NVIDIA's Next-Gen Flagship: RTX 50 Series
NVIDIA is set to unveil its latest flagship graphics card series, the RTX 50 line, at CES early next year. The series will encompass both desktop and mobile variants, including the high-end RTX 5090 and 5080 models.
RTX 5090: A Powerhouse Upgrade
The RTX 5090 will be the pinnacle of the series, featuring a new PCB design and the Blackwell GB202-300-A1 GPU core. Out of its 192 SMs, 170 will be active, resulting in an impressive 21760 CUDA cores.
Clock speeds have received a significant boost, with the RTX 5090 expected to run at 2900 MHz, a 30% increase over the RTX 4090.
In terms of memory, the RTX 5090 will sport 32GB of GDDR7 with a blistering 28Gbps speed. The 512-bit memory bus width translates to a colossal bandwidth of 1792 GB/s.
However, this performance comes at a cost. The RTX 5090's TBP (typical board power) is expected to reach 600W, a significant 150W jump from its predecessor.
Addressing the concerns raised by the RTX 4090's power connector issues, NVIDIA has opted for a new 12V-2x6 16-pin interface for the RTX 5090.
Pricing Expectations
Given the performance enhancements and GDDR7 memory, the RTX 5090's price is likely to increase substantially. While the RTX 4090 had a suggested retail price of $1,599, the RTX 5090 is projected to be priced between $1,800 and $2,000, or approximately 1.3-1.4 million yuan in Chinese currency.
The table below compares the key specifications of the RTX 5090 and 4090:
| RTX 5090 | RTX 4090 | |
|---|---|---|
| CUDA Cores | 21760 | 16384 |
| Boost Clock | 2900 MHz | 2520 MHz |
| Memory | 32GB GDDR7 | 24GB GDDR6X |
| Memory Bandwidth | 1792 GB/s | 1008 GB/s |
| TBP | 600W | 450W |
（举报）