The new cards are power hungry. The flagship RTX 2080 Ti draws 250 watts, while the “regular” RTX 2080 draws a minimum of 215 watts (225 for the Founder’s Edition), using 6-pin and 8-pin header connectors. NVIDIA is offering a “Founder’s Edition” card with much the same specs, but it can be overclocked to 1800 MHz, compared to 1710 Hz for the regular model.
The numbers don’t tell the whole story. NVIDIA has broken from its “GTX” naming scheme, with the “R” in RTX standing for “ray-tracing.” NVIDIA’s CEO Jensen Huang went into a lot of detail about it, but in brief, the new system does real-time ray-tracing that allows for realistic lights, shadows and cameras while gaming. We saw the result with a few incredibly realistic demos that showed light realistically glinting off of shiny surfaces that cost soft, life-like shadows.
If the new GeForce RTX 2080 or 2080 Ti would smash too big a hole in your budget, there’s good news. NVIDIA has unveiled a formidable second tier card, the GeForce RTX 1070, that gives you just about all the performance for quite a bit less money. It also packs 8GB of newfangled 14Gbps GDDR6 memory, more than the rumored 7GB, with 2304 CUDA cores. That should yield performance 40 percent better than the current GTX 1070, and 17 percent above the current GTX 1080.
As with the RTX 2080, the RTX 2070 will arrive first in a Founder’s Version that’s going to cost you a bit extra. However, it does offer a bit better overclocking than the regular model (1710 MHz overclocked compared to 1620 MHz). The RTX 2070 still draws a significant 175 watts of power, compared to 150 watts for the last model.
We don’t have all the pricing yet, but the Founder’s Edition GeForce RTX 2080 will cost $800 and start shipping on or around September 20th, 2018. The Founder’s Edition GeForce RTX 2070, meanwhile, will cost $600, but there’s no word on shipping yet. For now, NVIDIA is limiting sales of the RTX 2080 to two per customer, so if you have a multi-GPU rig larger than that in mind, you’ll have to wait a bit.
No comments: