is more likely somewhere between 0% and 15%, depending on application.

I actually did a quick benchmarks a year ago, feel free to look at it: https://forums.macrumors.com/threads/an-opinionated-cpu-benchmark-of-the-2016-15-mbp.2026475/

 

[mj_]

Its exactly what it means. If you have two equally clocked CPUs and code that can mot efficiently utilise the CPU's resources, the Skylake should complete around 10-20% faster than the Haswell. This has to do with additional execution unit as well as other architectural improvements. Of course, no applications will probably approach these ideal conditions since there is always some stalling here or there. Real-world difference between Haswell and Skylake is more likely somewhere between 0% and 15%, depending on application.

Again, it's not exactly what it means. Efficiency and performance are two different things. A CPU can be more efficient without being any faster at the same time. If CPU A and CPU B need a certain amount t to complete a certain calculation but CPU A's power consumption was 10% lower than CPU B's then CPU A would be considered more efficient than CPU B without being any faster at all. And it goes beyond that if you look at more than mere IPC (instructions per clock cycle): if CPU A is 5% slower than CPU B for a specific calculation but CPU A's power consumption is 10% lower than CPU B's it would still be considered more efficient since it offers more computation power per energy unit than CPU B.

Efficiency does not equal performance. Methinks we're talking about two different things here: efficiency vs. effectiveness.

 

[leman]

Again, it's not exactly what it means. Efficiency and performance are two different things. A CPU can be more efficient without being any faster at the same time.

[...]

Efficiency does not equal performance. Methinks we're talking about two different things here: efficiency vs. effectiveness.

Ok, fair enough, its a semantic issue English is not my native language. I used "efficiency" as in "can do more work in the same time", but I can totally see that this is ambiguous (or maybe even plainly wrong).

 

[jerry70450]

LOL nividia gpus. we're not gonna see those in macs for a very long time especially with intel - amd partnership