Gigabyte shows it is going all-in on AI with new AI hardware and software at Computex 2024
It didn't showcase any notable or new gaming products at the show.
Gigabyte also showcased its new 32-inch OLED gaming monitor, which was announced at CES 2024.
At Computex 2024 in Taipei, Gigabyte launched its AI TOP brand in a bold move to enter the artificial intelligence (AI) hardware market. This new lineup includes a variety of components designed to meet the demands of AI applications, ranging from workstation motherboards and graphics cards to more-robust SSDs and even power supply unit (PSU) models. While Gigabyte’s AI TOP offerings cater to enterprise-level users, they also have solutions that are targeted at AI novices – for both AI training and inference.
The TRX 50 AI TOP motherboard is built for AMD’s Threadripper Pro 7000 series of processors.
The workstation motherboards unveiled include the TRX 50 AI TOP for AMD’s Threadripper Pro 7000 series but if you prefer an Intel platform there’s also the W790 AI TOP for Raptor Lake CPUs. Gigabyte says both are engineered to support intensive AI workloads, and along with the new graphics card range that features hardware from both NVIDIA and AMD camps, like the GeForce RTX 4070 Ti Super AI TOP 16G and the W7900 AI TOP 48G and W7800 AI TOP 32G respectively, are optimised for AI applications.
Gigabyte also introduced the AI TOP 100E SSD, available in 1TB and 2TB variants, designed for the rigorous demands of AI training with a jaw-dropping 219,000 TBW (Terabytes Written) rating. By comparison, a consumer-grade SSD like the Samsung 990 Pro has a 1,200 TBW rating. This high durability is crucial for AI applications that require constant data writing and reading, and Gigabyte claims that the drive is “built to withstand substantial data throughput”, making it suitable for continuous AI training operations.
Gigabyte's AI TOP series only offer the GeForce RTX 4070 Ti Super GPU, but not NVIDIA's higher-end models.
The 1600W power supply, UD1600PM PG5 AI TOP, is another critical component of Gigabyte's AI TOP range. Built with server-grade components, the 80 Plus Platinum-rated PSU ensures reliability and efficiency and is crucial for systems running up to four GPUs - necessary for intensive AI workloads that require significant power.
Perhaps the most interesting of the AI TOP series is a set of “Gigabyte-recommended” AI hardware that the company says is tailored to different model sizes and budgets. Ranging from US$6,499 to US$35,999 for the most powerful system, they are targeted at enterprise users who need off-the-shelf solutions for their AI training needs. One of Gigabyte’s selling points is that these systems are essentially built with the above-mentioned components and so are easy to assemble and upgrade, without any specialised parts or expertise required. These are not your regular gaming PCs.
The TRX 50 AI TOP motherboard can support up to four AMD W7900 AI TOP 48G cards.
What sets these AI TOP components apart from their more mainstream counterparts is their design for "ultra durability" specifically for AI training. Gigabyte has equipped these parts with extended warranties and higher-quality components and emphasizes their reliability and longevity under the heavy workloads typical of AI applications – and hence the higher price point too.
But if regular workstations are too low-key for you then Gigabyte's AI servers were another highlight at Computex. Equipped with NVIDIA's latest Blackwell architecture, the B100 and B200 that were announced earlier this year, these servers are engineered for generative AI accelerations. The B200 model, in particular, includes a direct liquid cooling system to manage its thermal output.
In addition to hardware, Gigabyte is also dipping its hands in AI software as part of its end-to-end solution services. It showed the AI TOP Utility, a local (not cloud-based) AI model trainer with a “user-friendly interface” that integrates several open-source AI models such as PyTorch and OpenAI. It aims to help AI enthusiasts with limited programming skills perform AI training tasks without getting into complex coding. I didn’t get to spend time trying it out, but Gigabyte demonstrated it in action and the interface does seem pretty streamlined. Perhaps more importantly, I think it provides beginners with a less intimidating platform to dabble with AI, before leveling up to using more advanced software.
Gigabyte will also offer four fully built workstations that range from US$6,499 to US$35,999.
All in all, Gigabyte’s AI-centric offerings are designed to provide a comprehensive solution for local AI training, integrating high-performance hardware with user-friendly software. Their approach aims to create a seamless experience for users, whether they are just starting with AI or are seasoned professionals.
However, this shift also seems to signal a significant change in Gigabyte's strategy, even if they are not admitting it so. Traditionally known for its consumer and Aorus-branded gaming hardware, Gigabyte is now pivoting towards the AI business, potentially moving away from its core audience of gamers and PC enthusiasts. This new direction was evident at Computex 2024, where their extensive AI hardware and software lineup seemed to overshadow their usual gaming and consumer products. By contrast, its key competitors like ASUS and MSI showcased new PC concepts, gaming handhelds and even next-generation gaming monitors at their respective booths.
Whether this ultimately means a full-swing transition to an AI-first company or simply a strategic move to capitalise on the current AI trend remains to be seen. Fingers crossed.
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.