Return to Level1Techs.com

Nvidia RTX 20XX Thread


#82

I’d save it anyways, these GPUs sound good in performance but man their prices are very high.

I wonder what the performance gains are going to be here since it looks like it’s likely we may see an additional performance increase even not factoring in the additional hardware accelerated Raytrace.

The only thing Navi can really possibly do now is do what Polaris did and and sell on the mid tier GPUs and some budget GPUs, hopefully a mid-high tier GPU as well ($400).

AMD is going to have to undercut NVidia hard in price, much like how Bulldozer undercutted Intel CPUs in price later on when AMD needed to slash those prices. Besides, as much as I like powerful GPUs, I like my GPUs to at least be affordable, the RTX GPUs don’t qualify as affordable for me.


#83

AMD yet has a trick card up it’s sleeves. :wink:


#84

Tease :smiley:

Guessing something corny along the lines of

“The magic was inside them ALL along…”


#85

Yeah, in 2020.

I just don’t see AMD beating NVidia in performance anytime soon. It’s unfortunate too since I like the recent work AMD has done with open sourcing the GPU drivers (except for the firmware unfortunately).

I guess if AMD can use Navi to make a $200 GPU with GTX 1070 level performance (or a bit less, that’s kind of a tall order) and a $300 GPU with GTX 1080 level performance and a $400-500 GPU with GTX 1080 Ti level of performance. No idea how any of that is going to work though. Also we need the $100-150 GPUs actually have substantially better performance. The RX 560 and GTX 1050/Ti are not strong cards and are even weak GPUs compared to a RX 570 and GTX 1060 3 GB.

Although personally, I would look between $200 and $400 for a GPU for gaming.


#86

Yes and no, the market for the cards at the RTX price is tiny, like 10% or less or something… So they aren’t building the game around raytracing, so its going to be like physx, it looks great on the demos but in the game its tacked on, because no developer is going to eliminate 90% of the market.

And honestly none of the game demos looked like they had good graphics, raytracing or otherwise. Actually Nvidias own Raytracing DEMOs looked fantastic, the gameworks demos looked like trash.

Convinced gameworks is trash. For trashy developers that don’t want to pay for developing their own game… Its like those fake games that people put on steam using generic game assets…

Gameworks is generic game assets for EA and other shitty publishers…


#87

Jayz: Perspective


So what if its only running at ‘30 fps’ on tomb raider with ray tracing at 1080p, there is more to gaming that fps… Then proceeds to bitch about consumers bitching about price.

Which from a tech tuber on a 6 figure USD salary (and I know guys it doesn’t go far in cali, but that is his own doing, he could just as easily move to the mid west) who gets everything for free that is kinda an absoloute jokes…

So the fact that in two generations the flagship has gone from 600mm sq and 600usd to 750 sq mm and 1200USD, and gives only 30fps at 1080p is no reason to be upset… I would need like 2 of these and a new 1080p monitor for this to be ‘worth it’… And it would have to be a locked down, over priced nvidia G-synch monitor because of the low fps…

So it would cost like $3000 all up… Clearly Raytracing isn’t a consumer level tech yet, look the demo they had with star wars was running at like 22FPS… It looks great, but its like 2 or 3 generations away from getting that in a PC game… On a single flagship card…

And I appreciate that its likely going to be fast, and I don’t need to play it in that mode or with Raytracing at max, but cmon to say this isn’t a legit complaint. Even if performance is 50% higher than the 1080ti and the price is 71% greater…

And normally when we get a refresh on basically the same node and we get bigger dies because the node has matured and defect rates have dropped, normally we get a bigger die and they don’t have to increase the price… Certainly never increase the price more than the performance gain…

So we are just supposed to shut up and put up?


#88

Yeah, because it’s so easy to just pluck your loved ones and uproot your life.


There’s always going to be an early-adopter tax.

I agree with you here. There’s some cost for performance per dollar sheet somewhere. Diminishing returns.


#89

Really don’t get what everyone is upset about.

Only uninformed people would be upset about performance that hasn’t even been published yet.

¯\_(ツ)_/¯


#90

Nvidia is Hype training this.

Consoles and PC are all made up of AMD and GTX cards. RTX will be a single digit % of the market. Sure Nvidia will poor in money to have some games, what 25 is it some with only the new anti-aliasing. I think it was only 7-8 with new and old engines to support RTX fully.

This custom Nvidia only ray tracing could go the way of PhysX and just not get put into games. Its just a little more bling.


#91

Everyone please keep in mind that GP104-400-A1 (GTX 1080) is a 2014-2015 design that launched May 2016.

GP102-350-K1-A1 (GTX 1080 Ti) is a 2014-2015 design that was purposely held back over a year and launched on an improved node in March 2017.

What does everyone think they’ve been designing in the past 3 years with all the money everyone’s been throwing at Nvidia :wink:


#92

This. I really thing they’re just trying to engineer another “talking point” divide between themselves, Intel and AMD. I think they’re likely afraid of Intel taking a significant chunk of the market and AMD possibly bouncing back and doing the same. This is especially worrisome since AMD have been competitive on the CPU side for the first time in nearly a decade and finally won back a nice chunk of Intel’s market share.


#93

A bit of a Historical point of reference


#94

One video that sums up everything anyone not behind and NDA/embargo knows right now:


#95

Raytacing features are a standard by now in both DirectX 12 Raytacing API and Vulkan API.


#96

My opinion here should be taken with a grain of salt, as I do not believe this card is marketed toward someone like me.

While we wait on benchmarks, I do think it is safe to say that this card will have a performance gain over the previous gen. If it doesn’t, Nvidia really screwed up.

That being said, I think regardless of performance, Nvidia has priced many people, including myself, out of the market. My speculation is that, depending on price, this may push some people to buy less capable cards to save some money.


#97

Well we did have HBM but still, AMD didn’t charge $1200 for an HBM nor HBM2 GPU such as the Fury and Vega and those were expensive considering they fell short in performance.


#98

Nothing. Just like Intel did when they were the ruler of the market, they did nothing. Ya, 3 to 6% here or there and more cores to keep the server guys happy.
Nvidia, if they can count to 10, will do the same.


#99

Well Intel is definitely paying the price though. 10 nm is STILL a dumpster fire, they are losing BADLY in price/performance to AMD and even just losing in performance in general. Intel CPUs have more vulnerabilities so when they get patched they are going to lose performance.

Intel painted themselves in a terrible corner not expecting AMD to catch up with Ryzen. Given that NVidia responded to Vega 64 with a GTX 1070 Ti (even though a GTX 1080 was FINE), idk if NVidia will be as foolish as Intel when NVidia can see that Intel is not doing so well now compared to before.


#100

Just as ‘difficult’ to move between states and other suburbs. Anyway it’s up to him, the difference is probably tens of thousands in higher costs of living, doing business, and taxes… Anyway its his money, and its his decision…

Frankly I can’t believe anyone would pay a premium to live in that big government hellhole, but it depends what you value in life… There are people in Eastern europe who miss ‘the good old days’ of communism…

Figures has been publicized, the performance with raytracing in their demos is public, they posted graphs, also people counted frames. They also published graphs of the ‘performance’ of the RTX2080.

To pretend there is no performance figures is missleading. To claim ignorance about why people might be upset to go from paying $600 for a 600mm part to $1200 for a 750mm part in two generations is just ridiculous…

And frankly the reason to be upset over the pricing is no different from people being upset that a company making aids medicine that the government has issued a license (patent) banning people from producing anything that competes just jacks the price up like 10,000% over a few years…

Its the same thing… They are ripping people off… That should be like a $700-800 part. And if 7nm wasn’t around the corner, and if they supported adaptive refresh, if they were on point with Mgpu, and they were fairly priced then maybe I would buy two of them…

But I shouldn’t have to pay more for a poorly supported product because government monopoly…


#101

Here is actual performance gains between gens using TPU data:

Computed chart:

In order to have a median performance increase per dollar between the 1080ti->RTX 2080ti then the RTX2080ti would need to be 130% faster (2.3x the speed of the 1080ti). In order to be above the worst release in >decade in terms of performance gains per dollar that bar is >105%…

(This is at a 1200usd msrp.)

Interestingly and hear me out on this, the rumored launch price of the RTX2080ti was 800USD, the median price per sq mm puts the RTX2080ti at 815USD. Coincidence?

Overall its shaping up to be a poor showing in performance gains for the money… Worst in a long time at these prices…