We previously wrote about tinybox. The $15,000 AI server system is powered by AMD Radeon RX 7900 XTX graphics cards and can reportedly deliver 37% of Nvidia H100 compute performance.
However, it appears that the makers of tinybox have encountered issues with bugs affecting the Radeon-based platform. After parent company Tiny Corp posted several tweets expressing frustration with AMD’s AI acceleration toolkit – in which it brazenly tagged AMD rivals Nvidia and Intel – AMD’s CEO, Lisa Su, intervened and said her team was working to fix the issues unload.
Unfortunately, the solutions weren’t good enough for the small company, which fired off another round of frustrated tweets, asking AMD to “fix their basic problems” and suggesting the tech giant make its firmware open source so the small startup could do what AMD seemed unable to – namely “fix their LLVM leak bug and write a fuzzer for HSA”.
If AMD makes their firmware open source, I’ll fix their LLVM leak bug and write a fuzzer for HSA. Otherwise, it’s not worth putting a lot of effort into fixing bugs on a platform you don’t own. https://t.co/c4I2So27YGMarch 5, 2024
The dilemma
After things got even more heated, Su tweeted: “Thanks for the cooperation and feedback. We are all about providing you with a good solution. The team is working on it.”
While that could be good news for small businesses – only time will tell – Su could face a backlash for essentially stepping in to support the use of its consumer products in an enterprise-facing server.
As jlake3 pointed out Tom’s hardware, “tinybox buys consumer cards instead of data center models, but seems to expect a data center SLA? They got a revised firmware within six hours of the first linked tweet and a call to tech the next day, which is more than I would think a startup buying a bunch of gaming cards would qualify for if there were actually paying enterprise customers are, and she seems to be having a public meltdown that AMD isn’t doing more for a startup with fewer than 100 servers built and none sold yet (and pointedly avoiding using professional GPUs).
He also made another good point. “As for Nvidia, their EULA prohibits the use of GeForce products for data center CUDA applications, so they certainly wouldn’t direct anyone to talk to tinybox in this situation, except perhaps an attorney.”