NVIDIA GeForce RTX 5090: Leaked Specs and What to Expect

The tech world, especially the PC gaming community, has been buzzing with excitement over leaked details about NVIDIA's next-generation flagship GPU, the GeForce RTX 5090. Fresh leaks suggest significant hardware upgrades that could once again set benchmarks in the realm of gaming and graphical performance. If you're a gaming enthusiast, tech junkie, or content creation powerhouse, buckle up, because this GPU might just be what you're dreaming of.

The Heart of the Beast: GB202-300 GPU​

NVIDIA is preparing to debut its latest GB202-300 GPU, a part of the much-anticipated Blackwell architecture family. Known for pushing the limits of graphical processing power, Blackwell is poised to be the successor of NVIDIA's Ada Lovelace architecture utilized in the RTX 40 series. Early information indicates that this new GPU package will be no ordinary upgrade—it’s a massive leap.
Here are the juicy details so far:

• CUDA Cores: A jaw-dropping production of 21,760 CUDA cores. This significant increase marks a leap from the RTX 4090's 16,384 cores, guaranteeing a gaming and computational performance boost.
• Die Size: With a die size of approximately 24x21 mm, production is reportedly on a cutting-edge 5nm fabrication process. Smaller die size often implies better power efficiency, but in this case, it'll also house the beefed-up CUDA cores, enabling unparalleled parallel processing power.

For context: CUDA cores are the processing units within GPUs that handle calculations for rendering graphics, AI computations, and gaming physics simulations. A higher count allows the card to handle more visual or mathematical elements simultaneously, which translates to smoother gameplay and quicker workloads.

GDDR7 Memory – The Future of Graphics Memory​

The RTX 5090 isn’t just getting a GPU upgrade; it’s packing GDDR7 memory, a technology that doubles down on performance compared to GDDR6X found in the RTX 4090. The leaked configuration shows:

• 32GB of GDDR7 Memory: This makes the GPU future-proof for years to come, handling 4K, 8K gaming, and all the AI or machine learning frameworks one could expect.
• Speeds of 28 Gbps: This makes data transmission within the memory significantly faster, enhancing load times, in-game textures, and rendering speeds.
• Memory Bandwidth of 1,792 GB/s: To put this in perspective, the RTX 4090’s bandwidth tops out at around 1,008 GB/s. The jump demonstrates what seamless performance may look like when processing massive workloads.

Memory speed and bandwidth are critical for modern GPUs. They dictate the velocity at which the card can load textures or perform graphical calculations. With bandwidth closing in on 2TB/s, RTX 5090 ensures better framerates, reduced stuttering, and lightning-fast creative workloads like 3D rendering.

A Board Built to Dominate​

Based on the leaked PCB images and specifications from trusted sources, the GeForce RTX 5090 promises to be an absolute juggernaut of graphical performance. Here's what stands out:

• Massive GPU Package: The card boasts a physical package size of 63x56mm, making it colossal by GPU standards. Such a large area is reserved for housing robust components, likely including additional VRMs (Voltage Regulation Modules) to manage power delivery.
• Power Connectors: It uses a 16-pin PCIe Gen 5.0 power connector, ensuring it can sip up to 600W of power. Yes, this means you’ll want a sturdy PSU upgrade if you’re planning to adopt this beast.
• Advanced Display Outputs: 3x DisplayPort 2.1a and 1x HDMI 2.1 are built-in. The DP 2.1a compatibility ensures future-proofing for ultra-high refresh rates at resolutions beyond 8K.

The staggering power requirements are par for the course for cutting-edge hardware, but NVIDIA is continuing its push for energy efficiency. Though the RTX 5090 requires significant wattage, its generational architectural upgrades hint at better performance-per-watt ratios.

What Makes NVIDIA Blackwell Architecture Special?​

The Blackwell generation is expected to build on advancements in:

• Efficiency Gains: Aiming at both performance and sustainability by shrinking transistor sizes and optimizing core designs.
• Improved Ray Tracing: Blackwell might introduce better RT cores (responsible for real-time ray tracing), pushing lifelike reflections, shadows, and light scattering even further.
• AI-Powered Processes: Tensor cores could see revamped functionality, empowering AI-assisted graphics and computational workloads like DLSS 4.0.

DLSS 4.0 (Deep Learning Super Sampling)—if introduced—could make high-resolution, ultra-smooth gaming accessible even on lower wattage settings by upscaling graphics using machine learning.

---

Can the World Handle the Power?​

Of course, sticking this kind of power into an air-cooled GPU comes with challenges:

• Thermals Could Be Tricky: The monstrous performance may mean that NVIDIA needs equally impressive cooling designs. Expect triple or even quad-slot coolers with enhanced thermal dissipation techniques.
• Compatibility: At 600W and this physical size, you might need a case upgrade. Space and airflow will inevitably become critical.
• Price: Just like the RTX 4090 shook wallets, the RTX 5090 will likely fall in the same premium-tier price range. If you’re building a PC around this card, start saving now.

It’s also important to note that power-hungry rigs like this often carry supply chain risks. Manufacturing the GPU components and specialized GDDR7 memory might mean limited availability upon launch.

---

Why This Matters to Windows Users​

For Windows users, hardware like NVIDIA’s upcoming RTX 5090 ties into key workflows:

• Gaming: This GPU will redefine the boundaries, enabling settings cranked to ultra in resolutions as high as 8K.
• Productivity: Those in design, 3D modeling, or video editing ecosystems like Blender, Adobe Premiere, and Unreal Engine will see those time-intensive renders cut down drastically with this level of processing power.
• Windows 11 Optimization: Features like DirectStorage, an API in Windows 11 designed to reduce load times using NVMe drives and GPU decompression, will be fully utilized with such robust hardware.

---

Key Takeaways​

The NVIDIA GeForce RTX 5090 stands to be a game-changer in more ways than one. With the upgraded GB202-300 GPU, cutting-edge GDDR7 memory, jaw-dropping bandwidth of 1,792 GB/s, and an early 2025 unveiling at CES, it embodies NVIDIA's commitment to pushing technological boundaries.
Whether you’re gaming at the highest fidelity, using your PC to crunch creative workloads, or simply want bragging rights for owning the world’s most powerful GPU, the RTX 5090 seems primed for the title. Just make sure you’ve got the power supply, space, and funds to back it up.

---

What are your thoughts on this upcoming powerhouse? Are you upgrading for the RTX 5090, or is it simply too much for your workflow? Let us know in the comments below on WindowsForum.com—we want to hear all those GPU and gaming dreams!

---

Source: VideoCardz.com NVIDIA GB202-300 GPU and GDDR7 memory for GeForce RTX 5090 leaked - VideoCardz.com

 

[ChatGPT]

Hold onto your graphics cards, gamers and PC builders—2024 has dropped one final bombshell before the year fades out. It looks like the hardware scene is gearing up for a massive leap forward, as images of the alleged PCB (printed circuit board) for NVIDIA’s upcoming GeForce RTX 5090 have surfaced online. If these leaks hold any thread of truth—and that’s always a big “if”—the RTX 5090 could mark new heights in GPU design. Here’s everything you need to know about the leaked specs, the tantalizing possibilities for performance, and why this GPU might change how we think about gaming and computing in 2025.

---

What We Learned from the PCB Leak​

The leak reportedly originates from NVIDIA's AIB (add-in board) partner, possibly PNY, and showcases the PCB design for the yet-to-be-released RTX 5090. For a bit of context, AIB partners like MSI, ASUS, and others take the base technology of NVIDIA’s GPUs and often develop their customized editions of the cards, tweaking the thermal performance, clock speeds, and sometimes even aesthetics.
Now, onto the juicy details. At a glance, the leaked PCB suggests this GPU will boast a massive GPU package featuring a 63 x 56mm GB202 die. In simpler terms, that’s a wildly oversized chip designed for extreme levels of muscle-flexing in graphics performance. But that’s not all:

• Memory Configuration:
• The PCB layout hints at 16 memory module slots, arranged in what tech enthusiasts call an “Android” pattern: five memory blocks on each side, four at the top, and two at the bottom. Depending on the module sizes (likely 2GB, 3GB, or 4GB per module), this could translate into memory capacities of up to 64GB of GDDR7 RAM.
• Yes, you read that right—GDDR7. That’s next-gen VRAM technology that promises faster memory bandwidth, lower power consumption, and more impressive gaming and computing performance than GDDR6 or GDDR6X.
• Power Supply:
• The RTX 5090 seems to require just one power connector, specifically a 12V-6×2 pin. This could indicate a push for efficiency despite the card’s powerhouse specs. For context, NVIDIA’s recent 4000-series GPUs also made strides in reducing power requirements compared to predecessors like the RTX 3090.

---

Why GDDR7 Matters for RTX 5090​

The upgrade to GDDR7 memory could be one of NVIDIA’s game-changers in the RTX 50-series. GDDR7 is set to beat its predecessors with double the performance per watt. It accomplishes this with PAM3 (Pulse Amplitude Modulation), essentially allowing it to transmit more data in the same amount of time compared to the widely used PAM2 technology.
If gaming GPUs are like Olympic sprinters, GDDR7 is the new running shoe that shaves a full second off the 100-meter record. Expect blazing-fast frame rates, buttery-smooth 4K gameplay, and even realistic potential to handle 8K gaming—assuming, of course, your monitor can keep up.
On the productivity side, GDDR7 also holds possibilities for creative professionals. Rendering 3D models, editing high-definition video, or running machine learning workloads could see major speed bosts from the wider memory bandwidth.

---

Is This the Death Knell of Power-Hungry GPU Designs?​

This is where things get even more intriguing. Despite its high-end design, the RTX 5090 PCB suggests NVIDIA is sticking to just one power connector. Compare that to earlier 3000-series cards, which demanded higher wattage and heftier PSUs. While the RTX 4000-series began curbing that power-hungry trend, retaining efficiency at higher performance seems to be a major focus for the RTX 50-series.
One potential culprit for this focus on efficiency is competition—not just from AMD, but from the shift in energy standards worldwide and a growing push toward eco-conscious computing. The use of a single 12V-6×2 connector might make the card friendlier to those running high-end rigs without having to upgrade their power supply units (PSUs).

---

When Can You Get Your Hands on It?​

Rumors around the RTX 5090 point to an official NVIDIA unveiling during CES 2025 in January. However, don’t expect to snag one immediately. If industry patterns hold true, the RTX 5080 will likely land first, with the heavyweight RTX 5090 arriving later—potentially in late January or mid-February 2025.
Notably, this staggered rollout aligns with NVIDIA’s recent trend of launching slightly more affordable flagship cards like the 4080 or 3080 before delivering the absolute showstoppers like the 4090 or Titan variants.

---

Broader Implications for Gamers and the Industry​

So what does this mean for average gamers or PC enthusiasts? Here’s the lowdown:

• A Price to Pay:
• High-end NVIDIA GPUs are never kind to our wallets. Given how the RTX 4090 hit PC gaming enthusiasts with price tags north of $1,599, it’s reasonable to expect the RTX 5090 to cross the $2,000 mark for the deluxe editions. If you’re budgeting for a new GPU next year, start saving now.
• Smaller, Less Power-Hungry Builds:
• Smaller PCs are already becoming a trend, with compact ITX cases and more efficient cooling solutions gaining traction. The RTX 5090’s possible single power connector could inspire next-gen designs that make monstrous gaming power possible in compact builds.
• Futureproof Gaming:
• Support for technologies like DLSS and ray tracing will see a boost, and upcoming titles in 2025 and beyond (think Unreal Engine 5 games) could push RTX 50-series GPUs to their fullest potential.

---

Final Thoughts: Revolution or Evolution?​

Is the RTX 5090 shaping up to be a revolution in GPU design, or just a steady iteration on what we’ve seen in the 4000-series? From this leak, it seems like NVIDIA is opting for both—combining forward-looking tech like GDDR7 memory with practical changes like more efficient power use. But as with any rumors, we should take all the leaked information with a grain of salt. While promising, these specs might not reflect the final product.
If NVIDIA’s CES 2025 reveal goes as expected, though, the RTX 50-series will undoubtedly set new benchmarks for gaming and professional performance. Buckle up—2025 might just kick off the next golden era for PC gaming.
What do you think? Are you intrigued by the leaked design? Or do you think this is just another chapter in GPU overpricing while offering incremental changes? Jump into the discussion on our forums and share your take!

---

Source: Lowyat.NET Alleged NVIDIA GeForce RTX 5090 PCB With Large GPU Package Leaks