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

[Ottonomous]

NVIDIA GeForce RTX 2080 Ti and RTX 2080 ‘official’ performance unveiled

https://videocardz.com/77983/nvidia-geforce-rtx-2080-ti-and-rtx-2080-official-performance-unveiled

This is with no DLSS on, that's pretty good.

Bit better than expected but the AI focused INT/FP performance is showing, and the 2080 won't be a true 4K card. Enough of an incentive to push 1080ti users to the Ti.

 

[Ottonomous]

The rumored 2080+ might also make a comeback as maxxed out TU104. All that's left is to see how the 2060 will be positioned as a midrange RTX option

 

[Head1985]

Looks pretty good. 2080 looks more consistently above 1080Ti than I expected.

Of course some are jumping on some specific benchmarks to complain something is wrong with 1080Ti scores, but NVidia is obviously comparing to 1080ti FE cards, not faster AIB cards.

I am not going to track all the individual benches down, but I checked the overlap with the Anand 1080Ti FE review.
https://www.anandtech.com/show/11180/the-nvidia-geforce-gtx-1080-ti-review/10
https://www.anandtech.com/show/11180/the-nvidia-geforce-gtx-1080-ti-review/13

For the games I spotted in both sets: Witcher 3 and Hitman, there seems essentially the same 1080-1080Ti delta as in the NVidia guide numbers, so they don't appear to be sandbagging the GTX1080Ti numbers to make 2080 look better.

So you guys think 15% gap in some games and 23% average in those 7games in 4K SDR graphs is normal between 1080TI and 1080?1080TI have 40% more SP and 50% more bandwidth and 38% more rops.
Well lets see the gap in techpower up review.They didint test all games sadly.

Bf1 NV-15% gap between 1080 and 1080TI

Techpower up-32% gap between 1080 and 1080TI

Middle earth-NV-27% between 1080 and 1080TI

techpower up-38% gap between 1080 and 1080TI

With ultra textures-43% gap between 1080 and 1080TI

Another crazy results are SW Bf2-15% gap but techpower up didnt test it.And mass effect 19% gap again techpower up didnt test it.
You dont need to be rocket scientist to see that NV did something to 1080TI to rtx2080 look better.
Other 4K normal results looks +- OK but 2080 dont look that impresive there....

 

[alcoholbob]

So you guys think 15% gap in some games and 23% average in those 7games in 4K SDR graphs is normal between 1080TI and 1080?1080TI have 40% more SP and 50% more bandwidth and 38% more rops.
Well lets see the gap in techpower up review.They didint test all games sadly.

Bf1 NV-15% gap between 1080 and 1080TI

Techpower up-32% gap between 1080 and 1080TI

Middle earth-NV-27% between 1080 and 1080TI

techpower up-38% gap between 1080 and 1080TI

With ultra textures-43% gap between 1080 and 1080TI

Another crazy results are SW Bf2-15% gap but techpower up didnt test it.And mass effect again techpower up didnt test it.

2080 Ti and 2080 dual axial open air cards are being compared to blower founders cards and you have no idea what case or thermals conditions were used or how well those cards boosted, maybe they were duds.

Ultimately the numbers they put for 1080 Ti or 1080 they put are irrelevant, the only thing that matters is the absolute numbers of the 2080 Ti and all you have to do is compare those numbers in an apples to apples way like 2080 Ti open air vs 1080 Ti open air. It's pretty simple math, we are looking at around ~20% avg gain in DX11 and a bit higher in DX12.

 

[Head1985]

The numbers 1080 vs 1080TI is all that matter because that is how you can see they did something to 1080TI in some games where the gap is crazy low like 15% or 20%(Most 4k SDR results)
There is no way 1080Ti is only 15% faster than 1080 in 4K/ultra.No way.
BOTH 1080TI and 1080 ARE FE in NV and techpower up reviews

Edit: YES witcher3 is one of the few games where the performance of 1080TI looks normal- 32% faster than 1080.
Also rtx 2080 looks pretty bad there only 8% faster than 1080TI FE.AIB 1080TI will have same performance or even will be slightly faster

 

[PeterScott]

The numbers 1080 vs 1080TI is all that matter because that is how you can see they did something to 1080TI in some games where the gap is crazy low like 15% or 20%(Most 4k SDR results)
There is no way 1080Ti is only 15% faster than 1080 in 4K/ultra.No way.
BOTH 1080TI and 1080 ARE FE in NV and techpower up reviews

In a couple of games, and in a couple of other games there are essentially the same as other reviews.

Previously it was already assumed that 2080 was only about the same as 1080Ti. With these results it looks more like 2080 best 1080Ti. Not by a huge margin, but it's some margin and it will likely remain in 3rd party testing, even if it shrinks a couple of percent.

 

[AtenRa]

It doesnt look good not good at all, RTX2080 even if it is 10% faster vs GTX 1080Ti after both are OCed its still a big fail because perf/$ its taking a BIG dive. Not to mention that Pascal its 2.5years old by now. The only card that perhaps someone would even consider buying is the RTX2080Ti but at the MSRP price its not what many out of few people that are buying High-End cards will put their hand in the pocket for it.

NVIDIA should of wait for 7nm to bring any RT cards in the consumer market.

 

[Head1985]

In a couple of games, and in a couple of other games there are essentially the same as other reviews.

Previously it was already assumed that 2080 was only about the same as 1080Ti. With these results it looks more like 2080 best 1080Ti. Not by a huge margin, but it's some margin and it will likely remain in 3rd party testing, even if it shrinks a couple of percent.

Yes 1080Ti perfomance in couple of game is lower than it should be.And in those games RTX2080 looks better than it should be(BF1, F1 2018, mass effect andromeda,Shadow of war,SW BF2)
On other hand where the 1080TI performs how it should(30+ faster than 1080) rtx2080 dont look that good(witcher3, PUBG,Tomb raider, hitman)

Also is clear that turing will be faster in dx12/vulcan games(look at wolf2)because it support things like Fp16 and also in AMD games(rainbow six) so they didnt need fake 1080TI results in those games.

 

[happy medium]

The video cards article also shows the settings that were used in each game, do they match the settings used in the reviews you guys are comparing?

 

[Head1985]

The video cards article also shows the settings that were used in each game, do they match the settings used in the reviews you guys are comparing?

I was comparing 1080 vs 1080TI gap.Nv fake 1080TI numbers in some games so 2080 look better.
With normal performance(where 1080Ti is 30+faster than 1080) 1080TI was very close to 2080.With fake numbers(where 1080TI is only 15-20% faster than 1080) 2080 looks much better.
There is simple no way 1080TI is only 15% faster than 1080 in 4K/ultra.So they must have fake it.

 

[Thala]

I could see things like dense foliage being a stumbling block as well. Training an AI to recognize a single tree isn't too difficult, training it to recognize that there are individual trees among multiple trees with criss-crossing branches and a mesh of leaves is a lot more difficult.

Thats not how it works. You will not train the network to recognize the trees. You train the network to detect and generalize properties of the image ,which are much more abstract than detection of concrete objects.

 

[tviceman]

I'm LOLing at all the posts about performance failures and Nvidia sandbagging their old cards. The prices suck, but there are no conspiracies going on.

Die sizes ballooned because of the RT and tensor cores, AMD is holding their tiny man parts, and here we are with insane prices.

Nvidia will have 6 chips as fast or faster than Vega - GP104, GP102, GV100, TU102, TU104, and TU106.

 

[Muhammed]

There is no way 1080Ti is only 15% faster than 1080 in 4K/ultra.No way.

I now know why. VideoCardz posted the wrong 1080 numbers. 1080 scores lower @4K in the review guide.

 

[sze5003]

B&H has the evga ftw3 for preorder but you can't use the checkout until 8pm. It's for the preorder. I would definitely want a custom card though. I don't mind waiting until November actually if that's the case.

 

[Kenmitch]

Less than a week to go and you guys can see the good, the bad, and possibly some ugly. The reviews should be interesting, but I'm thinking the best stuff will be in the comment sections.

 

[Hitman928]

Thats not how it works. You will not train the network to recognize the trees. You train the network to detect and generalize properties of the image ,which are much more abstract than detection of concrete objects.

Yes, I know how training a neural network works, I've used them. My point still stands.

 

[Markfw]

I want to see F@H ppd before I buy any 208x video card for my folding.

 

[Elfear]

I'm LOLing at all the posts about performance failures and Nvidia sandbagging their old cards. The prices suck, but there are no conspiracies going on.

Die sizes ballooned because of the RT and tensor cores, AMD is holding their tiny man parts, and here we are with insane prices.

Nvidia will have 6 chips as fast or faster than Vega - GP104, GP102, GV100, TU102, TU104, and TU106.

Some of the comments were actually pretty spot on pointing out the disparity in Nvidia's numbers and independent reviewers. The performance gap between the 1080 and 1080Ti certainly looks strange considering they are very well known quantities by now. Can't blame you for questioning a conspiracy theory but did you have to add in the divisive comments about AMD? They weren't really part of that whole line of discussion so why fuel the flames when it's already hard enough having useful and constructive discussions here?

 

[Kenmitch]

I want to see F@H ppd before I buy any 208x video card for my folding.

Not sure if you'll find them on the initial reviews. You could see if anything in the posted review guide might mention it. Maybe a non sponsored independent review will satisfy your quest for knowledge.

 

[DaveSimmons]

It doesnt look good not good at all, RTX2080 even if it is 10% faster vs GTX 1080Ti after both are OCed its still a big fail because perf/$ its taking a BIG dive. Not to mention that Pascal its 2.5years old by now. The only card that perhaps someone would even consider buying is the RTX2080Ti but at the MSRP price its not what many out of few people that are buying High-End cards will put their hand in the pocket for it.

NVIDIA should of wait for 7nm to bring any RT cards in the consumer market.

True for people who already own a 1080 ti.

For the rest of us: if I decide to upgrade my 980 ti this year I'll probably get the 2080. Faster than a 1080 ti, 25 to 35 watt lower TDP, and some bleeding edge features I'll probably never benefit from. I can easily afford the 2080 ti but I'm not running at 4K and RT is probably going to be hardware PhysX 2,0.

Of course the post-NDA reviews might change that.

 

[AtenRa]

True for people who already own a 1080 ti.

For the rest of us: if I decide to upgrade my 980 ti this year I'll probably get the 2080. Faster than a 1080 ti, 25 to 35 watt lower TDP, and some bleeding edge features I'll probably never benefit from. I can easily afford the 2080 ti but I'm not running at 4K and RT is probably going to be hardware PhysX 2,0.

Of course the post-NDA reviews might change that.

Well, in your case I would go with a used GTX1080ti at 500-550USD, you will save 300+ USD and have the same performance and still the card will be under guaranty.
If RTX2080 was selling for 499-550 it would be the best choice but for 800USD with this level of performance it is not, especially since you are not going to use any of its new features.

 

[happy medium]

NVIDIA DLSS Explained – Much Higher Quality Than TAA or Much Faster Performance, Delivered by NVIDIA NGX

https://wccftech.com/nvidia-dlss-explained-nvidia-ngx/

 

[happy medium]

The NVIDIA Turing GPU Architecture Deep Dive: Prelude to GeForce RTX


https://www.anandtech.com/show/13282/nvidia-turing-architecture-deep-dive

 

[cmdrdredd]

Personally I am not buying a 2080ti but I want to see real gains in games I play now at 4K without DLSS etc. if I understand right the game has to have DLSS support or it doesn’t work. I’m not going to use Ray tracing at 4K and I’m not going to use 1080p for that. So my interest is how much they gain without the fancy proprietary features as it would be more relevant to me. Again I’m not going to buy any flavor of card this go around because my 1080ti still feels fine at the moment but just to satisfy my curiosity this is the info I’m interested in.

I imagine the gains will less than some of the slides suggest with base software (that is no super sampling etc)

 

[SPBHM]

I'm still very curious to see how viable the 2070 is at RT, being in many aspects almost half of the 2080 ti and with the reports of the 2080 ti struggling on early demos.

overall it is a very interesting release by Nvidia, the new tech is very exciting, performance itself seems to be solid but, it's not a 8800GTX moment or anything, so it's mixed, when you consider the pricing, not very positive, but when you take into account the huge dies and so on, maybe less bad.