How much performance increase (overall and 3D graphics-wise) will one gain if connecting the MBP 15" 2017 to an eGPU?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Waiting for the 2018 MBP
- Thread starter benface
- Start date
- Sort by reaction score
If it's optimized correctly? Tons.. Night and day difference. Search around on youtube..
On TB3 about 60-80% of the normal GPU power. It's gonna fly 
Apologies, I should have clarified that I meant having 4 cores instead of 2 with the introduction of coffee lake. I think this is the first time a 13" would have it.
With Coffee Lake, Intel should be able to provide still 15W dual- and now 28W quad- and 45W hexa-core chips, but given that's still on their same 14 nm process, something has to give in to stay in the same power envelope as current Kaby Lake chips.
Just as currently U-series chips either offer faster CPU but low-end HD iGPU, or slower CPU with higher-end Iris iGPU, I'd bet their solution to increase core count will be to offer 28W and 45W [slightly] slower CPU with low-end iGPU.
The rMBP13 is probably not designed to get a dGPU to compensate a worse iGPU, and the choice made with the redesign to get slimmer removing battery tells me the rMBP13 will in the future only get dual 15W chips (with Iris), and the rMBP15 switch to quad 28W chips with low-end iGPU but still a dGPU to counter-balance. The battery life gains would be sensible, the performances still good, and then with 10 nm Cannon Lake (and performances/watt growing again) the current design will reveal as a good choice.
Now, we can imagine Apple then capable to also offer a high-end rMBP15 (I doubt this would be a 17" model) with Coffee Lake hexa 45W chips + dGPU for high performances but still trading in exchange battery life.
Just as currently U-series chips either offer faster CPU but low-end HD iGPU, or slower CPU with higher-end Iris iGPU, I'd bet their solution to increase core count will be to offer 28W and 45W [slightly] slower CPU with low-end iGPU.
The rMBP13 is probably not designed to get a dGPU to compensate a worse iGPU, and the choice made with the redesign to get slimmer removing battery tells me the rMBP13 will in the future only get dual 15W chips (with Iris), and the rMBP15 switch to quad 28W chips with low-end iGPU but still a dGPU to counter-balance. The battery life gains would be sensible, the performances still good, and then with 10 nm Cannon Lake (and performances/watt growing again) the current design will reveal as a good choice.
Now, we can imagine Apple then capable to also offer a high-end rMBP15 (I doubt this would be a 17" model) with Coffee Lake hexa 45W chips + dGPU for high performances but still trading in exchange battery life.
Last edited:
We could see an increase in core count with either Coffee Lake or Cannon Lake... Why would Apple go to a 28W chip in the performance-focused 15" machine which has long accommodated a 45W chip? The only way to get real performance increases out of the Intel architecture seems to be to add cores - there is debate whether this is Moore's Law collapsing, or simply a particular architecture running out of steam, but per-core performance has been on a very slow slope for some time now.
This suggests that the 15" might get BOTH quad and hexa-core options. Assuming that chleusame is right that the way Intel gets power envelope for the extra cores is by reducing GPU power (sounds very logical, although I haven't seen elsewhere), the 45W power envelope will split into CPU focused hexa-core models with limited iGPU power and GPU-focused quad cores. The versions of the MBP without a dGPU will get a 45W quad with good integrated graphics. I would expect anything with the dGPU to get the highest CPU performance 45W chip available (with the possible exception of Xeons - but they might even stick a mobile Xeon in a top model).
The 13" is perhaps a trickier case - it can only take a 28W processor (and no dGPU). Assuming the 28W quad-cores have weak iGPUs, Apple will be stuck with 28W dual cores with the best GPU available - assuming Intel makes such a thing. If the best iGPU only appears on a 15W part, Apple will likely use a 15W dual core and boast about the battery life, since the 13" isn't a performance-focused machine.
This suggests that the 15" might get BOTH quad and hexa-core options. Assuming that chleusame is right that the way Intel gets power envelope for the extra cores is by reducing GPU power (sounds very logical, although I haven't seen elsewhere), the 45W power envelope will split into CPU focused hexa-core models with limited iGPU power and GPU-focused quad cores. The versions of the MBP without a dGPU will get a 45W quad with good integrated graphics. I would expect anything with the dGPU to get the highest CPU performance 45W chip available (with the possible exception of Xeons - but they might even stick a mobile Xeon in a top model).
The 13" is perhaps a trickier case - it can only take a 28W processor (and no dGPU). Assuming the 28W quad-cores have weak iGPUs, Apple will be stuck with 28W dual cores with the best GPU available - assuming Intel makes such a thing. If the best iGPU only appears on a 15W part, Apple will likely use a 15W dual core and boast about the battery life, since the 13" isn't a performance-focused machine.
I am waiting to buy the first 13" quad core computer from Apple.
Hopefully available when Coffee Lake is launched.
http://laptopmedia.com/news/intel-core-i5-8250u-is-coming-acer-swift-3-to-get-it-early/
I don't really see the point in a hexa-core in a laptop, but I guess a very small percentage needs that for their operation. May be worth it if you also have a eGPU docked at your desk, and do alot of video/3D work and want a single computer for all your work. Would still opt for iMac or Desktop PC if I was an editor. For dailys at a movie shot, it may be worth it to have a laptop with more power to see the "complete" picture.
Still a very tiny percentage of users. Laptops are simply not worth it anymore. I see very few road warriors with 15 inch laptops on cafes anymore. The exception may be the small percentage who don't have Internet access at home for various reasons. With mobile broadband and cheap handsets, I don't really see why people would buy a laptop with the exception of the STEM student, prosumer and professional. All of them are funded either by their business, through loans or by their private means.
Hopefully available when Coffee Lake is launched.
http://laptopmedia.com/news/intel-core-i5-8250u-is-coming-acer-swift-3-to-get-it-early/
I don't really see the point in a hexa-core in a laptop, but I guess a very small percentage needs that for their operation. May be worth it if you also have a eGPU docked at your desk, and do alot of video/3D work and want a single computer for all your work. Would still opt for iMac or Desktop PC if I was an editor. For dailys at a movie shot, it may be worth it to have a laptop with more power to see the "complete" picture.
Still a very tiny percentage of users. Laptops are simply not worth it anymore. I see very few road warriors with 15 inch laptops on cafes anymore. The exception may be the small percentage who don't have Internet access at home for various reasons. With mobile broadband and cheap handsets, I don't really see why people would buy a laptop with the exception of the STEM student, prosumer and professional. All of them are funded either by their business, through loans or by their private means.
Last edited:
This 15W CFL quad-core i5-8250U apparently only has a GT2 HD 620 iGPU.
CPU base clock would be either 1.6 or 1.8 GHz (depending on the sources), with turbo boost at 3.4 GHz.
And here is for GeekBench freaks: https://browser.geekbench.com/v4/cpu/3168934
The numbers are nice.
GeekBench OpenCL
Intel(R) HD Graphics 620
16898
A little bit above the Iris Pro 6100 (16648) in the MacBook Pro 2015.
If I can get the same power that the rMBP 15" 2015 possess in a 13" form factor I am back buying Apple. There must be a i7 model as well with more power.
Are those correct? My late 2013 13" MBP posts 19,897 in the openCL test.
The Iris Pro 6100 was in the march 2015 Broadwell rMBP13. Your Haswell 2013 model has the Iris 5100 (scoring 14907 for comparison).
The 2017 TB model got the Iris 650: its OpenCL score is 30173.
And the Iris 640 in the non-TB model gets 28739.
Well, that's just one benchmark anyway.
https://browser.geekbench.com/opencl-benchmarks
I don't think Apple will use this particular 15W CFL quad core chip in a future rMBP13 unless they have also room for a not-too-power-hungry dGPU.
Now, a 28W derivative, with a slightly faster CPU and a beefier Iris iGPU could do it, maybe. Is it feasible though? And how would it operate under load out of short benchmarks tests, in the rMBP13 enclosure and its cooling capabilities?
The 2017 TB model got the Iris 650: its OpenCL score is 30173.
And the Iris 640 in the non-TB model gets 28739.
Well, that's just one benchmark anyway.
https://browser.geekbench.com/opencl-benchmarks
I don't think Apple will use this particular 15W CFL quad core chip in a future rMBP13 unless they have also room for a not-too-power-hungry dGPU.
Now, a 28W derivative, with a slightly faster CPU and a beefier Iris iGPU could do it, maybe. Is it feasible though? And how would it operate under load out of short benchmarks tests, in the rMBP13 enclosure and its cooling capabilities?
Last edited:
As long as it is a 28 watt TDP CPU, the current design 13 inch MacBook Pro can cool it. TDP (thermal design power) defines what the cooling solution must be able to dissipate. The current 13 MBP with touch bar is already designed to dissipate 28 watts.
Yes. But KBL current dual CPUs base clock speed being faster, wouldn't it mean for many operations (not too threaded) the quad CFL model would need to run faster than its base speed more often, and then consume more power and generate more heat in comparison for equivalent performances?
An always hot thin laptop isn't really fun, if that's what it means to get a quad core at 14 nm in a 13" MBP.
An always hot thin laptop isn't really fun, if that's what it means to get a quad core at 14 nm in a 13" MBP.
Last edited:
My guess would be that with a base clock speed of 1.6-1.8Ghz, single core performance would take a hit on a quad core chip limited to a 28 watt thermal envelope. Unless it was able to turbo a single core very high and very quickly (which Kaby Lake can thanks to SpeedShift technology). I don't think it would necessarily translate to more heat generated, though I could very well be wrong.
After Skylake and Kaby Lake, maybe Apple will support Speed Shift indeed with Coffee Lake. I guess we won't know before the first CFL Mac and nothing will filtrate out of macOS betas before new hardware release, or even next WWDC.
CPUs, it's all about cores and reducing transistor size along with temps now. Most of the time you never max your CPU or even GPU, so default clock speeds like 1.8Ghz is more than enough to drive normal usage, especially if the OS utilize the new CPU instructions and optimizations. There is no need for a dGPU if you are not into video editing or gaming, which will require a desktop and external screen anyways if you want to enjoy all of the benefits a powerful setup can provide.
Last edited:
Well my score seems to be 30% above what is on there which makes me question the other numbers as there is a 30% increase and about 20% higher than a 2015 and possible 2018 model?
I have 16gb ram maybe that isn't accounted for? And all those are generic gpu tests and not Mac model specific.
That’s not an answer after your attempt to provide GPU benchmarks to serve your argument.
And what are you talking about? I’m not advocating for a dGPU solution, I want a decent iGPU solution for the rMBP13 (and I also wish the same for a low-end rMBP15...). I don’t want a quad core rMBP13 if it’s at the expanse of comfortable graphic capabilities that we now have with 15/28W intel chips.
We’re here discussing about a Macbook Pro, not some low-end PC laptop with crappy integrated GPU; Apple usually goes for balanced performances, and not the fastest CPU with only GT2 class iGPU. The solution here is similar, but rather than a faster CPU, it would be to lower speed to double the cores; that don't seem good to me.
You’re ok with 2015 performances for a 2018 laptop, I’m not open for this kind of regression; the experience with the internal retina screen wasn’t so good the first years before Skylake Iris in U-Series chips, and a prosumer laptop should be capable to drive an external screen or two without a dock+eGPU.
Now I repeat, if this particular Coffee Lake chip is a 15W model, we can imagine intel capable to make a decent 28W version with better iGPU and maybe even faster CPU base speed, and that could make it to a rMBP13.
Help yourself: https://browser.geekbench.com/v4/compute/search?q=macbook+pro+13-inch
Results fluctuates for some reasons but in general:
Iris 540/640 scores over 25k
Iris 550/650 scores just over 30k
Older rMBP13 are under or just reach 20k
But I repeat, that’s just a benchmark to be taken for what it is. doitdada will justify its validity as he was using it to try to make some point.
Last edited:
That’s not an answer after your failed attempt to provide GPU benchmarks to serve your argument.
And what are you talking about? I’m not advocating for a dGPU solution, I want a decent iGPU solution for the rMBP13 (and I also wish the same for a low-end rMBP15...). I don’t want a quad core rMBP13 if it’s at the expanse of comfortable graphic capabilities that we now have with 15/28W intel chips.
We’re here discussing about a Macbook Pro, not some low-end PC laptop with crappy integrated GPU; Apple usually goes for balanced performances, and not the fastest CPU with only GT2 class iGPU.
You’re ok with 2015 performances for a 2018 laptop, I’m not open for this kind of regression; the experience with the internal retina screen wasn’t so good the first years before Skylake Iris in U-Series chips, and a prosumer laptop should be capable to drive an external screen or two without a dock+eGPU.
Help yourself: https://browser.geekbench.com/v4/compute/search?q=macbook
Iris 540/640 scores just under 30k
Iris 550/650 scores just over 30k
Older rMBP13 are under or just reach 20k
But I repeat, that’s just a benchmark to be taken for what it is. doitdada will justify its validity as he was using it to try to make some point.
Thanks. Makes more sense than the previously posted numbers.
On the same page as you. Would love a quad core 13in with semi decent igpu. Luckily I'm not in desperate need of a new notebook as of now but want to upgrade in the next 12 months.
Was watching the recent keynote again last night. iMacs with USB 3.0, high-bandwidth SD slots, and Thunderbolt 3.0. It's just sort of a bummer the MBP doesn't have these also. The iMacs show that there isn't some dogmatic standpoint on ports, there's no reason why the MBP couldn't have them. But maybe the iMacs also light the way forward. I really hope USB 3.0 and SD come back to the MacBook Pro. If they can be on the iMac 'Pro', then why not Apples Pro portable? If going all in TB3 was a way of forcing the industries hand on pushing USB-C forward, i think it's worked. You can now buy cables and dongles everywhere. But that job is done now and Apple should bring these two ports back into the fray. It's just makes pros lives easier.
While I could see for many usages the value of a quad rMBP13, for myself included, you're the one willing an amputated quad core solution. Read again the last page of posts if you're not playing jerk here.
My bet was Apple won't make a quad core rMBP13 before at least Cannon Lake, if ever. I better see coming a Macbook line with the three sizes (simple rebranding of the 13" and 15" model, same design): on top of the fanless dual 12" model, a dual 15W Iris iGPU 13" model and a quad 28W 15" Iris iGPU (optional dGPU maybe) model, bringing back battery life and sufficient performances for the masses [with money/choosing macOS]. And to serve the higher-end market a 15" Macbook Pro model with better specs, 45W hexa core, dGPU, >4K screen, all that fits, to serve the most demanding prosumers wanting a powerful mobile Mac laptop.
If you can't understand the rest of the dialog, I can't help you.
You're not answering: how would you sell the regression in iGPU performances with your solution (still considering it's HD620 level of performances with the U-series Coffee Lake chip that was posted above)?
The Iris Pro (5200) found in the 2015 rMBP15 you take as reference scores 27k in the OpenCL benchmark you mentioned; the rMBP13 default resolution is now higher since the Skylake models, users expects to use 4K/5K external displays too.
My bet was Apple won't make a quad core rMBP13 before at least Cannon Lake, if ever. I better see coming a Macbook line with the three sizes (simple rebranding of the 13" and 15" model, same design): on top of the fanless dual 12" model, a dual 15W Iris iGPU 13" model and a quad 28W 15" Iris iGPU (optional dGPU maybe) model, bringing back battery life and sufficient performances for the masses [with money/choosing macOS]. And to serve the higher-end market a 15" Macbook Pro model with better specs, 45W hexa core, dGPU, >4K screen, all that fits, to serve the most demanding prosumers wanting a powerful mobile Mac laptop.
If you can't understand the rest of the dialog, I can't help you.
You're not answering: how would you sell the regression in iGPU performances with your solution (still considering it's HD620 level of performances with the U-series Coffee Lake chip that was posted above)?
The Iris Pro (5200) found in the 2015 rMBP15 you take as reference scores 27k in the OpenCL benchmark you mentioned; the rMBP13 default resolution is now higher since the Skylake models, users expects to use 4K/5K external displays too.
Last edited: