23

My club is considering buying a Raspberry Pi computer for tinkering. Can Mathematica be installed on and ran from it?

faysou
  • 10,999
  • 3
  • 50
  • 125
user13253
  • 8,666
  • 2
  • 42
  • 65

5 Answers5

44

Update: Yes, the Wolfram Language (and Mathematica) are now available free (for non-commercial use) on the Raspberry Pi.

Details:

http://www.raspberrypi.org/archives/5282

http://blog.stephenwolfram.com/2013/11/putting-the-wolfram-language-and-mathematica-on-every-raspberry-pi/

===

We are working on a project related to Raspberry Pi at Wolfram Research.

If you are interested in becoming a prerelease tester for this project, please send email to prerelease@wolfram.com. Please be sure to include your name, wolframid and any ideas you have on using Mathematica on a Raspberry Pi.

Arnoud Buzing
  • 9,801
  • 2
  • 49
  • 58
16

Mathematica is available for non-commercial use on the RPi now.

No. Mathematica is not available for Linux on the ARM architecture; only for x86/x86-64.

Even if it ran, performance wouldn't be very good, considering the Pi's weaker hardware. From the FAQ you linked to:

Overall real world performance is something like a 300 MHz Pentium 2, only with much, much swankier graphics.

while Wolfram recommends

Processor: Intel Pentium III 650 MHz or equivalent
Disk Space: 4 GB
System Memory (RAM): 512 MB required; 1 GB+ recommended

Renan
  • 731
  • 9
  • 21
13

Just for kicks, I ran the "Benchmark Mathematica" test:

  • Raspberry Pi model B
  • Over-clocked CPU @ 800 MHz
  • 64 of 256 MB RAM reserved for video
  • 16 GB Class 10 SSD (going to need swap space with only 256 MB)

Results:

Benchmark Result: 0.01

  1. Data Fitting 27.68
  2. Digits of Pi 11.30
  3. Discrete Fourier Transform 73.74
  4. Eigenvalues of a Matrix 126.18
  5. Elementary Functions 153.39
  6. Gamma Function 14.19
  7. Large Integer Multiplication 17.73
  8. Matrix Arithmetic 25.21
  9. Matrix Multiplication 1070.78
  10. Matrix Transpose 35.56
  11. Numerical Integration 33.52
  12. Polynomial Expansion 4.55
  13. Random Number Sort 24.37
  14. Singular Value Decomposition 410.94
  15. Solving a Linear System 720.21

  16. Total: 2749.33

I was impressed that the test actually finished. The desktop remained responsive (when I moved the mouse, the cursor always responded smoothly and immediately).

I didn't monitor temperatures, but all chips on the motherboard were comfortable to the touch.

If I did the test over again, I'd probably monitor the temperatures in software as well as monitor the size of the swap file.

cjpembo
  • 1,282
  • 1
  • 10
  • 16
  • 3
    Seems to me that this is actually quite a satisfactory level of performance. Apart from the numerical linear algebra tests (heavily memory bandwidth-bound and probably not using an optimized BLAS/LAPACK), the other timings are only about 10-20 times greater than for a typical desktop PC. Raspberry Pi is apparently a very reasonable proposition in terms of price/performance. Perhaps a future release of Mathematica for ARM could employ ATLAS and FFTW, which now both support ARM? The authors of OpenBLAS are also working on an ARM port. – Oleksandr R. Nov 27 '13 at 01:19
  • Can you include a link or description of the benchmark test? – bobthechemist Nov 27 '13 at 02:32
  • I benchmarked my system PC at 3Ghz and i measured also 10-20 times faster than rasp. For matrices operations it was even 1000 times faster but i suppose this has to do with the available memory. Model-B runs at 700Mhz so 10-20 times slower than a single kernel on 3000Mhz means that Pi's power is not exploited too efficiently by wolfram-engine. – tchronis Nov 27 '13 at 09:24
  • @cjpembo your link above is dead... – tchronis Nov 27 '13 at 09:25
  • @tchronis a straightforward ratio of clock speeds is not an accurate way to compare the capabilities of different processors. The Raspberry Pi's ARM processor is much simpler in design and has (very) much less memory bandwidth than desktop processors do, as it is optimized for low cost and power consumption, not outright performance. A factor of only 10-20 difference in performance is actually quite favorable toward the Raspberry Pi, IMO. – Oleksandr R. Nov 27 '13 at 13:36
  • @tchronis I formatted the link wrong... here it is again: http://reference.wolfram.com/mathematica/Benchmarking/ref/Benchmark.html – cjpembo Nov 27 '13 at 16:10
  • 1
    Wolfram also included a Remote Developemt Kit which allows: "Seamless development on the Raspberry Pi from any desktop version of Mathematica, including Home and Student Editions." http://www.wolfram.com/raspberry-pi/RemoteDevelopmentKitInstaller.cdf $50 Mathematica-powered remote "sensors" sounds like fun! – cjpembo Nov 27 '13 at 16:29
  • @cjpembo - slightly off-topic but hopefully you can help me here - would you mind trying to implement the v9 HDR code here on the Raspberry Pi? (it's the second answer from Simon which is much shorter). I can get the first step to work but the second tone-mapping step doesn't seem to want to. http://mathematica.stackexchange.com/questions/9342/programmatic-approach-to-hdr-photography-with-mathematica-image-processing-funct – Vincent Tjeng Dec 14 '13 at 13:43
11

Here's an article I wrote about Mathematica on raspberry pi. It turns out that pi version of Mathematica is v10 which allows us to get a preview of what's coming soon to the desktop (currently at 9.0.1). Also, speed comparisons with desktop version and python.

http://www.walkingrandomly.com/?p=5220

WalkingRandomly
  • 4,107
  • 1
  • 20
  • 36
  • I think that Raspberry Pi will never approach a desktop PC for numerical linear algebra performance, not due to the lack of any special instruction sets but simply because of the much simpler cache hierarchy and memory controller. High bandwidth, low latency buses have high power requirements and large, fast caches require a lot of die area. Nonetheless, I think that a factor of 100 difference is probably more realistic than 1000, i.e. there is potentially a factor of 10 to be gained over the current situation by using optimized libraries. – Oleksandr R. Nov 27 '13 at 14:04
10

Update

For those working in STEM education, there will be a presentation on running Mathematica on the Raspberry Pi on September 17th during the Virtual Conference for Education, see here for more details:

https://www.wolfram.com/events/virtual-conference/stem-education-2013/schedule.html

Arnoud Buzing
  • 9,801
  • 2
  • 49
  • 58
  • N.B.: one can watch the presentation videos on the WRI website, so although the date has passed, the answer is still relevant. – Oleksandr R. Nov 27 '13 at 14:08