According to MSI’s news post, the memory switch counts for a 5% performance increase over the otherwise identical 14Gbps RTX 2080 Ti Gaming X Trio. Both feature the same GPU boost clock of 1,755MHz, and the cooling solution remains the same beefy, slightly screwy, design. However, the tweaked memory on the Gaming Z Trio tips it over the line a little ahead of the 14Gbps Gaming X Trio.
That 5% could be enough to see the Gaming Z Trio outperform even Nvidia’s super-smart Titan RTX. The inference card built for AI and deep learning is more or less an RTX 2080 Ti with all the extras: the full TU102’s 4,608 CUDA Cores, 1,770MHz boost clock, and 24GB of 14Gbps GDDR6 memory.
With plenty of memory to keep it well-fed, and a slight CUDA and clock speed advantage, the Titan RTX has long been marginally faster than the Founders Edition RTX 2080 Ti. However, third-party coolers that allow Nvidia’s GeForce GPUs to really crank up the frequencies (thanks to Nvidia’s GPU Boost technology), and generous factory overclocks, have seen the gap shrink to just a couple of frames.
If MSI’s numbers are to be believed, the move to 16Gbps GDDR6 memory should take the highly competitive Gaming X performance and turn it up to… well, 16. In doing so it could allow the Gaming Z Trio to top the Titan RTX in-game by a small margin.
The value of memory bandwidth cannot be understated, either. Take the GTX 1660 Super. Identical in almost every way to the GTX 1660, the upgrade from GDDR5 to GDDR6 memory caused it to perform some 15-20% faster.
Perhaps that’s why MSI is keen to release an updated Turing card a year and a half following its initial release: it wants the championship title before the next generation of Nvidia’s professional and gaming chips tear up the rulebook once again later in the year.