Nvidia RTX 5000 graphics cards rumored to launch next year with massive performance increase

[midian182]

Rumor mill: It's fair to say that Nvidia's RTX 4000 series has been one of the most disappointing card families ever to come from team green. But could its successor, the RTX 5000 line, put right what Lovelace did wrong? According to rumors, we might find out next year.

Unlaunching the RTX 4080 12GB, the disappointing RTX 4060 Ti, and prices that don't correlate with their generational performance increase, the Ada Lovelace series hasn't been Nvidia's greatest triumph, which is why the company might be glad to see rumors about the successor picking up pace.

During Jensen Huang's keynote at Computex this week, the CEO mentioned the Hopper Next architecture; Nvidia previously launched the Hopper architecture for datacenters, enterprise, and AI use. The rumor is that Hopper Next could also be used in gaming cards.

YouTube channel RedGamingTech says that Hopper Next is "basically Blackwell," the codename for Nvidia's next line of gaming GPUs. The rumor is that Hopper Next will be used across different product segments, including RTX 5000-series GeForce GPUs and high performance computing (HPC) parts for AI and other enterprise use.

Blackwell is said to offer a massive performance uplift over Lovelace – some of our benchmarks with the RTX 4060 Ti 8GB show the previous-gen RTX 3060 Ti equivalent reaching higher FPS. With Blackwell, there's talk of a 2X improvement, and ray tracing will likely be an important element for Nvidia, as always.

The RTX 5000 series is expected to be based on TSMC's 3nm process node, though Huang did confirm that it could tap Intel for future chip manufacturing. The CEO said early test results for an Intel-manufactured Nvidia chip based on the former's s next-gen process node had yielded good results, and both companies are "evaluating the process" on how best to move forward.

One interesting aspect of the RTX 5000 is how much VRAM Nvidia will include in its next-generation cards. Modern video games are requiring an increasingly large amount, and plenty of RTX 4000 cards have been criticized for their stingy amount of VRAM, so maybe the company will do away with 8GB cards in the mid range. We'd also like to hope that Nvidia will rethink its pricing strategy, but that might be wishful thinking.

The RTX 5000 release date is just a rumor, of course, so take it with a heavy dose of salt. But Nvidia might want to put Lovelace behind it as quickly as possible.

Permalink to story.

https://www.techspot.com/news/98922-nvidia-rtx-5000-graphics-cards-rumored-launch-next.html

 

[MarcusNumb]

It's impossible to expect that NVIDIA will learn from their mistakes, cause they don't care xD Let's wait to see another record of VGA power consumption

 

[Puiu]

Don't worry, Nvidia will make sure that there will be only a small increase in cost per frame

 

[SirDigby]

I am planning to upgrade next year, whole new build, might even get new case and everything. If the 5000 series isn't a sh*tshow then that's definitely on the cards. I want stable 4K gaming on the 5070 level, maaaaybe 5080 if I can run to it.

 

[Nobina]

*Performance increase measured using DLSS 4 (it's now a video of gameplay)*

 

[DSirius]

"Nvidia RTX 5000 graphics cards rumored to launch next year with massive performance increase".
Would be nice, but I think that it is a typo error.
I am sure that will be rather Massive PRICE increase than performance increase.

 

[Achaios]

Τhe Leatherman will first eat his leather jacket piece by piece before he lowers GPU pricing.

 

[takaozo]

Massive power consumption and massive price increase maybe. Be prepared for 1KW power draw and have your credit cards ready.

 

[yRaz]

*Performance increase measured using DLSS 4 (it's now a video of gameplay)*

I think everyone is tired of DLSS talk at this point. It's great tech but they're segmenting it even amongst their own cards. "you can get extra frames with the AI accelerators we used to take up space on the silicon that you didn't ask for, and you wont have access to all the DLSS features once we release new a card with a new DLSS version."

To me, DLSS isn't a selling point if it's only going to work for a little while. Permanent features? maybe, but not the way they're selling it.

I'm so sick of hearing about DLSS.

 

[kiwigraeme]

Don't worry, Nvidia will make sure that there will be only a small increase in cost per frame

That's what they want to do - fortunately unstainable - ie double performance - increase price by 180% means pushing price of cards spiralling outwards - there has to be a limit for your base xx80 card that will get enough sales - as for xx90 and xx90 Ti - well whales will happily pay $6000 for bragging rights

Even 20% increase in price will mean double price in 4 new cycles 1.2 to power of 4

 

[Axeia]

I mean 2x the performance sounds good, sadly since it's NVIDIA it'll likely also be a 2x price increase.
Hopefully there will be a RTX 5030, or AMD and Intel simply make APUs/iGPUs on a level where anyone with a 1080p monitor doesn't need to bother with a graphics cards

 

[Neatfeatguy]

"Blackwell is said to offer a massive performance uplift over Lovelace"

Where the disclaimer here? Shouldn't there be an asterisk at the end of this sentence that tells us in very tiny, ant-sized, print at the end of the article that they used DLSS3 with Blackwell and didn't use any DLSS with Ada when they ran comparisons? This way they can still claim, like they did with Ada vs Ampere where Ada was up to 3X faster, that Blackwell is just that much faster than Ada!

 

[Paul Ivan]

I think, people are just blind on what Nvidia is planning on the 4000 series, though the performance leap isn't high but the power consumption is half, if you match the watt vs watt the performance of a 4000 series can be more or doubled. however the lack on vram is something concerning really even if the l3 cache is bigger

 

[cristianm]

I believe it will probably be a 25-30% general performance increase with a 25% power increase, at least for some of the tiers. Sure, it could be 2x in some scenarios, new dlss with 2 generated frames, some new/improved ray tracing mode. But for general performance, I don't think there is room for such big improvements.

 

[wiyosaya]

I'm with everyone else about a massive price increase. IMO, Nvidia's "head" has become massively inflated to the point where they will not learn that people are sick of paying through the (whatever) for their products. It might take an equally spectacular failure with sales of the 5000 series of cards, or perhaps even the 6000 series of cards before Nvidia gets the message.

 

[redgarl]

Really? RGT? Are you guys lowered yourself to the level of WCCF?

On a side note, the 4000 series was accomplished after a major boost in frequency on the best node in the world after being on one of the worst, and still, the only relevant model is a 1600$ 600mm2 4nm GPU, and it is barely 50% faster than a 3090TI at 2160p.

Don't hope for more than 40-60% performance increment over AD102. It is always the same story gen after gen.

https://I.ibb.co/BPk4Xx3/reality.jpg

 

[redgarl]

I think, people are just blind on what Nvidia is planning on the 4000 series, though the performance leap isn't high but the power consumption is half, if you match the watt vs watt the performance of a 4000 series can be more or doubled. however the lack on vram is something concerning really even if the l3 cache is bigger

Typical NVDelusional take... when performance is not there, just talk about efficiency like the good 2050 days...

 

[redgarl]

I believe it will probably be a 25-30% general performance increase with a 25% power increase, at least for some of the tiers. Sure, it could be 2x in some scenarios, new dlss with 2 generated frames, some new/improved ray tracing mode. But for general performance, I don't think there is room for such big improvements.

Even Pascal was not a 2X performance increase... and they are going from a TSMC node to another TSMC node. WIth ADA, it was different, they were going from a Samsung node to a TSMC node. And even there... 50% performance increase...

 

[ScottSoapbox]

2X is probably floating point computation.
50% real world gaming increase for the 5090.
??% increase for the normal cards.

 

[yRaz]

That's what they want to do - fortunately unstainable - ie double performance - increase price by 180% means pushing price of cards spiralling outwards - there has to be a limit for your base xx80 card that will get enough sales - as for xx90 and xx90 Ti - well whales will happily pay $6000 for bragging rights

Even 20% increase in price will mean double price in 4 new cycles 1.2 to power of 4

Maybe I'm reading your post wrong, but what's the point of releasing a new card with the same performance that's MORE expensive?

 

[LetTheWookieWin]

Frankly, it won’t take much to make the 5000 series look good by comparison to the 4000 series. The same thing happened with the Turing to Ampere transition.

I’m not optimistic about pricing, but I’m also not planning to upgrade from my 3080 12G anytime soon. If the “new normal” requires that I pay $700-$900 for a card that’s not almost immediately obsolete, then I’ll just dial down the settings over time and move to a 5 year replacement cycle. No way am I paying that amount every 2-3 years.

 

[tylerb1489]

I am planning to upgrade next year, whole new build, might even get new case and everything. If the 5000 series isn't a sh*tshow then that's definitely on the cards. I want stable 4K gaming on the 5070 level, maaaaybe 5080 if I can run to it.

I want a 5070 with 16gb vram and 4080+ performance level while using a 50-100 less watts for $599.

 

[RandomWAN]

"With Blackwell, there's talk of a 2X improvement, and ray tracing will likely be an important element for Nvidia, as always."

Twice the performance if you count fake frames, maybe. In terms of raw performance, it'll likely only be a 15-20% increase.

 

[wiyosaya]

There has always been talk of a 2X performance boost between generations. When it comes to benchmarks, it never pans out. This is Nvidia trying to generate that "I gotta have it" marketing/talking head drivel.

 

[techdude1]

No problem with horrific performance and overpricing. They invent frame generation version 2.0.