Nvidia Shadowplay Review & Benchmarks

[SaleemVFX]

Hey guys since Nvidia has announced their new feature to record gameplay off you GPU I was really excited for it. Now it is out I am doing a proper review on it. 

I will be testing to games Bioshock Infinite and Tomb Raider (2013). I will be recording using Fraps and Shadowplay while a benchmark is being done. In this way I can see what impact each recording methods have on the games.

First we have Fraps

 

As you can see it has a lot impact on in the FPS Min section.

 

Next we have Nvidia Shadowplay

 

Now we see a big difference in the game frame rate drops.

 

Now without any recording software

 

As you can see they are little to no impact on games while using Nvidia Shadowplay recording tool.

 

Now Comparing each game in Both Fraps and ShadowPlay

 

Tomb Raider

 

Bioshock Infinite

 

 

 

What about the file sizes? How big are they?

 

Now just watching the chart that is MEGA difference. The file sizes were

Tomb Raider: Fraps = 3.44GB ShadowPlay =349MB

Bioshock Infinite: Fraps = 5 GB ShadowPlay = 410MB

 

Now that is some amazing recording technology Nvidia has implemented for those who have the supported GPU's. The supported GPU's are 600's and 700's series and up.

My Video Review: 

[TheSLSAMG]

Nice review, but it's in the wrong section. It should be here http://linustechtips.com/main/forum/25-member-reviews/

[SaleemVFX]

Nice review, but it's in the wrong section. It should be here http://linustechtips.com/main/forum/25-member-reviews/

can i move it??? or repost it in that section??

[brownninja97]

Thats quite cool, you nvidia folk enjoy that. its great for people recording, for others it doesnt mean much

[TechFan@ic]

These charts are a pain to read, they should've put fraps & shadowplay graphs together.
I'm interested in more common gameplay recording programs like dxtroy.

[TheSLSAMG]

can i move it??? or repost it in that section??

It's okay, just report your original post asking a moderator to move it for you. We all make mistakes, like when I tried to insert an AGP card in a PCI slot.

[SaleemVFX]

thanks

[_ASSASSIN_]

Thats quite cool, you nvidia folk enjoy that. its great for people recording, for others it doesnt mean much

it means a lot for game streamers and recorders, and that is a large portion of the community, so it was an extremely smart move to do shadowplay

[rustikles]

Yeah, it would more effective to put fraps and shadowplay side by side per game in each category so that you can actually see the difference clearly. i.e. graph for each game. Otherwise, cool results!

[brownninja97]

it means a lot for game streamers and recorders, and that is a large portion of the community, so it was an extremely smart move to do shadowplay

Its awesome for those that can use it, i agree with your statement. Its nice to see nvidia doing something great with the premium prices they make people pay.

[SaleemVFX]

Yeah, it would more effective to put fraps and shadowplay side by side per game in each category so that you can actually see the difference clearly. i.e. graph for each game. Otherwise, cool results!

Hey i added the two games, comparing each one 

[rustikles]

Yay! Thanks for the update.

[CaptainSi]

Shadowplay looks like a fricken great little technology. Performance graphs are cool and all, but I'd really like to see some more investigation into the actual compression on the footage and the effect on image quality. To achieve such a small storage footprint it's obvious that Nvidia is quite deep into AVC's encoding levels, I'm thinking ~4 (which could explain the 30p limit). At that sort of level the footage has lost 9/10ths of it's potential colour depth, and the worrying issue at this level is to increase perceivable IQ ever so slightly requires almost an immediate tripling of the bitrate. Scott Michaud from Pcper even mentioned that the colour profile of the video was quite flat, completely obliterating yellows. At this implementation I think Nvidia has hit the best comprimise they could for storage vs performance degradation vs image quality, but I'm worried that at the same time they've dug themselves into a hole which leaves little room for Shadowplay to improve. If Nvidia is truly invested in the technology I think we'll very quickly see either specialized on PCB hardware for next gen cards or inclusion of a HEVC encoder, both solutions which won't help current Shadowplay hardware iterations unless Nvidia is prepared to stretch their current criteria for Shadowplay. It's because of these reasons that I think Shadowplay will not be a Dxtory killer (and the obvious, that AMD doesn't have their own respective technology), especially because of lossless algorithm support, even within the AVC codec. As for Fraps, well, that garbage should have died long ago, and I'll happily shoot bullets at that sinking ship to help it on it's way.

[TheProfosist]

The file size it just compression you could do it in something else also the only really nice thing is it uses the dedicated hardware encoder.

[legopc]

wow those stats are great oh wait i need a new gpu for that...

[TheProfosist]

it means a lot for game streamers and recorders, and that is a large portion of the community, so it was an extremely smart move to do shadowplay

Means nothing for streamers as you cant stream with it unless you can read the temp/cache file.

[TheProfosist]

Shadowplay looks like a fricken great little technology. Performance graphs are cool and all, but I'd really like to see some more investigation into the actual compression on the footage and the effect on image quality. To achieve such a small storage footprint it's obvious that Nvidia is quite deep into AVC's encoding levels, I'm thinking ~4 (which could explain the 30p limit). At that sort of level the footage has lost 9/10ths of it's potential colour depth, and the worrying issue at this level is to increase perceivable IQ ever so slightly requires almost an immediate tripling of the bitrate. Scott Michaud from Pcper even mentioned that the colour profile of the video was quite flat, completely obliterating yellows. At this implementation I think Nvidia has hit the best comprimise they could for storage vs performance degradation vs image quality, but I'm worried that at the same time they've dug themselves into a hole which leaves little room for Shadowplay to improve. If Nvidia is truly invested in the technology I think we'll very quickly see either specialized on PCB hardware for next gen cards or inclusion of a HEVC encoder, both solutions which won't help current Shadowplay hardware iterations unless Nvidia is prepared to stretch their current criteria for Shadowplay. It's because of these reasons that I think Shadowplay will not be a Dxtory killer (and the obvious, that AMD doesn't have their own respective technology), especially because of lossless algorithm support, even within the AVC codec. As for Fraps, well, that garbage should have died long ago, and I'll happily shoot bullets at that sinking ship to help it on it's way.

You can check what settings they used pretty eaily with media info. 4gb for 10min or even 4gb for 20min is not really all that small for compressed video. Im fairly sure they expect you to transcode it at least one befor utilizing it anywhere. Its using some quite odd settings here and there but that might be bacause their actually restricting it currently. Either way it uses lvl 5 and there isnt a 30p limit. Its recorded constant 60p but oddly using variable framerate which basically fortells improvements on the way. Its being recoded to standard 4:2:0 which unless you have professional stuff youll never be seeing anything higher than that for quite some time. This would be instead of a games native 4:4:4. Encoders like x264 cam encode h.264 using that profile bit the amount of things that can decode it is quite small and the amount of things that can properly decode then display it properly is even smaller. The way nvidia has it now you could just throw the video straight up onto the web if you please with a bit of trimming or just as is. I see no need for a specialized separate piece of hardware as that defeates the purpose of this which is described pretty clearly in the name. Also HEVC doesnt have a proper software encoder yet that is reasonable much less a hardware encoder. Besides that though HEVC is h.265 that doesnt just make it the next standard. There are a good amount of competeing technologies out there right now, sony currently having the one with the most pull. Stretch the criteria? Do you mean make more cards work with this? I find that highly unlikely as it uses a hardware encoder that it on the card and as far as we know kepler was the first tobhave it which would make sense since shield probably uses the same tech. Just a heads up lossless x264 isnt really h.264/AVC-1 and is as compatible as a pile of poop your much better off useing lag or huffy.

[ACatWithThumbs]

Means nothing for streamers as you cant stream with it unless you can read the temp/cache file.

Streaming comes with the final release it's still in BETA.

 

[_ASSASSIN_]

Means nothing for streamers as you cant stream with it unless you can read the temp/cache file.

they are working on streaming with twitch

[TheProfosist]

Streaming comes with the final release it's still in BETA.

they are working on streaming with twitch

I await to see how its implemented as the current implementation doesnt allow for it unless they allow streaming sites and programs access to the Shadow Play cache. Which I still need to find out where it is.

[CaptainSi]

You can check what settings they used pretty eaily with media info. 4gb for 10min or even 4gb for 20min is not really all that small for compressed video. Im fairly sure they expect you to transcode it at least one befor utilizing it anywhere. Its using some quite odd settings here and there but that might be bacause their actually restricting it currently. Either way it uses lvl 5 and there isnt a 30p limit. Its recorded constant 60p but oddly using variable framerate which basically fortells improvements on the way. Its being recoded to standard 4:2:0 which unless you have professional stuff youll never be seeing anything higher than that for quite some time. This would be instead of a games native 4:4:4. Encoders like x264 cam encode h.264 using that profile bit the amount of things that can decode it is quite small and the amount of things that can properly decode then display it properly is even smaller. The way nvidia has it now you could just throw the video straight up onto the web if you please with a bit of trimming or just as is. I see no need for a specialized separate piece of hardware as that defeates the purpose of this which is described pretty clearly in the name. Also HEVC doesnt have a proper software encoder yet that is reasonable much less a hardware encoder. Besides that though HEVC is h.265 that doesnt just make it the next standard. There are a good amount of competeing technologies out there right now, sony currently having the one with the most pull. Stretch the criteria? Do you mean make more cards work with this? I find that highly unlikely as it uses a hardware encoder that it on the card and as far as we know kepler was the first tobhave it which would make sense since shield probably uses the same tech. Just a heads up lossless x264 isnt really h.264/AVC-1 and is as compatible as a pile of poop your much better off useing lag or huffy.

Thanks for the info man. I'm still on a 6970 so I haven't been able to try out Shadowplay, I was just going off of information I had read about the place, not sure where I got the 30p thing from. I agree with Nvidia's tact of providing an "upload ready" video output, and I should probably clarify - I didn't mean a standalone hardware encoder, I said an on PCB hardware encoder, so it would be a part of the card. I don't think the current h.264 encoder is robust enough to be a proper replacement for Dxtory or discrete hardware capture solutions- yet, however taking advantage of upcoming codecs like HEVC may allow it to be. I disagree with your hesitation about HEVC become a well supported standard, as it is integrated immensely well with HTML 5. It may not be "the standard", but unfortunately I've not yet evolved the ability of foresight. Maybe you have. By stretch the criteria I meant allowing for higher bitrate footage and for lower performance hit thresholds. I'm very aware that the Largarith codec is vastly superior in terms of support compared to AVC lossless, perhaps the sentence structure confused you, but thanks for the "heads up" I suppose.

[Whiskers]

The only issue I have with Shadowplay so far is that the audio quality is very inconsistent.

 

It may be a result of it not working properly with surround sound, but I find any audio sources that are directed from the rear in a game sound awful when captured in Shadowplay. I will record a YouTube later to show what I mean.

[TheProfosist]

The only issue I have with Shadowplay so far is that the audio quality is very inconsistent.

 

It may be a result of it not working properly with surround sound, but I find any audio sources that are directed from the rear in a game sound awful when captured in Shadowplay. I will record a YouTube later to show what I mean.

I wouldnt doubt it would have problems with surround sound since its and mp4 and having surround in one of those can be done but not everything can play it back. This is my guess as to why they stuck with 2ch sound, also I dont believe youtube and other online sites support surround, though I may be wrong there.

For those of you interested here are the pastebin's that @Hazy125 sent me. These were recorded on Win7. The first one is Shadow Time and the second one is Manual.

http://pastebin.com/XAqk6ahD

http://pastebin.com/w17UfLaN

[CaptainSi]

I wouldnt doubt it would have problems with surround sound since its and mp4 and having surround in one of those can be done but not everything can play it back. This is my guess as to why they stuck with 2ch sound, also I dont believe youtube and other online sites support surround, though I may be wrong there.

For those of you interested here are the pastebin's that @Hazy125 sent me. These were recorded on Win7. The first one is Shadow Time and the second one is Manual.

http://pastebin.com/XAqk6ahD

http://pastebin.com/w17UfLaN

Thanks so much for posting the pastebin links, this is exactly what I'm after. At L5, AVC is capped at just over 72fps@FHD. I'm interested in knowing how shadowplay handles the extra frames higher than this threshold. If it's just cutting them out that may explain the audio synch issues people have been having, in tandem with the variable frame rate. It furthers my concern about a criteria hole they've dug themselves in, as we can see that the jump to L5 to L5.1 requires an almost doubling of the bitrate. Also, no 4k support unless they do make the jump to L5.1. Not so relevant now obviously, but will be in the future.

[Focux]

Hey guys I can't seem to find my ShadowPlay recordings..is it the file output in the software itself..? I press Alt-F10 and I see a slash over a circle