The RTX 3090 has been dethroned by the arrival of the new king of desktop GPUs. Nvidia’s freshly unveiled RTX 3090 Ti is currently the most powerful desktop GPU to date, outperforming its non-Ti sibling by around 13%.
The American hardware manufacturer has not yet published all of the data, but we have learned of a few specifications. The 3090 Ti will have the same 24GB GDDR6X memory as the 3090, but will run at 21Gbit/s instead of 19.5Gbit/s. The older sibling of the 3090 can compute 40 shader teraflops, 78 RT teraflops, and 320 tensor (AI) teraflops.
The standard 3090, on the other hand, could manage 35.6 shader teraflops, 69.5 RT teraflops, and 285 tensor teraflops.
We won’t know the full storey until Nvidia releases more information about the 3090 Ti, but in the meantime, we can make a few informed guesses about how Nvidia attained this performance. For instance, the 3090 only uses 82 of the GA102’s 84 streaming multiprocessors, thus the Ti sibling may be utilising all 84.
Nvidia has most likely increased the primary GPU clock speeds in addition to increasing the memory clock speeds. To achieve the 40-teraflop figure provided by Nvidia, the boost frequency on the 3090 Ti will need to be roughly 1,850MHz (up from 1,695MHz).
According to Nvidia, more information will be released later this month, implying that the 3090 Ti will be released at a small event. That’s where we’ll find out how ridiculous the pricing is. As a reminder, the base RTX 3090 is only $1500.