Intel Arc A770 (“Alchemist”): Developers observe that the memory clock is too low

On the GitHub developer platform, there are increasing reports of a memory clock that is too low for the fastest Alchemist graphics card, the Intel Arc A770 (test). In such cases, the affected developers report a memory clock of 2,000 MHz or 16.0 Gbps instead of the promised 2,187.5 MHz or 17.5 Gbps.

Intel Arc A770 Limited Edition runs at 16 Gbps only

The Intel GPU Community Issue Tracker (IGCIT), a project driven by the developer community on GitHub to collect, log and fix errors on Intel GPUs, lists a bug on the Intel Arc A770 (“Alchemist”) that leads to this It may be that the graphics card's GDDR6 graphics memory only runs at 16 Gbps.

In such cases, the memory of the affected Intel Arc A770 Limited Edition only reaches 2,000 MHz instead of the 2,187.5 MHz specified by the manufacturer, which would be necessary for a speed of 17.5 Gbps.

The error could be detected in this as well as in other cases via the Intel System Support Utility (SSU) and via GPU-Z in the latest version 2.50.0. The error therefore occurs when the graphics card is used properly with the latest graphics driver under Windows 11 2022.

• Application is not cracked, modded and use the latest patch
• Device is not undervolted nor overclocked
• Device is using the latest drivers

Intel GPU Community Issue Tracker (IGCIT)

Users and developers are also reporting this problem on the Intel Insiders Discord channel.

The Intel Arc A770 Limited Edition only clocks with a memory clock of 2 GHz (Image: IGCIT via GitHub)

Users affected can use the GitHub developer platform under the topic “Arc A770 VRAM speeds run at 2000 MHz” to debug or look for a solution.

Intel was already involved

The manufacturer has already been involved via the public Discord channel of the Intel Insiders Community. Intel itself has not yet commented on the problem. The project asks not to message, tag, or mention Intel's developers on GitHub.

Editors will update this article accordingly as soon as Intel has feedback or a fix on GitHub was found.

The editors would like to thank the community from the ComputerBase forum for the information on this message.