ALITEQ.

Nvidia finished building a better RTX 5070 months ago. Then it just... didn't ship it

18GB of VRAM, the same core, hardware reportedly already sitting with board partners — and still no release date. Here's the real reason, and what I'd buy instead this week.

Ravi MalhotraUpdated 1h ago6 min read
An Nvidia GeForce RTX 50-series graphics card

Nvidia's RTX 5070 Super reportedly exists. Board partners are said to have working hardware, the specs are locked in — 18GB of GDDR7, up from the standard 5070's 12GB — and it still isn't launching. Nvidia has apparently put the entire RTX 50 Super refresh on indefinite hold, and the reason isn't a design flaw. It's the same memory shortage that just pushed prices up on the RTX 50-series cards you can actually buy today.

What you'd actually be getting

The jump from 12GB to 18GB doesn't come from a redesigned chip — it comes from swapping 2GB GDDR7 modules for 3GB ones on the same 192-bit memory bus. Same bus width, same 28Gbps memory speed, same 672GB/s of bandwidth by the math (28Gbps times 24 bytes of bus width). CUDA core count reportedly moves from 6,144 to 6,400, a small bump, not a redesign. This is Nvidia taking an existing card and giving it more room to hold larger models or higher-resolution texture sets — not a faster GPU, a roomier one.

RTX 5070 vs. RTX 5070 Super (rumored)

RTX 5070

shipping now

vs

RTX 5070 Super

shelved

12GB GDDR7
VRAM
18GB GDDR7
6,144
CUDA cores
6,400
192-bit
Memory bus
192-bit
672 GB/s
Memory bandwidth
672 GB/s
$549 (actual)
Rumored MSRP
$549 (rumored, at risk)
Shipping since launch
Availability
Indefinitely on hold
5070 wins 1wins 2 Super

Why Nvidia buried its own upgrade

The 3GB memory modules the Super needs are reportedly the expensive part. Enterprise AI datacenters have been buying up memory inventory aggressively enough that GDDR7 pricing has climbed sharply for everyone, and a card that specifically needs the newer, larger modules gets hit hardest. Nvidia has apparently decided that shipping the Super at anything close to its rumored $549 price isn't viable right now, and there's no confirmed date for revisiting that math. Computex 2026 is the next plausible announcement window, and even that's being described as a long shot rather than a sure thing.

RTX 50-series China distributor price hikes, by card

RTX 5080 Vulcan W OC (Colorful)+59%

vs MSRP

RTX 5070 Ti Vulcan W OC (Colorful)+58.5%

vs MSRP

RTX 5080 Gaming Trio OC (MSI)+18.2%

9,999 to 11,999 yuan

RTX 5080 Shadow / Ventus (MSI)+20%

vs prior week

Is this China-only, or coming everywhere?

The specific 59% and 58.5% jumps from Colorful and MSI are confirmed only for Chinese distributor pricing so far — nobody's reporting an equivalent move in the US or Europe yet. But the underlying cause isn't regional. It's the same DRAM and memory-chip shortage that's already made its way into products that have nothing to do with gaming, and it's the same one behind every other price story we've covered this year. When a shortage is squeezing memory-chip supply globally, a distributor price hike in one country is usually the first domino, not the only one.

Close-up of memory modules on a graphics card circuit board
GDDR7 modules are the exact component both the shelved RTX 5070 Super and the China price hikes trace back to. · Unsplash

Verdict

Buy now, or wait for the Super?

Buy now, if you were going to buy a 5070-class card anyway. There's no confirmed date for the Super, the price protection it was rumored to have is already in doubt, and 12GB of VRAM on the regular card covers the overwhelming majority of 1440p gaming today.

Best for: Buy now: anyone who needs a GPU this year and isn't chasing 4K. Wait: only if you specifically need 18GB or more for local AI or heavy content work, and can tolerate an open-ended timeline — check what actually needs that much VRAM first.

RTX 5070 Super questions

Is the RTX 5070 Super coming out?
Nvidia has reportedly put it on indefinite hold. Hardware has apparently reached at least one board partner, but no launch date exists, and the next realistic announcement window would be Computex 2026 — with low confidence even then.
How much VRAM does the RTX 5070 Super have?
Reportedly 18GB of GDDR7, up from the standard RTX 5070's 12GB, achieved by swapping in 3GB memory modules instead of 2GB ones on the same 192-bit bus.
Why is Nvidia delaying the RTX 50 Super refresh?
GDDR7 memory costs have reportedly risen sharply as AI datacenter demand eats into consumer memory supply — the same dynamic that just pushed Chinese distributor prices on existing RTX 50-series cards up by as much as 59%.
Will the RTX 5070 Super still cost $549?
That was the rumored price, matching the standard 5070's launch MSRP, but it was rumored before the current memory cost spike. If it ships at all, a higher price is more likely than the original number holding.
Should I wait for the RTX 5070 Super instead of buying now?
Only if you have no immediate need. There's no confirmed release date, and everything about how this has played out suggests Nvidia sees it as optional.

I'll watch Computex for movement, but I'm not holding my breath, and I wouldn't plan a purchase around it either. If the regular RTX 5070 is sitting anywhere close to $549 where you shop, that's the card worth buying — not a maybe.

Hardware Editor

Ravi Malhotra

Ravi has been building and taking apart PCs since the single-core days — his idea of a good weekend is a repaste and a spreadsheet full of thermals. He covers GPUs, CPUs and the build decisions that actually move frame rates, and he'd rather hand you a benchmark than a press release.

The Aliteq brief

The tech worth knowing — hardware, AI, gaming, deals. No spam, unsubscribe anytime.

Keep reading