Nvidia ,Rtx2080ti,2080,2070, information thread. Reviews and prices September 14.

[TestKing123]

So when the hell do we see actual reviews with benchmarks with actual games? We have to wait 30 days for September 20th? Is that why they have "pre orders" up now, to get people's monies now without actual reviews first? You'd think there would be at least some samples given out for review purposes.

 

[sze5003]

I find it odd that every release they showed performance overall relative to the new cards. But this release they just showed the RTX operations.

I'd wait until we see how these do under the current games we have out now.

 

[beginner99]

When measured with NVs new RTX Share Price OPS benchmark!

When predicted with the new AI based on the 10 Gigarays!

 

[IntelUser2000]

I'll say this.

The pricing is the problem. The cards themselves are nice. The RT hardware is there so it can accelerate extra visual features without the performance penalty(maybe small).

 

[ginfest]

Anyone have an active link to a preorder? Must have one of these SAAP!

Nevermind, ordered one from NV

 

[LTC8K6]

New SM can do INT and FP ops concurrently.

Is that a big deal?

 

[LTC8K6]

01:57PM EDT - 'Designed for crazy amounts of overclocking'

01:57PM EDT - 'And it's just so quiet' ... 1/5 audio levels of 1080 Ti at max overclock

Heh...

 

[crisium]

Even the blower 2080 Ti by Asus is $1,210 on Newegg. $210 above "MSRP" and even $10 above FE.

Totally available "from $999".

What a load of false advertising nonsense.

 

[Headfoot]

Stoked for ray tracing. Despite the hype factor, it really is the holy grail of computer graphics and we've been steadily marching towards that reality. Since I have a 1080 Ti I will probably skip the first generation of RT hardware but its a chicken and the egg problem. The sooner they get RT hardware into the world the sooner devs who aren't tightly partnered will start baking it in.

This is relatively 'AMD like' of nVidia, pushing new features that require dev and not progressing a ton on today's games (presumably - there could be magic they didnt mention but I doubt it). Smell's like AMD's dx12 move, and AMD's earlier tesselation move. And of course the so far failed-on-PC move to the new Vega pipeline stuff.

So I think the RT stuff will be awesome, especially when its more deeply built in vs being an add-on effect in a few games. But it also presents an opportunity for competitors to come in and play the old nVidia strategy - double down on what make today's games go fast because day 1 benchmarks stick around forever. Given how long the less dramatic switch to DX12 has taken, I am not convinced RT adoption will be much faster. We're looking at a few years before it approaches anything that looks like mainstream

 

[antihelten]

Whatever all that means...

It basically means the same thing. It's an extension on Nvidias work on MRS. Previously with MRS, Nvidia could vary the frame buffer resolution of individual viewports, now they can also vary the shading resolution.

Again this is only really useful for foveated rendering (or VR headsets with lens distortion), due to the drop in IQ.

 

[SPBHM]

"ray tracing" stuff is very interesting but it's for the future and probably the next gen is the one that will be good enough to fully take advantage, also without the weight of a main console and without competitors being compatible with the same it's difficult to see it becoming a normal thing, so... we will see... still, exciting products, but pricing looks very wrong, specially if it's not massively faster in current games.

 

[LTC8K6]

IIRC from an earlier article all of NV recent cards can take advantage of hybrid graphics/ray tracing in some form.

 

[alcoholbob]

01:57PM EDT - 'Designed for crazy amounts of overclocking'

01:57PM EDT - 'And it's just so quiet' ... 1/5 audio levels of 1080 Ti at max overclock

Heh...

That thing better still fit the 10.5" form factor of the regular FE cards, otherwise there's no point buying it compared to an aftermarket cooler.

 

[SPBHM]

IIRC from an earlier article all of NV recent cards can take advantage of hybrid graphics/ray tracing in some form.

from the presentation you can see that they are VERY slow at it, so kind of useless, you will need a new RTX card to enable those effects with decent performance I would think.

 

[Headfoot]

Yeah I don't see RT really taking off into mainstream until at least one of the main consoles supports some form of it

 

[LTC8K6]

It basically means the same thing. It's an extension on Nvidias work on MRS. Previously with MRS, Nvidia could vary the frame buffer resolution of individual viewports, now they can also vary the shading resolution.

Again this is only really useful for foveated rendering (or VR headsets with lens distortion), due to the drop in IQ.

What about the "Concurrent FP and INT execution"?

 

[IntelUser2000]

01:57PM EDT - 'Designed for crazy amounts of overclocking'

They can say anything, but its not true until it is. They claimed the same with 1080.

 

[LTC8K6]

from the presentation you can see that they are VERY slow at it, so kind of useless, you will need a new RTX card to enable those effects with decent performance I would think.

https://www.engadget.com/2018/08/15/nvidia-turing-quadro-8k-video-playback/

That's the article I was thinking of. It's mostly about 8K and says it's code rather than hardware.

 

[Ottonomous]

Its silly math but
2070 - 45/6 = 7.5
2080 - 60\8 = 7.5
2080ti - 78\10 = 7.8
7.5 + 7.5 + 7.8 = 22.8\3 = 7.6

The Gigaray/s perf seems to scale linearly, does that mean we can factor out its TFLOP performance from the rest of the logic/processing package (excuse my ignorance) because the core counts are a bit off

edit: nvm, its RTX-OPS and I can't find a clue to its TFLOP performance, but since its linear it must mean the logic is totally independent right?

 

[antihelten]

What about the "Concurrent FP and INT execution"?

In theory, I suppose it could result in noticeable performance improvements, but only if developers significantly rewrote their games to take advantage of it, since I don't really think most games are limited by GPU INT throughput.

Again there's probably a reason why Jensen skipped over these things so quickly.

 

[beginner99]

Totally available "from $999".

What a load of false advertising nonsense.

have been doing this since the first FE release. It's no surprise at all and just one of the 1000 reasons NV will never see a dime from me till they have some serious management and strategy change. It's amazing how many gamers fall for this crap. I get a company paying $100k for their dx servers. You need 1 good model and you get the investment back >100 times. Why would I pay $1200 for a card that was <$400 only couple years ago? Because half it's die is useless for my gaming? just omg at who falls for this crap.

 

[Headfoot]

have been doing this since the first FE release. It's no surprise at all and just one of the 1000 reasons NV will never see a dime from me till they have some serious management and strategy change. It's amazing how many gamers fall for this crap. I get a company paying $100k for their dx servers. You need 1 good model and you get the investment back >100 times. Why would I pay $1200 for a card that was <$400 only couple years ago? Because half it's die is useless for my gaming? just omg at who falls for this crap.

Not so much "falling for it" as "there is literally no other choice that performs at this level"

 

[LTC8K6]

They can say anything, but its not true until it is. They claimed the same with 1080.

2080ti non-fe listed boost clock 1542
1080ti listed boost clock 1582

1080ti topped out at ~2.1-2.2
The non-fe 2080ti would have to overclock better just to tie that, since it's starting off slower.

 

[IntelUser2000]

Why would I pay $1200 for a card that was <$400 only couple years ago? Because half it's die is useless for my gaming? just omg at who falls for this crap.

If you can convince most to buy at the new price level, its a direct revenue boost.

No doubt why they've been posting record revenues. That's why prices are no longer about production costs and die sizes anymore.

LTC8K6:

I was talking about the regular 1080. The 1080 Ti reaches similar probably because they are using the same uarch and same circuitry, with the base clocks limited by TDP.

 

[sze5003]

I got benchmarks guys:

https://imgur.com/L7icF7t