How did the 48G version of 4090/4090D come about?
Explore the new trends in the graphics card market and uncover the technological changes behind special graphics memory upgrades.
Core content:
1. The popularity of RTX 4090 48GB and RTX 4090D 48G graphics cards and their price increases
2. The origin of 48G graphics card memory and PCB board modification technology
3. The development of GDDR6X technology and its impact on graphics card performance
Recently, two graphics cards with special specifications - RTX 4090 48GB and RTX 4090D 48G are very popular. The notable feature of these two graphics cards is that the video memory capacity is directly doubled compared to the regular version , which is very suitable for DeepSeek's current taste. However, their specific source remains a mystery.
From a certain e-commerce platform, we can see that the RTX4090 48GB version and the RTX 4090D 48G version are both selling like hot cakes.
The RTX4090 48G version is priced at 24,599 yuan , which is 2,040 yuan more expensive than the 24G version at 22,559 yuan, and the price increase ratio is 9%.
The RTX4090D 48G version is priced at 21,999 yuan , which is 4,040 yuan more expensive than the 24G version at 17,959 yuan, a price increase of 22%!
|| Source of 48G
In a nutshell: modified PCB board . Of course, the seller also calls this customization.
In fact, these two graphics cards have been quietly circulating in the market since the end of June 24.
In the previous article , the 4090D is coming at 12999 yuan. Will you buy it? We introduced the parameters of the 4090D:
Normally, the standard RTX 4090D is equipped with 24GB of GDDR6X video memory , which consists of 12 chips with a capacity of 2GB each. In the field of hardware modification, it is not uncommon to double the video memory capacity by replacing large-capacity chips. However, it should be noted that there are currently no 4GB GDDR6X chips on the market.
Therefore, these two graphics cards with special specifications are not simple memory upgrades. It is very likely that the PCB (printed circuit board) has been redesigned and modified, and the same number of memory slots have been added on the back of the graphics card , with 24 and 16 2GB GDDR6X chips installed respectively, to double the memory capacity.
|| GDDR6X
GDDR, the full name of which is Graphics Double Data Rate, where "G" stands for "Graphics" , means graphics. It is a random access memory specially designed for graphics processing units (GPUs). GDDR technology is derived from DDR technology. However, with the rapid development of graphics card technology and the unique properties of GDDR technology itself, it has gradually opened up an independent development path of its own.
As a memory designed specifically for GPUs, GDDR has obvious differences from traditional DDR memories in terms of clock frequency, operating voltage, and storage capacity. Moreover, the working mode of the GPU is also very different from that of the CPU. The main task of the GPU is to perform large-scale concurrent calculations, such as graphics rendering, AI computing, etc. This requires that the video memory that matches it must have the characteristics of high bandwidth, high capacity, and high bit width. At the same time, the requirements for latency are not strict.
Given the urgent need for high bandwidth in video memory, GDDR technology has continued to evolve, leading to the development of half-generation technologies such as GDDR5X and GDDR6X.
For example, in the RTX 30 era, GDDR6X technology came into being and was introduced. Although the burst length of GDDR6X is only 8 bytes, which is only half of the burst length of GDDR6, it uses advanced PAM4 encoding technology (compared with the NRZ or binary encoding used by GDDR6, PAM4 encoding doubles the data transmission volume in each clock cycle), coupled with a higher operating frequency, successfully increasing the pin bandwidth from 14Gbps of GDDR6 to 21Gbps.
It is precisely thanks to this major technological breakthrough that the memory bandwidth of NVIDIA GeForce RTX 4090 can successfully break through the critical threshold of 1TB/s, providing solid support for its powerful graphics processing performance.
GDDR6X was initially jointly developed by NVIDIA and Micron. Generally speaking, GDDR6 is more common and widely used in low-end and mid-range graphics cards and some devices that do not require high video memory performance. GDDR6X is mainly used in high-end graphics cards, such as NVIDIA's RTX 3080, 3090 and 40 series.
