Nvidia reportedly working on RTX 5050 with 9GB of VRAM on a 96-bit bus, featuring 28 Gbps GDDR7 modules — RTX 5060 with cut-down GB205 GPU also planned

News
By published

Moving away from 8GB GDDR6 modules.

Nvidia GeForce RTX 5070
(Image credit: Nvidia)

The global component shortage is apparently affecting Nvidia — one of the catalysts of this very crisis — just as hard as other manufacturers, with the consumer segment taking the hit. Memory and chip constraints are leading the company to think of interesting new products — a new RTX 5050 with 9GB of VRAM seems to be in the works, according to leaker @Zed__Wang.

The leaker goes on to mention that Nvidia could build 12GB variants of the RTX 5050 and 5060 with 3GB GDDR7 chips, but perhaps the company isn't concerned with value maximization like that, especially not in these times. The last time a desktop GPU from Nvidia rocked a 96-bit interface was back in 2024 with the RTX 3050, so it's actually not too distant of a memory.

The rest of the specs should remain identical; the same GB207 die with 2,560 CUDA cores, built on TSMC's 5nm process, rated at 130W. Clock speed differences (if any) will become public knowledge as we near the potential release of this SKU. The RTX 5050 is one of the few GPUs that basically saw no price hikes in the past few months, only going up about $10. The only other card with the same stability was the RTX 5060.

Article continues below

Speaking of which, @Zed__Wang also talked about a new RTX 5060 in the works with a cut-down version of the GB205 GPU — that's the silicon that powers the RTX 5070. Apparently, Nvidia has told AIBs to focus on the 8GB RTX 5060 Ti, which has led to a shortage of GB206 dies for RTX 5060 SKUs. Since both the RTX 5060 and 5060 Ti use the GB206, the chipmaker is now forced to switch gears.

Defective GB205 dies might be repurposed to fit into new PCB designs for the RTX 5060 with an 8-pin power connector. Otherwise, the RTX 5070 has a 12V-2x6 connector and the GB205 die inside features 6,144 CUDA cores — those would be reduced to 3,840 CUDA cores for an RTX 5060. Moreover, the bus width on the GB205 is 192-bit and that would also go down, to 128-bit, in order to match an RTX 5060.

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Hassam Nasir
Hassam Nasir
Contributing Writer

Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.

29 Comments Comment from the forums
  • Gururu
    Very interesting adjustment by nVidia. I have to say that the 5050-5060 cards are positioning themselves for long term domination.
    Reply
  • LordVile
    Gururu said:
    Very interesting adjustment by nVidia. I have to say that the 5050-5060 cards are positioning themselves for long term domination.
    Don’t really see that tbh, they’re better in budget builds for pricing but they’re massively impacted by the x8 configuration in anything that isn’t PCIE 5 which isn’t budget. The 9060 series 8GB cards that has the full x16 lanes is significantly less impacted on systems with PCIE 4 and especially PCIE 3 where the x8 8GB cards are effectively unusable a lot of the time.
    Reply
  • PEnns
    9GB of VRAM on a 96-bit bus?
    What's next. a 7.5 GB and 64-bit bus version??

    "The RTX 5050 is one of the few GPUs that basically saw no price hikes in the past few months,"
    Yeah, tough to raise prices on the bottom of the barrel stuff.
    Reply
  • Gururu
    LordVile said:
    Don’t really see that tbh, they’re better in budget builds for pricing but they’re massively impacted by the x8 configuration in anything that isn’t PCIE 5 which isn’t budget. The 9060 series 8GB cards that has the full x16 lanes is significantly less impacted on systems with PCIE 4 and especially PCIE 3 where the x8 8GB cards are effectively unusable a lot of the time.
    I hope not but I have no doubt they will compose 90% of Steam survey GPUs by end of 2027.
    Reply
  • LordVile
    Gururu said:
    I hope not but I have no doubt they will compose 90% of Steam survey GPUs by end of 2027.
    Nah mid range from the last couple of gens will be. 60 and 70 cards are typically the most popular
    Reply
  • Notton
    I find it hard to believe Nvidia of all companies would do this.
    Why waste perfectly good 3GB GDDR7 modules on a relatively low profit margin card, when they could milk more from an RTX Pro 6000?
    Did 2GB GDDR6 go up in price so insanely that it's cheaper to use ultra high demand 3GB GDDR7 chips?
    I thought they couldn't get rid of their 5050/5060Ti 8GB stock fast enough and are sitting on a stockpile of it?
    Reply
  • DS426
    Notton said:
    I find it hard to believe Nvidia of all companies would do this.
    Why waste perfectly good 3GB GDDR7 modules on a relatively low profit margin card, when they could milk more from an RTX Pro 6000?
    Did 2GB GDDR6 go up in price so insanely that it's cheaper to use ultra high demand 3GB GDDR7 chips?
    I thought they couldn't get rid of their 5050/5060Ti 8GB stock fast enough and are sitting on a stockpile of it?
    I don't see how the economics pans out either; I couldn't be more surprised, lol. I know AMD is churning through a lot of GDDR6, but that's still nothing compared to AI-related demand for nVidia GPU's and therefore GDDR7. Why there is ANY engineering and other resources going into a 5050 refresh is mind-boggling. Maybe it's more of a streamlining and simplifying effort as then nVidia doesn't have to source any GDDR6. Or how about this: they asked their AI and it told them it was a good idea. :ROFLMAO:
    Reply
  • beyondlogic
    Notton said:
    I find it hard to believe Nvidia of all companies would do this.
    Why waste perfectly good 3GB GDDR7 modules on a relatively low profit margin card, when they could milk more from an RTX Pro 6000?
    Did 2GB GDDR6 go up in price so insanely that it's cheaper to use ultra high demand 3GB GDDR7 chips?
    I thought they couldn't get rid of their 5050/5060Ti 8GB stock fast enough and are sitting on a stockpile of it?

    why would they do this simply to bump up there higher cards in price its a price bumping method.

    sell this card for 300 then bump the rest up lol.

    depending on if this is real though id buy it if it comes in small form factor

    amds Achilles heel is it has zero small form factor cards that are decent.
    Reply
  • thestryker
    Notton said:
    Did 2GB GDDR6 go up in price so insanely that it's cheaper to use ultra high demand 3GB GDDR7 chips?
    The 5050 is the only current card using GDDR6 which means their purchasing volume is much lower. It may actually be cheaper for them to use 3x 3GB GDDR7 packages than 4x 2GB GDDR6 even though on the open market it wouldn't be. Getting more cards out of fewer packages might be part of it as well.
    Reply
  • Gururu
    thestryker said:
    The 5050 is the only current card using GDDR6 which means their purchasing volume is much lower. It may actually be cheaper for them to use 3x 3GB GDDR7 packages than 4x 2GB GDDR6 even though on the open market it wouldn't be. Getting more cards out of fewer packages might be part of it as well.
    This is what I am getting from other tech reports. Not even AMD or Intel are using GDDR7 so nVidia is in fact the primary buyer of all GDDR7 RAM produced.
    Reply