Leaked Nvidia RTX 3090 benchmark score shows performance up to 26% faster than RTX 2080 Ti

[mongeese]

Highly anticipated: A next-gen Nvidia flagship has been spotted in the 3DMark benchmark database. It’s the first time we’ve seen a GPU in a performance bracket like this – if you discount the below photo that leaked a couple weeks back – with a performance output in 3DMark that appears to be up to 26% faster than the RTX 2080 Ti.

Let’s plunge into the numbers. The GPU achieved a score of 18,257 in the graphics component of 3DMark’s Time Spy benchmark. That score could be considered proof enough that this is a new GPU. A Titan RTX can’t generally get that high without exotic cooling like liquid nitrogen.

Relative to the average scores that 3DMark provides, it’s 21% faster than the Titan RTX and 26% faster than the RTX 2080 Ti. Thumbs up, Nvidia.

You should note, this ain’t no run-of-the-mill GPU – this is an alien from the exotic GA102 species. That’ll be the flagship die that a different set of GPUs could use: the next Titan, a stack of Quadro cards, and some gaming graphics cards potentially called the GeForce RTX 3080 and RTX 3090.

A few leakers are confident that the GA102 will have up to 5,376 cores, but there’s no verdict on the precise core configurations of specific models, although strangely, everyone agrees on the memory capacities as specified in the tweet below...

2nd Gen NVIDIA TITAN
GA102-400-A1 5376 24GB 17Gbps
GeForce RTX 3090
GA102-300-A1 5248 12GB 21Gbps
GeForce RTX 3080
GA102-200-Kx-A1 4352 10GB 19Gbps

— KatCorgi (@KkatCorgi) June 19, 2020

Parties that approximately agree with these numbers include: Rogame, Kopite7kimi, and Igor’s Lab.

It’s been assumed that this next-gen GPU is the flagship RTX 3090 because of how well it performs, but it could also be the RTX 3080. That will all depend on how Nvidia decides to position and market the GPUs ahead of release.

Other things to note: 3DMark recorded a 1935 MHz clock speed during the benchmark, a 25% increase over the RTX 2080 Ti’s 1545 MHz. In other words, the upcoming GPU doesn’t necessarily need a higher core count than the RTX 2080 Ti to reach that performance bracket. However, it is expected that Nvidia will release at least one card that does have a higher core count, most likely by the end of the year.

Dark times await AMD Radeon fans, unless big Navi delivers, of course.

Permalink to story.

https://www.techspot.com/news/85714-leaked-nvidia-rtx-3090-benchmark-score-puts-rtx.html

 

[yRaz]

26% faster on what is likely to be a $1000 graphics card isn't "putting it to shame"

If you look at ray tracing performance, a 26% increase means we'll have to wait for the 4000 series for "real time ray tracing" to actually be a practical thing

And "dark times await Radeon fans", what was that? I didn't know I was reading a BuzzFeed article

 

[VitalyT]

RTX 3090 is the replacement for RTX 2080 Titan. And if it is only 26% faster than RTX 2080 Ti, it puts itself to shame, and not the other way round.

This would mean the 3000 series will bring only about 20% of real performance improvement, which after all the hype isn't impressive at all.

 

[Eldritch]

RTX 3090 is the replacement for RTX 2080 Titan. And if it is only 26% faster than RTX 2080 Ti, it puts itself to shame, and not the other way round.

This would mean the 3000 series will bring only about 20% of real performance improvement, which after all the hype isn't impressive at all.

I won't be so sure about that. The improvements may very well be in the Ray Tracing department and the benchmark won't be swayed too much by that but in real gaming turning on ray tracing can be a showstopper, quite literally.

So delivering a smooth experience with RT will be the next big step as finally the devs will start properly implementing it.

By next next gen, ie 4xxx Ray tracing will hopefully become like SSAO, resource intensive but reasonable tradeoff considering results.

 

[Captain828]

We don't really know for sure if this is the top SKU yet so hard to tell what these numbers refer to but given the complexity of the cooler design & lack of competition, I'm fairly confident that the top SKU will be around $1300.

Even if this information turns out not being the top SKU, Nvidia could easily price this one above the RTX 2080 Ti MSRP so over $1200... at least until AMD's Big Navi launch.

 

[Lionvibez]

My Budget for GPU's usually doesn't exceed $600 so I won't be looking at any of these.

Hopefully when big navi is out it will provide some downward pressure on pricing for at least the previous gen NV cards.

 

[newcombcmb]

I'll keep my 1080ti for a while longer until something is released. Even then I'll wait for the reviewers to get a hold it them. EVEN then, I'll wait for the new cards to come back into stock, since all the early adopters will have taken all the cards. AND EVEN THEN... I'll have to ask the wife if I can justify spending that kind of money on a single component. And if your married, you already know the answer to that question! LOL

 

[MaxSmarties]

We don’t know what a “3090” is, so why are we speaking about this time spy result as if it was an RTX 3090 score ?

 

[EClyde]

Leaked? Ha Ha Ha. Come up with something new

 

[Splus]

Well, if it's 26% faster than the 2080 Ti and runs at 25% higher clock, then that's no more and no less than simply an overclocked 2080 Ti!

 

[Evernessince]

I don't know what the author is smoking but 26% is completely underwhelming. AMD have already shown in slides a 50% performance per watt improvement for RDNA2.

I'm pretty sure Nvidia won't allow itself to be noncompetitive.

 

[Jerry in WA]

I honestly doubt this is the top card. Even if it is, I also seriously doubt that card will be called a 3090 (rather than a 3080Ti)

Nvidia likes to release the 3080 (series) first. Something better than last gen's Ti but not their best. Then 3-6 months later, they will release the 3080Ti. A card I doubt is even ready enough to be spotted in the wild yet.

I say this is the 3080. Non Ti. (or, I'm a bit underwhelmed)

 

[Thanthan]

I really doubt this is the top card o.O would be rather underwhelming considering that it's launching 2 years after Turing. anything under 30% is a fail imo, and 40% is needed to be considered "good" by me.

 

[EEatGDL]

I came here expecting something in the ballpark of 40-50% more performance; 25% would be acceptable to me for a 3080, not a 3090. I guess our expectations of "putting to shame" and performance scaling differ.

 

[gigantor21]

The gap between the 780Ti and 980Ti was 40-50% just a few years ago--and the latter launched at around $650.

A 25% jump for a card that will likely cost over $1000 is a shame, alright, but not the way the author thinks.

 

[PurpleYoda]

I'll keep my 1080ti for a while longer until something is released. Even then I'll wait for the reviewers to get a hold it them. EVEN then, I'll wait for the new cards to come back into stock, since all the early adopters will have taken all the cards. AND EVEN THEN... I'll have to ask the wife if I can justify spending that kind of money on a single component. And if your married, you already know the answer to that question! LOL

Lol!! :-D

 

[grvalderrama]

Not exactly overwhelming, now, if that would be the score for a 3070, then that'd be pretty exciting

 

[krizby]

2080 Ti FE stock get around 13610 in Timespy Graphic score so that is a 34% increase, not 26%. Don't compare with manufacturer overclocked 2080 Ti model please.
https://www.guru3d.com/articles_pages/geforce_rtx_2080_ti_founders_review,33.html
I will be even happier if this score is from 3080 and not 3090, that means the 3090 would demolish the 2080 Ti just like how 1080Ti outdone the 980Ti

 

[grumblguts]

Where on earth did the 25% number come from. It's going to be more like 40%-75% and RTX will have a minor hit on performance.

https://imgur.com/R982azw

Well im just laughing... Told you so.

 

[grumblguts]

If the New RDNA card isnt twice as fast as the 2080ti I wont be buying that neither.
Not worth upgrading unless its triple digit gains or high double.

 

[Mr Majestyk]

26% faster on what is likely to be a $1000 graphics card isn't "putting it to shame"

If you look at ray tracing performance, a 26% increase means we'll have to wait for the 4000 series for "real time ray tracing" to actually be a practical thing

And "dark times await Radeon fans", what was that? I didn't know I was reading a BuzzFeed article

I think the author is confused, the 3080 leaks are showing improvements like this relative to 2080Ti. This is not 3090 performance at all, AMD would be laughing if this was 3090 benchmark.

 

[SpenReyn]

I'll keep my 1080ti for a while longer until something is released. Even then I'll wait for the reviewers to get a hold it them. EVEN then, I'll wait for the new cards to come back into stock, since all the early adopters will have taken all the cards. AND EVEN THEN... I'll have to ask the wife if I can justify spending that kind of money on a single component. And if your married, you already know the answer to that question! LOL

Yeah mine nearly had a fit when I spent $280 on an RX5700, can't imagine what she'd say to a $1000+ video card, except maybe NO!

 

[PEnnn]

"You should note, this ain’t no run-of-the-mill GPU – this is an alien from the exotic GA102 species."

"Dark times await AMD Radeon fans, unless big Navi delivers, of course."

As they say here in the regular version of TS: What just happened??

Did this website get hijacked by nVidia??

Or maybe the author is an undercover nVidia fan boy? This is so undignified, it's pathetic for a tech site to post such ludicrous drivel!

 

[grumblguts]

I'm gonna sell both my 2080Ti and buy the 3000Ti.

Who else is in with me?

well SLI sucks so does crossfire.
just foolish to waste money twice.
You could of just purchased a £500 card and do some 2.5k aka 2560x1440 gaming and had just as much enjoyment.
Dont do that again you only hurt yourself when its time to upgrade.

 

[amghwk]

Clearly the "...put to shame" nonsense is purely a clickbait for this article. I wonder if the writers get perks for attracting more readers/clicks/shares/views.