No, the 5070 is NOT a 4090 - NVIDIA RTX 5090, 5080, 5070 Ti CES Keynote Recap

Summary

At CES 2025, Nvidia unveiled its new RTX 50 Series GPUs, featuring four models with varying prices and performance enhancements. Key highlights include the introduction of DLSS 4.0 for smoother gameplay, improved cooling solutions, and the Blackwell architecture powering the GPUs. While Nvidia claims impressive performance boosts, especially with DLSS 4.0, independent reviews will be crucial to validate these claims. The GPUs, ranging from the $550 RTX 570 to the $2,000 RTX 5090, signal Nvidia’s continued push into both gaming and AI, though availability and pricing remain potential concerns. Overall, the new lineup promises significant innovation in gaming graphics and AI applications.

Takeaways

• 😀 Nvidia introduced the RTX 50 Series GPUs at CES 2025, with four new models: RTX 5090, 580, 5070 TI, and 570.
• 💸 The RTX 5090 is priced at $2,000, a $1,200 increase from its predecessor, the 490, while the RTX 580 will be priced at $1,000.
• 🔍 Nvidia's RTX 5070 TI is priced at $750 and claims to deliver performance on par with the RTX 4090, though this comparison is under scrutiny.
• 🖥 The RTX 570 is priced at $550 and is expected to outperform the RTX 4070 by around 25-30% in some tests.
• 🤖 DLSS 4.0, a major new feature, enhances performance with up to three additional frames per real frame, but this doesn't add real information from the game engine.
• 🌐 Nvidia's Blackwell architecture powers the RTX 50 Series, featuring 92 billion transistors and 4 petaflops of AI performance.
• 💨 New Vapor Chamber cooler design for the RTX 50 Series GPUs aims for improved cooling efficiency, possibly surpassing the performance of current Founders Edition models.
• 📊 Nvidia revealed performance comparison numbers showing the RTX 5090 outperforms the 4090 by 20-25%, but these numbers should be taken with caution.
• 🏢 Nvidia's AI hardware is also being marketed for data centers, with a new EnVLink 72 solution offering 130 terabytes per second of bandwidth.
• 💻 Project Digits is Nvidia's effort to shrink its DGX supercomputers into a desktop form factor for AI workloads, featuring a new GPU/CPU combo chip and ample memory.
• 🗓 The RTX 50 Series will launch in January 2025, though there is uncertainty about which model will launch first, but it's expected to be the RTX 580.

Q & A

What are the four GPUs announced in the RTX 50 Series at CES 2025?

The four GPUs announced in the RTX 50 Series are the RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070. These GPUs target different price points, with the RTX 5090 priced at $2,000, the RTX 5080 at $1,000, the RTX 5070 Ti at $750, and the RTX 5070 at $550.

How does the pricing of the RTX 50 Series compare to previous generations?

The pricing of the RTX 50 Series seems relatively reasonable, with the RTX 5090 costing $2,000, which is less than the expected $2,500 price. The RTX 5080 is priced lower than the original RTX 4080, and the RTX 5070 Ti and 5070 are also priced lower than their direct predecessors.

What performance claim did NVIDIA make regarding the RTX 5070?

NVIDIA claimed that the RTX 5070 performs at the same level as the RTX 4090, particularly with DLSS 4.0 features enabled. However, this claim should be taken with caution as DLSS 4.0 involves frame generation, which may not represent a complete performance comparison.

What is DLSS 4.0, and how does it affect performance?

DLSS 4.0 is a new AI-powered technology introduced with the RTX 50 Series that generates additional frames to improve gaming smoothness. It can produce up to three additional frames per rendered frame, but it does not provide additional information from the game engine, which could limit its impact compared to traditional rendering methods.

How does the performance of the RTX 5070 compare to the RTX 4070 Ti based on NVIDIA's benchmarks?

According to NVIDIA’s performance comparison, the RTX 5070 outperforms the RTX 4070 Ti by approximately 25-30%, though the data presented lacks clear axis markers, so exact comparisons should be approached cautiously.

What new cooling technology did NVIDIA introduce for the RTX 50 Series?

NVIDIA introduced a new **Vapor Chamber cooler** for the RTX 50 Series, which features two pass-through fans. This design is meant to provide better cooling by reducing the size of the PCB and improving airflow, which could lead to better thermal management for high-performance GPUs.

What is the Blackwell architecture, and why is it significant?

The Blackwell architecture is the underlying technology powering the RTX 50 Series. It features 92 billion transistors and boasts impressive AI capabilities with 4 petaflops of AI performance. Blackwell is designed to support advancements like DLSS 4.0 and real-time ray tracing.

What is the **Grace Blackwell NVLink 72** solution that NVIDIA unveiled?

The **Grace Blackwell NVLink 72** is an AI-driven supercomputing solution demonstrated at CES 2025. It consists of 72 Blackwell GPUs, 14 terabytes of memory, and 259 Grace CPU cores, providing 130 terabits per second of bandwidth. This system is designed for use in data centers and AI applications.

What is Project Digits, and what does it offer?

Project Digits is a miniaturized version of NVIDIA's DGX supercomputer. This smaller, desk-friendly device integrates a **GB10 Super Chip** that combines Blackwell GPUs with Grace CPU cores. It is designed for AI tasks and can handle cloud computing needs, making it a powerful tool for personal and small-scale AI workloads.

How does NVIDIA's focus on AI and robotics fit into the overall CES 2025 keynote?

NVIDIA's CES 2025 keynote emphasized the company’s expanded role in AI, robotics, and industrial applications. The keynote covered topics such as digital twins for factories, AI-driven automotive applications, and their work with robotics, signaling a broader vision beyond gaming hardware. NVIDIA also teased future developments in AI-powered robotics, including a potential 'ChatGPT moment' for robotics.

What concerns were raised regarding the availability of the RTX 50 Series GPUs?

While the RTX 50 Series is expected to launch in January 2025, there are concerns about the availability of these GPUs. Since NVIDIA initially planned to release these cards a few months earlier, there could be potential supply challenges or delays in meeting demand.