Originally Posted by Marvin
There's one here on the left, I'm sure it sells quite well:
Like the above test showed, Iris is faster than the 650M for OpenCL, sometimes by double. 3D work doesn't require the framerates of games so no problem there, plus it has shared memory so 1GB minimum for everyone (yes it's slower but there's a special Iris model with a fast GDDR5 memory cache). It has OpenGL 4 and OpenCL 1.2 support.
They require framerates. 3d app performance is entirely different. It's first a matter of what you're doing. If it's a specialized role, that role is what matters. Texturing requires a lot of memory if you want to display texture maps in 3d. Modeling can really lag if the poly count is too big for the card. Animation is just about being able to evaluate at 24 fps. The viewports themselves do not behave like game engines, and some cards can really suck. I won't say that about this one as I haven't seen it, but some of the developers have started to certify integrated graphics just because of how many machines ship with them. I'm not sure whether the users are more students or smaller businesses. There are quite a few areas where geometry doesn't get too crazy, archviz being a prime example.
If it can lower the price considerably, improve fan noise, improve battery life and sacrifice negligible performance, I'd say it's worth doing.
The CPU in the listing is a very expensive mobile GPU and the TDP is higher than the CPU the 13" uses so it would considerably lower battery life.
I am not prone to trusting intel's claims, but your suggestion that they codename it "eclipse" still makes me laugh.
For raw performance, the 2013 NVidia and AMD options would be better but not for power consumption and the performance boost would be unnoticeable for actual professional work and not gaming.
I kind of wonder what the usage breakdown is like on the notebooks.