In 2008, AMD announced the Radeon HD 4850 desktop graphics card with a TDP of 110W, a price tag of $199, 0.5 GB of GDDR3 memory, and a GPU frequency of merely 625 MHz. This was the first 1 TERAFLOP graphics card ever produced. It took 14 years to multiply this accomplishment by 100.
NVIDIA has now released their RTX 4090 GPU, which can be overclocked and achieve up to 100 TFLOPS. The RTX 4090 is the first graphics card to utilize 4nm Ada Lovelace architecture and is a powerful 4K gaming card. The card is set to 450W TDP by default, although it looks like NVIDIA could have drastically reduced the power consumption.
At 300W, performance drops by 9%, but at 350W, it drops by only 6%, according to ComputerBase. However, this is reciprocal. Numerous media sources have reviewed the RTX 4090 Founders Edition today. It has a power restriction of 600W. With power viruses such as Furmark, the power usage can surpass TDP by as much as 496W and even 667W when overclocked. What if power consumption was unimportant, and the objective was to attain 100 TFLOPS at any cost? This appears to have been addressed by Wccftech, who overclocked their sample to 3150 MHz and completed the AIDA64 GPGPU single-precision computation test at 100976 TFLOPS.
According to speculations, the RTX 4090 can surpass 100 TFLOPS and brings a new gaming graphics era. The only uncertainty that remains at this time is what plans AMD has for RDNA3. Are comparable accomplishments anticipated, or will AMD demonstrate dominance in alternative metrics? The answer should show up by the end of the next month.