Nvidia ,Rtx2080ti,2080,(2070 review is now live!) information thread. Reviews and prices

[krumme]

TechPowerUp's review shows the GTX 2080 only marginally beating the previous-generation GTX 1080 in perf/watt. Kepler->Maxwell and Maxwell->Pascal were much bigger jumps.

I agree that this was probably meant to be on a smaller node and was back-ported due to unavailability, immaturity, and/or cost. I think Nvidia wanted the full TU104 to draw 185W or so like all the other '4'-series chips before it, but had to bump that to 225W because of the node regression. That said, I think Nvidia does deserve at least some credit for doing a successful backport instead of just delaying or cancelling like Intel did with 10nm products.

Yeaa. In general there is much good to say about the tech in these cards for sure. As there was with fermi. Its forward looking cards with perspective and we havnt seen that from nv in years. I respect that deeply. They have the same power troubles as fermi but what pisses me off is still that crazy price. Its the principle.

Granted i would buy it to play bf5 and nv is not strong here and i dont care for rt and dlss or whatever its called is not used in bf5 as i understand it.
Might as well take an aib 1080ti even if its only say 15% faster than an vega 64 in bf series.
I was hoping the 2080 would do it but imo its the same old. I will cancel that for sure. Hold onto the 2080ti even if i have my doubt now. No zero fan idle is a huge minus in my setup. Gaaa. Will wait and see.

Didnt see any bf5 bm skimming some of the bm. Anyone?

 

[railven]

Well, they're not really wrong. If you're just looking at a $700 1080 Ti vs an $800 2080, the latter is probably the better choice. They were also cautionary on buying the new cards based on the promise of new features.

We're just not used to seeing a GPU released at almost the same price point and with similar performance and price/perf as a previous generation card released 18 months earlier. GM can get away with releasing a 2019 that has virtually the same price and performance as a 2018 model and people will still buy them because they need a new car. GPUs don't really wear out in the same way though, and there's only so many cards you can sell when you offer consumers no incentive to upgrade in their price bracket. We'll see how it works for Nvidia.

Kind of the analogy I was making in the other thread. You don't need a new GPU to play current or newer games. This product is clearly aimed at the disposable income buyers, not the mainstream.

Outside of random issues, it's a one time purchase with negligible maintenance, so its a flat $1200 that will most likely last me two years. I spent about the same going from 980 Ti -> 1080 -> 1080 Ti. Told the wife for the next series I'll probably just snag the Titan from the get go and be covered for the duration of the product life. NV must have heard me because they're selling me the x80 Ti at the Titan price.

The outrage is all over the internet, and rightly earned. I don't think NV really cares as they're going to sell these cards, period. 7nm might bring things back to normal, or not. Either way, I can't even confirm if I'm getting a card now haha.

It is always amusing though. Watching the flip flopping positions people take.

 

[zinfamous]

Another example from the FF demo:
https://www.kitguru.net/components/...idia-rtx-2080-founders-edition-8gb-review/13/

DLSS is more AA, but softer. Look at the instrument panel.

Yeah, I don't agree that DLSS is better in that cropped image. The softening removes all the texture. Look at the seat and the hair.

 

[badb0y]

These cards are basura.

20-30% boost and you want me to fork over $1200...lol.

 

[realibrad]

Just look at the headrest and see all the texture it removes. Lots of aliasing with TAA vs DLSS, but, DLSS makes everything too soft in this shot for me.

 

[Head1985]

Something fishy in techpower up review.Looks like 1080TI is slower than it was in their latest review.Maybe Nvidia "magic" drivers?Btw good way how compare both MSI gaming trio cards.
Witcher3-100 vs 97,8fps

GTA5-162 vs 157

Hitman-128 vs 120fps

Prey-130 vs 127fps

anyway more results AIB vs AIB

 

[MrTeal]

What's your Pascal SLI experience been like? I am going back to SLI after skipping it for the first time in a decade last cycle due to 4k 144hz needs

It's hit or miss, but when it works it's great. Unfortunately it's been a pretty busy year, and really the only things I've devoted much time to are F1 2016 and finally finishing Brood War, neither of which really call for the kind of horsepower I have available.

 

[PeterScott]

Just look at the headrest and see all the texture it removes. Lots of aliasing with TAA vs DLSS, but, DLSS makes everything too soft in this shot for me.

I have seen DLSS screens from a number of reviews now. Slight edge to TAA, but it is very sensitve to precise shot alignment and motion. Here is one that has three shots: TAA/DLSS/OFF:
https://pclab.pl/art78828-21.html

Both TAA and DLSS soften textures. Look at the girls jacket/jeans with AA off vs either AA method.

While actually playing the game, I doubt much difference would be spotted between DLSS/TAA. They both soften things and reduce Aliasing by a similar amount.

 

[Guru]

Yeah, I don't agree that DLSS is better in that cropped image. The softening removes all the texture. Look at the seat and the hair.

That's probably why it is faster, because it actually removes the detail and reduces texture quality.

I don't know what cool aid the writer was drinking, but as you said the TAA image looks so much better, due to the fact that DLSS removes actual details. Details on the seats, hair and dashboard all removed.

The RTX series is really fast, I mean generally the 2080 beats the 1080ti by 3% to 5% on average, the 2080ti is over 30% faster over the 1080ti at 4k, about 20% faster on 1440p(though it's clearly bottlenecked by cpu), so from a purely performance perspective they are solid, the issue is they are atrocious value, in fact when compared to their counterparts they hold NEGATIVE performance per dollar.

 

[Malogeek]

Something fishy in techpower up review.Looks like 1080TI is slower than it was in their latest review.Maybe Nvidia "magic" drivers?Btw good way how compare both MSI gaming trio cards.

Could be a result of Intel Spectre/Meltdown patches?

 

[crisium]

Just look at the headrest and see all the texture it removes. Lots of aliasing with TAA vs DLSS, but, DLSS makes everything too soft in this shot for me.

Yeah, Nvidia oversold DLSS (regular, still have yet to see 2x). That being said, DLSS is +37% framerate. It looks to be a viable choice if you cannot run native res + TAA at the framerate you'd like.

I have seen DLSS screens from a number of reviews now. Slight edge to TAA, but it is very sensitve to precise shot alignment and motion. Here is one that has three shots: TAA/DLSS/OFF:
https://pclab.pl/art78828-21.html

Both TAA and DLSS soften textures. Look at the girls jacket/jeans with AA off vs either AA method.

While actually playing the game, I doubt much difference would be spotted between DLSS/TAA. They both soften things and reduce Aliasing by a similar amount.

Well DLSS looks the best at that link. Maybe due to the static nature? Who knows.

 

[Qwertilot]

DLSS is going to be faster because it uses all the 'extra' bits of the rx cards.

 

[railven]

I have seen DLSS screens from a number of reviews now. Slight edge to TAA, but it is very sensitve to precise shot alignment and motion. Here is one that has three shots: TAA/DLSS/OFF:
https://pclab.pl/art78828-21.html

Both TAA and DLSS soften textures. Look at the girls jacket/jeans with AA off vs either AA method.

While actually playing the game, I doubt much difference would be spotted between DLSS/TAA. They both soften things and reduce Aliasing by a similar amount.

Opening the images for each one of the 3 in that link, woof at TAA and DLSS. Those two are both horrendous to me.

TAA - the trees/rocks in the background are smeared so bad, it looks the worst to me.
DLSS - textures on Cid are definitely smeared, but compared to the TAA the background looks a bit crispier.

I'd personally use neither of these and let the edges cut my eyes.

Actually, looking again - is there some DoF screwing things up? The no-AA screen shot also has rather muddy background, but then looking at Cid's bag on her belt, the DLSS doesn't smear as much as TAA.

Glancing over the 3 images, I can't tell which of the 3 is better to be honest. Each one has weird muddy textures at different spots, but just using those 3 images, TAA is by far the worst to me.

 

[crisium]

DLSS is going to be faster because it uses all the 'extra' bits of the rx cards.

It's because it runs the game at a lower resolution with the regular CUDA cores. 3840x2160 DLSS is not actually 3840x2160.

 

[crisium]

Something fishy in techpower up review.Looks like 1080TI is slower than it was in their latest review.Maybe Nvidia "magic" drivers?

Or the FE 1080 Ti throttled more in the 2080 review. The room could have been a lot hotter with a marathon benchmark session.

The important thing is at max OC both in other reviews, the cards perform identical.

 

[Muhammed]

No.
Though Vega is a compute oriented GPU, so the results aren't surprising

Double No. Sandra is an opencl mess, just look how bad GPUs scale in that bench. TitanXP faster than 2080Ti? WTF?

For a consistent compute benches look no further than Anand, Turing is a beast in compute, smokes Vega by a large margin.

https://www.anandtech.com/show/1334...x-2080-ti-and-2080-founders-edition-review/15

 

[maddogmcgee]

It's hit or miss, but when it works it's great. Unfortunately it's been a pretty busy year, and really the only things I've devoted much time to are F1 2016 and finally finishing Brood War, neither of which really call for the kind of horsepower I have available.

Please tell me you are mainly playing Starcraft Broodwar on those?

 

[sze5003]

I guess the argument for keeping it would be:

--performance @1440p with Ray Tracing
--games now/future with meaningful Ray Tracing functions, on current generation (is the hardware for this, specifically, going to be better with 3xxx and 4xxx?)
--potential for gaining performance with the current 2xxx with better drivers.

At 1440p only, I'm not sure you would be getting much benefit above your 1080ti which, I imagine, is giving you ~120-140fps with minimums above 100fps?

I think the weirdness in the numbers right now across games for this generation, is due to early gen drivers which, understandably, are going to need some ironing out with the brand new toys. I think it's possible driver improvements will give 10-15% better performance within a year or 1.5years for these cards, simply because there are a lot of new functions here unlike the previous generation (which were mostly foundationaly mature by Pascal).

So yeah, if raytracing does make for really good pretties within a year, and the games are there, and you are looking at minimum 30% uptick with potential of 45% off of what you have now, then it was a good purchase if you simply don't care about the cost.

Actually at 1440p ultra in BF 5 I was getting around 95 and a bit lower fps, Deus ex mankinded divided I got around 74 fps benched.

Tomb raider I get avg 83. Just a few games I have installed right now.

So I would see more with this new card. I gotta go read these reviews tonight.

 

[MrTeal]

Please tell me you are mainly playing Starcraft Broodwar on those?

Yeah, I downloaded the Blizzard client to claim my free copy of Destiny 2, and I saw I could get Brood War for free. I never did play it when it was new even though I'd run through the original campaign a few times, so I played the full campaigns of the vanilla and expansion.

 

[maddogmcgee]

Yeah, I downloaded the Blizzard client to claim my free copy of Destiny 2, and I saw I could get Brood War for free. I never did play it when it was new even though I'd run through the original campaign a few times, so I played the full campaigns of the vanilla and expansion.

Nice. The last zerg mission is a challenge. There is still a professional scene in Korea for the multiplayer which is kinda crazy all these years later.

 

[Ajay]

Double No. Sandra is an opencl mess, just look how bad GPUs scale in that bench. TitanXP faster than 2080Ti? WTF?

For a consistent compute benches look no further than Anand, Turing is a beast in compute, smokes Vega by a large margin.

https://www.anandtech.com/show/1334...x-2080-ti-and-2080-founders-edition-review/15

AT doesn’t provide a summary of its testing methods. I don’t know if these results contain double precision tests; except f@h which is clearly SP. There is also something weird going on with the Geekbench results which scale non-linearly.

 

[Muhammed]

AT doesn’t provide a summary of its testing methods. I don’t know if these results contain double precision tests;

Consumer Turing doesn't support DP.

There is also something weird going on with the Geekbench results which scale non-linearly.

Probably due to the separate integer and fp paths on Turing.

 

[happy medium]

NVIDIA GeForce Game Ready 411.63 WHQL Drivers For GeForce RTX 2080 Ti and RTX 2080 Launched

https://wccftech.com/nvidia-geforce-game-ready-411-63-whql-drivers-geforce-rtx-20/

 

[PeterScott]

NVIDIA GeForce Game Ready 411.63 WHQL Drivers For GeForce RTX 2080 Ti and RTX 2080 Launched

https://wccftech.com/nvidia-geforce-game-ready-411-63-whql-drivers-geforce-rtx-20/

Why don't you just use "Paste as text".

Posting in Large Bold, is bad form, worse than ALL CAPS SHOUTING!

 

[zinfamous]

Just look at the headrest and see all the texture it removes. Lots of aliasing with TAA vs DLSS, but, DLSS makes everything too soft in this shot for me.

yes, I agreed as well (first!)

DLSS seems to deal (sorta) with one problem, at the expense of creating more by removing generational gains in fidelity.

The more you think about it, this release is just nothing but swallowing compromises in order to accept some unproven "new hotness."

--Oh look at the 30% gains at 4k! which, really, is great....but at what price?

--Yeah, but we added these gigarays that bring giga to your rays and according to our scientists, are powered by dliethium crystals (but seriously, don't ask them). Oh, you want the new effects? They are great! You will love them! You will especially love them at SHOCKING! 2010 resolutions @60fps minimums! 1080p never looked so great!

--Oh hey, that AA was never perfect and always very resource intensive. So here's this new one called DLSS, which is, like, 20% closer to perfect (seriously, don't look too closely at the lines, boss!) and with LESS performance hit (seriously--there are some numbers somewhere!)! How much, like, almost less that what was previously basically none! And it's great! Almost less jaggies and none of those bothersome sweet textures that you were getting used to over the last 10 years. KEEN!

I mean, it's exactly like an AMD release, and all of things that AMD is constantly piled on about, but at shockingly intolerable prices. Just bring the prices back to reasonable and this would be a boring, if not good (ti only) release. Considering that the new hotness really seems to be kinda bad in the worst case, a bit of a dud in the best, it would at least give some sort of value to 2080 and the as-yet-unknown 2070 if the prices had remained equal or, at least, rational.

...I'm starting to wonder if the 2070 is going to come out significantly below the 2070ti in overall performance, and maybe match 2070 AIB cards.