Solutions from Hostkey are extremely multifarious, so that each customer can pick exactly what they need – at the most favorable prices. In particular, this provider offers a wide variety of dedicated servers, including a range of GPU servers.
For example, if your company, lab, or scientific institution deals with technical calculations or scientific researches that take advantage of machine learning, using GPU-based dedicated servers becomes almost a must. CPU-based servers perform times worse in such a case. So, the question is how to use GPU for machine learning optimally, rather than whether to use them or not.
Quite predictably, renting a top-notch server with the best GPU for machine learning is far from being cheap. To save a considerable sum of money without sacrificing performance, you should just choose the right configuration. So, what can you cut down on?
- First, the central processor unit needn’t be the latest one.
- Second, the volume of RAM of the server is not vitally significant: 16GB is enough in most cases; sometimes even 8GB could be quite ok. The rule of thumb is that it should be about twice the RAM on the video cards.
- Third, the volume of the SSD is not critically important as well.
- Fourth, when it comes to the number of video cards, you won’t actually need four or eight of them. Two will be sufficient in nearly all cases, that’s why dual GPU machine learning servers are so popular.
Although there are video card manufacturers other than nVidia, machine learning GPUs used most often, usually come from this brand. Some Tesla cards are a good fit for these tasks, but they are pretty expensive. It means that if you are after the most cost-effective solution, nVidia GeForce is what you need.
Hostkey offers you a great number of configurations with optimal price-quality ratio. If you are in search of the cheapest variant, you can even get along with a VDS (Virtual Dedicated Server) instead of a physical one. For instance, consider this one if your tasks aren’t time-critical: VDS: 1 x nVidia GeForce GTX 1080Ti, CPU 4 cores x 3.5GHz, 16Gb RAM, SSD 240Gb. It is capable of doing the job quite well and comes in handy when it doesn’t matter whether the particular job is going to take a second or a couple of minutes.
One more rather cost-effective option is the following: VDS: 1 x nVidia GeForce GTX 1080, CPU Xeon 4 cores x 2.6GHz, with 16Gb RAM and 240 GB SSD
The next variant is a dedicated physical server – a well-performing but still budget-friendly one, with the following configuration: 2 x GTX 1080Ti, Xeon E3-1230v6, 32Gb RAM, 480Gb SSD.
But if you want nothing less than the fastest-performing solution (e.g.for processing the constant flow of data online), the following option will definitely suit you: 2 x nVidia GeForce RTX 3080. CPU AMD Ryzen 9 3900X, with 64Gb RAM and 1024Gb SSD
Each of these options, in spite of their diversity, could be the best fit – everything depends on the requirements of a particular customer.
By the way, you may try any of the dedicated servers from Hostkey before renting. The online trial is completely free of charge.
As to the physical location of the servers, there are three Tier III data centers to choose from. NY1 is in New York City, the USA; the other one, named Serverius II, is located in the town of Meppel, not far from Amsterdam, the capital of the Netherlands. The third datacenter, DataPro, is in the Moscow region. All three facilities are secure, well-equipped and staffed with highly qualified personnel.
Should you face any difficulty in making the right decision, you are always welcome to directly contact Hostkey’s customer support service. You can email the support team, or get in touch with them via the live chat form on the website, or through messengers such as Telegram and WhatsApp.