Nvidia RTX 60-Series Specs Could Follow a Familiar Pattern

Nvidia’s next gaming GPUs don’t have official RTX 60-series specs yet, but this PC Gamer analysis argues the company’s past cards point to a clear direction. Nvidia has spent the last decade shifting from the GTX 10-series to ray tracing and AI-heavy designs, and that matters because the next generation may lean on the same habits: similar die sizes, more transistors, and only modest gains in cache. For players, that could mean faster cards without a dramatic jump in power draw or board size.

The article looks at where Nvidia’s future GeForce cards could land, with the RTX 60-series expected “in the near future” and likely built around TSMC’s N3 process rather than the more advanced N2 node. That choice would shape everything from transistor density to shader counts, and it also explains why the author thinks Nvidia may keep L2 cache close to Blackwell levels. If you care about upgrade value, this is the part that decides whether the next round of cards feels like a real leap or just another careful step.

About Nvidia’s Next GeForce Generation

According to the source, Nvidia’s gaming GPUs have changed fundamentally since the GeForce GTX 10-series, with ray tracing and AI now central to the company’s hardware strategy. The article says the current top tier is the 90-class, though Nvidia used the Titan name for that slot in the GTX 10-series and RTX 20-series eras. That context matters because the piece compares four tiers of cards: 60-class, 70-class, 80-class, and the highest-end model.

The analysis also points out that Nvidia does not manufacture its own GPUs. Instead, it relies on TSMC, and the process node choice shapes the size and density of each chip. In practical terms, that means Nvidia’s next consumer cards may inherit the same design philosophy as Blackwell: efficient use of silicon, careful cost control, and as much performance as the company can squeeze out without making the chips too large.

Die Size, Density, And What That Means For RTX 60-Series Cards

The source says Blackwell gaming chips use a custom version of TSMC N5 called 4N, while the RTX 40-series also used that node and the RTX 30-series used Samsung’s 8LPH. Before that, Nvidia’s GTX 10-series and RTX 20-series used a custom TSMC N16 node, and three prior GTX generations used N28. The author believes Nvidia will likely stick with TSMC N3 for RTX 60-series chips instead of jumping to N2, mainly for cost reasons.

That process choice would raise die density sharply. The article says TSMC’s N3 sits around 200 million transistors per square millimetre, while Blackwell and Ada Lovelace sit at about 120 million transistors per square millimetre on 4N. In plain terms, that gives Nvidia room to pack in more logic, but not every part of the chip scales the same way, so the gains won’t translate cleanly into a 66% jump in every feature.

Nvidia’s own design habits matter here too. The piece says the company has steadily favoured small dies for most gaming products because that improves wafer yields and profit margins, while the RTX 5090 sits near the upper limit of what TSMC can make. The author expects RTX 60-series parts to keep similar die sizes to Blackwell, which would mean roughly 60% to 70% more transistors than Blackwell without a major size increase.

• TSMC N3 could give RTX 60-series chips about 66% more density than Blackwell.
• Logic scales faster than cache and PCIe/VRAM circuitry, which only improves by around 5% at best.
• The RTX 5090 already sits close to the practical size ceiling for TSMC’s equipment.
• The author expects Nvidia to keep similar-sized dies for the RTX 60-series.

CUDA Cores, FP32 Throughput, And Cache

On shader counts, the article says the jump from RTX 20-series to RTX 30-series was dramatic, but 60-, 70-, and 80-class cards have not always grown much from one generation to the next. The author argues that raw CUDA core totals don’t tell the whole story because newer cores run faster and do more work per cycle. That means a future RTX 60-series card could feel much quicker even if the shader count doesn’t explode.

Peak FP32 throughput tells a more useful story, according to the source. The article says there was a noticeable gap between RTX 20-series and RTX 30-series cards, then another between RTX 30-series and RTX 40/50-series cards, driven by higher clocks and changes to the CUDA cores themselves. The author expects RTX 60-series cards to see a healthy increase in shader count and FP32 performance, but not on the same scale as die density growth.

Cache is the other big constraint. The source says RTX 40/50-series GPUs carry far more L2 cache than earlier chips, which helps reduce pressure on VRAM bandwidth and improves compute and ray tracing performance. Because SRAM scales poorly with process shrinks, the author expects Nvidia to keep L2 cache close to Blackwell levels rather than stuffing in a huge new cache block that would bloat the die.

• The author expects roughly a 30% to 50% increase in shader count versus Blackwell.
• RTX 40/50-series GPUs use far more L2 cache than Pascal, Turing, or Ampere.
• Large L2 cache helps both compute and ray tracing performance.
• RTX 60-series chips will probably not get a dramatic L2 cache increase.

Tensor Cores, Ray Tracing, And VRAM Limits

The article also turns to Tensor cores and ray tracing. It says Nvidia’s quoted AI TOPS figures can be misleading because each generation supports different data formats, and the raw numbers don’t compare cleanly across eras. Even so, the author believes RTX 60-series cards will probably gain Tensor performance mostly through higher core counts rather than better efficiency per cycle.

That matters for DLSS. The source notes that even the RTX 5060’s INT8 figure sits only 6% below an RTX 2080’s, so more Tensor cores should help DLSS performance even if the exact frame-generation format remains unclear. At the same time, the article warns that Nvidia can’t simply throw a huge pile of Tensor cores at the problem without running into register-file limits inside each SM.

Ray tracing should also improve. The source says each SM in every RTX GPU contains one RT core, and those units have changed significantly across generations. The author expects next-gen RT cores to become more capable without taking up much more space, which is the kind of upgrade players actually feel in supported games: better lighting and reflections without a giant trade-off elsewhere on the chip.

VRAM is the murkiest part of the prediction. The article says Nvidia may use 3 GB GDDR7 modules instead of 2 GB chips in some tiers to avoid another round of criticism over memory capacity, but it does not expect major changes to bus width. In the author’s forecast, the RTX 6060 would still use a 128-bit bus and the RTX 6070 a 192-bit bus, which keeps module counts down and helps margins, even if memory prices remain a headache for buyers.

What This Means For Players

This is a cautious forecast, and that makes it more believable than the usual spec-sheet fan fiction. The article’s central argument is that Nvidia will use the extra transistor budget to improve performance, but it won’t waste silicon on huge cache jumps or oversized dies when profit margins matter so much. For players, that suggests a generation built around steady gains rather than a headline-grabbing redesign.

That approach makes sense, especially with memory supply still under pressure. If Nvidia leans on 3 GB GDDR7 modules in some tiers, it could soften the blow from the RTX 5060 and 5060 Ti backlash, but the source also leaves open the ugly possibility that VRAM capacities won’t rise much at all. Nobody buying a new GPU wants to hear that, but the article’s logic is hard to dismiss.

What stands out most is how little room Nvidia has to keep cramming in more cache. The piece makes a strong case that L2 growth won’t match die-density growth, which means the company will likely chase performance through more shaders, better clocks, and incremental RT and Tensor gains instead. That’s a smart move if you want efficiency; it’s less exciting if you were hoping for a clean generational reset.

For now, the final unknown is clock speed, and the source cuts off before finishing that thought. Even so, the broad shape of the RTX 60-series is already clear enough: similar die sizes, higher density, modest shader growth, and memory decisions shaped as much by cost as by ambition. If Nvidia confirms these cards, the real question won’t be whether they’re faster. It’ll be whether they’re fast enough to justify the usual upgrade tax.