Manli has launched the trade’s first Nvidia GeForce RTX 3060 graphics card with solely 8GB of reminiscence. The brand new board has the identical compute efficiency as fully-fledged GeForce RTX 3060 12GB product, nevertheless it cuts down the reminiscence capability and bus width, which in flip reduces reminiscence bandwidth. All of these modifications ought to have a tangible impact on efficiency.
Manli’s GeForce RTX 3060 8GB GDDR6 graphics card (through VideoCardz (opens in new tab)) presumably makes use of the identical GA106 graphics processing unit because the 12GB card. It comes with 3584 CUDA cores with a 1777 MHz increase clock, the identical clocks because the reference RTX 3060 12GB.
In contrast to the everyday GeForce RTX 3060 12GB board that encompasses a 192-bit reminiscence bus with 360 GB/s peak reminiscence bandwidth, the GeForce RTX 3060 8GB solely encompasses a 128-bit reminiscence interface. That likewise cuts the height reminiscence by 33%, down to simply 240 GB/s. In circumstances the place reminiscence bandwidth issues, the brand new board will find yourself being considerably slower than the present GeForce RTX 3060 with 12GB of GDDR6 reminiscence.
Apart from the narrower reminiscence bus, the GeForce RTX 3060 8GB has the identical specs because the 12GB card. It comes with a 170W TBP (Complete Board Energy) score, a dual-slot cooling system, and 4 show outputs (three DisplayPort 1.4, one HDMI 2.0).
Neither Manli nor Nvidia have given an MSRP on the brand new RTX 3060 8GB mannequin, although because it ought to land between the GeForce RTX 3050 8GB ($249) and GeForce RTX 3060 12GB ($329), it’ll seemingly be priced accordingly. We count on different graphics card distributors will comply with Manli shortly with their very own variants of the RTX 3060 8GB.
The impetus behind the cardboard seems to be a need to supply one thing between the slightly anemic GeForce RTX 3050 and the present RTX 3060 12GB. The 3050 will get pummeled by AMD’s RX 6600 in addition to Intel’s Arc A750, each in value in addition to efficiency. Nonetheless, whereas trimming the reminiscence capability could cut back the worth relative to the 12GB card, it’ll additionally drop efficiency.
Our greatest guess is that Nvidia in all probability has a great deal of GA106 graphics processors, a few of which can not have three totally useful 64-bit GDDR6 controllers however which nonetheless have practically all the potential 30 Streaming Multiprocessors (SMs) accessible. This seems to be like a strategy to attempt to bump up the worth relative to the 3050 whereas offering a modest increase to efficiency.
Nvidia’s aim will probably be to make use of any and all alternatives to dump as many Ampere GPUs as it could actually within the subsequent a number of months in a bid to clear the street for its next-generation GPUs based mostly on the Ada Lovelace structure. We won’t assist however suppose it will discover extra success if it minimize costs, slightly than making an attempt to fill the hole between the present 3050 8GB and 3060 12GB playing cards.