The Nvidia RTX 5090 and RTX 5080 specifications have been a hot topic lately, fueled by numerous leaks and rumors over the past months. Details ranging from memory capacity to CUDA cores have emerged, but the reliability of this information has varied. Recently, reputable leaker @kopite7kimi has provided some fresh insights into the specs of the RTX 5090 and RTX 5080, helping to clarify the situation.
New Nvidia RTX 5090 Specifications Leak Indicates Liquid Cooling Requirement
According to @kopite7kimi on Twitter, the Nvidia RTX 5090 will utilize the GB202 die and feature a total of 21,760 CUDA cores. This leaker has established credibility by accurately predicting the specs of multiple Nvidia GPUs in the past; however, it’s prudent to approach this information cautiously.
The full GB202 die is rumored to encompass a total of 24,576 cores, suggesting the Nvidia RTX 5090 is a downscaled variant. The complete GH202 die may be earmarked for a potential RTX 5090 Ti, although this seems less likely with AMD stepping back from the high-end GPU market.
If AMD remains absent from the competition, Nvidia has little incentive to launch the complete GB202 die. Additionally, @kopite7kimi posits that the RTX 5090 will come with 32GB of memory, configured on a 512-bit bus using advanced 32Gbps GDDR7 memory modules, as previously reported.
Furthermore, this leaker indicates that the TDP for the RTX 5090 will sit at 600W, which is significantly higher than earlier speculations, suggesting that maintaining adequate cooling for this powerful GPU will be a formidable task. Nvidia may need to explore liquid cooling solutions to manage the GPU’s thermal performance, potentially necessitating users to upgrade their PSU as well.
Here’s a comparison of the Nvidia RTX 5090 specs against the RTX 4090:
Specifications | RTX 5090 | RTX 4090 |
GPU variant | GB202-300-A1 | AD102-300-A1 |
CUDA cores | 21,760 | 16,384 |
L2 cache | 128 MB | 72 MB |
Memory capacity | 32 GB | 24 GB |
Memory bus width | 512-bit | 384-bit |
Memory module | GDDR7 | GDDR6X |
TDP | 600W | 450W |
The specifications indicate that the RTX 5090 significantly outperforms the RTX 4090 in terms of CUDA cores, L2 cache, and memory capacity, among other specs. This upcoming GPU is shaping up to be a formidable contender, but its pricing remains a critical factor.
Nvidia has consistently raised the prices of its flagship graphics cards since the introduction of the first Nvidia RTX graphics card, and rumors suggest that the pricing for the RTX 5090 could exceed $2,000. Should Nvidia incorporate liquid cooling for this model, it may justify such a price hike.
New Nvidia RTX 5080 Specs Leak
In addition to the RTX 5090 rumors, leaker @kopite7kimi has also divulged the specifications for the Nvidia RTX 5080. According to these leaks, the RTX 5080 will feature 10,752 CUDA cores, an improvement over its predecessor, although not as substantial. It is expected to have 16GB of memory with a 256-bit bus and utilize fast 32Gbps GDDR7 memory modules.
Here’s how the Nvidia RTX 5080 specs compare to the RTX 4080:
Specifications | RTX 5080 | RTX 4080 |
GPU variant | GB203-400-A1 | AD103-300-A1 |
CUDA cores | 10,752 | 9,728 |
L2 cache | ? | 64 MB |
Memory capacity | 16 GB | 16 GB |
Memory bus width | 256-bit | 256-bit |
Memory module | GDDR7 | GDDR6X |
TDP | 400W | 320W |
The RTX 5080 showcases a reasonable upgrade over its predecessor, although not to the same extent as the leap from the RTX 4090 to the RTX 5090. The TDP for the RTX 5080 has been leaked at 400W, which, while higher than the previous 320W mark, should still allow for air cooling. The Nvidia RTX 4090, despite its 450W TDP, remains air-cooled.
Comparing Nvidia RTX 5080 and RTX 5090
The specifications between the Nvidia RTX 5090 and RTX 5080 reveal a significant performance disparity, with the latter possessing half the specs of the former. This performance gap could only be bridged by the potential introduction of an RTX 5080 Ti, although no leaks currently hint at its availability.
Can the Nvidia RTX 5080 Outperform the Nvidia RTX 4090?
With the latest specs leak indicating that the Nvidia RTX 5080 includes around 10,752 CUDA cores paired with a 256-bit memory bus, it falls significantly short of the specifications presented by the RTX 4090. Although changes in Nvidia’s new Blackwell architecture and the integration of GDDR7 memory modules may provide enhancements, they are unlikely to completely close the performance gap.
If we accept the information regarding RTX 5080’s specifications as accurate, the RTX 4090 will maintain a lead with over 50% more CUDA cores. It’s unlikely the RTX 5080 will surpass the RTX 4090 in performance, although it may deliver comparable results in certain scenarios.
Leave a Reply