ShadowPlay - GVR - Mirillis Action - Fraps - Show your performance hit

[pgi947]

Good idea greg and well done.

Looking forward to the results.

@Pgi,

Iirc, you have Assetto Corsa???

If yes, a special shadowplay request for you-driving a white Abarth 500 EsseEsse and post it up for me please?


If Pgi doesn't have it, would one of the great oCuk forum members here maybe manage to do a run for tommy, please?

Yeah I have it, will have to fraps the results of any hit though so unless the gap is massive we could be looking at margin of error stuff.

 

[FloppyPoppy]

Can you try bioshock infinite Greg? I did a similar thing the other day and lost 22% of my frame rate, looking at your results in other games I'm thinking something's a miss at my end.

My situation is very similar. I actually stumbled upon a video about dxtory and for me it actually feels smoother than shadowplay

 

[FloppyPoppy]

I decided to go and test shadowplay/dxtory/fraps after reading this thread, I did this because as i said above dxtory always feels smother to me personaly than shadowplay

reading this and seeing such small performance hits i knew i suffered much larger performance hits around 20-30% with shadowplay

Im not sure if this is anything to do with SLI/resolution/drive speed. Genuinely don't know why but it always did feel faster after i watched this video on it

System

3970x @ 4.8Ghz. 2400Mhz HyperXBeast, 2x Gtx780, MSI Big Bang xpower II, 2x 1TB evo 840Ssd's, 1x 4TB Wd Black, Corsair 900D, Gpus&Cpu&mobo Watercooled


I have only benchmarked hitman absolution for this. If anyone would like i can try with other games. what i want to understand is where our differences in performance are coming from. I have a feeling it could be a resolution thing. As shadowplay did only recently add support for 1440p

Ok so for game settings

Recorder settings ( please note all recorders were set to record to a samsung evo 840 )

Shadowplay

Manual mode
resolution 2560x1440
Bitrate 50
framerate 60


Fraps

Full Size
1000fps ( to remove the frame limiter ) ( Max on shadowplay is 60fps but this does not cap your fps )


Dxtory
Codec - Lagarith lossless codec
Quality - 100
framerate - 60 ( Like shadowplay this does not cap the frame rate it just records to the drive at 60fps )
format AVI
Scaling 100% ( 2560x1440)



I have used these settings to try and keep it as fair as possible - Where i can match settings i have done and when i cannot i have put them as close as possible.

As you can see by the video i am getting quite a big performance difference between Shadowplay and Dxtory, and fraps is still awfull

I am aware that others have not quite the advertised performance from shadowplay

The performance is still great compared to fraps but for me at the moment il keep using Dxtory


EDIT

One more thing i thought will be of interest is file size

bearing in mind all videos are the same length within the margin of 3 seconds

Shadowplay : 340MB
Fraps : 6.95 GB
Dxtory : 3.08 GB

 

[Rusty0611]

File size isn't a massive thing unless you're not doing anything with the files afterwards. More often than not you're going to compress / encode and edit the raw footage anyway.

Shadowplay doesn't have the best quality raw footage and the file size reflects that. I think it's more than good enough if you're uploading to YT though.

 

[LtMatt]

I noticed the same with GameDVR. I can produce such better quality and smaller file sizes using Mirillis Action and then converting with VirtualDub. I have to convert the MP4 to AVI before i can use it in VirtualDub and it looks pretty crap once uploaded to the tube. Going to have to try and find a better way to do it, but I'm not sure how.

 

[ubersonic]

Meh, so the performance hit show offs take when recording themselves for Youtube has been lowered, wasted R&D time that could have been better used elsewhere IMO.

 

[pgi947]

I noticed the same with GameDVR. I can produce such better quality and smaller file sizes using Mirillis Action and then converting with VirtualDub. I have to convert the MP4 to AVI before i can use it in VirtualDub and it looks pretty crap once uploaded to the tube. Going to have to try and find a better way to do it, but I'm not sure how.

What's your process just now?

Mp4 (source) > covert to avi > edit in vdub > publish

Or

Mp4 (source) > convert to avi with vdub > publish

what decoder codec is used by mirillis? H/.264 or mpeg? Also what are you using that requires avi over mp4?

 

[LtMatt]

What's your process just now?

Mp4 (source) > covert to avi > edit in vdub > publish

Or

Mp4 (source) > convert to avi with vdub > publish

what decoder codec is used by mirillis? H/.264 or mpeg? Also what are you using that requires avi over mp4?

With GameDVR output file is MP4. File size is quite big, too big for my connection to upload. I usually have to convert it somehow. I stopped using MP4 via Mirillis for this exact reason. So far with DVR I've been using something else to convert the MP4>AVI then encode using H264 using VirtualDub. However the quality is not very good and uploading to youtube kills it.

Prior to this i used Mirillis and recorded in AVI, then used VirtualDub to encode using H264. This gave great quality and decently sized files. Uploading to youtube kept the quality the same too.

 

[Gregster]

Youtube does tend to trash the file quality. It is free and easy and I feel we all know that youtube is a no frills video player. I have almost finished my SLI testing and massive performance hits with ShadowPlay

 

[LtMatt]

Youtube does tend to trash the file quality. It is free and easy and I feel we all know that youtube is a no frills video player. I have almost finished my SLI testing and massive performance hits with ShadowPlay

It does not trash it at all if you record in AVI, then encode using H264. Quality is great. It's only when you upload an MP4 file to youtube that it goes crappy.

 

[Varsh]

I use AVI via Mirillis because my i7 doesn't support QuickSync - this means I can't capture MPEG at anything above 720P.

As the AVI files can't be edited using the Mirillis codec, I have to get Action to convert them first but leave them uncompressed, then I use Handbrake (VirtualDub is a bit old hat now) to make highly compress files with a very crisp picture. So far any videos I've done and placed on YouTube are crystal clear.

If you want help with some Handbrake settings than I can help if you like.

 

[Gregster]

It does not trash it at all if you record in AVI, then encode using H264. Quality is great. It's only when you upload an MP4 file to youtube that it goes crappy.

That's something I will get into on my next 2 weeks off from work (5 weeks time). I am loving the recording feature and want to learn more (a good part of the reason for this thread as well).

 

[LtMatt]

I use AVI via Mirillis because my i7 doesn't support QuickSync - this means I can't capture MPEG at anything above 720P.

As the AVI files can't be edited using the Mirillis codec, I have to get Action to convert them first but leave them uncompressed, then I use Handbrake (VirtualDub is a bit old hat now) to make highly compress files with a very crisp picture. So far any videos I've done and placed on YouTube are crystal clear.

If you want help with some Handbrake settings than I can help if you like.

I'd definitely be interested in your handbrake settings. I tried using that but i found the quality to be worse than what i could achieve with VirtualDub. The problem i had with handbrake was it made writing/text a bit blurry where as VirtualDub was clear. I'm sure it was just an option i was missing or forgot to enable.

 

[pgi947]

With GameDVR output file is MP4. File size is quite big, too big for my connection to upload. I usually have to convert it somehow. I stopped using MP4 via Mirillis for this exact reason. So far with DVR I've been using something else to convert the MP4>AVI then encode using H264 using VirtualDub. However the quality is not very good and uploading to youtube kills it.

Prior to this i used Mirillis and recorded in AVI, then used VirtualDub to encode using H264. This gave great quality and decently sized files. Uploading to youtube kept the quality the same too.

I'll have a look tonight, you should just be able to downsample the raw file and use something like aac on the audio to reduce file size, handbrake is alright, I'll try remember what tools I used to use though I know I did demux and work on raw files then remux to the container.

Youtube does tend to trash the file quality. It is free and easy and I feel we all know that youtube is a no frills video player. I have almost finished my SLI testing and massive performance hits with ShadowPlay

I wonder if shadowplay is allocating a lot of resources from the second gpu then? As my bioshock hit I mentioned earlier was with sli enabled

 

[Rusty0611]

Surely a percentage is a percentage?

( x / y -1) * 100

Or you can use the more accurate method detailed in the other thread.

 

[Rusty0611]

I noticed the same with GameDVR. I can produce such better quality and smaller file sizes using Mirillis Action and then converting with VirtualDub. I have to convert the MP4 to AVI before i can use it in VirtualDub and it looks pretty crap once uploaded to the tube. Going to have to try and find a better way to do it, but I'm not sure how.

You don't. You just need the right plugin for vdub.

 

[LtMatt]

You don't. You just need the right plugin for vdub.

Which plugin?

 

[Gregster]

SLI performance hits are a big hit and when recording, neither GPU really gets over 70% usage, so I feel for SLI, ShadowPlay needs some work.

To make it easier on the eye's, here is an averages only bench:

As you can see, some massive performance hits.

Spoiler

Sleeping dogs averages with ShadowPlay (SP) on were 113.5 and with SP off, 152.5. That equates to a hit of 29.3%

Spoiler

Sniper Elite V2 got an average of 143.97 with SP on and with it off got 207.65. That equates to a 36.2% hit on performance.

Spoiler

Thief up next and averages with SP on are 85.3 and off are 111.1. This is a hit of 26.2%

Spoiler

Tomb Raider gave averages with SP on of 104.3 and with SP off, 149.8. This is a hit of 35.8%

Spoiler

Batman with SP on gave 123 fps averages and with SP off, gave 186 averages. This is a massive 40.7% performance hit

Spoiler

Hitman has never been a great game for scaling on SLI systems but does see the lowest performance hit of 15.5% with averages of 96.5 (SP on) and 82.6 (SP off). Poor game for SLI users overall.

Finally, we have Bioshock Infinite and with SLI getting averages of 202.84 with SP off and a measley 132.43 with SP on, this is a big hit of 42%

So to summarise:

• Tomb Raider - 35.8%
• Thief - 26.2%
• Sniper Elite V2 - 36.2%
• Hitman Absolution - 15.5%
• Batman Arkham Origins - 40.7%
• Sleeping Dogs - 29.3%
• Bioshock Infinite - 42%

When we compare this to the single card results, there is quite a contrast.

 

[pgi947]

Ok so after scouring forums, it turns out shadowplay and sli has a bug causing the performance hit, comment from nvidia as recently as the end of July says it will be fixed 'soon' in an upcoming driver release. Though the bug has been known since November 2013 so I wouldn't hold your breath as it doesn't seem high up on nvidias list of priorities.

 

[GoogalyMoogaly]

( x / y -1) * 100

Or you can use the more accurate method detailed in the other thread.

Surely it'd be:
((x / y) - 1) * -100

Not sure you can get more accurate than that if you're calculating a perfomance hit (x = DVR fps, y = normal fps).