Separate names with a comma.
Discussion in 'MacBook Pro' started by KensaiMage, Nov 18, 2017.
What do you expect?
I think the lifespan of Kaby will be relatively short-lived. I predict the next iteration of MBP will be coffee in June 2018. But, what the hell do I know?
Only Apple knows. Don't hold your breath waiting on it to announce anything.
Intel will push the 8th generation processors. These Kaby Lake-R, Coffee Lake, and Cannon Lake. I expect the MacBook Pros Kaby to use KabyLake-R and CoffeeLake . The Kaby Lake-R for 13", Coffee Lake for the 15". The 13 will be quad cores, and the 15" hex core.
Based on other laptop manufactures results, expect a considerable performance boast because Intel has done a good job balancing performance and power consumption. Lisa Gade does a good job explaining how in this video.
OPs question here is if it'll use the Intel-Radeon combo module, so it's not so simple as just the CPU generation. Coffee Lake's 'G' version could be much later, and Apple has prioritized graphics over a smaller CPU bump many times (320M instead of gen 1 i series, Iris 650 instead of a bump to the current 8th gen ulv quads, etc).
I'd like two extra cores, but the combo module is pretty sweet already, we're going from 80GB/s graphics memory to 180, 200 more shaders, etc.
Ideally both, but if I were to guess between the two, I think Apple would go with Kaby Lake G for this combo module that shrinks the MoBo space. Hopefully fit in more batteries for that.
The 8th gen quad-core chips were not available when Apple released the current MacBook Pro. Also, we're not speaking about a "smaller CPU bump" this time - it's huge! These 65W Kaby Lake G chips don't even offer GPU performance advantages, since AMD offers the GPUs used there separately, and considering the package power draw, I can't imagine that they need much more than a 35W TDP.
There is no mobile Vega with HBM2 outside of this Intel interposer, and dedicated packages also have less compute units than this (and correspondingly ROPs and TMUs). The shared package itself is 18% more efficient and there's no GDDR5, so this chip would still be the most ideal even if mobile Vega exists outside of it, and how it's trading blows with Pascal efficiency for once.
If we could get both the 'G' Package with Coffee Lake Hexacores, all my money would be flying at the screen, don't get me wrong. All I'm saying is that past history would indicate if Apple saw both a GPU advantage and a CPU advantage that could be used, they're leaning GPU as of the last decade.
There is a mobile Vega GPU - AMD showed it during CES. They didn't show any specs, but the chips looks awfully similar to the Vega 20 / 24 units on Kaby Lake G, and due to economic reasons alone I'd expect them to be similar. They probably will come with HBM2 memory too. It doesn't even have to be as efficient - the current MacBook Pro already uses an 80W CPU + GPU combination, they can stay at that.
And I don't think Apple values GPU performance more than CPU performance. Sure, they did use older CPUs in the past to avoid either the Intel GMA chipsets or the need for a dedicated GPU due to the lack of integrated ones. But Apple, in the past six years at least, always used the fastest CPUs available within their thermal limits, and arguably compromised on the GPU compared to a lot of the competition (Radeon Pro 560 vs GTX1060, for example).
I'm sure I've read somewhere the steady state performance settles in around 65W for CPU+GPU on a load for the 15, so the 65W package would already be ideal and would of course use any thermal capacity for turbo. Particularly with 18W within the system saved from power delivery, any bit that uses power = heat.
If indeed they wait for hexacores, that would be appealing, but I'd still hope it would be in a package like this - why turn their nose up at an 18% shared power efficiency gain as well as the ability to use more space for other things - like more cooling to be able to maintain higher performance longer, like they did by removing the HDD option from the iMac Pro - or a bit more battery capacity which is always nice, or microSD, etc etc. The smaller package still enables a lot of things regardless.
I bet on cannon lake G next month
But honestly, the Kaby Lake G would already be rather sweet. With dynamic power balancing the resulting performance is probably going to be a solid improvement from the last gen.
Intel probably won't offer any G-Series hexa-core chips this generation, I guess they already would have announced it if they wanted to sell them. So the option is to go for a 23% higher TDP for a 40% higher CPU performance or to skip on the additional performance. I sure know what I'd prefer them to do. Battery life shouldn't be far off either since the H-series CPU itself is more efficient, and the GPU will only be accessed when necessary anyway.
Also, the 15" MacBook Pro can supply up to 89W of sustained power. And even though it can, there aren't many situations outside of synthetic benchmarks which stress both the GPU and the CPU at 100% for a longer period of time, so it should be fine just as it currently is.
It's supplying power to a lot more than the CPU and GPU - the screen, antennas, chipsets, etc. By TDP I'm referring to the power use the CPU + GPU settle down on that the heatsink and fans can dissipate constantly. Look at the power consumption figures, even including everything else in the system it's not drawing the 89W you claimed (drawing 51W on the witcher 3, certainly loading both chips, and that's including powering the screen and all), and after the first few tests throttles down
The extra PSU wattage is one for overprovisioning as they lower in efficiency over time, and two probably for plugged in peripherals, the GPU and CPU in their throttled state cannot keep using 89W because they hit temperature limits first.
As for not having much use for both the CPU and GPU over time - well, I'd wonder why one is getting a 15" comparably large and expensive system? Video editors are certainly a core demographic and long exports over dozens of minutes or longer certainly do use both chips. And if you look at an iMac that benchmarks similar to it, you can see how the macbook throttles down over time and is slower in the long run despite a similar benchmark score in the short run. If an opportunity to make one part smaller in exchange for improving cooling, batteries, and much else existed, I'm not sure why we wouldn't want that.
(it was in here, can't find where you can replay it yet...)
Of course I'm aware of the physics of the matter and a laptop will never dissipate heat as well as even a large AIO desktop, but literally any percentage improvement there would be appreciated, certainly a lot of people who get a 15" rMBP get it to...Well, make use of a 15" rMBP.
As for Intel announcements, well, yeah, that's fair enough. Though Apple has also surprised us with dual FMAs and AVX-512 down to the 8 core Xeon up through the top for example (FMAs scale with core count outside of these Apple requested Xeons), maybe they can coax more 'slightly custom' packages out of Intel, even if not as semicustom as AMD lets them.
I guess you're right, but even video editing doesn't constantly stress the CPU and GPU to a 100% for longer periods of time. While the video is rendering, yes, but not while working on it.
Also I think the word "throttling" is thrown around far too much lately. Most people seem to say "throttling" as soon as a notebook can't keep the maximum turbo frequenzy forever, but that's pretty much any notebook. A six-core chip with the same TDP as current quad-cores would certainly help with that.
And don't get me wrong - I'd love to see a six-core G-series chip, I just can't see it happening. From what I gather, these chips are already at least as expensive as a separate CPU and GTX1050, so if Intel made six-core versions soon, they'd need to price them even higher, otherwise those quad-core versions would instantly be redundant. And with even higher prices, most PC OEMs would rather go the traditional route. Apple may be an exception here, but I'm not convinced yet that the actual efficiency gains are as high as Intel claims. We'll have to wait until Vega Mobile is launched before we can say for sure though.
As I see it, and as Intel markets them, G-series chips are aimed at comparably thin and light gaming notebooks, not at MacBook Pro type ones. And it makes sense - six-core CPUs provide a much smaller benefit for most games as opposed to working in a multitude of apps.
I would have said kabylake G a couple of weeks ago, but having seen the release information 65W is way above the thermal limits for a 13 inch MBP and 100w is way above the TDP for the 15 inch and the lack of a quadcore and hexcore version respectively makes me think it may be unlikely.
Coffee Lake in March, Cannon Lake late this year/early next.
Just a guess.
Those SDP numbers are not carved in stone.
maybe not but if they down-clock them enough to get within TDP then the performance hit and the lack of quad-core with radeon and hex-core with radeon will put apple off I'm afraid. If I was designing them I would much rather have a quad-core with iris graphics in the 13 inch and a hex-core with a dGPU in the 15 inch, in thin and light pro laptops.
We still don't know, unfortunately, when the hexacore mobile chips are landing. In the meantime, a 80-85W Kaby Lake G will be a solid upgrade in both the CPU and the GPU performance (if the shared power balancing really works well).
Since there's no rumors, it seems that we will not see an upgrade on march, therefore will probably not be kaby lake G...
A 4 cores with a 4Tflops Gpu would make a macbook a beast on FCPX and more than ok for gaming.
Seeing that other companies are releasing laptops with kaby lake g, I think its possible. I really hope they release a macbook pro with Vega M GH, going from 2.6 gflops to 3.6 gflops would be a great improvement.
Coffee Lake mobile CPUs are not even announced yet whereas Windows laptops equipped with Kaby Lake G CPUs are starting to appear and will continue doing so in the first half of this year. I think it's inevitable that a MacBook Pro with 4 cores and the Radeon graphics package will appear and I'd say it'll be sooner than later.
I think Apple could be saving the 6 core Coffee Lake MacBook Pros for after they have released a 3rd generation of the present design. A redesign with the Coffee Lake CPUs could arrive early next year - the earliest iteration that could see Apple address keyboard issues and touchbar expense.
--- Post Merged, Mar 6, 2018 ---
Two complaints about the MacBook Pro 2016/17 can be addressed with relative ease by Apple with the new combined CPU+GPU.
1. Battery life
I would have thought that the 65w Kaby Lake G off the shelf packaged combinations offer significant power savings over a Kaby Lake + AMD Pro graphics solution and thus improving battery life on the same units through lower power draw. The fact that there's an i5-8305G with smaller cache and slightly lower clock speed might also allow Apple to reduce the entry price of the 15" MacBook Pro if they choose to use it.
Indirectly speaking, depending on Apple's plans for the MacBook Pro follow-up they might have been able to engineer better keyboard experience by the time of the third generation if they can make room with a smaller motherboard and battery (smaller battery because of using 65W TDP parts). The 65w parts only do 2.6T Flops whereas the 100w parts (which aren't going into the MacBook Pro have 3.7TFlops)
--- Post Merged, Mar 6, 2018 ---
Where do you have these numbers from? From all benchmarks I've seen, the Kaby Lake G GL is slower than the desktop GTX 1050, which has a floating point performance of 1.8 TFLOPs, or just about the same performance as AMDs Radeon Pro 560 (notebook). Judging by this, a 65W Kaby Lake G would just about equal the current 15" MacBook Pro in both CPU and GPU performance, which is not a desirable thing considering the option of a much faster Vega-based GPU and a six-core CPU.
Battery life and price are also question marks rather than facts. Intel did for whatever reason not specify the price of the Kaby Lake G chips, but their NUCs using these chips are VERY expensive compared to the previous ones, which might make this option not even cheaper than using separate components. Same for the power draw - while it might be lower under load compared to the current 45W CPU + 35W GPU setup, a refreshed CPU + GPU setup would certainly provide more power. And if you actually want to increase battery life while you're at it, just swap the 45W CPU for a 15W one – this would bring down the total CPU + GPU TDP to 50W while offering the same CPU and superior GPU performance compared to Kaby Lake G.
My source was single precision T-Flops quoted from Ars Technica - forgot to link the page - but obviously benchmarks can be picked and chosen. The HBM memory on that package will surely help too - I don't recall seeing any standalone VEGA benchmarks/prices in the wild yet.
I guess we'd have to see what Apple's ultimate plans are if they did indeed have a hand in specifying their needs to Intel - they are clearly a customer that Intel are aiming for with this and they'll have their own priorities going forward.
Everyone's battery life will vary but you have to say if the new combination will offer 20% better power consumption along with a generational increase in numbers that Phil Schiller can reel off during a presentation then I suspect Apple will be happy enough. I also note that the Kaby Lake G CPU range uses DDR4-2400. Not the low power version - so there will be some extra penalty to pay there unless it can use Apple's preferred LPDDR3. Intel's page suggests 64Gb RAM max but I think that's with 4 slots. With 2 slots Apple could then provide - and power - the 32Gb of RAM that some users have been craving.
The current CPU/GPU combination in the 2017 MBP is bound to be closer to 85w which accounts for the some decent performance. What if Apple are about to drop a 32Gb MacBook Pro because all of a sudden they have the power budget to provide it?
Apple aren't going to wholesale drop the class of CPU used in a particular line unless they have a good reason.
I'm dubious as to whether or not they'd use the i5 version of the Kaby Lake G CPU but for the fact that I think they are seeing good sales figures for discounted 2016 and 2017 models judging from the perpetual deals going on at the moment.
I agree on the fact that mobile coffee lake aren't yet out and apple usually arrives 6 month after it's officially used by every other manufacturer.
And I also don't think apple could justify going from nearly 4Tflops + 4 Cores on a generation, and then 6 cores and less than 4Tflops.
Improvements have been made on mobile GPU for sure, but there's no way a 6 core mobile CPU + GPU could be better than the Kaby lake G on power consumption.
Maybe Apple will use it for a new generation of Mac Mini?
Actually it would be very good for them to sell it for 1000$. Would love to see something like that.
I need more CPU power but I also need GPU power, yet I really don't know what I would prefer for the Macbook...
At the same time the eGPU market is STILL not mature enough : we need a solution that's 100% compatible with every macbook, every software, bootcamp, etc. It doesn't exist yet.
Actually, they are rather speedy in adopting new CPUs. Its just that the CPUs are often released in waves and the tier that Apple uses is not available yet .
A separate mobile Vega will certainly be as fast as the on-package Vega M GH, just less power- and space-efficient. So yes, these packages are ideal for Apple. In fact, I hope for a Coffee lake + Vega package. Should't be a problem for Intel probably, they just need to replace the CPU die...
You are being too optimistic here. The CPU/GPU package alone will probably cost more then $500...