Question / Help Stuttering in Local Recordings on HDD, encoder lag

[MrCakenom]

I'm looking for some help for a problem I have been experiencing with my local recordings with OBS. Any help and suggestions would be greatly appreciated.

When I look back at my recordings I notice some stuttering or where the video freezes for a second or two. I investigate my the obs stats window to see a high percentage of lost frames due to encoding lag. This would vary from 3% to 11%. Not sure what the reason could be so I test a range of recording settings. I test a bunch of bit rate and quality settings but still get encoding lag. I also test this on a range of games and results are consistent for demanding and less demanding games, lost frames from encoding lag.

Then I tested recording on one of my SSDs instead of my hard drive and behold encoding lag was gone. I tested another hard drive I use games for rather than the one I record to and there was still encoding lag (I was running a game from my SSD at the time so it shouldn't affect the results), so it seems the issue seems to be with using hard drives in general.

At the end of the post I've posted a log file where I did a test session where I played in minecraft and recorded 5 minutes onto my hard drive then 5 minutes to my SSD. Stuttering gameplay on hard drive and smooth gameplay on the ssd. I use a dedicated hard drive for the recordings so no other tasks should be using the recording hard drive. This hard drive is a Seagate ST2000DM008. The drive was checked for errors and defragemented before recording.

In the recording I also overlayed task manger to see if there were any patterns. Disk usuage would be low then spike high around 95% - 80% where high number of lost frames occur, though I suspect frames are being lost consistently judging by the smoothness of the SSD gameplay compared with the HDD gameplay.

I'm quite confused of why the hard drive is producing such results since as far as I know the hard drive should be handling the settings just fine and many guides on forums and youtube have suggested even higher settings for recording on hard drives.

Here's the log file: https://obsproject.com/logs/UhYiKS1YO4S30UqP
I changed the output to my SSD where the log file states
"21:27:06.007: Settings changed (outputs)"

Many Thanks

 

[Narcogen]

21:25:52.737: Video stopped, number of skipped frames due to encoding lag: 1483/18157 (8.2%)

I don't believe a hard drive performance issue presents in the OBS log as encoding lag, but I could be wrong.

https://obsproject.com/wiki/General-Performance-And-Encoding-Issues

 

[MrCakenom]

21:25:52.737: Video stopped, number of skipped frames due to encoding lag: 1483/18157 (8.2%)

I don't believe a hard drive performance issue presents in the OBS log as encoding lag, but I could be wrong.

https://obsproject.com/wiki/General-Performance-And-Encoding-Issues

This article is targeted at encoding with x264. The issue is encoding lag from nvenc is rare but for what ever reason I keep encountering it despite lowering to 720p and 30fps. But I shouldn't need to lower setting this much with nvenc.

Log file of recording at 720p and 30fps: https://obsproject.com/logs/qRvyttMkbAb_qs_Z
I get 2.4% skipped frames

 

[R1CH]

The ST2000DM008 is an SMR drive, these have extreme performance issues since to write data, they must first read existing data, modify it then write. Most drives have some small amount of cache so the performance drop isn't too noticeable until you write enough data. At high bitrates an SMR drive may not be able to keep up with an OBS recording.

 

[MrCakenom]

The ST2000DM008 is an SMR drive, these have extreme performance issues since to write data, they must first read existing data, modify it then write. Most drives have some small amount of cache so the performance drop isn't too noticeable until you write enough data. At high bitrates an SMR drive may not be able to keep up with an OBS recording.

Thank you very much for your help! This issue has been driving me mad! I will be looking to buy a drive with cache.

 

[Narcogen]

The ST2000DM008 is an SMR drive, these have extreme performance issues since to write data, they must first read existing data, modify it then write. Most drives have some small amount of cache so the performance drop isn't too noticeable until you write enough data. At high bitrates an SMR drive may not be able to keep up with an OBS recording.

R1CH, just for my edification, if OBS has trouble writing to disk when recording, is this included in the listing of frames dropped due to encoder lag?

I haven't been including this as a possible cause for users experiencing encoder lag, it would be helpful to know if this is sometimes the case.

 

[MrCakenom]

Just been looking at hard drives and I assume hard drives marketed for performance and for creative work loads are unlikely to be SMR. Something like a WD black HDD. I'm considering a Seagate Barracuda Pro (considering 2tb or 4tb depending on the price difference) but I wanted to double check on this forum to confirm if anyone knows of any issues with this drive.

Many thanks

 

[R1CH]

Generally the cheapest drives are SMR, or will have only 1-2 platters instead of 3-4 of other drives. You should be able to tell from the spec sheets in some way, or just Google the model name + SMR to see if anything shows up.

 

[koala]

Where did you actually find the information that the ST2000DM008 is a SMR HD? I mean reliable and definite information. It's not that I distrust your information, I'm simply confused. Seagate itself does not give this information, neither in the product info page at https://www.seagate.com/de/de/internal-hard-drives/hdd/barracuda/ nor in the actual data sheet of the product line: https://www.seagate.com/www-content/datasheets/pdfs/3-5-barracudaDS1900-7-1706US-en_US.pdf
In the data sheet, the "max sustained data rate" is given with 220MB/s, so I assumed this rate is delivered even when the cache is full, and 220 MB/s is well enough for any video recording. You don't find the recording format in any product series data sheets. And if it is not mentioned anywhere, it does not seem to be a major performance hit. And if this is the case, I wonder if the HD of the OP may simply be defective instead of just slow.

I'm quite interested in this, since I plan to build a new computer this year and intend to not buy a SMR device. With the arrival of SMR for desktop devices it becomes quite difficult to choose a proper hard disk. Up to now (2016), I selected HDs by fitness for 24/7 and NAS-approval, but today every manufacturer has so much HD series it becomes quite difficult if you want more then just the cheapest desktop HD whose MTBF is calculated in a way where the user is not supposed to switch it on for more than 8 hours a day.

 

[MrCakenom]

Where did you actually find the information that the ST2000DM008 is a SMR HD? I mean reliable and definite information. It's not that I distrust your information, I'm simply confused. Seagate itself does not give this information, neither in the product info page at https://www.seagate.com/de/de/internal-hard-drives/hdd/barracuda/ nor in the actual data sheet of the product line: https://www.seagate.com/www-content/datasheets/pdfs/3-5-barracudaDS1900-7-1706US-en_US.pdf
In the data sheet, the "max sustained data rate" is given with 220MB/s, so I assumed this rate is delivered even when the cache is full, and 220 MB/s is well enough for any video recording. You don't find the recording format in any product series data sheets. And if it is not mentioned anywhere, it does not seem to be a major performance hit. And if this is the case, I wonder if the HD of the OP may simply be defective instead of just slow.

I'm quite interested in this, since I plan to build a new computer this year and intend to not buy a SMR device. With the arrival of SMR for desktop devices it becomes quite difficult to choose a proper hard disk. Up to now (2016), I selected HDs by fitness for 24/7 and NAS-approval, but today every manufacturer has so much HD series it becomes quite difficult if you want more then just the cheapest desktop HD whose MTBF is calculated in a way where the user is not supposed to switch it on for more than 8 hours a day.

I've read across many threads online complaining about that many brands don't advertise if their HDs use SMR or not. I think R1CH is right about platters since SMR can store more information on fewer platters if I'm not mistaken. You usually find this information by looking up the hard drive's data sheet. This one is for the ST2000DM008:
https://www.seagate.com/www-content...a-fam/barracuda-new/en-us/docs/100817550h.pdf
However I could not find out the number of platters on the ST4000DM006 which I have ended up buying:
https://www.seagate.com/files/www-c.../en-us/docs/barracuda-pro-ds1901-6-1701gb.pdf
I suspect HD manufacturers don't want to confuse consumers with a lot of technical information so they throw alot of terms like "home use" and "professional use", etc to try to guide consumers to buy the right product for them. Thus I suspect the Barracuda pro drives are not SMR while the non pro versions have SMR.

Like R1CH said usually a google search of "[INSERT DRIVE HERE] SMR" is a quick way of identifying SMR drive, as people will usually complain online that it's a SMR drive. I didn't see anyone claiming that ST4000DM006 has SMR which is why I choose the drive. In the unlikely event that it is SMR I will let everyone know

 

[R1CH]

Seagate in particular are quite bad for hiding this information. I personally got tricked myself, buying what I thought was a good drive (ST4000DM004) but it turned out to perform like shit and of course after some digging I find out it uses SMR. Apparently the "Barracuda Pro" line at 7200 RPM are mostly safe, the "Barracuda Compute" and 5400 RPM drives are much more likely to be SMR. Googling for other people's feedback is the safest bet.

 

[MrCakenom]

I've tested some recordings with the new harddrive and only suffered around 0.3% skipped frames when I tested for 4 minutes and rewatching the footage I don't notice any stutter. I also tried out the old Nvenc and got fewer skipped frames (approx 0.045%). Either way I'm happy with the result.

Thanks again R1CH for your help