Next year, industry giants like AMD, Intel, Nvidia, and others are set to release a wave of new chips. These chips, including Nvidia's Blackwell architecture, promise drastic improvements over their predecessors, marking a significant leap forward.
A First Look at Nvidia’s RTX 50 Blackwell Series
Today, we offer an early glimpse at the upcoming ‘Next-Gen GeForce RTX 50’ Blackwell laptop lineup, which will feature GDDR7 memory. It's crucial to note that the information in this article stems from a leak following a ransomware attack on CLEVO—a Taiwanese OEM/ODM laptop manufacturer. As a result, the details are unofficial and should be taken with a grain of caution. Despite the uncertainty, we’ll explore the main aspects, potential specifications, and other technological advancements expected from these GPUs.
The Leaked Information: Margin of Error and Chip Tags
Given the nature of the leak, the margin of error is considerable, with variations across available sources. However, a strong foundation of evidence suggests we’ll see RTX 5090, RTX 5080, RTX 5070, and RTX 5060 models, potentially accompanied by Ti and/or Super variants.
Leaked Chip Tags:
GN22-X11 (16 GB GDDR7)
GN22-X9 (16 GB GDDR7)
GN22-X7 (12 GB GDDR7)
GN22-X6 (8 GB GDDR7)
GN22-X4 (8 GB GDDR7)
GN22-X2 (8 GB GDDR7)
Specifications:
Tentative Blackwell GPU Specifications
| Features | RTX 5050 | RTX 5060 | RTX 5070 | RTX 5080 | RTX 5090 alt | RTX 5090 |
| Architecture | GB207 | GB206 | GB205 | GB203 | GB202 | GB202 (2x GB203?) |
| Process Technology | TSMC 4NP? | TSMC 4NP? | TSMC 4NP? | TSMC 4NP? | TSMC 4NP? | TSMC 4NP? |
| Transistors (Billion) | NA | NA | NA | NA | NA | NA |
| Die size (mm^2) | NA | NA | NA | NA | NA | Full reticle? |
| SMs | 20? | 36? | 50? | 84? | 160? | 192? |
| CUDA Cores (Shaders) | 2560? | 4608? | 6400? | 10752? | 20480? | 24576? |
| Tensor Cores | 128? | 144? | 200? | 336? | 640? | 768? |
| RT Cores | 32? | 36? | 50? | 84? | 160? | 192? |
| Boost Clock (MHz) | 2500? | 2500? | 2500? | 2500? | 2500? | 2500? |
| VRAM Speed (Gbps) | 20? | 36? | 36? | 36? | 36? | 36? |
| VRAM (GB) | 8? | 12? | 18? | 16? | 24? | 32? |
| VRAM Bus Width | 128? | 128? | 192? | 256? | 384? | 512? |
| L2 / Infinity Cache (MB) | 24? | 32? | 48? | 64? | 128? | 128? |
| Render Output Units | 32? | 48? | 80? | 112? | 192? | 192? |
| Texture Mapping Units | 128? | 144? | 200? | 336? | 640? | 768? |
| TFLOPS FP32 (Boost) | 12.8? | 23.0? | 32.0? | 53.8? | 102.4? | 122.9? |
| TFLOPS FP16 (FP8) | 102? (205?) | 184? (369?) | 256? (512?) | 430? (860?) | 819? (1638?) | 983? (1966?) |
| Bandwidth (GBps) | 320? | 576? | 864? | 1152? | 1728? | 2304? |
| TDP (watts) | 125? | 175? | 225? | 320? | 450? | 450? |
| Launch Price | $299~$399? | $449~$549? | $599~$799? | $999~$1,199? | $1,599? | $1,999+? |
What We Know About the Blackwell Architecture
Blackwell is an evolution of the Ada Lovelace design, introducing significant upgrades over the previous generation. While specific details remain scarce, we can anticipate improvements in ray tracing, continuing Nvidia’s trend of advancing this technology. The exact nature of these enhancements is still unknown, but they’re expected to be as impactful as previous innovations.
Tensor Cores, AI Performance, and TSMC 4NP Process
The Blackwell B200 GPU is expected to feature updated tensor cores capable of handling new FP4 and FP6 formats, optimizing the GPU for AI tasks. This likely means that consumer GPUs will also benefit from these advancements. If Nvidia continues to use the TSMC 4NP process, major changes may be limited, though high-end models could see the introduction of exciting updates like NV-HBI.
Release Timeline and Pricing Speculation
According to sources, these laptop GPUs are expected to launch at CES 2025, powering the latest gaming laptops alongside next-generation CPUs like AMD’s Fire Range and Intel’s Arrow Lake. More details about these laptop GPUs will likely emerge following the release of the desktop versions, anticipated in late 2024 or early 2025.
There is speculation that Nvidia might price these new GPUs slightly lower than the RTX 40-series at launch. Given the potential for higher AI performance and increased demand from non-gaming sectors, the next generation could launch at similar or even lower prices compared to the current generation. This pricing strategy could respond to the stagnant growth in gaming demand over the past few years.
Conclusion
For a more in-depth analysis, check out our full article on Blackwell. As we get closer to the official release, more details will likely surface, providing a clearer picture of what to expect from Nvidia’s next-gen GPUs.