RT and prebaked doesn't set in the same sentence. Ray Tracing has nothing to do with pre-baking ..
RT and prebaked doesn't set in the same sentence. Ray Tracing has nothing to do with pre-baking ..Pre-baked RT and DLSS demos.
From my understanding, the DNN is making huge supersampled versions to determine the ground truth so that for particular scenes or geometry (?) it knows ahead of time what is the ideal subsample pixel location for a pixel when it hits a geomtry boundary that requires subsampling. What I don't understand is what the training consists of and how that data is then finally stored and provided to the driver.
With sometime like infiltrator demo, there's no interaction so you can sample each frame and it's identical each time. With a game it's all dynamic, you don't know ahead of time where the camera is and what it's facing. Do they have monkeys at Nvidia playing the game at 64x SSAA running around everywhere for days on end? Does the game developer have to provide a method of interaction with the game engine that the DNN can hook into and render all possible scenes and angles? That's an insane amount of data to analyze and then what exactly is provided to the driver for the game?
What I think is more likely is that it's using the same principles behind AI image upscaling. In this case the images used for training would be specific to the game itself. The more likely limitation (or difficulty) is games with more varied visuals compared to Infiltrator. For example a game that takes place in very dark in door environment and a very bright outdoor environment with completely different visual aesthetics I'd assume you need to train for both cases. If you expand that then even more.
Starts page 19 - https://developer.download.nvidia.c...DEWpLlr_s6o7LeUkA8O_AJ8QP0yixtreyL1D3CCsWzzsm
Millions? Likely in the thousands, not even over tens of thousands. 2080's were always available for pre-order since announced and 2080ti's were sold out rather quickly. How many people you think worldwide would be buying $1300 US GPUs?millions of cards are preordered
Yea it's a money grab it feels like. There's really no reason for it other than they can say the drivers aren't ready. People are paying $1200+ for a card that's marketed for tech that isn't fully fleshed out yet. If you do want to use it you will have to drop to 1080p. You can't even say it's future proofing to buy one because next year or after Ray tracing is even better implemented, you will have to buy another card anyway.I almost don't even have the words for this release. Nvidia STILL isn't allowing benchmarks when millions of cards are preordered? They're literally releasing performance data ONE DAY before shipments go out? Total insanity. Then take into account the massive, massive price increases and this is just crazy.
RT and prebaked doesn't set in the same sentence. Ray Tracing has nothing to do with pre-baking ..
Yea it's a money grab it feels like. There's really no reason for it other than they can say the drivers aren't ready. People are paying $1200+ for a card that's marketed for tech that isn't fully fleshed out yet. If you do want to use it you will have to drop to 1080p. You can't even say it's future proofing to buy one because next year or after Ray tracing is even better implemented, you will have to buy another card anyway.
It's not a benchmarks "leak". They're numbers provided by Nvidia for marketing slides authorized for release today. There are no numbers available from independent reviewers yet and won't be till the 20th apart from the usual stupid omgz timespy numbers, and some other random numbers from personal leakers with no frame of reference or settings.
It's in the first paragraph. Let me bold for others here.If legit (videocardz is usually pretty reliable with this kind of stuff)
The data we are sharing with you today comes from official Reviewers’ Guide. The numbers in this guide are only a reference for further benchmarking. It is probably an important thing to say that those numbers are should not be taken very seriously. Each reviewer has a different testing methodology (different scenario, different testing equipment, a different list of games).
Some numbers are way off.If legit (videocardz is usually pretty reliable with this kind of stuff), then I think it falls right in line with expectations (I'm assuming the 1080Ti Nvidia used was FE which is what they've been using for comparison so far).
Basically for "traditional" gaming, 1080 Ti AiB ~ 2080 FE. The 2080 Ti looks to be about 40-50% faster on average than the 1080Ti but that will drop to 30% or less when using a non-reference 1080Ti.
*Note, this math was done quickly in my head, I could be a bit off and can run the actual numbers in a bit.
It's in the first paragraph. Let me bold for others here.
Some numbers are way off.
Bf1-only 15% gap between 1080TI and 1080 at 4k???
F1 2018-again only 21% gap.In guru3d review there is 33% gap https://www.guru3d.com/articles_pages/f1_2018_pc_graphics_performance_benchmark_review,6.html
Mass effect-only 19% gap....
SW battlefront2-15% gap??????
1080TI is average 35% faster in 4K than 1080.
Your opening sentence indicated you didn't know if the data source was legitimate. I was pointing out where the data source was coming from whilst also emphasizing that these are from Nvidia marketing and thus should not be taken as results from independent analysis.I'm confused, what is your point?
Your opening sentence indicated you didn't know if the data source was legitimate. I was pointing out where the data source was coming from whilst also emphasizing that these are from Nvidia marketing and thus should not be taken as results from independent analysis.
It looks more like Random numbers generator for 1080Ti they must downclock 1080Ti to like 1500mhz to be only 15-20% faster than 1080 in 4k...Meh, assuming these are real numbers from Nvidia, they'd want the 20xx series to look as good as possible, so I'm sure they used FE models which means the 1080Ti has a higher tendency to thermally throttle vs the 1080, especially if they put it in a cramped case vs. open air. When I was running the numbers in my head I was giving 1080 Ti +15% for an after market model, perhaps it will end up being more like +20% or more in the hands of reviewers, we'll have to wait and see.
Some numbers are way off.
Bf1-only 15% gap between 1080TI and 1080 at 4k??? -36% in techpowerup review https://www.techpowerup.com/reviews/MSI/GTX_1080_Ti_Gaming_X_Trio/7.html
F1 2018-again only 21% gap.In guru3d review there is 33% gap https://www.guru3d.com/articles_pages/f1_2018_pc_graphics_performance_benchmark_review,6.html
Mass effect-only 19% gap....
SW battlefront2-15% gap??????
1080TI is average 35% faster in 4K than 1080.
Edit: also in rainbow six gap is 30% in nv chart but again 36% in techpower up FE vs FE and 45%!! vs AIB 1080TI https://www.techpowerup.com/reviews/MSI/GTX_1080_Ti_Gaming_X_Trio/20.html
Haha so lets wait for proper reviews...
really?Numbers from these official slides (4K, SDR), are showing GTX 1080 Ti is ~30% faster than GTX 1080
You're using the HDR YUV422 numbers, not the SDR numbers.Correct me if I'm worng, but according to these numbers, GTX 1080 Ti will give you ~30% higher FPS than GTX 1080