Geforce GTX1180/2080 Speculation thread

Vega · 22 Aug 2018 at 19:10

EastCoastHandle said:
Take another look. That's not what its saying when comparing 2080 vs 1080. It's more like around 1.5. When tensor cores are used vs no tensor cores they are saying it's 2.0x.

And look at the HDR results LOL. Besides, the community said time and time again to take manufacture performance charts with a grain of salt.

Uhh, 1.5 is 50% increase. Not sure what you are on about. I've never found the NVIDIA performance slides that far off in the past.

EastCoastHandle · 22 Aug 2018 at 19:13

Vega said:
Uhh, 1.5 is 50% increase. Not sure what you are on about. I've never found the NVIDIA performance slides that far off in the past.

There are the other suite of game that weren't mentioned. This was cherry picked best case scenerio.
Don't be "marketed".

Vega · 22 Aug 2018 at 19:15

Ya ten games is cherry picked? How many games you want on one chart not to be cherry picking, 50? :rolleyes:

LoadsaMoney · 22 Aug 2018 at 19:17

A'right guys, hows it going.

EastCoastHandle · 22 Aug 2018 at 19:17

Vega said:
Ya ten games is cherry picked? How many games you want on one chart not to be cherry picking, 50?

There are 6 games where effects won't be available to them. Leaving 4 that could be on equal terms. So yeah, get the other games in the suite that won't be touched by tensor effects
:rolleyes:

kevman · 22 Aug 2018 at 19:18

EastCoastHandle said:
Take another look. That's not what its saying when comparing 2080 vs 1080. It's more like around 1.5. When tensor cores are used vs no tensor cores they are saying it's 2.0x. Far from the 6x claimed in the keynote. hmmm...

And look at the HDR results LOL. Besides, the community said time and time again to take manufacture performance charts with a grain of salt.

So the tensor cores do some sort of new fangled AA right?

Vega · 22 Aug 2018 at 19:19

Let me guess this slide is also no good in your universe?

Gerard · 22 Aug 2018 at 19:21

Invisible slide

chaparral · 22 Aug 2018 at 19:21

Gerard said:
Remember the 1080 presentation? Ooh here it is running at 65c and over 2ghz!! Was a week or so later people found out it was overclocked and the cooler was running full tilt, something else they conveniently "forgot" to mention.

I can't remember it all but years ago there was a gtx 480 they showed made out of wood and had wood screws or something like that...

found about it

WHAT DO YOU DO when you have a major conference planned to introduce a card, but you don’t have a card? You fake it, and Nvidia did just that.

https://semiaccurate.com/2009/10/01/nvidia-fakes-fermi-boards-gtc/

LoadsaMoney · 22 Aug 2018 at 19:21

Here comes the turnaround, lmfao!

JediFragger · 22 Aug 2018 at 19:22

LoadsaMoney said:
A'right guys, hows it going.

*Triggered*

N19h7m4r3 · 22 Aug 2018 at 19:23

LoadsaMoney said:
Here comes the turnaround, lmfao!

If the 2080Ti smooshed my V, I'll let Jensen smash.

EastCoastHandle · 22 Aug 2018 at 19:24

kevman said:
So the tensor cores do some sort of new fangled AA right?

Yup, they are a jack of trades.
Nvidia Clarifies - RTX in games doesn't mean Ray Tracing

Comparing it to a card that doesn't have that capability is meaningless.
People want to know what the performance will be for the games they are playing now. Not games that will involve RTX use.

Gerard · 22 Aug 2018 at 19:24

chaparral said:
I can't remember it all but years ago there was a gtx 480 they showed made out of wood and had wood screws or something like that...

Yeah the Fermi quadro card that had PCB traces that went nowhere and seemed to be made out of a random pcb they chopped up so they would have something to show. Why they let the press get close enough to it to take detailed pics is mind boggling, as if none of them were tuned in enough to know it was a dummy card.

Tom|Nbk · 22 Aug 2018 at 19:29

As said if it smashes my ti then it's still a done deal. Just let reviewers get their hands on them ffs

CAT-THE-FIFTH · 22 Aug 2018 at 19:31

EastCoastHandle said:
Take another look. That's not what its saying when comparing 2080 vs 1080. It's more like around 1.5. When tensor cores are used vs no tensor cores they are saying it's 2.0x. Far from the 6x claimed in the keynote. hmmm...

And look at the HDR results LOL. Besides, the community said time and time again to take manufacture performance charts with a grain of salt.

Also look very carefully at what is being tested.

Loads of HDR titles. The GTX1080 unlike the Vega64 drops 10% to 15% of its performance in HDR:

https://www.computerbase.de/2018-07/hdr-benchmarks-amd-radeon-nvidia-geforce/2/

What is more interesting is Wolfenstein 2 which is DX12/Vulkan which means that is doing what we expect with Nvidia cards.

Expect more DX12/Vulkan games in the next few years now.

Still if we say 40% overall,its still means the RTX2080 is 50% more expensive than a GTX1080.

Not really uber value.

TNA · 22 Aug 2018 at 19:35

LoadsaMoney said:
A'right guys, hows it going.

Hahahahahaha. Well done sir

KentMan · 22 Aug 2018 at 19:37

Vega said:
Let me guess this slide is also no good in your universe?

At what settings though? Simply saying 4k HDR 60hz is still not painting the full picture, that could be with tons of stuff turned off

Rroff · 22 Aug 2018 at 19:38

Vega said:
So that looks like around ~45% average FPS increase, not too bad. The 2080Ti over 1080Ti should be about the same. So that would put the 2080Ti around 10% faster than the Titan V if it scales linearly.

The approach to VRAM on the 2080 probably suggest atleast a 40% performance uplift over the 1080 anyhow - otherwise they've gone to a lot of expense needlessly if the ROPs, etc. could perform with lesser support.

FloatingPoint · 22 Aug 2018 at 19:38

Anyone know why it’s called Turing compared to Maxwell, Pascal, Fermi, Volta? Is it to do with Cryptography?