Message Boards Message Boards

GROUPS:

Mathematica speedup from 2019 iMac over late 2015 iMac?

Posted 1 month ago
510 Views
|
16 Replies
|
1 Total Likes
|

My current computer, a late 2015 iMac 27” has a 4 GHz Core i7 processor, 32 GB RAM, 3 TB Fusion drive, and Radeon R9 M395X with 4 GB VRAM, Retina display.

What kind of speedup with Mathematica, if any, might I expect from a new, 2019, iMac 27” with the following configuration?

  • 3.6 GHz Core i9 processor (8-core)
  • 64GB RAM
  • 3 TB Fusion drive
  • Radeon Pro Vega 48 with 8GB HBM2 memory

Although this new CPU has Turbo Boost up to 5 GHz, I’m concerned that the “default” of only 3.6 GHZ, being below my current 4 GHz, might impede perfornance.

Presumably the graphics rendering will be faster, right?

Or do the additional cores make up for that difference? In fact, can Mathematica take advantage of those additional cores (without my explicitly coding for parallel kernels)?

16 Replies

Murray,

The best way to compare machines is here- geekbenchmarks

The multiple cores will not help much unless you use the parallel code. In default, Mathematica uses two cores.

Regards

Neil

No benchmarks at geekbenchmarks yet for the 2019 iMacs, of course, since they have not quite yet been released to market.

Murray,

If you go to apple's website they give the speed up for the new iMac's in Mathematica. They claim a 40% speedup for the 6 core 4K machine and a 50% speedup for the 8 core 5k machine with the following footnote:

Testing conducted by Apple in February 2019 using preproduction 3.2GHz 6-core Intel Core i7-based 21.5-inch iMac systems with 32GB of RAM, and preproduction 3.6GHz 8-core Intel Core i9-based 27-inch iMac systems with 64GB of RAM; and shipping 3.6GHz quad-core Intel Core i7-based 21.5-inch iMac systems with 32GB of RAM, as well as shipping 4.2GHz quad-core Intel Core i7-based 27-inch iMac systems with 64GB of RAM. Tested using Mathematica v11.3.0 with built-in benchmark, WolframMark. Performance tests are conducted using specific computer systems and reflect the approximate performance of iMac.

You would have to look at the details of the WolframMark to decide how relevant it is to you. You can then get the comparisons of your existing machine to the quad core baseline on GeekBench to see how your speed up will compare. Alternatively you can run the WolframMark on your existing machine and make that comparison.

Update: Actually, I just realized that your current machine is just slightly slower than the baseline machine Apple used to compare so they are claiming a 50% faster performance for the two machines as you describe them.

Regards,

Neil

Neil Singer: I missed those benchmark comparisons on the long Apple iMac info page. This information is very useful to me, so thank you for pointing to it!

Parallel code on a single multi-core computer in Mathematica is great and it is good on a cluster that is properly set up. See my progress with Mathematica since 2000 and parallel code since 2012 here.

If you are referring to the brand new 2019 iMac, I would not be surprised if it were slower!

I read that this machine is equipped with an old fashioned HDD of Fusion Drive. And not with a SSD. This will slow down everything badly.

Beware that the new Mathematica Version 12 will be over 10 GB on disk space. And although not all is loaded at the same time into memory, considerable portions are. And this is beyond user control.

I would assume that my going from 32 GB RAM to 64 GB would help there.

The 2019 iMac I'm considering, like the late 2015 iMac I'm using now, has a 3 TB Fusion drive, which includes 128 GB flash storage.

There is an option for SSD only, but the max size Apple will offer, pathetically, is 2 TB for that.

I find this machine as disappointing as the upgraded Mac Mini from 2018. The Mini has a poor graphics card. Personally, I am waiting for the upcoming 2019 Mac Pro desktop machine.

But undoubtedly many, many thousands of dollars more expensive than the newest iMacs, and probably thousands more than the newest iMac Pros. So that's like comparing apples with onions.

The RAM increase in size won't do much for speed, unless you ended up using virtual RAM from your HD in the old computer because of too little RAM, then RAM size will make a great difference. RAM speed is what you should look at. Your new computer will probably have a great increase in that!

Posted 1 month ago

I missed those benchmark examinations on the long Apple iMac data page. This information is profitable to me, so thank you for demonstrating it!

I'll report comparative Mathematica benchmarks once my new iMac arrives and I've finished configuring it.

Now that Mathematica 12.0 has been released and is installed on my new 2019 iMac, I can now report the results, both running macOS 10.14.4:

  • Late 2015 iMac — 4 GHz Core i7 CPU, 32 GB RAM, 3 TB Fusion drive: 3.10
  • 2019 iMac — 3.6 GHz Core i9 CPU, 64 GB RAM, 2 TB SSD: 3.75
Posted 6 days ago

Could you perhaps also report the Mathematica 11.3 benchmark for both machines, if you still have version 11.3 installed?

No, sorry: 11.3 is gone from my new machine and I don't have an 11.3 benchmark from the old one (which was traded in).

Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract