Overclock.net - An Overclocking Community - Reply to Topic
Thread: [Nvidia] DLSS boosts Port Royal Benchmark performance by up to 50% Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
02-11-2019 02:12 AM
ToTheSun! I wouldn't mind running simulations for nVidia during the winter.
02-11-2019 01:46 AM
epic1337
Quote: Originally Posted by ILoveHighDPI View Post
If anyone would use "AI Cores" for actually running NPC AI, that sounds awesome, heck yeah I'll pay extra for that. I keep thinking that Tenser Cores have a much brighter future in gaming than RT cores do.
or a full blown AI bot.
https://www.engadget.com/2019/01/24/...tion-tlo-mana/

though self-learning features isn't limited to AI, they could be used for map creation, obstacles or unit placements.
e.g. imagine a map or scenario that gets progressively harder, instead of relying on randomized or predefined results.
02-11-2019 12:25 AM
ILoveHighDPI If anyone would use "AI Cores" for actually running NPC AI, that sounds awesome, heck yeah I'll pay extra for that. I keep thinking that Tenser Cores have a much brighter future in gaming than RT cores do.
02-10-2019 08:06 PM
ZealotKi11er
Quote: Originally Posted by senileoldman View Post
Would be interesting to see if the next generation of consoles start using Nvidia hardware, or if AMD puts AI cores on their cards.
1) Next gen as far as we know uses Navi which does not have DLSS/Ray Tracing.
2) This technology makes zero sense for Consoles. You are allocating die space for this. Nvidia just trying to make use of the idle tensor cores and ray tracing cores.
02-10-2019 05:19 PM
UltraMega
Quote: Originally Posted by epic1337 View Post
erm, you make it sound like community contribution is a waste of resources.
but sure, only Nvidia users that are interested in obscure games would benefit from this, the others could just simply ignore DLSS altogether.
This wouldn't be "community contribution", it would be a free super compute cluster for Nvidia. You might be interested in using your hardware as a node in an Nvidia cluster but it would be a bad move for Nvidia for obvious reasons like the pure desperation of the idea in the first place. Nvidia taking the position of releasing $1200 dollar GPUs and then asking users to help them compile the data needed to properly utilize these GPUs would just be laughable from a marketing stand point.
02-10-2019 03:48 PM
senileoldman Would be interesting to see if the next generation of consoles start using Nvidia hardware, or if AMD puts AI cores on their cards.
02-10-2019 03:30 PM
guitarmageddon88
Quote: Originally Posted by ToTheSun! View Post
https://www.reddit.com/r/nvidia/comm..._does_it_work/

What stunned me was that I got a 50% (!) performance increase at 1440p in Port Royal with DLSS, while the graphical result was subjectively better overall. It introduced some artifacts, but the quality of reflections was much better.

Now, this is a best case scenario, of course, but if we can get even just 20-30% performance increases for future games with (quasi) completely raytraced graphics (which 7nm nVidia cards might be able to realize), I really don't see how machine learning, with Tensor Cores and AMD analogues, can ever not be part of graphics cards again.
Yes I agree. You can tell that there was some anti-aliasing happening but it did look Superior in my opinion. Maybe the smoothness of the rendering has to do with that just as much as literal image quality (which some have said is not as good , as in the final fantasy examples)
02-10-2019 07:12 AM
ToTheSun!
Quote: Originally Posted by guitarmageddon88 View Post
So what I'm confused about is what we are exactly training to "learn". Do my results get stored in a server? Is there a DLSS info file somewhere within the program directory that keeps learning? How does it continuously improve and is that data stored with the user?
https://www.reddit.com/r/nvidia/comm..._does_it_work/

What stunned me was that I got a 50% (!) performance increase at 1440p in Port Royal with DLSS, while the graphical result was subjectively better overall. It introduced some artifacts, but the quality of reflections was much better.

Now, this is a best case scenario, of course, but if we can get even just 20-30% performance increases for future games with (quasi) completely raytraced graphics (which 7nm nVidia cards might be able to realize), I really don't see how machine learning, with Tensor Cores and AMD analogues, can ever not be part of graphics cards again.
02-09-2019 09:31 PM
epic1337 yes, the AI is trained in Nvidia's servers to learn and compare between native 4K and upscaled output.
by doing this the upscaled output would be much closer to native 4K in quality.
02-09-2019 09:14 PM
huzzug
Quote: Originally Posted by guitarmageddon88 View Post
So what I'm confused about is what we are exactly training to "learn". Do my results get stored in a server? Is there a DLSS info file somewhere within the program directory that keeps learning? How does it continuously improve and is that data stored with the user?
You run nothing other than the game when you play it. The game is run by Nvidia, Jensen to be exact in his backyard garage at 4K, the findings of which are compiled with their monthly PC Pro Drivers which you can use if you're subscribed to. Your card merely analyses those findings and runs that game on your 4K.
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off