Wednesday, September 10th 2025

NVIDIA GeForce RTX 50-Series SUPER GPUs Could Arrive at CES 2026

Widespread speculation surrounding the possible launch of the NVIDIA GeForce RTX 50 SUPER series has shifted, with industry sources confirming that the official timeline for the enhanced GPU variants has been moved by a few months. According to Hongxing2020 hardware leaker, known for his ties with GPU AIBs, he claims that the RTX 50 SUPER series has been delayed. Recent online reports have claimed Christmas as the release date; however, these may not be accurate. NVIDIA remains committed to its current RTX 50 Blackwell series for the following months, which continues to operate at full production capacity with no discontinuation plans in sight.

According to VideoCardz, a possible mid-cycle "Blackwell" refresh in the form of SUPER GPUs could hit at CES 2026, when the announcement is expected. NVIDIA plans to increase the memory sizes of key performance-segment and high-end graphics card SKUs to boost their performance in higher resolutions, particularly with ray tracing and neural rendering enabled. Specifically, the company plans to leverage new 24 Gbit density GDDR7 memory chips, which allow it to create memory sizes such as 18 GB over a 192-bit wide memory bus or 24 GB across a 256-bit wide memory bus. Besides the added memory, NVIDIA is expected to increase shader counts and GPU clock speeds marginally.
Source: VideoCardz
Add your own comment

69 Comments on NVIDIA GeForce RTX 50-Series SUPER GPUs Could Arrive at CES 2026

#1
GodisanAtheist
Still a ways off then.

I figured we were slated for a pre-christmas launch but I guess the standard 5xxx series must be selling ok (or NV just doesn't care about consumer anymore, maybe a combination of the two).
Posted on Reply
#2
Philaphlous
GB205 is where it's at. Can't wait to get my 5070ti laptop...OC the crap outta that thing...maybe shunt mod....
Posted on Reply
#3
Tomorrow
25-55W higher TDP. That's more than i expected. Coupled with AMD using second sourced 20W more efficient memory on RX 9000 series cards this means that Blackwell completely loses it's efficiency edge.

As for the cards themselves it seems performance improvement will be miniscule, if nonexistent in games.
If they somehow manage to keep the price the same then at least 5080 will have the amount on ram it should have launched with, but 24GB 5070 Ti Super will immediately undercut it anyway. Just like 5080 is pointless now. As for 5070 Ti then i doubt most people buying it felt limited by 16GB.

5070 moving from 12GB to 18GB is good assuming Nvidia will not price it horribly compared to 9070.
Nvidia should also discontinue existing 5060 8GB and 5060 Ti 8GB models and make 12GB models of them instead at same prices as current 8GB models.
Posted on Reply
#5
Bwaze
I could imagine these Super models with extra memory (but otherwise no extra performance increase, which at least for RTX 5080 is extremely disappointing - it's basically RTX 4080) could be marketed directly for AI / LLM / prosumer market, again iterating that Nvidia is a grown up business now that caters mostly to serious users, not joystick wankers.

With price tag fitting the market.
Posted on Reply
#6
mb194dc
It's a massive yawn anyway isn't it? No massive difference to what's on the market now. Just prices are still high compared to the 2020 and before period.
Posted on Reply
#7
MrDweezil
I figured they'd be launching these alongside some new RTX tech that makes use of the extra memory, but seems like there's no word on that. Otherwise the only thing that stands out is that the 5070 SUPER is a better match for non-XT 9070.
Posted on Reply
#8
phints
These wattages are hilarious. Imagine a 70 tier at 275W. They are badly in need of a new gen still using TSMC 5N ain’t cutting it. This whole gen is a waste waiting for 6070.
Posted on Reply
#9
TheinsanegamerN
mb194dcIt's a massive yawn anyway isn't it? No massive difference to what's on the market now. Just prices are still high compared to the 2020 and before period.
And prices will remain there. We printed 50 percent of our currency in 2020 alone and have spent additional trillions since.
Posted on Reply
#10
N/A
I can't settle for anything less than 12800 CUDA.
Posted on Reply
#11
LastDudeALive
N/AI can't settle for anything less than 12800 CUDA.
Such a silly way of looking at things. The 7168 cores in my 4070 Super do the same work as the 10496 cores in a 3090.
Posted on Reply
#13
AusWolf
TomorrowNvidia should also discontinue existing 5060 8GB and 5060 Ti 8GB models and make 12GB models of them instead at same prices as current 8GB models.
Those would be way more interesting than any of the above speculated models, imo.
Posted on Reply
#14
Why_Me
Tomorrow25-55W higher TDP. That's more than i expected. Coupled with AMD using second sourced 20W more efficient memory on RX 9000 series cards this means that Blackwell completely loses it's efficiency edge.

As for the cards themselves it seems performance improvement will be miniscule, if nonexistent in games.
If they somehow manage to keep the price the same then at least 5080 will have the amount on ram it should have launched with, but 24GB 5070 Ti Super will immediately undercut it anyway. Just like 5080 is pointless now. As for 5070 Ti then i doubt most people buying it felt limited by 16GB.

5070 moving from 12GB to 18GB is good assuming Nvidia will not price it horribly compared to 9070.
Nvidia should also discontinue existing 5060 8GB and 5060 Ti 8GB models and make 12GB models of them instead at same prices as current 8GB models.
Let's compare apples to apples such as the RX 9070 XT vs the RTX 5070 Ti.

Posted on Reply
#15
Tomorrow
Why_MeLet's compare apples to apples such as the RX 9070 XT vs the RTX 5070 Ti.
This is the current state and in this table 9070 XT uses Hynix G6. Not Samsung G6.

You really think that 5070 Ti Super will have the same 279W power consumption with 8GB more G7?
Meanwhile you can subtract 20 from the 314 number and likely add 25 to 5070 Ti number and that's the chart we will be looking after 50 super series launches.

Not that i expect TPU to retest 9070 XT at that time. I mean the last one they tested was nearly four months ago, but they have already added 50 super series to the performance chart with zero testing.
Posted on Reply
#16
AusWolf
TomorrowThis is the current state and in this table 9070 XT uses Hynix G6. Not Samsung G6.

You really think that 5070 Ti Super will have the same 279W power consumption with 8GB more G7?
Meanwhile you can subtract 20 from the 314 number and likely add 25 to 5070 Ti number and that's the chart we will be looking after 50 super series launches.

Not that i expect TPU to retest 9070 XT at that time. I mean the last one they tested was nearly four months ago, but they have already added 50 super series to the performance chart with zero testing.
Modern graphics cards run up to their specified TDP. A 304 W 9070 XT (the Powercolor Reaper is the only such model that I know of) will always consume 304 W at full load. Whether the VRAM eats 20 or 40 W out of that 304 doesn't affect the overall power consumption of the card, only the headroom left for the GPU chip.

On the other hand, I do think that arguing about 20-ish W between the 5070 Ti and 9070 XT is pointless.
Posted on Reply
#17
Onasi
AusWolfModern graphics cards run up to their specified TDP. A 304 W 9070 XT (the Powercolor Reaper is the only such model that I know of) will always consume 304 W at full load.
It’s not the only one, all the base frequency non-OC models are, on paper, adherent to the base TBP. How it works out in practice is another thing entirely - some are actually MORE efficient due to better power delivery systems, some lose in the silicon lottery and go a tad above. It’s not a zero sum game.
Posted on Reply
#18
AusWolf
OnasiIt’s not the only one, all the base frequency non-OC models are, on paper, adherent to the base TBP.
I know that, but I mean the PC Reaper is the only base TDP model that I know about.
OnasiHow it works out in practice is another thing entirely - some are actually MORE efficient due to better power delivery systems, some lose in the silicon lottery and go a tad above. It’s not a zero sum game.
Sure, but the total continuous power consumption of the card during 100% load will always be the manufacturer specified TDP. The only thing those variables change is how much headroom the GPU chip has left.
Posted on Reply
#19
Tomorrow
AusWolfModern graphics cards run up to their specified TDP. A 304 W 9070 XT (the Powercolor Reaper is the only such model that I know of) will always consume 304 W at full load. Whether the VRAM eats 20 or 40 W out of that 304 doesn't affect the overall power consumption of the card, only the headroom left for the GPU chip.

On the other hand, I do think that arguing about 20-ish W between the 5070 Ti and 9070 XT is pointless.
Recent 9060 XT testing with Samsung G6 disproves what you're saying. 9060 XT was consuming 20W less. Having more efficient VRAM on board does not mean that the extra power will simply go to the GPU. It's dGPU. Not an iGPU. The GPU consumes what it needs regardless of how efficient the VRAM is.

The GPU will not suddenly consume 20W more because the VRAM is consuming 20W less. Most 9070 XT cards have TBP around 350W. 314W is not even the maximum. My card for example has TBP of 363W (+10% manual adjustment included) and i believe the AsRock Taichi model goes to 404W.
Besides on cards using Samsung G6 the TDP is not lowered from 304W to 284W. It's still 304W.
Posted on Reply
#20
AusWolf
TomorrowRecent 9060 XT testing with Samsung G6 disproves what you're saying. 9060 XT was consuming 20W less. Having more efficient VRAM on board does not mean that the extra power will simply go to the GPU. It's dGPU. Not an iGPU. The GPU consumes what it needs regardless of how efficient the VRAM is.

The GPU will not suddenly consume 20W more because the VRAM is consuming 20W less. Most 9070 XT cards have TBP around 350W. 314W is not even the maximum. My card for example has TBP of 363W (+10% manual adjustment included) and i believe the AsRock Taichi model goes to 404W.
Besides on cards using Samsung G6 the TDP is not lowered from 304W to 284W. It's still 304W.
That could be true if the GPU reaches its own limits before reaching those of the whole card (or if it's RDNA 2 or earlier). I haven't seen it in practice lately, though.
Posted on Reply
#21
Tomorrow
Yep. If the card is already consuming it's maximum TBP then more efficient VRAM will likely not make it consume 20W less, but mostly i dont see my card being constantly peaked at the maximum 363W.
Posted on Reply
#22
Why_Me
TomorrowThis is the current state and in this table 9070 XT uses Hynix G6. Not Samsung G6.

You really think that 5070 Ti Super will have the same 279W power consumption with 8GB more G7?
Meanwhile you can subtract 20 from the 314 number and likely add 25 to 5070 Ti number and that's the chart we will be looking after 50 super series launches.

Not that i expect TPU to retest 9070 XT at that time. I mean the last one they tested was nearly four months ago, but they have already added 50 super series to the performance chart with zero testing.
At the moment the RX 9070 XT uses 40W more than the RTX 5070 Ti. The RTX 5070 Ti Super is due to use 50W more than the current 5070 Ti.
Posted on Reply
#24
Tomorrow
Why_MeAt the moment the RX 9070 XT uses 40W more than the RTX 5070 Ti. The RTX 5070 Ti Super is due to use 50W more than the current 5070 Ti.
I wasn't talking about the current moment. As i said in my post "this means that Blackwell completely loses it's efficiency edge".
It was implied that Blackwell loses it's efficiency edge. Not that it already lost it.

Tho i dont believe 50W rise in TDP will means actually 50W increase in power consumption during gaming. Im guessing around 25W maybe because i dont want to overstate the eventual increase. The number of memory chips on board stays the same but obviously higher capacity modules consume more power.
Posted on Reply
#25
AusWolf
TomorrowYep. If the card is already consuming it's maximum TBP then more efficient VRAM will likely not make it consume 20W less, but mostly i dont see my card being constantly peaked at the maximum 363W.
Is the maximum TBP of your card reported as 363 W in HWinfo?
Posted on Reply
Add your own comment
Nov 16th, 2026 13:43 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts