NVIDIA RTX Voice: Real-World Testing & Performance

[Vayra86]

Still not a proof. But at least adds some statistical significance.

And yes, I proposed an experiment which would probably provide some interesting results and smear some scientific method on this discussion.
But maybe it's too much for this community.
I apologize for the idea. Let's go back to the usual "Nvidia is evil" points.

Your experiment only serves to expand even further on a totally irrelevant point though, one you have been dragging on for two pages for no tangible reason except Nvidia saying 'it runs over Tensor'... and that is what I meant with selective reading too. Even best case your net result is what, 2-3% of performance gaps. Why bother. Besides, nothing is stopping you from doing a test yourself, right? Is that too much for you, too?

Another telltale sign here is that first you apparently found 'lots of people' with problems using non Turing cards with RTX voice... and yet, no evidence or source was provided... I mean, why, man?

 

[nico_80)]

Here.

 

[TheGuruStud]

Physx part deux!

 

[Dimi]

The amount of hate towards Nvidia in this comment section is just unreal.

I've watched several youtube demonstrations of this technology and it looks very useful. Mind you, this is still in beta.

 

[TheGuruStud]

The amount of hate towards Nvidia in this comment section is just unreal.

I've watched several youtube demonstrations of this technology and it looks very useful. Mind you, this is still in beta.

It's another feature no one asked for and is suited for all the CPU horsepower we have, now. Any video perf penalty is not acceptable. It's stupid.

 

[Vayra86]

It's another feature no one asked for and is suited for all the CPU horsepower we have, now. Any video perf penalty is not acceptable. It's stupid.

So you don't install it?

 

[heky]

I am not against Nvidia, i just think it is misleading to market a program as only running on RTX technology, which is clearly not the case. Even the name itself is misleading. Sure Nvidia made the program and can do with it what they want, but that does not change the fact the marketing is misleading.(Making people think they have to buy an RTX card to get such functionality)

 

[the_wtfunicorn]

I use Nvidia maximum fps in the driver to cap the fps to monitor refresh - 2fps. The fps is extremely stable compare to ingame fps limiter, which is in my opinion more crucial to aiming accuracy rather than slightly lower input lag. I have been playing almost exclusively BR games like PUBG and COD Warzone and had done lot of testing to increase my KDR (around 2.0 KDR so far).

My point here regarding RTX Voice is that the reduction in performance is related to Tensor Cores sucking available power from the shader cores (~30watt), which capping FPS will instantly solve as your card would not reach its power limit. Undervolting here should work too.

I tested this feature with BFV with RTX ON and surprisingly I found that this new app actually uses the same hardware as ray tracing does since my fps were in the mid 30 with an RTX 2080, but as soon as I quit RTX Voice the FPS went up into the 70s (keep in mind I play at 1440p ultra settings, except for RTX which is on medium, with DLSS). My theory on why RTX Voice works with GTX cards would be that CUDA Cores are emulating (of some sorts) the tensor cores. This would be possible since Nvidia brought ray tracing to SOME GTX cards through emulating those exact cores I am talking about.

Nvidia RTX Voice: Sehr gute Ergebnisse, aber es reicht auch GeForce GTX

Entgegen Nvidias Angaben funktioniert die Geräuschunterdrückung RTX Voice auch auf GTX-Grafikkarten. Die Ergebnisse lassen sich hören.

www.computerbase.de

TL;DR : no difference between RTX and GTX. Just pure marketing bluff.

I tested this feature with BFV with RTX ON and surprisingly I found that this new app actually uses the same hardware as ray tracing does since my fps were in the mid 30 with an RTX 2080, but as soon as I quit RTX Voice the FPS went up into the 70s (keep in mind I play at 1440p ultra settings, except for RTX which is on medium, with DLSS). My theory on why RTX Voice works with GTX cards would be that CUDA Cores are emulating (of some sorts) the tensor cores. This would be possible since Nvidia brought ray tracing to SOME GTX cards through emulating those exact cores I am talking about.

 

[toilet pepper]

I tried it in my rig and it surprisingly works. However, the moment you run it the card's core clock and mem ramps up to its game clock.

 

[Fergutor]

You're worried about voice communication latency or game latency? 2 separate beings.

Voice comm latency.

 

[Xzibit]

GTX 1080 tested with the mod

 

[ZoneDymo]

The amount of hate towards Nvidia in this comment section is just unreal.

I've watched several youtube demonstrations of this technology and it looks very useful. Mind you, this is still in beta.

It really isnt about it being useful or not, its about Nvidia making it arbitrarly RTX only.
Its just very fake to me to promote a new product with new features when really there is nothing new...they just simply lock out existing/older products from doing the exact same thing.
Ez money......

Imagine if they actually were straight about this:
"we wont give you, our loyal customer, this new tech that we pushed forward for people working at home because of the pandemic, because we want more money out of you first, we dont really value that you bought our previous product, thats old, we want NEW money, oh and for the next cards we will make some other tech that can totally work on the card you are about to buy but that will be old money by then so yeah screw you, more money pls"

If this tech ACTUALLY had to rely on tensor cores, on AI that only runs well on tensor cores, then fine, its legit connected to new technology only found on the latest cards, but it doesnt soooo yeah, bunch of anti-consumer horsesh!t from big N in my opinion.

 

[the_wtfunicorn]

The amount of hate towards Nvidia in this comment section is just unreal.

I've watched several youtube demonstrations of this technology and it looks very useful. Mind you, this is still in beta.

A VERY VERY EARLY beta!

 

[Vayra86]

I tested this feature with BFV with RTX ON and surprisingly I found that this new app actually uses the same hardware as ray tracing does since my fps were in the mid 30 with an RTX 2080, but as soon as I quit RTX Voice the FPS went up into the 70s (keep in mind I play at 1440p ultra settings, except for RTX which is on medium, with DLSS). My theory on why RTX Voice works with GTX cards would be that CUDA Cores are emulating (of some sorts) the tensor cores. This would be possible since Nvidia brought ray tracing to SOME GTX cards through emulating those exact cores I am talking about.



I tested this feature with BFV with RTX ON and surprisingly I found that this new app actually uses the same hardware as ray tracing does since my fps were in the mid 30 with an RTX 2080, but as soon as I quit RTX Voice the FPS went up into the 70s (keep in mind I play at 1440p ultra settings, except for RTX which is on medium, with DLSS). My theory on why RTX Voice works with GTX cards would be that CUDA Cores are emulating (of some sorts) the tensor cores. This would be possible since Nvidia brought ray tracing to SOME GTX cards through emulating those exact cores I am talking about.

This kind of echoes the FPS hit we saw in the review and it will probably get much worse (like in your case) when tensor cores are already at max load... I think this rabbit hole is even deeper than we thought and CUDA might even end up performing better in that specific situation (because CUDA is not taxed to capacity by the game itself). The conclusion then, is that this beta version is not built right, in the sense that it does not dynamically allocate the RTX Voice load to whatever resource is available the most. Which makes sense because if they don't allocate to Tensor first, why call it RTX voice eh

Here's what I see when I put two identical Valley runs side by side. The first run is RTX Voice OFF. The second run is RTX Voice ON, with Audacity recording input over RTX Voice and filter over both in and output. Note that I have a GTX 1080, both runs are over CUDA entirely.

Pretty big hit for RTX Voice no matter where it runs, it seems, and from your example, I don't directly get the impression Tensor is more efficient. The only efficiency edge it might have, is power usage. I'm seeing a 19% power load with RTX Voice active over CUDA. Also translates to about 30W though which is rather similar to the review.

The FPS hit seems remarkably similar to the power load RTX Voice requests. 119 > 101 FPS.

 

[Fiendish]

It really isnt about it being useful or not, its about Nvidia making it arbitrarly RTX only.
Its just very fake to me to promote a new product with new features when really there is nothing new...they just simply lock out existing/older products from doing the exact same thing.
Ez money......

Imagine if they actually were straight about this:
"we wont give you, our loyal customer, this new tech that we pushed forward for people working at home because of the pandemic, because we want more money out of you first, we dont really value that you bought our previous product, thats old, we want NEW money, oh and for the next cards we will make some other tech that can totally work on the card you are about to buy but that will be old money by then so yeah screw you, more money pls"

If this tech ACTUALLY had to rely on tensor cores, on AI that only runs well on tensor cores, then fine, its legit connected to new technology only found on the latest cards, but it doesnt soooo yeah, bunch of anti-consumer horsesh!t from big N in my opinion.

Based on the Gamers Nexus testing, their GTX hardware gave worse performance AND quality compared to RTX hardware. Seems like it DOES only run well with tensor core aid.

 

[ZoneDymo]

Based on the Gamers Nexus testing, their GTX hardware gave worse performance AND quality compared to RTX hardware. Seems like it DOES only run well with tensor core aid.

Yeah well Gamers Nexus by own admission made a completely rediculous unrealistic torture test.
Would be nice if they tried an actual....ya know...realistic test in which case I would imagine both do fine.

 

[Wintyer]

I noticed power draw is missleading in this test as if you card was already under load the power draw at least for me on a 1660 super is 1 W. But in this artical they are saying the power draw is higer because it takes it out of idle and not being in idle is the real power draw. with RTX voice turn off and you took your card out of idle would use all the same power minus about 1w. Making it missleading please correct this.

 

[Tech We Love]

So if my buddy has his kids talking in the background on his mic, and I apply the RTX Voice to Output, will that suppress them, just like it suppresses my background noise ?

If so, this is genius.

 

[Wintyer]

So if my buddy has his kids talking in the background on his mic, and I apply the RTX Voice to Output, will that suppress them, just like it suppresses my background noise ?

If so, this is genius.

yes but RTX Voice output does do a bit more of a performance hit then on impute