NVIDIA has officially unveiled their RTX 50 series GPUs at CES 2025, with the lineup focusing heavily on AI capabilities and performance improvements. The announcement showcases significant upgrades across the board, particularly in AI processing power and memory configurations.

Key features of the 50 series include:

• New Blackwell architecture
• DLSS 4 technology with improved upscaling and reduced artifacts
• GDDR7 memory across all models, offering up to 40% faster memory speeds than GDDR6X
• Ninth-generation video encoders for enhanced streaming capabilities
• Updated 16-pin power connector with angled design to prevent cable issues

Performance comparisons to 40 series according to NVIDIA:

• RTX 5090: 2x faster than 4090 in standard rasterization, 2.5x faster in ray tracing
• RTX 5080: Matches or exceeds 4090 performance at half the power consumption
• RTX 5070 Ti: 35% faster than 4080
• RTX 5070: Roughly equivalent to 4090 when using DLSS 4

The initial lineup and pricing is as follows from NVIDIA's Australian website:

• RTX 5090: $4,039 AUD (American RRP: $1,999 USD, 32GB GDDR7)
• RTX 5080: $2,019 AUD (American RRP: $999 USD, 16GB GDDR7)
• RTX 5070 Ti: $1,509 AUD (American RRP: $749 USD, 16GB GDDR7)
• RTX 5070: $1,109 AUD (American RRP: $549 USD, 12GB GDDR7)

Desktop cards will be available starting January 2025, with laptop variants featuring these GPUs expected to launch in March 2025. The laptop models will incorporate Max-Q technology for improved power efficiency.

While the performance claims are impressive, particularly the RTX 5070 supposedly matching the RTX 4090's performance, these figures likely depend heavily on DLSS 4 and other AI features. Independent benchmarks will be crucial for verifying these claims. Early testing suggests the most significant gains are seen in AI-accelerated workloads and games utilizing DLSS 4 technology.