In computer graphics too much is never enough and in the case of memory speed, that is even more true. Graphics lives and dies by memory speed and size. Micron just broke the speed limit and took GDDR6 from 768 GB/s up to 1 TB/s with their new GDDR6X
The new Micron’s GDDR6X graphics memory doubles input/output (I/O) performance while minimizing the cost of memory by enabling multi-level signaling in the form of four-level pulse amplitude modulation (PAM4). PAM4 multilevel signaling technique helps transfer data much faster, doubles the I/O rate, increasing the capability of each memory die from 64 GB/s to 84 GB/s. Micron says their GDDR6X memory dies are also the only graphics DRAM that can be mass-produced while featuring PAM4 signaling. And, Micron's GDDR6X needs 15% less power per transferred bit compared to the previous generation GDDR6 memory.
What is interesting is that Micron says that its GDDR6X memory can hit speeds of up to 21 Gbps whereas we have only seen 19.5 Gbps in action on the GeForce RTX 3090. It is likely that AIBs could utilize higher binned dies as they are available. Micron also confirms that they plan to offer speeds higher than 21 GB/s moving in 2021 but we will have to wait and see whether any cards will utilize them.
The full details of the various generations of GDDR are shown in the following table.
|Density||From 512 MB to 8 GB||8 GB||8 GB, 16 GB||8 GB, 16 GB|
|VDD and VDDQ||Either 1.5 V or 1.35 V||1.35 V||Either 1.35 V or 1.25 V||Either 1.35 V or 1.25 V|
|Data rates||Up to 8 GB/s||Up to 12 GB/s||Up to 16 GB/s||19 GB/s, 21 GB/s,|
|Access granularity||32 bytes||64 bytes|
|2× 32 bytes in pseudo 32B mode||2 ch x32 bytes||2 ch x32 bytes|
|Burst length||8||16/8||16||8 in PAM4 mode|
|16 in RDQS mode|
|Package||BGA-170 14 mm × 12 mm 0.8 mm ball pitch||BGA-190 14 mm × 12 mm 0.65 mm ball pitch||BGA-190 14 mm × 12 mm 0.65 mm ball pitch||BGA-190 14 mm × 12 mm 0.65 mm ball pitch|
|I/O width||x32/x16||x32/x16||2 ch x16/x8||2 ch x16/x8|
|Signal count||61 – 40 DQ, DBI, EDC – 15 CA – 6 CK, WCK||61 – 40 DQ, DBI, EDC – 15 CA – 6 CK, WCK||70 or 74 – 40 DQ, DBI, EDC – 24 CA – 6 or 10 CK, WCK||70 or 74 – 40 DQ, DBI, EDC – 24 CA – 6 or 10 CK, WCK|
|PLL, DCC||PLL||PLL||PLL, DCC||DCC|
|CRC||CRC-8||CRC-8||2× CRC-8||2× CRC-8|
|VREFD||External or internal per 2 bytes||Internal per byte||Internal per byte||Internal per byte|
|3 sub-receivers per pin|
|VREFC||External||External or Internal||External or Internal||External or Internal|
|Self-refresh (SRF)||Yes Temp. Controlled SRF||Yes Temp. Controlled SRF Hibernate SRF||Yes Temp. Controlled SRF Hibernate SRF VDDQ-off||Yes Temp. Controlled SRF Hibernate SRF VDDQ-off|
|Scan||SEN||IEEE 1149.1 (JTAG)||IEEE 1149.1 (JTAG)||IEEE 1149.1 (JTAG)|
|Feature comparison of Micron GDDR|
In the blog post, Micron suggests they could offer denser G6X capacities later. They mention that 8 GB dies will be available initially whereas 16 GB dies will be added in 2021.
Nvidia is the first to use this new super-fast GDDR6X in the new RTX30xx series.
If you want to dig deeper, Micron has a very thorough technote here.
GDDR6X is Micron’s extension to the JEDEC GDDR6 standard. The key differentiator is that with GDDR6X, Micron increased data rates using their novel approach of implementing PAM4 multi-level signaling in memory (DRAM I/O). This is not proprietary to Micron and the company sees PAM4 potentially being considered for other memory standards, too. Micron is promoting GDDR6X to other industry players who need high-performance, ultra-bandwidth solutions and are open to bringing this to JEDEC, too. In addition, when Micron launched GDDR5X, while it was not yet an officially published JEDEC standard, it had actually completed the JEDEC standardization process. In fact, Micron’s last GDDR5X collaboration with Nvidia for the GeForce GTX 1080 Ti GPU was so successful, it also drove the specification of the next JEDEC-standard GDDR6.