Lazy8s said:
Intel's Menlow chipset is being fabricated in their 45nm process, so it's purported to be affordable and battery efficient despite performing nearly as powerfully as modern laptops.
The SGX530 core was the first one PowerVR had available for the portable, handheld, market, so that's the one that the OMAP3430 is probably using considering the 3430 is among the first portable SoCs using SGX.
It would be nice if the OMAP3430 is using the SGX530. It has an effective fill rate of 1200M pixels/s and 13.5M polygons/s. It's not the highest mobile SGX but has more than enough power for our uses IMO.
i.MX31 comparison:
135M pixels/s and 1.7M polygons/s
I wouldn't expect that kind of performance from any embedded processor. There are two things that stand in the way of achieving those "magic" numbers:
1. Memory interface - must have dedicated RAM on a dedicated channel which no SOC has, it is always shared with main memory. For example, yes the MBX lite is rated up to 1.7M polys but the i.MX31 is rated at only 0.8M polys realistic throughputs with effects applied. If the OMAP SGX core can do 13.5M polys, maybe we can see 3M? (still very impressive and beats Nvidia or ATI's embedded graphics chips) I think some of those MIPS chips with GPUs use 64-bit interfaces to double date throughput, but it's MIPS, so nevermind.
2. Clock speed of GPU core - I believe that clock can be up to 200MHz but isn't likely to run at that speed once the core is inserted into a chip design and has to scale back according to the architecture and transistor layout it needs to mesh with.
Having said that, the OMAP3430 should kick the crap out of the I.MX31 in many applications.
h34r: