The Firestorm Viewer version 6.4.21 had been out for a bit now. People are reacting to it. The thread Newest Firestorm version – lag monster popped up 8/4. Which made me curious because that is NOT my impression. So, I went and ran my tests, which I rate at one half-a-posterior.
This version is covered in detail over on Inara’s blog here: Firestorm 6.4.21. I’m not going to repeat what she has already written. I just want to consider what people are experiencing based on what I can see and measure.
My experience is: this version is better.
The big thing is Firestorm’s viewer now uses multiple threads for decompressing textures. Decompressing is a task that has bottle-necked performance forever. But since it has had little effect on Frames per Second numbers it doesn’t seem to have been a high priority to change. The major effect of texture decompression is in scene render time. I think scene render is much faster. But actually, measuring it well is quite complicated. I suspect most of us hit a region and consider how long it takes for the place to render. A few of us may now and then even time it. Which is practical and simple. Neither is precise and measurements are highly influenced by our location in the region, which way we are looking, our cache – when did we last visit, the region’s construction, and avatar occupancy.
One thing that can affect our perception is this install clears our viewer’s cache of downloaded content. We have to reload EVERYTHING. I think this should only affect your first-time post-install-visits to places you frequent.
A while ago the Firestorm Viewer was upgraded to use the simplified cache developed by the Lab. I think this is the first release of Firestorm that has this upgrade. Beta testers for Firestorm will have had the upgrade for some time.
From the changes the FS Team has made, I would expect this viewer to be faster at rendering a scene. I wouldn’t expect a higher FPS (Frames per Second) rate. I guessed that I might even see FPS dip while textures were decompressed as more CPU cycles would be getting used by the texture processes. That is what I experience… sometimes… There are some oddities happening when one teleports. So, detecting a slowing or a speed-up due to the new multiple threads is tricky.
There are 20 to 30+ Firestorm threads running at anyone time. My four cores are all carrying about the same load but the overall use is about 50%, which is similar to past versions.
Getting precise measurements and eliminating the effects of regions, network, CPU, GPU differences… to focus on a single thing is tedious. How about, does it seem faster in overall use? That is what I went for with my front porch measurements. Here is what I found. (Eyeball averages)
2021-08 – v6.4.21
Front porch: 10 FPS @ 512m DD
Front porch: 21 FPS @ 128m DD
Green Room @ 2100m: 125 FPS @ 128m DD
” ” ” : 160 FPS w/o Atmo-Shader
Nelsonia Safe Hub @ 128m DD w/26 avatars: 18 FPS
2021-03 – v6.4.13
Ground-level on my new porch @ 128m DD – 24.1 – 24.6 FPS
Old porch @ 128m DD – 50 -53 FPS
Green Room @ 2200m @ 128m DD – 110 – 115 FPS
Old site of Green Room @2200 @ 128m DD – 102 – 109 FPS
Nelsonia Safe Hub @ 128m DD w/15 Avatars – 31± FPS
I changed to my ‘new’ front porch earlier this year when I moved into Bellisseria. I did my tests at my old home in Fishergate and my new one in Coyote to provide a comparison on location. It obviously takes more power to render Bellisseria.
I added a change in Draw Distance. I was curious as that wasn’t mentioned in the thread. I did another on Atmospheric Shaders. Both have a significant effect on FPS.
The rest of the numbers are similar to the previous version. The decrease in Nelsonia’s w/avatars number is significantly lower. BUT… I can’t control the complexity of the avatars in the scene. So, while it is an interesting number at best it can only hint at what is going on. Without days of repeated measurements on both versions over a number of days, I consider it a pretty meaningless number.
My conclusion is this viewer runs about the same as Firestorm has for the last couple of years. I see only minor changes and I consider those within the margin of error for my measuring as these are *NOT rigorously controlled tests. I just walk out on my porch and check the FPS.