| Captain_Yuri said: Yea this is what I meant when I said it probably won't be usable at all pre-4000 even if you could use it. Cause one thing you don't want is worse than native input latency and DLSS 3 in performance mode can give you better than native input latency but worse than DLSS 2. In some situations, that would be a good reason to enable it because of the smoother framerates. So while I am sure they will improve the artifacting with later iterations and continue to improve on the input latency, you really need the super fast Tensor Cores in Lovelace that is able to process the rendered frames fast enough. I do think Hardware Unboxed continues to be too AMD bias though such as in that video where they mainly focused on negativity where as Digital Foundry did a better job of presenting the pros and cons of DLSS 3. Tune in when AMD makes a worse version where Hardware Unboxed will praise it. |
True, I couldn't imagine running frame generation on lesser hardware and getting good results. Even more added more input lag would vastly diminish the benefits of having extra frames. So probably the right call on Nvidia's part, at least for now.
Yeah I noticed that as well, that they are heavily AMD biased. Which is quite apparent in their views and testings which usually favor AMD platforms. On the other side.. I don't trust DF so much as they do paid previews for other companies; so there may be partial or even (tin foil hat on) some clauses that enable or not allow they to say certain things. But yeah its probably best to take data from these guys and leave out the rest. And yeah I wouldn't be surprised if AMD has a go at this and make a worse version and HWU praising it, claiming its open source so its better lol.
