Up from Nehalem: Westmere-based Dell Precision workstation wows

In July of last year, I reviewed a group of new Nehalem-based workstations from Dell, Hewlett-Packard, and Lenovo. The outstanding characteristic of these machines was the terrific performance conferred by the new Intel processor architecture. In this review, I examine an update to one of the models, the Dell Precision T5500 workstation. It now ships with the latest incarnation of Intel’s Xeon family, code-named Westmere. Benchmarks show it to be stronger and faster, while consuming less power.

The result is an excellent 12-core workstation that combines top performance, excellent graphics rendering, and miserly energy consumption, all in a desktop form factor. Let’s look at the details.

Dell Precision T5500: The workstation

The Dell Precision T5500 workstation is the midrange machine in the vendor’s lineup. That lineup could be viewed as serving three tiers: value (sub-$5,000), which includes the T1500 and T3500 models; midrange ($5,000 to $10,000), consisting of the T5500 model; and the high-end (greater than $10,000), which includes the highly expandable T7500 model.

In past years, the midrange has been a somewhat forgotten tier; customers wanted either value or all the power they could possibly get, regardless of cost. As a result, they favored the ends of the spectrum, rather than the middle. However, there is a growing appreciation that the high-end machines command a very high premium for the extra power and scalability. Today, the midrange machines, which are remarkably powerful systems in their own right, are emerging as a good blend of value and power.

The Westmere-based T5500 model reviewed here is in most ways identical to its older brother. It has the same case and desktop form factor. The physical attributes such as USB ports, PCI slots, and network adapter are all the same. Likewise, the graphics card is the same Nvidia Quadro FX 4800 model. This particular graphics adapter has been a staple of workstation vendors as it combines excellent performance with a good price. In addition, it is energy efficient, and with 1.5GB of video RAM, it has plenty of room for handling memory-consuming graphics.

The FX 4800 is notable for another reason: It has only one DVI port. The remaining two ports are DisplayPort, a technology that rose to fame in the Apple universe. These ports are quickly emerging as the new norm on graphics adapters due to the fact that they support more colors (up to 1 billion hues) than dual-link DVI. They also support longer cables, and they can send both audio and video signals on the same cable.

Dell Precision T5500: The processor Westmere is a die shrink of the previous generation of Nehalem processors. Die shrink means that the processor is literally etched onto the silicon using smaller connections (in this case, 32 nanometers). The upshot is that the same processor takes up less room, consumes less power, and runs faster. Intel chose to use some of this newfound room to add two cores on the processor die. This feature — six cores — is the distinguishing attribute of Westmere. How much this additional capacity delivers when compared with Nehalem’s four cores is examined in the next section.

Test Center Scorecard
	35%	20%	20%	15%	10%
Dell Precision T5500 Tower Workstation	10	9	9	7	9	9.1 Excellent

The Westmere processor is also faster, running at 3.3GHz (vs. 2.93GHz Nehalem in the previous review). This new clock speed is an important number. When Intel and AMD were competing at the turn of the century to deliver the fastest single-core processors, their offerings topped out at roughly this same 3.3GHz. At that point, the companies realized that pushing the clocks even a little faster would require so much more power that PCs would burn up from the radiated heat. This realization drove them to move to multiple cores — providing processing power in parallel. But when they did so, they dropped the clock rates back below 2GHz.

For many applications that ran on just one core, the result of this redesign was decreased performance. Both companies used numerous tricks to increase execution speed, such as larger caches, better branch prediction, and so on. Now, with Westmere, the clock speed of each core has finally worked its way back to where it was when the multicore revolution began. However, because of the added technologies and refinements, a single-threaded app on a 3.3GHz Westmere core runs considerably faster than it did on the last of the Pentium 4 chips with the same clock speed.

In addition to adding two cores to Westmere and increasing its clock speed, Intel bumped up the L3 cache to 12MB per chip (from 8MB). This keeps the cache/core ratio the same. However, because all cores can access all the L3 cache, active cores have as much as 50 percent more cache into which to store data and code for quick access when other cores are dormant.

Dell Precision T5500: Performance A side-by-side comparison of last year’s Nehalem-based T5500 with this year’s model is shown in the performance sidebar. As always, these numbers need some explanation. There appears to be a big jump in the SPEC Viewperf 10 results. This benchmark measures straight graphics output capacity. The number on last year’s machine is an anomaly due to how the test was conducted. (It’s due in part to Dell’s default vsync setting on Nvidia cards, which optimizes image quality but limits performance.) Additional tests of the same graphics subsystem done at that time showed results of around 85.

The pure graphics processing power is the same — not surprising, given that the Nvidia graphics adapter is the same. ViewPerf 11 was not used last year, so there are no comparative results. It’s included here so that you can compare against other workstations using the new version of the benchmark. Going forward, ViewPerf 10 is being retired in favor of version 11.

The difference in the Cinebench results is significant, however. Cinebench measures the system’s ability to render graphics. As expected, a 50 percent increase in the number of cores translates into an almost 50 percent jump in rendering capability. This number encapsulates the special use case for this workstation: Although most workstation software is not threaded, the rendering engine is the primary exception. With this machine’s rendering capability now significantly faster, designers can both design and render on the same machine. Rendering locally was certainly possible before, but the new level of performance makes it even more viable.

Westmere does have the option of Hyper-Threading Technology, which converts this system’s 12 cores into 24 execution pipelines. On some tasks, Hyper-Threading will provide a boost, but not on rendering. The Cinebench results are lower when Hyper-Threading is enabled. In general, it’s better to leave it off, which is the default configuration.

The remaining benchmarks show greater RAM bandwidth and a faster RAID controller. What is striking, though, is that adding a total of four more cores wrapped in a faster system can consume less power. This win-win provides a glimpse into how fast technology is improving power efficiency across all components of desktop systems.

The other notable numerical difference is that the current T5500 model runs about $200 more than its predecessor. As such, it slides in under the $10,000 cut-off for high-end workstations with some room to spare. Fortunately, you get a lot for the two extra C-bills. Looked at another way, the Dell T5500 combines the processing capability and memory capacity of a server, the graphics engine of a high-end workstation, and the power consumption of a desktop. For $9,000, that’s pretty good.

• View Dell Precision T5500 specs (as tested)
• View Dell Precision T5500 benchmark test results (3.3GHz Westmere vs. 2.93GHz Nehalem)

This story, “Up from Nehalem: Westmere-based Dell Precision workstation wows,” was originally published at InfoWorld.com. Follow the latest developments in workstations, servers, processors, and other hardware at InfoWorld.com.