10-03-2013 03:28 AM
I'm debugging some code I'm working on for utilising the 3D picture to display some data.
I've noticed that certain 3D settings have a large influence on performance as expected but I've also noticed something weird.
On my work PC (Core i7 2770, 8GB RAM, Win7 64-bit LV 2012 SP1 f5, AMD 3450 Graphics) I have a nice fluid display with several datasets rendering simultaneously. Zooming and panning is smooth as butter. On my Home-office PC (AMD X5 1090T (only 4 cores enabled), 8GB RAM, Win7 64-bit LV2012 SP1 f5, Nvidia GTX 460 Graphics) I have way more problems with my display. This is kind of unusual because the graphics card seems like it should be MUCH faster and Furmark (OpenGL Benchmark) seems to be delivering expected values for my card.
Are there known combinations of Graphics cards / drivers and/or CPUs with which the 3D picture control work better than others? I find it hard to understand how my code can run butter smooth on a weak graphics card like the 3450 but has trouble on a much beefier card like the GTX 460. Using GPUShark utility, I also see that the GPU really is pegged to 100% load even though the workload being requested of it is really small in absolute 3D terms.
Any experience anyone can share with me? I've tried installing the latest drivers. That doesn't help.
Shane.
10-03-2013 03:32 AM
Oh and another thing. There's a noticeable "lag" when trying to use the built-in camera controller. The first display change (After a mouse down) seems to take up to one second to update whereas the following changes (Panning / Zooming) are again smooth as butter. This is part of the reason I think a CPU scheduling issue might be present. I know the X6 1090T is less performant than the i7 2770 on a core-for-core basis but they're both quite fast CPUs. I don't see anything here which should be slowing my PC down this much.
I have also removed nearly all programs from the background. Especially any which may be trying to use the Graphics card. No luck.
Shane
10-04-2013 01:53 AM
I just tried the same software and the same data on both machines. Here's a valid comparison of what I'm seeing.
Computer 1: Intel Core i7 2770, Win7 64-bit 8GB RAM, Win2012SP1 f5, AMD 6350 Graphics
Computer 2: AMD 1090T, Win7 64-bit 8GB RAM, Win2012SP1 f5, NVidia GTX460 Graphics
Both computers have all recent patches and drivers installed. According to any bencahmarks I have found the GTX 460 should be a little over twice as fast as the AMD 6350 card.
For the same data set of a 1024x1024 profile (displayed as a heightmap) with texture applied (also 1024x1024) plus a small coordinate object (Three arrows in X,Y and Z directions) I see the following.
Computer 1: Update rate around 17-18 ms (VSync at 60Hz)
Computer 2: Update rate around 120 ms (Much slower)
I would have initially assumed that Computer 2 is NOT using the 3D capabilities of the graphics card. I checked OpenGL functionality by running Furmark and the results for my card match what a GTX 460 should deliver. While the LV software is running GPUChark shows that the GPU is running at 100%, so something IS running on the graphics card but it seems to be horribly inefficient.
At ths stage, while I love the functionality of the 3D picture, the not-quite-understandable differences seen between PCs has me worried. I don't want to propose changing out display routines from 2D to 3D and have the overall update speed deteriorate.
I know there was a change made to the 3D picture a while back because it didn't work well with AMD graphics. I now have the opposite problem. The same behaviour was observed in a vanilla 2012 SP1 installation (without any patches, f0 so to speak) so I don't think it's a side-effect of the AMD patch.
Shane.
10-15-2013 08:02 AM - edited 10-15-2013 08:02 AM
I bought a new computer recently (Core i7 4770) and even the integrated graphics of this processor (with only Windows default drivers) is three times faster than my "fast" graphics card.
If noone has news ont his or if this is deemed pointnless, can someone at least let me know?