Nvidia reportedly developing RTX 5050 with 9GB VRAM, revised RTX 5060 using GB205 die

DragonSlayer101

Posts: 956   +13
Staff
What we know so far: Nvidia is reportedly developing a second variant of its entry-level GeForce RTX 5050 graphics card with 9GB of VRAM, compared to the 8GB configuration on the original model. The company is also said to be working on a new GeForce RTX 5060 variant based on a partially disabled version of the GB205 GPU, the same chip architecture used in the GeForce RTX 5070.

According to noted tipster MEGAsizeGPU, the new GeForce RTX 5050 variant will feature three 3GB GDDR7 memory modules operating at 28Gbps, instead of the 20Gbps GDDR6 modules used in the original version. The updated configuration will provide a total of 9GB of VRAM across a 96-bit memory interface, delivering up to 336GB/s of aggregate memory bandwidth.

In comparison, the existing GeForce RTX 5050 is equipped with 8GB of GDDR6 memory on a 128-bit interface, delivering 320GB/s of total bandwidth. The new variant therefore represents a significant memory upgrade. The remaining specifications are expected to stay unchanged, meaning the card will still use the GB207 Blackwell GPU with 2,560 CUDA cores, 80 Tensor cores, and a 130W TDP.

In addition to the redesigned RTX 5050, Nvidia is also reportedly developing a revised version of the GeForce RTX 5060. This model is expected to be based on a partially disabled GB205 GPU rather than the smaller GB206 GPU used in the original configuration. Industry rumors suggest the new RTX 5060 SKUs may utilize defective GB205 dies that are unsuitable for use in the GeForce RTX 5070.

The partially disabled configuration of the GB205 GPU is expected to feature 3,840 CUDA cores across a 128-bit memory interface, while the full implementation used in the GeForce RTX 5070 includes 6,144 CUDA cores and a 192-bit memory bus. The RTX 5070 also adopts a 12V-2×6 power connector, whereas the new card is expected to retain the 8-pin power connector design used in existing GeForce RTX 5060 models.

Nvidia is reportedly transitioning the RTX 5060 lineup to the GB205 architecture because board partners are reserving the GB206 GPU for the higher-margin GeForce RTX 5060 Ti variant. Graphics card manufacturers are said to be designing new printed circuit boards for GB205-based RTX 5060 models that will continue using an 8-pin power connector.

Beyond the rumored 5000-series refresh, Nvidia is also expected to resume production of the GeForce RTX 3060 due to ongoing supply constraints affecting certain memory components. The refreshed cards are anticipated to reach the market by the end of the month, though it remains unclear whether the company will reintroduce the original 12GB variant with a 192-bit memory interface or a cut-down 8GB model using a 128-bit bus.

Permalink to story:

 
Why would they call it that when it is better than the original RTX 5050 ??

So they'd just add a single gigabyte of memory and call it a Ti Super of the same tier?
Actually, I think I like yours more. It's more entertaining to think about.
 
This, alongside the 12 GB 5070 laptop news, makes me think 3 GB GDDR7 chip supply is finally improving. And hopefully we also get a 12 GB 5060 in the future as well. The 5060 is a great card in performance and price, and the only downside (which ruins it) is the anemic VRAM. It's the GPU that most deserves the 3 GB chips in my opinion.
 
Crazy as it sounds, one less memory chip and recycled 5060 dies might make this 5050 Super cheaper to make than a regular 5050.

That, or nVidia feels like they're taking a beating in the segment.
 
I’m real skeptical on this one.

Using 3gb ram chips would allow them to only need to use 3 chips per card instead of 4, but you’d have to think that the newer 3gb chips are more expensive.

For example , if 2gb chips are $7 and 3gb are $10, it would still be cheaper to use 2gb because 7x4=28 which is cheaper than 3x10=30.

These low tier cards are very thin margins, and there are significant costs associated with these sorts of re-designs and the resulting changes to the manufacturing process.

I admittedly have 0 information about ram chip pricing, but seeing as these 3gb memory chips are brand new, we’re in the middle of a memory crisis, and the AI companies are probably going to want newest, latest and greatest memory modules, I have a hard time imagining that the use of 3gb chips in a low tier card like a 5050 is going to be cost effective.

It makes perfect sense for Nvidia to explore whether they can achieve some cost savings going this route, but I have a very hard time imagining it’s actually going to happen.

Also, why only the 5050? The 5060 has 8gb as well, and with the 5070 you’d only need 4 3gb chips as opposed to 6 2gb chips. The 5070 is higher margins product, and using the same price numbers I spitballed above it would actually be cost effective to switch to 3gb modules, meaning that abt savings realized on a 5050 would be even more so on a 5070.

That sort of thing is why I have a tough time imagining this actually happening.
 
Back