On paper, Nvidia’s RTX 6000 Ada professional graphics card has more compute power than its GeForce RTX 4090 flagship board for gamers. Still, when I tested it on the 3DMark TimeSpy Graphics benchmark, it didn’t really beat me. reddit Post noticed by @harukaze5719.
Nvidia’s RTX 6000 Ada 48GB graphics board is based on the AD102 GPU with 18,176 CUDA cores enabled. Nvidia and its partners haven’t specified clocks for his GPU, but have revealed that the card has a peak computing performance of 91.1 FP32 TFLOPS (which suggests a clock of around 2,505 MHz). . In contrast, Nvidia’s GeForce RTX 4090 24GB runs at up to 2,520 MHz and uses an AD102 GPU with 16,3840 CUDA cores rated up to 82.575 FP32 TFLOPS.
There is one big difference between the two cards (although they appear to use the same PCB). Gaming cards can consume up to 450W of power, while professional cards can consume up to 300W of total board power. It can run much longer and at higher clocks than the former. While this limits performance, it extends the lifespan of the RTX 6000 Ada and GeForce RTX 4090.
This is exactly the performance of the two cards on the 3DMark Time Spy Graphics benchmark. The RTX 6000 Ada’s graphics score was 30,158 points in the latest Windows update (see screenshot), but before the update he may have been 36,844 points, according to the individual who posted the score. Healthy Blood-54On the other hand, the GeForce RTX 4090 scores easily 40,964 point.
The reason for such relatively poor performance is of course the relatively low GPU clock, which drops from a whopping 3 GHz or more in the first few minutes of the test to around 1.5 GHz afterwards. Interestingly, the owner says the board doesn’t overheat, so it probably starts dropping clocks at relatively low temperatures.
“At full use, the core temperature is about 85°C and the fan is about 70%,” explains Healthy-Blood-54. [RTX] A6000 [based on GA102]Remember, though, that these are battle cards. They can probably run 24/7 on a core at 95°C for 10 years before problems arise. VRAM isn’t the only cost. ”
The RTX 6000 Ada 48GB graphics card can’t beat the GeForce RTX 4090 despite its high number of enabled CUDA cores and high peak compute throughput in FP32, but with 18,176 stream processors enabled. The rumored GeForce RTX 4090 Ti could significantly outperform the current flagship. .
