The fastest desktop PC ever tested - but is it any good?

Armari showcases the latest Intel Xeons, making this the most powerful rendering workstation we’ve ever seen, says James Morris.

Just a decade ago, fast desktop computing was all about high clock speeds. But since then the number of cores in processors has risen on an annual basis. This issue, we get our first experience of the latest chip iteration: the new Intel Xeon E5, sporting a whopping 12 cores. We take a look at the new CPU in Armari’s flagship dual-socket workstation, the Magnetar M32-SW1200.

The Armari system comes equipped with a pair of Xeon E5-2697 v2 CPUs, which is currently the top of the dual-socket range from Intel. This uses Intel’s Ivy Bridge-EP core, rather than the Haswell core, as the latter has only been released as a uniprocessor quad-core so far. The E5-2697 v2 runs at a nominal 2.7GHz, and incorporates Hyper- Threading, so each 12-core processor is presented as 24 virtual cores. The processor also supports Intel Turbo Boost, allowing a single core to reach 3.5GHz, and all 12 cores to hit 3GHz.

Memory performance

Armari has partnered these powerful processors with a healthy 64GB of 1,866MHz DDR3 SDRAM in eight modules. These take up all the available slots, but you probably won’t need more than 64GB of RAM for some years to come.

The graphics are similarly high quality, in the form of Nvidia’s Quadro K5000. This ultra high-end card from the Kepler generation sports 1,536 CUDA cores, although these run at half the speed of the previous Fermi generation’s cores.

The Armari also boasts 173GB/sec of bandwidth to its sizeable 4GB of GDDR5 frame buffer. Despite packing in 24 physical cores and high-end professional graphics, the Armari is remarkably quiet in operation (after the initial boot sequence), and the custom water-cooling does a great job of dissipating processor heat, even under heavy load.

The Armari system rounds off the main specification with a 256GB Corsair Neutron solid state disk, with a 2TB Seagate Constellation CS ST2000NC001 conventional 7,200rpm hard disk handling general data storage.

Exceptional results

If any proof of how far Intel processors have come in the last decade is needed, it is found in the fact that Windows has some issues with systems having more than 32 threads, although it is possible to force most applications that can use over 32 threads to use everything available.

In the case of Maxon’s Cinebench R11.5 rendering test, this adjustment to run on 48 threads achieved the stunning result of 33.39 - 30 per cent faster than any other system we’ve tested. However, the OpenGL score of 62.68 is merely good, as modelling benefits more from clock speed than multiple cores. Similarly, the SPECviewperf 11 lightwave-01 result of 60.04 is a good one, but not exceptional, as is 94.92 in maya-03 and 52.33 in the SolidWorks sw-02 viewset.

The Armari Magnetar M32-SW1200 is the fastest workstation we have ever tested for rendering by a country mile, although a system with a faster core clock speed and the same graphics would provide more modelling performance. But if you want the most rendering power ever available in a single box, alongside very good modelling abilities, then the Amari Magnetar M32-SW1200 is ideal.

Verdict

Score: 8/10

Amari’s Magnetar M32-SW1200 is a supreme workstation for rendering, and very good for modelling, although you pay a sizeable premium for the new 12-core Intel Xeon processors.

Uppers

  • The most rendering power we have ever seen in a single workstation
  • Good modelling; very quiet

Downers

  • Expensive

James Morris has tracked the rise of every new development, from OpenGL accelerators to multiprocessor workstations, over 15 years of testing 3D content-creation hardware.

This article originally appeared in 3D World issue 177 - on sale now!

Liked this? Read these!

Topics

3D