Over-thinking PC game capture

[porina]

I'm debating recording some gaming for later analysis. There are occasions I think I did something correctly, but the game says otherwise. I might get into uploading edits on YouTube as low priority in future - no live streaming plans.

 

The game is FFXIV so not particularly demanding on GPU, but I play at a non-standard resolution close to 1440p 60 Hz (windowed approximately 16:9 on 3440x1440 ultra-wide). Debating what my options are here to have reasonable quality recording without too much configuration, and also not impact gaming performance.

 

Current main system is 6700k, so from a CPU perspective that might not be sufficient to record and game without impacting performance.

 

Option 1: switch gaming to a Ryzen system. I already have a 1700 and could move a more recent nvidia GPU into it, and use nvidia recording software. CPU also needs to be re-overclocked as it has been temporarily running stock while I've been playing with bios update/ram settings. I don't know yet if nvidia capture works on Windowed games. I'm a little concerned that even with overclock, if the CPU is fast enough to maintain 60fps since the game isn't heavily threaded. Past testing has shown even my 6600k (OC to 4.2) is much faster than 1700 (OC to 3.6) on this game but I didn't analyse lows, only averages.

 

Option 2: Keep playing on 6700k system, but use 2nd PC to do capture. I currently have to connect my current monitor by DP as it only runs an old version HDMI (1.x) and can't keep 60 Hz at native resolution. Are there affordable DP passthrough capture cards? If I go to HDMI I think I can still run full screen/pillarbox 1440p 60Hz.

 

Do cheap HDMI capture devices even support above 1080p resolutions? Haven't looked yet.

 

I know it depends on the settings, but what sort of disk usage are we looking at? I have used fraps in the distant past and recall even at then typical 1024x768, it liked eating disk space. However that also used a low compression ratio codec and I'm sure more modern ones can maintain adequate quality at moderate bitrates. I don't need to have it pixel perfect.

[tom_w141]

12 minutes ago, porina said:

windowed approximately 16:9 on 3440x1440 ultra-wide

How do you do this? I have recorded 21:9 but it looks stretched/squashed. How do I know my window is 16:9? Sorry to thread hijack  

[porina]

Just now, tom_w141 said:

How do you do this? I have recorded 21:9 but it looks stretched/squashed. How do I know my window is 16:9? Sorry to thread hijack  

The game has cutscenes which are rendered in 16:9 with borders if you're out. I know my window aspect ratio is near enough matching that.

 

I have been doing more searching since writing the above. It looks like all the cheap HDMI capture devices only support 1080p. The only one I found "supporting" higher resolutions would downscale the recording to 1080p anyway. I'm not gaming in 1080p! What year is this? It is 2017 right, not 2007?

 

Can't find any DP capture devices at all. Not to say they don't exist, but apparently not in any consumer context.

 

I think this is pushing me more towards software capture. I have found shadowplay supports higher resolutions, and can capture windowed but might do the whole desktop in the process. Not ideal but I could work around that. The question remaining is what is the CPU impact, and if I need more CPU? Can you actually buy Skylake-X yet?... or maybe I can upgrade my X99 Xeon setup. ThreadRipper like Ryzen will be too slow so not an option. I think the best optimisation here is some more cores without sacrificing clock, not a lot more cores at low clock.

 

Before I throw any money at it, I suppose I should actually try it maybe...

[KenjiUmino]

7 minutes ago, porina said:

I have found shadowplay supports higher resolutions, and can capture windowed but might do the whole desktop in the process.

shadowplay uses the NVENC hardware encoder so there won't be that much CPU impact either.

 

i say NVENC is fine for recording. it sucks at low bitrates but if you are only recording you might as well set the highest quality possible and then the difference between software and hardware encoding isn't a problem any more.

[porina]

1 minute ago, KenjiUmino said:

shadowplay uses the NVENC hardware encoder so there won't be that much CPU impact either.

 

i say NVENC is fine for recording. it sucks at low bitrates but if you are only recording you might as well set the highest quality possible and then the difference between software and hardware encoding isn't a problem any more.

Ok, I might have incorrect got the assumption that CPU might limit "software" solutions and performance. Then again, if it is consuming GPU resource, would it impact gameplay? I think I am not GPU limited at current settings, and if the video encoding uses a separate processing block that might not have an impact anyway. Depending on the bitrate I have a mechanical HD for bulk storage, but could also move in another SSD if needed.

[KenjiUmino]

6 minutes ago, porina said:

Ok, I might have incorrect got the assumption that CPU might limit "software" solutions and performance. Then again, if it is consuming GPU resource, would it impact gameplay? I think I am not GPU limited at current settings, and if the video encoding uses a separate processing block that might not have an impact anyway. Depending on the bitrate I have a mechanical HD for bulk storage, but could also move in another SSD if needed.

AFAIK, the NVENC is a seperate section somewhere on the GPU chip with the sole purpose of encoding video.

 

that means recording your game with this is not "stealing" any cuda cores away from the actual game itself or anything like that. 

 

the overall performance impact (if any) should be so small you won't even notice ... i use nvenc all the time and wold probably not know if it's on or off if i didn't turn it on myself (i can tell you if the software encododer kicked in or not though )

 

same with the HDD - i mean yes ... compared to a SSD, a mechanical drive is slower ... but unless you got a 12 year old 5400rpm drive it should be fast enough to keep up with your video. 

[porina]

3 minutes ago, KenjiUmino said:

the overall performance impact (if any) should be so small you won't even notice ... i use nvenc all the time and wold probably not know if it's on or off if i didn't turn it on myself (i can tell you if the software encododer kicked in or not though )

Sounds good, and I'm looking forward to testing it after work. Maybe I'm confusing it with things like xsplit or obs but I assumed it was more CPU based.

 

3 minutes ago, KenjiUmino said:

same with the HDD - i mean yes ... compared to a SSD, a mechanical drive is slower ... but unless you got a 12 year old 5400rpm drive it should be fast enough to keep up with your video. 

Well, it is a 1TB 2.5" WD Blue so it is not a high end drive at all. I only added it for bulk storage when SSDs were getting full. I have a still unused 512GB SSD after I RMA a dead one so I could move that in. #firstworldproblems

[tom_w141]

50 minutes ago, porina said:

Ryzen will be too slow

I can game at 100Hz on my PG348Q so I highly doubt you will find it too slow. (100Hz is Panel Limit)

[porina]

6 minutes ago, tom_w141 said:

I can game at 100Hz on my PG348Q so I highly doubt you will find it too slow. (100Hz is Panel Limit)

If it isn't on FFXIV it is irrelevant. I have tried hitting 144 Hz before and this game is CPU limiting before it is GPU limiting.

[tom_w141]

Just now, porina said:

If it isn't on FFXIV it is irrelevant. I have tried hitting 144 Hz before and this game is CPU limiting before it is GPU limiting.

Haven't played that game in years sorry. Nice concept but combat was just so slow compared to other mmos. Went back to SWTOR and WoW which are both also cpu limited due to being single threaded but can still hit 100 fps. I don't know if higher because I don't turn g-sync off and frames above 100 are diminishing returns anyway. The difference in perception between 60-100 is noticeable but 100-144 not so much.

[KenjiUmino]

8 minutes ago, porina said:

Sounds good, and I'm looking forward to testing it after work. Maybe I'm confusing it with things like xsplit or obs but I assumed it was more CPU based.

obs can use the nvenc as well - and has a rather low CPU usage when it does. maybe 12-15% on my core i5

[porina]

12 minutes ago, tom_w141 said:

Haven't played that game in years sorry. Nice concept but combat was just so slow compared to other mmos. Went back to SWTOR and WoW which are both also cpu limited due to being single threaded but can still hit 100 fps. I don't know if higher because I don't turn g-sync off and frames above 100 are diminishing returns anyway. The difference in perception between 60-100 is noticeable but 100-144 not so much.

Still irrelevant if I want to play FFXIV and not WoW or others. When I tried it, WoW was already technically ancient anyway and could probably run on a potato. Dunno if requirements went up in later expansions.

 

12 minutes ago, KenjiUmino said:

obs can use the nvenc as well - and has a rather low CPU usage when it does. maybe 12-15% on my core i5

Good to note, I'll give shadowplay a try first since I already have it.

[tom_w141]

@porina btw I found this which you might appreciate for setting up 16:9 recordings at your required resolution. Set this as your wallpaper, resize your game accordingly and then set desktop wallpaper back.

[PlayStation 2]

Someone else already mentioned OBS using NVENC. That would be worth a try. If you can record at 2160p at 60fps fine (which I'll assume is the case) then your card should have no trouble with 3440x1440 at 60fps with a nice bitrate.

[porina]

5 hours ago, KenjiUmino said:

obs can use the nvenc as well - and has a rather low CPU usage when it does. maybe 12-15% on my core i5

 

30 minutes ago, Dan Castellaneta said:

Someone else already mentioned OBS using NVENC. That would be worth a try. If you can record at 2160p at 60fps fine (which I'll assume is the case) then your card should have no trouble with 3440x1440 at 60fps with a nice bitrate.

 

I've just tried shadowplay. First recording had practically zero CPU usage, but I was disappointed to find it rescaled my output to 1080p even though I selected "use game resolution". I forced 1440p and then it was fine. At 1080p it was doing about 3 minutes per GB. I've yet to test out 1440p in anger.