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

[Hitman928]

No, your exaggerated numbers are a lie that misrepresents reality. Nothing better. Stop exaggerating to misrepresent my position, and misrepresent the situation.

Ok, since you keep calling me a liar and disingenuous without actually addressing any of my points, would you kindly explain your point clearly and how my example is a twisted lie of your point.

Misrepresenting through exaggeration is disingenuous.

No it's not, it's taking a flawed argument and showing why it's flawed because it can so easily be taken to an extreme that no one would agree with while maintaining the same argument. Instead of just constantly trying to smear me, why not address the argument instead?

Pretending that an unadvertised internal part numbers represent some kind of consistent marketing, is disingenuous.

Again, you're arguing that marketing names (which can be changed at the drop of a hat) and die size are what matters for pricing, correct? That's the only consistent argument I've seen you make. I'll wait for your response before disagreeing because I don't want to be called a liar for disagreeing with you.

Complaining is fine, but try to keep it honest.

Where did I complain? I'd like to see you quote me complaining once in this thread about Nvidia's pricing. Who's being dishonest here?

The 2070 should be considered this generations 1070, by any reasonable comparison.

The pretenses used to argue that it's this generations 1060, are disingenuous games.

You keep repeating the same thing and ignoring any counter argument, that's what seems pretty disingenuous to me.

I'll let you know (again) that I'm not upset about Nvidia's pricing. You keep trying to assign a false emotional motive to my arguments to try to just dismiss them. I can evaluate Nvidia's new pricing structure without getting upset or emotionally invested in it.

Truth is, if someone is upset about Nvidia's new pricing, they should be upset with AMD for not competing and forcing a more consumer favorable market. That doesn't prohibit me (or anyone else) from analyzing the new prices and concluding that it's less consumer friendly and is consistent with a pattern that has developed over the last 6 years or so.

 

[EXCellR8]

If anyone really wants ray tracing, and that's fine, I would wait for 3000 series cards which will most likely be refined versions; by that time more games will support RT.

Also, still can't find any information on GTX cards... is it possible they will release GTX 2080 without the ray tracing?

 

[Hitman928]

If anyone really wants ray tracing, and that's fine, I would wait for 3000 series cards which will most likely be refined versions; by that time more games will support RT.

Also, still can't find any information on GTX cards... is it possible they will release GTX 2080 without the ray tracing?

No real info yet. The rumor is that GTX cards without RT come in below the 2070. So 2060 and below will be GTX. Could be that their essentially rebranded 10xx cards (i.e. 1070 becomes 2060) or could be that they are Turing without the RT. Also could be that they don't release anything else and just keep selling 10xx series cards at their current prices.There's been no official indication yet.

 

[GodisanAtheist]

Another thought (I haven't had my morning joe so bear with me here): Nvidia took this radical departure from their established execution cycle because... their arch is starting to max out.

Aside from the addition of the RT/TN cores, the word on the web has been that the SM structure of Turing is not a radical departure from either Pascal or Maxwell.

The evolution of NV's arch is dead, because as far as rasterization workloads are concerned, they have effectively perfected it. From this point forward, it was going to be brute force gains reliant on die shrinks, which we have seen are becoming a less and less reliable as foundries begin knocking on God's door with their lithography process.

AMD, with fewer resources and playing second fiddle in a number of markets, capped out their current arch with the Vega configuration and simply doesn't have the funds for a major overhaul.

Perhaps NV finds itself staring at the same door (in their case, their arch is already so good that overhauling it just doesn't pass the $/reward muster). So instead they go all in with RT/TN, leveraging the gap they've placed between themselves and their competitor.

 

[crisium]

They are claiming gains to the SMs of Turing though.

2080 vs 1080 Ti will reveal if there's any truth. FE vs FE, 2080 will boost higher but it would have to boost 21% higher to match 1080 Ti TFLOPs. It also has less bandwidth and fewer ROPs. I am very interested in this comparison.

 

[tamz_msc]

Turing SMs are significantly different from Pascal SMs, but not so much different from the Volta SMs, if you ignore the FP64 cores.

 

[PeterScott]

You keep repeating the same thing and ignoring any counter argument, that's what seems pretty disingenuous to me.

What counter argument? That some internal part number on a card, is the true representation of a cards market segment?

Not its power usage, not it's actual marketing designation, not it's price, not it's performance. All of which indicate this a the next generation x70 card.

But apparently for some of you, none of those things matter. Instead it's internal part number of one of it's chips.

I haven't ignored this "counter argument". I openly ridiculed that "argument" for the utter nonsense that it is.

 

[PeterScott]

They are claiming gains to the SMs of Turing though.

2080 vs 1080 Ti will reveal if there's any truth. FE vs FE, 2080 will boost higher but it would have to boost 21% higher to match 1080 Ti TFLOPs. It also has less bandwidth and fewer ROPs. I am very interested in this comparison.

I don't think 2080 will match 1080Ti TFLOPs. NVidia keeps working in more optimizations to get better gaming performance with less brute force TFLOPS.

This is why AMDs raw compute performance has been better across segments (ex: GTX 1060 vs RX 580) where gaming performance is similar. AMD has less gaming optimizations, so then need more raw brute force TFLOPs (usually throught bigger chips, but obviously with RT tax not now), to compete in gaming.

There is another thread about whether AMD can exploit how much bigger RTX dies have become.

Hopefully AMD should have some new consumer cards early next year. They likely won't carry the "RT Tax" on die space/price, and thus can devote die space more to gaming. It should present some opportunity for AMD to have cards with clearly better price/performance on traditional Raster Games, and a chance for people here to vote against Raytracing with their wallets.

 

[Hitman928]

I haven't ignored this "counter argument". I openly ridiculed that "argument" for the utter nonsense that it is.

Let's see:

it's how the product line stacks up regardless of the marketing name. You know, the chips Nvidia actually designs and orders mask sets for and puts on wafers. I wouldn't say that's arbitrary

Then there's this:

If it performed like a 1050Ti, then at best it would be a next generation 2050, and sell for less than the 1050Ti MSRP.

Aren't you arguing against the 20xx series line up here? Isn't this the point I was making (albeit mine was taken to an extreme to prove a point). According to the best estimates we have, the 2080 will have the performance of ~1080 Ti and yet cost more. So according to you, the 2080 should be priced under the 1080 Ti. . .

Then:

It wasn't a straw man, I just took your argument and exaggerated the price increase to show why the 20xx series doesn't make sense from a consumer stand point. Yes, the number is exaggerated, but the principle of the argument still holds.

Yes, Kepler to Maxwell didn't bring a large SM increase at each level, but it was a thoroughly reworked architecture which brought large clock speed and overall performance increases. We don't have that in this case, from what we know so far. If it turns out that Turing brings significant IPC increases (we already know clock speeds are basically the same), I'll be happy to adjust my opinion.

As far as shaders go, they're not on the same node. Pascal is on 16 nm, Turing is on 12 nm. It's not as large a difference as we would normally see in a node shrink, but there is a density increase as well as a reticle limit increase. Also, part of the reason the 2070 and 2080Ti are so close is because the 2080Ti is more cut down from the full die than the past few generations (going by the full Volta core) and most likely the 2080 is cut down as well. Looks like boost clocks also went down compared to GP107 also. So if you compare that to prior gens, it's not as big of a leap for the 2070 as it seems.

would you kindly explain your point clearly and how my example is a twisted lie of your point.

No it's not, it's taking a flawed argument and showing why it's flawed because it can so easily be taken to an extreme that no one would agree with while maintaining the same argument. Instead of just constantly trying to smear me, why not address the argument instead?

Again, you're arguing that marketing names (which can be changed at the drop of a hat) and die size are what matters for pricing, correct? That's the only consistent argument I've seen you make.

Also:

Complaining is fine, but try to keep it honest.

Where did I complain? I'd like to see you quote me complaining once in this thread about Nvidia's pricing. Who's being dishonest here?

All of the above you either ignored or replied with, "stop lying / being disingenuous."

It's perfectly fine if you disagree with me, I have no problem with that. I do take issue with you dismissing all arguments that you don't like and calling me a liar for disagreeing with you.

 

[LTC8K6]

I think the 2080 will be right with the 1080ti in Tflops.

 

[PeterScott]

All of the above you either ignored or replied with, "stop lying / being disingenuous."

I ignored the stuff, which I already agreed with in the past. I have no issue with rationally looking at the product stack and pointing out the prices have gone up. That is blatantly obvious, and something that I don't think anyone disagrees with, and neither do I.

This recent nonsense I am arguing against is your ridiculous claim that the 2070, is really a x60 series card, because of an internal part number.

Ignoring the 2070, power usage, actual marketing designation, price, performance, while only using the internal part number seems like a clearly disingenuous argument to me.

When I said "what counter argument". It was what counter argument that makes the case for your ridiculous claim that the 2070 is really a x60 series, not any other random RTX series issues you want to divert away with.

Are we clear now?

 

[railven]

I randomly pop in to check if there is any useful new info, and still this minutia. You got a mod liking a post by a poster attacking another poster.

We're back to stupid names too? Should we return to Nvidiots and FanATIcs while we're at it?

Am I just getting too old!?

However, good to see good ole '90% of GTX 1080 in DX12 for $200' prophet is still prophesying.

NV's going to continue to dominate, and the same people will continue to find every thing under the Sun to complain about. Glad I stopped buying coffee at work, because it sure is getting expensive to upgrade. If AMD brings us back to normalcy, kudos, but as most others will do - just buy NV at lower prices.

Only hope I have left is Intel. AMD won't ever see a dime from me again, and Nvidia is just the defacto king now. Here is hoping Intel delivers. 2020 can't come soon enough.

 

[railven]

Another thought (I haven't had my morning joe so bear with me here): Nvidia took this radical departure from their established execution cycle because... their arch is starting to max out.

Aside from the addition of the RT/TN cores, the word on the web has been that the SM structure of Turing is not a radical departure from either Pascal or Maxwell.

The evolution of NV's arch is dead, because as far as rasterization workloads are concerned, they have effectively perfected it. From this point forward, it was going to be brute force gains reliant on die shrinks, which we have seen are becoming a less and less reliable as foundries begin knocking on God's door with their lithography process.

AMD, with fewer resources and playing second fiddle in a number of markets, capped out their current arch with the Vega configuration and simply doesn't have the funds for a major overhaul.

Perhaps NV finds itself staring at the same door (in their case, their arch is already so good that overhauling it just doesn't pass the $/reward muster). So instead they go all in with RT/TN, leveraging the gap they've placed between themselves and their competitor.

Talking about this with the wife, and I basically think it's NV's attempt to usurp AMD's foot hold in consoles. Consoles this gen were very under powered, which led to PC ports with relatively lax PC requirements. Something that's bad for any company trying to sell you overpriced toasters. IF, NV can get MSFT to work up Ray Tracing, and can sell it to enough devs through their dev-rels (they have a better chance of RT catching on than Rapid Math or that word that should not be mentioned (primitive shaders)) they create an opening down the road to either A) bolster PC requirements for games (even consoles) leading to being able to sell more expensive toasters, or B) shaking AMD's grip on the consoles.

I mean practically everyone and their mother here confirmed Switch was going to be an AMD based SoC. It wasn't and then the Switch was dubbed "the worst console ever" yet it's a run away success (basically, most of the pro-AMD guys here are too blind to listen when even the President of Nintendo said "we're going ARM" but what does the president know, ATF posters are experts).

I don't mean NV is going to be in the next gen consoles, but I do mean NV can shake the buck enough that they might be considered for the gen after, or at least shaking the PC/API landscape enough that AMD's limited resources are but through more strain as they now have to try to catch up on that front.

NV is playing 4D chess here, and RT can fizzle or become the new standard. Interested to see where it goes.

 

[Hitman928]

What counter argument? That some internal part number on a card, is the true representation of a cards market segment?

The internal part number isn't just a number, it's a designation for actual chips that must each be designed and validated separately as well as separate mask sets purchased and separate wafers ordered. The internal part number tells us what the chips actually are and how they rank in the full product stack. How is this arbitrary?

Not its power usage, not it's actual marketing designation, not it's price, not it's performance. All of which indicate this a the next generation x70 card.

But apparently for some of you, none of those things matter. Instead it's internal part number of one of it's chips.

It's strange to me that you mention marketing name and price as reasons why it's a x70 card. Nvidia can make any card it wants the x70 card and price it that way.

Marketing names can be anything and change all the time. Nvidia's top gaming card used to be x80, then it was Titan, then it was x80Ti. Go back before that and there were x80 Ultras and x900XT's and all sorts of labels at every performance tier.

What would have stopped Nvidia from not releasing the GP102 card at all? Why couldn't they have released the 1080 as the 1080 Ti at $700? It would have been the fastest gaming card in the world, so why not call it the 1080 Ti? They've done it before in the past (GK104 took the established top tier marketing name and price at the time despite being the cut die).

Of course performance is key to where a card sits in the product stack (I personally don't care as much about power), but that's tied directly to which die it is which ties directly to it's internal code name. You have a brand new product stack full of cut down chips (most likely for yield reasons) but the smallest chip isn't cut down so it's fine that it gets to step up to the traditionally higher level marketing name? If you want to say that it provides a relative better value to the higher cards than in the past, then I'm fine with that, but then to be consistent, you also have to recognize that the higher level cards are the cut down cards of the past and therefore provide even less value (relative) themselves.

No matter which way you look at it, Nvidia took it's 4th tier chip (by design and fabrication) and gave it a 3rd tier marketing name while pushing the entire product stack up in price. If you want to say that you think it's performance will justify this, that's fine, but to call someone a liar and disingenuous for pointing this out most certainly rubs me the wrong way.

 

[PeterScott]

The internal part number isn't just a number, it's a designation for actual chips that must each be designed and validated separately as well as separate mask sets purchased and separate wafers ordered. The internal part number tells us what the chips actually are and how they rank in the full product stack. How is this arbitrary?

The name of a chip, is just as arbitrary as the name of a card. It's just a name, assigned by NVidia.

They are simply splitting to more chips this time, x80 and x70 get separate chips. A different chip at the X70 level, doesn't mean it's an x60 card. That is just nonsensical.

When 1080 had 314 mm2 die, it was economically reasonable to still sell a higher volume, lower priced x70 based on the same chip.

Now that that the 2080 uses a MASSIVE 545 mm2 die, it has no place in the higher volume, lower priced(relatively) 2070. They NEED the die size reduction this time since the parent die is so large.

One more thing. Note that they maintained the 256 bit memory bus as well. Not going to the same 192 bit bus used on the last x60 card.

So the 2070:
Guzzles power like x70 series,
Performs like a x70 series,
has a 256 bit memory bus like a x70 series,
priced like a x70 series
Actually named an x70 series.

But it's really a nefarious NVida plot selling a x60 as x70 series because of the name stamped on the chip.

Because it's not like there is a perfectly logical reason why they need another chip this time specifically for x70 serise. Oh wait, there is.

No matter which way you look at it, Nvidia took it's 4th tier chip (by design and fabrication) and gave it a 3rd tier marketing name while pushing the entire product stack up in price.

Uh no. How can it be a "4th tier chip" when they only have 3 chips and 3 card tiers. Only by twisting your view of reality to fit your strange idea that the 2070 is a x60 card.

 

[Elfear]

I randomly pop in to check if there is any useful new info, and still this minutia. You got a mod liking a post by a poster attacking another poster.

We're back to stupid names too? Should we return to Nvidiots and FanATIcs while we're at it?

Am I just getting too old!?

However, good to see good ole '90% of GTX 1080 in DX12 for $200' prophet is still prophesying.

NV's going to continue to dominate, and the same people will continue to find every thing under the Sun to complain about. Glad I stopped buying coffee at work, because it sure is getting expensive to upgrade. If AMD brings us back to normalcy, kudos, but as most others will do - just buy NV at lower prices.

Only hope I have left is Intel. AMD won't ever see a dime from me again, and Nvidia is just the defacto king now. Here is hoping Intel delivers. 2020 can't come soon enough.

Maybe you should consider what fuel you're adding to the fire and instead contribute to a more meaningful discussion.

 

[tamz_msc]

Talking about this with the wife, and I basically think it's NV's attempt to usurp AMD's foot hold in consoles. Consoles this gen were very under powered, which led to PC ports with relatively lax PC requirements. Something that's bad for any company trying to sell you overpriced toasters. IF, NV can get MSFT to work up Ray Tracing, and can sell it to enough devs through their dev-rels (they have a better chance of RT catching on than Rapid Math or that word that should not be mentioned (primitive shaders)) they create an opening down the road to either A) bolster PC requirements for games (even consoles) leading to being able to sell more expensive toasters, or B) shaking AMD's grip on the consoles.

I mean practically everyone and their mother here confirmed Switch was going to be an AMD based SoC. It wasn't and then the Switch was dubbed "the worst console ever" yet it's a run away success (basically, most of the pro-AMD guys here are too blind to listen when even the President of Nintendo said "we're going ARM" but what does the president know, ATF posters are experts).

I don't mean NV is going to be in the next gen consoles, but I do mean NV can shake the buck enough that they might be considered for the gen after, or at least shaking the PC/API landscape enough that AMD's limited resources are but through more strain as they now have to try to catch up on that front.

NV is playing 4D chess here, and RT can fizzle or become the new standard. Interested to see where it goes.

Consoles are underpowered w.r.t. the CPU, not so much the GPU, especially the Xbox One X or PS4 Pro, with the GPUs in them basically equivalent to a GTX 1060 3/6GB. The next gen console design will in all likelihood use Zen 2 along with whatever AMD's next GPU is going to be. Nvidia got to supply the SoC for the switch simply because AMD doesn't make mobile SoCs.

 

[ozzy702]

The name of a chip, is just as arbitrary as the name of a card. It's just a name, assigned by NVidia.

They are simply splitting to more chips this time, x80 and x70 get separate chips. A different chip at the X70 level, doesn't mean it's an x60 card. That is just nonsensical.

When 1080 had 314 mm2 die, it was economically reasonable to still sell a higher volume, lower priced x70 based on the same chip.

Now that that the 2080 uses a MASSIVE 545 mm2 die, it has no place in the higher volume, lower priced(relatively) 2070. They NEED the die size reduction this time since the parent die is so large.

One more thing. Note that they maintained the 256 bit memory bus as well. Not going to the same 192 bit bus used on the last x60 card.

So the 2070:
Guzzles power like x70 series,
Performs like a x70 series,
has a 256 bit memory bus like a x70 series,
priced like a x70 series
Actually named an x70 series.

But it's really a nefarious NVida plot selling a x60 as x70 series because of the name stamped on the chip.

Because it's not like there is a perfectly logical reason why they need another chip this time specifically for x70 serise. Oh wait, there is.




Uh no. How can it be a "4th tier chip" when they only have 3 chips and 3 card tiers. Only by twisting your view of reality to fit your strange idea that the 2070 is a x60 card.

I'm with you on everything except priced like an X70 series. Absolutely not the case, especially since the MSRP this go around will NEVER be the price we can actually pick up a card at. Pricing for the 2000 series is across the board absurd, with the 2070 being the worst of the bunch.

 

[Headfoot]

Yeah I can see CPU being underpowered, but the GPUs in the latest console refreshes are really pretty powerful as far as consoles go

 

[tviceman]

Talking about this with the wife, and I basically think it's NV's attempt to usurp AMD's foot hold in consoles. Consoles this gen were very under powered, which led to PC ports with relatively lax PC requirements. Something that's bad for any company trying to sell you overpriced toasters. IF, NV can get MSFT to work up Ray Tracing, and can sell it to enough devs through their dev-rels (they have a better chance of RT catching on than Rapid Math or that word that should not be mentioned (primitive shaders)) they create an opening down the road to either A) bolster PC requirements for games (even consoles) leading to being able to sell more expensive toasters, or B) shaking AMD's grip on the consoles.

I mean practically everyone and their mother here confirmed Switch was going to be an AMD based SoC. It wasn't and then the Switch was dubbed "the worst console ever" yet it's a run away success (basically, most of the pro-AMD guys here are too blind to listen when even the President of Nintendo said "we're going ARM" but what does the president know, ATF posters are experts).

I don't mean NV is going to be in the next gen consoles, but I do mean NV can shake the buck enough that they might be considered for the gen after, or at least shaking the PC/API landscape enough that AMD's limited resources are but through more strain as they now have to try to catch up on that front.

NV is playing 4D chess here, and RT can fizzle or become the new standard. Interested to see where it goes.

If Nvidia wants RT to succeed much sooner than later, they need to go all in on it. Not half-assed physx, but they need a significant number of AAA titles to support it. They will have to look past short term shareholders and take a temporary hit on their margins.

Already, based on Turing pricing, it doesn't look like any of the above is going to happen. Turing adoption will be slower and lower than Pascal was over Maxwell, less than Maxwell was over Kepler, etc. It's hard to push features for hardware that is priced out of 80% of the potential market.

Turing is likely going to end up too forward looking, too expensive, and too underpowered for most of it's newest features.

 

[LTC8K6]

The new features of NVIDIA Turing architecture

https://videocardz.com/77895/the-new-features-of-nvidia-turing-architecture

 

[ub4ty]

Must be a 1080 Ti owner...
Been gaming a lot longer than you have. I like 4K. I like 4K to be locked at 60FPS with some eye candy on. Still need AA on a 43 @ 2ft.
Show me a single card that can do that.
Hint: it’s not the 1080 Ti

Also, we haven’t paid any luxury tax. These are the new prices. nVs top tier cards have NEVER gotten cheaper.
Do I like the price? Of course not, it sucks. But I like 4K. I’ve been waiting for this gen since early 2015 as I knew the 10 series wasn’t the gen that could do 4K well.

Then don't buy it if the price sucks...

> Atari
> Sega Genesis
> DinoPark Tycoon
> Doom / Quake / Duke Nukem (Dos Mode reboot)
> SC 1 / CS / CS source
> Fortnite / Dota / etc etc

Yeah, I never paid for the most high end GPU. You don't need to. The majority don't. You don't need 60 FPS.
You most definitely don't need 4k. The majority of the major game titles people play can be played on a 1060 or lower. I play @ 2560x1440 on a $140 2GB GPU. It's a champ.

The minority (between 1-2%) own a 1080ti and no sane game developer makes a game for such an exclusive group. They'd go bankrupt if they did.

Resolution?
https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam
1024 x 768 - 0.65%

1280 x 720 - 0.42%

1280 x 800 - 0.85%

1280 x 1024 - 2.19%

1360 x 768 - 1.97%

1366 x 768 - 14.18%

1440 x 900 - 3.62%

1536 x 864 - 0.31%
1600 x 900 - 3.68%

1680 x 1050 - 2.63%

1920 x 1080 - 60.66%
~91% of people are at OR below 1920x1080

No one's gaming at 4k .. Again, less than 2% of people fall into such use cases.
I've been gaming for a long time and what I long for are fun games that I enjoy not Umbra, penumbra and antumbra and reflections I never pay attention to when I'm trying to get the best K/D ratio. The 1%/2% will never determine the broader market. Nvidia has made a huge mistake by setting a precedent whereby they establish such prices. This is why people are mad. Hardware accelerated ray tracing was not done by Nvidia. It was done by imagination technologies years ago in a mobile phone processor envelope. I guess the 1-2% will be buying these cards and I don't mean that in terms of wealth. Working in tech where people make gobs of money producing this technology, I see some of the most dated cellphones and personal computer hardware. Meanwhile, I see people who can least afford it with $1,000 iphones likely paying installment payments. It's not about that though. Do as you please with your money. People are upset about Nvidia trying to sneakily set a precedent of far more increased prices for gaming hardware. There are consequences to this. If you claim you've been gaming forever, you're recall the deathly period of the PC and PC gaming in which people flocked to consoles. Every game developer remembers it. And its the reason why the most popular games don't need such hardware to run. You can run fortnight at 120fps @ 1920x1080 on a $150 2400G APU from AMD. Guess what is all the rage?

Everyone should be completely bashing Nvidia for this move. Thankfully only a small minority (1-2%) are screaming : Take my money

 

[ub4ty]

I'm still gonna wait to see benchmarks. I may not upgrade at all this year either.

In 2 days, I am going to parse their Turing micro-architecture presentations to see what kind of special low R&D costing gimmicks are behind this new card release. In a week, (thanks for the delay Nvidia putting out reviews a day before shipment), I'll be looking at benchmarks to get laughs in.

 

[ub4ty]

Another thought (I haven't had my morning joe so bear with me here): Nvidia took this radical departure from their established execution cycle because... their arch is starting to max out.

Aside from the addition of the RT/TN cores, the word on the web has been that the SM structure of Turing is not a radical departure from either Pascal or Maxwell.

The evolution of NV's arch is dead, because as far as rasterization workloads are concerned, they have effectively perfected it. From this point forward, it was going to be brute force gains reliant on die shrinks, which we have seen are becoming a less and less reliable as foundries begin knocking on God's door with their lithography process.

AMD, with fewer resources and playing second fiddle in a number of markets, capped out their current arch with the Vega configuration and simply doesn't have the funds for a major overhaul.

Perhaps NV finds itself staring at the same door (in their case, their arch is already so good that overhauling it just doesn't pass the $/reward muster). So instead they go all in with RT/TN, leveraging the gap they've placed between themselves and their competitor.

And mature products that are hitting their limits see prices fall not rise.
Slapping in a tiny asic portion Ray trace pipeline and meme cores is a business gimmick to try to keep margins and profits alive. Magically, they decided to increase them even further. SM structure is basically Volta with FP64 cut out w/ a tiny RT pipeline added in.

 

[PeterScott]

I'm with you on everything except priced like an X70 series. Absolutely not the case, especially since the MSRP this go around will NEVER be the price we can actually pick up a card at. Pricing for the 2000 series is across the board absurd, with the 2070 being the worst of the bunch.

Fair point. It's priced above a x70 series historicals. But that doesn't help the argument that it is really an x60 card masquerading as x70 series.