-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
NVIDIA RTX 20 Series Reviews
- Thread starter Mecha Meister
- Start date
so a rtx 2060 averages ~50fps at 1440p with rtx high? that doesn't seem too bad at all
I can't watch right now, and maybe I'm misunderstanding, but are you saying your GPU utilization is at 60% or you're getting 60fps on your rig?
Oh sweet, thanks for the info about the code!
As for a price match, I know amazon has that specific card cheaper by like $20. Bestbuy.com appears to now have it @ $380 now as well it seems!
Oh ok, that is what I paid for it. I just wanted to make sure I wasnt missing something. It came out to 411 with tax.
If I hook it up and want more, I hope I can return it and get a 2080.
YES. This as well. Digital Foundry is usually very good at showing any artifacts when objects are in motion from TAA/Upscaling methods so hope this continues when they look at DLSS-supported games.
He's running a 2080 on 1080p @ 60fps. Of course the GPU should easy be able to run that, and thus, why it's only 60% usage. I already told him to use DSR and downscale since he has extra GPU power.
Both, the 2080 has no problem running Metro Exodus on ultra, hairworks, physx, and rtx on high with DLSS off at 1080p, most of the time my fans don't even spin up!
I tend to not like using DSR unless I can do a pretty big bump in res. I wouldn't be able to maintain 60 FPS at 2x res...
Last edited:
Given that DLSS borks HDR, I've been playing Metro at 4K with 0.6 pixel rate (or whatever the crap that option is called). That seems to run about as well as DLSS, but I'm not surrendering HDR. The game looks so much better with it. Whatever they're doing for scaling looks pretty clean too, at least on my monitor. Ray tracing on high, naturally vs ultra.
The ray tracing isn't in your face or as obvious as it is in BF5, instead the lighting looks that much more accurate and everything sits together much better in the environment. I really like the aesthetic.
Oh I should have said, my framerate seems to generally be hovering around 58 to 60 fps with these settings. I've seen higher and lower than that, but traversing outside that's where it seems to be. I've not been benchmarking so I can't speak to what I'm averaging, but it's been around that ball park. Fortunately I haven't had any trouble getting Gsync to work, so I'm quite content with that performance level.
The ray tracing isn't in your face or as obvious as it is in BF5, instead the lighting looks that much more accurate and everything sits together much better in the environment. I really like the aesthetic.
Oh I should have said, my framerate seems to generally be hovering around 58 to 60 fps with these settings. I've seen higher and lower than that, but traversing outside that's where it seems to be. I've not been benchmarking so I can't speak to what I'm averaging, but it's been around that ball park. Fortunately I haven't had any trouble getting Gsync to work, so I'm quite content with that performance level.
I was curious if DLSS had an effect on hdr, thanks for the info. Now that modern games are thankfully offering resolution scale option i wonder if DLSS is even necessary.
plagiarize do you have Asus or Acer 4k monitor?
plagiarize do you have Asus or Acer 4k monitor?
I have the Acer Predator. Stupidly pricey, but very nice.
Right now at least in Metro Exodus, they don't work together.
I spent a few hours with Metro on my Vega and a few hours on my 2080ti.
Different PCs, no cloud saves (thanks Epic Store!), so I played the game from the beginning everytime.
Old rendering approach vs new rendering approach, we could say.
Now I understand why Dictator was so excited, and I can only imagine the struggle between expressing what he thought and not doing so given the inevitable hate from, well, haters.
If you are even remotely interested in gaming tech, I think it's impossible to be indifferent to the first RT game out there. How can you not be excited? I know I am.
I think we reached a milestone here.
And as I stongly believe that, I'd like to thank Nvidia and 4A Games for making this possible.
Thanks guys, I mean it, I'm so happy for what I witnessed at home, in real time.
Can't wait to see what will come next (Remedy I'm watching you!), and can't wait for Team Red to jump in.
Different PCs, no cloud saves (thanks Epic Store!), so I played the game from the beginning everytime.
Old rendering approach vs new rendering approach, we could say.
Now I understand why Dictator was so excited, and I can only imagine the struggle between expressing what he thought and not doing so given the inevitable hate from, well, haters.
If you are even remotely interested in gaming tech, I think it's impossible to be indifferent to the first RT game out there. How can you not be excited? I know I am.
I think we reached a milestone here.
And as I stongly believe that, I'd like to thank Nvidia and 4A Games for making this possible.
Thanks guys, I mean it, I'm so happy for what I witnessed at home, in real time.
Can't wait to see what will come next (Remedy I'm watching you!), and can't wait for Team Red to jump in.
I will consider the 1660 Ti an elaborate industry troll until Nvidia announces it.
Anything less than the RTX 2070/GTX 1080/Vega 64 wouldn't be a good enough upgrade, I think, and even those 3 wouldn't be a big upgrade either.
The GTX 980 Ti is close to the GTX 1070 level of performance, so the RTX 2060/GTX 1070 Ti/Vega 56 would be just a small upgrade over that.
Wait for reviews and pricing on the Nvidia RTX 3080. The 2000 series isn't worth it at your power level.
I went from a 1070 (a bit better than your 980 Ti) to a 2080 and I really appreciate the difference. So if you're going to upgrade, go with at least a 2080 or 2080 Ti IMO.
NVIDIA DLSS: Your Questions, Answered
https://www.nvidia.com/en-us/geforce/news/nvidia-dlss-your-questions-answered/
TL;DR taken from reddit
https://www.nvidia.com/en-us/geforce/news/nvidia-dlss-your-questions-answered/
TL;DR taken from reddit
- DLSS is not enabled for every resolution because if the game is already running at high frame rate, you don't get much performance gain from DLSS due to DLSS needing a fixed amount of GPU time per frame to run the neural network. If the GPU processed the frame faster than DLSS operations, it won't get performance boost.
- Blurry frames in lower resolution is the product of having less source data which makes it challenging for DLSS to detect the input frame and predict the final frame. NVIDIA is working to add more training data to improve image quality
- Battlefield V Update: Focusing testing and training to improve 1080p and Ultrawide with DLSS
- Metro Exodus Update: Game update coming to improve DLSS sharpness and overall image quality across all resolution. Also training more sections of the game and looking into other reported issues such as HDR.
same here. i had a 1070 and now have a 2080. i don't think either a 2060 or 2070 would have been a worthwhile upgrade. i would've got a 2080 ti but yeah i ain't paying £1,100+ lol.
Seems like crazy ammount of work is required for this and results are of questionable quality. I don't think DLSS will ever become something that is mainstream and integrated into every game. It will probably always remain something for select high profile titles, assuming, ofcourse, Nvidia doesn't abandon this in a couple of years just like they abandoned many of their proprietary antialiasing methods.
Poor nvidia never doing anything right.
I hope they improve in the future.
Come on guys don't give up!
seems more like a desperate attempt to shoehorn in some type of use for hardware designed for an entirely different market
The thing that could be problematic with DLSS is the time it might take to produce good results.
Most people could be finished with a game (especially if it's single player) before that
Most people could be finished with a game (especially if it's single player) before that
Download the first experimental RTX On Octane Benchmark (expected to be released with 2019.2 or later) - Reddit
RTX 2019 OctanceBench Preview - OTOY Forum, more info+download
Apparently, this is a just a chunk of the environment seen in the GDC 2015 Brigade demo.
2015: The completed scene/Brigade demo was ~3200 Octane Bench score, on 147.2 FP32 TFs worth of GTX 680 class gpus.
2018: About the same OB score, but now on ~118.4 FP32 TFs, 8 Titan V's/GV100s, using a newer version of Octane with Brigade optimizations/code merged in.
2019: A user on the Octane forums achieved a score of 3659.67 in this benchmark by using 4 RTX Titans
Scene stats from Octane 4:
GPU Memory:
RTX 2019 OctanceBench Preview - OTOY Forum, more info+download
I got this working on 417.71, Win 10 1803 (GTX 1080 Ti)
Apparently, this is a just a chunk of the environment seen in the GDC 2015 Brigade demo.
2015: The completed scene/Brigade demo was ~3200 Octane Bench score, on 147.2 FP32 TFs worth of GTX 680 class gpus.
2018: About the same OB score, but now on ~118.4 FP32 TFs, 8 Titan V's/GV100s, using a newer version of Octane with Brigade optimizations/code merged in.
2019: A user on the Octane forums achieved a score of 3659.67 in this benchmark by using 4 RTX Titans
anifex3D said:
Scene stats from Octane 4:
GPU Memory:
- Used: 1658.8MB
- LDR RGB: 71
- HDR RGB: 0
- LDR greyscale: 0
- HDR greyscale: 0
- Total Triangles 1709905
- Emitting Triangles: 1668
- Emitting Instances: 9
2080 at least. You'll get +40-50% performance on average if you won't be CPU limited.
And for roughly the same price I paid for the 1070 years ago. What the hell is going on with the GPU market? We've made literally zero price/performance progress in three years?
With a rtx 2060 from a GTX 980 ti he will gets 20-30% more performance. With a 2080 ti he will get 100-120% more performance.
1060 has almost identical price/performance metric as 2060. There has been no progress at all. This GPU generation is screwed up on multiple levels.
A rtx 2060 oc scanned offers 20-30% more performance in every game. So it is already something. Most of games 25-30% more.
My GTX 980 ti scores 5400 on timespy. My rtx 2060 scores 7970.
1060 has almost identical price/performance metric as 2060. There has been no progress at all. This GPU generation is screwed up on multiple levels.
Is this chart based on today's prices or launch prices for the respective cards?
There's no reason why DLSS should look worse, so yeah, whatever training they've done on Metro so far, it isn't nearly enough. Glad to hear that they recognize it isn't where it needs to be.
Once cards have enough gigarays, it won't be nearly as necessary, so I'd imagine it's got a shelf life. Even though training will take less times as their supercomputers get more and more powerful, that you've got to basically train for every type of environment in that game and what have you is going to ensure that it remains work intensive both on the side of the PCs handling the training and the people setting up the image sets in the first place.
I hope it works really well in Serious Sam 4 though :)
how does nvidia release updated dlss data, does it come with driver updates or through game patches?
When you launch a game supporting DLSS the driver should get the latest data according to this answer in the link posted before :
EDIT : looks like model are stored in "C:\ProgramData\NVIDIA\NGX\models"
So we should be able to see when one is updated.
Last edited:
The 1660 Ti's rumored MSRP for the US is $280, rumored Euro price is 300€ (I never seen the 1070 below 350€, actually years ago that's how much the 1060 costed here, so at least here there's a clear price/performance progress, 1070 performance for possibly lower than 1060 price at launch), but I really have no idea about the UK prices in general.
It's been discussed over and over. Die space is being used to introduce new features. There has been no die shrink, so the die is generalliy massive relative to the previous cards.
There are also other metrics than price/performance. When I play my games I only experience performance.
Isn't timespy a dx12 based test? Firestrike (dx11) a well OCd 980 ti can hit 20k graphics, which is pretty much right where a 2060 is iirc.
Here's a 1450mhz 980 ti (a pretty easy OC for that card) vs a 2050mhz 2060. Pretty similar results accross a range of titles with the 980 ti actuallly besting the 2060 a fair amount. They pretty much trade blows. I wouldn't upgrade from a similar performing 6gb card to another 6gb card but that's just me. I guess if you really want dxr but the performance at 1080p on the 2060 isn't that great with dxr. Minimum upgrade from a 980ti at this point for me yeah would probably be a 2080 or Radeon vii, or used 1080 ti.
https://youtu.be/KFjWi0_g2Rs
Last edited:
Well here most of the games get 25-30% diff.
https://youtu.be/aIHrdc38hQU
Firestrike is a old test and doesn't make sense anymore for the new gpus.
A rtx 2080 ti gets 22-23k but blow 980ti perf by 100% or more.
Yeah so a stock 980 ti, a card that ocs very well. A 2080ti scores 35-40k depending on oc graphics score, which pretty much lines up with it's dx11 performance vs a Radeon vii or 2080 that score about 28k-33k graphics score depending on OC.
Firestrike is a a pretty good gauge of relative performance you can expect as you don't run a ton of stuff in dx12.
The 2080 ti pretty much does score exactly double that of a 980 ti.
Last edited:
980Ti = 1070, 2060 is about 10% faster than 1070, this would be a pointless upgrade. Hence why I've said "at least 2080". And 2080Ti is just really bad value.
10% faster then an 1070ti. Rtx 2060 oc scanned performs on pair with a GTX 1080. So 25-30% more performance in games.
It is the same Chip from 2070 cut a bit. Overclocked performs like a stock 2070
No.
"OC Scanner" works on 1080 as well as 2060.
Nope. 980Ti is also a much better overclocker than 1070/2060.
Why are people always talk about overclocking as if it's only possible to do on one GPU and not all of them?