15" MacBook Pro (Touch Bar) dGPU questions

[sleim031]

Hey everyone,

I believe that I am not the only one wondering about the 15" GPU department...

To start, I've ordered a 2.9 GHz, Radeon Pro 460 4GB, 1TB SSD 15" MacBook Pro with the built in Intel HD 530 iGPU.

So how does the HD 530 compare to the previous generation's Iris Pro (iGPU in high end 15")? On paper it does look weaker but is it THAT weaker than previous gen Iris Pro? I understand that they opted for it because of a better power efficiency (?) but does the whole Skylake thing (considering the 2.9 GHz processor) make the HD 530 a bit more powerful than purely on paper? HD 530 will probably be used for less demanding tasks like browsing the www, word processing and all that and it should deliver a performance (in those areas) like if not better than Iris Pro of the last years 15"?

Secondly, the dGPUs are the new Pro series as you know. From what I've been looking at and from the available data on the Pro series, the lowest end Pro (500) is similar (?) to the R9 M370X of the previous gen 15". I believe it should be at least a bit more powerful than the previous generation (even though this is a comparison between this year's BASE level 15" and last years top uncustomised 15").

For someone that works in Premiere Pro CC, Lightroom, Photoshop et al, dealing mostly with 1080p video (I plan on doing 4k soon), I believe that the 15" should prove useful for the task since the last generation did for so many YouTubers who use it as a daily driver and editing their stuff.

I want to hear what you think on all this guys?

 

[maflynn]

So how does the HD 530 compare to the previous generation's Iris Pro (iGPU in high end 15")? On paper it does look weaker but is it THAT weaker than previous gen Iris Pro?

Why do you think the dGPU in the new MBP is weaker then the iGPU in the older MBP? Where have you seen information that causes you to think that?

 

[sleim031]

Why do you think the dGPU in the new MBP is weaker then the iGPU in the older MBP? Where have you seen information that causes you to think that?

Im comparing the integrated Intel HD 530 in the new 15" with the integrated Iris from last year; not considering the dGPUs for this specific comparison.

So presumably the HD 530 will be used for the same stuff the Iris was used for, day-to-day, without kicking in the dGPU. I'm wondering will the HD 530 deliver the same if not better performance in those areas than last year's Iris. From what I've read thus so far, people don't have much love for the HD 530 here.

 

[maflynn]

Im comparing the integrated Intel HD 530 in the new 15" with the integrated Iris from last year; not considering the dGPUs for this specific comparison.

I see, sorry for the mixup. I thought you were looking at the dGPU vs. the iGPU.

I've heard mixed things regarding the new iGPU, but I'd like to see some solid numbers first

 

[NickPhamUK]

The iGPU is weaker than the one in MBP 13" and it is supposed to drive more pixels! I'd rage if I spend that much of money to see my computer lagging when loading heavy safari pages.

 

[sleim031]

The iGPU is weaker than the one in MBP 13" and it is supposed to drive more pixels! I'd rage if I spend that much of money to see my computer lagging when loading heavy safari pages.

I see what you're saying but I'm really hoping that there's more here than meets the eye. It doesn't seem logical (esp. for Apple) to underpower a new laptop, especially top-end.

The iGPU performance depends, to some extent, on the rest of the hardware and its speed; there's only benchmarks from Windows laptops so far online so I don't know whether that will be the same as in the MBP

 

[lympero]

The iGPU is weaker than the one in MBP 13" and it is supposed to drive more pixels!

Still can't understand how they decided something like that.

 

[jun180]

Interesting, I didn't realise the 15" didn't use the iris iGpu.

Have you seen any benchmarks for the iris 540/550 vs HD 530?

 

[lympero]

Interesting, I didn't realise the 15" didn't use the iris iGpu.

Have you seen any benchmarks for the iris 540/550 vs HD 530?

Benchmarks will show that hd 530 (15 model) will be weaker than hd 540 (base 13 model) and you'll have ui lag in your shiny 15 inch mac forcing you to buy the updated next year model which will solve all these first generation problems.

 

[torana355]

So how does the HD 530 compare to the previous generation's Iris Pro (iGPU in high end 15")? On paper it does look weaker but is it THAT weaker than previous gen Iris Pro? I understand that they opted for it because of a better power efficiency (?) but does the whole Skylake thing (considering the 2.9 GHz processor) make the HD 530 a bit more powerful than purely on paper? HD 530 will probably be used for less demanding tasks like browsing the www, word processing and all that and it should deliver a performance (in those areas) like if not better than Iris Pro of the last years 15"?

"For Skylake, Intel also gave non-Pro Iris GPUs an eDRAM cache—a smaller 64MB pool of it, yes, but any cache at all is better than none. The result is a 13-inch MacBook Pro with an integrated GPU that’s a little bit faster than the integrated GPU in the 2014/2015 15-inch MacBook Pro"

Quote from the Arstechnica review, it seems the new Iris 540 is actually more powerful then the Intel GPU in last years 15" model. Im surprised the iGPu in this years 15" model is less powerful then the 13" model even if it does have the DGPU.

"Apple doesn't use these GPUs in MacBook Pros anymore, though; the 15-inch models that used to use them now use Intel HD 530 GPUs without any eDRAM, leaving the heavy lifting to the AMD Polaris dedicated GPUs."

It seems the dedicated GPU might take on more responsibility so to speak.

 

[dallas112678]

According to wikipedia the 530 (15") has a little over HALF of the performance as the 540 (13"). What's comical is they put a dedicated GPU (radeon 450) in the base 15" that's only 25% more powerful than INTEGRATED graphics in the 13". I imagine they didn't put the stronger iGPU in the 15" is because they were already putting in a barely more powerful dGPU.

Apple truly managed to mess up the graphics so horribly, all to shave 1 lb and .1" of thickness off of this computer. It's absolutely astounding.

This makes the laptop so much more unprofessional than the lack of 32GB of RAM.

https://en.wikipedia.org/wiki/Intel_HD_and_Iris_Graphics#Skylake

 

[torana355]

So if OSX only uses the 530 to drive the UI and only uses the 460 for more intense tasks there will probably be UI lag on the 15" Glad i got the 13" model now!

 

[enzoshadow]

According to wikipedia the 530 (15") has a little over HALF of the performance as the 540 (13"). What's comical is they put a dedicated GPU (radeon 450) in the base 15" that's only 25% more powerful than INTEGRATED graphics in the 13". I imagine they didn't put the stronger iGPU in the 15" is because they were already putting in a barely more powerful dGPU.

Apple truly managed to mess up the graphics so horribly, all to shave 1 lb and .1" of thickness off of this computer. It's absolutely astounding.

This makes the laptop so much more unprofessional than the lack of 32GB of RAM.

https://en.wikipedia.org/wiki/Intel_HD_and_Iris_Graphics#Skylake

Are you comparing GPU performance of different GPU architecture purely by FLOPS? Are you for real?

 

[dallas112678]

Are you comparing GPU performance of different GPU architecture purely by FLOPS? Are you for real?

Seeing as that's a very popular method of comparing general performance between cards (even those of various brands), yes. Now if the point you are trying to make that Flops don't really matter, which you are, then why would AMD even bother to come out and report those numbers?

There's no point in reporting those numbers if it's not supposed to give people a general idea of performance. Simple.

 

[enzoshadow]

Seeing as that's a very popular method of comparing general performance between cards (even those of various brands), yes. Now if the point you are trying to make that Flops don't really matter, which you are, then why would AMD even bother to come out and report those numbers?

There's no point in reporting those numbers if it's not supposed to give people a general idea of performance. Simple.

http://www.guru3d.com/articles_pages/msi_radeon_rx_480_gaming_x_review,17.html
http://www.guru3d.com/articles_pages/msi_radeon_rx_480_gaming_x_review,25.html
RX 480: 5.8 TFLOPS
GTX 980Ti: 5.63 TFLOPS
Notice how much performance difference they can be depending on applications?
And NO, no one will tell you it's a popular method of comparing performance. You are embarrassing yourself.
Do you compare Pentium 4 4.0GHz to Skylake Quad 2.7Ghz as well? Do you really need to know why they need to list that out?
It makes perfect sense why you already bashing it. You only judge the final product by spec sheet, and you don't even know how to read them correctly.

 

[kokhoong0624]

So if OSX only uses the 530 to drive the UI and only uses the 460 for more intense tasks there will probably be UI lag on the 15" Glad i got the 13" model now!

Certainly hope this is not the case because I really want the 15 inch model

 

[Softwarez]

From what I've read here and there, the Polaris dgpu will be far more active in the new MacBook than previous gens.

That aside though, even the iGPU is capable of rendering UI animations with no dropped frames. I don't understand how people can possibly think it would lag. The UI animations aren't Crysis 3 people. An iGPU can even handle 4K UI'a without a problem. The MacBook is no problem.

On a more technical note, the advantage of the Polaris dGPU is its GDDR5 vram. The iGPU only has a measly 64MB of edram which is much slower than GDDR5. And when that cache runs out, it will have to use system ram which is even slower.

Polaris' architecture also makes it particularly good at FP8 and FP16 compute tasks, which is what the MAJORITY of MacOS' animations consist of (since it's OpenGL based). In fact, AMD architectures are usually better at compute than nvidia GeForce cards, which is what most of these AMD cards will be seeing anyways (accelerated photoshop, video rendering)

 

[leman]

1. The HD 530 in the new 15" is substantially slower than the Iris Pro 5200 in the 2015 model and also slower than the Iris graphics in the 13". Not that it matters at all. The iGPU will only be used for desktop compositing and its more then able to deal with that task comfortably.

2. The Pro 450 will most likely be flightily faster than the M370X (similar theoretical performance peak, but increased efficiency and higher memory bandwidth). But we will need benchmarks to quantify the difference.

Still can't understand how they decided something like that.

Most likely, they didn't have much of a choice. It fairly evident by now that Intel was unable to manufacture a sufficient quantity of Iris Pro 580-enabled Skylake CPUs.
[doublepost=1478156215][/doublepost]

On a more technical note, the advantage of the Polaris dGPU is its GDDR5 vram. The iGPU only has a measly 64MB of edram which is much slower than GDDR5. And when that cache runs out, it will have to use system ram which is even slower.

AFAIK, the HD 530 doesn't have any eDRAM.

 

[dallas112678]

http://www.guru3d.com/articles_pages/msi_radeon_rx_480_gaming_x_review,17.html
http://www.guru3d.com/articles_pages/msi_radeon_rx_480_gaming_x_review,25.html
RX 480: 5.8 TFLOPS
GTX 980Ti: 5.63 TFLOPS
Notice how much performance difference they can be depending on applications?
And NO, no one will tell you it's a popular method of comparing performance. You are embarrassing yourself.
Do you compare Pentium 4 4.0GHz to Skylake Quad 2.7Ghz as well? Do you really need to know why they need to list that out?
It makes perfect sense why you already bashing it. You only judge the final product by spec sheet, and you don't even know how to read them correctly.

You do realize that comparing Flops is not the same thing as comparing CPU clock speeds, right?

You do realize that there are other things such as drivers that can affect performance, right? How many times have we seen a driver release drastically improve the performance of the same card? Lots of times.

You do realize the 980Ti was release a full YEAR before the RX 480, meaning it will surely have more completely developed drivers for certain applications, right?

I guess the folks over at Top 500 who rank the most powerful super computers along with writing all sorts of articles on new and advanced technologies are just a bunch of dummies as they rank them by flops, right? Oh, and no, those super computers also don't have the same architecture and component composition just incase you were wondering.

Remember, when they tell you Flops, they are saying PEAK PERFORMANCE, it's written wright on AMD's slide. Clock speeds were never a measure of peak performance, and manufacturers never said they were.

Trust me, the only one embarrassing themselves here is you.

 

[Softwarez]

2. The Pro 450 will most likely be flightily faster than the M370X (similar theoretical performance peak, but increased efficiency and higher memory bandwidth). But we will need benchmarks to quantify the difference.

Most likely, they didn't have much of a choice. It fairly evident by now that Intel was unable to manufacture a sufficient quantity of Iris Pro 580-enabled Skylake CPUs.

Note that the M370X is Cape Verde GCN 1.0. And Polaris is GCN 4.0. I would imagine the architectural improvements are significant. But of course, benchmarks are needed to confirm or refute it.

As for why no iris in the 15 inch. Easy. Every 15 inch now comes with a dGPU. There's simply no need for a powerful iGPU since the dGPU would take over in graphics intensive tasks.

 

[leman]

Note that the M370X is Cape Verde GCN 1.0. And Polaris is GCN 4.0. I would imagine the architectural improvements are significant. But of course, benchmarks are needed to confirm or refute it.

Yep. Polaris is much more efficient at utilising its performance (meaning that it will more easily reach its theoretical peak) and it also has advanced color compression, meaning that it get get much more out of its VRAM bandwidth.

P.S. In fact, it seems that Polaris has an edge over Pascal in color compression, which would make it a better choice in a bandwidth-limited device such as a laptop?

 

[Softwarez]

You do realize that there are other things such as drivers that can affect performance, right? How many times have we seen a driver release drastically improve the performance of the same card? Lots of times.

You do realize the 980Ti was release a full YEAR before the RX 480, meaning it will surely have more completely developed drivers for certain applications, right?

I guess the folks over at Top 500 who rank the most powerful super computers along with writing all sorts of articles on new and advanced technologies are just a bunch of dummies as they rank them by flops, right? Oh, and no, those super computers also don't have the same architecture and component composition just incase you were wondering.

Remember, when they tell you Flops, they are saying PEAK PERFORMANCE, it's written wright on AMD's slide. Clock speeds were never a measure of peak performance, and manufacturers never said they were.

Trust me, the only one embarrassing themselves here is you.

Like you said, flops is a good way to gouge the theoretical max performance of various cards in terms of the peak raw throughout they're capable of. But reality is much more complicated than that.

Things like compression algorithms, optimized data paths for specific compute tasks, and various other architecture level design decisions can alter the performance of cards depending on different workloads.

So in terms of real world benchmarking, flops isn't the most accurate way of comparing real world performance. Its useful as a quick comparison metric, but any conclusions based on flops alone is silly.

Also, your reference to super computers isn't really relevant. The way they function is completely different. And the programs they run are written to be optimized for the specific super computer being used.

 

[torana355]

That aside though, even the iGPU is capable of rendering UI animations with no dropped frames. I don't understand how people can possibly think it would lag. The UI animations aren't Crysis 3 people. An iGPU can even handle 4K UI'a without a problem. The MacBook is no problem.

The 2015 rMBP had UI lag, some people don't notice it but I certainly do.

 

[Softwarez]

The 2015 rMBP had UI lag, some people don't notice it but I certainly do.

Strange. I've never experienced it. Which iGPU and which version of MacOS?

Maybe there's things running the background slowing it down.

 

[dallas112678]

Like you said, flops is a good way to gouge the theoretical max performance of various cards in terms of the peak raw throughout they're capable of. But reality is much more complicated than that.

Things like compression algorithms, optimized data paths for specific compute tasks, and various other architecture level design decisions can alter the performance of cards depending on different workloads.

So in terms of real world benchmarking, flops isn't the most accurate way of comparing real world performance. Its useful as a quick comparison metric, but any conclusions based on flops alone is silly.

Also, your reference to super computers isn't really relevant. The way they function is completely different. And the programs they run are written to be optimized for the specific super computer being used.

I never said Flops was a perfect way to compare performance, as I already mentioned something as simple as drivers could change it. All I said was what you just did, that it gives a decent idea of the possible performance, with some variations.