NIOH2 doesn't support ray tracing. NVIDIA is advertising it for the DLSS support.
Nvidia's 3000 Series RTX GPU [3090s with different memory capacity]
- Thread starter Shortbread
- Start date
-
- Tags
- nvidia
It looks just like it does on consoles....
NIOH2 screen from Nvidia geforce UK fb page, looks amazing with ray tracing. The amount detail is on another level.
Frenetic Pony
Veteran
No no, see raytracing added the valve from the creepy Valve guy. You know, for the steam version.
One really has to wonder how much of it is actually RT being amazeballs versus RT being pushed primarily by Nvidia.
One has to wonder how much RT is being downplayed due to AMD's DXR performance deficit...I think we have seen it all by now but my favourite is still screen space reflection with obvious flawed pushed as better than RT-reflections...
One has to wonder why AMD wasted time to implement DXR features when it is soooooo useless.
(Or not...)
I’m pretty sure everyone knew RT was amazeballs long before Nvidia started talking about it.
Pretty far from it, try playing both, 20mins the console version then switch to pc and vice versa. The pc version maxed out, with the addition of ray tracing using reshade RT and appropiate framerate sure does look better then the ps5 version.
I'm not sure how amazing current implementations can even be when people aren't able to discern when a game has it.
Well, DF does find it a amazing tech and a huge difference in CP2077 for example. Even early games like BFV did show some nice improvements using dxr tech. But yea since the consoles basically lack it, we have to hate it, right?
Quite a turn from the game supposedly looking so incredible and detailed with ray tracing, while it actually doesn't use ray tracing, to suggesting that now we're supposed to mod it with software RT before comparing to consoles
Quite a turn from the game supposedly looking so incredible and detailed with ray tracing, while it actually doesn't use ray tracing, to suggesting that now we're supposed to mod it with software RT before comparing to consoles
Nah, even without ray tracing, the pc version maxed does look better, while also performing better. The addition of ray tracing just lifts it even more above the ps5 version. Anyway, this isnt the thread for console warring, i posted a screenshot of Nioh 2 using RTX tech, in the NV thread, then some like you have to bring up consoles, i have no idea why but keep it elsewhere.
DegustatoR
Legend
These shots were about DLSS only though, no RT is happening there.
DLSS does work considerably better than game's own TAA though so there's that.
DLSS does work considerably better than game's own TAA though so there's that.
No. You posted a screenshot describing how good it looks thanks to ray tracing, while it does not use ray tracing at all. NVIDIA is promoting the game for it's DLSS support, it doesn't have ray tracing support.
Also I didn't bring in the consoles.
D
Deleted member 2197
Guest
RTX 3060 vs Jetson AGX for BERT-Large (naut.ca)
March 16, 2021
March 16, 2021
This article compares the performance and energy efficiency of the RTX 3060 vs the Jetson AGX when running BERT-Large training and inference tasks. This article also looks at the effects of batch size and Automatic Mixed Precision, i.e. FP16 mode, on performance. The benchmark task is the Hugging Face CoNLL-2003 token-classification example, with the PyTorch backend.
....
On paper, the RTX 3060 appears to have 8x the FP32, 4x the GP FP16, and 3.5x the Tensor Core performance compared to the Jetson AGX. However, we will see that the performance changes drastically with different PyTorch modes, and that the RTX 3060 is surprisingly efficient for inference-only tasks.
...
TLDR: the RTX 3060 is great. Get it! Also, isn't it strange that the Jetson AGX, a device designed for low-power inference is more efficient at training than the RTX 3060, but less efficient at inference?
The RTX 3060 is 3x faster than the Jetson AGX at FP16 training, and 4x faster at general-purpose FP32 (TF32=False) training.
Inference testing was done with FP16 only. The RTX 3060 was approximately 7x faster than the Jetson AGX at FP16 inference.
Isn't Jetson using Volta with 1st gen tensor cores, which aren't quite a match for 3060's 3rd gen tensors?
D
Deleted member 2197
Guest
Yeah, though despite that it was still going to be an "apples to oranges" comparison with interesting results.
I can only guess how they were able to measure perf drops in games where resizable bar is not supposed to work at all without according game profile.
It seems they have measured run-to-run variance of their tests or something along these lines in many games without game profiles or NVIDIA's profiles don't work at all or there are some HW issues when ReBAR is enabled on their testing PC.
I assume they would run the tests multiple times to confirm the results so its probably one of the latter issues.
