Will a gtx1080ti make much perf. difference over a 1070?

[invictius]

Say in frozen time at 4k? (Or similarly demanding maps)

[wildweasel]

As I understand it, the bottleneck there is not with the GPU, but I don't know much more than that to be able to explain why.

[Graf Zahl]

In terms of GPU load, a map like Frozen Time is actually peanuts to process - if all the data was on the GPU and could be processed in one quick go.
But the entire scene processing runs on the CPU and this is what takes most of the time. Due to the high detail of the main area, traversing the BSP, and preparing every single linedef, several 1000 of them, is what makes this slow.

[Medicris]

When it comes to GZDoom, it's not actually very demanding of the GPU past a certain point. Even bottom-end cards will do fine, even my (very) old HD 5670 did the job no sweat at 1080p and I know my old RX 460 as well ran supersampled 4k just fine. Your card's several times faster than those, you should already be way past the point of necessity.

To get higher frames out of GZDoom, I'd suggest a CPU upgrade instead. What are you running now? Like most older games, GZDoom likes higher clock speeds more than cores, you won't get any speed by going with a lower-clocked octacore vs. a higher clocked quad core.

I know that there is a fair bit of overhead in GZDoom to render the maps and environments, that will be helped by higher clock speeds. The CPU still does the vast majority of the rendering work, it just passes the info to the GPU to draw it onscreen. Additionally, to make them moddable, there a lot more scripts running at once for each entity, like monsters. In Frozen In Time's case, there aren't many vision-blocking lines so the entire map is often being drawn at once. We're talking thousands of sectors being calculated and drawn by the CPU every frame, here.

Before buying anything though, try this. Try running the map in prboom+. From my understanding, prb+ is much lighter on the CPU side, as it doesn't deal with any kind of modding capabilities and so each monster is running on far simpler code, and its GL renderer there is also more simple. If it runs fine in prb+, it's probably just GZDoom's relative rendering complexity and scripts demanding more of the CPU.

[Arco]

http://gpu.userbenchmark.com/Compare/Nv ... 3918vs3609

Whether or not you would want to spend an extra ~$300 on a graphics card with better frame rates as opposed to practicality is your preference.

[invictius]

http://gpu.userbenchmark.com/Compare/Nv ... 3918vs3609

Whether or not you would want to spend an extra ~$300 on a graphics card with better frame rates as opposed to practicality is your preference.

I already have a 1070, was just curious.

How about when you start adding things like bloom and tonemaps?

[Graf Zahl]

If you want to talk performance, the features that really need some power are SSAO and shadowmaps. Those may benefit from a 1080.

[invictius]

If you want to talk performance, the features that really need some power are SSAO and shadowmaps. Those may benefit from a 1080.

Are shadowmaps under opengl settings? I've not noticed them. SSAO either for that matter.

[Topazy]

Hey,1080ti way better then 1070.
here easy compare: http://gpu.userbenchmark.com/Compare/Nv ... 3918vs3609
Do not look at the price.Its show wrong.The benchmark shows 50% perf.
regards Topazy.

[Graf Zahl]

You really should read what this is about before posting. Hint: CPU-limited renderer cannot work fast enough to fully use the card's power.

[Hypersonic]

Seems best to first test with a very low resolution to see if you're CPU/Ram limited, factoring out the GPU performance. Then ramp up the resolution to see if the framerate starts to lower significantly to determine if the GPU is a limiting factor.

[Rachael]

My 860M, which is a really weak GPU compared to both listed, barely even sweats with "vid_scalefactor 2" at 1680x1050. GPU is not a limiting factor and like Graf said, a low clock CPU is your biggest enemy here.

As far as GZDoom is concerned, there's literally absolutely no difference in performance between an 1080 and 1070.

Sorry, but you're benchmarking the wrong game for the difference between the two GPU's.

[Hypersonic]

Does GZDoom have something like Quake 1's r_speeds variable? There were some other performance variables as well listed under '1.8 Technical Graphics Commands' in http://www.gamers.org/dEngine/quake/QDP/qconcmds.html

[Graf Zahl]

'stat rendertimes' and 'stat renderstats'.


Regarding performance, compared to a Geforce 1060, my old 550 ti performed at the same level if the postprocessing effects were disabled and shadowmaps were off.
However, with postprocessing on and shadowmaps the 1060 runs circles around the old card, here the more than quadrupled performance really shows.

On a map like Frozen Time, which is about pushing polygons to the GPU there's no difference to be seen.

[Hypersonic]

Thanks for the info on the stat parameters, just noticed that if you type just stat it'll show you all of the possible parameters as well which is very useful.

I was comparing vid_maxfps 60 and vid_maxfps 200 on a 60hz display, and noticed that with vid_maxfps 60 there were subtle but noticeable stutters, while vid_maxfps 200 was totally smooth. A similar phenomenon occurs with Quakespasm https://sourceforge.net/p/quakespasm/di ... /a180cafc/