Home / PC Tech News / Nvidia releases RTX 2080 Super with reduced Samsung GDDR6 spec

Nvidia releases RTX 2080 Super with reduced Samsung GDDR6 spec

RTX 2080 SUPER released officially

In July, Nvidia officially launched the RTX SUPER series as an update over the standard RTX lineups. This was released before the AMD Radeon RX 5700 and RX 5700XT. AMD personnel did claim they rickrolled NVIDIA with the pricing changes just before its launches. Of Course, we could argue AMD made this MSRP price revision after NVIDIA RTX Super series release and performance capability is known. Allegedly, AMD discontinues a very young and confused Radeon VII graphics card. Now we have the RTX 2080 Super Series graphics card with an army of non-reference models behind it. Competition is good, it keeps the pricing lowered and incentives companies to provide for the same or lesser price.

Coming back to Nvidia, the following specs are the comparison:

Geforce RTX 2080 Super FE Geforce RTX 2080 FE
Architecture Turing
CUDA Cores 3072 2944
Base Clock 1650MHz 1510MHz
Boost Clock 1815MHz 1800MHz
Memory Type GDDR6
Memory Capacity 8GB
Memory Speed 15.5Gbps 14Gbps
Memory Bandwidth 496GB/s 448GB/s
Memory Bus Size 256-bit
NVLink Yes
MSRP $699 $799

The non-reference version is also available at the time of launch. At the time of writing, SCAN Computers listed non-reference graphics card from most of its manufacturers. But we should see the listing with retail pricing in online and brick-and-mortar stores

Nvidia underclocked its SAMSUNG GDDR6 modules

TDP seems to be an acceptable bump as we have updated SMs, CUDA Cores, RT Cores, Texture and higher memory bandwidth. But it is revealed that RTX 2080 SUPER’s memory performance is lesser than advertised when the paper specs are compared to SAMSUNG GDDR6’s module specifications.

According to SAMSUNG GDDR6 specifications, they are made to run at 16 Gbps, with NVIDIA listing its RTX 2080 SUPER’s memory speed at 15.5 Gbps. This is reported by ANANDTECH who reminds its reader base that SUPER series is nothing more than Turing refresh models, and NVIDIA may have not wanted to push the memory speeds too high. Nvidia wanted to have this implementation done easily on the existing PCB design of the RTX graphics cards- for both FE and non-non-reference models. The PCB of the original RTX cards is made to operate GDDR6 memory at 14Gbps. This is the best they can do without compromising anything on the card. The pre-test is likely to be redundant for those who manufacture non-reference PCB, just to be on the safe side. Its suffice to say, a true 16Gbps RTX cards would require a completely new PCB design.

This is fair enough because refresh models phase out the original units with certain upgrades. True to the words ‘refresh models’ it has to be made that can qualify for the existing architecture, manufacturing and assembly process.

Leave a Reply

Your email address will not be published. Required fields are marked *

*