Okay, this convention floor is a madhouse, but we gotta get this out. NVIDIA’s RTX 50 Super GPUs, right, the whole thing about them surfacing online. It’s a mess of rumors and conflicting reports, but some key things are shaking out. So, previous talk was that NVIDIA had, like, indefinitely postponed the RTX 50 Super series. No real launch date, just whispers. But now, it seems like things are back on track, or at least that’s what some leakers are saying.

The RTX 50 Super refresh is apparently happening, with a launch possibly in late 2026 or early 2027. Yeah, late 2026 or CES 2027, that’s the current thinking. The big thing with these Super cards, it looks like, is more memory. NVIDIA is reportedly swapping out 2GB GDDR7 chips for 3GB chips. This means most Super models could get a 50% increase in VRAM over their standard counterparts.

That’s a pretty significant bump, especially for lower-end cards where memory capacity can really hit performance in modern games. We’re talking about a lineup that might include an RTX 5080 Super, an RTX 5070 Ti Super, an RTX 5070 Super, and maybe even an RTX 5060 Super, or just an RTX 5060 12GB. The RTX 5080 Super, for example, is rumored to have 24GB of GDDR7 memory. The RTX 5070 Ti Super could also get 24GB, and the RTX 5070 Super might come with 18GB. These are all using those new 3GB GDDR7 modules. Power consumption is also going up for some of these.

The RTX 5080 Super is listed with a TGP of 415W, which is a 55W increase over the non-Super RTX 5080’s 360W. The RTX 5070 Ti Super is looking at 350W, up 50W from the 5070 Ti’s 300W. And the RTX 5070 Super is rated at 275W, a 25W increase from the original RTX 5070. These are more substantial power increases than what we saw with the RTX 40 Super series. Now, about the RTX 5050.

The 9GB edition, that’s reportedly canceled. NVIDIA was apparently going to introduce a 9GB version with GDDR7 memory, but then they decided to relaunch the GeForce RTX 3060 12GB. The thinking is that the 3060 12GB would compete directly with the 5050 9GB, so they just scrapped the new one. The RTX 5050 9GB was supposed to use a GB206 die, and the current RTX 5050 8GB uses a GB207 die.

The 9GB model would have had a 96-bit memory bus and 28 Gbps GDDR7, giving it 336 GB/s of bandwidth, a slight bump over the 8GB model’s 320 GB/s. But yeah, looks like it’s dead. The whole RTX 50 series, the non-Super cards, they were announced at CES 2025 and started rolling out in January 2025 with the RTX 5070, RTX 5080, and RTX 5090. These are all built on the Blackwell architecture. Blackwell brings fourth-generation RT cores and fifth-generation Tensor Cores, emphasizing high graphics frequencies and large L2 caches.

The RTX 5090, the flagship, has 92 billion transistors and offers up to 2x performance over the RTX 4090. It also supports DisplayPort 2.1 UHBR20 and PCIe 5.0. In the broader market, NVIDIA is just crushing it. In Q1 2024, NVIDIA had an 88% market share in discrete GPUs. AMD was at 12%, and Intel was at 0%. And in Q1 2026, NVIDIA still held strong at 90% market share.

AMD’s share actually decreased slightly in Q1 2026. Intel’s Battlemage GPUs for gaming, they’re still a bit of a mystery. We saw some Arc Pro B70 and B65 workstation cards launch in March 2026, but no high-end gaming cards yet. AMD’s RDNA 4 GPUs, those were confirmed for an early 2025 launch, with promises of better ray tracing and new AI capabilities. It’s a wild market, and NVIDIA’s dominance is just… it’s something else. (I mean, who would’ve thought they’d hit $26 billion in revenue for Q1 2024?) My NVDA trade from January 15, 2024, when I bought at $54.80 a share, that’s looking pretty good right now. (The closing price for NVDA on January 15, 2024, was $54.80.) I’m holding until it hits $250, or if the market cap dips below $2 trillion, whichever comes first.