A startup hailing from Finland, Flow Computing, is claiming to revolutionize the world of silicon engineering. With their proprietary companion chip, they promise to double the performance of any CPU, and even increase it by up to 100 times with software adjustments. This groundbreaking technology could potentially help the industry keep up with the ever-increasing demand for computing power from the creators of AI.
This ambitious venture is a spinout of VTT, a research organization backed by the Finnish government, akin to a national laboratory. The Parallel Processing Unit (PPU) that they are commercializing is a result of extensive research conducted at VTT. While VTT is a primary investor, Flow owns the intellectual property.
“The claim, Flow is first to admit, is laughable on its face. You can’t just magically squeeze extra performance out of CPUs across architectures and code bases. If so, Intel or AMD or whoever would have done it years ago.”
Despite this admission, Flow has been diligently pursuing something that has theoretically been possible but hasn’t been achieved yet. CPUs have come a long way since the days of vacuum tubes and punch cards, but they still have some fundamental limitations. They are single-threaded processors, meaning they can only handle one task at a time. Although they can switch between tasks a billion times per second across various cores and pathways, this is simply a workaround for their inherent limitations. In contrast, GPUs, or graphics processing units, are specialized in performing multiple calculations simultaneously.
Flow’s co-founder and CEO, Timo Valtonen, acknowledges this limitation, stating, “The CPU is the weakest link in computing. It’s not up to its task, and this will need to change.” While CPUs have become incredibly fast, there’s still a significant amount of inefficiency due to the sequential nature of their processing. The PPU developed by Flow aims to remove this limitation, essentially transforming the CPU from a one-lane street into a multi-lane highway. The CPU can still only perform one task at a time, but the PPU manages nanosecond-scale traffic on-die, enabling tasks to move in and out of the processor at a much faster rate than previously possible.
- “This is a fundamental game-changer that could revolutionize the industry,” says Valtonen. “It could potentially double the performance of CPUs and accelerate them by up to 100 times.”
Valtonen further explains this concept with an analogy, “Think of the CPU as a chef working in a kitchen. The chef can only work so fast, but what if they had a superhuman assistant who could swap tools and ingredients in and out of their hands, removing all tasks that don’t require the chef’s skills? The chef still has the same capability, but they can now work ten times faster.” While not a perfect analogy, it gives a general idea of the innovation Flow has achieved.
The company’s research and demonstrations have captured the attention of industry players, with Flow engaging in talks with several major companies. The PPU doesn’t increase the clock frequency or push the system in ways that would create extra heat or power consumption. In other words, it’s not asking the chef to chop twice as fast, but rather using the CPU’s existing cycles more efficiently.
However, as Valtonen points out, “This type of technology isn’t entirely new. It has been studied and debated in high-level academia. The issue has always been that implementing it required rewriting all the existing code, making it practically useless.”
The PPU also provides significant performance gains when combined with code that has been modified to take advantage of its technology. Flow has seen improvements of up to 100 times with their technology, and they are working on developing tools that make it easier for software makers to optimize their code for Flow-enabled chips.
Kevin Krewell, an analyst from Tirias Research who was briefed on Flow’s technology and offers an external perspective states, “The biggest market right now is AI acceleration, which is achievable with specialized silicon like Nvidia’s H100. While a PPU-accelerated CPU would have advantages across the board, chipmakers may not want to deviate from current trends. Additionally, it’s a significant investment of resources to implement Flow’s technology, with potentially substantial impacts on their existing five-year plans.”
Will Flow’s technology become an essential component for every chipmaker out there, catapulting them to success? Or will they face challenges from those hesitant to invest in unproven technology? Like all startups, their future will depend on their customers’ acceptance.
In conclusion, Flow has recently come out of stealth mode, securing €4 million (about $4.3 million) in pre-seed funding from investors like Butterfly Ventures, FOV Ventures, Sarsia, Stephen Industries, Superhero Capital, and Business Finland.
Image Credits: Flow Computing
