The Moore Threads MTT S8 is a general consumer graphics card designed and manufactured in China. This is precisely what makes it special, that it is a new step by that country to achieve its long-awaited technological independence from the West. A priori, its characteristics were promising, since this graphics card, based on the MUSA architecture and manufactured in the 12 nm node, has 4,096 shaders, 16 GB of graphics memory and supports PCIe Gen5.
Alas, those impressive specs they are nothing more than a mirage. A South Korean medium has had the opportunity to test the Moore Threads MTT S8, and the first thing they have highlighted is that the installation and configuration process is not easy, since this graphics card is only compatible with certain motherboards, and with certain monitors.
To these restrictions we must add, in addition, that it does not support the latest versions of Windows and that it is not compatible with many current games either, something understandable since, as if this were not enough, its support is limited, for the moment, to DirectX 9and it doesn’t support tessellation in games, a feature that was revolutionary at the time and that today is essential to avoid the archaic flat textures that were common in games more than 15 years ago.
In the performance test that the Moore Threads MTT S8 has carried out, it has faced two relatively old graphics cards, although still functional today, the GeForce GTX 1060 and the GeForce GT 1030. The former was a mid-range model and is still quite capable at 1080p today, while the latter was low-end and is unable to handle today’s demanding games smoothly.
The results that the Moore Threads MTT S8 has achieved are terribly bad, since in most cases performs less than half that of a GeForce GT 1030 and gets 252 wattswhile the NVIDIA graphics card consumes 30 watts. We could say that it is a driver issue, but the source clearly indicates that the MUSA architecture is very inefficient, and that it seems to have major shortcomings. Added to this poor performance is also a total lack of support in professional environments and the lack of AV1 decoding.
The truth is that it is a first step, and as such you have to understand that developing a GPU from scratch is not exactly easy. Now that Moore Threads has a base to work from, that R&D stage should begin to polish and improve this architecture, and we may see more competent products in the future, both in terms of performance and efficiency.