Monday, January 4th 2021

NVIDIA GeForce RTX 3080 Mobile Maxes Out "GA104" Silicon

With the desktop GeForce RTX 3080 being based on the "big" GeForce Ampere silicon, the "GA102," we wondered how NVIDIA would go about designing the RTX 3080 Mobile. It turns out that the company will max out the smaller "GA104" silicon on which the desktop RTX 3070 and RTX 3060 Ti, are based. An unreleased ASUS ROG Zephyrus Duo (GX551QS) gaming notebook's Geekbench online database entry reveals the name-string and streaming multiprocessor (SM) count of the RTX 3080 Mobile.

The Geekbench online database entry lists out the OpenCL device (GPU) name-string as "GeForce RTX 3080 Laptop GPU," and OpenCL compute unit (SM) count as 48. This corresponds with the maximum SM count of the "GA104," which features 6,144 Ampere CUDA cores spread across 24 TPCs (48 SM), 48 2nd generation RT cores, 192 3rd generation Tensor cores, 192 TMUs, and 96 ROPs. The Geekbench entry also reveals the video memory amount as 16 GB, maxing out the 256-bit wide GDDR6 memory interface of the "GA104," likely using 16 Gbit memory chips.
Sources: Geekbench Database, TUM_APISAK (Twitter), VideoCardz
Add your own comment

6 Comments on NVIDIA GeForce RTX 3080 Mobile Maxes Out "GA104" Silicon

#1
owen10578
I'm gonna guess the mobile RTX 3080 is gonna be like a desktop RTX 3060Ti in performance. Just like the old RTX 2080 mobile was more like an RTX 2060 in performance.
Posted on Reply
#2
hgk87
I doubt they'd get away with that "with max-q design" suffix lightly this time as it's an entirely different silicon being used here.
Posted on Reply
#3
Vya Domus
Really a 3070, probably even slower than that. I see they're back to naming their GPUs something which they aren't.
Posted on Reply
#4
Anymal
Nothing new regarding that.
Posted on Reply
#5
nguyen
So these are the highest binned GA104 die, does it justify selling them at same price bracket as the previous 2080 Super Max-Q which use the 545mm2 TU-104 die ?
Posted on Reply