Nvidia set to unveil the GeForce RTX 5090 and RTX 5080 graphics cards today at CES, watch it live here

midian182

Posts: 10,633   +141
Staff member
Highly anticipated: What has felt like the longest-ever wait in tech history is almost over. In around 12 hours, Nvidia will finally unveil the GeForce RTX 5000 series desktop graphics cards during its CES keynote. Leaks for consumer-grade Blackwell GPUs have been coming thick and fast over the last few days, but we'll soon have the official details, including the all-important prices. You can watch the entire event right here, starting at 6:30 pm PT / 9:30 pm EST / 3:30 am CET.

Update: Nvidia has now unveiled the new GeForce 50 series, along with DLSS 4 technology. Now, it's time to share our thoughts on what was revealed during the CES 2025 keynote – including all the specs, pricing, and performance claims.

A most recent leak claimed to show a screenshot from a European retailer listing the RTX 5080. It was priced at 1,699 Euros, around $1,749, though that included 20% VAT, so the actual price would be a still very high $1,399. The listing also appeared to confirm the use of GDDR7 memory and the introduction of Nvidia's next-generation of upscaling tech, DLSS 4.

Keeping with tradition, Nvidia still hasn't officially confirmed that the RTX 5000 series is on the way, but there have been plenty of hints. The fact the gaming festival it's currently running is called GeForce LAN 50 and the company is talking about looking "to the future" on social media pretty much confirms what's coming. Moreover, Jensen Huang is delivering the keynote, marking the first time that the CEO has appeared at the event since 2019, suggesting something big is in store.

It's expected that the RTX 5090 and RTX 5080 will be unveiled at CES, while the RTX 5070, RTX 5070 Ti, and China-only RTX 5090D will likely be on show. In the last few hours, a marketing image of the Inno3D RTX 5090 iChill X3 appeared online. It features a massive 32GB of GDDR7 and a 3.5-slot cooler.

The RTX 5090 is expected to feature 21,760 CUDA cores, nearly 1.8TB/s of memory bandwidth, and a TDP of 575 watts, 125W more than the RTX 4090. It's also the only card in the RTX 5000 line to use the GB202 GPU.

The RTX 5080 is believed to be the first card in the RTX 5000 series to launch, on January 21. It's expected to come with 16GB of GDDR7, half that of the RTX 5090, and a 256-bit memory bus.

Remember to come back here at 6:30 pm PT / 9.30 pm EST / 2:30 am BST to watch the whole event.

Permalink to story:

 
I wonder if he will actually have the guts to announce the price during the presentation or he will chicken out and leave the customer to find out by going to the Nvidia pre order page....
 
RTX sales slogan:
"Buy our expensive upscaling machines...
Our upscaling is so great, you only need to buy a 5060 and upscale to your content. We also sell the 5070 and higher for d0mb people who don't understand how good upscaling is.

For the really gullible people, we also sell a $1,500 RTX5080 for those who don't like latency while upscaling."


Sorry for the satire, but that^ is exactly how I feel NVidia will be able to talk about, is upscaling & gimmicks. Because NV's Blackwell architecture is so heavily based on non-gaming die space, so nVidia will have to ultra-market "upscaling everything" and use ultra-gaming gimmicks to sell these cards, bcz their Gaming architecture can't push frames, unless it uses upscaling.
 
Last edited:
RTX sales slogan:
"Buy our expensive upscaling machines...
Our upscaling is so great, you only need to buy a 5060 and upscale to your content. We also sell the 5070 and higher for d0mb people who don't understand how good upscaling is.

For the really gullible people, we also sell a $1,500 RTX5080 for those who don't like latency while upscaling."


Sorry for the satire, but that^ is exactly how I feel NVidia will be able to talk about, is upscaling & gimmicks. Because NV's Blackwell architecture is so heavily based on non-gaming die space, so nVidia will have to ultra-market "upscaling everything" and use ultra-gaming gimmicks to sell these cards, bcz their Gaming architecture can't push frames, unless it uses upscaling.
Neh, at least not the top card. With over 20 000 cuda cores, the raster performance will also be really good, it will blow anything else on the marked off the board long before upscaling comes into play.
The reason why they're using upscaling techniques is simply because they've hit a performance wall with the current tech and they're trying to be innovative. 5090 will have a TDP close to 600w - that card will be a spaceheater.
Next step is to come up with something brand new - they can't just keep upping the power requiremens. DLSS looks pretty damn awesome in 4k as it renders from 1440p - but looks pretty crap on 1080p monitors. Maybe they've done something smart with the "neural" thing that makes it look better on lower tier monitors for the mid tier cards
 
We know you can't afford our new RTX but we also know you will find a way to pay a scalper for your next nVidia GPU vs buying AMD or Intel GPUs -- now that's the way nVidia was meant to play.
 
At launch the best bang for buck 4000 series card was the 4090. I suspect nvidia will set up the same value proposition for the 5000 series, much to our chagrin.
 
Neh, at least not the top card. With over 20 000 cuda cores, the raster performance will also be really good, it will blow anything else on the marked off the board long before upscaling comes into play.
The reason why they're using upscaling techniques is simply because they've hit a performance wall with the current tech and they're trying to be innovative. 5090 will have a TDP close to 600w - that card will be a spaceheater.
Next step is to come up with something brand new - they can't just keep upping the power requiremens. DLSS looks pretty damn awesome in 4k as it renders from 1440p - but looks pretty crap on 1080p monitors. Maybe they've done something smart with the "neural" thing that makes it look better on lower tier monitors for the mid tier cards
That the whole point.

NVidia's "Top Card" is for graphic artists & those who need CUDA cores, NOT Gamers. NV still relies & uses dlss in top card's marketing, because that is all Blackwell offers. Bigger, hotter and more money.

Gamers do not care about CUDA, they want Price/Performance. (Upscaling is a crutch)



Secondly, to your point of 1080p... if the RTX5060 can't handle 1080p native (without dlss), then RTX5000 is a failed architecture. And if the rtx5060 can't upscale to 1440p, then it's failed architecture.

Contrary, if the current outgoing 4070 can't upscale to 1440p, then why do you need a 5000 series..?
 
I would be pretty surprised if I am so bored this evening that watching this reveal is the most entertaining thing I can find to do... but I guess there's a chance...
 
My 2080 Ti is showing it's age, and since I'm both a gamer and a data scientist who likes to dabble in Gen AI, I'll probably buy the 5090 if it's under 2 grand, that extra memory will be wonderful to have. But, like I did with the 2080 Ti, I won't upgrade for 6+ years, making the cost per year quite reasonable.
 
I wonder if he will actually have the guts to announce the price during the presentation or he will chicken out and leave the customer to find out by going to the Nvidia pre order page....
Haven't they always announced the price at the same time?
 
I wonder if he will actually have the guts to announce the price during the presentation or he will chicken out and leave the customer to find out by going to the Nvidia pre order page....
lol you do realize CEOs don't deal with pricing? There's an entire department that deals with that.

Besides, gaming is a paltry $2 billion a quarter, their AI Data centre business has become $26 billion a quarter. So... you'll have to excuse him if he doesn't really give a flying skunk.
 
$549 for the 5070.

4080 super and under just lost 25% of its value.
 

Attachments

  • RTX-5090-HERO-1536x825.jpg
    RTX-5090-HERO-1536x825.jpg
    72.7 KB · Views: 1
Last edited:
If a 5070 indeed matches a 4090 then the 4090 just lost it’s value by $1200 on avg.

Nvidia knows how to tell you to get screwed and laughs while you enjoy it!

Aaaaand. AMD Is screwed on pricing. This means they won’t have a gpu to match the 5070
 
5090 for $1999
5080 for $999
5070Ti for $749
5070 for $549

And the 5070 claims an AI workload performance passing the 4090 (remains to be seen).

Sounds to me like a lot of credit cards are about to get exercised.
 
5090 for $1999
5080 for $999
5070Ti for $749
5070 for $549

And the 5070 claims an AI workload performance passing the 4090 (remains to be seen).

Sounds to me like a lot of credit cards are about to get exercised.
Prices so reasonable that I am genuinely surprised.

I am guessing a lot more people will upgrade their ancient 3000 or even older cards to 5070 or better
 
Considering the 5070 matching the 4090 is actually 5070 DLSS performance mode and FG enabled vs 4090 native, it's not huge leaps ahead of where the lower 4000 series cards are now. It's bringing 4070TiSuper performance down to $550. Not bad, but not revolutionary.

Edit: Definitely explains why they cut off 4000 series production so quickly.
 
Considering the 5070 matching the 4090 is actually 5070 DLSS performance mode and FG enabled vs 4090 native, it's not huge leaps ahead of where the lower 4000 series cards are now. It's bringing 4070TiSuper performance down to $550. Not bad, but not revolutionary.

Edit: Definitely explains why they cut off 4000 series production so quickly.
I agree, except everyone was anticipating higher pricing.
To be at 4070Ti S in rasterization for $200 less ... AMD has some work to do.
 
If a 5070 indeed matches a 4090 then the 4090 just lost it’s value by $1200 on avg.

Nvidia knows how to tell you to get screwed and laughs while you enjoy it!

Aaaaand. AMD Is screwed on pricing. This means they won’t have a gpu to match the 5070
If you actually listened Jensen made it clear it's only possible with AI. Meaning the 5070 can only match the 4090 with multi-frame generation.
 
Back