NVIDIA ‘Ampere’ 8nm Graphics Cards

[Gerard]

Of course it was a tech demo, you are stating the obvious. However if they have found a way to dramatically improve RT lighting then why would it take years to implement or why would it never appear? It was first announced 3 years ago so I doubt it is now so away from being able to be implemented, especially when Ampere with it's vastly increased RT prowess is released https://research.nvidia.com/publication/2017-07_Spatiotemporal-Variance-Guided-Filtering:

Because they have to convince (pay) game devs to put time into implementing it. Yet another checkbox feature that will only be usable for a minority of players. Look at the amount of engine tech demos that are around that look highly impressive vs the actual end result when a game is built around it.

Unreal has been churning out these kind of demos for years and it's rare that a game ever comes near to that level of fidelity.

As I said it's one thing to do it for a demo which is just a series of non interactive and occasionally interactive scenes, a game has more finite resources to make use of and it has to cater for lower end hardware, whereas a demo can be built around a fixed spec and make the most of that hardware.

 

[Richdog]

Because they have to convince (pay) game devs to put time into implementing it. Yet another checkbox feature that will only be usable for a minority of players. Look at the amount of engine tech demos that are around that look highly impressive vs the actual end result when a game is built around it.

Unreal has been churning out these kind of demos for years and it's rare that a game ever comes near to that level of fidelity.

As I said it's one thing to do it for a demo which is just a series of non interactive and occasionally interactive scenes, a game has more finite resources to make use of and it has to cater for lower end hardware, whereas a demo can be built around a fixed spec and make the most of that hardware.

In that case we can agree to disagree as I see no reason why the benefits of this can't be seen in this generation for the reasons I said above. I will keep an eye on it and post back here when I see any updates.

 

[Gerard]

In that case we can agree to disagree as I see no reason why the benefits of this can't be seen in this generation for the reasons I said above. I will keep an eye on it and post back here when I see any updates.

Will keep an eye out for it .

 

[Calin Banc]

Can you please explain why "it will take years, if ever" to see any in-game implementation?

Depends how easy is to implement and how much it hogs performance. As you can see, the assets of the demo, besides lighting, are quite ... bad. Put all of that in a highly detailed environment like Metro Exodus, RDR 2 or even older ones such as The Witcher 3, ArmA, GTA , etc. and then let's see how it performs.

To be widely implemented it requires an extremely wide install base - think in consoles volumes. UE 5 at least ran on a consoles compared to other demos out there, so the chance to get something like that is higher. The chance to get this type of RT, with proper materials and the likes, is rather low. Except, of course, if is something you can do in a week or so.

 

[nashathedog]

Hmm.. You seem to have confused the P for precision with penetration...

If the shoe fits wear it

You seem to have confused the shoe for the sock, of sorts.

 

[Richdog]

Depends how easy is to implement and how much it hogs performance. As you can see, the assets of the demo, besides lighting, are quite ... bad. Put all of that in a highly detailed environment like Metro Exodus, RDR 2 or even older ones such as The Witcher 3, ArmA, GTA , etc. and then let's see how it performs.

To be widely implemented it requires an extremely wide install base - think in consoles volumes. UE 5 at least ran on a consoles compared to other demos out there, so the chance to get something like that is higher. The chance to get this type of RT, with proper materials and the likes, is rather low. Except, of course, if is something you can do in a week or so.

Yeah the graphics were nothing special for sure, it was all about the lighting physics and it was impressive. The video claims that they have greatly simplified the ability to use it with a unified language so I don't see any major hurdles with next-gen performance being so good.

 

[ANDREW GREEN]

Yeah the graphics were nothing special for sure, it was all about the lighting physics and it was impressive. The video claims that they have greatly simplified the ability to use it with a unified language so I don't see any major hurdles with next-gen performance being so good.

It'll depend on:

A) Is it in next gen consoles
B) Is it easy to add in after the fact, as games are primarily designed for consoles. If the answer is, no and no, you'll see the same none existance of ray tracing we've seen so far.

Also the fact the textures aren't top notch, mean Nvidia can't combine top notch textures with ray tracing even in a demo. Meaning it's still some way off.

 

[Rroff]

Put all of that in a highly detailed environment like Metro Exodus, RDR 2 or even older ones such as The Witcher 3, ArmA, GTA , etc. and then let's see how it performs.

Path tracing techniques tend to scale well with scene complexity as far as things like mesh and texture detail goes - what slows it down is additional bounce/ray counts, caustics, reflections to a degree, etc.

Also the fact the textures aren't top notch, mean Nvidia can't combine top notch textures with ray tracing even in a demo. Meaning it's still some way off.

Doesn't mean they can't - Quake 2 RTX has some fairly decent textures albeit constrained by the old geometry they are pasted on top of.

 

[Phixsator]

Wrong on both accounts. What you and I believe is irrelevant. I am giving you the facts my opinion.

Look at the die size of the Turing RTX cards and the Turing GTX cards. Then look at the die size of the Pascal GTX cards. You might actually gain some understanding.
*

*fixed that for you*

And as usual melmac, you come off as condescending and arrogant. What amuses me to no end is that you think that everyone has to value/see things the same way that you do or they are simply wrong. You completely disregard anything written that doesn't fit your narrative in your pursuit to win an argument and while that is all fine and dandy, you do this while also trying to belittle whoever you are having a disagreement with. I find this behavior in extremely poor taste and frankly don't care much for it. I'm fine with being proven wrong from time to time but it seems most other people can do just that while also keeping a proper tone. My opinion on Nvidia's dies has not changed, so your attempt failed.

 

[scubes]

Still waits patiently for ampere news .

 

[Stu999]

Still waits patiently for ampere news .

+1 and Big Navi / RDNA2 also.

 

[scubes]

E

+1 and Big Navi / RDNA2 also.

Exiting times ahead both out on September too .

 

[TNA]

Still waits patiently for ampere news .

Well we are unlikely to get the cards any earlier than September, even if we get news now. Seems you are looking forward to upgrade that 2080Ti after all

 

[scubes]

Well we are unlikely to get the cards any earlier than September, even if we get news now. Seems you are looking forward to upgrade that 2080Ti after all

depending on price matey if 50% faster than 2080ti and its priced at £1000 yes.

 

[Fyter]

+1 and Big Navi / RDNA2 also.

Same here, willing to go with either. Would it be safe to assume that the RTX 3070 ampere would match if not perform better than the current 2080 ti due to the 7nm fab process? These tends to yield the highest generational leaps in performance for GPUs historically.

 

[Tommy_Here]

Same here, willing to go with either. Would it be safe to assume that the RTX 3070 ampere would match if not perform better than the current 2080 ti due to the 7nm fab process? These tends to yield the highest generational leaps in performance for GPUs historically.

I'm hoping the 3070 offers very close to the Ti performance and exceeding it when overclocked/newer games. Really looking forward to it!

 

[TNA]

I'm hoping the 3070 offers very close to the Ti performance and exceeding it when overclocked/newer games. Really looking forward to it!

I hope so Tommy my boy, I hope so

My main worry right now is they will only slap 8GB of RAM on it. Would like to see 12GB on 3070/3080 and 16GB on 3080Ti.

 

[Troezar]

I'm hoping the 3070 offers very close to the Ti performance and exceeding it when overclocked/newer games. Really looking forward to it!

Going by the AdoredTV analysis it looks like Ampere will be faster than RDNA2 by a decent margin which gives a good chance of the 3070 being 2080Ti performance.

I reckon the idea that big Navi is 2080Ti + 30% seems reasonable and that the 3080Ti is going to be 2080Ti + 50%. If that's the case I'm expecting big Navi at £700ish with the 3080Ti similar to current 2080Ti prices. No reason to drop them if you're still number one. All complete speculation of course

 

[Fyter]

Going by the AdoredTV analysis it looks like Ampere will be faster than RDNA2 by a decent margin which gives a good chance of the 3070 being 2080Ti performance.

I reckon the idea that big Navi is 2080Ti + 30% seems reasonable and that the 3080Ti is going to be 2080Ti + 50%. If that's the case I'm expecting big Navi at £700ish with the 3080Ti similar to current 2080Ti prices. No reason to drop them if you're still number one. All complete speculation of course

As long as there is good cards for the money it will be good. We are in need of a good generational leap. The last gen from both manufactures did not have the best of leaps.

 

[Troezar]

As long as there is good cards for the money it will be good. We are in need of a good generational leap. The last gen from both manufactures did not have the best of leaps.

Agreed I'd have liked something faster than the 5700XT for the money but that's all that was available at the time. Still a bit of an increase on my 1070.