Harness AI for design and art using your very own workstation

Coming up with ideas is one thing, but executing can be a time-consuming chore. While such an exercise is a labour of love for artists, when it comes to pitching potential ideas to potential clients, generating images to express concepts that can’t all be chosen down the line, can be disheartening, mentally exhausting and even soul destroying. This is just one area where generative Ai can help you.

Of course, getting into Ai can present some issues. It’s simple enough to ask ChatGPT a question or ask an app like Photoshop to make a rough-and-ready picture but, after that, the learning curve points north while experienced practitioners start running into cost, hardware, privacy and intellectual property problems that require more than a basic PC or expensive cloud service. And here’s where Gigabyte’s new AI TOP technologies will help you.

A complete ecosystem

AI TOP is an ecosystem spread across three areas: the AI TOP Utility, the AI TOP Tutor and a range of AI TOP hardware. The possibilities offered by AI art were recently demonstrated by world-class artists at an exhibition at the Computex trade show called, “AI New Era: Humanity X Art X Technology” which showed-off new media installations and artworks using the new toolsets and capabilities afforded by Ai tech.

It included a ‘living’ animated work by the Dimension + new media art team, led by Taiwan’s Escher Tsai and Hong Kong’s Keith Lam, titled ‘Ecological Pool.’ This included a demonstration where the artists showed how changing a few parameters could change the appearance and behaviour of virtual, complex, primordial beings.

Also in attendance was Brazil’s Ygor Marotta, from the VJ Suave studio in São Paulo, who is known for ‘projecting love onto city corners using tricycles’ and who created the highly colourful and high contrast ‘microdosys.’ This is an ongoing, long-term, Ai art creation project that uses code, text and imagination to create a series of Ai paintings and Ai animations.

Another key figure was emerging Taiwanese new media artist, Tim Wei, whose work, ‘Chaos Grammar β’ generated 3D models, characters and sounds in real-time based upon text inputs. Participants could freely type text on a keyboard and the results, generated by Ai, would float and collide arbitrarily in a virtual, 3D arena.

In addition to displaying Ai-generated artworks there was a ‘Vs Ai Street Battle’ interactive arcade game installation plus demonstrations from the artists of how they’d used machine learning applications to create and modify static and moving artwork on the fly, in real time.

However, other demonstrations showed how old black and white photos could be intelligently colourised in a single click and how a laptop and a webcam could transform visitors into superheroes as they watched themselves, live.

Hardware designed for Ai

Everything was created and operated using Gigabyte Ai PC hardware which is available in both component and pre-built PC forms. The former are engineered with Gigabyte’s Ultra Durable engineering technologies and come with extended warranties. This is particularly useful as Ai workloads can take anything from seconds to thousands of hours depending on complexity.

The components go beyond standard PC devices without straying into expensive, certified workstation territory, in order to remain easily configurable and upgradable. AI TOP Dual-slot graphics cards, from Nvidia and (uniquely) AMD’s Radeon Pro series, are optimised to run continuously in quad-card systems with their solid-copper heatsinks, special airflow channels and turbo fans.

The AI TOP motherboards can take four GPUs, up-to six SSDs, handle up-to 2TB RAM, use embedded Thunderbolt 4 ports and have dual, 10GbE LAN sockets.

The AI TOP SSDs can handle 125 times more writes and rewrites (TDWs) than standard SSDs.

The AI TOP power supply uses server-grade components to supply reliable and efficient power for twice the lifespan of competitor PSUs while still only requiring a single wall-outlet connection.

You can get going now

However, Gigabyte also offers pre-built systems to suit different industries and different price points.

The entry-level model is aimed at Post-grad students, YouTube Live streamers and those with personal studios who want to create models with eight billion (up to 30B) parameters. It comes with two Nvidia GeForce RTX 4070 Ti Super cards and takes 288 hours to train a 100K-parameter sample.

The intermediate system is aimed at SMBs, personal studios and educational institutes. It is ideal for 13B-to-70+B model sizes and can perform the same task in 32 hours thanks to its dual Nvidia RTX 4090 cards.

The next system makes use of four Nvidia 4070 cards and is aimed at larger studios, research institutes and medium-sized businesses which seek to create 30B-to-110+B-parameter models that can perform a similar task in 17 hours.

At the top end is the model for enterprise and vertical industry usage that’s ideal for medical, financial, engineering and geometry tasks and uses four W7900 AI TOP GPUs to support LLMs up-to 236B parameters in size. It can perform a 100,000-strong workload in 15 hours.

By using such hardware locally, creators can ensure that their IP is secure and not being compromised in a public cloud. Also, at 2.5-tokens per Watt, Gigabyte AI TOP hardware is significantly cheaper and more efficient to operate than on standard, unoptimised PCs which typically manage under half-a-token per Watt and much cheaper than the cost of cloud.

So, whether you’re an expert or a novice, Gigabyte’s AI TOP tech can make making Ai simple and efficient whatever your budget. The output can save countless hours of painful donkeywork when pitching while simultaneously opening-up a whole new creative toolbox that your peers and competitors won’t yet have. To find out more click here.