Friday, January 24th 2025
New Leak Reveals NVIDIA RTX 5080 Is Slower Than RTX 4090
A set of newly leaked benchmarks has revealed the performance capabilities of NVIDIA's upcoming RTX 5080 GPU. Scheduled to launch alongside the RTX 5090 on January 30, the GPU was spotted on Geekbench under OpenCL and Vulkan benchmark tests—and based on the performance, it might not make it among the best graphics cards. The tested device was an MSI-branded RTX 5080 labeled as model MS-7E62. This setup had AMD's Ryzen 7 9800X3D processor, which many consider one of the best CPUs for gaming. It also included an MSI MPG 850 Edge TI Wi-Fi motherboard and 32 GB of DDR5-6000 memory.
The benchmark results show that the RTX 5080 scored 261,836 points in Vulkan and 256,138 points in OpenCL tests. Compared to the RTX 4080, its previous version, the RTX 5080 has a 22% boost in Vulkan performance and a small 6.7% gain in OpenCL. Reddit user TruthPhoenixV found that on the Blender Open Data platform, the GPU got a median score of 9,063.77. This score is 9.4% higher than the RTX 4080 and 8.2% better than the RTX 4080 Super. Even with these improvements, the RTX 5080 might not outperform the current-gen top-tier RTX 4090. In the past, NVIDIA's 80-class GPUs have beaten the 90-class GPUs from the previous generation, but these early numbers suggest this trend might not continue for the RTX 5080.
The RTX 5080 uses NVIDIA's latest Blackwell architecture, with 10,752 CUDA cores spread across 84 Streaming Multiprocessors (SMs) versus the 9,728 cores in the RTX 4080. It has 16 GB of GDDR7 memory on a 256-bit bus. NVIDIA says it can deliver 1,801 TOPS in AI performance through Tensor Cores and 171 TeraFLOPS of ray tracing performance using its RT Cores.
That said, it's important to note that these benchmark results have not been fully verified so we should wait for the review embargo to lift before concluding.
Sources:
DigitalTrends, TruthPhoenixV
The benchmark results show that the RTX 5080 scored 261,836 points in Vulkan and 256,138 points in OpenCL tests. Compared to the RTX 4080, its previous version, the RTX 5080 has a 22% boost in Vulkan performance and a small 6.7% gain in OpenCL. Reddit user TruthPhoenixV found that on the Blender Open Data platform, the GPU got a median score of 9,063.77. This score is 9.4% higher than the RTX 4080 and 8.2% better than the RTX 4080 Super. Even with these improvements, the RTX 5080 might not outperform the current-gen top-tier RTX 4090. In the past, NVIDIA's 80-class GPUs have beaten the 90-class GPUs from the previous generation, but these early numbers suggest this trend might not continue for the RTX 5080.
That said, it's important to note that these benchmark results have not been fully verified so we should wait for the review embargo to lift before concluding.
260 Comments on New Leak Reveals NVIDIA RTX 5080 Is Slower Than RTX 4090
So really?They're not, that user seems to be trying at the yanking of chains..
I have 5080 as well, how's your 3d mark scores? Port Royal? Or speedway? You will be in Hall of Fame for sure!
Let me light the case. Passmarks GPU tests are completely CPU bound. For example in DX9 10 11 12 tests my 4090 was chilling at 30 to 60%. So he probably has a 9800x 3d. The only test that actually stretches the GPU is Compute, and in that one I scored over 30k vs his 23k. That's a 30% difference btw. So yeah, he hasn't "unlocked" 100% of his 5080s brainpower.
I'm not here to believe. I'm here to learn.
An artificial driver limit to disable parts of a perfectly working chip doesn't make any sense from any standpoint, imo, but if he can support his claim with actual evidence, I'm willing to listen.
What NV does do to segment GeForce to their professional cards is selectively disable optimizations that target certain professional suites, and disable certain esoteric features like 30-bit color SDR. The pro-viz optimizations that target things like specviewperf, Autodesk suites, CATIA, etc. - was famously enabled specifically on Titan X Pascal, Xp, V and RTX, with all other professional features disabled, NVIDIA did this as an answer to the Vega Frontier, which initially explicitly supported both Radeon Pro Software and Adrenalin - nowadays this still works but it's a registry leftover and has to be toggled manually by the user.
Vega FE likewise didn't really enable everything that WX 9100 supported (stereoscopic 3D, ECC, deep color SDR, genlock etc.) are all disabled and hidden, although if you flash a WX 9100 BIOS on that GPU all of these features will be restored and fully functional as the core is exactly identical, and so is the HBM memory used, with the exception of genlock since the Vega FE board physically doesn't have the syncing connector. Only other catch is that since WX 9100 has 6 mDP and Vega FE is 3 DP + 1 HDMI, the HDMI port gets knocked out and DPs 1-3 get detected as the first three ports, with no way to connect anything to 4, 5 and 6 as that physically doesn't exist on the FE board. Since AMD bailed out of the "prosumer" deal with the Radeon VII, NV just released the 3090 as a pure gaming card, buried the Titan line and kept it that way until now. RTX 5090 is... a purebred gaming card. No extra features extended to it.
IF, and only IF this dude is telling the tiniest bit of truth, what he came across is likely the lock on pro-viz optimizations, which to the best of my knowledge, do not affect Cinebench but you should see significant gains in the specviewperf benchmarks.
gwpg.spec.org/benchmarks/benchmark/specviewperf-2020-v3-1/
www.techpowerup.com/gpu-specs/nvidia-gb205.g1074
Given the poor price-to-performance ratio of the 5070 compared to 9070/9070 XT, there's a good chance that next year we'll see the full-die 5070 Super replacing the 5070 at $550 MSRP.
Its the internet. The baseline response I have when someone says anything that isnt common sense is 'yeah, whatever'. Turns out to be the correct response in 99,99% of the exchanges you have on this medium. One needs only a brief look at social media discourse to get proof of that.
2. If we assume that the 5080 works with 40% of its parts disabled by default, that means that the chip is capable of performing 40% better than the 4080 Super with a similar number of components running at similar clock speeds, or it has a 12% higher power consumption while using 40% fewer components. Neither of these is possible.
I'm still willing to listen to anyone who wants to prove these points wrong simply on the basis of the above.
I ain't fallin for that trap anymore. I like to use history as the biggest teacher. Everything we see has been done before and got its reality check before. Miracles don't really happen anymore. Boring, but true. Its a bit like an adblocker; the blacklist keeps growing, and the internet keeps getting better that way. Less is more.
For a while, I used to think that TPU was different, that this was a place for people who really understand tech, but I have to admit sadly that it isn't.Very true.