The Nvidia GeForce RTX 5090 specs have simply leaked, in line with the most recent phrase within the rumor mill, together with the RTX 5080 specs too. If these leaks are real, the brand new flagship Nvidia gaming GPU seems to be like a severely highly effective monster, however the RTX 5080 seems to be comparatively weak.
Not that this could fear Nvidia. With AMD anticipated to solely be specializing in the mid-range GPU market when it launches its RDNA 4 structure, Nvidia ought to find yourself having the flagship greatest graphics card market all to itself. Nvidia additionally hopefully realized a lesson with the launch of the GeForce RTX 4080 again in 2022, which is that you would be able to’t really cost no matter you need for a graphics card if the worth is horrible – individuals will simply select to not purchase it. Nvidia later addressed this error with the cheaper and extra highly effective RTX 4080 Tremendous earlier this 12 months.
Let’s begin with the RTX 5090 specs, which have seemingly simply been dropped by common tech leaker kopite7kimi. In a publish on X (previously Twitter), kopite7kimi revealed a string of numbers, which seem so as to add as much as an extremely highly effective new graphics card, utilizing the GB202-300-A1 Blackwell chip. Based on the leak, the RTX 5090 will function 21,760 FP32 (32-bit floating level) items, which means it has 21,760 CUDA cores, as Nvidia’s core design can course of each FP32 and INT32 calculations.
With 128 CUDA cores in every streaming multiprocessor (SM) unit, this spec would give the RTX 5090 170 SMs, with every SM anticipated to have one RT core for ray tracing. When it comes to numbers, that is considerably larger than the RTX 4090, which has 16,384 CUDA cores unfold over 128 SMs, though this isn’t a like-for-like comparability on the subject of efficiency, because the RTX 5090 will use the Blackwell structure somewhat than the Ada Lovelace structure utilized by the RTX 4090.
The leak additionally exhibits a fats reminiscence system, when it comes to each capability and bandwidth. If this leak is real, then the RTX 5090 has an enormous 32GB of GDDR7 VRAM connected to a 512-bit interface. As some extent of comparability, the RTX 4090 has 24GB of GDDR6X VRAM and a 384-bit interface.
There’s no point out of the frequency of the VRAM within the leak, but when it’s 28Gbps, which is definitely possible, then the entire reminiscence bandwidth can be an enormous 1.79TB/s, in comparison with simply over 1TB/s on the RTX 4090. The leak additionally claims the RTX 5090 will draw 600W of energy, which corresponds with earlier leaks concerning the RTX 5090 energy draw.
This had beforehand led some individuals to take a position that the RTX 5090 has two 12Vx6 energy connectors, however tech web site Benchlife has seemingly put that hypothesis to mattress in a latest report, claiming that “there aren’t any two units of 12v on the GeForce RTX 5090, as reported by some media.” Benchlife additionally claims that the RTX 5090 could have a 14-layer PCB and use the PCIe 5.0 interface.
Nvidia GeForce RTX 5090 specs estimate
That is our greatest estimate for the Nvidia GeForce RTX 5090 specs, based mostly on rumor and hypothesis, as Nvidia hasn’t made an official announcement but.
| Nvidia GeForce RTX 5090 | |
| GPU | GB202-300-A1 |
| CUDA cores | 21,760 |
| SMs | 170 |
| RT cores | 170 |
| Interface | 16x PCIe 5.0 |
| VRAM | 32GB GDDR7 |
| Reminiscence interface | 512-bit |
| Energy draw | 600W |
| Energy connector | 1 x 16-pin |
Subsequent up is the RTX 5080 which, very like the RTX 4080 when in comparison with the 4090, seems to be prefer it has a lot much less energy than the 5090, though it ought to nonetheless be succesful. Based on kopite7kimi, this new GPU relies on the GB203-400-A1 chip and has 10,752 CUDA cores, lower than half the variety of the 5090, equating to 84 SMs (and 84 RT cores).
In the meantime, the reminiscence setup seems to be similar to that of the RTX 4080, with 16GB of VRAM connected to a 256-bit interface, however with GDDR7 reminiscence doubtlessly growing the bandwidth. On the plus facet, it reportedly attracts much less energy, with a score of 400W.
Nvidia GeForce RTX 5080 specs estimate
That is our greatest estimate for the Nvidia GeForce RTX 5080 specs, based mostly on rumor and hypothesis, as Nvidia hasn’t made an official announcement but.
| Nvidia GeForce RTX 5080 | |
| GPU | GB202-300-A1 |
| CUDA cores | 10,752 |
| SMs | 84 |
| RT cores | 84 |
| Interface | 16x PCIe 5.0 |
| VRAM | 16GB GDDR7 |
| Reminiscence interface | 256-bit |
| Energy draw | 400W |
| Energy connector | 1 x 16-pin |
In fact, none of this has been formally confirmed by Nvidia, so take all this with a pinch of salt proper now. Nevertheless, if these are genuinely the specs of the RTX 5080, then let’s at the least hope that Nvidia doesn’t give it a ridiculous value, because it did with the RTX 4080 when it first launched. It at present seems to be as if there’s an infinite gulf between the 5080 and 5090 on the subject of each GPU energy and the reminiscence setup, and it might be unhealthy information for avid gamers if there isn’t a lot of a distinction in value between the 2 chips.
It could even be disappointing if Nvidia’s second-in-command did certainly solely have a 256-bit reminiscence interface and 16GB of VRAM. As some extent of comparability, even AMD’s current-gen Radeon RX 7900 XT has 20GB of VRAM and a 320-bit interface. If this GPU finally ends up costing $999 or extra, then I’d anticipate it at the least to have a 20GB/320-bit setup to assist see it by the subsequent few years of recreation releases at excessive settings.
For extra details about what’s coming with Blackwell for avid gamers, try our Nvidia GeForce RTX 5090 information, the place we share the whole lot we at present know concerning the new chip.
