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

[LTC8K6]

Since NV said it on their blog, I am now expecting a massive jump in performance outside of ray tracing.

 

[tamz_msc]

https://blogs.nvidia.com/blog/2018/08/20/geforce-rtx-real-time-ray-tracing/

They repeat the same thing a bit further down:

GeForce RTX 20-series are the first GPUs that will play AAA games at 4K 60 FPS HDR.

2X compared to what? I thought that the Titan Xp or 1080 Ti already does 4K 60 FPS HDR in supported titles, like Destiny 2?

 

[sze5003]

Yea I wonder it's 2x to what? And if it's true than that means it will blow the 1080ti away. If true I'll finally be able to get a 4k monitor. But I still smell marketing talk.

 

[BenSkywalker]

I was absolutely right in assuming that you have no idea what you're talking about.

Alright, you at least diamond league in Overwatch? Get your team together and let's go.

 

[Malogeek]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles."
https://blogs.nvidia.com/blog/2018/08/20/geforce-rtx-real-time-ray-tracing/

Translation: It will do 20fps instead of 10fps in corner cases in one game at 4K.

 

[SMOGZINN]

You don't invite people over to your house for a party you are throwing and then make sure to announce to everyone that your kid accidentally sneezed in the punch bowl right before everyone showed up. You ignore it and move on baby.

The trouble is he can't hide the snot in this punch bowl. Eventually we are all going to see the benchmarks, and then if he was deceiving us to try to sell pre-orders Nvidia is going to take a huge hit to their reputation and probably get a bunch of those cards returned.

 

[crisium]

RT needs massive die space. If RT is the future of games, what is the fate of GTX 1060, 1050 Ti, 1030 successors? Are they too small to get ray tracing cores and keep them cost effective? Will they get them eventually? 1060 class at 7nm, 1050 Ti class at shrink, 1030 class at shrink after that?

What does this mean for the future of lighting in games? More than just legacy cards keeping legacy settings, we are poised for current cards not having RTX cores (2060 and under rumour), and possibly even subsequent generations.

Ultra becomes RT, and High and below are resterization? So from now on and indefinitely in the future, all games will require programing for both types of light sources? Devs have their work cut out for them?

Back in the day, it was pretty common for a new generation to have a new DX version (Radeon 9700 with DX 9.0) or new SM (Nvidia 6000 series with SM 3.0). But these didn't require massive new die space. Within a single generation, sometimes two (lower end has always been often refreshed), basically all cards supported the tech. And even then, it would take years for a game to absolutely require the latest DX or SM. Now we have hardware RT that takes a lot of space. We have consoles that don't have it now, and may not even for next-gen (who knows).

I believe this will be an even slower crawl of adoption than previous advancements. Nvidia will work with devs to push them hard, of course. In the end game many years from now RT should be better than if we continued without it, but man, right now it's hard to get excited when looking at these prices.

 

[tamz_msc]

According to this video, a Titan X (Pascal) does 4K Ultra HDR at 60FPS in all the tested games except for Assassin's Creed Origins. So, it is not clear what they're comparing against Turing.

 

[sze5003]

The trouble is he can't hide the snot in this punch bowl. Eventually we are all going to see the benchmarks, and then if he was deceiving us to try to sell pre-orders Nvidia is going to take a huge hit to their reputation and probably get a bunch of those cards returned.

So many people are gonna be so mad, yet so many who waited so happy. But I think the people that pre ordered won't care. You could probably sell them crap on a stick and they would have pre-ordered anyway since people that jump on things like this don't care about money vs value.

 

[rainy]

Since NV said it on their blog, I am now expecting a massive jump in performance outside of ray tracing.

Yes, because Nvidia is famous for saying only the truth./s

 

[moonbogg]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles." I don't know what this means, but what they want me to see is that 2080Ti will give me 120fps if my 1080Ti gives me 60fps in existing games. Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? I smell a mile high turd and its leaning toward me right now.

 

[Malogeek]

Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? A smell a mile high turd and its leaning toward me right now.

From some replies here, yes the marketing is succeeding. Brains have been turned off.

 

[crisium]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles." I don't know what this means, but what they want me to see is that 2080Ti will give me 120fps if my 1080Ti gives me 60fps in existing games. Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? I smell a mile high turd and its leaning toward me right now.

If 2080 Ti delivers 2x the frames of a 1080 Ti in an average benchmark suite, I'll order one.

I'm not getting my hopes up.

 

[sze5003]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles." I don't know what this means, but what they want me to see is that 2080Ti will give me 120fps if my 1080Ti gives me 60fps in existing games. Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? I smell a mile high turd and its leaning toward me right now.

I highly doubt this will be the case. 2x the performance is a very bold statement to make. They need to say anything to get people to throw money down before actual reviewers get the cards in hand to run the tests we all are waiting for.

 

[PeterScott]

What I don't get is, why not just offer lower prices and really push volume + value for its core gaming audience? Even if they couldn't manage tangible efficiency gains due to an already efficient Pascal arch + basically the same node (see marginal Ryzen 2000 gains from 12nm.. ie basically nothing), then why not come in swinging with a 2070 that brings 1080 performance down to the $300 level?

Why don't people give away free stuff? Look at Chevy Volt. EV + Gas engine. Arguing that they should just sell it the same price as a gas engine only car because that is the most important part is nonsensical. It cost more money to have the EV part in there, and it cost more money to have the RT/Tensor cores in there. This is a business run with profit motive. Not charity giving fans what they want for the price they want.

12nm is very, very cheap for AMD/NVIDIA. Its basically a mature 16nm so you can be sure, especially given Samsung's increased competition for wafer design wins, that TSMC has lowered wafer costs versus when 16FF was bleeding edge tech.

It's not cheap, each process shrink increases the cost/mm2. The 2080Ti has the largest, most expensive die ever in a consumer GPU. So it follows that it is the most expensive consumer GPU ever.

People trying to get defensive about ray tracing tech.. look its cool! It's fine I mean... no one really cares about it either way. The core problem here is seemingly jacked prices akin to peak-crypto fever levels and zero performance benchmarks. Its 100% NVIDIA's fault that people assume the worst even if that isn't the case. What else could they have expected?

The problem here is huge dies and expensive GDDR6 memory. Those things increase the card costs, and NVidia is a business, not a charity, they will try to maintain their margins, so when selling something that is more expensive to produce, they will ask more from the buyer to compensate.

Whenever anyone attempted to add this kind of new HW that eats a lot of die space, there is going to be a hiccup on price/performance, for older games that don't use the new HW.

There are two options:

a) You live with that hiccup.
OR
b) You never have this kind of advance.


I think (a) is a lot better way to proceed. Because there is no (c) where it comes for free.

 

[LTC8K6]

Translation: It will do 20fps instead of 10fps in corner cases in one game at 4K.

How would that cover the 60fps bit though?

 

[realibrad]

I'm no video card salesman, but to me, the best way to sell these cards would have been to show them spanking Pascal cards in traditional gaming, then show the ray great tracing possibilities.

If I had seen a 2070 keeping up with a 1080 in traditional gaming, I'd have wanted a 2070 really bad.

Being able to do RT would have been a cherry on top.

So, when you are competing against yourself, you usually don't want to attack your own product. Nvidia is the premium option, and, want to clear out the glut of their old hardware which is still in the pipeline. Going way above on the price of the new allows you to market the old stuff to get it cleared out because the new stuff is at a different price level.

You remember how people were trying to return chips to Nvidia, well, this would fit as a solution. Offer something new that is at the high end, price it well above the old so you get some revenue from people that are not price sensitive. Offer the old at a slight discount but with healthy margins. Clear out the old product and sell some new high end to hold you over until 7nm is ready.

This also gives them time to refine their 7nm product even further which will help keep them ahead of Intel which wants to enter the market. Its a nice position to be in.

 

[LTC8K6]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles." I don't know what this means, but what they want me to see is that 2080Ti will give me 120fps if my 1080Ti gives me 60fps in existing games. Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? I smell a mile high turd and its leaning toward me right now.

It's puzzling. If it's even close to true, why couldn't they realize that a demo of it would have cards flying off the shelves?
It would have taken only a minute out of their presentation.

 

[alcoholbob]

I give up on this thread, i'll be back in a month i guess.

ITT: people completely unwilling to discuss things that are known, but love talking about things that they demonstrably do not know.

Pretty much the entire basis of the modern investment banking works like that, and this type of thinking was pioneered by nuclear physics--it's called Bayesian probability, and works better than any other model we currently know of.

You should look it up, it's quite a good probability model by looking at past events, and it's absolutely fair to look at the history of Nvidia and see that outside of Maxwell, we haven't seen appreciably performance gains outside of pure cuda core increases. While it's not a sure thing, it's pretty fair to bet that the IPC increases, if any, aren't going to be spectacular.

 

[PeterScott]

"For existing games, Turing will deliver 2x the performance and 4K HDR gaming at 60 FPS on even the most demanding titles." I don't know what this means, but what they want me to see is that 2080Ti will give me 120fps if my 1080Ti gives me 60fps in existing games. Does anyone actually expect that to happen? Really guys? A 100% increase in traditional gaming performance? I smell a mile high turd and its leaning toward me right now.

What is that quote from?

 

[LTC8K6]

What is that quote from?

https://blogs.nvidia.com/blog/2018/08/20/geforce-rtx-real-time-ray-tracing/

 

[LTC8K6]

Also this:

GeForce RTX 20-series are the first GPUs that will play AAA games at 4K 60 FPS HDR.

 

[24601]

Pretty much the entire basis of the modern investment banking works like that, and this type of thinking was pioneered by nuclear physics--it's called Bayesian probability. You should look it up, it's quite a good probability model by looking at past events.

"Modern investment banking" works by insider trading, infinite free money from [central bank] / license to print infinite national money by yourself from the government + law of large numbers * martingale betting.

The rest of it is made up garbage they feed people to keep them happy (aka things like insider traders pretending that they got the stock tip from their stuffed dog or the infamous shoe shine boy story)

 

[Malogeek]

How would that cover the 60fps bit though?

Marketing language. It's two separate statements, one being 2x performance (in what/where? relative to what?) and that it does 4K HDR 60fps (which Pascal already does).

Replace "and" with "in" and you have an inclusive statement.

 

[Hitman928]

When he lands the shadows are being cast in front of him. He didn't inspect where the sun was positioned, but based on that- in every single one of the shots I mentioned, the light source is behind him.

Incorrect again. I don't know why you keep trying to argue this point when it's so demonstrably false. Here's the first scene in question.

As he tops the hill, the light source (sun) is at the top of the screen a little off center to the right. The enemy player is almost dead center, slightly to the left. If you notice, all the shadows from the trees, structures, etc., are being cast back and slightly left. That means the player's shadow would be cast back and slight left, behind the player. The player stands between the enemy and his own shadow. To say RT would help the enemy here is crazy talk and just you grasping at straws trying to justify your initial argument that was silly to begin with, which is why you have to stretch the truth to the point of breaking to try to justify it.

Further, have you never played paintball with hills? Yeah, seeing someone's shadow crest a hill is actually extremely noticeable- if they jumped with that angle of light down the hill surrounded by snow their shadow would be like a beacon.

You seemed to be confused as to how shadows work. If someone is cresting a hill with a light source behind them, you won't see their shadow until, you know, their body gets in the way of the light. This is such a silly argument. I especially love your jumping down a snowy hill scenario. Are you really trying to get us to believe that if someone is literally jumping down a snowy hill that the FIRST thing you notice is their shadow?

A realistic scenario where a shadow is a giveaway for a person's position would be something more like a blind corner in a hallway where there is a dimly lit light source on your side of the corner and a brighter light source on their side of the corner and they are between the light source and the corner. Their shadow would cast past the corner to where it would be visible for you and you would know someone was there, though how far from the corner would be tricky. Even then, I don't see how RT would improve this scenario as the shadow would still be cast (most likely even more clearly) under the current light/shader engines.

As opposed to using the marker floating over the players head, enemy soldiers use that in real life all the time.......

You're arguing against your own point without realizing it. Games like this don't try for 100% realism. They include and exclude certain visual features to make the game easier to play. That's why most people playing these types of games turn shadows to low or even off. More often than not, they are a detriment to competitive play rather than an advantage.

Let's look at it from an ever so slightly different angle if people don't like that. Threadripper 2950x is $899 and smaller than the 2070, the 2990wx is slightly bigger than the 2080Ti for $1800- and neither one of those come with their own PCB and 8/11GB of GDDR6 RAM, so what's the deal with that?

We know for a fact thanks to SEC filings that nVidia spent more on R&D then AMD did, they are offering roughly comparable die sizes with a healthy amount of very high tech RAM and a PCB for quite a bit less than what AMD is offering.

Both of these arguments are completely irrelevant to the discussion. CPUs and GPUs are different markets with different pricing elasticity and expected functionality. R&D expenditure is also irrelevant. Should we as consumers say, oh company X spent 10x the R&D so I am willing to pay 10x the price even though the performance is the same? Price as a function of R&D expense is relevant for the company executives, not for the end consumer.