Samsung showed off its GDDR7 VRAM chips at Nvidia’s GTC 2024 event. The South Korean manufacturer showed off modules with a capacity of 16 GbE, or 2 GbE, and a bandwidth of up to 32 Gb/s per pin.

From images of Samsung promotional materials, Which appeared on the HardwareLuxx websiteIt has been shown that GDDR7 modules are about 20 percent more efficient than GDDR6 modules, which operate at 1.35 volts. GDDR7 modules will operate at 1.1V. According to Samsung, this is also lower than the industry standard for GDDR7, which is 1.2V.

The chips are also said to generate 70 percent less thermal resistance compared to GDDR6. According to the company, this is due to optimization Circuit design. GDDR7 modules are also said to generate less heat high–Speed Operations. It’s not clear what testing methods Samsung used for these numbers.

claims technology website VideoCardz Samsung’s GDDR7 chips will also be used in Nvidia’s upcoming RTX50 GPUs. The modules in these video cards will also have lower bandwidth, according to previous rumors. according to Trusted Nvidia leaker @kopite7kimi The RTX 50 GPUs will have 28 Gbit/s GDDR7 modules, instead of 32 Gbit/s. Samsung and Nvidia have yet to confirm these claims.

Samsung announced its first generation of GDDR7 memory chips with a bandwidth of 32 Gbit/s per pin in the summer of 2023. At the beginning of this year, the South Korean company also reported the arrival of faster GDDR7 memory chips with a bandwidth of 37 Gbit/s per pin. SK hynix also offered a GDDR7 vram at the beginning of this year. The SK hynix was a more power efficient variant with a bandwidth of up to 35.4 Gbit/s.

GDDR7 specifications were previously announced. The standard was created at the beginning of this year. The memory type switches to PAM3 encryption for data transfer, among other things. This is in contrast to using pam2 encoding of GDDR6 or pam4 encoding in GDDR6X. With PAM3, three voltage levels are used to transmit the signal. This makes it possible to send more bits per clock cycle without having to increase the clock speed. According to rumors, both AMD and Nvidia are using the new memory type in their next generation video cards. The Nvidia GPU Technology Conference will be held March 18-21, 2024.