As gamers and hardware enthusiasts, it’s easy to get jaded by Extreme Edition processors and multi-card CrossFire configurations. They push performance full-bore. And although you’ll never hear AMD, Intel, or NVIDIA recommending against using their enthusiast offerings in higher-end configurations, there’s no denying that Xeons, Opterons, Quadros, and FireGLs belong to a different pedigree. It’s The Hamptons versus South Beach. You can’t go wrong either way, but there is a distinct difference.

So when it comes time to pass judgment over workstation-class hardware, you might be tempted to turn your nose up at a list of stuffy-sounding specs. A quad-core Xeon running at under 2 GHz? Two gigs of system memory? A single-slot graphics card with a 128-bit memory bus? Are we kidding here? Ah, but check out the uptime. Bet you can’t get one of those boxes to crash, even under full load. Sure, servers and workstations prioritize performance, but there are other metrics used to determine if a critical machine passes muster.

Systemax recently sent over one of its Endeavor workstations, featuring a long list of those high-end parts that gamers might otherwise once-over. The business-class folks, on the other hand, will be much more likely eat up the machine’s build quality, its composition, and—hopefully, after our battery of tests—its dependability. 

Upon unwrapping Systemax’s Endeavor box, it was clear we were working with a true workstation. The all-black chassis (black is the new beige) is pretty much par for the course. In fact, you might recognize the enclosure as Chenbro’s SR10566 Black. Having had hands-on time with some of the company’s rackmount server enclosures, Systemax’s choice to go with Chenbro for its Endeavor makes good sense.

The box is plenty sturdy. Crafted from .8mm steel, the chassis alone weighs more than 23 pounds (total shipping weight for the whole workstation is a staggering 64 pounds). Nothing about the external enclosure is particularly eye-catching, which most workstation customers will appreciate since this is no enthusiast configuration. The only real weakness we picked up on was a somewhat flimsy side panel, which popped in and out when pressed. Nobody’s ever going to notice something like that once the workstation is in place and running. Even still, .8mm steel shouldn’t do that.

A plastic, mesh-covered door masks the enclosure’s externally-accessible drives and connectors. Hinges on the door feature limited range, so you’ll only be able to swing it open 90 degrees from front bezel. We've seen plenty of instances where a less constrained door comes in useful, so the door's 90 degree hinge is a bit of a bummer. The SR10566 has three 5.25” bays, two of which come populated by a floppy drive and dual-layer DVD writer. To be honest, we could have done without the floppy, especially since Windows XP comes pre-installed on the workstation and anyone planning a move to Vista can install storage drivers straight from an optical disc.

Two USB 2.0 ports constitute all of the chassis front-panel connectivity and the rest of the bezel is consumed with air intakes and an 80mm fan that pulls air in through the front door and blows it over the workstation’s hard drives. Classy, yet simple. Even still, front access to the installed hard drives would have been nice, had the workstation featured a hot-swap cage.

Spin the chassis around and you’ll find fairly average I/O. The Silverstone ST75ZF power supply incorporates a fan pushing air from inside the case out through a vent in the back, so there’s no visible cooling from the back. That setup also seems to help the supply’s acoustic characteristics. Unfortunately, there’s a non-adjustable 120mm cooler right below the power supply cranking at full blast, generating plenty of noise.

The Supermicro rear I/O panel offers up PS/2 ports, one serial port, a single parallel port, four USB 2.0 connectors, twin RJ-45 jacks, and enough 1/8” mini-jacks for 7.1-channel analog output. Only one peripheral is plugged into the board’s expansion slots—a Quadro FX 1700. Its dual-link DVI outputs are, of course, available back there as well. 

As a whole, nothing on the Endeavor's exterior screams eye-candy. Whereas tier-one boxes generally have their own unique touches, whether that is an integrated memory card reader, proprietary styling, or prominent branding, Systemax has adopted a less-is-more philosophy. On one hand, that’s great for the professional who isn’t trying to stand out with deep blue lighting or extra plastic pieces. On the other hand, you’re left with the feeling that the Endeavor is yet another whitebox workstation built using parts readily accessible to enthusiasts online. Thus, we’ll be looking for the system’s value somewhere other than its superficial shell.

We ran the Endeavor as it came configured from Systemax. The platform included a 32-bit installation of Windows XP—apparently, Vista still isn’t considered suitable in the professional space, though we do have a Vista workstation here that runs well now that software drivers are much more prolific. The Endeavor was missing some of the latest patches from Windows Update, so those were all installed in order to match our reference configuration.

The hardware in our reference workstation is based largely on Intel’s “V8” content creation platform, minus one of the Xeon X5365 quad-core processors. The platform also shipped with 4GB of FB-DIMM memory. However, we yanked two of the 1GB modules and added an NVIDIA Quadro FX 3450 for comparison’s sake. The 3450 is technically a higher-end card, but because it centers on an older graphics architecture and includes 256MB of memory, the match-up may be closer than expected. 

Test Systems Intel Inside!

Preliminary Testing with SiSoft Sandra XI Synthetic Benchmarks

We started off with a popular synthetic suite of benchmarks able to measure the CPU performance, multi-media alacrity, and memory bandwidth of both Systemax’s Endeavor and our reference workstation platform. SiSoft’s SANDRA, optimized for threading, will demonstrate the potential of going from one quad-core CPU to a pair operating at nearly twice the frequency.

Futuremark PCMark05 and 3DMark06 Synthetic CPU, memory, and graphics benchmarks

PCMark05 has been superseded by Futuremark’s PCMark Vantage suite. However, Vantage only runs on Windows Vista. Given Systemax’s decision to sell the Endeavor with XP, we were left leaning on the aging PCMark05 suite. We ran the workstation through the full battery of 11 system tests designed to mimic real-world usage patterns. Everything from XP startup time to file decryption, Web page rendering, and multi-tasking are measured here.

Despite its weaker storage subsystem, the reference machine’s superior processing power again pushes it through this synthetic-gone-real-world benchmark. The more economical take-away here is that the Endeavor, which costs signficantly less than the reference box, is able to crank out such respectable numbers. How's this for perspective? The street price on a pair of Xeon X5365 CPUs, the processor in our machines, costs as much as Systemax's entire workstation.

Given the pedigree of Systemax’s Endeavor, it’s fair to say that the box isn’t optimized for gaming. However, Futuremark’s 3DMark06 still gives us a good idea of how the machine’s graphics subsystem handles today’s DirectX-based workloads.

Amazingly, the Quadro FX 1700’s unified shader architecture and 512MB of onboard graphics memory pushes its way past the Quadro FX 3450 in 3DMark06. The Endeavor shows that when your workload is decidedly graphics-heavy, processing horsepower will only take you so far. The Quadro FX 1700 costs less than the FX 3450, but its architecture and memory configuration is better suited to 3DMark’s DirectX tests.

As expected, 3DMark06’s CPU test swings back in favor of the Xeon X5365.

LAME MT MP3 Encoding Test Converting a 622MB .wav into .mp3

Now we’re getting into the real-world testing—the types of usage models you’d see with some hands-on time with the Endeavor. First up is LAME MT a threading-aware encoding application used to convert a .wav file into .mp3 format. Naturally, this is going to be another processing-intensive metric that scales according to clock speed and the number of cores you throw at the project. For our purposes, we used a 622MB copy of Tchaikovsky’s 1812 Overture.

For the discrepancy in frequency and front side bus speed, Systemax’s Endeavor fares well enough against our reference config. It still takes more than a minute longer to finish the task of turning .wav into .mp3. As a clear play on value, at least, the Endeavor does its job remarkably well.

Windows Compression 500MB file folder, zipped

Next up, we measured the time it took for Windows to compress a 500MB folder of music, movies, Web pages, and documents of various sizes. On the Endeavor, we performed the test on its RAID 0 array, where higher theoretical I/O throughput would show off the machine’s higher-end storage subsystem.

Windows compression nevertheless favored brute processing strength over disk performance as Systemax’s Endeavor took 20 seconds longer to cram our test folder into a .zip file.

Cinebench R10  3D modeling and rendering

Cinebench, based on Maxon’s CINEMA 4D rendering tool, is another processor-intensive test benchmark very relevant to the workstation world. The test is both multi-processor and multi-core aware, so you can expect to see scaling according to the speed of our CPUs and the number of cores put up against the test. Release 10 of the benchmark features a new scene that better taxes today’s graphics cards and processors and incorporates light sources, procedural shaders, ambient occlusion, and multi-level reflections.

In what’s proving to be a pattern, the superior processing power of our reference box flexes its muscle yet again. In the single-threaded test, you see the result of when a 1.6 GHz core goes up against a 3 GHz core based on the same architecture. The multi-threaded test demonstrates similar scaling.

POV-ray 3.6  Single-threaded ray-tracing

For our next test, POV-Ray, we opted against using the latest beta version 3.7, which adds symmetric multi-processing support to the graphics creation tool. Instead, we used 3.6, the latest official version of the software.

This is another benchmark where faster processors are going to give you better results. True to our expectations, the 3 GHZ Xeon X5365 handily outperform the E5310 in Systemax’s Endeavor workstation.

By now you’re probably wondering why we’re being so gosh-darned hard on the plucky little Systemax setup, when in fact that isn’t our intention at all. In fact, as we’ve pointed out several times already, the Endeavor is putting up quite the fight given its price point and composition. While it’ll naturally lag behind a workstation specifically designed to showcase processing power, the Endeavor’s graphics and storage performance offset the CPU somewhat. And remember, there’s still an empty processor socket, ready to take another four cores running at 1.6 GHz, doubling the Endeavor’s available processing resources. And if you're looking for a little more juice right out of the gate, you can buy the Endeavor with up to a Xeon E5345 running at 2.33 GHz on a 1333 MHz bus.

picCOLOR Image analysis benchmarking

You’ll find two more workstation tests on this page. The first is called picCOLOR, described as a modular system designed for scientific and industrial image processing and analysis. Here we are, faced with yet another test of processing prowess.

picCOLOR scores a reference 1 GHz Pentium III on a 440BX-based platform as a 1. Everything beyond that is scored in reference to that system. Systemax’s Endeavor scores a 5.26 and our own content creation workstation delivers a 9.04. The difference in core frequency and front side bus speed is once again apparant.

SPECviewperf 10 Graphics performance benchmarking under OpenGL 1.5

SPECviewperf should be a familiar sight to graphics workstation professionals. Version 10 of the benchmark incorporates viewsets from 3ds max, CATIA, EnSight, Maya, Pro/E, SolidWorks, UGS Teamcenter, and UGS NX. It features the option to test full-screen anti-aliasing in a battery of tests that take several hours to run, but the standard batch simply runs through the viewsets. We chose to test our workstations at 1600x1200.

SPECviewperf 10 doesn’t favor the Endeavor’s Quadro FX 1700 as decisively as 3DMark06 did, but there’s a definite case for the workstation’s value as a graphics powerhouse here. In five of the eight viewsets, Systemax’s box puts down better numbers than our reference machine equipped with NVIDIA’s $999 Quadro FX 3450. When it comes to bang for the buck, that FX 1700 really gives the Endeavor some serious oomph.

We can rave all night and day about the torque afforded by super-fast Xeons running at 3 GHz, with 8MB of cache and access to a 1,333 MHz bus. However, there’s a price to be paid for that level of performance. You first pay it upfront as a heftier price tag (the X5365s still go for $1,200 or so). Then, you pay a higher energy bill. Our Xeon chip belongs to the first stepping, rated at 150W (they’ve since come down to 120W). But the Xeon E5310 wielded by Systemax’s Endeavor sips just 85W. Further, the Quadro FX 1700 doesn’t need the assistance of an auxiliary power connector, whereas the Quadro FX 3450 guzzles more juice than the PCI Express bus can crank out. The numbers add up quickly. And if the Endeavor is going to show its grace anywhere, it'll be here.

Power Consumtion Testing under load and idle

We connected both Systemax’s box and our own reference platform to an Extech power analyzer, cranked up 3DMark06 and POV-ray simultaneously, and recorded the peak consumption while both tests ran. Then, we let both workstations idle for 10 minutes and recorded power draw.

Here’s where the Endeavor really shines. True, the box features business-class hardware—mainly Intel’s Xeon, Supermicro’s 5000X-based motherboard, Kingston FB-DIMM memory, and NVIDIA’s Quadro FX 1700 workstation card—but it doesn't consume much more energy at load versus its 170W idle. Our reference box, on the other hand, sucks down more than 100 extra watts once its weighed down with a heavy load.

Naturally, adding a processor, dropping in more RAM, or substituting in a higher-end graphics card is going affect the Systemax box’s power consumption, but it’s fairly clear that the advantage here lays with the Endeavor, no matter what you add to it.

At the beginning of this evaluation, we observed that Systemax’s Endeavor looked a lot like a whitebox (pardon the cliché—we know the chassis is black) any enthusiast could piece together with parts sourced from e-tail. Indeed, at the end of that day, that remains true. There’s no proprietary molding with extra front-panel connectors or fancy branding. Instead, you get a well-built platform that, in our experience, ran as stably as you could hope from a workstation machine.

Seeing as the box itself was built as well as we could have expected, our eyes turned next to support. Systemax protects select platforms with a three year parts and labor warranty. Ours, however, appeared as though it was covered for one year instead, according to Systemax’s personalized support site. The site also looked like it was set up to provide knowledgebase articles and driver downloads, though it wasn’t populated with any additional information beyond the workstation’s basic specs. We were hoping for a bit more there, seeing as though support would be a great place for a system builder to really step things up a notch.

The performance story isn’t a concern for us. The Endeavor cranks along very well. Its graphics card delivers seriously impressive speed given entry-level billing, and the addition of a second CPU could really turn things up a notch if you’re running a lot of threaded software. The striped storage array and DDR2-667 FB-DIMM memory modules balance out an otherwise competent workstation.

Would we buy one? As enthusiasts, it’d be hard to commit to a box so seemingly similar to what we are already comfortable piecing together. We would, however, have no problem recommending the Endeavor to someone more interested in working on a workstation than trying to build one.

Reasonable Price Excellent Craftsmanship Plenty of flexibility Extremely stable	Relatively noisy Average performance as configured