AMD and Nvidia have been going back and forth as the main manufacturers of consumer graphics cards for years. Although Intel have added to the race with their Intel Arc series they are still a long way behind. The Nvidia GeForce RTX 4000 seires continue to dominate the market and provide the best and most powerful graphics cards available all while being the most expensive, and for good reason. Not only do Ada Lovelace Graphics Cards provide top quality ray-tracng capabilities to the table, but you also get to experience smooth frame rates at high resolutions thanks to DLSS 3.5 and frame generation.
The upcoming RTX 5000 series from Team Green is rumored to surpass the current generation. Codenamed "Blackwell" after the renowned mathematician David Blackwell, the new Nvidia graphics cards are speculated to feature high-bandwidth GDDR7 memory, PCIe 5.0 support, and various other enhancements. This article will provide a comprehensive recap of the latest rumors and leaks surrounding the GeForce RTX 5000 family.
Nvidia GeForce RTX 5000 series: When we're hoping to see it and how much we want to pay
Nvidia typically follows a 2-year product cycle for its consumer graphics cards, with a mid-cycle hardware refresh around the 1-year mark. Given that Nvidia launched the Super variants of its current RTX 4000 GPUs in January 2024, the company is expected to unveil the next-generation Blackwell series between late 2024 and early 2025.
Earlier, the German tech website HardwareLUXX reported that the "Ada Lovelace-Next" graphics cards may not arrive until 2025. However, YouTuber Moore's Law Is Dead claimed the newer cards could launch as early as Q4 2024. According to the leaker who tipped them off, the release window for the new series depends on the sales of the RTX 4000 lineup and competition from AMD's upcoming RDNA 4 GPUs. In a recent online post, leaker kopite7kimi speculated that the new GPUs may be released "at the end of the year."
The pricing for the new graphics cards remains uncertain. The RTX series is notorious for being overpriced, and this trend is expected to continue with the upcoming Blackwell family. Assuming the flagship model is an RTX 5090 rather than a 5090 Ti, it will likely carry a launch price similar to or slightly higher than the current GeForce RTX 4090, which retails for $1,599.
Nvidia GeForce RTX 5000 series: What we could see
While AMD has embraced a Multi-Chip Module (MCM) approach for its graphics cards, Nvidia has maintained the use of monolithic dies. A monolithic chip integrates all circuitry onto a single die, limiting the ability to shrink transistors without increasing the overall chip size. In contrast, the MCM approach utilizes separate chiplets connected via interposers on a single substrate. This modular design enhances the scalability of MCM-based chips, enabling manufacturers to pack in more powerful specifications.
According to leaker kopite7kimi, Nvidia's upcoming data center chips, the GB100 and GB102, will utilize a multi-chip module (MCM) design rather than a traditional monolithic architecture. Similarly, the high-end GB202 chip is expected to bring this MCM approach to consumer GPUs. However, Nvidia is likely to retain the monolithic design for its budget and mid-range GPUs, including the GB203, GB205, GB206, and GB207 models.
Higher bandwidth GDDR7 memory
Many GPU enthusiasts were disappointed that Nvidia's Ada Lovelace GPUs utilized GDDR6X and GDDR6 memory from previous generations rather than the latest memory technology. However, rumors suggest the flagship RTX 5000 series will feature the cutting-edge GDDR7 memory.
For comparison, the GeForce RTX 4090 had a memory bandwidth of 1,008 GB/s, matching the last-gen RTX 3090 Ti. In contrast, the rumored memory bandwidth of the RTX 5090 is an impressive 1,536 GB/s - a substantial 52% upgrade over the previous generation.
The details around the GeForce 5090's memory bus have been the subject of much speculation. Initially, prominent leaker kopite7kimi claimed the flagship "Blackwell" GPU would have a 512-bit memory interface. However, he later backtracked, stating the RTX 5090 would instead utilize the same 384-bit bus as the RTX 4090 and 3090 Ti. Then, in March 2024, kopite7kimi reversed course again, asserting the top-tier Blackwell GPU would in fact feature a 512-bit memory bus.
The leaker revealed that the GB203 and GB205 models will not have the same memory interface as the flagship GPU. Instead, they will be capped at 256-bit and 128-bit memory buses, respectively.
Aside from rumors suggesting the RTX 5090 may feature 36GB of GDDR7 memory, there have been few updates on the VRAM capacity of the upcoming GPUs. However, the flagship model is expected to have at least 24GB of VRAM, similar to the current RTX 4090.
As for the other high-end variant, kopite7kimi previously stated the GB203 card may have half the memory of the GB202 model. Assuming it's not a multi-chip design, this could mean a theoretical RTX 5080 may come with 12 to 18GB of VRAM.
New TSMC 4N manufacturing process (with a higher core count)
Nvidia has consistently leveraged newer manufacturing processes across its recent GPU generations, and the upcoming Blackwell series appears poised to follow suit. Rumors suggest that Team Green's next-gen GPUs will be built on TSMC's advanced 4N process node, building upon the TSMC 5 fabrication technology. Prominent leaker kopite7kimi has also confirmed that the flagship GB202 chip will utilize the same manufacturing process as the GB100 chip designed for data center GPUs. However, the leaker notes that the GB202 SKU is expected to boast a 30% increase in chip density.
In a previous tweet, kopite7kimi revealed that the new GPUs are unlikely to see a significant increase in the number of Graphics Processing Clusters (GPCs) and Texture Processor Clusters (TPCs). Additionally, leaker panzerlied claimed that the GeForce RTX 5090 could see a 15% frequency boost, 78% more cache, and a 50% improvement in scale, potentially translating to a higher 24,576 CUDA core count.
Support for PCIe 5.0 and DisplayPort 2.1
The Ada Lovelace GPUs are compatible with PCIe 5.0 motherboards, but they cannot fully leverage the capabilities of this high-speed interface standard. In contrast, the upcoming Blackwell GPUs are expected to be able to take advantage of the lightning-fast data transfer speeds offered by PCI Express Gen 5.0.
Similarly, the rumored GeForce RTX 5000 graphics cards are anticipated to offer significant improvements in display capabilities compared to their predecessors. According to kopite7kimi, the new GPUs will support the DisplayPort 2.1 standard, which provides much higher bandwidth, refresh rates, and resolutions than the previous DisplayPort 1.4 specification.
Nvidia GeForce RTX 5000 series rumors: To believe or not to believe
Despite the plausibility of some recent leaks, it's prudent to view them with cautious skepticism. For instance, the rumored 800W power draw of the flagship RTX 4090 from Nvidia's Ada Lovelace lineup sparked significant backlash, only to be quelled when the actual 450W TDP was revealed at launch.
While there are compelling reasons to expect Nvidia's Blackwell GPU family this year, the company has remained tight-lipped so far. Nevertheless, more details are likely to emerge in the near future. In the meantime, the existing RTX 40-series GPUs are still worth considering, with the GeForce RTX 4070 standing out as a compelling option that delivers solid performance without breaking the bank.
All Image and Content credit - https://www.xda-developers.com/nvidia-rtx-5000-rumors
