Notebooks soon to be able to support Nvidia Shadowplay

So anyone who downloaded the brand new drivers and got the updated Nvidia GeForce Experience, maybe got a bit disapointed because we can't use Nvidia Shadowplay, yet. But, when you go to this page and scroll down a bit, you may notice this under requirements: "GeForce GTX 650 or higher desktop GPU required ( notebook GPUs are not supported at this time)". This means that we will be able to use Shadowplay soon! Yey!

 

Hold your horses. Nvidia said the exact same thing about streaming to Shield months ago when Shield was released, but it's never happened for us.

 

Oy, totaly forgot about that. Thanks for the headsup.

 

Would you mind trying something for me? Add this to your GFE shortcut and see if you can get ShadowPlay recording to work on your laptop.



BTW this worked for somebody else: http://forum.notebookreview.com/ide...-owners-questions-thread-110.html#post9427685

 

Holy , just tried it and it seems to work (haven't test it in-game yet). I wish I could test it right now, but I gotta go. Will definitely check it out when I get back though!

 

Yeah this is pretty amazing. First read about it a few hours earlier on the GeForce forums, too bad I'm at school. I'm just dying to get home and try it out.

 

Is there any way Nvidia could configure Shadowplay to record multiple audio inputs ? (like system sound+ mic input for example)
It would really help tons of gamers everywhere.

 

This is good news. But watch your temps. There must be some reason they disabled it.

 

I hope for official support in a future update. Or at the very least that they don't kill off that workaround. The GTX mobile Keplers should be more than able to handle Shadowplay. Will try it out tomorrow.

 

Probably won't be supported by machines with Optimus, just like everything else.

 

For me it starts the Shadowplay interface but does not actually record.

 

Coming in the future.

No difference in temps and GPU usage. Everything is offloaded to the separate H.264 hardware encoder.

Probably because you have Optimus.

 

Hmmm I wonder what's so different abut your setup that you only get 7% performance loss while I get 20-25%. I don't know if it can be explained by the difference in GPU power, since based on what I know about how ShadowPlay works it shouldn't really matter. What's most surprising about your recordings is that they're perfectly smooth and stutter-free. Almost all of the ShadowPlay footage I've seen on YouTube, regardless of system specs, had that same rhythmic stutter that I saw during playback but wasn't there during actual gameplay while recording.

 

I wish i could explain it. Maybe its because i forgot it was recording on my SSD or that its affected by the GPUs total ram. Hopefully its just some kind of software bug thats easily patched.

 

Well I was recording to a separate drive from the one the game was running off of and my 2GB VRAM never got remotely close to being used up, so I don't think it's that. I just hope this gets officially supported on laptops soon and the performance improves for me.

 

NVENC uses a mix of CUDA and specialized h 264 hardware. The encoding hardware does nothing other than compress video. Everything else is handled by CUDA. The results we are seeing is because of this. Audio isn't the big hit. I suspect the big hit is the requirements a MP4 container brings with it. I think if they had stuck to their original plan of using a ts container, we'd be seeing less CUDA use and less of a GPU impact.

 

So have you messed around with the CUDA - GPUs setting in the Nvidia Control Panel? When I tested ShadowPlay I had mine set to All. I noticed that there was 39% video engine load on the first GPU and 0% on the second one. So maybe the brunt of the performance hit is being taken by the first GPU? Maybe if some load can be offloaded to the second GPU it would improve performance? Not sure what all of this means exactly, just throwing it out there.

 

I think the number of Cuda cores available on the GPU probably has a direct influence on the difference in performance when recording. My framerate measurement with 780M SLI with FRAPS as well as what I can see with on-screen display appears to be unchanged whether recording or not. I recorded a Borderlands 2 video with on-screen display and then ran through the same game map without recording using FRAPS benchmark and the results were basically identical.

Spoiler

 

Let's see if I can explain this in common English. Keep in mind I'm not an expert in h 264. Background before turning to System Administration is in Chemistry. More specifically, a decade of primary research in a EM lab that included analyzing images on a fundamental level in both 2d and 3d. I had a client a couple years ago who wanted to get into video editing, so in order to get her up to speed I had to spend more time than I wanted reading about various h 264 packages, encodes, and editing software.

Alright, the reason why low/medium/high has no affect on game performance is because h 264 encoding is done in hardware that's not used for gaming. It doesn't matter what bit rate you tell it to encode at, it's done on a dedicated chip that has nothing to do with gaming performance. In something like a MP4 file, there's more there than just H 264 compressed video. The video and audio are contained in a package container. MP4 is a package container. Think of the Video (h 264) and Audio (AAC/AC3) in a video file as being like a roof and floor. The container package is like the rest of the house that is holding the two together.

There are other containers than MP4 that allow for H 264 with AAC/AC3 audio. Quite a few actually. They are designed for different purposes. For example, over the air TV transmissions are done via a MPEG Transport Steam. That container package has all sorts of extra things so that if the signal degrades, then it is still possible to receive the video.

There's things like frame normalization, gamma correction, and whatever the hell else you need to do before before you start compressing the video. In other words, there's more that goes into H 264 video encoding than compressing a bunch of sequential frames into a video. You have to pre-process the frames, then post processes the video, you have to link the video and audio.

That extra stuff that's needed is being run on gaming part of the GPU. It's using CUDA to do this. This is where there's a loss of gaming performance. The more GPU cores you have running at a higher clock speed, the lower the percent of the total GPU cores you have which are needed to get the job done. The better your GPU, the lower the performance cost.

Judging from what I've seen. The audio and video compression for the recorded video have no measurable affect on gaming performance. It's everything else needed to make a MP4 file that is causing this performance cost. Also, the CUDA part of the video compression appears to have no SLI optimization at all.

The reason why I said it might be faster if they weren't using MP4 as the final format is really simple. MP4 isn't made for dynamic streaming. The beginning and end of the video is constantly changing. Obviously it can be made to work with pretty good results, since NVIDIA has done just that. I suspect m2ts or something similar would have a lower affect on gaming performance. I haven't tested this, so I can't say for sure.

The amount of CUDA processing it's using now clearly assumes a high end desktop GPU. Most notebook GPUs aren't clocked high enough nor have enough CUDA cores to get the job done without impacting gaming performance. Which is probabally why it's disabled by default on their Mobile GPUs.

 

Thanks for the very detailed explanation. I'm been reading on other forums that even people with the bare-minimum GTX 650 desktop GPU, which is no better than what's in my laptop considering my overclock, are getting little to no FPS drop with ShadowPlay. Hopefully what I'm getting can be chalked up to the current "unofficial" status of ShadowPlay on laptops and once we get official support from Nvidia performance will improve.

On the other hand there are people with high-end desktops that are getting the same kind of performance loss I'm experiencing:

https://forums.geforce.com/default/...dowplay-feedback-thread/post/3991353/#3991353

https://forums.geforce.com/default/...dowplay-feedback-thread/post/3991383/#3991383

But the real reason ShadowPlay is unusable for me right now is the stutter that affects all my recordings. This seems to be a widespread issue. This software is definitely in beta as there are a lot of teething issues.

 

I think people who are reporting low performance impact haven't actually done benchmarks. It was my first impression too, then I actually measured the benchmarks and frame rate drops. There's a clear 20-30% loss in FPS while recording. Drop from 60 to 45 fps isn't as obvious as it seems. In games where a single 650m is over kill, there would be no affect at all since you're aren't using the majority of the GPU.

I'll lay money that if they ran 3dmark with it on and off, they would see an immediate and measurable performance hit.

The reason why I'm certain of this is because I have read NVIDIA's developer intro to NVENC, which is the name of the technology used in the Kepler GPUs for hardware based video encoding. It's pretty clear on what the technology does. It's both hardware H 264 and CUDA running at the same time. I've seen the CUDA part of NVENC directly hit the primary GPU on my laptop while game is "idle" (not being used to actually render in a 3d environment), but recording with Shadowplay. It uses between 20-30% of that GPU.

In something like TF2 and DOTA, there's no visual differece at all when it's recording. In games where I'm pushing 99% on both GPUs and already running at a "low" frame rate (like 30-40), it's pretty damn obvious that there's a measurable performance hit.

 

Scaling up with CUDA cores makes sense. But I've noticed that with ShadowPlay enabled my GPU usage actually drops along with FPS in those "idle" game situations you're describing. For example, sitting on the singleplayer main menu of BF3 I normally get 165 FPS and 99% usage on both GPU's. With Shadow Mode on I get 90 FPS but GPU usage is only 80%.

Huh, why is this?

What's the performance impact of ShadowPlay?

 

I'm guessing it's because at worst it takes up 10% of a single card and the second card is artificially limited to 90% to prevent microstutter/frame pacing issues, so you lose 20% perf overall.

 

Kepler's NVFBC / NVIFR / NVENC...ShadowPlay must-read Guru3D.com Forums - View Single Post - MSI Afterburner 3.0.0 Beta 16(2013-10-25)

RivaTunerStatisticsServer v5.5.0 HW accelerated / NVIFR (Inband Frame Readback) allows windowed videocapture support.
Try MSI Afterburner 3.0.0 Beta 16 Download with this version of RTSS Rivatuner Guru3D.com Forums - View Single Post - MSI Afterburner 3.0.0 Beta 16(2013-10-25) compare results

Prerecording Guru3D.com Forums - View Single Post - MSI Afterburner 3.0.0 Beta 15(2013-09-16)

 

thx mate!, it worked awesome! Can i donate or something? You are just wayyyy too awesome

 

I tried forcing shadowplay and experience to use the 660m instead of defaulting to intel still didnt work.

But im guessing because experience is grayed out in control panel thats why shadowplay isnt working on optimus.

 

Is there anyway any of us with optimus can force geforce experience to use the nvidia gpu?

There has to be a way.

 

Happy it worked for you. Have fun!

I can't help you with this since I don't have Optimus, but don't give up hope yet even if this "hack" doesn't work for you: https://forums.geforce.com/default/...s/about-the-shadowplay-/post/3989511/#3989511

Well, I managed to extract this generic statement out of Nvidia:

 

Anyone...?

 

Already saw you post this in another thread. So what are you trying to say with all those links?

 

Compare results.

 

Well the Intel GPU is physically disabled here so no Quick Sync. And the guys in that thread who tried this are reporting no performance increase while recording compared to the previous version of RTSS. Apparently even Unwinder explained that while the new version is supposed to be faster it still can't be as fast as ShadowPlay.

 

Yeah he can't use NVFBC due to no windowed videocapture support but being his first attempt results look promising https://www.youtube.com/watch?v=bP-Fdy8n3ik&feature=youtu.be

 

Can you understand the language he's speaking? I'd like to know the results of the video but can't figure it out just by watching.

 

Min. 50
Max. 117
Avg. 63

ShadowPlay
Min 45
Max 82
Avg. 59

RTSS
Min. 39
Max. 89
Avg. 59

 

He was testing Arkham Origins? But Unwinder says it only works in D3D9 games, unless the guy in the video is doing something else.

Guru3D.com Forums - View Single Post - MSI Afterburner 3.0.0 Beta 16(2013-10-25)

 

I can't understand what the guy is saying just a few words, maybe someone here who speaks Russian could help...probably it worked only with the benchmark or he was using RTSS 5.4.1 hopefully he can find a solution and NVIFR will not be limited to D3D9 or use NVFBC(faster / ShadowPlay), the licensing issues I mentioned earlier are with NVENC not NVFBC differently to Intel QuickSync H.264 encoding...

 

I have a GTX 675 in a Samsung NP700G7C-S01 and this activates fine for me..........

 

But it does not record.

 

Just got this working on my alienware 18 with dual 780's. Seems to be a minimal impact on fps if at all.

 

I have a GeForce 770M and works fine when using an external monitor through HDMI:thumbsup:!!!

Penalty hit its around 10-12% way lower that FRAPS or MSIAfterburner (Different drive for game and for capture folder):







Content lower that 1080p is recorded as 1080p (resize), video quality is good but we need more youtube friendly settings (720p, 30p) and multichannel support.

 

Right, it does not record. I am not on optimus, nor have the ability, but when I go to experience.

 

That's normal.

 

So looks like Optimus will be supported:
https://forums.geforce.com/default/...-a-priority-for-nvidia-/post/3992754/#3992754

I just hadn't seen anything definitive.

 

For me all works fine. I am on a Retina macbook with a 650m and after starting Experience with -shadowplay and opening the Settings of shadowplay I can use it every time i am in a game. And I can not see any Frame Drops at all. I tried it with MWO.

 

What game is MWO?

 

me not had a chance to read through this thread yet but the latest whql released today says Enables GeForce ShadowPlay

http://forum.notebookreview.com/gam...82-geforce-driver-released-19th-november.html

 

Doesn't work with Optimus, at least not on my system. Specs still say desktop card a requirement.

 

ShadowPlay has been enabled since the last 331.65 WHQL driver, but it's only officially supported on desktop Kepler GPU's. There is a hack to get it to work on notebooks, but it only works for non-Optimus GPU's.

 

Not entirely true, hack works too on Optimus GPUs when using HDMI Output.