Hi everyone,
I would like to understand how the choice of the GPU/graphics card is influencing the performance of XPression software. I have built a quite powerful machine in terms of CPU, RAM, diskspeed, pcie bandwidth. It made a great leap regarding the former 10 year old HP Z800 machine. But when I put my old GTX 970 into that machine, I can hardly sense an improvement on how XPression itself is performing (ok, its still 32 bit software so RAM is not used beyond 4GB) away from the overall better Windows performance. I do assume that the Framebuffer is not cruicial, is it? I would expect that the FB is delivered always ONE HD stream for one output as a result of the renderings before - regardless how many streams are combined in one image before output, right?
So can anyone tell me what the main bottleneck is for example in the number of simultaniously played full HD streams? The old machine would begin to drop frames here and there when I would play say 3 full HD streams in parallel. I was expecting a proportional scaling when I have 4 times the pcie bandwidth an more than 20 times the diskspeed. So I assume that the bottleneck seems to be the graphics card. Am I right?
But what properties of the graphics card are most crucial for XPression performance? Is it
- the gpu core count
- the gpu core clock
- the dedicated graphics memory size
- the memory bandwidth
I could just buy a RTX 3090 for about 3000 € and see what happens ... but still I would not know if the amount of RAM, the core count, the core clock, the memory bandwidth would be the reason for a performance boost. I would never know if a GTX 3060 (around 500 €) for example would have been sufficiant or not (and vice versa, if the performance would not be boosted with a 3060).
So: what can I expect to take the highest impact on video playpack performance?
I would like to avoid spending 3000€ in vain, when it would not outperform the 500 € GPU. I mean it is not about gaming with dozens of powerful rendering tricks like GPU raytracing etc. because XPression does not take advantage of those modern high end features, does it? In my understanding XPression playpack performance is about sheer throughput bandwidth, right?
For example if you compare the GTX 3060 with 12GB but only 192 Bit memory bandwidth and less cores against a GTX 3060 ti with 8GB but with a 256 bit memory bandwith and more cores... what would improve the playback prrformance?
Apart from the fact, that we do not utilize many 3D Objects in our designs... As far as I know XPression does not provide a sophisticated usage of lightning, does not use reflection, refraction, fog, raytracing, scene/depth based blur or other things It is straight foreward 3D with some basic lightning - period. Right?
I would appreciate any "light" in this for me "shady topic" of GPU capabilities and its dependancy on XPression performance.
If all my above thoughts are crap, please let me know. I also don't quite understand the many "memory allocation" settings on the XPression prefs (I did not touch them so far).
And if someone could share his/her expierience with a 64 bit edition of XPression (and what difference it makes) this would be very interesting too.
Thanks in advance
------------------------------
Sefa Aydemir
On Air Visuals / Switchers and Live Graphics Specialist
MMC Studios Cologne
Cologne Germany
------------------------------