NVIDIA GeForce RTX 2080 Ti features 4352 CUDA cores, and new 20 series Card From Gigabyte pictured!

[AdmiralMeowmix]

https://www.pcgamer.com/leaked-and-possibly-fake-images-of-geforce-rtx-2080-ti-cards-are-popping-up/

 

I mean I believe it's faked tbh, 11 GB's for one seems small for this generations TI, I would presume they'd want to go with a largewr amount of memory to make it sound more compelling over the current gen to people who don't know the different in GDDR's and such.

[Firewrath9]

6 hours ago, Froody129 said:

A 1060 with a decent chip is fairly easy to get to about 80w already if you undervolt. 

 

Personally I'm going to be sticking with my 980ti for a few years still. I'll pick up a used card again in the future but for now I'm very happy.

 

This looks to be the usual cycle of dropping the performance one tier down. So 1080ti performance at 2080 price etc. 

If its like the past, 1070=980 ti, 780 ti = 970,

1080 ti = rtx 1070 or 1070 ti

 

 

 

 

 

 

hopefully.

[PrinnyExplodes]

4 hours ago, AdmiralMeowmix said:

https://www.pcgamer.com/leaked-and-possibly-fake-images-of-geforce-rtx-2080-ti-cards-are-popping-up/

 

I mean I believe it's faked tbh, 11 GB's for one seems small for this generations TI, I would presume they'd want to go with a largewr amount of memory to make it sound more compelling over the current gen to people who don't know the different in GDDR's and such.

I think it may actually be pretty accurate. Today on the WAN show apparently Linus and Luke were talking about how a spec leak a few days before the official Nvidia new gen launches have been true for most the GTX series. They speculated that the size might be due to the lack of competition in the market and Nvidia just giving a "that will do". 

 

I am actually curious to see what the actual performance difference of GDDR6 can do

 

EDIT: We know the pricing now that PNY has posted it on their website and took it down. GTX 2080 ti is $1000 USD. Here is some screenshots before they took it down according to pcmr users https://drive.google.com/drive/folders/1apXOnZCfaccafMEw2qw-Q7WXtzkrWTOx ....don't take it for fact but it does look very believable

 

 

[Maslofski]

so they dont even beat the 5120 cuda cores of the titan v

[Lathlaer]

I have a problem with 2080ti right now. I'd get those but there is a part of me that don't trust NVIDIA not to release something bigger in 6 months time

[Cedimedi]

7 hours ago, Locutus494 said:

And how can you know that yet?

WHY do people keep insisting that SLI is dead?! Nothing could be further from the truth!

How could you know that too? 

 

PNY just leaked specs of the TI and it is indeed true that SLI is gone and replaced by NVlink 2 way. 

 

[Lathlaer]

10 minutes ago, Cedimedi said:

PNY just leaked specs of the TI and it is indeed true that SLI is gone and replaced by NVlink 2 way. 

You know, whenever people say that SLI is dead, it being replaced by another form of multi GPU bridging is not what they have in mind.

[Bananasplit_00]

8 hours ago, Locutus494 said:

WHY do people keep insisting that SLI is dead?! Nothing could be further from the truth!

because its garbage and NVIDIA are trying to kill it? scaleing isnt great and after NVIDIA dumped the responsebility for how well it works onto the devs its gone down the drain tbh. NVLink would make a lot of sense and pretty much all the leaks we have would point towards NVLink being "the new SLI" NVLink is a lot better in pretty much all ways and has been on the quadros for quite a while now so it would make sense to bring it to the consumer cards as its a much better technology

[pas008]

39 minutes ago, Bananasplit_00 said:

because its garbage and NVIDIA are trying to kill it? scaleing isnt great and after NVIDIA dumped the responsebility for how well it works onto the devs its gone down the drain tbh. NVLink would make a lot of sense and pretty much all the leaks we have would point towards NVLink being "the new SLI" NVLink is a lot better in pretty much all ways and has been on the quadros for quite a while now so it would make sense to bring it to the consumer cards as its a much better technology

Dead?

Nvidia dumped the responsibility? Think dx12 kinda did that

Heres sone numbers for you

https://babeltechreviews.com/the-gtx-1070-ti-sli-vs-gtx-1080-ti-performance-review-35-games-tested/3/

[leadeater]

3 hours ago, Lathlaer said:

I have a problem with 2080ti right now. I'd get those but there is a part of me that don't trust NVIDIA not to release something bigger in 6 months time

Same, I kind of want to upgrade but realistically don't need to but as things stand right now I'm very hesitant to buy anything. Just feels like I'll get screwed in 6 to 12 months and with the way prices are right now I don't want to waste that kind of money.

[Taf the Ghost]

14 minutes ago, leadeater said:

Same, I kind of want to upgrade but realistically don't need to but as things stand right now I'm very hesitant to buy anything. Just feels like I'll get screwed in 6 to 12 months and with the way prices are right now I don't want to waste that kind of money.

We had legitimate leaks that Ampere was coming soon, which now seems clear to either be the next generation on 7nm or the Compute Cards on 7nm. It's all but assured the 2080 Ti is launching at or near the same time as the 2080, which means the RTX Titan would need to be a cut-down version of the TU100 GPU. Nvidia is thus not leaving themselves any room for cutting down their large dies for high-end consumer.

 

AdoredTV's leak has been right on the nose for nearly all of the details, which means it points to there only being one more actual GPU Design for the Turing family. The TU106/116/117 is probably going to be the exact same die. (Should be 15, 10 and 7 SMs.) Which means that die is going to be replaced on 7nm well before the normal 2 year GPU generation is up, but Nvidia will wait on AMD so as to get as much money out of the Turing design as possible. (Ampere for consumer is likely just a die-shrink Turing.)

 

So, for people wanting to get into High-end GPUs, especially those still on the 900 series, should be great for them. If you're sitting on at least a 1070, it's a lot of money for not a huge uplift. Unless you go all the way up the SKU stack to the Ti.

[Lathlaer]

@Taf the Ghost yeah but I don't want to underestimate NVIDIA. The whole Titan XP -> 1080ti -> Titan Xp sequence is still fresh in my memory ;-)

[Taf the Ghost]

2 minutes ago, Lathlaer said:

@Taf the Ghost yeah but I don't want to underestimate NVIDIA. The whole Titan XP -> 1080ti -> Titan Xp sequence is still fresh in my memory ;-)

While I agree, we'll be able to know more after the official spec releases. If they simply don't have any larger dies available, they aren't going to be able to slice up the market more. Which points to thoughts from a number of people that we're going to get a hybrid generation next year, with a replaced lower tier and the upper end not changed. 

 

Which is to say the next RTX Titan, not the one that'll be out by year's end for 3k USD, is the one they're going to really abuse like the 1080 Ti generation.

[GoldenLag]

1 hour ago, pas008 said:

Dead?

Nvidia dumped the responsibility? Think dx12 kinda did that

Heres sone numbers for you

https://babeltechreviews.com/the-gtx-1070-ti-sli-vs-gtx-1080-ti-performance-review-35-games-tested/3/

So dead due to dx12?

 

SLI and Crossfire is dead in the way it scales poorly and creates extra issues to worry about. Not to mention the minority who buy high end cards and the minority of the minority who runs them im SLI.

 

Polaris was meant to be good in SLI, but you dont see people running dual 480s or 580s do you?

[Taf the Ghost]

3 minutes ago, VegetableStu said:

dear EVGA,

 

BARF. WTF?

 

sincerely,

1997

Good point. It looks like a wider version of what you'd get in the late 90s, as heatsinks used to need to be a lot smaller. All it needs is some render of a female on it to jump to the mid-2000s.

[GoldenLag]

7 minutes ago, VegetableStu said:

dear EVGA,

 

BARF. WTF?

 

sincerely,

1997

I think they are going for the "seethrough feel" of the nintendo devices back in the day. Nothing beats a seethrough gameboy

[GoldenLag]

1 minute ago, VegetableStu said:

but that look did not age well at all!

I want my next phone to have that kind of look. The sort of thing jerryrigeverything does. I know its niche but to me its amazing. Also seethrough ninendo 64 controller with green tint.

[Taf the Ghost]

2 minutes ago, VegetableStu said:

 

but that look did not age well at all!

I imagine it's supposed to light up with the RGB.

 

 

Not quite sure that's the effect they're looking for, though.

[GoldenLag]

3 minutes ago, VegetableStu said:

...

 

NO. NO.

How about blue LEDs?

 

[Amazonsucks]

Not a single mention of literally the only thing worth buying this card for. Does it have RT and Tensor Cores like the Quadros???

[Taf the Ghost]

1 minute ago, Amazonsucks said:

Not a single mention of literally the only thing worth buying this card for. Does it have RT and Tensor Cores like the Quadros???

Yes, as it'll be from the same die.

 

We're not sure if those RT Cores do anything in 99% of games, but we'll find out soon enough.

[Amazonsucks]

3 minutes ago, VegetableStu said:

the 2080ti would probably be the same "card", but are they doing the Big Turing Little Turing thing like what they did with Pascal? o_o

I would assume, and i could be wrong here, that the Turing with the 384bit bus is the GT102 medium chip, the 256 bit is the small GT104, and we havent seen the big chip which is the GT100.

 

Thats how it is now with Pascal, GP100, GP102 and GP104. Volta only ever had the one big GV100 chip.

 

Ever since they moved to HBM on their high end chip, the 102 has been basically a bigger 104(no ECC registers and no FP64).

[Taf the Ghost]

17 minutes ago, VegetableStu said:

the 2080ti would probably be the same "card", but are they doing the Big Turing Little Turing thing like what they did with Pascal? o_o

At the moment, it looks like there is 4 dies: TU100, TU102, TU104 and TU106/116/117. The 100/102/104 are the Quadro Cards that were announced. By dies, it looks like TU100 will be the RTX Titan for ~3k USD. TU102 will be 2080 Ti. TU104 will be the 2080 + any future 2070 TIs. The TU106 looks to be the 2070 along with all of the 2060/2050 parts. (TU106 at 15 SM would let them cut down all the way to the 2050 without having to make a separate die.) 

 

What it looks like is that Nvidia has moved the Titan up-market, again, so they can sell off their supply of partially defective Quadros into a somewhat larger market. That means the Titan has moved up a Die Size. They've pushed the Ti up a die size as well. That's what happened with Titan X(p) -> 1080 Ti -> Titan XP. 1080 Ti was just a higher clocked version of the first Titan, while the second was the fully active die.

 

Without waiting on the 2080 Ti, it points to this generation being shorter for Nvidia. We can assume 7nm parts are taped out already, so they can bring online the 7nm generation whenever it's feasible in the market. That's going to be before H2 2020, which is the time frame they'd normally launch another generation.

[Taf the Ghost]

29 minutes ago, VegetableStu said:

the 2080ti would probably be the same "card", but are they doing the Big Turing Little Turing thing like what they did with Pascal? o_o

Other thing to keep in mind is this "12nm" node is actually semi-custom to Nvidia. They paid TSMC for certain advances on the 16nm FinFET node, so we can expect Nvidia to exploit the node to its fullest possible extent.  This points to some funky stuff with some of these designs, as we can expect them to have a longer life. The TU104 die I fully expect to be still sold by Nvidia to possibly the 4000 generation. Especially if it has those RT cores and Ampere is not much more than a die shrink. It just becomes the 3070 & 3060.

[leadeater]

14 minutes ago, Taf the Ghost said:

At the moment, it looks like there is 4 dies: TU100, TU102, TU104 and TU106/116/117. The 100/102/104 are the Quadro Cards that were announced. By dies, it looks like TU100 will be the RTX Titan for ~3k USD. TU102 will be 2080 Ti. TU104 will be the 2080 + any future 2070 TIs. The TU106 looks to be the 2070 along with all of the 2060/2050 parts. (TU106 at 15 SM would let them cut down all the way to the 2050 without having to make a separate die.) 

 

What it looks like is that Nvidia has moved the Titan up-market, again, so they can sell off their supply of partially defective Quadros into a somewhat larger market. That means the Titan has moved up a Die Size. They've pushed the Ti up a die size as well. That's what happened with Titan X(p) -> 1080 Ti -> Titan XP. 1080 Ti was just a higher clocked version of the first Titan, while the second was the fully active die.

 

Without waiting on the 2080 Ti, it points to this generation being shorter for Nvidia. We can assume 7nm parts are taped out already, so they can bring online the 7nm generation whenever it's feasible in the market. That's going to be before H2 2020, which is the time frame they'd normally launch another generation.

I'm still a little dubious that the RTX 8000 is actually using the biggest Turing die, without an announced Tesla card with HBM I'd say be cautious in assuming any products are actually using a TU100. Not that it's a great indicator but the given Turing die specs which are smaller than Volta and that it's using the same process there could be a different die configuration for the Tesla cards, not that it matters greatly if that is the case because if true no gaming card will be based off of it.