New alleged 3DMark benchmarks for the GeForce RTX 4080 (16 GB) have surfaced from Chiphell, alluding to a new graphics card from NVIDIA that can hit scores of up to 28,929 in the benchmarking suite’s Time Spy test. The score implies that the second-ranking Lovelace option features performance that is up to 62% faster than the GeForce RTX 3080 (10 GB), according to a few of the comparisons shared online. Screenshots also indicate that it was running at a GPU clock speed of 3.0 GHz, a 490 MHz increase over NVIDIA’s official number of 2.51 GHz.
From a Wccftech report:
We can’t say for sure if this is an overclocked variant or if it is running at stock clocks. One can also make the graphics card run at higher clocks simply by raising the power limits. The graphics card is seen running at a TBP of 333W while the TGP is rated at 290W. This is within the 320W spec that all reference variants should offer. The GDDR6X memory is operating at 22.4 Gbps so it looks like the memory clock was not touched.
In terms of performance, the graphics card scored 28929 points which marks a 32% increase over the RTX 3090 Ti and a 62% increase over the RTX 3080 10 GB graphics card. The card also pushes up to 57% higher performance than the 12 GB RTX 3080.
Here’s what the performance of the GeForce RTX 4080 (16 GB) might look like when pitched against the GeForce RTX 3090 Ti, according to the various 3DMark scores that have leaked thus far.
- GeForce RTX 4080 16 GB vs. RTX 3090 Ti 24 GB in Time Spy Extreme: 20% Faster
- GeForce RTX 4080 16 GB vs. RTX 3090 Ti 24 GB in Fire Strike Ultra: 29% Faster
- GeForce RTX 4080 16 GB vs. RTX 3090 Ti 24 GB in Port Royal (DXR): 16% Faster
NVIDIA plans to release the GeForce RTX 4080 (16 GB) next month alongside a 12 GB variant, which features some further differences such a reduced CUDA Core count (7,680 vs. 9,728) but higher clocks. The 16 GB and 12 GB versions will start at $1,199 and $899, respectively.
Discussion (4 replies)
Join Discussion →Again a far shout for 2x to 3x. But not bad just not.. wow.
It's a normal generational leap, more than we've had previously to be honest, I remember some generations only giving maybe 30% per generation.
Close to the old days where we used to get double the performance every generation.
Waiting for real benchmarks
I remember the lap with Maxwell... that was pretty cool.