Yes, it is discussed a bit here: http://www.fractalforums.com/new-theories-and-research/categorising-fractals/msg22585/#msg22585
It's an example of a class of fractal of which there are so few examples of that there was no obvious name, so I called them shells. I'd like to build one out of card, but it would be an arduous job!

And an alternative: http://www.fractalforums.com/ifs-iterated-function-systems/fun-with-koch-fractals/msg2630/#msg2630

Desperate for a Koch???

Thanks for all the replies, I already knew that there is an huge number of possibilities, and I already have seen those you posted and others but I liked the most the Squarry Koch and the pyramid.

David can you share the exact 3D transform you used (edventually with a drawing)? It may come very handy (but if it is made of more than 10 matrix don't care )

I have ported this algorithm on MB, the result is cute but the noise is annoying.

  if abs(x)>1 || abs(y)>1
  x = abs(x)
  y = abs(y)
  z = abs(z)
      x=abs(x);y=abs(y);z=abs(z);
      if(x-y<0)
      x1=y;y=x;x=x1;
      endif
      if(x-z<0)
      x1=z;z=x;x=x1;
      endif
      if(y-z<0)
      y1=z;z=y;y=y1;
      endif
  if (x-xlim)>y
   x = xsub1 - x
   y = yzsub1 -y
   z = yzsub1 -z
   x,y,z *= scale1
   else
   x = (xsub2 - x)*scale2
   y = (yzsub2 -y)*scale2
   z = (yzsub2 -z)*scale2
   endif
  endif

Scale1=Scale2 does not plot, so the only way is to set em differently

The cubes are not perfectly displaced, the formula is still approximated

Params used

.Double Scale 1 = 4
.Double Scale 2 = 3
.Double x limit = .75
.Double dx1 = 1.25
.Double dyz1 = 0
.Double dx2 = 2
.Double dyz2 = 1.5

I am attaching the formula too, that gives nice results but is still noisy

The wrong part is when I do y z second folding unconditionally, plus wrong values, what looks good in 2d may be horrible in 3d. More time needed, sir

by chance i have found this amazing video showing a sqkoch
<a href="http://www.youtube.com/v/SbSo7onX9qI&rel=1&fs=1&hd=1" target="_blank">http://www.youtube.com/v/SbSo7onX9qI&rel=1&fs=1&hd=1</a>

Okay, the test parameter. I enabled very high quality (slow render), too bad the structure is hidden by some annoying artifacts in normal quality.

Mandelbulb3Dv16{
O.....S....O/...w....2A..............1.......s1E................................
................................OaNaNaNadz1........0./..................y.2.....
................/MU0A....6kHN...i/....E2.....QzIXLua2PrD...........c./...w1.D8Qx
w.EnAnQD1E../2.swsc1au1xz0r2VJpErBuD..IpM1t49x9..................y1...sD...../..
.wHnAngD2Rw3Oc.cazXAWreAEZc2zSWyooeAVOyjTcFzUvMS3t97C1JzzzzvzUNehUisqclj.Zh4j8Hc
aznAWreAEZc2zconlUV0UOyD......o90.....................sD..kz0...................
..............................MRS4.4qtN.kObb/UoRS4.srtN.sVbb/.FSS4..............
.....................wzzz1.U..6.P....M4...EB....W....61....F....8/...I1....UJl22
...U.iVFwxDE./ozPM2Tzz7lz16.mc..zXCc.El18XGQeGyDjvIRhrVAkz1............28.kFrA0.
FWb96aAIVzvh1se7Umvxz0........../6U0.wzzz1................................E.0c..
zzzz.................................2U.8.kzzzD.................................
/6U0.wzzz1...................................2CcN/UvNP6.eeWCNq0.yRii.EJJUk1f..XR
S1.mx3CcN/UvNP6.QsLsUa3.ibhV..bTV1OK.sSq40.ly3CcN/UvNP6.MwLsUa3.ibhV.kqTV1OK.sSq
40.kz3CcN/UvNP6...EsUa3.eeWCNq0.IJ36wk8.wyLsUa3.................................
E.E..2....E.....I....6....kOjB4OTBKRWJ4.7NoI............................4MU/4MU/
4M.................0./........zD...................0./..........................
................................................................................
........................}

Ok! what about these two ones?
It's a mix between KIFS and msltoes' approache. The DE is not perfect: there are still some discontinuities but it's good enougth. Unfortunately those discontinuities are more severe if one changes the scale, offset or apply rotation. That may mean that hybrids would give some sand effect.