Sony vs. MainConcept

cris wrote on 3/21/2018, 3:19 PM

Just want to share the experience. I'm soon out with a new video for the band as, since we're gonna do some serious promotion this time, I want it to look its best. Therefore I tried the two encoders (Sony and Mainconcept) at different rendering settings to check out what's best. Since this is going both on terrestrial tv and of course youtube, there's really no limit - it can be a large file (YT will of course re-encode on upload, so I have to test what looks better there).

Now for anything from 16Mbps up in general I think very few viewers would actually notice much difference - one really has to look and know what to look for to see them. Obviously within large, reasonable limits most viewers are far more focused on the content (thank godness!) than the visual detail. The video is filmed and colored as an indie movie so I wanted the final look to match that vibe. I also haven't really disabled resample almost anywhere in the project because none of the exports have suffered of major problem in any way. Differences are minimal and any of the renderings would work fine.

I found that both encoders do a reasonably good job so long they have enough bitrate to work with. With stock setting, didn't like too much the 24fps/16Mbps Mainconcept - there was some (subtle but perceptible) stuttering on slightly slowed down scense (all the footage is 50p to allow for slowdowns). Sony (at 24fps/25Mbps) was better, but still some movements were still blurred in a not so nice way. So no "film" rate for this one. Sony 30fps/16Mbps wasnt less better than I expected, while going for the Sony at full throttle - 50fps and 25bps - eliminated all the movement imperfections and gave a very "digital" and detailed look - TV-documetnary style. Perfect for precision but not so great for vibe. Also the color rendering was a little less good than I liked, with still a little noticeable artifacts in the fuill-white walls hit by oblique lights. Could live with that tough - and the size was 650Gb for about 4 mins video.

Then I ran the Mainconcept again, this time with pretty heavy options - NSTC fps (29,970), variable bitrate with average 24mbps and 50 max, two pass and 2 reference frames - and it really knocked it! Good color, smooth movements but much less clinic looks. Quite impressive. The result wasn't so different in size, about 750Gb.

Without it being an exhaustive test, and granting that both encoders produce good results when set alright, the rule of thumb for me now is to use the Sony for detail and accuracy and the Mainconcept for film-like stuff.

Cheers,

-cris

Comments

3POINT wrote on 3/21/2018, 3:29 PM

(all the footage is 50p to allow for slowdowns). Sony (at 24fps/25Mbps) was better, but still some movements were still blurred in a not so nice way.

Probably you forgot to disable resampling!

cris wrote on 3/21/2018, 3:32 PM

(all the footage is 50p to allow for slowdowns). Sony (at 24fps/25Mbps) was better, but still some movements were still blurred in a not so nice way.

Probably you forgot to disable resampling!

No, these were the only bits where I tried that. :-)

cris wrote on 3/21/2018, 3:33 PM

There's to say, the mainconcept with the heavy settings takes a bit longer to render (dont remember if I had the GPU acceleration set or not). But that's really no big deal unless one's rendering for the day after at 3:00am :D

Musicvid wrote on 3/21/2018, 7:00 PM

And at those bitrates, MC and Sony are every bit the equal of x264, although the latter may be still faster.

So the true test remains--at what normal DELIVERY bitrates do these encoders retain the most bearable quality? Easy to forget that connections and CPUs of the average family are well below those being optimized here by hobbyists.

If we merely halve the delivery frame rates being discussed, to 30 and 25, optimized bitrates are halved as well, putting them almost within reach of the average five-figure household, and at which x264 is arguably better in many (most?) cases than the commercially licensed codecs.

Fact: the average pre-i5 processors of a few years back have trouble keeping up with sustained 20 Mbps bitrates even with optimized players such as VLC (they invented x264 btw).

Your tests remind me of the tests run several years back by master photographer Jerry am Ende, who tested those same three encoders for the lowest deliverable bitrate, using low-motion source (talking heads against a static background.

http://www.jazzythedog.com/testing/lowbitrate.htm

You do need Flash to play those old uploads, sorry.

Thanks for running those much-needed tests, cris!

Tim L wrote on 3/21/2018, 8:10 PM

...Could live with that tough - and the size was 650Gb for about 4 mins video. ...

...The result wasn't so different in size, about 750Gb. ...

Surely those file sizes are Megabytes, not Gigabytes, right?

(And now I just did the math: (25,000,000 bps / 8 bits-per-byte) x 4 minutes x 60 sec = 750,000,000 (i.e 750 MB) )

Whew!... 650 - 750 GB had me scared at first...

Musicvid wrote on 3/22/2018, 6:06 AM

Here is a simplified version of Tim's formula. Note that Mb = Megabits and MB = MegaBytes,

Bitrate (Mbps) x Time (Sec) x .125 = File Size (MB)

Saves me the Whew! factor because I can remember it and usually do all the work in my head; that is, when I'm not enjoying Colorado's variety of exotic flora.

cris wrote on 3/22/2018, 1:03 PM

...Could live with that tough - and the size was 650Gb for about 4 mins video. ...

...The result wasn't so different in size, about 750Gb. ...

Surely those file sizes are Megabytes, not Gigabytes, right?

(And now I just did the math: (25,000,000 bps / 8 bits-per-byte) x 4 minutes x 60 sec = 750,000,000 (i.e 750 MB) )

Whew!... 650 - 750 GB had me scared at first...

Haha yes, MB of course :D

cris wrote on 3/22/2018, 1:07 PM

And at those bitrates, MC and Sony are every bit the equal of x264, although the latter may be still faster.

So the true test remains--at what normal DELIVERY bitrates do these encoders retain the most bearable quality? Easy to forget that connections and CPUs of the average family are well below those being optimized here by hobbyists.

If we merely halve the delivery frame rates being discussed, to 30 and 25, optimized bitrates are halved as well, putting them almost within reach of the average five-figure household, and at which x264 is arguably better in many (most?) cases than the commercially licensed codecs.

Fact: the average pre-i5 processors of a few years back have trouble keeping up with sustained 20 Mbps bitrates even with optimized players such as VLC (they invented x264 btw).

Your tests remind me of the tests run several years back by master photographer Jerry am Ende, who tested those same three encoders for the lowest deliverable bitrate, using low-motion source (talking heads against a static background.

http://www.jazzythedog.com/testing/lowbitrate.htm

You do need Flash to play those old uploads, sorry.

Thanks for running those much-needed tests, cris!

A very good point with the speed. In my case it's for broadcast (so they'll likely process the file further) and for youtube upload (and the YT encoder will rework as well), but of course if the idea is to "consume" the video directly on a PC, yes, some old one may not have enough steam. Hadn't thought of that!

Of course when size and playability performance are important it's a different matter. But fun to see how good results we can have with MS (14, I bought 15 but haven't yet dared to install it :-)

 

BrianK wrote on 3/22/2018, 2:25 PM

Does either one play nicer with a Nvidia GPU than the other?

Musicvid wrote on 3/22/2018, 3:23 PM

Would you mean CUDA or NVENC?

BrianK wrote on 3/22/2018, 4:01 PM

Either or both I guess. My card is a Pascal generation Quadro. I have some familiarity with CUDA but NVENC is a new term to me.

BrianK wrote on 3/23/2018, 1:10 AM

OK. Just used NVENC. Finally some benefit to having a Nvidia GPU. I wouldn't say earth shattering but it was faster and definitely took some of the load off my CPU.

cris wrote on 3/23/2018, 2:26 PM

Does either one play nicer with a Nvidia GPU than the other?

I checked and yes, I have the GPU acceleration set up to that PC's Nvidia card, a GForce GTX 770.