Cyberpunk 2077: PC version with raytracing, first screenshots

10 months away? Have a feeling Nvidia will time their Ampere release with it.

 

This.

Well, That's nice gameplay w/o DX-R. I hope they can continue having same IQ on release date with DX-R disabled. (except some bugs)
But it should always be remembered that this is not visual showcase. (As if DX-R could be visual showcase any time soon...) This is game with story where players can do a lot of stuff.

 

I get the feeling that is exactly what will happen. Game is going to be epic, that's for sure. I have no doubt about that.

 

I somewhat doubt they'll release a new product generation in April for a single game. And if they do, I doubt it will be that much faster in anything than current Turing lineup. Although I'd like it tbh, I doubt it.

 

I'm just wrapping up Witcher 3, so I'm really looking forward to playing this game in 2022 when affordable hardware catches up to it.

 

I would not be surprised if this game will be bundled with Ampere.

 

I don't think it will be faster because the cost per transistor is still rather limiting on 7nm but I think they'll probably be shipping a new architecture early next year. They've stated that they were waiting for Samsung 7nm EUV - which just hit mass production now and typically its 8-12 months after a process hits mass production that Nvidia starts shipping GPUs on it. It also wouldn't surprise me if their next architecture used chiplets to some degree - they've been experimenting with scalable inference chips with up to 36 chiplets. It would be interesting if they broke out the tensor cores on it's own chiplet in an effort to improve yields. Idk if they are at that level yet though.

As for the RT performance - a large part of the BVH optimization is algorithmic. It's possible that the next architecture has the same number of RT cores but the performance per core is increased due to new techniques. That's not to mention that game developers are getting better in general at optimizing for it. A large part of the overhead seems to be syncing the raster/BVH environment and starting the BVH as early as possible in the frame. I think as developers get their hands on DXR, especially now that next generation consoles are supporting it, the performance will just naturally improve.

 

I think they won't increase core count, but focus more on rt core count. Gpu is already fast enough by normal raster.

 

Well then they'd introduce a new process (7nm EUV), and a new architecture? Like I said, I doubt it somewhat (they weren't able to milk the current generation enough I thought), but if you are right I'm not that unhappy. And I know there's a lot of people out there wanting higher 4K performance without RTX, and RTX will only really kick off once it performs better. It would just feel so un-Nvidia to do that... although I've been wrong in the past as well about that.

edit: On second thought, it does make sense. Milked first Turing iteration due to higher prices, throw 2nd one rather "rapidly" to get RTX to have a grip on the ray tracing market and once again get people to buy new cards when they were already holding back on Turing.

 

I reckon they will include dx11... and in that case just using the witcher 3 sli profile ought to work.

If not then i am going to very sad... 4k with a single 1080... not going to be a good experience.

 

i just want to know if it will support NVLink because I'm ready to buy another GPU if I have to

 

It will but you'll need to enable DLSS so really it will be running at 720p

 

The Nvidia 600 series comes to mind, honestly - Kepler was a pretty radical architecture shift and they went from 40nm to 28nm all in the same jump. I think as the hardware development tools improve, as Nvidia's R&D cost increases and they can do more engineer sampling, it becomes less of an issue to transition to a smaller node with an architecture change.

 

Nvidia needs 2 sku+ to meet the mid/high-end gaming market demands.

A lot of people don't want RTX, so it's obvious that GTX will be the cheaper option. Meanwhile RTX will be the new high-end with all the bells and whistles.

For me there are two questions;

1) Will Nvidia release both a GTX3080 and RTX3080?
2) If so, will they actually have the same performance, minus RTX stuff?

I would prefer for them to make two separate sku, rather than rely on disabling the RTX portions of the chip for GTX.

What's interesting to me is how in general, on retailer sites; the specs of RTX2080ti to RTX2060 do not list the number of RT cores as a spec yet. There's only the mention of "Real-Time Ray Tracing". The number of CUDA cores, ram and gpu clock-speeds are still the most well-known and important specs that people care about. I wonder how Nvidia will market RTX performance to the public with meaningful numbers/performance that consumers can understand, especially in comparison to the previous (Turing) generation of cards.

 

80Gb? i'll even remove windows to make room for CP2077 if necessary

 

Why didn't you just buy a better gpu in the first place then ?

 

As the topic is around gpu in this thread lately, i personallywonder what will be the impact of the SUPER line-up release on Ampere release.
My guess being a slight 'delay' compared to what we expected?

 

While I do believe your educated judgement on this, I just thought they'd try to keep Turing around to earn more with it. On second thought I have to agree Nvidia might have already made the money they wanted, upping the pace now for more steady revenue. I'm not worried that they couldn't, I just thought it might leave them with too little income from Turing to feed their targets since Jensen was under fire for low sales numbers.
From a technological point of view I honestly have no doubt Nvidia could blast us all away with performance, RTX or not, just that they didn't see the need or chance to make lots of money if they drop Turing after such a short time.

Yes I'm wondering about similar things. While I don't expect them to do more than they already did (seperating RTX as top line and GTX like 1660TI as mid line), I don't think they will produce an extra SKU like a GTX3080. I could be off, but I think they'd more likely use the fab capacities to trie and out as many big, full chips as possible, and if they can't they laser cut it down. Which brings us the RTX3070, TI maybe there, 3060 etc. but I somehow doubt they'd go through the trouble of effectively doubling their lineup. But as I said, I've been wrong in the past as well, they might just do something like this.

 

Yeah, probably means that it will be another year before we see ampere... at this rate i am going to have had my 1080's far longer than any other gpu's. Have already had them for 3 years.

 

Value purchase then?