by Why does AMD decide not to follow Nvidia with faster VRAM?
Updated December 7, 2022 4:42 p.m
The first graphics cards to use AMD’s new RDNA 3 architecture are the RX 7900 XTX and 7900 XT, which are part of the company’s next generation of graphics cards. By using a chiplet design, they are comparable to the company’s Ryzen processors.
People have been eagerly awaiting AMD’s upcoming line of graphics cards to see how they would stack up against Nvidia’s latest RTX 4000-series GPUs. The company’s flagship RX 7900 XTX is advertised as being up to 1.7x faster than the RX 6950 XT card it currently offers for 4K gaming.
It packs 24GB of GDDR6 memory and 96 compute units clocked at 2.3GHz. It also includes a 384-bit interface. AMD seems to be all about performance per watt here.
Although it draws more power than Nvidia’s base RTX 4080, AMD could have an efficiency advantage here with an output of just 350 watts versus the RTX 4090’s 450 watts.
The RX 7900 XT is a less powerful choice; It features 20GB of GDDR6 memory running at a slightly slower 320-bit and 84 compute units running at 2GHz, along with 300W of board power.
editor pick
Editor’s choice
Save $800
Samsung S95B 65″ QD OLED TV
With the S95B, Samsung introduces the latest QD-OLED TV technology – with impressive color accuracy, brightness and contrast. In addition to these properties, the S95B also has functions
Regarding not using the faster specification because it consumes less power, the manufacturer claims that it uses GDDR6 memory instead of GDDR6X for both cards. Both cards require two standard 8-pin connectors, so you don’t need to use any special connectors or adapters.
For the 4090, Nvidia used a new 12VHPWR connector, which has since faced a melting issue, an issue the company is still investigating. For this reason, AMD has discarded the more power-hungry high-speed modules. Instead, the company is aiming for efficiency gains with its new RDNA 3.
When is GDDR7 coming?
Compared to Micron’s current 24Gbps GDDR6X standard, the new 36Gbps GDDR7 standard offers potentially 50% faster rates compared to GDDR6X. The maximum GDDR7 bandwidth could reach 1.7 TB/s with a 384-bit bus.
Samsung Electronics, a leader in advanced semiconductor technology, unveiled a new generation of storage solutions and plans for the future at Samsung Tech Day 2022 to dramatically boost performance for data center, server, mobile, gaming and automotive applications.
Samsung is accelerating the 1b production process for the DRAM industry, enabling scaling beyond 10nm with innovations like High-K. The company will soon release DDR5 chips with 32GB density, which is a 2x improvement over the current 16GB chips and a 33 percent improvement over 24GB chips.
The use of 8.5Gbps LPDDR5X DRAM solutions for smartphones and ultrabooks is also expected to increase in the coming year. Development of customized DRAM products such as HBM-PIM, AXDIMM and CXL will continue to offer faster processing for neural networks and AI applications.
Faster RAM will benefit next-gen GPUs, the South Korean firm revealed GDDR7 specs with transfer rates up to 36 Gbps. Compared to Micron’s 24 Gbps GDDR6X chips used by some Nvidia RTX 4000 graphics cards on the market now.
This new standard aims to increase speed by 50%. In theory, GDDR7 can deliver 1.7 TB/s of bandwidth with a 384-bit bus, although a 256-bit bus can still surpass the 1 TB/s limit.
Samsung hinted that the new 36 Gbps GDDR7 could debut alongside upcoming RTX 5000 and RDNA 4 GPUs. However, a release date for the technology has not yet been set. In 2023, we’ll likely learn more about and see more of Samsung’s upcoming 36Gbps GDDR7 memory.
