When Nvidia announced its RTX family, it launched three GPUs, each with their own specific underlying design. The RTX 2070 is based on the TU106 GPU (2,304 cores), the RTX 2080 uses the TU104 GPU (3,072 cores), and the RTX 2080 Ti uses the TU102 GPU (4,608 cores). In two of these cases, however, Nvidia opted not to enable the full measure of cores available on the card. The RTX 2080 ships with 2,944 cores active, while the RTX 2080 Ti has just 4,352 cores. This is fairly standard practice in GPU design since it gives a company breathing room to recover bad die if an early chip isn’t perfect, and allows for some potential follow-up products if a manufacturer needs to bump performance by a bit in a midmarket refresh.
Nvidia, having already launched the RTX 2070, 2080, and 2080 Ti, is moving forward to push the RTX Titan into view. This new chip is the full-fat version of the RTX 2080 Ti, with 4,608 CUDA cores, 12GB of memory clocked at 14Gbps, and a 384-bit memory interface instead of the RTX 2080 Ti’s 352-bit interface. The total number of tensor cores and RTX cores are both slightly higher, at 576 and 72, as opposed to 544 and 68. In all cases, these improvements work out to a 5.8 percent gain in available resources. The GPU clock of 1350MHz matches the RTX 2080 Ti, while the maximum boost clock of 1770MHz is ~8 percent faster than Nvidia’s current top-end card. Video RAM is also doubled, at 24GB altogether. This matches the RTX 6000, which also has a 24GB RAM buffer. The RTX Titan is capable of full-speed FP16 w/FP32 accumulate performance, other cards in the GeForce family only perform this at 50 percent speed.
Ziggy the kitten helping me upgrade my GPU in the kitchen.
A post shared by Gavin Free (@gavinfree) on Nov 30, 2018 at 12:24pm PST
Nvidia teased the GPU through “accidental” reveals on Instagram and a streamer’s account before revealing it officially. Turing appears to have been selling fairly well in its early days, but the series launch hasn’t been without controversy. Nvidia’s decision to raise prices without commensurate performance improvements has not been popular. The justification that everyone should embrace higher prices because of the amazing potential of ray tracing in games hasn’t fallen on particularly fertile ground, either — especially given that RTX carries approximately a 60 percent performance penalty. The RTX 2070, a $500 card, maintains ~45fps in Battlefield V with RTX enabled at 1080p. Meanwhile, early problems plaguing the RTX 2080 Ti were confirmed to be caused by Nvidia inadvertently shipping defective cards from the factory. The company has been replacing and solving this issue, but it’s been another controversial situation within the GPU community.
The RTX Titan certainly won’t come cheap. While it offers performance improvements and full-speed FP16 perf w/FP32 accumulate, it’s also going to sell for $2,500, more than double the price on the RTX 2080 Ti. This GPU isn’t really intended for gaming, but between the clock gains and the additional cores, overall performance should be 10-15 percent higher than the RTX 2080 Ti.
Now Read: Nvidia Stock Plummets on High Inventory Fears, The First Battlefield V DXR Ray Tracing Benchmarks Have Dropped, Nvidia RTX 2080 and RTX 2080 Ti Review: You Can’t Polish a Turing