Welcome to Fractal Forums

Hello !
I am new here ! My question is about the hardware (win PC). What do you recommend ? CPU and so on ... What are the minimum power I need to calculate fractal and enjoy it !
Thank you.  :dink:

Depends on what you can afford.

Anything that doesn't use Microsoft software ;)

am attempt of a serious answer: it depends on what you want to do...
the deep zoomers here want high integer power, the cpu 3d renderers (and most other fractal renderers) need high float performance and open cl or cuda engines need a fast graphics device.

A 16-core Bulldozer and two Tesla cards should do...  :dink:

A no more helpful, but more serious answer:

Unfortunately, the various fractal generator programs out there have very different needs. This means that for some programs, AMD processors will offer the most value. Many other programs run much better on Intel processors. And finally there is a new crop of software that uses the GPU to do computation; these will like beefy graphics cards.

If you are undecided, the safest bet is probably a midrange quad core from Intel (no Hyperthreading required) with built-in "better than nothing" graphics. If it turns out you like using programs that run on the GPU, you can later add a graphics card in entry level "enthusiast" range (ideally a discounted model from last year's generation).

Do not buy the absolute high end. The price premium is generally not worth the performance increase. The graphics card doesn't really need to be compatible with the latest buzzwords, as long as it is a relatively recent model. The important capabilities (computational uses, double precision floating point throughput) of GPUs are usually not advertized, so you best ask again more specifically before you spend a whole lot of money. Both AMD and NVidia offer very similar capabilities in that regard; it depends on the specific models.

An AMD processor can be a good option if you already know that you want to do extremely deep zooms with multi precision arithmetic. For anything else, an Intel processor at a similar price point is probably more suitable.


In the unlikely case that you are a graphics designer who can live off of fractal animations, you can go crazy and buy a workstation with two or four multicore processors and two or more high end GPUs. You will probably overload even that machine eventually. :-) But mere hobbyists should not have any need to waste that kind of money.

More seriously than my previous reply - as others have said, unless you have an unlimited budget then it very much depends on what you want to do with fractals/fractal art e.g. render stills, render animations, or do realtime animations, and which software you are going to use (especially if creating your own in which case the OS and source code type are even now as important as the hardware).

If on a minimum budget then in the UK I'd go for a minimum quad-core system (with or without hyper-threading but better with) as far as CPU is concerned and at least the video card I have here (or similar) i.e. an AMD/ATI Radeon 5870 - but I'd buy a second-hand system - such systems will definitely be around for under £300 if you look *but* do not buy sight-unseen or untested as most will be ex-gamer's systems of the sort that get hammered for 12 months and then replaced due to being out-of-date ;)

Having said that, even on a budget I would never buy a system now in which the hardware wasn't 64-bit and was at least capable of taking a minimum of 8 GB of RAM - though of course the amount of RAM or even hard-disk space actually installed on a second-hand system is pretty irrelevant given the cost of RAM and internal hard-drives nowadays.

Also having said all the above - to answer the original question, the minimum required to "actually enjoy" working with fractals and fractal art would be equivalent to a fast Commodore Amiga 68040 - e.g. say an old PC with a 486DX or Pentium 1 running at 200MHz or better ;)

i.e. any system that can render the original Mandelbrot say at 100 max. iterations and bailout of say 32 at a resolution of 640*480 for (-2.5,-1.5) to (1.5,1.5) in under a minute using either fixed-point on CPU or on FPU using single-precision floats.