New Gigabyte AI Top local AI training solution revealed at Computex 2024; makes local AI training possible on your own desk
Training large language models (LLMs) tends to be an extremely costly and technically challenging affair that is fraught with security risks but Gigabyte has revealed a novel solution dubbed as Gigabyte AI Top that enables companies to locally train their AI models, potentially nipping all these problems in the bud while also making AI training accessible to smaller or less well-heeled organisations and people.
What is Gigabyte AI Top
Revealed in a gala launch at Computex 2024, Gigabyte AI Top is an all-round AI model training solution that consists of a conceptual triumvirate of intermeshing technologies. In essence AI Top is built primarily around an AI Top Utility which offers an informative yet intuitive user interface and acts as the means to train AI models.
The AI Top Utility itself has been tweaked to be easy to use for novices and experts alike, supports 236B LLM local training with a vast degree of customisability and control to ensure optimised outcomes. It also hosts an intuitive setup and dashboard to modify parameters without the need for coding and can work off a standard electrical system without the need for an upgraded grid. Needless to say, AI Top Utility’s ability to locally train an LLM locally means that it is both private and secure. Of note is that AI Top Utility works with most popular LLMs like GPT-2, StableBeluga2, Gemma, Llama2, Deepseek, Google’s Bert and more.
When you get stuck or encounter a problem, an AI Top Tutor solution allows users to troubleshoot, help with setting up an AI training programme and offer a modicum of technical support short of a physical help desk.
Last but not least, Gigabyte is introducing a range of AI-training optimised AI Top hardware solutions ranging from GPUs to PSUs that are more upgradable and versatile as well as being able to offer more affordable training compared to the usual training solutions on the cloud.
What are Gigabyte AI Top hardware solutions
Where Gigabyte AI Top Utility and AI Top Tutor are software solutions, Gigabyte also launched AI-training optimised CPUs, GPUs, memory and storage solutions to create a full spectrum solution that encompasses both hardware and software.
At the launch, Gigabyte unveiled the TRX50 AI Top motherboard which offers a whopping 8 R-Dimm slots for up to 2TB DDR5 RAM, space for 4 SSDs in Raid 0 configuration, up to 4 GPUS via 16 PXIe 5.0 slots and dual 10GbE LAN with 25 KV surge protection.
Gigabyte also revealed the W790 AI Top motherboard that also has 8 R-Dimm slots for up to 2TB DDR5 RAM, up to 6 SSDs via the provision of 4 PCIe 5.0 slots and 2 PCIe 4.0 slots, the ability to mount up to 4 GPUs and two Thunderbolt 4 ports too.
In terms of AI Top GPUs, Gigabyte revealed the RTX 4080 Ti Super AI Top 16G GPU which has a single PSU and is designed to be ideal as part of a 4-GPU solution, features an all-copper heatsink, has optimal airflow via the built-in turbo fan and support for up to 64GB VRAM.
Gigabyte is also leveraging their deep collaboration with AMD to create the AMD W7900 AI Top 48G and 32G which offer up to 192GB vRAM with the option to mount up to 4 GPUs for AI model training and has optimal airflow design.
Also unveiled at the Gigabyte launch was their AI Top 100E SSD which comes in 1TB and 2TB sizes, offers up to 7200/6,500 MB/s sequential read/write speeds as well as 150W TBW making it optimised for AI Model Training.
The brand also revealed the Ai Top UD1600PM PG5 power supply with double the lifespan of a conventional model, the ability to support up to 4 GPUs for AI model training and has an 80 Plus Platinum rating for power efficiency. 
For more on Gigabyte’s AI Top solutions and hardware, check out their official page at https://www.gigabyte.com/WebPage/1079/index.html If you’re looking for something more unusual and entertaining, the Gigabyte AI New Era beckons with its fun yet educational blend of art installations that also meld Aorus and Gigabyte AI PCs in new, interactive ways. In the meantime, stay tuned for more news and updates from Computex 2024!