After many rumors throughout the year, a big livestream finally revealed the new RTX3000 line. The RTX3090 will be available from September 17, the RTX3080 will launch September 24, and the RTX3070 will follow sometime in October.
The graphic cards are based on NVIDIA's Ampere technology, there is new SM architecture, 3rd gen Tensor cores, 2nd Gen Ray Tracing cores, and all three of them are twice as fast as the last generation. It also uses Samsung 8nm fabrication instead of TSCM.
RTX3090 replaces the Titan card, followed by the RTX3080 and RTX3070. We strongly expect RTX3060 to be launched later this year as well. All cards have HDMI 2.1 and Displayport 1.4a. The rumors turned out to be true, and the power consumption is at a staggering 350 watt. We do, however, not need to worry about 12 pin connection, 2x8 pins are still used for the RTX3090 and the RTX3080.
NVIDIA has given out CUDA cores like it was candy, 10.496 for RTX3090, 8.794 for RTX3080, and a RAM configuration is now 24GB DDR6X for the RTX3090, 10GB DDR6X for the RTX3080, and 8GB DDR& for the RTX3080. If NVIDIA is to be believed, RTX3070 almost starts where the RTX2080TI started, but at a totally different price. $499 is the price of the RTX3070, which is $100 less than the starting price of the RTX2070.
RTX3090 comes with 19.5 Gbps RAM, 384 bit 935 GB/s bandwidth, 82 RT cores, 328 Tensor cores and 96 ROPs (Render Output Units).
Pricing is 1499 Euro for the RTX3090, 699 Euro for the RTX3080, and 499 Euro for the RTX30700. It's actually cheaper than its predecessor. That we did not expect, but we are very delighted.
If you have got the money for an RTX3090, it will give you the option of 8K HDR content and capture, however, this is while using DLSS, AI-based image, and FPS improving technology, and something that has impressed us so far. While not all games are able to achieve 60FPS even at this level, don't worry, you will still need to find the money for a 4K gaming monitor, which start at the same price as a used car.
Some new things were presented, like the NVIDIA Reflex that works with 1000 series cards as well, offering up to 30% improvement on total system latency. How this is done is not clear, but it does add "free" performance to new and existing hardware from NVIDIA. Dedicated video encoding on the card called NVECNC will improve and conquer a large part of the streaming market as well, especially as it can even take some load of the CPU by doing some of the work that is normally dedicated to the CPU.
However, we still have the issue of the new cooling design. This time NVIDIA has opted for a separate dual intake-output design, with the PCB having two cooled areas, but nothing in the middle. Both intake fans are placed in the case, while one output is at the I/O, and the other is inside the case, facing the RAM and CPU. NVIDIA claims 55% higher airflow over blower-style coolers, 30% increased efficiency, and 3x the silence, however, the cost is direct heat on the CPU and CPU cooler, and even with a liquid solution, hot air is still blown more or less directly on the tubes, and an air cooler will have hot air fed directly into the intake.
We are looking forward to trying out both NVIDIA and 3rd party cards because this is an upgrade many have been waiting for in a very long time.
Loading next content