1. Introduction2. Test Hardware3. Test Software Packages4. Summary5. View All Pages

From the advent of affordable internet in the 90s to the explosion of broadband services over the past 5 years, content has been getting better and better. From building designers to game developers, from training companies to broadcasting organisations, everyone is processing larger and larger media files and regularly sharing them with colleagues on the other side of the planet. With so many of us working in/around the digital content creation (DCC) zone, there seems a paucity of intelligent, accessible buying advice for professional systems. Today, we begin the first in a new series at KitGuru, which aims to give you a complete buying guide to workstations – beginning with the options available to you and the differences in the kind of load that each package places on your hardware. 

Given the applications you want to run and the type of data you are working with, where should you spend your hard-earned money?

When considering which hardware vendors to work with on this project, we remembered something from the PC Specialist ‘online system configurator' that impressed us a while back when we visited their old production centre. If you created a system that only draws 350w, but then opted for a 750w PSU, the configurator would flag the potential mis-match and ask if you were certain that you wanted the more expensive power supply unit.

PC Specialist wanted to make sure that you were not spending money unnecessarily. Which is exactly how we are approaching this guide – so PC Specialist was a natural choice. Given the lion's share of the workstation market is owned by Intel and NVIDIA, choosing Xeon/i7 processors and Quadro graphics cards built onto an Asus mainboard was also a simple decision.

What do we mean by ‘Spending money unnecessarily'?

If you have a modelling package that scales with the number of CUDA cores available, then it makes sense to buy as many CUDA cores as possible within your budget. But what if your package is sensitive only to the number of CPU cores available?  Then the focus of your spend should be to increase the amount of parallel processing available in your CPU.

While on the subject of the CPU, you need to consider how ‘multi-thread aware' your package is. There is no point in buying a 48-thread CPU if the performance of your package only scales with CPU clock speed. During our research, we found one specialist workstation company that regularly pairs expensive NVIDIA Quadro cards with Intel's highly overclockable Anniversary Edition G3258 for some customers, because a well-cooled G3258 running close to 4GHz can outperform a £1,000 Xeon at 2.6GHz in some professional applications.

This series will look at many different options and present you with a ton of data, as well as expert analysis on the options available and the choices we think you should make.

But first we want to give you some background on our hosts for this series, PC Specialist.

As the largest custom PC builder in the UK, PC Specialist has a wide range of customers. In recent times, the company has been getting more requests for business and professional systems than ever before. One area where PC Specialist’s reputation for quality control and customisation for performance comes into its own is in the area of workstations. For those unfamiliar with the company, let us merely say that the company has just had to relocate to larger premises for the third time in four years, and if you watch any of England's recent international friendlies, then you would have see a lot of PC Specialist branding around the football grounds.

The question on the table is, when buying a workstation for professional use, where should you spend your money? To help answer this question, KitGuru (KG) sat down for a chat with John Medley (JM), the head of sales and marketing for PC Specialist and the driving force behind the company’s winning formula for PC specifications.

KG: “Why this particular line up of systems?”

JM: “Every minute counts for professional users, so if we can help to increase the amount of work they can do in a day – we’re saving them valuable time and money. They need to be looking for powerful systems so, on the processor side, the choice is between Intel's Haswell-E and Xeon.”

KG: “What is the trade-off between these processors?”

JM: “Xeons have huge scalability in terms of the number of cores and threads available, plus some turbo capability. On the other hand, the Haswell-E processors offer a really stable overclocking platform, which is perfect for software applications that respond well to clock speed.”

KG: “Looking at all of the test rigs you have assembled at our request, each of these boxes looks very similar, can you tell our readers why?”

JM: “For the initial bank of benchmarks, KitGuru wanted to focus on the performance difference that the CPU and GPU can make in numerous different applications. By using the same core components (chassis, motherboard, power supply, CPU cooling, SSD and HDD) , we are deliberately removing a lot of variability between these systems so we can be confident in the results.”

Having checked carefully before we began, we can confirm that all three rigs supplied off the PC Specialist production line had vanilla set ups, with no factory overclocks and only a basic installation of the operating system before we started. KitGuru then made sure that all of the drivers and test software were installed.

Once we did a baseline benchmark test on the Intel Core i7 5820K system at stock speed, we then brought it up to 4.4GHz – in line with the kind of overclocking service that PC Specialist would offer a professional customer who requested an overclocked i7 workstation.

JM: “We’ve found the Core i7 5820K to be a very reliable processor and, with a high quality board like the Asus X99-E WS, it clocks to 4.4GHz on air with no stability issues.”

The cards made available ranged from the entry-level NVIDIA Quadro K620 through to the K5200.

KG: “Do all professional users choose Quadro?”

JM: “Not necessarily, no. Some software applications respond better to the more mainstream GTX Series and others prefer Quadro. This is why we aim to keep our workstation solutions as customisable as possible, so we can produce each system to specifically meet our customers' needs.” 

KG: “The memory on our initial array of workstation systems is standard, in the sense that it is not ECC (Error Correction Code), why is that?”

JM: “You will find some users that want to specify ECC memory, but we find that they tend to be server purchasers. We install a huge number of memory modules every year and for us, Kingston HyperX is the best solution – and our customers agree – both in terms of performance and reliability.”

Enough about the concept and hardware choices, let's move on to the benchmarking software.

This is a comprehensive list of the software we used, as well as our observations about what each package is testing for. We start with the standard system tests, including CrystalDiskMark and why we included CPU-Z screen grabs – and move onto the really serious tests like SPECapc.

Maxon Cinebench R15
This is a great initial performance test for a workstation, or indeed any PC (or Mac!). The two halves of the test simulate modelling and rendering performance. The modelling test uses OpenGL and is mostly relying on graphics performance; the rendering test is highly multi-threaded so takes full advantage of a multi-core CPU.

Stressed: GPU speed and cores; CPU cores; CPU clock speed

Un-Stressed: Nothing!

CPU-Z
We run CPU-Z so we can get a reading of the CPU and memory specification – and present that to our readers.

GPU-Z
What CPU-Z does for CPU and memory, GPU-Z does for graphics, allowing us to check the GPU specification and present that.

Crystal Disk Mark
We ran Crystal Disk Mark to show another aspect of the baseline performance. The drive on which the operating system and applications reside has very little effect on modelling or rendering performance. We ran this test for information only, although the results do provide a mildly interesting story.

Stressed: SSD or HDD

Un-Stressed: Everything else

SPECviewperf 12.02
This is the industry-standard test for workstation graphics, now in its 12th incarnation. It runs eight different viewsets that replicate a comprehensive selection of the kinds of software you might run on a 3D content creation workstation. Many of the viewsets are based on code from the most popular software in each respective area, such as Autodesk Maya.

Stressed: GPU speed and cores; some tests are affected slightly by CPU clock

Un-Stressed: CPU threads

Cadalyst Benchmark Test for Autodesk AutoCAD 2015
The popularity of AutoCAD for computer aided design led us to run the Cadalyst Benchmark Test for AutoCAD 2015. It uses a variety of 3D models to test 3D graphics, 2D graphics, disk, and CPU.

Stressed: GPU speed and cores; CPU clock speed

Un-Stressed: CPU threads

SPECapc for 3ds Max 2015
With a single iteration taking an hour, SPECapc is one of the most in-depth tests for 3D intensive applications. It uses huge array of modelling and rendering tasks to put GPU and CPU through a full workout.

Stressed: GPU speed and cores

Un-Stressed: CPU threads

So there you have it: The baseline to KitGuru's complete guide to buying the best workstation for your needs.

While we will be stressing ‘value for money' at every stage, that doesn't mean that there has been any compromise in the hardware chosen. In terms of pricing, the range of choices tested start with a Core i7 5820K system and an nVidia K620, which will set you back just over £1,000 – right through to a high end Xeon with a Quadro K5200 that will be pushing the £5,000 mark.

What we want to make sure is that, if you are in the market for a professional workstation (something that you will be using to do work that will be invoiced to a client), then you are certain that your budget has been spent in the right place. This is the kind of analysis that F1 teams do 365 days of the year. Within the parameters of weight and performance, where should we put our focus?

As the series develops, we will be putting more and more topics under the microscope as we present the most comprehensive set of workstation tests that we have ever assembled.

We hope you find the series enjoyable and a useful tool for your purchasing decisions. As we expand the series in the future, we're hoping that you find it an invaluable reference, but at the same time we are always keen to get feedback from our audience.

Let us know what you think over on Facebook, and read on for Part 2 where we perform some initial testing and draw some early conclusions.

KitGuru says: We are brand agnostic. The choices that we investigated with PC Specialist could just as easily have been investigated with a different system builder.  The important part is that the systems chosen formed a ‘Reference Platform' for our tests that remained constant in the two days that we were Up North. The exact results you get from another workstation manufacturer will be slightly different, but the ‘performance ratios' between each platform will remain similar.

Become a Patron!