34 reported in 1975
The most elaborate electronically-generated visual simulations
available at the present timeare one developed for NASA in
connec- tion with the lunar module (Lockwood, 1971) and the
space shuttle vehicle (SSV) program, and another currently
being developed for the Air Force. [...] It consists of 320 edges. Its 20 Hz.
scanning rate and a 600 line raster were designed to gain
spatial resolution at the expense of temporal resolution.
As yet I've been unable to chase down the reference:
Lockwood, L. W. Visual simulators for moon men. Optical Spectra,. 1971, 5, No. 8, 32-36.
It is possible, but I think unlikely, the Apollo display was more capable than the Space Shuttle one
320 edges gives around 212 polygons in a tri-mesh.
Edit to add NASA Technical Note D-7112 (1973):
Two general techniques have been used for target-vehicle
simuiation: a direct analog system of closed-circuit TV and
models, and an electronically generated (drawn) image. In both
systems, the input to the display system was through a CRT
in the infinity optics systems. The electronic image
generator (EIG) was used successfully in one of the Gemini
mission simulators, in the Gemini part-task trainer, and in
the LMPS. In the EIG system, the target vehicle was drawn
on the face of the CRT. The outline or envelope of the
target was drawn at a 60-hertz rate; however, the surface was
filled in at a
15.75-kilohertz rate. The image generation contained nine degrees of freedom and produced such phenomena as line-of-sight blanking, illumination 20 shadowing, and perspective distortion. Simple target shapes (cylinders, cones, and others)
as well as combinations of these shapes were readily
simulated with simple surface markings and details
Both the TV/model and electronic- image techniques have
produced satisfactory displays for stationkeeping and docking.
In the EIG technique, complex shapes cannot be drawn;
therefore, realism is significantly less. Conversely, the EIG
is a much simpler system to maintain and operate.
To address the notion of using early (integrated circuit) transistor technology to produce a "modern" simulation, NASA's 1971 ILLIAC IV
managed a staggering 50 MFLOPs on it's 64 cores, but it only had 16k memory available to each core. You'd need 14 cores just to hold the frame buffer
But let's be generous and say with 16 bit addressing it could access a full 64kB, and by some magic of massively parallel computing they can instantly share each others memory you'd need 512 ILLIAC IV's to make a 2GB graphics card. At $31 Million each, that is around three fifths of the entire Apollo budget.