Hardware & Technical 2080 Teased

[[VR] optimal_909]

So, the 2080 (see hints in the video) has been teased, full reveal expected Aug 20.
I'm so looking forward to it! I guess it will be sold out quickly at a higher price, so I will try to preorder, possibly a Founders Edition.

[video=youtube;F7ElMOiAOBI]https://www.youtube.com/watch?v=F7ElMOiAOBI[/video]

 

[Caramel Clown]

Kinda not too excited...looks like we will only see leap in this new ray tracing fab, which will probably not show up in mainstream gaming for years. My 1080ti is safe for at least 2 more years.

 

[[VR] optimal_909]

Simply put: we don't know yet.
I think it will deliver the usual jump in performance.

 

[Patrick_68000]

I can not imagine the price of this jewel

 

[Gun Star]

Im good with my 1070 OC for at least 2-4 years. I'll upgrade when they drop in price and my PC cant run 60 FPS on new games. The only games that are in the works I plan on getting in the near future is Cyber Punk, ES VI, and Starfield

 

[Iskariot]

So, the 2080 (see hints in the video) has been teased, full reveal expected Aug 20.
I'm so looking forward to it! I guess it will be sold out quickly at a higher price, so I will try to preorder, possibly a Founders Edition.

I am only looking forward to this, because it will make the 1080 cards cheaper.
For now my GTX 980 still does all I need.

 

[Morbad]

I think it will deliver the usual jump in performance.

Slightly less than usual, most likely.

2080 or whatever they call it will probably be ~10% faster than a 1080 Ti as an optimistic estimate.

DX12 raytracing stuff will be considerably faster, but that will remain niche for another generation or two of hardware and by the time heavy raytracing is mainstream a 2080 will be old and slow.

I can not imagine the price of this jewel

599 USD seems probable for the MSRP of the reference cards.

 

[Patrick_68000]

599 USD seems probable for the MSRP of the reference cards.

Looks a little light, but hey we'll see

 

[Morbad]

Looks a little light, but hey we'll see

I suppose they could try for the $699 MSRP that the initially released the 1080 Founders Edition at, but since they cut the price just a few months later and this gen is almost certainly not going to give the same advantage over prior gen parts, I'm thinking $599 MSRP is at least as likely.

Then again, new cooler and the PCB looked pretty robust...

 

[[VR] optimal_909]

this gen is almost certainly not going to give the same advantage over prior gen parts

Where do you get that info? Because during the presentation, Jensen highlighted that the computing advantage will be there in 'legacy' rendering as well. I.e. day 1 advantage, and a significant one.

 

[Jon flint]

Well, if it can beat out an OC'd 1080ti by 15% or more it'll be a day 1 purchase for me. Here's hoping

 

[GJ51]

Certainly overkill for ED unless you use VR, but 1080(Ti) pricing seems to be falling to reasonable levels. 1080 series could be a bargain purchase after the new cards are released if there's sufficient price difference.

 

[Morbad]

Where do you get that info?

Extrapolation from the Quadro RTX 5000 specs.

If the GeForce RTX 2080 is a fully enabled GT104 with modestly higher clocks than the Quadro part, it will have very similar shader performance, moderately worse pixel/texture fill rate, and slightly worse memory bandwidth than a GTX 1080 Ti. Assuming there are modest enhancements that allow it to do more work per-clock per-functional unit, it should be faster than a 1080 Ti, but not by much.

Because during the presentation, Jensen highlighted that the computing advantage will be there in 'legacy' rendering as well. I.e. day 1 advantage, and a significant one.

He's talking architectures, specifically the big die parts of each. That's GP100/102 vs. GT100/102, where the new part is 40-50% faster than the old one. The Quadro RTX 8000/6000 will be much faster than the P6000, and the RTX Titan or RTX 2080 Ti will be much faster than the Titan Xp or 1080 Ti.

However, the RTX 2080 is a GT104 part...it's to Turing as the non-Ti 1080 was to pascal.

NVIDIA has zero competition currently and has spent a large portion of their transistor budget on Tensor and RT cores that are going to do exactly nothing in 99.9% of games upon the release of the RTX 2080, so the bar they have to hit is not high. If it's any faster than the 1080 Ti in 'legacy' rendering, they'll be successful.

This is the perfect time for NVIDIA to push raytracing. If they had to face a credible competitor architecture from AMD (and/or Intel soon enough) they'd almost certainly have fallen behind in such legacy apps, but with no such competitor, they barely have to improve over their own extant parts with the initial launch.

The GTX 1080 was quite a bit faster than the 980 Ti, but Pascal was a straight upgrade over Maxwell 2, had much higher clocks than previous parts, and didn't spend any transistors on new features that couldn't be leveraged in existing games.

The GTX 980 was often only slightly faster than the GTX 780 Ti, and I think that sort of improvement is a more reasonable expectation with the RTX 2080 over the 1080 Ti, at least for legacy apps. Newer DX12 DRX apps, especially ones using NVIDIA's GameWorks RTX will either be much faster on Turning or flatly won't be able to use all of their features on older archs...however, broad adoption, if it ever happens, is likely to take some time.

 

[[VR] optimal_909]

Extrapolation from the Quadro RTX 5000 specs.

If the GeForce RTX 2080 is a fully enabled GT104 with modestly higher clocks than the Quadro part, it will have very similar shader performance, moderately worse pixel/texture fill rate, and slightly worse memory bandwidth than a GTX 1080 Ti. Assuming there are modest enhancements that allow it to do more work per-clock per-functional unit, it should be faster than a 1080 Ti, but not by much.



He's talking architectures, specifically the big die parts of each. That's GP100/102 vs. GT100/102, where the new part is 40-50% faster than the old one. The Quadro RTX 8000/6000 will be much faster than the P6000, and the RTX Titan or RTX 2080 Ti will be much faster than the Titan Xp or 1080 Ti.

However, the RTX 2080 is a GT104 part...it's to Turing as the non-Ti 1080 was to pascal.

NVIDIA has zero competition currently and has spent a large portion of their transistor budget on Tensor and RT cores that are going to do exactly nothing in 99.9% of games upon the release of the RTX 2080, so the bar they have to hit is not high. If it's any faster than the 1080 Ti in 'legacy' rendering, they'll be successful.

This is the perfect time for NVIDIA to push raytracing. If they had to face a credible competitor architecture from AMD (and/or Intel soon enough) they'd almost certainly have fallen behind in such legacy apps, but with no such competitor, they barely have to improve over their own extant parts with the initial launch.

The GTX 1080 was quite a bit faster than the 980 Ti, but Pascal was a straight upgrade over Maxwell 2, had much higher clocks than previous parts, and didn't spend any transistors on new features that couldn't be leveraged in existing games.

The GTX 980 was often only slightly faster than the GTX 780 Ti, and I think that sort of improvement is a more reasonable expectation with the RTX 2080 over the 1080 Ti, at least for legacy apps. Newer DX12 DRX apps, especially ones using NVIDIA's GameWorks RTX will either be much faster on Turning or flatly won't be able to use all of their features on older archs...however, broad adoption, if it ever happens, is likely to take some time.

Thanks for the great answer.
I have a 1060 6gb which I'm planning to upgrade as I'm also planning to get into VR.

I will see the prices and availability, if the 1080ti comes it a much better value, I will get that. In any case, it feels like bad timing to upgrade now - but now I have the chance to do so.

 

[Flash_Baggins]

Well my graphics card has been playing up recently so Im all for getting myself a 1080 at a cheaper price!

 

[[VR] optimal_909]

Well, I was on crossroads this spring.
I had my whole rig upgraded last year, rocking an i7700k. Then I thought I'll upgrade from 960 4gb, but then again the GPU shortage/pricing hit me. I ended up with a 1060 6gb FE, even though I was rather aiming for a 1070ti.

As I want to buy a Rift, I need a better card... and here is hope, that nVidia will not have the balls to launch a 2080 after 2 years that only barely beats the 1080ti.

 

[VR StarLightPL]

If the pascal launch had taught me anything, it's "Wait for the Ti version". I bought a 1070, then 1080Ti come out and price-obsoleted everything. If mining cryptocraze didn't happen in the meantime, we would all be treated to cheap 1070s right now. Luckily I managed to score a 1080Ti and sell my 1070 recently. I'd advise waiting a bit, unless you are in dire need. Also, buying a blower? Why? The non-reference designs are usually better at cooling and noise levels.

EDIT: you should be okay in VR with a 1060 6GB. I would know, I was able to run Elite on DK2 and GTX 670. Granted, much has happened since then ;-) But you can enjoy VR with 1060 too, especially if it's the one with more VRAM.

 

[[VR] optimal_909]

Yes, it's the 6gb version.
I think I'll wait until the reveal, and see. If the performance gain will be substantial, I'll probably pull the trigger, if not, then I might wait. But I could wait forever as there will be always the 'next big thing'. Some say this 12nm generation won't last long as 7nm production is imminent.

Edit: The blower noise isn't that bad, i.e. this issue is 'overblown'.

 

[VR StarLightPL]

Yes, it's the 6gb version.
I think I'll wait until the reveal, and see. If the performance gain will be substantial, I'll probably pull the trigger, if not, then I might wait. But I could wait forever as there will be always the 'next big thing'. Some say this 12nm generation won't last long as 7nm production is imminent.

Edit: The blower noise isn't that bad, i.e. this issue is 'overblown'.

Yeah, but it always can be better, too. I had a Palit Gamerock 1070, mainly because I build my rig to be a silent PC hooked up to the TV in the living room. This thing had a substantial cooler on it. It was as quiet as a mouse despite operating on 2.012GHz OC'd and temps around 60-somethings Celsius. Now I have a Gigabyte 1080Ti, and this thing sounds like a vacuum cleaner. I picked it up as a bargain I couldn't refuse, just to tide me over till the next Ti version comes out. But really, your card isn't bad at all. It is more than enough for Elite. You will get some framedrops in the stations in VR but it's not that bad, and with Oculus ASW/ATW you probably won't notice it much, or at all. I'd at least advise to wait for non-reference systems to come out, they usually blow (pardon the pun) the founder's editions out of the water ;-)

Of course step from a 1060 to next gen will be a shock. Between my 1070 and my current 1080Ti there is a whopping 70% performance difference and they're both Pascal cards. The Ti is a beast, and as the mining craze is hopefully over, I expect their price to drop to a very sensible level. We'll see. Thus, you also should wait for initial comparisons of the next gen to the current consumer pinnacle which is 1080Ti (I consider TitanX outside of sane consumer market). It might turn out that the next gen is not really that much faster (I expect 30% at most), and the revolution will probably happen with next gen Ti, as it did previously.

My advice - pull the trigger on the Oculus now, use the happy days of VR honeymoon to wait for new card, then decide to pull the trigger or not
PS: Third camera for Oculus is recommended. I have a Vive and know joys of roomscale ;-) Think Skyrim VR for example

 

[Patrick_68000]

I suppose they could try for the $699 MSRP that the initially released the 1080 Founders Edition at, but since they cut the price just a few months later and this gen is almost certainly not going to give the same advantage over prior gen parts, I'm thinking $599 MSRP is at least as likely.

Then again, new cooler and the PCB looked pretty robust...

However I read that the interest of nvidia is to sell a lot of 10xx when Turing will arrive because the stocks are still important