GeForce RTX 3080 tests in CUDA and OpenCL tasks – an increase of 68% compared to RTX 2080
After the announcement of the GeForce RTX 3000 series, video card tests are gradually emerging. We have previously written about the performance evaluations of the RTX 3080 versus the RTX 2080 by Eurogamer. Now it’s time for similar RTX 3080 scores in CUDA and OpenCL compute tasks using the Compubench benchmark suite.
THE GeForce RTX 3080 graphics card has been tested using 456.16 drivers. We are talking about a variant with 10 GB of video memory and 68 computing units (streaming multiprocessors) with a frequency of 1710 MHz. This is either the Founders Edition model tested by one of the reviewers or a third-party variant with the same clock speed. The CUDA info page confirms that this is a 320-bit memory configuration with 19 Gbps modules.
These types of tests are computationally-oriented and have nothing to do with gaming performance directly. Especially when you take into account all the new blocks for speeding up calculations (for example, tensor kernels and ray tracing blocks).
VideoCardz staff collected data for the CUDA and OpenCL benchmarks. The average for both tests is more or less the same. GeForce RTX 3080 on average can offer 68% more performance than RTX 2080 and 38-41% more than RTX 2080 Ti.
Recall: NVIDIA GeForce RTX 3080 is equipped with an 8nm GA102-200 graphics processor with 8704 CUDA cores, works in conjunction with 10 GB of the new generation GDDR6X memory, and uses a 320-bit memory bus. The graphics card will officially go on sale on September 17th. Reviewers are expected to publish their Founders Edition benchmark tests on September 14th, with third-party model reviews coming out 3 days later.
