Media Center server update highlights CPU evolution
In that same February 2017 column, I mentioned that a notable number of Windows Media Center aficionados were abandoning the default program guide for an independently enthusiast-developed and paid (albeit very economical) alternative called EPG123. And shortly after crafting that earlier writeup, I too took the EPG123 plunge, after yet another program guide outage, albeit cautiously at first. I didn't want my experiment to muck up what I'd already set up, in case EPG123 didn't work out. Fortunately, I had another headless Windows 7-based PC available to me that I could use as an EPG123 test platform.
As background (for all but the longest-time and longest-memory readers, who've heard this before), to date my Media Center Server had been based on a "headless" Foxconn R10-S4, built on an Intel Atom 330 CPU foundation. I originally purchased the R10-S4 for $89.99 (not including the HDD, RAM, optical drive and other extras) in March 2011; the Atom 330 was launched in Q3 2008, according to Intel's website. Here are a few photos:
The other system available to me is Foxconn's nT-i2847, which I purchased in May 2014 for $109.99 and is based on Intel's Celeron 847 CPU introduced in Q2 2011. Here are a few "stock" photos of it (mine happen to be white) and its freebie (at the time, thanks to a promotion) companion DVD drive:
And here's another photo of the two magnet-mated together and powered up, along with a banana on top for scale:
Obviously, the nT-i2847 is much smaller than its R10-S4 predecessor; to be precise, it has dimensions of 7.48"×5.31"×0.98", while the companion DVD burner is 7.48"×5.31"×0.79" (and the R10-S4 is 11.1"×12.4×3.75"). But that's not what I'm particularly focusing on here. Whereas the Atom 330 CPU employs a 1.6 GHz dual-physical-core (quad-virtual-core, thanks to HyperThreading support) architecture, the Celeron 847 is a 1.1 GHz dual-physical-core processor with no HyperThreading virtual-core augmentation capabilities. Both systems are also based on Seagate "hybrid" HDDs (rotating magnetic storage paired with a front-end flash memory cache); a 2 TByte 3.5" ST2000DX001 for the R10-S4, and a 1 TByte 2.5" ST1000LM014 for the nT-i2847.
On paper, the Celeron 847 would seemingly be slower than the Atom 330. But in reality, the nT-i2847 runs rings around its R10-S4 precursor. Take a look, for example, at the Windows Experience Index reports for the R10-S4:
And that of the nT-i2847:
And in action, whereas the Windows Media Center program guide takes around a minute to pull up when my Xbox 360 is connected to the R10-S4, the guide launches near-instantaneously on the nT-i2847; the Windows Media Center UI is otherwise similarly much snappier on the newer system. Part of the reason may be (although I doubt it) that the nT-i2847 includes a GbE wired Ethernet subsystem, whereas the R10-S4 only has a 100 Mbit Ethernet transceiver. The network chain (router, switches, cabling) between both PCs and any of my Xbox 360s is otherwise fully GbE-capable, but the Xbox 360s only offer 100 Mbit Ethernet support (therefore my doubt).
Operating system differences also bear mention; the R10-S4 is running 32-bit Windows 7 Professional, whereas the nT-i2847 employs 64-bit Windows 7 Home. And DRAM technologies and capacities also vary between the two systems; the R10-S4 includes 2 GBytes of DDR2-667 SDRAM, whereas the nT-i2847 uses 4 GBytes of DDR3-1333 SDRAM. But the bulk of the system performance improvement in the R10-S4 to nT-i2847 transition, I suspect, comes from CPU evolution, in spite of the newer processor's slower clock speed and lack of HyperThreading capabilities; among other things, the Celeron 847 supports out-of-order instruction execution, leading to greater IPC (instructions per clock) efficiency. PassMark benchmark results also bear out this hypothesis; a single-thread rating of 537 for the 1.1 GHz Celeron 847, versus 252 for the 1.6 GHz Atom 330. That's an impressive architecture-driven improvement, which has presumably continued with even newer processor generations from both Intel and AMD.
Post-transition, all's pretty much well. The new system's smaller, quieter, cooler, and faster than its predecessor. I even tethered an inexpensive 2 TByte external HDD to it as an alternative recorded-program repository, to supplement its comparatively diminutive integrated storage capacity (the nT-i2847 offers USB3 ports; the R10-S4 is USB2-only). And while EPG123 isn't perfect (it refuses to recognize that latest Season 8 broadcasts of Shark Tank are new programs, not repeats, for example), it at least updates reliably. Consider me pleased.
—Brian Dipert is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company's online newsletter.
- Set-top box replacement options
- Foxconn R10-S4: a competent Windows Media Center server, and more
- Microsoft Describes Next Xbox SoC
- Catching heat for partners’ shortcomings
- Overcoming the embedded CPU performance wall
- Optimizing high-performance CPUs, GPUs and DSPs? Use logic and memory IP
- Evaluating value-oriented x86 CPUs: The price of falling prices