View Full Version : First Nvidia Encoder results :)
CruNcher
10th January 2009, 20:47
http://s11b.directupload.net/images/090110/8pz3j4we.png <- Software (Cyberlinks Software Encoder)
http://s10.directupload.net/images/090110/dzeabnxp.png <- GPU ? (Nvidia PureVideo Encoder)
http://s10b.directupload.net/images/090110/ywb3gaev.png <- GPU ? (Nvidia PureVideo Encoder)
I'am not really sure if it runs on the GPU with that high CPU utilization it's definitely much higher (70%) then what a Badaboom Encode would consume.
The Final output is High Profile @ Level 4.0 (Badaboom can't do this yet)
Visually Nvidias Encoder Produced a much nicer result in less time here then Cyberlinks Encoder :)
+ Audio
Cyberlink: 5:39s
Nvidia : 2:02s
Im using Nvidias Pre-Alpha 185 Forceware might be that it caused a Software Fallback or maybe a Encoding option still have to figure that out :)
shon3i
10th January 2009, 21:23
where we can find that encoder??
CruNcher
10th January 2009, 21:48
http://www.nvidia.com/object/winxp_181.20_whql.html
http://www.cyberlink.com/multi/download/patches_4_en_US.html
Nvidia doesn't provide yet a Interface to it for free only usable by 3rd Party ISV's currently
PS: It's far from good it fails miserably on most test cases Badaboom easily beats it even being Main Profile only, not worth to test this any further just a waste of time, compared to x264 it looks like a First Day Encoder not sure what to say about Cyberlinks Encoder it's even more in the Direction of 0 Day :P. But maybe the Cyberlink engineers just did wrong implementing it inside their application it looks really bad it has B-frame stability issues in motion scenes, im to scared to encode Touhou with those jesus that must look awful.
rica
11th January 2009, 01:10
Hi cruncher,
I've just paid to Cyberlink so as to give it a go.
I've seen %100 CPU utilization, what's wrong?
FW 180.84 plus CUDA Tools with 260 AMP 2 Edition card.
CruNcher
11th January 2009, 01:20
I feel sorry though @ high bitrates for sure it's much more temporal stable (these are for sure RC bugs anyways) but don't expect as good Detail Preservation as with x264 even @ High Bitrates and yeah this high CPU Utilization is fishy also on my 8800 GT G92 as Badaboom doesn't consume so much CPU and gives 100000 times better results.
You should try http://www.nvidia.com/object/winxp_181.20_whql.html as these are the official Drivers for the PureVideo Encoder but from what i see now i would say this isn't doing any Encoding on the GPU but is just a Software Encoder like ATI's Encoder was @ the beginning of the HD1000 series it seems to me Nvidia rushed this out to have something Marketing wise for CES vs ATI.
And marketing wise it works wonderful for sure you have this already awful Cyberlink Encoder now you check the Hardware Encoding and boom it's much faster over 2x and the result is also much better though far away from what is SOA nowdays and what you can easily get for free with x264. But on a Consumer show it works people wont ask or look @ the CPU utilization they would just see the Result on screen imho it's FRAUD.
rica
11th January 2009, 01:39
OK, forget it :)
Anyway, i gonna share my shots tomorrow more detailled.
it seems to me Nvidia rushed this out to have something Marketing wise for CES vs ATI.
But i suppose doom9 has started before CES. :sly:
CruNcher
11th January 2009, 01:45
Nope they released it the second day of CES and it's almost over now so people that saw it for a AVCHD High Bitrate Encode believe now that it's awesome stuff, i hate such Marketing which fools Consumers, but maybe i'm wrong and it works indeed as promised on the GPU anyway it's not a very good Encoder, and it makes absolutely no sense to position that besides badaboom except making more profits from the licensing :(
rica
11th January 2009, 01:52
Sorry, i was thinking CES at the end of this month ;)
CruNcher
11th January 2009, 10:19
Here are Visual Results
http://mirror05.x264.nl/CruNcher/force.php?file=./dshow-batchframework/Cyberlink.m2ts
http://mirror05.x264.nl/CruNcher/force.php?file=./dshow-batchframework/Nvidia.m2ts
but hold on to your chair or it will blast you off ;) (i don't mean that in a good way)
rica
11th January 2009, 12:59
Could you upload the master clip?
CruNcher
12th January 2009, 11:59
It's the Parkrun sequence from the VQEG Evaluation process :)
http://forum.doom9.org/showthread.php?t=135034
spyder
12th January 2009, 23:53
Is this encoding support just tools like GPU ME and transforms, etc? I don't think the GPU would be implementing the full encoder so it's likely not NVIDIA's fault if the output is crappy.
rica
12th January 2009, 23:57
Is this encoding support just tools like GPU ME and transforms, etc? I don't think the GPU would be implementing the full encoder so it's likely not NVIDIA's fault if the output is crappy.
%100 agree. It can not use nVidia's HW encoder.
CruNcher
13th January 2009, 01:34
i doubt too the high cpu utilization imho is strange as i stated badaboom utilizes only 30% for the Encoding process and the result is 1000x times better i also tested Arcsofts current Encoder which seems to be getting GPU support soon too (it already has a GPU checkbox though doesn't work on it yet) :)
Arcsofts Encoder Visual Results in Software Mode are already much much better then Cyberlinks or the Nvidia Encoder results :)
Dark Shikari
13th January 2009, 02:48
i doubt too the high cpu utilization imho is strange as i stated badaboom utilizes only 30% for the Encoding process and the result is 1000x times betterIf Badaboom is 1000x better than this, I fear for how bad this must be.
wyti
13th January 2009, 05:23
Simply see the results you will be amazed.
Here are Visual Results
http://mirror05.x264.nl/CruNcher/force.php?file=./dshow-batchframework/Cyberlink.m2ts
http://mirror05.x264.nl/CruNcher/force.php?file=./dshow-batchframework/Nvidia.m2ts
but hold on to your chair or it will blast you off ;) (i don't mean that in a good way)
V-tec
13th January 2009, 08:43
how we can compare quality between this two clips?
the bitrate difference is near to 50%...
CruNcher
13th January 2009, 09:46
hehe technically you are of course right, that isn't my fault though but i guess comes from the ultra high speed mode Nvidias "Software" Encoder uses :)
anyways, especially as the bitrate is 50% higher don't you could expect a much better visual result from the Nvidia Encode sure, but does it deliver that ? no it delivers still temporal problems.
This test @ this bitrate is though heavy to master especially in 1 pass only the top of the top does it @ acceptable quality and that's why im so surprised about Arcsofts Result which im gonna publish next :)
And i still have to find out if this really is working on the GPU and if not (what i suspect from the evidence same case as back the time with ATIs HD1000x Encoder release) why doesn't it as it says clearly GPU HARDWARE on the Checkbox and is advertised also by Cyberlink as such (big press release big mumbo jumbo) also imho it shows how low quality Cyberlink in terms of Codec Development is (they should better license one again) ?
Was it just a CES PR stunt to have something to show even if it isn't really doing what it should, just to have something vs ATI in terms of Marketing ?
And once again the Consumer gets fooled by the ISV and his Partner to believe something which is not the truth ?
rica
13th January 2009, 19:06
What i know i paid 127 USD to a garbage.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.