Guru3D Reviews: GeForce RTX 2080 and 2080 Ti Founder Edition

You base your buying decisions on ethics. Others base it on getting best bang for the buck regardless of maker. Currently Nvidia rulez with the 1080's price vs alternatives in its price bracket.

 

Until those features you talk about are also tasked on the 1080 ti, then obviously your snarky comment about 1080p 60fps seems pointless. But since we can't test either of those features and see what real performance we get from those features on the new cards as well as the old cards, it's a pointless comment to make.

Not only that, but no card performs the same as the 1080 ti, the 2080 performs better more often then not according to this review then the 1080 ti, and yes, sometimes worse. If you're only going to focus in on the titles that see similar performance or worse, and then claim that's the only ones that matters, thats your own fault and your own doing.

That being said since you talked about the 1080p @ 60fps with the new features i can actually only imagine you're talking about the 2080 ti vs the 1080 ti, not the 2080 vs to the 1080 ti, since the only performance numbers we have seen at all for the "new features" (again can only assume new features means ray tracing since obviously the DLSS improves performance across the board, which wouldn't make sense if you were trying to say that is the 1080p @ 60fps performance....) is of the 2080 ti. And if that's the case......where's any of your even remote amount of logic to make a statement such as a 2080 ti is the same performance of a 1080 ti? lol?

It's like you didn't even read the article.

 

Sure. Because the 600 games in my Steam library are gonna get RTX.

But OK. 2 years from now, when RTX is available on the majority of games, it might be worth it then. But if I want to play my existing games at higher resolutions and higher frame rates, the RTX cards seem like a disaster to me.

Actually, no. It's not OK even then. I don't want to play games at 1080p 60. You may enjoy that. Have fun, I won't stop you. But I'll pass. "Oh but a 1080 Ti would not even run 720p 60 with those features on." Yeah, and? I don't see how that would make me want to play 1080p 60. Still a disaster.

 

Ha ! That would be a miracle to see.

More likely there will be two dozens or so of RTX games, and nobody would actually PLAY them with RTX ON because of severe performance degradation.
Wasn't there another NV technology that did something similar...?

What was the name... hmm, damn... can't remember... it's so lost in history.. something with P? Power... umm, no; Panda ... no, that's not it.
Oh, I know!!!! \o/... PhysX

Hey look, in 2018 The majority of games are using NV PhysX. All 30 of them or so... all using it for extremely important gameplay stuff like... debris and shattered glass.
K

 

As far as performance goes, I expected nothing, and I'm still disappointed. Yet not too surprised.

Regardless I'm super happy/relieved that ray tracing and something like DLSS are finally being pushed forward. Hopefully they won't just abandon ray tracing like PhysX.

 

There's also the one with "G". They might as well have labeled the gameworks options in the game menus as "30FPS lock".

NVidia seems to not be capable of introducing features that would allow them to say "buy NVidia instead of AMD, because we have features that will make your games run super fast and look super nice." Nope. What they're doing instead is "we have features that will make your games run like crap, but be happy because on AMD they run even shittier."

I believe they need to rethink their strategy.

 

Anyone know what the hash rates are on these puppies?

 

They probably need 7nm until they can make significant progress. Until then, they can use the 20-series to test out new features - and there's no better time for that. Even Jim from AdoredTV praised the innovation in RTX. Aside from ray-tracing, it's amazing how Nvidia has managed to put their experience with AI to good use with DLSS. Whether it becomes a long-lived feature or dies off like PhysX, only time will tell.

There's been plenty of RTX bashing but let's look at it from another angle. Performance jumps are always nice, but every now and then we need companies taking risks with new technology. Even if the 20-series cards aren't particularly good at ray-tracing they at least give developers something to play with - and hopefully what they produce can then be properly applied when RTX goes mainstream (with the 30-series cards and beyond).

And I say this as someone who would prefer a more powerful card for traditional rendering. I game at 4K and the limitations of the 1080 Ti at that resolution are well known. Would I prefer a GPU that is 100% faster than my 1080 Ti? Absolutely. I can hold out for a while longer though, and 7nm GPUs are probably worth the wait. Meanwhile, those customers who want to try out RTX can feel free to buy the Turing cards (as I said before, they are the beta testers).

If you're referring to mining, it's mediocre. 56 Mh/s at 306 watts, with the memory overclocked and 75% power. Calculated ROI = 100 years, LOL. It's safe to say that miners won't be touching this.

http://www.legitreviews.com/nvidia-geforce-rtx-2080-ti-and-rtx-2080-benchmark-review_207896/15

 

I found that in the Infiltrator demo, the IQ is worse with DLSS activated. Here's a comparison: http://screenshotcomparison.com/comparison/120754 (just mouse over and wait for the other image to load in, it's about 2mb big so can take a few seconds depending on your connection)

I noticed lesser texture definition, aliasing, worse LOD levels of objects and some shading/shadowing abnormalities in the one using DLSS ON. I marked the screenshots with red and green differentiate the two.

It would be much better if Hilbert could provide us with raw in-engine screenshots from both demos to make comparisons easier, despite youtube offering a higher bitrate tolerance for 4K videos, it's still quite a bit of compression which makes details harder to spot. I guess Digital Foundry will be all over this as soon as they can, they love close inspection and comparisons lol.

 

Awesome review Hilbert, a massive thanks for all the effort/work, as always super job man! I really love the MSI GeForce RTX 2080 TI GAMING X TRIO card it looks epic, it will be mine... oh it will be mine

Thanks again man, I hope you get some sleep today.

 

The link already answered it.

If you have a current GPU, able to render 4K@60 FPS with 8xAA now, adding a second GPU will not necessarily increase the FPS but can increase the AA levels.

With 2x GPU, there is a 16xSSAA selection in the drivers, with each GPU rendering 8xAA Max.

I am corrected in that 32xAA was QuadSLI feature, now disabled. I remember it being available but forgot no more than 2 GPU are now supported.

So my point was that the GPU companies could have invested more die space to ROPs for improved AA quality in lieu of these tensor or RT cores since those features are not ready "yet". The benefit is that improved AA can be available at a higher resolutions for more games NOW.

Based on current pricing, current users can add a second gpu to assist. Even if you cannot get 8xAA at 4K on one GPU, you can probably get away with 8xAA on 2 GPU (4xAA each) and it would still look better than TAA, to me.

For games that don't support AA, there are still options to downsample - VSR or DSR (shader based post processing downsampling) and again increasing ROPs budget could have improved here. Both work with 2x GPUs.

This is nearly, not 100%, applicable to more games than specific in game support.

You can check back all the charts from 2005 till now on ROP count. That stayed stagnant (probably due to resolution output) while most HW investment went to increase and improving shader operations for effects and post processed AA.

I had CFX and ran up to 16xAA when possible and can see it clearly. If you have never seen SLI or CFX AA, then I suggest you try it to understand the overall quality impact.

@ Hilbert, when you prepare the inevitable SLI review, can you also test SLI AA and DSR modes as so we can compare performance to DLSS? It would give users an indication on how to use that power now instead of waiting for features to ripen. Maybe we should have a separate thread about this...

 

Worse doesn't even describe it.
It's a blurry mess.

For example the shape of metal bars flooring on the right side platform vanishes completely,
All rails everywhere become very "soft",
The detail on the Mech's plating vanishes completely in a pile of smudges,
Various small objects are erased out of existence.
The text "DAN(GER)" on the complete right side becomes a complete smudge and is unreadable.

If that's what it takes to have higher framerate, why even play at 4K ? Play at 720p on Medium settings and be done with it, instant +200% boost in performance. It will probably still look more accurate than this DisasterLSS.

Jesus, this is what NoVidia asks $1200->1500 Euro for ???

 

Well, with that kind of motion blur used in Infiltrator, you would have hard time noticing visual difference to FXAA on screenshots as even those would be affected by blurring.
There are very few moments where camera moves just slightly. Therefore presence or absence of visual kinks is pretty hard thing to confirm.
And even playing 4K video on 1080p... it looked pretty blurred. Doing comparison of AA techniques apparently requires disabling motion blur and other blur methods, which did not happen here.

 

Impressive performance. I must say that here in South Africa the online retailers are clearing stock of the GTX 10XX and man you can get a GTX 1080 for $435 if you revert back to USD. Not bad if you don’t want to put your hands too deep in your pockets for the 20XX cards.

 

Compared to Hardware Unboxed FC5 test(Ultra/SMAA) since they use 8700k@5GHz also, I manage to close the gap to 10%(120 vs 132) and 22%(63 vs 77) in 1440p and 4K respectively.

Yes this was oc'ed 1080Ti(~2038/+520mhz mem) vs stock 2080Ti.

Not bad.

 

Yeah, DLSS is much lower quality... very apparent that it's being upscaled from lower res (hence the huge performance boost - TAA uses 2-5% gpu power, not 25+%), no matter what nvidia / fanboys claim. DLSS looks like upscaling, with taa + sharpening filter... nothing more, nothing less.

 

It really makes you realize that TAA is a blurry mess, the only thing giving away DLSS in the 2nd picture is the thin white line around the rear window, otherwise it's a clear winner, the rest of the picture is perfectly clear and sharp.

Look at the writing on the left "..surath basin", it's day and night really.

 

These prices though............................... You can literally build a nice little gaming pc for the price of one of these.