Well, it looks like Micron has officially started Mass production of next generation GDDR6X memory chips with speeds up to 24Gbps.
Micron’s 24Gbps GDDR6X memory enters mass production and will power NVIDIA’s next-generation GPUs later this year
micron publication Earlier this year, they were preparing new 24Gbps GDDR6X memory modules for future graphics cards. The announcement came as NVIDIA’s GeForce RTX 3090 Ti was entering the market where Micron is making 16Gb DRAM modules with speeds up to 21Gbps.
Micron now offers faster delivery speeds of up to 24Gbps at 16Gbps. This means I’m looking at 2GB for video and up to 24GB with a 384-bit bus interface. Like the GeForce RTX 3090 Ti, which houses all memory modules on the front of the PCB, Next-gen GeForce RTX 4090 With 12 GDDR6X modules on the front, it allows for more cooling compared to solutions with memory modules on the back of the PCB such as the RTX 3090 (Non-Ti).
The NVIDIA GeForce RTX 4090 (rumored to offer up to 21Gbps die speed) isn’t expected to use the full 24Gbps speeds, but the 24Gbps GDDR6X memory modules will deliver up to 1,152 TB/s bandwidth can provide 1,008 TB/s is available with today’s 21GDDR6 memory modules.
Here are the bandwidth numbers you can expect from a 24Gbps DRAM solution:
- 512-bit solution – 1.5TB/sec
- 384-bit solution – 1.1TB/sec
- 320-bit solution – 960GB/sec
- 256-bit solution – 768GB/sec
- 192-bit solution – 576GB/sec
- 128-bit solution – 384GB/sec
- 92-bit solution – 276GB/sec
- 64-bit solution – 192GB/sec
The NVIDIA GeForce RTX 4090 will be one of the first next-gen graphics cards to use the latest GDDR6X memory modules launching later this year. Given that these chips are likely to be cut back to hit their power targets, there’s a huge amount of overclocking potential available to enthusiasts.
AMD, on the other hand, could rely on its partner Samsung for its next-generation RDNA 3 lineup.samsung too Works with 24Gbps GDDR6 memory It is expected to reach mass production soon.
| graphic memory | GDDR5X | GDDR6 | GDDR6X |
|---|---|---|---|
| Workload | graphic | AI inference accelerator | AI inference accelerator |
| Articles of incorporation (example) | titanic | Titan RTX RX5700XT |
GeForce® RTX™ 3090Ti GeForce® RTX™ 3080Ti |
| number of positions | 12 | 12 | 12 |
| gigabytes/sec/pin | 11.4 | 14-16 | 19-24 |
| GB/s/mode | 45 | 56-64 | 76-96 |
| GB/s/system | 547 | 672-768 | 912-1152 |
| Configuration (example) | 384 (12pcs x 32 IO packs) |
384 (12pcs x 32 IO packs) |
384 (12pcs x 32 IO packs) |
| framebuffer in the model system | 12GB | 12GB | 24GB |
| Average device power (pJ/bit) | 8.0 | 7.5 | 7.25 |
| General I/O channel | PCB (P2P SM) |
PCB (P2P SM) |
PCB (P2P SM) |
News source: Harukazi 5719
Products introduced in this article
READ GoldenEye007’s revival may be near
–>