Nvidia's mid-range RTX 4070 is getting a new sub-variant featuring slower GDDR6 memory modules. Videocardz reports that Nvidia AIB partner Galax is preparing to launch a new aftermarket RTX 4070 dubbed the "OC 2X," featuring downgraded 20 Gbps GDDR6 memory modules. For extra clarification, this change only affects the RTX 4070, not the RTX 4070 Super, as far as we know.
The new GDDR6 RTX 4070 gets a new Ada Lovelace codename to accompany the memory change—AD104-251. This new AD104-251 RTX 4070 is slightly less potent than its (AD104-250) GDDR6X-equipped counterpart, featuring a measly 5% reduction in memory bandwidth. The original GDDR6X variant features 21 GBps modules, giving the original RTX 4070 504 GB/s of memory bandwidth. The new GDDR6 model reduces total memory bandwidth from 504 GB/s to 480 GB/s.
Beyond the switch from GDDR6X to GDDR6, the new AD104-251 RTX 4070 reportedly resembles its preceding twin, including the 192-bit memory interface and 5,888 CUDA cores. The only spec not verified is the TDP, which could increase or decrease by several watts.
| Specs: | RTX 4070 GDDR6X | RTX 4070 GDDR6 |
|---|---|---|
| Die Nominclature | AD104-250 | AD104-251 |
| Core Count | 5,888 | 5,888 |
| Memory Interface | 192-bit | 192-bit |
| Memory Type | 21 Gbps GDDR6X | 20 Gbps GDDR6 |
| Memory Bandwidth | 504 | 480 |
Nvidia has not officially disclosed why it has decided to make a GDDR6 variant of the RTX 4070. However, a rumor from late last month suggested that the availability of all RTX 40 series GPUs with GDDR6X memory modules would temporarily hiccup due to a quality control failure of Micron GDDR6X memory affecting GDDR6X production.
If this rumor is legit, introducing a GDDR6 RTX 4070 makes complete sense. This GDDR6 variant will help Nvidia overcome this memory shortage, ensuring that RTX 4070 production stays consistent until Micron gets its GDDR6X quality control situation sorted.
The good news is that this memory change is unlikely to affect the RTX 4070's performance. A minimal 5% memory bandwidth reduction and the RTX 4070's large L2 cache make memory bandwidth less critical, making this change even less relevant.
