Sign in with
Sign up | Sign in
Dell Precision T5600: Two Eight-Core CPUs In A Workstation
By ,
1. Inside And Outside Of Dell's Precision T5600 Workstation

Last month, Intel introduced its Ivy Bridge-EP-based Xeon E5 v2 processors, including the dual-socket-capable -2600-series CPUs. While we've already spent some time previewing the company's latest efforts in Intel's 12-Core Xeon With 30 MB Of L3: The New Mac Pro's CPU? and continue testing the company's latest and greatest in our lab, let's not forget that there are still Sandy Bridge-EP-based workstations out there compatible with the same LGA 2011 ecosystem, and now available at discounted prices.

Dell sent over one such system for us to look at: its T5600, which starts at around $1400 from the company's online store. Of course, as you'd expect, there are copious options to choose from, facilitating truly high-end configurations. When everything was said and done, we ended up with a monster of a machine priced in excess of $8000.

What do you get for just shy of ten grand? Two eight-core Xeon E5-2687W CPUs (the same ones we reviewed in Intel Xeon E5-2600: Doing Damage With Two Eight-Core CPUs), 16 GB of DDR3 memory, a Quadro K5000 card, a SAS-capable RAID controller with a couple of Samsung SSDs hooked up, and Windows 7. Naturally, most of the cost comes from those CPUs, which still sell for almost $2000 each on Newegg. But at $1800, the Quadro card isn't cheap, either.

Dell's Precision T5600 is fairly compact, particularly considering that it houses two 150 W CPUs. It bears the industrial aesthetic a business professional might expect from a workstation, but also isn't just a boring black box, either. Front-panel I/O includes four USB ports, though only one of the is 5 GT/s-capable, audio output, and a microphone input.

Our review sample included one slim 8x DVD+/-RW drive, though Dell does have an option to add a second optical drive in the full-size 5.25" bay next to it.

Up top, in front and in the back, you'll find integrated handle designed to make the heavy machine a little more portable. It's silver on both ends; however, each handle's shape maintains the top panel's lines. Under the front panel's handle, the workstation recesses, and you find a grated cover for ventilation. Around back, there's an exhaust vent directly below the handle, right where most desktop cases blow CPU-warmed air out.

The Precision T5600's enclosure can be used laying flay on a desktop, presumably with a monitor on top, or standing up next to your desk in a pedestal configuration. If you're standing it up, an indentation on the top panel functions as a tray you can use for holding external storage, for instance, or maybe your keys and wallet. Flipped over the other way, you'll see the flat side panel up top with a handle for easy access to the workstation's internals. Not aesthetically pleasing, we'd say, but certainly more functional than a row of screws.

Pop the side of Dell's Precision open for a better look at the system's internals. Whereas our reference workstation from iBuyPower benefits from closed-loop liquid cooling, the T5600 goes traditional with big air-cooled heat sinks. Exhaust moves from right to left, so the waste heat from Xeon E5 number one is largely what cools Xeon E5 number two. The processors are offset a bit, so the far-right one also blows across two memory slots.

Clearly, the principles of building and cooling workstations are different from gaming PCs. That is to say, Dell's setup works the way it should, and both high-end processors run completely stably. But a lot of airflow is required, and the outcome is not a particularly quiet configuration under load (idle acoustics are much more favorable). Granted, with two 150 W Xeon E5-2687W processors, this is as taxing as it gets. Intel's lower-voltage CPUs won't get as hot and, consequently, won't require the same airflow.

Dell's motherboard hosts eight DDR3 slots, or four per processor interface to tap into Sandy Bridge-EP's quad-channel memory controller. The company smartly populates all of them with 2 GB modules to yield 16 GB of ECC-capable DDR3-1333, maximizing memory bandwidth.

Under the processor interfaces and memory slots you'll find peripheral connectivity. Each Xeon offers 40 lanes of 8 GT/s PCI Express, and Dell exposes two 16-lane PCI Express 3.0 slots, one third-gen x16 slot wired for x4 signaling, one second-gen x16 slot also wired for x4 transfers, one single-lane PCIe 2.0 slot, and one legacy 32-bit PCI slot.

Note that the PCI-based FireWire card is not installed in the corresponding slot. Neither is Dell's add-in storage hardware, the PowerEdge RAID Controller based on LSISAS2008 silicon. The eight-lane PCIe 2.0-connected board exposes four ports of SAS or SATA connectivity, and is optionally available with 1 GB of cache on-board for an extra $385. Incidentally, this probably wasn't a necessary upgrade on our workstation sample, which came with two SSDs. Using the card instead of Intel's PCH-based storage controller added $35 of cost and as much as 45 seconds to each boot sequence as the ROM initializes. Because Dell bundled the PERC H310, though, we added some storage-oriented testing to our normal suite.

Despite its dated core logic, which continues to see use in the Ivy Bridge-EP platform, this Precision T5600 offers four USB ports up front and six around back. Unfortunately, only one in each location is USB 3.0-capable. You'll also find audio I/O, gigabit Ethernet, PS/2 peripheral connectivity, and an old-school serial port on the rear panel. The image above shows the workstation with its 635 W power supply option. However, stepping up to the hardware in our machine necessitates upgrading to 825 W. Dell only charges $7.50 for this, so it's fairly inconsequential. The company's configurator will let you know if you exceed the smaller unit's capacity before ordering.

Multiple fans lined across the front of the Precision T5600 are complemented by ducting to force air into the removable power supply and to shroud the CPU heat sinks. The enclosure's compactness and all of this duct work means that room for internal storage is very limited. It'll accept two 3.5" or four 2.5" drives, and again, up to two optical drives. This is plenty for most desktops. However, it's fairly limiting in a professional environment. Presumably, Dell would recommend hooking up to a file server. Otherwise, you're limited to 6 TB of slow mechanical storage when it comes to maxing out capacity.

2. Test System And Benchmarks

iBuyPower provides our reference workstation hardware, and we continue using the company's P500X as the baseline point of comparison in our workstation coverage. Check out our review of that system in iBuyPower P500X And P900DX Workstations, Reviewed

Test System Specifications

iBuyPower P500XDell T5600
CPUIntel Xeon E3-1270 v2 (Ivy Bridge), 3.5 GHz, Quad-Core, LGA 1155, 8 MB Shared L3, Hyper-Threading enabled, Power-savings enabled2 x Intel Xeon E5-2687W (Sandy Bridge-EP), 3.1 GHz, Octa-Core, LGA 2011, 20 MB Shared L3, Hyper-Threading enabled, Power-savings enabled
CoolerAsetek 550LC2 x Dell Heat Sink and Fan Combo
MotherboardAsus P8B WS, Intel C206 PCH, BIOS 2009Dell 0Y56T3, Intel C602 PCH, BIOS A08
RAM2 x Kingston KVR1333D3E9S/4G, ECC DDR3-1333 CAS98 x Hynix HMT325R7CFR8C, ECC DDR3-1333 CAS9
GraphicsPNY Quadro 2000 1 GB
625 MHz GPU, 1300 MHz Memory
128-bit GDDR5, 42 GB/s Memory Bandwidth
192 CUDA cores
PNY Quadro K5000 4 GB
706 MHz GPU, 1350 MHz Memory
256-bit GDDR5, 173 GB/s Memory Bandwidth
1536 CUDA cores
RAID ControllerN/ADell PowerEdge RAID Controller H310, SAS/SATA RAID, PCIe 2.0 x8
SSDKingston Hyper-X SH100S3B/120G, 120 GB MLC SSD2 x Samsung PM830 MZ-7PC256D, 256 GB MLC SSD
Hard DriveHGST HDS732020BLA642 2 TB, 7K3000, 7200 RPMN/A
OpticalLite-On iHAS124-04(C) 24x Dual-Layer DVD±RW WriterSamsung/Toshiba SN-208 Slimline 8x DVD+/-RW SATA

Sound

Echo Digital Audio AudioFire 2 (not included in price)

Echo Digital Audio AudioFire 2 (not included in price)
Creative Labs Sound Blaster Recon3D PCIe (not used in testing)
NetworkingIntegrated Intel 82574LIntegrated Intel 82579
FireWireIntegrated VIA 6308SLSI L-FW323-07 Three-Port PCI FireWire Card
Power SupplyCorsair TX650 V2, 80 PLUS Bronze, 650 WDell H825EF-00, 825 W
CaseCooler Master Silencio 550Dell T5600 Case
Software
Operating System
Windows 7 Professional x64Windows 7 Professional x64
Graphics DriverQuadro Driver 320.49Quadro Driver 320.49
Audio Driver5.85.8
ASIO DriverIncluded in audio driverIncluded in audio driver
Warranty and Price
WarrantyThree-year labor, one-year partsThree-year basic hardware service with three-year NBD on-site service after remote diagnostic
Price As Configured$1999$8012

Since our last workstation review, Adobe shipped its Creative Cloud suite, while Autodesk released the 2014 versions of their apps. So, the workstation tests are updated to reflect those changes.

Application Benchmarks and Settings
Compression/Archiving
7-ZipVersion 9.28: THG-Workload (1.3 GB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5"
WinRARVersion 4.2: THG-Workload (1.3 GB) to RAR, command line switches "winrar a -r -m3"
WinZipVersion 17.0 Pro: THG-Workload (1.3 GB) to ZIP, command line switches "-a -ez -p -r"
Content Creation
Newtek Lightwave 3D 11.5Custom workload: High-polygon-count Tom’s Hardware logo, Modeler test: Scripted cloning of surface details across a segment of the logo, Render test: 1920x1080 render of logo with photoreal motion blur, ray-traced shadows, global illumination, OpenGL Test: Generate OpenGL preview of animation for real-time playback on screen
BlenderVersion: 2.68a
Syntax blender -b thg.blend -f 1, Resolution: 1920x1080, Anti-Aliasing: 8x, Render: THG.blend frame 1, Cycles renderer and internal tile renderer (9x9)
e-on SoftwareVue 11 Infinite PLECustom workload: Landscape (generated in Vue 8 full version and imported into PLE)
Autodesk 3ds Max 2014Space Flyby Mentalray, Frame 248, 1440x1080 Tom’s Hardware Logo render in V-Ray, 1920x1080, global illumination, photorealistic motion blur, ray-traced shadows, Create Nitrous preview of logo scene, to Y: RAM drive, Autodesk chair scene in iray, 1920x1080, 250 passes, GPU (CUDA) only rendering, Car render in V-Ray RT, 1920x1080, 256 passes, CUDA-only
Autodesk Maya 2014Tom’s Hardware Logo render in mental ray, 1920x1080, global illumination, photo-realistic motion blur, ray-traced shadows, OpenGL Test: Generate Playblast (OpenGL preview) animation to Y: RAM drive
Maxon Cinebench r11.5

3D Rendering and OpenGL Benchmarks, built-in benchmarks with default settings

Adobe Premiere Pro CCCustom Workload: Edit of 59.94 fps 720p DVCProHD footage, with transitions and some color correction, Render To H.264 720p
Adobe Photoshop CCFilter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates filters
Adobe After Effects CCCustom Workload: SD motion graphics sequence with three picture-in-picture frames sourced from 720p HD QuickTime, Same scene rendered using a frame sequence instead of from QuickTime sources, HD redo of the project using frame sequences, to 1080p
Adobe Acrobat XIVersion 11: Print PDF from 115 Page PowerPoint, 128-bit RC4 Encryption
Reaper v.4.402DAWBench Universal 2012: Test number of simultaneous copies of ReaXComp that the system can effectively run, Custom Workload: Render and mix down to .wav custom score project, multiple tracks of audio, VST synthesizers and effects
Visual Studio 2010Compile Chrome project (1/31/2012) with devenv.com /build Release
Encoding Benchmarks and Settings
HandBrake CLIVersion: 0.9.9
Video: Big Buck Bunny (720x480, 23.972 frames) 5 Minutes, Audio: Dolby Digital, 48,000 Hz, Six-Channel, English, to Video: AVC Audio: AC3 Audio2: AAC (High Profile)
TotalCode Studio 2.5Version: 2.5.0.10677: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, 2 Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV
LAME MP3Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s)
Synthetic Benchmarks and Settings
SPECviewperf 11Default GUI options; Workloads: CATIA, EnSight, LightWave, Maya, Pro/E, SolidWorks, Teamcenter Visualization Mockup, Siemens NX
LuxMark 2.0OpenCL-based rendering benchmark, default settings
CASE Euler3DCFD simulation over NACA 445.6 aeroelastic test wing at Mach 0.5
Thesycon DPCLatDPC Latency Checker, run with default settings
SiSoftware Sandra CPU Test=CPU Arithmetic/Multimedia, Memory Test=Bandwidth Benchmark, Cryptography
Iometer 1.1.0Workers = 1, 16 GB repeating data, 4 KB random, 128 KB sequential
AS SSD 1.7.4739Sequential, 4 KB Random, and 4K-64 THRD tests
3. Results: Adobe Creative Cloud

After Effects

We have a long-standing issue with our After Effects test, where it doesn't respond positively to more than four processing cores. In fact, we've seen instances where adding more compute horsepower actually yields a negative effect.

As you can see, with nine cores (the maximum we can allocate on the Precision T5600 workstation due to its 16 GB of memory) the workload takes two seconds longer to complete than with four cores. Hours have gone into troubleshooting this since Chris first noticed it, and I finally figured out the source of the issue: QuickTime. The source for the three window overlays are QuickTime footage encoded using the QuickTime PNG codec, making them effectively lossless. So, I changed them out for PNG image sequences.

The result is a slight improvement on the baseline machine, but a huge improvement favoring Dell's Precision T5600. It literally renders the test in one-third of the time using the same nine cores.

The side effect, of course, is that the After Effects motion graphics test we've been using to benchmark is finishing too fast on the latest hardware. I decided to increase the workload's demands by upping the project to 1080p. This was achieved by using Illustrator to convert the background graphic into a structured drawing, and then resizing and repositioning the video overlays for a 16:9 aspect ratio.

This was a good idea at first. And it worked great until I moved the test from the baseline machine's hard drive to an SSD, at which point the complete time went from over two minutes to 71 seconds. Then I ran it on the Dell workstation, which chewed through the task in 24 seconds. That's just one extra second more than the 720p-based benchmark, even though the frame being rendered is six times larger.

On the other hand, the HD version is now much more sensitive to storage performance. That's not a bad thing, given how hard it can be to demonstrate the real-world performance benefits of SSDs. Given more system memory, the T5600 could have utilized its additional execution cores, and the test might have finished even faster.

Adobe Premiere

This test involves our standard Hollywood Sequence project featuring DVCProHD 720p footage. Since the footage is only 100 Mb/s, it isn't as sensitive to storage performance. And as a result of optimizations for GPU acceleration, it's much more affected by graphics performance. Remember, this is the latest version of Premiere Pro, so it is OpenCL-enabled, so it also works with AMD's cards.

The encode to H.264, however, is still an entirely CPU-based affair, which is more taxing. Previously, I was rendering transitions and encoding separately so that both tasks could be evaluated on their own. But to make my tests consistent with Premiere benchmarks on other hardware, I used the same settings. Still, the Dell proves itself just under three times faster than our baseline.

Adobe Photoshop

Our two Photoshop tests are based on a series of filters applied to a very large image. They were selected for their prodigious use of threading (in the case of the CPU test) and OpenCL acceleration (for that corresponding benchmark).

The CPU-based Photoshop test falls right about where we'd expect it, just under three times faster over the baseline machine. But the OpenCL result is more difficult to explain. This behavior shows up in our WinZip tests, too. I tried a couple of different Quadro driver revisions and saw little or no difference.

4. Results: Autodesk 3ds Max And Maya

Autodesk 3ds Max 2014

Space Flyby

This is the 3ds Max test used for benchmarking CPUs here at Tom's Hardware. It's a fairly straightforward mental ray render with very little in the way of advanced settings. Thus, it finishes pretty quickly, even on our baseline machine. Dell's Precision T5600 winds up 2.24x faster than that system.

3ds Max: V-Ray

This is our Tom’s Hardware logo scene, which was originally created in LightWave 3D, imported into Max via FBX format, retextured, and then output with new settings in the popular V-Ray renderer from Chaos Group. Four frames are sampled from the animation to reflect a quartet of different frame content types. As you can see, they behave uniquely depending on the number of polygons in the scene, how much motion blur is applied, and whether the motion blur is linear. Dell shows up 4.3-4.6x faster than our baseline Xeon E3 box, thanks to its two eight-core CPUs and much higher memory bandwidth.

3ds Max: V-Ray RT

While I was messing around using V-Ray RT for another article, I did some digging in the 3ds Max settings and found that the color space wasn't set correctly for our benchmark, and the orientation of the light used to illuminate the car wasn't right, either. I also darkened up the tires, taking us from this:

To this:

It also affected our render times, though just slightly.

The update runs slightly slower on one machine and slightly faster on the other. Overall, Dell finishes 3.1x faster than the baseline box. That's pretty much the result we were expecting from a CUDA-accelerated workload shifting from Quadro 2000 to K5000. As a side effect, we're using a better render now, too.

3ds Max: iray

Our iray benchmark is a GPU-accelerated version of mental ray, using a scene provided by Autodesk. The Precision T5600 comes out 2.2x faster, which is a narrower victory than the other CUDA-based workload, presented above.

3ds Max DirectX Preview

This benchmark tests the performance of 3ds Max’s 3D display by playing back a preview of the entire THG Logo animation to our machine's RAM drive. It's a fairly accurate representation of 3ds Max's viewport, which is DirectX 11-based. While it benefits slightly from the T5600's faster GPU, it's largely bottlenecked by the process' single-threaded nature.

Autodesk Maya 2014

Maya: mental ray Rendering

Maya ships with mental ray as its renderer, and since our complex render test for 3ds Max is done in V-Ray, we're using mental ray in Maya. It’s the Tom's Hardware Logo scene again, retextured and with different settings (remember, these don't translate across apps). Autodesk's 2014 apps do come with an updated version of mental ray, but those features aren't being tested here. For example, it's now possible to offload your global illumination calculations onto the GPU, essentially generating a GI "pass" on the graphics card and using it in the software render. It's an interesting option to explore, though this benchmark is still entirely CPU-based.

The results are closer between these two machines than the other renders we've presented. The new version of mental ray seems to be much more efficient at calculating motion blur than its predecessor, more than halving the render time for frame 500. Faster render times on the same scene are always a good thing!

Maya: Playblast

Maya’s Playblast feature records a viewport to storage (or a RAM drive in our case) by grabbing the preview windows and spooling them out. Even with Maya 2014’s new DirectX 11 preview windows, the Playblast function is still so single-threaded that it limits the T5600's performance, allowing the baseline machine's higher-clocked Ivy Bridge-based processor to take a lead.

5. Results: NewTek LightWave 3D 11.5, E-on Vue 11, And Blender

NewTek LightWave 3D 11.5

If you've followed our revamped workstation coverage, you know that our 3D application testing started with LightWave. This is based purely on my familiarity with that product.

LightWave 3D Rendering

This scene originated in LightWave and is probably best optimized for it. It has numerous n-gons (polygons with more than four points) that keep geometry manageable, but then get necessarily tripled before getting exported to other applications. In all four frames, the difference between the two machines is close to a factor of three.

LightWave 3D OpenGL Preview

The OpenGL preview in LightWave serves the same purpose as the animation preview in 3dsMax and the playblast in Maya and works similarly. We run the preview test in the older multitextureshaders and the newer GLSL modes. The multitextureshaders are designed to work with more antiquated OpenGL systems and are incredibly fast for generating previews, while the GLSL technique uses shaders that evaluate much more of the surface options, including bump mapping and procedurals. Naturally, the GLSL mode takes a lot longer than the older mode, and the single-threaded nature of the preview means that the baseline machine comes out faster in GLSL. Thanks for that, higher clock rates and Intel's Ivy Bridge architecture.

LightWave 3D Modeler

This test uses a script within LightWave to clone polygonal geometry across the surface of our Tom's Hardware logo object. That, in a few separate stages, is how the detailing for the logo was created. The metric also reports how much time it takes to complete this operation. Even though Modeler is largely single-threaded, the test responds extremely well to the T5600's additional memory bandwidth, wrapping up in less than half the time of our baseline system.

E-on Vue 11 PLE

E-on’s Vue is a piece of 3D software designed for rendering landscapes. Our scene was originally created in Vue 8, and we load it into the current Personal Learning Edition for rendering. Vue makes good use of the Xeon E5-2687W's extra cores and memory bandwidth, yielding a 3.5x-faster result than the baseline.

Blender

Blender’s new Cycles renderer is more modern and efficient than the previous Tiles renderer. It supports newer features, including GPU-based rendering. But in this test, we’re performing a purely CPU-based benchmark. As we’ve seen with several other render tests, the Precision T5600 comes out slightly more than three times faster than iBuyPower box.

6. Results: Digital Audio Workstation

Using a computer as a DAW stresses a platform differently than typical content creation and animation apps. Latency becomes a critical factor, and many background tasks can affect that.

Thesycon DPCLat

Thesycon, a software engineering and systems consultation firm in Germany, produced a small application it calls DPCLat, which measures the total system latency caused by Dynamic Procedure Calls. It runs continuously, allowing you to see the low and high points of system latency. Sometimes device drivers can cause massive latency spikes that make the system unsuitable for use as a DAW.

The T5600’s overall latency is still well within the range of acceptability, but still a little higher than the iBuyPower baseline system. Anything under 500 microseconds is still considered good.

Tom’s Score

We created an excerpt meant to approximate a finished piece, mixed, equalized, and ready for mastering. In order to get more analog sound out of it, there are multiple copies of the same synth detuned against one another by very slight amounts. DiscoDSP Discovery Pro, KX-SYNTH-16 (V4), and Synth1 are the primary software synthesizers, DSK Choirs adds the background choral pad, and the drums come from old Emulator samples.

The test itself is a simple mixdown with the finished track output to 44.1 kHz and 192 kHz 24-bit WAV files. Dell's Precision T5600 yields just under twice the performance of our baseline machine in this measurement of pure number-crunching DSP power.

DAWBench 2012 RXC

As opposed to our Tom's Tune test, which evaluates a system's ability to render out a finished product at its own pace, DAWBench is real-time and interactive. Basically, it plays a loop of music and you keep adding a signal processing plug-in until the audio breaks up, signifying that the platform can no longer keep the audio buffers filled. This is a good overall determinant of a machine's utility as a DAW, factoring in variables like audio drivers and hardware.

Unfortunately, it also gives up performance to atypical factors, like poorly-written USB drivers or a slow FireWire card. Dell's Precision does well, though its PCI-based FireWire card might have artificially held it back a bit.

7. Results: LuxMark, Cinebench, SPECviewperf, and Euler3d

What do the metrics on this page have in common? They are all based on code from real-world applications turned into benchmarks.

LuxMark 2.0

This is an OpenCL-based benchmark based on the LuxRender render engine. In both tests, Dell's Quadro K5000 is about twice as fast as the baseline system's Quadro 2000.

Cinebench

Cinebench Rendering

Over the years, we've seen Maxon use several different scenes for its Cinebench test. The current iteration is no different, and it employs global illumination, soft shadows, and reflections, which are the attributes of a "modern" rendering engine. In this case, the engine comes from Cinema4D.

The render is run in both single- and multi-threaded modes. It's actually a little surprising to see the Precision's Xeon E5-2687W outperform an Ivy Bridge-based Xeon E3 in the single-threaded measurement. We're less shocked to see the T5600 decimate our baseline in the more parallel version of the test.

Cinebench OpenGL

The OpenGL-accelerated benchmark uses the viewport engine from Cinema4D to show a car chase animation. The T5600 is around two times faster than the baseline.

SPECviewperf 11

SPECviewperf has been around a long time. It uses the viewport code and tests designed for the viewport of several different 3D animation and CAD programs, running reasonably complex models through multiple iterations, recording frame rate, and reporting the results as a weighted mean.

The tests seem evenly split between single- and multi-threaded workloads, and some of them incur little or no hit from AA, which points to something other than the GPU bottlenecking performance. In fact, SolidWorks performs better with AA on. How odd is that?

Euler3D

Euler3D is a benchmark developed by the CASE Lab at Oklahoma State University. It simulates a Mach 0.5 airflow over a AGARD 445.6 aeroelastic test wing. This test is 32-bit because the developers feel that 64-bit compilers aren't quite as mature, and prefer to keep results standardized. However, the benchmark is fully threaded and multiprocessor-aware, automatically detecting the number of cores and launching the appropriate number of threads. Dell's Precision T5600 comes out 4.75x faster than the P500X in the largest performance delta we've seen.

8. Results: Media Encoding And Productivity

TotalCode Studio

The encoder formerly known as MainConcept is well-threaded, and a fast SSD array helps this dual-processor workstation achieve the best score seen in this test thus far.

HandBrake

HandBrake demonstrates similar results. Fast storage, plenty of memory, and two powerful CPUs deliver impressive numbers in this threaded encoder.

Lame

The same cannot be said for Lame, which is not a threaded test. Rather, it's sensitive to architectural efficiency and clock rate. You can certainly run multiple file conversions at a time to achieve some degree of parallelism, but that's not what we're measuring here.

Acrobat XI Pro

Also a single-threaded workload, converting a PowerPoint presentation to PDF taxes one core. The faster you can run it, the sooner this task finishes.

Visual Studio

Compiling is certainly a productivity-oriented task that legitimately justifies workstation-class hardware. This is our standard Chrome compile test, and you really can’t shake a stick at these results. Two and a half times faster than our baseline workstation and a third faster than Chris’ results here.

9. Results: Synthetics

Sandra 2013 Arithmetic

Dell turns in a 3.8x-higher result in Sandra's Dhrystone test, which is in the neighborhood of its theoretical peak, and around a 3.55x-better number in the floating-point module.

 Sandra 2013 Multimedia

The T5600's results are consistently about 3.5x faster in the Multimedia benchmark.

 Sandra 2013 Cryptography

The AES-NI-accelerated part of this test is memory bandwidth-bound, so naturally the Precision workstation is way in front, despite the fact that both CPUs benefit from support. The hashing test taxes processor performance, and the Xeon E5s turn out to be around 2.81x faster.

 Sandra 2013 Memory Bandwidth

Two CPUs with four-channel memory controllers offer a ton of aggregate memory bandwidth, as you can see.

10. Results: Compression

WinZip 17 Pro

WinZip doesn't like the Dell setup. The ultra-compression EZ test kept asking for more virtual memory. Also, as with our Photoshop-based OpenCL-accelerated workload, the higher-end workstation turns in worse numbers than the baseline box.

WinRAR

The Precision T5600's extra cores and memory bandwidth don't help much in an application not optimized to exploit all of those resources. Instead, the Ivy Bridge architecture running at higher clock rates wins the race.

7-Zip

The same can't be said for 7-Zip, which we have seen utilize lots of cores in the past. Dell's submission ends up 62% faster than the baseline. 

11. Results: Storage

Storage makes up our final battery of tests. Previously, we weren't specifically benchmarking our workstations' storage subsystems, but I felt like we should add more targeted metrics, particularly given the add-in RAID card and SSDs included with this T5600.

AS-SSD

AS-SSD results, T5600AS-SSD results, T5600

AS-SSD Results, P500XAS-SSD Results, P500X

Sequential read speeds as reported by AS-SSD aren’t that different, but write performance is dramatically better on the Precision. Dell's drive array writes almost eight times faster.

Iometer

Iometer gives us similar results. The 4 KB random numbers turn out to be pretty similar; it's the sequential I/O where Dell's older Samsung drives take off. In many workstation tasks (like video and audio editing), sequential access is more important than random access, especially when gigabytes of video footage need to be cut.

12. The Precision T5600 Is Still At The Top Of Dell's Workstation Portfolio

I believe it's easier to review workstations than most gaming-oriented desktop PCs. There's a lot of choice in the professional world; Dell makes its Precision T5600 available with 18 different combinations of Xeon E5 CPUs in single- and dual-socket configurations. But if you're adding somewhere between $168 and $3683 for something above the base-line Xeon E5-2620, then you're also probably in tune with why you'd need to spend so much more in the first place. Some professional apps use all of the processing resources thrown at them. Others are less taxing, and simply need to run stably without any downtime. Know your workload and configure accordingly.

Knowing that there are hundreds of combinations of hardware you can find to go into Dell's Precision T5600, this story is less about the performance of our specific setup and more about Dell's system in general. The T5600 is well-designed and compact. It packs a lot of power into a relatively small chassis, even if we're not the biggest fans of the way its two CPUs are cooled (and the noise they consequently make under load). The system is audible from 10 feet away with a television and window-mounted air conditioning unit running at the same time. That's the workstation market for you, though; we've certainly heard servers that were far worse. But we also admire iBuyPower's closed-loop liquid cooler on the P500X, which does its job almost silently.

Another critique is the T5600's lack of space for additional storage. Maxing out at two 3.5" mechanical disks does give you up to 6 TB of capacity, which is pretty gargantuan. But if you're dealing with massive datasets, anything extra is going to require a networked storage device. Even worse, if you want a tiered setup with SSDs and conventional hard drives, the online configurator stops you at one 256 GB SSD and one 3 TB disk. That's it; workstation full. "At least give us another hard drive for backup," we'd ask. Apparently, we weren't the only ones to bring this up, because the newer T5610 adds another drive bay, in addition to employing Ivy Bridge-EP-based Xeons.

Intel's C602 platform is pretty old now, and it never natively supported USB 3.0. So, manufacturers like Dell that want to add this functionality have to lean on third-party controllers. That's why the Precision T5600 only has one compatible port up front and another in the rear. Frankly, that's a lack of I/O you're going to run into on any dual-processor box, at least until Intel updates its core logic for servers and workstations. It's at least nice to see four total USB ports among the T5600's front-panel I/O and six around back.

With those few critical points addressed, Dell's Precision T5600 remains a compelling option for professionals shopping for massive processing power. Yes, Intel's Ivy Bridge-EP-based Xeon E5-2600 v2 CPUs are available now, and Dell does sell a Precision T5610. However, as of this writing, there's only one CPU option: the Xeon E5-2609 v2. That's a quad-core chip far slower than the two octa-core battleships in our T5600. No doubt, more Ivy Bridge-EP-based boxes will emerge soon. But there's a reason we were so bullish on Sandy Bridge-E (and, by extension Sandy Bridge-EP) when it first surfaced: it remains a great architecture that even Intel's latest offerings have a hard time besting.