Tuesday, March 10th 2026

NVIDIA GeForce RTX 5050 9 GB Variant Comes with 130 W TDP

NVIDIA is reportedly preparing to update its GeForce RTX 5050 GPU with a new variant featuring 9 GB of GDDR7 memory. A well-known leaker on X, @kopite7kimi, confirmed that the upcoming card will maintain the same 130 W TDP and thermal envelope as the current GeForce RTX 5050, which has 8 GB of GDDR6 memory. The current RTX 5050 uses 8 GB of 20 Gbps GDDR6 memory on a 128-bit bus, providing 320 GB/s of bandwidth. The new model is expected to use three modules of GDDR7 memory, each with 3 GB of capacity, resulting in a total of 9 GB across a 96-bit memory bus. While the narrower bus decreases the interface width, the switch to new 28 Gbps GDDR7 memory would increase total memory bandwidth to 336 GB/s, a roughly 5% improvement, along with a 12.5% boost in VRAM capacity. The new leak also claims that NVIDIA is using GB206 as a GPU base, while the older RTX 5050 8 GB with GDDR6 used GB207 die. However, the core count remains at 2,560 CUDA cores, suggesting that lower-binned GB206 dies found in RTX 5060 and other mid-range SKUs are repurposed for the new RTX 5050.

NVIDIA's switch to GDDR7 memory likely helps the company manage supply chain procurement better, as memory modules are in short supply. Instead of using four GDDR6 modules with 2 GB capacity each, NVIDIA is switching to three modules of GDDR7 with 3 GB capacity each, reducing the number of memory modules needed for this GPU. Interestingly, memory makers like Samsung, Micron, and SK hynix could now produce more of the 3 GB GDDR7 modules, with GDDR6 being in much shorter supply, forcing manufacturers to focus on the newer memory technology while GPU makers have to adapt. This has resulted in a situation where procuring three modules of 3 GB GDDR7 memory is now easier than finding a sufficient supply of GDDR6 modules, which NVIDIA needs to secure for each RTX 5050 GPU. As we approach Computex 2026, this GPU is expected to arrive sometime during that period.
Source: @kopite7kimi
Add your own comment

25 Comments on NVIDIA GeForce RTX 5050 9 GB Variant Comes with 130 W TDP

#1
Bwaze
Blackwell Super refresh with all the cards getting about the same quantity of VRAM with fewer memory modules, lower bus width but about the same bandwidth?
Posted on Reply
#2
ebivan
BwazeBlackwell Super refresh with all the cards getting about the same quantity of VRAM with fewer memory modules, lower bus width but about the same bandwidth?
Dont forget that this will come with a price hike which is guaranteed to be a lot steeper than the 5% of increased bandwidth...
Posted on Reply
#3
dgianstefani
TPU Proofreader
Not bad for $250, but for most the 8 GB 5060 for $40 more would still be a better buy.
Posted on Reply
#4
azrael
This feels like a joke ...at our expense.
Posted on Reply
#5
Quicks
96 bit bus, what a waste of good 3GB Gddr7.

The level these companies would go to screw their customers is unquestionably bad.
Posted on Reply
#6
Bwaze
ebivanDont forget that this will come with a price hike which is guaranteed to be a lot steeper than the 5% of increased bandwidth...
Apply Jensen's Law.

"Moore's Law is dead … It's completely over, and so the idea that a chip is going to go down in cost over time, unfortunately, is a story of the past."
Posted on Reply
#7
jabbadap
OEM only, or?
The GPU base remains the same with GB206 die featuring 2,560 CUDA cores
RTX 5050 has GB207 die, GB206 at 2560 ccs would be cut down core from 4608 cc. Frankly quite normal usage for broken dies from nvidia.
Posted on Reply
#8
AleksandarK
News Editor
jabbadapRTX 5050 has GB207 die, GB206 at 2560 ccs would be cut down core from 4608 cc. Frankly quite normal usage for broken dies from nvidia.
Whoops updated the article now, thanks!
Posted on Reply
#9
Quicks
BwazeApply Jensen's Law.

"Moore's Law is dead … It's completely over, and so the idea that a chip is going to go down in cost over time, unfortunately, is a story of the past."
These people writing their own laws now as well.

Going by that logic everything will become too expensive and tech will just stagnant and kill itself. Seems to be happening already, if you look at the new GPU releases that didn't bring much performance to the table from previous generations.

That means space travel will never happen because it's will simply be too expensive and technology would never evolve, at least not quick enough to warrant research into space travel.
Posted on Reply
#10
Bwaze


They're not even considering a market of flesh and blood consumers as worth pursuing any more, what we have are the remnants of products designed before the great AI takeover of everything.
Posted on Reply
#11
Philaphlous
9GB? LOL they should have make the TDP value 131W instead of 130 just to keep things on the odd side....
Posted on Reply
#12
lemonadesoda
Philaphlous9GB? LOL they should have make the TDP value 131W instead of 130 just to keep things on the odd side....
121W would be better. 3^2 GB, and 11^2 W. Odd primes squared.
Posted on Reply
#13
notoperable
Just, Jensen - could you just f****ff finally
Posted on Reply
#14
wNotyarD
If only the 5050 was slot-powered. If only.
Posted on Reply
#15
Warigator
seriously ..... GeForce X050 card prices and energy demands are too high

130 watts for ..... a 50-class card? I would understand 60-class card, but 50?
Posted on Reply
#16
NoneRain
And, as always, this will sell enough. Sad world.
Posted on Reply
#17
TheinsanegamerN
Warigatorseriously ..... GeForce X050 card prices and energy demands are too high

130 watts for ..... a 50-class card? I would understand 60-class card, but 50?
550ti was 116w for stock, 130+ for OC models.

Old men have been yelling at that cloud for almost 15 years now. There's no law that a certain class is limited to certain power restrictions.
NoneRainAnd, as always, this will sell enough. Sad world.
IF this is all you need, why pay extra for more hardware? People cry when there is no low end hardware then cry more when low end hardware is made with parts bin hardware.
Posted on Reply
#18
Athena
It's clearly a cost cutting maneuver, it makes total sense for Nvidia, less chips are needed, so they can have more cards out there, but, the prices won't really change much, if at all

the question is, if AMD & Intel will follow, or will they go for 12GB, even though this will be much more expensive, so they won't be able to compete on price
Posted on Reply
#19
gord787
AleksandarKNVIDIA is reportedly preparing to update its GeForce RTX 5050 GPU with a new variant featuring 9 GB of GDDR7 memory. A well-known leaker on X, @kopite7kimi, confirmed that the upcoming card will maintain the same 130 W TDP and thermal envelope as the current GeForce RTX 5050, which has 8 GB of GDDR6 memory. The current RTX 5050 uses 8 GB of 20 Gbps GDDR6 memory on a 128-bit bus, providing 320 GB/s of bandwidth. The new model is expected to use three modules of GDDR7 memory, each with 3 GB of capacity, resulting in a total of 9 GB across a 96-bit memory bus. While the narrower bus decreases the interface width, the switch to new 28 Gbps GDDR7 memory would increase total memory bandwidth to 336 GB/s, a roughly 5% improvement, along with a 12.5% boost in VRAM capacity. The new leak also claims that NVIDIA is using GB206 as a GPU base, while the older RTX 5050 8 GB with GDDR6 used GB207 die. However, the core count remains at 2,560 CUDA cores, suggesting that lower-binned GB206 dies found in RTX 5060 and other mid-range SKUs are repurposed for the new RTX 5050.

NVIDIA's switch to GDDR7 memory likely helps the company manage supply chain procurement better, as memory modules are in short supply. Instead of using four GDDR6 modules with 2 GB capacity each, NVIDIA is switching to three modules of GDDR7 with 3 GB capacity each, reducing the number of memory modules needed for this GPU. Interestingly, memory makers like Samsung, Micron, and SK hynix could now produce more of the 3 GB GDDR7 modules, with GDDR6 being in much shorter supply, forcing manufacturers to focus on the newer memory technology while GPU makers have to adapt. This has resulted in a situation where procuring three modules of 3 GB GDDR7 memory is now easier than finding a sufficient supply of GDDR6 modules, which NVIDIA needs to secure for each RTX 5050 GPU. As we approach Computex 2026, this GPU is expected to arrive sometime during that period.
Hmm, 9 GB is such a weird number for VRAM, haha, feels like someone just glued an extra stick on there and called it a day. I get the GDDR7 bump helping bandwidth a bit, but on a 128 bit bus this thing still sounds pretty entry level honestly. Guess the real question is price, if it lands cheap then sure, if not… yeah people will just grab a used 3060 or something.
Posted on Reply
#20
efikkan
AthenaIt's clearly a cost cutting maneuver, it makes total sense for Nvidia, less chips are needed, so they can have more cards out there, but, the prices won't really change much, if at all
No, it's a matter of improving yields. This is salvaging tens of thousands of perfectly good chips which would otherwise go to waste.

This is the second thread within a week about this subject, full people complaining about this product, as a testament to their lack of knowledge or need to complain for no rational reason.

Nvidia and AMD does this all the time, using leftover bins from other chips to create variants of the "same" product, with minor variations, or OEM versions. And it shouldn't be a problem unless they have different performance characteristics, like the RTX 3050 6 GB.

We should expect to see cut down GB205 and GB203 as RTX 5060 too.

-----

I would wish both Nvidia and AMD refreshed their current lineup to provide the best availability for consumers, even if it meant using fewer larger chips.
E.g. replacing the RX 9070 (non-XT) with a 12 GB version would be a brilliant move.
Or a RTX 3070 "Super" with still 12 GB and more bandwidth would be welcome in the current market conditions, and a 3070 Ti "Super" with 18GB?. I know many are obsessed with VRAM they don't really need, but I would rather have more great GPUs in the hands of gamers with sufficient VRAM for their intended usage.
Posted on Reply
#21
Athena
efikkanNo, it's a matter of improving yields. This is salvaging tens of thousands of perfectly good chips which would otherwise go to waste.
of course all companies bin GPUs (and CPUs), and salvage what they can to use on lower end products

It's still a cost cutting measure, like I said
Posted on Reply
#22
notoperable
Jensen is gonna tell us next week that 6.40gb is enough for everyone. Mark my words.
Posted on Reply
#23
mechtech
hmmm paid $170usd for the RX6600 I got a few years ago. 28% more performance to $ would be about $218.............see what actual store prices are.
Posted on Reply
#24
NoneRain
TheinsanegamerNIF this is all you need, why pay extra for more hardware? People cry when there is no low end hardware then cry more when low end hardware is made with parts bin hardware.
I agree, but I also think Nvidia takes advantage of consumers buying something cheaper and fit for their needs right now, to push lower RAM, for example, in not so basic GPUs while not really cutting the equivalent price. Overpriced and downgraded, because competition is always behind. Less RAM today, and in 2 years you already want/need a better one.

BUT, RAM is gold rn, so...
Posted on Reply
#25
OneSorcerer
if it sells for 250 during black friday, I will grab one. good enough for windowed path of exile 2 @1440p on a 4k monitor and just in time for the game release (supposedly)
Posted on Reply
Add your own comment
Nov 16th, 2026 04:24 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts