Microstation - CPU, GPU or DirectX?

[fostertom]

Now that I have made a small 3D model, of 837K, in Microstation (attached),navigation using SpacePilot (not Pro) is sometimes painfully slow - frame by frame. It improved with latest Nvidia driver 307.32 for my Quadro FX580 but still must do better. The CPU is i3-2120 @3.30GHz.

With SpacePilot connected, but not using it, using MS's own Rotate view>Dynamic instead, rotation is good and fast - but I don't want to live, without my SpacePilot!

The difference may be, that with SpacePilot, 'in progress' things get recomputed frame by frame,
whereas with MS's own Rotate view>Dynamic, the 'in progress' thing disappears until the rotation is completed and then is recomputed just once for the new orientation.

So, if you have say placed one end of a smartline, then rotating with SpacePilot to see the the target for the other end, the line remains visible and can be converged on the target by mouse - very handy, if it would work fast. And shadows remain visible and change as you rotate.

Whereas with MS's own Rotate view>Dynamic, both the half-placed Smartline, and the shadows, disappear until the rotation is completed and then are recomputed just once for the new orientation. I do prefer the info-rich SpacePilot behaviour.

The closer I'm zoomed in it gets worse exponentially, by either method. MS's own Rotate view>Dynamic is fast enough that this is not a problem - but can only get worse with bigger models?

My question is, using SpacePilot which is apparently compute-intensive beyond the capabilities of my machine, what is it that Spacepilot is using, that is deficient - is it CPU computation speed, or is it all happening on the graphics GPU, or maybe needs more graphics memory, or do I have to wait for improved software algorithms (which I guess uprating the Nvidia driver from DirectX 9 to 11-capable amounts to - even though my FX580 itself only does DirectX 10)?

Or is it just unresolved incompatibility between SpacePilot and Bentley's implimentation of same in MicroStation?

[jwick]

It most likely isn't the 3D mouse. In fact, the 3D mouse requires less computation than the 2D mouse. MS has optimized the Rotate view-Dynamic with the 2D mouse by disabling some computations during movement, as you noticed. They should extend these optimizations for the 3D mouse-handling code.

The implementation was done by Bentley. Please contact them to ask them to extend this optimization to the 3D mouse. They are usually very responsive. Many MS users would be happy to see this improvement.

[fostertom]

Thanks for answering.

I v much like the way that these computations aren't disabled during 3D mouse rotation - it's great to be able to converge (using right hand 2D mouse) the second point of a half-placed element with its target, as it rotates (using left hand SpacePilot) into view - and it's nice to see shadows changing as the model rotates too.

Are you saying that having these computations remain active creates an impossible load on the system, so the rotation goes slow and jerky, frame by frame? That the only remedy is to disable such computations? That users would be glad if Bentley did that in their 3D mouse implimentation? I would hope it would be an option, switchable on the fly.

Does such disablement happen in other quality 3D programs' implimentations?

If there is no such disablement, which part of the system is being over loaded? Would faster CPU clock help. or GPU ditto, or more graphics memory?

[jwick]

You really have to take up computation performance issues with Bentley. Certainly they are aware of places in which the application should be made faster. They may also have specific suggestions for your configuration. They may even have tools for profiling and tuning your setup.

If they give you some useful tips, please post them here so other MS users can benefit.