Rumor? nVidia RTX 2080 & 2080 Ti Reviewers Guide Leak? Lots of 4K

[leadeater]

9 hours ago, bowrilla said:

I don't get why people are still comparing the 1080ti with the 2080ti. With this release NVIDIA x-ed the initial Titan card for enthusiasts. Therefore the Titan Xp successor is the 2080ti.

Yea no, if it were a Titan it would be called a Titan. The Titan normally comes first but just because this time the Ti is here first time round doesn't make it a Titan replacement, the only factors people are using to say that is price and release timing which is not conclusive of anything. It's simply more likely Nvidia needed to release the Ti sooner to keep generational performance increase at release and this is the price of it because it's a huge expensive die.

 

2080 Ti is not the maximum GT102 SM/CUDA core configuration, that is the Quadro RTX 6000 and if (unlikely) the Titan is going to be based off GT102 it'll use that configuration not the 2080 Ti configuration.

 

If there is going to be a Turing Titan it'll be a GT100 part or GT102 (fully unlocked) part likely at similar prices to the Titan V, the 2080 Ti is what it is and it's time to get used to it.

[mr moose]

18 minutes ago, leadeater said:

Yea no, if it were a Titan it would be called a Titan. The Titan normally comes first but just because this time the Ti is here first time round doesn't make it a Titan replacement, the only factors people are using to say that is price and release timing which is not conclusive of anything. It's simply more likely Nvidia needed to release the Ti sooner to keep generational performance increase at release and this is the price of it because it's a huge expensive die.

 

2080 Ti is not the maximum GT102 SM/CUDA core configuration, that is the Quadro RTX 6000 and if (unlikely) the Titan is going to be based off GT102 it'll use that configuration not the 2080 Ti configuration.

 

If there is going to be a Turing Titan it'll be a GT100 part or GT102 (fully unlocked) part likely at similar prices to the Titan V, the 2080 Ti is what it is and it's time to get used to it.

I think most of the claims about the 2080ti being a titan replacement were more to address the issues with people trying to compare only the price difference. Whether Nvidia intend for this to be it and they discontinue the titan name or if they release a titan card later on will remain to be seen.

 

Personally I don't see the 20 series as being a replacement to anything as such, but a new product stack in it's own right.  

[leadeater]

15 minutes ago, mr moose said:

Personally I don't see the 20 series as being a replacement to anything as such, but a new product stack in it's own right.  

Other part yes but this I don't see how it can be called that at all, new product stack how? This is not the first time new technology has been released in GPU hardware, even significant ones such as this time, but it's never been called or considered a new product stack. This is GeForce 20 series, it's the next series after GeForce 10 series but whether people want to buy in to it is totally up to the buyer but I don't see how it's a new product stack.

 

What exactly makes it a new product stack? RTX instead of GTX? Shift away from Shader performance? Increase in pricing? All of these types of things have happened before but it never made it "a new product stack". Some, but not all, Pascal owners will just find the GeForce 20 series disappointing.

 

The more sad thing is if there isn't a significant uptake in these new hardware features that would make GTX 1080 and GTX 1080 Ti owners actually buy in to this series we're talking about another 12 to 18 months of Pascal for these owners which would be the longest generation by a ridiculous margin.

 

Edit:

If this is a new product stack or what ever else people want to call it i.e. not a replacement for GeForce 10 series then Nvidia has simply failed as a graphics card technology company, they would of failed to be able to release a new generation of product and failed their customers.

 

No I don't think that is the case btw.

[Tech Enthusiast]

2 hours ago, mr moose said:

3. if so will people accept that the added tensor cores and associated performance effects (assuming they are positive) are legitimate performance upgrades or will they get stuck looking at the number of cuda cores and ram size and complain it's not worth the money?

1

Haters gonna hate. You can't stop them.

They would never accept any added performance beyond cuda performance.

And I stopped caring. Having a discussion with people that just want to hate, leads to nothing. They claim every little detail is invalid because [reason], every feature is invalid because [reason], or even worse, they leave out the reason and just claim you are an idiot if you don't see it yourself.

 

The funny part about this:

If they had just named it TITAN and not 2080ti, most of the hate would not be happening.

No matter if it is meant to be a titan, it is really just the name that seems to upset people. the 80ti name may not have a price tag like that is the basic reasoning. If we just ignore the names for a min. the whole picture looks less bad honestly. A minor uplift in pricing, a slightly less huge step forward in raster performance, but with plenty of new features added on top. Now we insert back the names and all we have left is: "omg a 80ti going for titan pricing!". #NamesMatter

[mr moose]

1 minute ago, leadeater said:

Other part yes but this I don't see how it can be called that at all, new product stack how? This is not the first time new technology has been released in GPU hardware, even significant ones such as this time, but it's never been called or considered a new product stack. This is GeForce 20 series, it's the next series after GeForce 10 series but whether people want to buy in to it is totally up to the buyer but I don't see how it's a new product stack.

 

What exactly makes it a new product stack? RTX instead of GTX? Shift away from Shader performance? Increase in pricing? All of these types of things have happened before but it never made it "a new product stack". Some, but not all, Pascal owners will just find the GeForce 20 series disappointing.

 

The more sad thing is if there isn't a significant uptake in these new hardware features that would make GTX 1080 and GTX 1080 Ti owners actually buy in to this series we're talking about another 12 to 18 months of Pascal for these owners which would the the longest generation by a ridiculous margin.  

 

Personally RTX is to GTX what xeon is to the core series.   Yes they do very similar jobs and to the end consumer are basically the same inside for the most part, but there are differences. This is a new product with not just a new/larger architecture, but an entirely new method of processing/programing.

[leadeater]

8 minutes ago, mr moose said:

Personally RTX is to GTX what xeon is to the core series.   Yes they do very similar jobs and to the end consumer are basically the same inside for the most part, but there are differences. This is a new product with not just a new/larger architecture, but an entirely new method of processing/programing.

Well no because Xeon and Core are essentially the same thing with a sprinkle of enabled features but are the same architecture, I'd say it's Core (Core 2 Duo/Quad) to Core i series.

 

I would make the Phenom/K10 to Bulldozer comparison, the way it's the same thing but different, but I'd get murdered for that because of how bad Bulldozer was and people would think that is what I was imply rather than the architecture shift. Sometimes architecture shifts doesn't lead to immediate performance increases or see the benefits realized immediately and when that happens the disappointment can be rather harsh. 

[WMGroomAK]

1 hour ago, leadeater said:

2080 Ti is not the maximum GT102 SM/CUDA core configuration, that is the Quadro RTX 6000 and if (unlikely) the Titan is going to be based off GT102 it'll use that configuration not the 2080 Ti configuration.

 

If there is going to be a Turing Titan it'll be a GT100 part or GT102 (fully unlocked) part likely at similar prices to the Titan V, the 2080 Ti is what it is and it's time to get used to it.

This has been my take on it...  I would not be surprised if nVidia releases an RTX Titan sometime in 2019 (maybe in Q1?) that consists of a fully unlocked TU102 part with 12 GB of RAM.  My best guess is that nVidia will either decrease the MSRP of the 2080 Ti to 900 or 1000 and release any RTX Titan at $1300 or keep the 2080 Ti price the same and release a Titan part at around $1500...  It would be interesting to see the performace of two 2080 Ti linked via NVLink versus a single Quadro RTX 6000.  Would you be able to get equivalent or better performance out of the $2500 for the 2080 Tis and the NVLink bridge than the $6300 for the Quadro? 

[leadeater]

4 minutes ago, WMGroomAK said:

This has been my take on it...  I would not be surprised if nVidia releases an RTX Titan sometime in 2019 (maybe in Q1?) that consists of a fully unlocked TU102 part with 12 GB of RAM.  My best guess is that nVidia will either decrease the MSRP of the 2080 Ti to 900 or 1000 and release any RTX Titan at $1300 or keep the 2080 Ti price the same and release a Titan part at around $1500...  It would be interesting to see the performace of two 2080 Ti linked via NVLink versus a single Quadro RTX 6000.  Would you be able to get equivalent or better performance out of the $2500 for the 2080 Tis and the NVLink bridge than the $6300 for the Quadro? 

Will be interesting to see what is done with a Titan, was the Titan V GV100 a one off and the Titan will return to Gx102 or will it continue with Gx100. I'm torn as to which because I think the Titan V only used GV100 so it could have Tensor cores plus there was only one die anyway so choice was that or none, there's no GV102 or below to make any judgments off as to how it will play way. Volta was weird....

[mr moose]

5 hours ago, leadeater said:

Well no because Xeon and Core are essentially the same thing with a sprinkle of enabled features but are the same architecture, I'd say it's Core (Core 2 Duo/Quad) to Core i series.

 

I would make the Phenom/K10 to Bulldozer comparison, the way it's the same thing but different, but I'd get murdered for that because of how bad Bulldozer was and people would think that is what I was imply rather than the architecture shift. Sometimes architecture shifts doesn't lead to immediate performance increases or see the benefits realized immediately and when that happens the disappointment can be rather harsh. 

I know fuck all about the xeons because I don't use them, but I used them as the example because because they physically have more cores, options that the core series don't (like ECC) and come on a completely different socket ( FCLGA3647) essentially making them a different product stack.  

 

[leadeater]

22 minutes ago, mr moose said:

I know fuck all about the xeons because I don't use them, but I used them as the example because because they physically have more cores, options that the core series don't (like ECC) and come on a completely different socket ( FCLGA3647) essentially making them a different product stack.  

Anyway, this is either a similar turning point like when hardware Shaders got introduced (GeForce 3 series) and there will be a sustained shift towards ray tracing in more significant ways or it'll stay in the realm of post effects only, similar situation to PhysX but fair amount more used than that.

 

I really don't know where it's going to go. There needs to be a lot more in depth information on things like what it takes to do a more complex full real time render using ray tracing, where are we at now, way more detail on the RT cores like how much die area and how easy is it to scale them out and also how efficient they are when scaled out (to replace more raster/shader pipeline).

 

Nice dose of put marketing aside and honest let's be real where are we at and how far do we need to go basically.

 

I'm also sure that now there is much more spotlight on ray tracing more techniques on the software/API side will come out to lower the hardware compute requirements as well.

[mr moose]

6 minutes ago, leadeater said:

Anyway, this is either a similar turning point like when hardware Shaders got introduced (GeForce 3 series) and there will be a sustained shift towards ray tracing in more significant ways or it'll stay in the realm of post effects only, similar situation to PhysX but fair amount more used than that.

 

I really don't know where it's going to go. There needs to be a lot more in depth information on things like what it takes to do a more complex full real time render using ray tracing, where are we at now, way more detail on the RT cores like how much die area and how easy is it to scale them out and also how efficient they are when scaled out (to replace more raster/shader pipeline).

 

Nice dose of put marketing aside and honest let's be real where are we at and how far do we need to go basically.

 

I'm also sure that now there is much more spotlight on ray tracing more techniques on the software/API side will come out to lower the hardware compute requirements as well.

I don't think anyone knows where it is going, suffice to say it has a very healthy dose of developer support.  Given not too many new technologies in the past have enjoyed this much attention on release let alone before the product even hits the shelves.

[JediFragger]

All nVidia have done is move the Ti into the old £1200-ish Titan segment and imo are intending to keep the Titan in the near-£3k bracket intended for the 'prosumer' market. Makes sense from that point of view for me.

[Brooksie359]

On 9/16/2018 at 2:20 AM, mr moose said:

I don't think anyone knows where it is going, suffice to say it has a very healthy dose of developer support.  Given not too many new technologies in the past have enjoyed this much attention on release let alone before the product even hits the shelves.

From what luke said on the wan show and what nvidia was saying during the press conference it seems that it's very easy to implement meaning it will likely have much larger adaptation than other technologies that aren't easy to implement. 

[TahoeDust]

So what day/time est are the reviews going to drop?  I have s 2080ti FTW3 pre-ordered, but it's an easy cancel.