The Silicon Review
A processor is an integral part of any electronic device. The new and old data center may have similar looks from a rack standpoint, but the inner components have radically changed. It’s a norm to use multicore CPUs that have dozens of cores and they are placed side by side with FPGAs, GPUs, and even custom accelerators. Processor speed affects the performance of a device strongly, faster processors can perform simple tasks very quickly and it also makes sophisticated applications quick enough for practical use.
Performance of conventional processors is not efficient and it is a cost burden for several businesses, to counter this Tachyum Inc. has come up with a measure to improve productivity and deliver efficient performance in various applications like machine learning, big data analytics, and large scale computing. Founded in 2016, Tachyum’s product was specifically designed as a replacement for the majority of the chips that are widely used and existing in several hyper-scale data centers, it offers significant improvements in server utilization, energy consumption, and space requirement.
The company has office locations in San Jose and Slovakia.
In conversation with Rado Danilak, Founder and CEO of Tachyum Inc.
Why was the company set up?
Tachyum was incorporated to commercialize a disruptive processor architecture to deliver ultra-high performance and ultra-low power operation across all disciplines of computing, including Data Center, AI, and High-Performance Computing (i.e. supercomputing) workloads.
What challenges did you face in your initial years?
As a startup, funding is always the most elusive challenge, with staffing running a close second. We tackled the funding challenge with self-funded hard work on an initial architecture design (getting our product development past the vaporware stage) and a near cycle-accurate simulation of the product. This work enabled us to validate the performance potential of our product, which in turn, enabled us to raise Pre-Series A and Series-A funding of over $30M.
Fostering a culture of feedback is crucial to the success of every organization. How is this true with your company?
Internally, Tachyum encourages frank discussion, feedback, and idea contributions from all levels of the organization. We actively seek out critiques of everything we do, because we believe “the more eyeballs, the better”. We have noticed that this type of open and empowered culture imbues each and every employee with a sense of “ownership” in the company, in ultimate product, and in the sub-projects which occur along the way to final product release.
About adaptability, how do you stay relevant to the consumer needs in this highly volatile market?
We dedicate significant effort to maintaining situational awareness of our customers, our investors, our competitors, and our own organization. Only by understanding the “ground truth” in these areas, management can be prepared to adapt and deliver excellence. Commonly used techniques include frequent dialogue with potential customers, usually based on sharing progress under a bilateral NDA. Other methods are active participation in relevant industry associations and attending pertinent trade shows within our industry.
How do you and your company contribute to the competitive global IT platform at large?
Tachyum’s Prodigy Universal processor provides a definitive solution to the most pernicious challenges faced by our customers. Without a breakthrough in processor power efficiency, in 10 years data centers will consume 8% of the planet’s electricity. This is not sustainable; therefore, Prodigy represents a path for the data center industry to increase their capacity by a factor of 10 while maintaining the same level of processing power consumption.
A second facet to our contribution to the global IT platform is derived from the fact that unused Prodigy processors can be powered up, to seamlessly create CAPEX free AI networks; delivering low cost / Big-AI on demand. Today, GPUs in data centers comprise approximately 6% of the total rack space. A Prodigy powered data center can deliver 10x more AI at much lower cost than the current approach.
What are the factors that make your brand most ‘innovative,’ and why?
In our primary market (hyper-scale data centers, with a ~$25B TAM / 15% CAGR for our processor product), performance, power, and cost are the most critical metrics to our customers. Tachyum’s Prodigy Universal Processor outperforms all of our competitors in normal Data Center workloads by a significant margin. Prodigy also outperforms our competitors in AI and Machine learning applications. With Prodigy, our customers can power up unused servers to handle AI/ML training tasks (typically addressed today by dedicated GPUs, which are expensive and difficult to program). Prodigy enables data center operators to deliver Big-AI to their customers, CAPEX free, using idle servers that are already bought & paid for.
Do you have any new products ready to be launched?
The company’s flagship product, code-named Prodigy, a 64-core “Universal Processor”, is currently in detailed physical design with tapeout to fabrication on track for end of 3Q20, with product sampling in 4Q20 and volume production in 2021.Prodigy development was undertaken to address two major global challenges: 1) the need for improvements in processor performance, which has plateaued for nearly a decade, and 2) the need to decrease data center power consumption, which currently consumes 2% of the planet’s electricity, and is doubling every 5 years.
As a question on sustainability, where do you see your company a couple of years from now?
Given the value proposition of our flagship product, Tachyum anticipates a revenue run rate approaching $1B within 3-4 years after product launch. Hyper-scale data centers are an increasingly commoditized product, with fierce competition in term of features and price. This, coupled with Prodigy’s compelling value proposition bodes well for a fairly steep revenue ramp. We intend to build a multi-billion dollar revenue company with a market cap exceeding $10B within the next 5 years.
“Rado” Danilak has over 25 years of industry experience and over 100 patents designing state-of-the-art processing systems. Rado was founder and CEO of Skyera, a supplier of ultra-dense solid-state storage systems, acquired by WD in 2014 and was co-founder and CTO of SandForce acquired by LSI in 2011 for $377M.At Wave Computing, Rado architected the 10GHz Processing Element of deep learning DPU. Rado pioneered enterprise and consumer MLC flash controllers and solved endurance limited by device physics. He was a chipset and GPU architect at nVidia, a CPU architect at Nishan Systems and Toshiba, and chief architect of 64b x86 CPU at Gizmo Tech.
Tachyum will unlock unprecedented performance, power efficiency, and cost advantages, to solve the most complex problems in big data analytics, deep learning, mobile, and large scale computing.