Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

casesam

macrumors newbie
Original poster
Oct 11, 2017
5
1
Hi All,
I am curious about the performance/throttling of the 2018 i7 and i9 laptops when used for tasks that are not video editing. I'm now using a 2015 MBP and I would like to upgrade for the extra memory and cores. My work flow consists of extensive Mathematica use (solving large systems of ODE/PDEs), finite element analysis, fluid dynamics sims, running sims inside docker containers, matlab, etc.

These are CPU intensive tasks, NOT CPU+GPU intensive at the same time. Has anyone tested out these sorts of tasks on the 2018's and seen the severity of throttling as seen for the video compression and such?

I have seen the review by the NASA engineer with the i9 (http://hrtapps.com/blogs/20180712/) , unfortunately there is no comparison to the 2018 i7s, so its hard to tell if the i9 is worth it. Also, the scaling he sees with the i9 is much lower when more than 4 cores are used, which to me suggest the throttling is kicking in.

Thanks in advance.
 
  • Like
Reactions: The Mercurian
You can follow the relevant thread, but from my experience with extensive CPU usage the i7 can match if not even surpass the i9 if you limit the wattage.

The i9 is better for tasks that max out all the cores for not more than few seconds. To get an estimate of your usage, its better if you look at the watts that the CPU uses with your typical analysis. If you are maxing it out ~45W then you will definitely will see throttling with the i9, in which case the i7 will perform just as well.

Keep in mind that could of course change in the feature if Apple provides a firmware with some thermal optimization.
 
You can follow the relevant thread, but from my experience with extensive CPU usage the i7 can match if not even surpass the i9 if you limit the wattage.

The i9 is better for tasks that max out all the cores for not more than few seconds. To get an estimate of your usage, its better if you look at the watts that the CPU uses with your typical analysis. If you are maxing it out ~45W then you will definitely will see throttling with the i9, in which case the i7 will perform just as well.

Keep in mind that could of course change in the feature if Apple provides a firmware with some thermal optimization.
Ah, yes, that makes sense to me. Looking at the wattage is a good idea for thinking about this. Thanks!
 
These are CPU intensive tasks, NOT CPU+GPU intensive at the same time. Has anyone tested out these sorts of tasks on the 2018's and seen the severity of throttling as seen for the video compression and such?
Many of the test done are mainly CPU intensive and not using the GPU much, so throttling will come with the i9.
This is some great findings, and shows that the i9 can be managed better then it is out of the box, gives hope to a firmware update
 
For such sciencists and engineers I recommend Dell Workstation 5820 with 8 core or 10 core CPU Intel Xeon W-2133 which has CPU PassMark result over 21,000 pts and with 1TB PCIe NVMe drive. On the market W2133 cost around $1300 and Dell Partner program you can get it for $700 which is still expensive but $600 discount is a good deal. I made a lot of savings selecting some basic Nvidia Quadro because I did not needed GPU processing (you need it possibly). Windows is also a more suited platform with much more variety of engineering tools and kits. That is my point of view using Matlab, Ansys, Mathematica, Mathcad, Scilab etc over a time. Also native MS Office is an advantage and full support of VBA in Excel (all my scripts did not work in MS Office on Mac and custom fonts also). Doing this things on laptop is not efficient however if you are a student I can understand that you want a mobile device on university so some compromise is needed. If you need Mac because all software is bought now I suggest to wait what will be Apple reaction and test results with official patch if it will be released. If you need it now I suggest to find a right balance between number of cores and CPU frequency. For example you can buy mentioned Dell T5820 workstation with 22 cores Intel Xeon, but more cores = lower base frequency (in case of app that use single core and scale with freq you loose processing power) and assuming that CPU cooler has the same design (capability to dissipate heat) there is a high risk that when you load all cores throttling will be more perceived. So never go with the highest CPU config especially on thin mobile laptop because you miss performance sweep spot. Like I mentioned most of the time i5 offered more stable and close performance to i7 where clocks were reduced (cooling for i5 and i7 is the same).
 
Last edited:
For such sciencists and engineers I recommend Dell Workstation 5820 with 8 core or 10 core CPU Intel Xeon W-2133 which has CPU PassMark result over 21,000 pts and with 1TB PCIe NVMe drive. On the market W2133 cost around $1300 and Dell Partner program you can get it for $700 which is still expensive but $600 discount is a good deal. I made a lot of savings selecting some basic Nvidia Quadro because I did not needed GPU processing (you need it possibly). Windows is also a more suited platform with much more variety of engineering tools and kits. That is my point of view using Matlab, Ansys, Mathematica, Mathcad, Scilab etc over a time. Also native MS Office is an advantage and full support of VBA in Excel (all my scripts did not work in MS Office on Mac and custom fonts also). Doing this things on laptop is not efficient however if you are a student I can understand that you want a mobile device on university so some compromise is needed. If you need Mac because all software is bought now I suggest to wait what will be Apple reaction and test results with official patch if it will be released. If you need it now I suggest to find a right balance between number of cores and CPU frequency. For example you can buy mentioned Dell T5820 workstation with 22 cores Intel Xeon, but more cores = lower base frequency (in case of app that use single core and scale with freq you loose processing power) and assuming that CPU cooler has the same design (capability to dissipate heat) there is a high risk that when you load all cores throttling will be more perceived. So never go with the highest CPU config especially on thin mobile laptop because you miss performance sweep spot. Like I mentioned most of the time i5 offered more stable and close performance to i7 where clocks were reduced (cooling for i5 and i7 is the same).

Tower systems not exactly carry-on luggage friendly. Some folks need power and portability
 
If you need a power and portability then you buy HP EliteBook Workstation or Dell Mobile Workstation. Please do not tell me they you make fluid dynamics simulations of RR Trent engine during the flight or during the lectures ;) Also how do you realize to present of numerical computations for a client if your system does not have RAM with ECC? Do you take into account possible errors? Let's be professional. 64GB RAM is a minimum for such tasks. Macs are even not a perfect device to fully handle Netflix or Youtube because 4K and HDR is not supported :) If you want to wait zillion hours for simulation result then go with Mac or some other consumer laptop designed for entertainment. I can present results on Mac for my customer but doing daily simulation is like reverting current of river using stick.
 
I use an Asus GL703GS for engineering purpose as it's extremely fast and scalable. At less than 3Kg and a 17.3 I find it adequately portable, travelling globally with the notebook.

I have no issue with notebook's getting thinner & lighter, however I do have issue with compromised usability & performance with the GL703GS being significantly faster than any current Apple notebook, even desktops most importantly performance is optimum with no throttling.

Q-6
 
Asus make a great gamers laptops and sure that performance difference will be noticeable better than on MBP because they use a plastic enclosure and very efficient cooling system is much larger. For a home or small company definitely better however still not comparable with Dell, HP mobile workstations when you have a good durability due to magnesium chassis and 3-5yrs additional support. Even AppleCare cannot be comparable since Apple use a repaired/used parts like optical drives and Apple cannot sent a technician to replace a part at customer. When time is money it is better to get typical mobile workstation designed for engineering tasks and with good next day support in case of failure or even accidental damage.
 
Thanks for all the replies!
I agree that MBP is not meant to be a primary machine for the tasks I mentioned initially, and indeed I run the real heavy stuff on clusters. But I use my MBP for development and smaller simulations and I like macOS, so I'm not so interested in moving to a windows machine.

The reason I started this thread was mostly because of my past experience with running stress tests on machines I have built, and I have found that not all stress tests are created equal. Even if the stress tests are fully CPU intensive, the resulting temps depend on the type of stress. Stress tests doing a lot of floating point calculations seemed to generate much more heat than stress tests that were doing millions of file calls.

So, its unclear to me, for example, how similar the cinebench testing, which is so popular, is to numerically integrating a large systems of equations, or doing repeated matrix calculations, or doing file calls.

But I also agree with the above, just have to wait until more people besides streamers get their hands on these things.

Thanks!
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.