For years, AMD and Nvidia have been the primary contenders in the consumer-grade graphics card market. Although Intel’s Arc series has entered the competition, Nvidia’s GeForce RTX 4000 series maintains its dominance in our rankings for several reasons. The upcoming Ada Lovelace GPUs promise exceptional ray-tracing capabilities alongside smoother frame rates, courtesy of DLSS 3.5 and advanced frame generation techniques.
Moreover, Nvidia’s forthcoming RTX 5000 series, codenamed “Blackwell” after mathematician David Blackwell, aims to surpass the current generation. Anticipated upgrades include high-bandwidth GDDR7 memory, support for PCIe 5.0, and a suite of additional features. This article will provide an overview of the latest rumors and leaks surrounding the GeForce RTX 5000 family.
Anticipating the Arrival and Pricing of Nvidia’s GeForce RTX 5000 Series
Nvidia’s release patterns for consumer graphics cards typically follow a two-year cycle, with a refresh occurring around the one-year mark. Following the introduction of the Super variants of its RTX 4000 GPUs in January 2024, speculation abounds regarding the unveiling of the next-generation Blackwell series, expected to occur between late 2024 and early 2025.
Reports from sources like HardwareLUXX, a German tech website, suggested that the “Ada Lovelace-Next” graphics cards might not surface until 2025. However, conflicting information from sources like YouTuber Moore’s Law Is Dead suggested a potential release as early as Q4 2024. According to insiders, the release timeline for the new series may depend on the sales performance of the RTX 4000 lineup and competition from AMD’s forthcoming RDNA 4 series GPUs. Recently, leaker kopite7kimi hinted on X (previously called Twitter) that the new GPUs could potentially arrive “at the end of the year.”
As for pricing, the landscape remains uncertain. Nvidia’s RTX series has earned a reputation for high prices, and it’s likely that this trend will continue with the Blackwell family. Assuming the flagship GPU is the RTX 5090 (rather than a 5090 Ti), it is expected to debut with a launch price comparable to or slightly higher than the current-gen GeForce RTX 4090, sitting at $1,599.
| Specs | RTX 5090 | RTX 4090 |
|---|---|---|
| Interface | PCIe 5.0 x16 | PCIe 4.0 x16 |
| Memory | X GB (possibly 36GB) GDDR7 | 24GB GDDR6X |
| CUDA cores | 24,576 | 16,384 |
| L2 cache | 128MB | 72MB |
| Boost clock | 2.9GHz | 2.52GHz |
| Architecture | Blackwell | Ada Lovelace |
| Memory bus width | 512-bit | 384-bit |
| Memory bandwidth | 1,536 GB/s | 1,008 GB/s |
What to Expect from Nvidia GeForce RTX 5000 Series: Potential Innovations
Nvidia has long adhered to monolithic die designs for its graphics cards, but recent leaks suggest a shift towards Multi-Chip Module (MCM) architectures, potentially revolutionizing its upcoming GeForce RTX 5000 series.
In contrast to AMD’s approach of utilizing chiplets connected via interposers, Nvidia traditionally integrated all circuitry onto a single die, limiting transistor miniaturization without enlarging chip size. MCM, on the other hand, involves separate chips interconnected via interposers on a single substrate, enhancing scalability and accommodating better specifications.
According to notable leaker kopite7kimi, Nvidia’s forthcoming data center chips, the GB100 and GB102, are set to adopt an MCM design. Similarly, speculation suggests the high-end GB202 chip will introduce MCM architecture to consumer GPUs. However, budget and mid-range offerings like the GB203, GB205, GB206, and GB207 are expected to retain monolithic architecture.
Potential Technological Upgrades
1. Higher Bandwidth GDDR7 Memory
Disappointment among GPU enthusiasts stemmed from Nvidia’s decision to carry over GDDR6X and GDDR6 memory to the Ada Lovelace family. However, rumors abound regarding the flagship RTX 5000 series graphics cards leveraging the latest GDDR7 memory.
For instance, the memory bandwidth of the RTX 5090 is rumored to soar to 1,536 GB/s, marking a substantial 52% improvement over its predecessor, the RTX 4090. Initially, speculation surrounding the memory bus of the RTX 5090 varied, with conflicting reports from leaker kopite7kimi. While earlier indications suggested a 512-bit memory interface, subsequent updates hinted at retaining the 384-bit bus, akin to the RTX 4090 and RTX 3090 Ti. However, in a twist, kopite7kimi later reverted to the 512-bit bus assertion.
Additionally, it’s speculated that lower-tier models like the GB203 and GB205 may feature reduced memory interfaces of 256-bit and 128-bit, respectively. While rumors swirl around the VRAM size of the RTX 5090, with some suggesting 36GB of GDDR7 memory, it’s likely to offer at least 24GB, akin to the RTX 4090. Moreover, there’s speculation that the GB203 variant might boast half the memory of the GB202 model, potentially offering between 12 to 18GB, assuming it doesn’t adopt a “multi-chip package” design.
In summary, Nvidia’s prospective GeForce RTX 5000 series appears poised to introduce significant architectural innovations, potentially revolutionizing the landscape of high-performance graphics processing. While concrete details remain elusive, leaks and rumors hint at a promising future for gaming and data center applications alike.
Nvidia’s upcoming Blackwell series of GPUs is poised to once again push the boundaries of performance and efficiency, thanks in part to their adoption of the cutting-edge TSMC 4N manufacturing process. This process, derived from TSMC’s 5nm node, promises enhanced chip density, with the flagship GB202 chip boasting a 30% increase in this regard compared to its predecessor, the GB100 chip designed for data center GPUs.
Leakers such as kopite7kimi have shed light on key details of the upcoming GPUs. While there won’t be significant changes in the number of Graphics Processing Clusters (GPCs) or Texture Processor Clusters (TPCs), the GeForce RTX 5090 is expected to feature a 15% boost in frequency and a substantial 78% increase in cache size. Moreover, rumors suggest a staggering 50% improvement in scale, potentially resulting in an impressive 24,576 CUDA core count.
In terms of connectivity, the Blackwell GPUs are set to embrace the latest standards. While previous Ada Lovelace GPUs were compatible with PCIe 5.0 motherboards, they couldn’t fully exploit the capabilities of this high-speed interface. However, the Blackwell series is expected to capitalize on the lightning-fast data transfer speeds offered by PCIe Gen 5.0, promising enhanced performance and efficiency.
Additionally, the new GPUs are rumored to elevate display capabilities to new heights. According to kopite7kimi, they will support the DisplayPort 2.1 standard, which boasts significantly higher bandwidth, refresh rates, and resolution compared to the previous DisplayPort 1.4 standard. This enhancement ensures that users can enjoy cutting-edge visuals and immersive experiences without compromise.
Amidst the flurry of leaks surrounding Nvidia’s anticipated GeForce RTX 5000 series, it’s prudent to approach them with a dose of skepticism. While some of these leaks have stemmed from reputable sources, it’s essential to maintain a cautious stance. Remember the debacle surrounding the purported max TDP of 800W for the flagship RTX 4090 from the Ada Lovelace family? The community was abuzz with concern, only to be relieved when the RTX 4090 was unveiled with a more moderate 450W TDP.
Anticipation swirls around the possibility of Nvidia unveiling the Blackwell family later this year, but the company has opted for silence thus far. Nevertheless, it’s safe to assume that more details will emerge about the upcoming GPU family in due time. If the wait for the Blackwell series proves unbearable, the RTX 40-series GPUs still offer compelling options. Among them, the GeForce RTX 4070 stands out as a favorite, delivering commendable performance without breaking the bank.
Source: https://www.xda-developers.com/nvidia-rtx-5000-rumors/
To read more articles, visit: HWRIG
