Powerful graphics and general workstation performance in a portable package for a reasonable price, says James Morris

PRICE: £1,401

SPECIFICATION

• 2.7GHz Intel Core i7 3740QM processor
• 16GB PC3-1600 DDR3 RAM
• Nvidia Quadro K3000M graphics with 2GB GDDR5 memory
• 240GB Corsair Force GS solid-state disk
• 750GB Western Digital Scorpio Black 7,200rpm SATA hard disk
• Gigabit Ethernet
• Windows 7 64-bit
• Two years premium collect and return warranty

MANUFACTURER: Scan http://3xs.scan.co.uk

Powerful mobile workstations normally cost a significant premium. If you want serious 3D content creation on the move, you have to be prepared to pay for it. Scan doesn’t seem to believe in this, however. At just over £1,400, the 3XS-MGW-10 doesn’t cost much more than a consumer-grade laptop. So what has Scan reduced to achieve this level of affordability?

Thankfully, the processor isn’t an area Scan has economised on. The MGW-10 is based around an Intel Core i7 3740QM, a quad-core CPU that runs at a nominal 2.7GHz. But Hyper-Threading and Turbo Boost 2.0 are on hand respectively to make eight virtual cores and to allow a single core to hit 3.7GHz. This powerful processor is partnered by a healthy 16GB of 1,600MHz DDR3 memory, although you can specify 32GB for just over £70 more.

More significantly, this is the first laptop I’ve reviewed to include the Kepler generation of Nvidia Quadros. In this case, it’s the mid-range K3000M that’s been included, and is in fact the only option for the 3XS-MGW-10. This sports a highly impressive 576 CUDA cores, which is a big step up from the previous-generation Quadro 4000M’s total of 336, although the CUDA units now run at GPU core speed rather than double, so the performance increase won’t be quite so big. The GPU is backed by a decent 2GB of GDDR5 frame buffer. Storage is generous too. The operating system and applications are installed on a 240GB Corsair Force GS solid-state disk, while general data is held on a 750GB Western Digital Scorpio Black conventional 7,200rpm hard disk. You could also specify an mSATA drive for main storage and two 2.5-inch disks. Even the Sony BC-5550H-01 optical drive is a cut above the average, adding Blu-ray reading to the DVD rewriting facilities (although this costs an extra £13). The 17-inch screen offers the 1,920×1,080 resolution you would expect at this size. The image is clear and colourful, and viewing angles are decent. However, this is still a standard laptop screen rather than the premium Super IPS unit found in the Dell Precision M6700.

VERY CAPABLE

All these quality components mean the 3XS-MGW-10 has impressive performance in the standard benchmarking applications. Its rendering score in Maxon Cinebench R11.5 of 7.06 is slightly behind the Dell Precision M6700, but still excellent for a laptop. The Cinebench OpenGL score of 60.81 is similarly slightly lower than the Dell, but capable nonetheless. The Nvidia graphics show their worth in SPECviewperf 11, however. The result of 60.39 in the lightwave-01 viewset is desktop calibre, and 62.72 in maya-03 is similarly capable. The SolidWorks sw-02 score of 38.83 shows this laptop will be equally at home with CAD product design as it is with 3D modelling and animation.

Despite the competitive price, there doesn’t really seem to be a weak point in the Scan 3XS-MGW-10. The screen can’t compete with the Dell Precision M6700’s, but otherwise the Scan offering is a close match, and notably better with some key 3D applications, for less than half the cost of the M6700. It also managed 137 minutes of intensive OpenGL animation on battery, which means it will give a useful couple of hours of work on the move. Overall, you get a huge amount of power for your money for modelling and even rendering with this laptop.

PROS

• Desktop-grade modelling
• Good rendering performance
• Great value

CONS

• The 17-inch screen could be better

The MGW-10 showcases the Kepler-generation Nvidia Quadro well, providing loads of graphics performance in an affordable package

RATING 4

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

Discover 35 free Photoshop brushes that every creative needs, over at Creative Bloq.

Nvidia’s new mid-range Quadro offers a huge amount of professional 3D acceleration for the money, says James Morris

PRICE: £384 / $599

OPERATING SYSTEM: Windows / Mac OS X

MAIN FEATURES

• 2GB GDDR5 RAM • DirectX 11
• OpenGL 4.3
• Shader Model 5.0
• 384 CUDA cores
• 2 x DVI-I
• Mini DisplayPort
• 4,096×2,160 resolution (DisplayPort 1.2)

MANUFACTURER: Nvidia

Nvidia’s latest graphics generation, code-named Kepler, arrived with less fanfare than expected. The first taste came in the shape of the Quadro K5000, supplied in Armari’s Magnetar M32-AW750R, which provided a pretty impressive performance. But only now do we have the official launch of the entire range, from entry-level upwards. The Quadro K2000D is arguably the most interesting card in the range. It occupies the lower end of the professional market, yet sports a specification similar to high-end cards from the previous generation.

The K2000D’s headline feature is how many CUDA cores it has. Where the previous Quadro 2000 had 192, and the Quadro FX 1800 before that had just 64, the K2000D has 384. This is actually more than the previous-generation Quadro 4000 and 5000, although the K2000D only offers 64GB per second memory bandwidth from its 2GB of GDDR5 frame buffer, where the Quadro 4000 offered 89.6GB per second. With its wealth of CUDA cores, the K2000D boasts 733 GFLOPS of single-precision performance – slightly more than the Quadro 5000 – and yet it still only draws 51W, one third of the power. However, as with the other members of the Nvidia Kepler generation, while the main GPU core runs at a faster 954MHz than previous models, the CUDA units now operate at the same frequency, rather than twice as much, so each one will be slower than before. Hence, the overall performance increase is not as dramatic as it could have been, and double-precision capability is actually decreased for the K5000 compared with the Quadro 5000, so I expect a similar situation for the K2000D.

PRODUCTIVITY BOOST

We tested the Nvidia Quadro K2000D in the Workstation Specialists WSX4-V2 workstation

The K2000D comes in two versions, depending on connection configuration: the K2000 sports two DisplayPort and one DVI connection, while the K2000D I tested opts for a single Mini DisplayPort plus two DVI connections. Whichever version you choose, up to four displays can be driven at once, and they are otherwise identical. There’s also Bezel Correction available if you want to create a single desktop from multiple monitors.

I tested the K2000D in a Workstation Specialists WSX4-V2 workstation, sporting an Intel Core i7 3770K and 16GB of RAM, making it an ideal partner for the budget-conscious card. The K2000D managed 60.12 in the OpenGL portion of Maxon Cinebench R11.5 – almost the same as the Quadro 4000-sporting Boston Venom 2300-7T.

The SPECviewperf 11 results were similarly stunning. The score of 94.42 in the lightwave-01 viewset is 50 per cent quicker – one of the best I have seen. The result of 85.31 in maya-03 is also amazing, with only the Quadro K5000- based systems I’ve tested being faster. In the SolidWorks sw-02 viewset the K2000D managed 54.63, which is again very competitive with previous-generation high-end cards.

Overall, the Nvidia Quadro K2000D is a pretty exciting card, and not just because it offers similar, if not better, performance than the previous high-end cards. Most significant of all is the price – at under £400, the K2000D is more than £200 cheaper than the Quadro 4000. This means that now even more keenly priced workstations can offer serious modelling performance, which is a very exciting proposition for productivity.

PROS

• Great OpenGL performance
• Keen price
• Low power consumption

CONS

• Kepler GPUs offer reduced double-precision performance

The Nvidia Quadro K2000D brings you greater performance than previous high-end cards at a sub-£400 price

RATING 5

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

Discover 20 top free 3D models at our sister site, Creative Bloq.

The Dell Precision M6700 provides true workstation-grade performance in a portable package, says James Morris

PRICE: £3,769

SPECIFICATION

• 2.9GHz Intel Core i7 3920XM processor
• 8GB PC3-1600 DDR3 RAM
• AMD FirePro 6000M graphics with 2GB GDDR5 RAM
• 512GB Samsung P830 solid state disk
• Matshita UJ252 BD-RE Blu-ray rewriter
• Gigabit Ethernet networking
• Windows 7 64-bit
• 3 years RTB parts and labour warranty

MANUFACTURER: Dell

Putting workstation-class performance into a portable package is a bit like squaring the circle – technically possible, but not exactly an easy problem to solve. Dell’s Precision M6700 mobile workstation definitely errs on the side of the ‘workstation’ half of that phrase, rather than the mobile half. But it’s still a system you could carry between locations, and even do some serious work on while on the move.

The M6700 is a solid 17in unit. The weight of 3.52kg is heavy by notebook standards, but it’ll fit into a large laptop bag. Inside, the components are resolutely desktop level. The processor is from Intel’s latest Ivy Bridge generation, and is the second-fastest mobile chip currently available. The Core i7 3920XM runs at a nominal 2.9GHz with four cores, but Turbo Boost 2.0 means a single core can operate at 3.8GHz, and Hyper-Threading provides eight virtual cores for improved rendering. Dell has partnered this top-end processor with an adequate if not outstanding 8GB of 1,600MHz DDR3 (You can upgrade to 32GB for an extra £240.)

However, the area where this M6700 sample makes some compromise is in the graphics. It’s possible to specify this notebook with Nvidia Kepler-generation graphics, but this costs a significant premium. So this unit comes with the entry-level AMD FirePro M6000 option. It sports 640 stream processors – around a third of the desktop FirePro W8000 – and 2GB GDDR5, which will provide good performance, but not up with the latest from Nvidia.

Storage, on the other hand, is top of the range. There’s a 512GB Samsung P830 solid-state disk provided for main storage, giving plenty of space and performance, alongside miserly power consumption. There’s a Matshita UJ 252 BD-RE Blu-ray rewriter included for removable storage, too. The M6700 offers a variety of 17.3in display options, starting at a resolution of 1,600×900. However, this unit has the most appropriate choice for the 3D content creator – a 1,920×1,080-pixel IPS panel with crisp detail, faithful colour and excellent viewing angles. There’s a 3D display available too, for even more outlay.

BENCHMARK TESTS

Performance is as fast as expected. The score of 7.2 in Maxon’s Cinebench R11.5 render test is the fastest I’ve seen from a notebook, and only about 40 per cent behind the quickest single-socket desktop workstations I’ve reviewed. You could definitely use the M6700 for a bit of rendering on the move.

The Cinebench R11.5 OpenGL score of 68.01 is also phenomenal for a portable. However, the Workstation Specialists WS-M1510 (issue 152, page 102) was slightly quicker. Similarly, the FirePro M6000 graphics can’t quite compete in SPECviewperf 11. The scores of 33.45 in lightwave-01 and 29.16 in the SolidWorks sw-02 viewset are good, but no match for the WS-M1510, and the result of 17.28 in maya-03 implies that one of the Nvidia upgrades could be worth the extra money if you run Maya primarily.

So the Dell Precision M6700 isn’t the most powerful mobile workstation I’ve tested, even if it could be with its top graphics upgrade option. It’s very capable, and lasted through 83 minutes of intensive OpenGL work running off the battery. The price isn’t exactly cheap compared with a similarly capable desktop workstation, but if you do need this class of power in a portable package, the M6700 is worth serious consideration.

VERDICT

PROS

• Powerful rendering for a portable
• Decent modelling
• Useful battery endurance

CONS

• Not the strongest graphics processor

If you need workstation-grade performance on the move, and you can afford the price premium, the M6700 is a high-quality option

RATING 4

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

Discover 30 free Photoshop brushes that every creative needs, over at Creative Bloq.

Does the debut of this real-time card and Maya plug-in mean the future has arrived for 3D artists? Antony Ward finds out

PRICE: $1,495

OTHER MODELS: Caustic Series2 R2100, $795

REQUIREMENTS

• Windows Vista or 7 (64-bit)
• Maya 2012 SP2 or Maya 2013 SP2

SPECIFICATION

• Two Caustic RTU (ray-tracing unit) chips
• 1 6GB on-board RAM
• Up to 100 million incoherent rays per second

DEVELOPER: Imagination Technologies

We’ve all been there. The deadline is looming and you’re at the stage where your scene needs its final tweaks. The shadows, lighting and shaders must be perfect before you produce your final render. The problem is, of course, each slight change results in a tortuous wait while that section or element re-renders.

There are obvious clunky workarounds, but this is supposed to be the future, isn’t it? Granted, the science world failed us on the promise of flying cars, hover-boards and free energy, but come on: where are our real-time ray-traced viewports?

The Visualizer for Maya plug-in dramatically upgrades the standard viewport

As it happens, Caustic Professional, part of the Imagination Technologies group, has recently come to our aid with the release of its suite of Caustic Series2 ray-tracing cards, with the accompanying Visualizer for Maya plug-in. (3ds Max support will be with us later in 2013.) With these tools, you may finally have what you’ve been hoping for: the ideal preview render, sitting neatly inside your viewport. Sounds too good to be true, right? Well, it gets better. On top of this, the viewport is still active, meaning you can continue to edit your scene in that viewport, with the ray-tracing preview updating on the fly at lightning speed.

The Series2 cards accomplish this by using their foundations in Imagination Technologies’ PowerVR OpenRL technology to take control of the ray-tracing and scene-geometry calculations. The shading of the scene objects is passed over to the CPU, unlike with most GPU cards, and textures are thrown into the system memory so that the Caustic cards are free to focus on each ray and polygon. With the R2500 card capable of handling up to 100 million rays, larger scenes are taken in their stride. While visiting Imagination Technologies, I took the opportunity to try some of my own work with the hardware. I know these scenes and I remember how much of a pain they were to set up and adjust, so what better test of the setup?

My ‘Serena the Genie’ scene loaded, and I activated the Visualizer for Maya viewport (which is as easy as switching to Viewport 2.0): in less than a minute, I was looking at a near render-quality version of Serena the Genie. This was impressive to see, but I also wanted to try it on my own ageing Dell T3400 in a real-world setting.

Change shaders and reflections on the fly, like the lenses in this demonstration model

By default, the Visualizer plug-in will runs through 10 passes of the viewport, starting with a rough interpretation and refining the image with each pass. With the hardware disabled, 10 passes of Serena took four minutes and 30 seconds to complete. Enabling the R2500 Series2 card reduced this to a mere one minute and 33 seconds. If you consider this factor over the duration of a project, the time saved could be immense, especially if you removed all those tedious test renders you would normally need to endure.

The Series2 boards come in two versions, the R2100 and the R2500, containing 4GB and 16GB of memory. The first is suited to single-processor machines, whereas its beefed-up brother is aimed at dual-processor systems.

With that in mind, you don’t initially need to buy one of these boards to experience the Vizualizer plug-in: you can buy it separately for $299, and a 30-day trial is available. Obviously, the Series2 cards offer a significant speed improvement, especially on larger scenes, but it doesn’t hurt to try before you buy.

CAVEAT EMPTOR

Before you empty your piggy bank and hit the online stores, there are a few things to consider. First, the software and drivers aren’t yet Windows 8 friendly, although the developers assure me that they are working on this and an update will soon be released. Second, when I say this is a fully ray-traced viewport, there are some limitations. Lighting, shadows, shaders and reflections look great, but character artists may miss the lack of subsurface scattering and nHair support. I also noticed that it seemed to struggle with a texture pathed into a ramp node. However, Caustic is listening to its users, and plans to continue to develop the software to incorporate more shaders, systems and effects in the future. It also hinted at the possibility of more application support: ZBrush and Mudbox, maybe? We can hope! Which reminds me to point out that the OpenRL SDK is currently available as a free download for anyone wanting to investigate support.

The R2100 only has 4GB of memory and a single chip, but it’s still a powerful model

So, should you buy one of these cards? Whether you’re a professional 3D artist or more of a hobbyist, downloading the plug-in is a no-brainer. It’s free for 30 days, which gives you plenty of time to decide whether to purchase a full licence, or boost your viewport’s speed with the addition of a Series2 card. If you’re a heavy renderer, though, this may just be the tool you’ve been looking for.

VERDICT

PROS

• Affordable real-time viewport ray-tracing
• Good support and future plans
• Could drastically improve your workflow

CONS

• No Windows 8 support yet
• Limited shader and dynamics support
• The plug-in alone is slow on old machines

With the Caustic Series2 Cards, Caustic brings real-time ray-tracing to the everyday artist at an affordable price

RATING 4

Antony Ward has been provoking pixels since the early 1990s. He has worked for some of today’s top studios

James Morris finds that this updated configuration is simply the most powerful modelling workstation we’ve ever tested

PRICE: £3,245

SPECIFICATION:

• 3.2GHz Intel Core i7-3930K processor (clocked to 4.3GHz)
• 32GB 1,600MHz DDR3 SDRAM
• Nvidia Quadro K5000 graphics with 4GB GDDR5 memory
• 240GB Intel 520 Series solid-state disk
• 2TB 7,200rpm Seagate Barracuda 7200.14 SATA hard disk
• Gigabit Ethernet networking
• Windows 7 64-bit
• Three years onsite warranty

MANUFACTURER: Workstation Specialists

All the excitement over Intel’s eight-core Xeon E5 processors has taken the focus away from more everyday, affordable workstations. A system with 32 virtual processor cores is a very powerful piece of hardware, but the price will be proportionate. If you primarily focus on modelling, a workstation based on the Intel Core i7 desktop processor can still make a lot more economic sense, particularly when it has been tweaked for extra performance, as is the case here.

CORE VALUES

In general, modelling benefits from higher clock speeds while rendering gets a boost from more cores. The WSX6 V2 comes with a single Intel Core i7 3930K. The top Xeons run at 3.1GHz, and even with Intel Turbo Boost only manage 3.8GHz from a single core. The Core i7 in the WSX6 V2 runs nominally at 3.2GHz, with a 3.8GHz Turbo Boost mode. But the WSX6 V2’s processor is permanently set to run at 4.3GHz when under load, which gives it a significant advantage over a Xeon. However this is still only a six-core CPU, even if Hyper-Threading turns it into 12 virtual cores, so it should still offer a decent amount of rendering power when needed. The processor has been partnered with a healthy 32GB of 1,600MHz DDR3 SDRAM, which comes in the shape of eight 4GB modules. This takes all the available slots, so would require replacing if you want to increase the memory.

Workstation Specialists hasn’t spared any expense in the graphics department. The latest Kepler-generation Nvidia Quadro K5000 is supplied, which I have already been highly impressed with in other systems. This sports 1,536 CUDA cores and 4GB of GDDR5 frame buffer, making it significantly more powerful than the Quadro 5000 it replaces – although only in single-precision calculations, as double-precision abilities are reduced.

Storage is similarly cutting-edge: a 240GB Intel Series 520 solid-state disk for operating system and applications, plus a 2TB conventional hard disk for general data in the shape of a 7,200rpm Seagate Barracuda 7200.14.

When it comes to performance, where Xeon E5-based workstations are more like jacks of all trades, the WSX6 V2 is very much the master of one. The Cinebench R11.5 rendering score of 12.47 is good, but the fastest Xeon alternatives are achieving twice that. Modelling is a different story, though: the Cinebench R11.5 OpenGL score of 99.81 is quite simply the highest I have ever seen, beating the next best score by 14 per cent. Similarly, the SPECviewperf 11 scores of 88.34 in lightwave-01, 134.18 in maya-03, and 73.17 in sw-02 are all significantly ahead of the fastest I have seen. The 4.3GHz clock speed and Quadro K5000 make this the most capable modelling workstation I’ve tested.

The WSX6 V2 gives you all-conquering modelling power and decent enough rendering for trying out ideas before the final output. If your focus is mainly on modelling, it’s the king of the hill.

PROS

• Fastest modelling performance to date
• Reasonable rendering performance
• Decent price

CONS

• Not suited to intensive offline rendering

The frequency-enhanced Intel Core i7 processor and Quadro K5000 graphics make this the most powerful workstation for modelling we’ve tested

RATING 5

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

Up your skills with these 25 Cinema 4D tutorials – over at Creative Bloq.

Streaming large volumes of media is becoming increasingly common. Rob Redman tests the G-RAID to see if it can keep up

PRICE: £450 / $600

OTHER EDITIONS: 8TB, £575 / $800

MAIN FEATURES:

• Dual-drive storage
• 7,200rpm
• Cache up to 32MB per drive

MANUFACTURER: G-Technology

You can buy several terabytes of storage for less than £200 these days, so why pay a major premium for this particular device? There are a few factors that make it stand out, and the first is that the G-RAID is a Thunderbolt-enabled drive.

Thunderbolt is an I/O interface for Apple computers, providing speeds that outperform all other interfaces. It’s eight times the speed of FireWire 800, and twice as fast as USB3. This speed makes it ideal as a media drive, because it’s possible to stream multiple video or audio files at once, which is a great ability for multi-cam editing or for creating an asset library for audio projects – or both combined.

I had four HD movies streaming to Final Cut Pro X, along with eight audio tracks that were all stored on the G-RAID, and there was no noticeable lag, which is pretty impressive. The only downside with Thunderbolt is that older Macs will need to use an adapter to connect the drive (available from Apple for £29).

Second, the device is a RAID array and would make a solid base for Time Machine backups, with some redundancy built in. I’ve had backups fail before, so moving to an array makes a lot of sense to me, and should be considered by anybody working in digital content creation.

The third selling point is its good looks: the G-RAID is solidly built and looks great (if you’re a fan of Apple’s design), sitting comfortably alongside any Mac from the last decade.

A well built, quiet and fast drive array, particularly well suited to digital content storage and local streaming

RATING 4

Rob Redman is a 3D artist and trainer, and is 3D World’s technical editor

Discover 20 top TV commercials at our sister site, Creative Bloq.

Whether you’re modelling or rendering, this workstation has power to spare. James Morris tries it out for size

PRICE: £4,995

SPECIFICATION:

• 2 x 2.6GHz Intel Xeon E5-2670 processors
• 32GB PC3-1600 DDR3 RAM
• Nvidia Quadro K5000 graphics with 4GB GDDR5 memory
• 240GB SSD 335 Series solid-state disk
• 2TB 7,200RPM SATA hard disk
• Gigabit Ethernet networking
• Windows 7 64-bit
• 3 years RTB parts and labour warranty

MANUFACTURER: InterPro www.ipworkstations.com

InterPro’s IPW-DX8 may have a rather unassuming appearance, but lurking inside its sober black chassis is an expansive range of cutting-edge technology. We have become used to the Intel Xeon E5 processor in recent times, and although it’s still the pinnacle of workstation performance, there are some even more recent components in here that I just have to shout about.

The core of the system is a pair of Intel Xeon E5 processors, in this case the 2670 edition. These run at 2.6GHz nominally, and have eight processing cores. However, Turbo Boost technology means that a single core can reach 3.3GHz, and multiple cores 3GHz when required. Hyper- Threading splits every physical core into two virtual ones, so the system presents a total of 32 virtual cores, which is great for parallel tasks such as rendering. These are partnered by 32GB of PC3-12800 DDR3 SDRAM, arranged in two banks of two 8GB DIMM modules each. This leaves 12 slots free for upgrade, up to the maximum of 512GB.

Even more significant are the graphics. This is only my second sighting of Nvidia’s latest Kepler generation of Quadro cards: again it comes in the form of the K5000 model, which ostensibly replaces the Quadro 5000. However, where the previous version offered 352 CUDA cores, the new K5000 upgrades this to a whopping 1,536, and ties this with 4GB of GDDR5 memory rather than the 2.5GB of its predecessor. So with this under the bonnet, it promises a hefty increase in performance, although this increase is tempered by a notable reduction in double-precision processing.

SOLID GEAR

Storage takes the now de-rigueur strategy of providing a solid-state disk for the operating system and applications, with a mechanical hard disk for the more general data. The former disk is a 240GB Intel SSD 335 Series. The SSD 335 Series is Intel’s first solid-state disk to feature 20-nanometer technology, allowing it to maintain a throughput of 500MB/ sec when reading and 450MB/sec when writing. The SSD 335 Series’ mechanical partner is a 2TB 7,200RPM SATA hard drive, which InterPro says will come from either Western Digital or Seagate, depending on stock. Either way, it’s a pretty adequate amount of secondary storage for most reasonably sized 3D or video assets.

I’m also pleased to report that performance is pretty impressive all round. The Xeon processors may not be the very top models, but the result of 22.41 in the Maxon Cinebench R11.5 rendering test is still very quick – weighing in at a credible 10 per cent behind the fastest systems I’ve seen.

The OpenGL score of 79.2 is similarly excellent, again with only a few systems faster. With Nvidia’s Quadro K5000 on board, the InterPro’s SPECviewperf scores were never going to disappoint. Highlights include a lightwave-01 viewset score of 67.04, maya-03 result of 109.28, and SolidWorks sw-02 score of 57.78. All are close to the best I have seen.

So the InterPro IPW-DX8 has phenomenal modelling performance available, and some very impressive rendering capabilities as well. The £4,995 price is also relatively reasonable, considering the high price of the graphics and Xeon processors. This is a super-powered workstation that will be equally at home with every stage of the 3D content creation process.

PROS

• Powerful rendering
• Supreme modelling performance

CONS

• Nvidia Quadro K5000 has reduced double-precision processing

With eight-core Xeon processors and the latest Nvidia Quadro K5000 graphics, this workstation will make short work of modelling and rendering 3D

RATING: 4

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

Discover 20 famous buildings that you have to see, over at Creative Bloq.

James Morris finds that the RS-D2850 is a cost-effective and compact way of providing network rendering for a small studio

PRICE: £3,195

SPECIFICATION:

• 2 x 2.2GHz Intel Xeon E5 2660 processors
• 32GB 1,600MHz DDR3 SDRAM
• 120GB PNY Prevail Series solid state disk
• Gigabit Ethernet networking
• Windows 7 64-bit
• Three-year on-site warranty

MANUFACTURER: Workstation Specialists

When I first looked at Workstation Specialists’ RS-D range in the shape of the RS-D2600, I rather liked the idea. This low-profile box is essentially a personal render farm that you can sit on top of your regular workstation tower. Once it’s set up, you can use it as a network render node, and everyone on the network can send jobs to it. The RS-D2850 is the most recent incarnation, sporting the latest Intel Xeon processors.

The benefit of this arrangement is that a render node doesn’t need expensive graphics, and you can scrimp on the storage too. Since a long render will have lots of parallel threads, you’re better off with more cores rather than a fast clock speed – unlike modelling – so you can make economies here too.

As a result, the RS-D2850 comes with the latest Xeon E5 processors, but not the top-of-the-range models. Instead, a pair of 2660 models are included. These still offer eight physical cores apiece, but run at 2.2GHz, where the top models run at 3.1GHz. Single cores can still hit 3GHz thanks to Turbo Boost, plus there’s 20MB of cache for each CPU, and Hyper-Threading means there are 16 virtual cores each, making a grand total of 32. As an added bonus, each Xeon has a 95W rating compared with 150W for the top Xeon E5, so you should save on power consumption too.

HEALTHY SPECIFICATION

Workstation Specialists has partnered the Xeons with 32GB of 1,600MHz DDR3 memory, supplied as eight 4GB modules. This leaves eight DIMM slots free for future upgrade, and the motherboard supports a whopping total of 512GB, so there’s plenty of room for growth. The other major component worth noting is the 120GB PNY Prevail Series solid-state disk, which houses the operating system and rendering software. This might seem a bit small, but assets would likely be sent over the network for rendering.

The RS-D2850 does have integrated graphics with a VGA connection, plus a couple of USB ports, so you can hook up keyboard and mouse to use it as a standalone system. But most of the time it would be accessed via the twin Gigabit Ethernet connections using a remote desktop or as a headless render node.

When it comes to performance, things have moved on apace compared with the RS-D2600. Where the latter managed just 14.11 in the rendering portion of Maxon Cinebench R11.5, the RS-D2850 leaps to 20.23. The top-end 3.1GHz Xeons are only about 25 per cent quicker than the 2.2GHz Xeons, so you get 75 per cent of the rendering power for a little over half the price. Of course, graphics co-processors like Nvidia’s Tesla offer an alternative to a whole dedicated render box, but can only benefit a single user at a time. You could kit out your studio with cheaper systems or workstation notebooks, and share one or two RS-D2850 units. For a smaller studio, the Workstation Specialists RS-D2850 could make a lot of economic sense.

PROS

• Top-end rendering performance
• Cheaper than a modelling workstation
• Compact size

CONS

• No modelling abilities

The RS-D2850 can’t take the place of a modelling workstation, but it can supply heavyweight rendering power to a network for a keen price

RATING: 4

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

ZBrush 4 R3 reviewed – over at our sister site, Creative Bloq.

The latest Venom provides all-conquering performance in every department – but you pay for the privilege, says James Morris

PRICE: £9,995

SPECIFICATION:

• 2 x 3.1GHz Intel Xeon E5-2687W processors
• 32GB PC3-1600 DDR3 RAM
• Nvidia Quadro K5000 graphics with 4GB GDDR5 memory
• Nvidia Tesla K20 CUDA co-processor with 5GB GDDR5 memory
• 256GB Crucial RealSSD solid-state disk
• 2 x 2TB 7,200rpm Western Digital Caviar RE4 SATA hard disk
• Gigabit Ethernet networking
• Windows 7 64-bit
• 3 years RTB parts and labour warranty

MANUFACTURER: Boston

The concept of using a GPU for things other than real-time 3D graphics processing has been around for a while, but it’s still far from widely utilised. One technology that is making the idea easier to harness is Nvidia Maximus, which helps you choose dynamically how you want to use your graphics hardware. Showcasing the latest incarnation is Boston’s Venom 2305-7T, an update of the Venom 2300-7T I tested in 3D World 161.

The graphics hardware in the new Venom has been upgraded to Nvidia’s latest Kepler generation, so it comes with the Quadro K5000 for real-time 3D plus a Tesla K20 co-processor. The K5000 is in theory the replacement for the Quadro 5000, but whereas the latter has 352 Cuda cores and 2.5GB GDDR5 memory, the Kepler upgrade increases the figures to 1,536 Cuda cores and 4GB. Similarly, the K20 compares very favourably with previous Tesla cards. The top Fermi generation Tesla M2090 sports 512 Cuda cores, but the K20 nearly quintuples this to 2,496. Although Kepler’s weaker double-precision floating point performance tempers this slightly, even in this area the K20 is still around twice as powerful as the previous Tesla generation. For single-precision processing, performance is closer to three times as much.

COMPREHENSIVE SPEC

The rest of the specification is equally top-end. Twin Xeon E5 2687W processors are included, each with eight cores and a nominal 3.1GHz clock speed, although Turbo Boost means a single core can jump to 3.8GHz when required, and multiple cores to 3.4GHz. Hyper-Threading divides each physical core into two virtual ones, so the system can offer 32 virtual cores in total. Backing this up is 32GB of 1,600MHz DDR3 SDRAM, supplied as eight 4GB modules, leaving eight slots free for upgrade. Storage is comprehensive too, with a 256GB Crucial RealSSD solid state disk for the operating system and apps, plus a pair of 2TB Western Digital Caviar RE4 mechanical hard disks configured as RAID 1, so they’re mirrored to keep data secure in case of failure.

The 2305-7T’s top-end spec includes a 256GB SSD and 2 2TB SATA hard drives

The Venom’s performance is about as stunning as any we’ve seen. Maxon Cinebench R11.5 rendering is the same as the previous model – on a par with other workstations using the same Xeons, which are the fastest currently available. Graphics performance is phenomenal, with 86.08 in the Cinebench OpenGL test, streets ahead of systems using any other graphics than the Quadro K5000. The key SPECviewperf results of 74.63 in lightwave-01, 121.65 in maya-03, and 64.77 in sw-02 are all among the highest we’ve recorded.

Then, of course, there’s the Tesla card, which provides acceleration for Cuda- and OpenCL-enabled renderers. We tried a test scene in Bunkspeed’s Pro Suite, an advanced CUDA 3D renderer: the scene took 530 seconds with just the CPUs, and 403 seconds with just the Tesla card. Adding the Quadro as well reduced the render time to 252 seconds, and adding all three took this down to 196 seconds.

So the Tesla card and GPU can more than halve the time for a Cuda-enabled render. The Boston Venom 2305-7T is the most powerful workstation we have ever tested, but you do pay a hefty premium for performance – just under £10,000. This is just about what you would expect considering the K5000 graphics and Tesla card. But you’ll need to have a Cuda-enabled rendering workflow to make the most of this top-end system.

PROS

• Unbeatable rendering performance, particularly if your rendering uses Cuda
• Supreme modelling performance

CONS

• Expensive

Eight-core Xeon processors and Quadro K5000 graphics make this the most powerful workstation we’ve ever tested

RATING: 5

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

Maya 2013 reviewed – over at our sister site, Creative Bloq.

The Quadro K5000 offers a big leap in modelling performance, but James Morris finds that it’s not a clear winner in every area

PRICE: £1,511 / $1,800

MAIN FEATURES:

• DirectX 11
• OpenGL 4.3
• Shader Model 5.0
• 1,536 CUDA processing cores
• 4GB GDDR5 RAM
• 2 x DisplayPort
• DVI-I, DVI-D
• 4,096 x 2,160 resolution (DisplayPort 1.2)

MANUFACTURER: Nvidia

Nvidia launched its Fermi generation of Quadro professional graphics cards over two years ago. Although the range still stands up well, two years is a long time in the world of integrated circuits. There have been two generations of Intel processors in that time, for example, and AMD launched its W series of FirePro cards earlier this year. So, at long last, we can formally introduce you to Nvidia’s Kepler generation.

Kepler is currently only available in the Quadro K5000, which essentially replaces the Quadro 5000. If you’re selective with your reading of the specification, the K5000 will wipe the floor with the 5000, and every Quadro that went before it. Where the 5000 has 352 CUDA processing cores, the K5000 has a gobsmacking 1,536. But things aren’t quite that simple – and an explanation of the theoretical processing throughput is required.

The Quadro 5000 is capable of 718 gigaflops of single-precision processing, where the K5000 manages a huge leap to 2,150 gigaflops. However, the story is the complete opposite when it comes to double-precision (64-bit) processing, with the Quadro 5000 achieving 359 gigaflops, while the K5000 can only muster 90 gigaflops. These figures should be contrasted with AMD’s recent FirePro W8000, which can produce 3.23 teraflops of single-precision and 806 gigaflops of double-precision performance. For the majority of 3D work, however, the single- precision increase will far outweigh the drastic fall in double-precision figures.

Benchmark tests

3D World tested Nvidia’s Quadro K5000 in the Armari Magnetar M32-AW750R workstation

We tested the K5000 in Armari’s Magnetar M32-AW750R, giving us a direct comparison with the AMD FirePro W9000 – which is a little more expensive but broadly comparable in the target market. The K5000 managed 87.36 in the OpenGL portion of Maxon Cinebench R11.5, compared with the W9000’s 74.19.

SPECviewperf 11 results were even more impressive, with the K5000 managing 77.33 in catia-03 compared with 21.14; 75.41 in ensight-04 compared with 55.11; 72.06 in lightwave-01 compared with 51.97; and 118.17 in maya-03 compared with 53.19. So the K5000 beats the W9000 across the board, and by significant quantities in the all-important 3D modelling viewsets lightwave-01 and maya-03.

We ran the same Bunkspeed CUDA-enhanced rendering test as we did for Boston’s Tesla-powered Venom 2300-7T. The test scene took 154 seconds with the CPU alone – almost the same as the Venom – which fell to 106 seconds with the K5000 helping out. However, the Venom took 72 seconds with the Tesla and Quadro 4000, implying that the K5000’s modelling abilities aren’t quite as stunning for CUDA-powered rendering.

We also tested the K5000 with Windows 8. Scores were mostly comparable, but a few SPECviewperf viewsets were behind the Windows 7 results, in particular 61.78 in catia-03, 12.42 in proe-05 and 53.45 in tcvis-02. So sticking with Windows 7 until drivers are optimised would appear to be the safest option.

Overall, the Nvidia Quadro K5000 is a significant new release for 3D content creators. It’s the fastest card for modelling on the market in every test we’ve tried, by quite some margin. As a CUDA or OpenCL co-processing unit, however, the K5000 isn’t such a clear winner. For 3D rendering it can offer a valid contribution, but any application requiring double-precision grunt should be run on different hardware.

PROS

• Fastest 3D accelerator for modelling
• Boost for CUDA-based 3D rendering
• Comparatively reasonably priced

CONS

• Lower 64-bit processing than Fermi cards
• Not ideal as a scientific CUDA co-processor

The Nvidia Quadro K5000 showcases Kepler’s modelling abilities but shows that it’s not ideal as a CUDA co-processor

RATING: 4

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

The Quadro K5000 was tested on the mighty Armari Magnetar M32-AW750R. To read more about the workstation, click the following link: Review: Armari Magnetar M32-AW750R