If you can run games at native 120 FPS, sadly LSFG isn't worth it.

[ChaosBuster]

Image quality and stability of LSFG has vastly improved with updates, but the biggest issue with LSFG I've noticed is that it consumes almost double the GPU power when running 60 to 120 FPS Frame Gen than native 120 FPS. If I can run games at native 120 FPS it consumes much less power. Even if I can't get 120 FPS in a game, it'll still use far less power than LSFG 60 to 120 FPS.

Example: (1080p, V-Sync On, G-Sync Off)
Ni no Kuni Wrath of the White Witch Remastered
Native 60 FPS: 24 W
Native 120 FPS: 32~36 W
LSFG 60 to 120 FPS: 48~52 W

I've noticed similar trends with other games as well so my results should be accurate.

Sadly, this makes LSFG far less useful than I was hoping for. If native 120 FPS uses less power, why would one use LSFG? And why does LSFG consume so much power? Shouldn't it be the other way around?

I can think of LSFG only being useful if the game has locked 60 FPS mode.

[RelaxForGoodAim]

Diposting pertama kali oleh ChaosBuster:

Example: (1080p, V-Sync On, G-Sync Off)
Ni no Kuni Wrath of the White Witch Remastered
Native 60 FPS: 24 W
Native 120 FPS: 32~36 W
LSFG 60 to 120 FPS: 48~52 W

I think that the test is not conclusive. Is it GPU or CPU power consumption? Can you measure whole system power consumption? Then CPU only and GPU separate?
It would be interesting to see.

[ChaosBuster]

Diposting pertama kali oleh RelaxForGoodAim:

Diposting pertama kali oleh ChaosBuster:

Example: (1080p, V-Sync On, G-Sync Off)
Ni no Kuni Wrath of the White Witch Remastered
Native 60 FPS: 24 W
Native 120 FPS: 32~36 W
LSFG 60 to 120 FPS: 48~52 W

I think that the test is not conclusive. Is it GPU or CPU power consumption? Can you measure whole system power consumption? Then CPU only and GPU separate?
It would be interesting to see.

It's the GPU power only. CPU only uses 7~9 W. So just add this to above

I've noticed similar trends with several different games (Monster Hunter World, Yakuza, Titan Quest & more, where GPU is using almost double power with LSFG than native) so I do think my findings are conclusive. 

[RelaxForGoodAim]

Since around 40W this days to run a game is not much at all, and on average games are using 200W upwards more often then not i decided to test it on my RTX 3080.

Genshin Impact, rough average:

No LSFG, 4k

60 FPS - 235 W
45 FPS - 140 W
30 FPS - 87 W

With LSFG, 4k (no upscaling, only FG)

120 FPS - 300 W
90 FPS - 265 W
60 FPS - 150 W

It is just one game that originally is locked to 60 FPS and have options for 45 FPS and 30 FPS.
I think that more power hungry the game, the more saving on energy when using LSFG. On Low power system like yours it seams that more efficient is to run native, but in my tests 60 FPS native 235 W vs 60 FPS using LSFG 150 W it is apparent that LSFG is more efficient option. I might do more tests with games that can run natively in 120 FPS.

[Xavvy]

Diposting pertama kali oleh RelaxForGoodAim:

Since around 40W this days to run a game is not much at all, and on average games are using 200W upwards more often then not i decided to test it on my RTX 3080.

Genshin Impact, rough average:

No LSFG, 4k

60 FPS - 235 W
45 FPS - 140 W
30 FPS - 87 W

With LSFG, 4k (no upscaling, only FG)

120 FPS - 300 W
90 FPS - 265 W
60 FPS - 150 W

It is just one game that originally is locked to 60 FPS and have options for 45 FPS and 30 FPS.
I think that more power hungry the game, the more saving on energy when using LSFG. On Low power system like yours it seams that more efficient is to run native, but in my tests 60 FPS native 235 W vs 60 FPS using LSFG 150 W it is apparent that LSFG is more efficient option. I might do more tests with games that can run natively in 120 FPS.

Thank you for posting this. I had done similar test after he posted saying it uses insane power and I didn't find on my 3080ti+ rigs that it was inefficient either. Another thing that people seem to forgot is.. Can your CPU+GPU even hit 120fps at a desired setting without LSFG 1.1 or higher and the answer is usually no.

At 4K Gaming even the most powerful hardware(I have 4090/4080 rigs) struggle to run 120fps in the most demanding games.. This software enables this in not only games this demanding but it also enables games that are incredibly heavily modded to run at a fluid 60-120fps which wouldn't be possible otherwise.

So the calculation can't just be about power consumption. If you want target 'A' FPS then the cost has to come from 'B' somewhere. That's just a fact. I built originally a budget rig for just gaming for myself(ended up gifting it to the wife) but anyways it's a 7950x3d with a 7900xtx. Like that hardware runs hot.. much hotter than my nvidia rigs with literally identical cooling and I find when not running at peak there power usage is higher than intel/nvidia based rigs at idle.

I tested AFMF and LSFG 2.0 on it and even with the best gaming CPU on the market there are games that are impossible to run at 4k max settings @120fps. Like it will get close but constantly dips because of CPU intensive games. This is where LSFG 2.0 comes in. It costs a bit more power yes but it allows you to hit FPS targets that just wouldn't be possible with even the strongest hardware currently.

And TO BE VERY CLEAR AFMF < LSFG 2.0.. Hell LSFG 1.1 is better than dog crap AFMF(I've never been more disappointed with an AMD feature. It's honestly not terrible but it's basically high end Sony TV frame gen with less input latency. It looks much sloppier than LSFG 2.0 and I really wish I could get LSFG 2.0 on my consoles). It's funny consoles having this tech now is nice but yeah AMD needs to really improve AFMF..

I really wish Nvidia would get off their ass and make a counter solution because.. when NiS and FSR1 battled it out it was great because both techs improved. AFMF is just remaining stagnate because they have no real financial reason to improve it since the competition didn't bother to counter it.

[Xavvy]

In some games AFMF is superior but it seems to depend on game implementation. IF FSR2.2 is present in the game then AFMF performs well if it has nothing or FSR2.0 base then it looks like dog crap imo. LSFG 2.0 is more consistent and has straight up better frame pacing( there is some minor ghosting but latest versions seem to sort most of that out for me) and almost no hitching. AFMF has slightly better input latency apparently but.. I dunno with allow tearing on I notice the difference to be imperceptible. LSFG 2.0 also seems to have some fast sync support which in most games performs quite well and you get literally no tearing.

[Ser_Chonks]

Diposting pertama kali oleh ChaosBuster:

It's the GPU power only. CPU only uses 7~9 W. So just add this to above

I've noticed similar trends with several different games (Monster Hunter World, Yakuza, Titan Quest & more, where GPU is using almost double power with LSFG than native) so I do think my findings are conclusive.

Frame generation workloads don't scale up/down with game graphics, run those tests with high power GPUs and truly demanding games and they'll dwarf frame generation requirements.

[On Vam Ne Steem]

Your opinion is true, but i think is heavy related to what specs and comfort visual/performance settings anybody want use.For 1080p and 120 fps programm will have small impact, it's true.But there a lot of experiments with games which have hard fpslock like bad ports of game, or software for play games from consoles, like a ps 1,2,3, switch.You really should see more advantage thanks to developer, than you see now