The SPEs are capable of handling geometries just fine but this serves nothing
if WipEout HD is already IO bounded, which seems to be the case after reading
the article.
All the fat PS3s are using an Silicon Image VastLane SiI9132 transmitter
(SiI9132CBU) which isn't able to decode DTS-HD MA. Initially, Sony wanted
to go with a dual HDMI ported SiI9133 transmitter, but decided against it.
The interesting fact here is that the SiI9134, which was available at the
same time, supports DTS-HD MA unlike the SiI9132. [Ref]
I've found no official specification of the SiI9132 transmitter. This chip was
specifically built for game consoles and as such might be under NDA by Sony.
However, Silicon Image says that the SiI9132 is equal to the SiI9134 in terms
of video processing, that is to say; both have the same maximal resolution
of 1080p, the same maximal color depth of 36 bits/pixel, and the same
maximum bandwidth of 6.75 Gbps. [Ref]
From Silicon Image website we can read off the specification of the SiI9134
transmitter, which is just an DTS-HD MA enhanced version of the Sil9132.
Let us consider the bandwidth again.
For a 24bit RGB signal, the required HDMI bandwidth computes as follows;
Bandwidth in Gbps = (resH+blankH)*(resV+blankV)*24bit*Hz*10/8.
The HDMI 1.4 specification provides specific timing requirements for the
horizontal blanking interval and vertical blanking interval which amounts to
280 pixel for the horizontal blanking interval and 50 pixels for the vertical
blanking interval for a resolution of 1920x1080@60Hz. The factor 10/8
amounts for the conversion of an 8bit color value to 10bit.
We have,
(1920[resH]+280[blankH])*(1080[resV]+50[blankV])*24[bit]*60[Hz]*10/8
~= 4.47 Gbps.
Hence, a 1080p@120fps signal would require a bandwidth of ~ 8.94 Gbps.
And since the bandwidth of the PS3's HDMI transmitter is limited to 6.75 Gbps,
it is rather unlikely that we will see 1080p@60fps in 3-D coming out of a PS3.