I was recently asked “Why is Renderstorm getting into the hardware business? There are so many other companies making hardware now. Besides, isn’t it a commodity business?” Said differently – why should you consider buying hardware from a specialty provider, instead of a large manufacturer, like Dell or HP?
And the answer is, simply put, if you are after the absolute latest release of hardware at the lowest possible purchase price, because you are just going to lease it and buy again in 18 months – you shouldn’t. If prices like $5,000 for a lower-end render server don’t scare you, and double that for the larger, more capable machines, then HP or Dell is the place to go.
But we see a totally different world, running one of the world’s largest high-performance custom-workflow render farms – a world where the latest gear gets you little more than 1-2% performance advantage over n-1 or n-2 generation hardware. Just take DDR3 memory as one example, which costs fractions of what you will spend on DDR4, required by lots of current gen processors – and performs at 95-97% of DDR4 when used in rendering workflows. Frankly, maxing out the memory in the server or workstation is far more important than using DDR4 instead of DDR3 – but you may have known that with your own testing.
Each component has a similar story, because “the devil is in the detail”, and in this particular case, it is true. There are 3 main types of rendering workloads – memory intensive, CPU intensive, and storage (or I/O) intensive. In fact, each software package and each plug-in treats the render server or workstation differently, and utilizes the compute assets in a slightly different way. We know this, as we work with them all and have at some point or another over the past few years, rendered very large jobs of each aforementioned type – CPU, storage, or memory intensive. Building the fastest machine that offers the most flexibility – high I/O capability, large memory capacity, fast dual CPU configurations, often is not about who can spend the most money and buy the latest and greatest hardware; it is about how to leverage the most cost-effective components in the right configuration, in volume, to address the computing task at hand.
And yes, cost-efficiency does matter. In fact, it is the paramount design point for us. We study the pricing across all the major components relative to their performance, and we never waste our dollars on the latest and greatest, when slightly older works just as well for 20% of the total cost. After all, your server sits in a rack and works all day – no one sees the CPU inside of it or the DDR4 memory that you just got talked into purchasing, unlike the hood ornament on a new Porsche sitting in the parking lot of a fashionable restaurant. And if our older ‘Mercedes’ works as well as the brand-new Porsche for a fraction of the total cost, isn’t that the point? Having run thousands of jobs of all types, I can promise you that our equipment will do exactly that – work just as well as anything you buy from one of the major server vendors.
Last but not least, there is the issue of service – for me the most important of all. When something goes awry, as it may, who do you call and how do they respond? When we build your machine, we make sure everything is tested, burned in, and running for hours before we ship. When you have a problem you call us directly, and we ship a replacement part or if you prefer, we can ship a replacement unit. You don’t have to sit on hold for hours and speak to people 6,000 miles from your facility, and explain the problem in tremendous technical detail. We understand your economic equation – if that machine isn’t working for you, you aren’t making money. That is something neither of us wants.
So as you consider your next hardware purchase, upgrade, or new render farm build-out, think about buying from a company that has custom-designed its own farm, for a fraction of the cost of the big cloud providers. You can benefit from that efficiency, from that expertise that comes after running tens of thousands of render jobs, and the service from people who will answer the phone personally when you call.