RTX 20 Series Cards Confirmed With Pricing

[dizmo]

1 minute ago, Spotty said:

I didn't think of import duties. Just put it in to the Australian Customs import duty calculator and it's AUD $81.50 for the import duties on the RTX2080ti putting it at around $1880 AUD. So you're right, more or less that means the pricing lines up with the USA pricing. By the time you add in shipping it works out about the same.

Aye. You're minimum wage is a lot higher too, no? $16/hr?

[Spotty]

2 minutes ago, dizmo said:

Aye. You're minimum wage is a lot higher too, no? $16/hr?

Yeah about $16-$17 AUD per hour ($11-$12 USD), but general cost of living is higher. Housing/renting is expensive here as well (too many 'baby boomers' living too long). Probably going to go very much off topic going in to it in too much detail.

Once you look at tax and import fees the Australian pricing works out pretty reasonable compared to the USA pricing. If I were to order one from USA and pay in USD to have it shipped to Australia it would work out about the same, maybe a little bit more expensive than buying locally, depending on shipping options (express, insurance, etc).

[ne0tic]

14 hours ago, SteveGrabowski0 said:

Jen-Hsun can go fuck his mother then. $600 for a 70-series card is ridiculous.

I would be happy for that price, in Sweden a RTX 2070 costs about $850!

[unsubscribed]

While not unexpected the move is bold and nasty to most of us. Few points to note from this:
the rumor of large existing stock is pretty much confirmed hence the high price, expect it to drop a good bit around January or worst case scenario when AMD anounces something new; perfect timing for NVidia

speed increase is only in RTX titles, in anything pre RTX which is 99.9% of existing games the increase will be the usual 25% or so compared like with like on both generations

above plus no competition in the market means NVidia can once again do whatever the heck they want; you guys do know that this means NVidia will be piling cash to get even bigger lead against AMD

 

So yeah, if you need fastest and greatest - start saving. These prices are not going to change anytime soon.

 

[Taf the Ghost]

2 hours ago, IhazHedont said:

I genuinely wonder if ray tracing won't be, again, a gimmick.

 

I mean it's not the first time new graphics cards are advertised around a new tech: DX10/11 and probably 12, exclusive Nvidia features in Gameworks and so on.

 

Each time these technologies were supposed to bring a new revolution to gaming, but at the end of the day they were nothing else but a gimmick.

 

Unless the entire industry is going to the same direction this time ...

It's not a gimmick and the core tech has been worked on for well over 20 years. The issue is replacing the old technology. It's going to be a while. Raytracing is one of the things they want to get to, yet it's probably not ready for primetime. In 5-7 years, that might be a different story, and Nvidia will have the GPU base to make it work. Plus, well, they can do certain types of compute tasks really, really well, which Nvidia can sell for a lot of money.

[asus killer]

1 hour ago, mr moose said:

?? I don't thinks so.   There is always variance in price depending on market forces, quality, market demand etc. 

But this is not an elastic demand. If you build on a budget you have 500€, if graphics cards are more expensive you dont spend more you get a 2200g and no gpu like we seen recently. You get a 1050ti instead of a 1060 and so on.

 

Its their right to do it, i just think they miss how this markets work.

We seen this on consoles for example.

[Trixanity]

1 hour ago, dizmo said:

Poor argument.

Yeah, being called out for stupidity sucks indeed. Easier to dismiss than to learn.

[Taf the Ghost]

2 minutes ago, Sampsy said:

For now I imagine it will be a gimmick - the cards just simply aren't fast enough to use it extensively. All of the tech demos showing it off were very simple scenes, with fairly basic geometry, very low view distance, few objects on the screen and generally very little going on. And even then the frame rates achieved were very low. It's exciting as it does look like more extensive real time ray tracing will be achievable in a couple of generations, but we're simply not there yet. 

 

Otherwise I wasn't very impressed with the presentation. The fact that Nvidia didn't actually say how much faster these cards are than Pascal makes me think that they probably aren't much faster at all. That's why they are hyping up the small amount of RT work the cards will be able to do. And the pricing is extremely high - the 1080TI is easily twice as expensive as it should be, but that's what a lack of credible competition does to the market I suppose. 

Other thing is that the Raytracing hardware doesn't have a massive amount of man-hours spent optimizing for it & with it. These are new, dedicated computational units. It just takes time to develop the infrastructure to take advantage of it. So, I imagine even this generation of Hardware has a lot of power available, but it's going to take years of learning to deal with it before the real power is unleashed.

 

Reminds me of all of the hype around Quantum Computers. You have to develop the entire infrastructure around the exploitation of the computational power. That just takes years to build.

[OriAr]

I know this is controversial but I don't think Nvidia's profit margins are as big as people here, while obviously a part of the price increase is because they can, people need to remember two things:

1. DRAM prices have been going crazy.

2. The die is huge, 750 square mm die can't be easy to produce and it's reflected in the price.

 

Hopefully with 7nm the die will be smaller and the price will get somewhat more normal.

[cj09beira]

1 hour ago, dizmo said:

 

You say that as if AMD wouldn't have done the same thing. No, they'll just spin off the same card for 3 or 4 generations, slightly tweaking it and saying it's new.

The prices for the lower tier cards, if AIB's stick to the prices provided, isn't really that bad once you factor in increases in memory costs and such.

AMD can't compare performance wise, unfortunately.

 

amd probably wouldn't, even when all they could do was rebrand they usually at least ask less for it, as it should be, (eg: ryzen 1 vs ryzen 2, 7970 vs 280x, etc)

do you think they spin off the same card for a long time because they want to, its because they can't afford to design 3 gpus per generation so they have to do 1-2 at a time, sometimes they make a new large gpu sometimes a new small gpu, 

1 minute ago, Taf the Ghost said:

Other thing is that the Raytracing hardware doesn't have a massive amount of man-hours spent optimizing for it & with it. These are new, dedicated computational units. It just takes time to develop the infrastructure to take advantage of it. So, I imagine even this generation of Hardware has a lot of power available, but it's going to take years of learning to deal with it before the real power is unleashed.

 

Reminds me of all of the hype around Quantum Computers. You have to develop the entire infrastructure around the exploitation of the computational power. That just takes years to build.

considering its probably renamed tensor cores, there is probably not much to be optimized though, probably more on the table from the software side 

[Taf the Ghost]

It's kind of funny we're starting to get the first round of, "if they're not talking about improved gaming performance, what's the actual increase?". Async Compute and faster memory should provide a good uplift per SM, but it's not looking to be huge compared to already OC'd Pascal cards.

[Taf the Ghost]

1 minute ago, cj09beira said:

amd probably wouldn't, even when all they could do was rebrand they usually at least ask less for it, as it should be, (eg: ryzen 1 vs ryzen 2, 7970 vs 280x, etc)

do you think they spin off the same card for a long time because they want to, its because they can't afford to design 3 gpus per generation so they have to do 1-2 at a time, sometimes they make a new large gpu sometimes a new small gpu, 

considering its probably renamed tensor cores, there is probably not much to be optimized though, probably more on the table from the software side 

That reminds me: has anyone seen any technical description of what these RT Cores are? They have to be some sort of vector math unit, but it has to be something non-standard. 

 

Sounds like something to go looking into.

[OriAr]

3 minutes ago, Taf the Ghost said:

That reminds me: has anyone seen any technical description of what these RT Cores are? They have to be some sort of vector math unit, but it has to be something non-standard. 

 

Sounds like something to go looking into.

I bet AnandTech will have a detailed description of them in their RTX review.

[cj09beira]

Just now, OriAr said:

I bet AnandTech will have a detailed description of them in their RTX review.

lets hope so

[Taf the Ghost]

11 minutes ago, cj09beira said:

considering its probably renamed tensor cores, there is probably not much to be optimized though, probably more on the table from the software side 

Go poking around and look what you find...

 

https://www.reddit.com/r/nvidia/comments/97qogl/former_nvidia_gpu_architect_what_is_rt_core_in/

 

Quote

The RT core essentially adds a dedicated pipeline (ASIC) to the SM to calculate the ray and triangle intersection. It can access the BVH and configure some L0 buffers to reduce the delay of BVH and triangle data access. The request is made by SM. The instruction is issued, and the result is returned to the SM's local register. The interleaved instruction and other arithmetic or memory io instructions can be concurrent. Because it is an ASIC-specific circuit logic, performance/mm² can be increased by an order of magnitude compared to the use of shader code for intersection calculation.

Some more discussion of it here:

 

https://www.reddit.com/r/nvidia/comments/8mzfsh/nvidias_rtx_denoising_is_not_based_on_ai_thats/dzua8t4/

[Taf the Ghost]

5 minutes ago, OriAr said:

I bet AnandTech will have a detailed description of them in their RTX review.

There'll be a lot of discussion of it, I imagine.

4 minutes ago, cj09beira said:

lets hope so

Let's.

 

Though I think I get how it's approaching things, and it makes sense. It's dedicated hardware units that spit out the data the rest of the SM needs to produce the scene. This isn't going to bring any extra performance, but things should look better. Which isn't "nothing".

[Helly]

Loving the bitching about the price people, good job. Doesn't matter though, both you and Nvidia know that you're going to buy these cards anyway. Because you got to have a new card and AMD "can't compete" and has nothing new.

 

You either let your wallet do the talking now or you accept these prices as the new standard. Again both you and Nvidia know which it will be. Good job everyone, thx for the $1000 top tier GPU's. Coming up next year, $1000 xx70. Year after that, $1000 xx60, or are you actually thinking intel will ask anything less then nvidia when the performance comes close enough?

 

They asked for this and you all gave it to them, the only ones to blame are the people buying this crap, not some company that can't keep up. They can't sell what you don't buy.

[Taf the Ghost]

@cj09beira

 

Thinking about it, the Raytracing will probably kill FPS for a long while before they get all of the kinks worked out. It's going to be a Crysis-like effect for a while. We are likely to see some massive jumps in performacne with driver updates as they work out the bottlenecks the Raytracing will cause.

[mr moose]

52 minutes ago, asus killer said:

But this is not an elastic demand. If you build on a budget you have 500€, if graphics cards are more expensive you dont spend more you get a 2200g and no gpu like we seen recently. You get a 1050ti instead of a 1060 and so on.

 

Its their right to do it, i just think they miss how this markets work.

We seen this on consoles for example.

Are you sure nvidia don;t know how markets work? because I have watched the value of cards almost double since mining and ram prices became a market force. I have watched laptop prices fall when demand shrunk.  I have seen the vega's sell for $1200 while the 1080ti was still at $1100,  why? because market forces and supply/demand change the price a company can charge for a product.   There is nothing yet I have seen that makes the price premium they are charging insane or out of the ordinary.   They know it will sell, and when it no longer does they will drop the price, just like all companies who want to make money.

 

 

[mr moose]

27 minutes ago, Helly said:

Loving the bitching about the price people, good job. Doesn't matter though, both you and Nvidia know that you're going to buy these cards anyway. Because you got to have a new card and AMD "can't compete" and has nothing new.

 

You either let your wallet do the talking now or you accept these prices as the new standard. Again both you and Nvidia know which it will be. Good job everyone, thx for the $1000 top tier GPU's. Coming up next year, $1000 xx70. Year after that, $1000 xx60, or are you actually thinking intel will ask anything less then nvidia when the performance comes close enough?

 

They asked for this and you all gave it to them, the only ones to blame are the people buying this crap, not some company that can't keep up. They can't sell what you don't buy.

It's not solely the buyers who are to blame, Yes they play their role, but markets self level, they can only increase the price so far before the number of people who can afford to buy is so small that they would make more money from selling more units at a lower price.   Also there are other factors still at play here.

[Taf the Ghost]

1 minute ago, mr moose said:

It's not solely the buyers who are to blame, Yes they play their role, but markets self level, they can only increase the price so far before the number of people who can afford to buy is so small that they would make more money from selling more units at a lower price.   Also there are other factors still at play here.

The high-end GPU market is quite small, though it brings in large margins. This is mostly Nvidia operating on the window with no RX Vega 96 on the way to exploit that segment of the market, because that segment has to pay significant amounts of money for upgrades. Yet they're actually getting the higher clocked but smaller SM dies, while the best dies are going to the Quadros.

[mr moose]

3 minutes ago, Taf the Ghost said:

The high-end GPU market is quite small, though it brings in large margins. This is mostly Nvidia operating on the window with no RX Vega 96 on the way to exploit that segment of the market, because that segment has to pay significant amounts of money for upgrades. Yet they're actually getting the higher clocked but smaller SM dies, while the best dies are going to the Quadros.

You mean supply/demand and market forces are allowing them to charge a premium for a premium product?  Not too sure how this shows nvidia doesn't understand how markets work.  If anything it exemplifies that.

[Taf the Ghost]

3 minutes ago, mr moose said:

You mean supply/demand and market forces are allowing them to charge a premium for a premium product?  Not too sure how this shows nvidia doesn't understand how markets work.  If anything it exemplifies that.

I didn't make the argument Nvidia doesn't understand what they're doing. I was commenting on the nature of the high-end market. 

 

This pricing structure, to me, actually says two things about the high-end market. Firstly, Nvidia is in absolutely no rush to bring 7nm dies to market. A straight shrink puts these around 400 mm², which is still really expensive. Nvidia can hang out on 12nm for 2 years if they see reason to, but it looks like probably 18 months or less.

 

The second thing it suggests is that xx70 series is actually fairly low volume relative to the xx80 class. A quick trip to the Steam Hardware survey and we have...

 

1070 - ~3.6%

1070 Ti - ~.6%

1080 - ~2.2%

1080 Ti - ~1.2%

 

So that's 3.4% of xx80 class to 4.2% of xx70 class. With obvious caveats about the Steam survey's data quality noted, we still have a pretty good picture of the relative sales volume. Nvidia is going to extra out of those xx70 class buyers the extra $100USD they wanted them to pay for the xx80 class. One way or the other.

[asus killer]

i can pre order them in my country (cheapest store) they range from 800 to 1350 euros. That's 1560 USD 

[JediFragger]

1 hour ago, Taf the Ghost said:

but it's not looking to be huge compared to already OC'd Pascal cards.

Makes you wonder if nVidia will start the Pascal 'performance decreasing' in drivers early this time