Re Gsync: when the Main Thread is the source....

[Cpt_Piett 4,801]

7 minutes ago, Noel said:

You can't improve on perfect.

I was going to say this but you beat me to it, lol. You've found your Nirvana. 

Remember the test I did in the other thread? I don't think you ever commented on the results I posted. Did you want to assess how FTV is with G-Sync?

 

[RJC68 823]

1 hour ago, Ianrivaldosmith said:

Try it on Amazon, test it for yourself. Don’t like it? Send it back. 
I don’t believe words can express G-Sync. It’s like me trying to explain the taste of chocolate to you, if you’ve never ever had chocolate before. 

Or Newcastle Brown Ale 😀

[Noel 4,557]

3 minutes ago, Cpt_Piett said:

I was going to say this but you beat me to it, lol. You've found your Nirvana. 

Remember the test I did in the other thread? I don't think you ever commented on the results I posted. Did you want to assess how FTV is with G-Sync?

Yes Sir I did see the test, but I thought the test was confounded since you stated you also used RTSS method I am using along with Gsync:

...here are the results. I'll upload the video once it's done processing. 

Settings: SU10 4K SDR DX12 TAA TLOD130 rest mostly ultra

Sim: 737-700 with livery, otherwise empty community folder, live weather and time, no AI traffic

Situation: Takeoff from KPSP 31L then CATH1.PSP departure up to cruise altitude

Display: 3840x2160 @ 60Hz

RTSS: fps limited to 30 using front edge sync

nVIdia control panel: G-Sync enabled (TV reports steady 60Hz during the entire session)

Sorry I should have responded then.  What would be more useful to see is a CFX analysis of unlocked frames using Gsync!

[Republic3D 1,232]

1 hour ago, Poppingcork said:

HAGS?

Hardware Accelerated GPU Scheduling. 
It's under your Graphics Settings in Windows.

[Cpt_Piett 4,801]

3 minutes ago, Noel said:

Yes Sir I did see the test

Hey no worries, I was just curious of what you thought. I probably misunderstood something since I set locked fps. 

[Noel 4,557]

2 minutes ago, Cpt_Piett said:

Hey no worries, I was just curious of what you thought. I probably misunderstood something since I set locked fps. 

Hey, YOU have been a kind fellow to do these tests!  Next time up go ahead and run that baby unlocked and set up a capture and we can see how CFX evaluates this with Gsync.  I wonder how Gsync varies from Gsync Ultimate?

And Ian, do the same, using Gsync Ultimate and CapFrameX and we can see how CFX evaluates frametime variance in Gsync Ultimate.

 

Edited November 8, 2022 by Noel

[Noel 4,557]

Techno wizards if you will:  if I bump up the pixel count by 25% on my 9900K/380Ti machine, what kind of impact would we expect to see on the CPU with this change in pre-rendering load to the CPU?  Is it roughly 25% more demand by the GPU resulting in 25% more demand to the CPU, or something less than that?

Edited November 8, 2022 by Noel

[Ianrivaldosmith 5,317]

1 hour ago, RJC68 said:

Or Newcastle Brown Ale 😀

One of Californias favoured drinks, and yet tastes different there to Newcastle. ( I am from Newcastle). And here it is called wife beeta! 

[Ianrivaldosmith 5,317]

1 hour ago, Noel said:

And Ian, do the same, using Gsync Ultimate and CapFrameX and we can see how CFX evaluates frametime variance in Gsync Ultimate.

Once I am back up and running, aim is the next 7-10 days now.

[Cpt_Piett 4,801]

38 minutes ago, Noel said:

Techno wizards if you will:  if I bump up the pixel count by 25% on my 9900K/380Ti machine, what kind of impact would we expect to see on the CPU with this change in pre-rendering load to the CPU?  Is it roughly 25% more demand by the GPU resulting in 25% more demand to the CPU, or something less than that?

AFAIK increasing the pixel count would have minimal impact on the CPU. I'd like to hear other opinions on this though. 

I'll add my 5 cents on mainthread/GPU heavy settings as I don't have that much else to do now that I (finally) got the Fenix airborne from Toronto Pearson en route to Logan Intl. By the way it was a stuttery experience - @Noel you would have hated it 😉 I was running TLOD150. But I'm testing real-time online traffic in SU11 and at big hubs (Pearson is huge) the amount of AI traffic is insane. And apart from AIG and FSLTL, there's no way of reducing the amount of AI traffic using real-time online. 

EDIT: Oh, and the Fenix only adds insult to injury to the poor mainthread. 

Mainthread:

• Terrain LOD. Probably also object LOD. 
• AI traffic - aircraft, road traffic, sea traffic, airport ground vehicle traffic
• Photogrammetry
• Certain aircraft addons. The Fenix seems to have one of the biggest impacts on mainthread. PMDG 737 is much, much better in this regard. 
• Glass cockpit refresh rate

GPU:

• Display resolution (pixel count)
• Render scaling
• Clouds (I only see any benefit of DLSS in very cloud heavy situations)
• Texture resolution
• Shadow size
• Anisotropic filtering
• Ambient occlusion
• Reflections

What about off-screen terrain pre-caching? CPU or GPU?

1 hour ago, Noel said:

Next time up go ahead and run that baby unlocked and set up a capture and we can see how CFX evaluates this with Gsync.  I wonder how Gsync varies from Gsync Ultimate?

I'm tempted to give it another try. Won't bother with a video probably, unless you think it would be super-useful. Not in this flight though as I can hear the mainthread screaming and moaning in the background. This will only worsen as I approach FT's KBOS... 

Next flight perhaps. If @Noel can confirm this is the correct setup:

• G-Sync (obviously)
• No RTSS fps limit
• No V-Sync
• Some sort of scenario that doesn't put a heavy load on the mainthread - I'm thinking a PMDG flight somewhere in Alaska without AI traffic. PAKT to PAJN perhaps?

Edited November 8, 2022 by Cpt_Piett

[RJC68 823]

14 minutes ago, Ianrivaldosmith said:

One of Californias favoured drinks, and yet tastes different there to Newcastle. ( I am from Newcastle). And here it is called wife beeta! 

I am from the UK Ian and I know Newcastle very well, spent many good times working there over the years, one of my favourite places (And I am a southerner lol). The Newcastle Brown is not too bad here in Canada but still way better in it's home. I recently discovered how to really confuse my Chilean wife when she sat down and attempted to watch an episode of Auf Wiedersehen Pet with me. I had to put subtitles on for her even though her English is fantastic 🤣

[RJC68 823]

17 minutes ago, Cpt_Piett said:

AFAIK increasing the pixel count would have minimal impact on the CPU. I'd like to hear other opinions on this though. 

I'll add my 5 cents on mainthread/GPU heavy settings as I don't have that much else to do now that I (finally) got the Fenix airborne from Toronto Pearson en route to Logan Intl. By the way it was a stuttery experience - @Noel you would have hated it 😉 I was running TLOD150. But I'm testing real-time online traffic in SU11 and at big hubs (Pearson is huge) the amount of AI traffic is insane. And apart from AIG and FSLTL, there's no way of reducing the amount of AI traffic using real-time online. 

EDIT: Oh, and the Fenix only adds insult to injury to the poor mainthread. 

Mainthread:

• Terrain LOD. Probably also object LOD. 
• AI traffic - aircraft, road traffic, sea traffic, airport ground vehicle traffic
• Photogrammetry
• Certain aircraft addons. The Fenix seems to have one of the biggest impacts on mainthread.
• Glass cockpit refresh rate

GPU:

• Display resolution (pixel count)
• Render scaling
• Clouds
• Texture resolution
• Shadow size
• Anisotropic filtering
• Ambient occlusion
• Reflections

What about off-screen terrain pre-caching? CPU or GPU?

You should have stopped by for coffee lol

I am about 15 mins north west of YYZ, I go there regularly with a coffee and my scanner. I have the excellent FlyTampa CYYZ but it really taxes my system with traffic. I have even resized my FSLTL textures & Fenix liveries to 2K which helps, the only way I can get decent stutter free performance there with FSLTL is using the 30Hz, 30fps & Max Judder control on the TV methodology. Before I get yelled at for 30fps lol with the Samsung 4K TV I am testing on it actually works surprisingly well. The judder control really smooths out the panning at 30fps. Maybe the nextgen AMD 3D processors will help with performance at these big hubs.

[Noel 4,557]

40 minutes ago, Cpt_Piett said:

I'm tempted to give it another try. Won't bother with a video probably, unless you think it would be super-useful. Not in this flight though as I can hear the mainthread screaming and moaning in the background. This will only worsen as I approach FT's KBOS... 

Next flight perhaps. If @Noel can confirm this is the correct setup:

• G-Sync (obviously)
• No RTSS fps limit
• No V-Sync
• Some sort of scenario that doesn't put a heavy load on the mainthread - I'm thinking a PMDG flight somewhere in Alaska without AI traffic. PAKT to PAJN perhaps?

Sounds perfect!  Set NVCP Low Latency to ON for this as well if you don't mind, and no other limiters anywhere you got it.  Yes, we want rule out mainthread interference w/ Gsync Nirvana!  Go up in something easy to process, anything, then we'll have a chance first to see how CapFrameX deals w/ Gsync and FTV.  If it looks like it's interpreting reality the you could amp up the load a little more.  All my tests were in the CRJ700 or PMDG 738.  Or pick a lightweight airport to fly out of and capture the first 15 minutes from gate to altitude or what have you.  Thanks!  Can't wait!!

[Poppingcork 24]

Would a FreeSync monitor yield the same results as one with G-snyc?  I think I know the answer but wanted to ask anyways....

[Ianrivaldosmith 5,317]

6 minutes ago, Poppingcork said:

Would a FreeSync monitor yield the same results as one with G-snyc?  I think I know the answer but wanted to ask anyways....

Same as g-sync compatible probably but not proper g-sync or g-sync ultimate .