UniPod speed testedBy Andrew Hill and Chris Williams. Published: 2nd Jun 2004, 21:08:53 | Permalink | Printable
IDE and ethernet with a need for speedAnnounced at the very end of March, the UniPod is the new combo podule from STD that features two USB ports, an IDE interface and a 100MBit ethernet network port. Seeing as the UniPod was touted as "the only RiscPC podule you'll ever need", we've spent the past week or so speed testing the upgrade to see how it fares.
Now, this article isn't intended to be an in-depth study, nor is it a comprehensive review of the product - it merely looks at how fast the UniPod's IDE and network interface can transfer data, compared to existing hardware. The data collected and presented below will vary depending on individual setups, but the figures should give an overall indication of performance. As always, we've tried our best to be as fair and accurate as possible, but if you spot any anomalies, feel free to point them out.
The testing begins
A RISC OS 4.39, StrongARM powered Kinetic RiscPC, with a Radeon Viewfinder and 192MB of RAM fitted, was used to test the UniPod, while Dave Holden's HDspeed application was used to test the speed of the podule - a copy of which can be found on the VirtualRiscPC CD. HDspeed employs a number of techniques to test a hard disc and the interface between the hard disc and the host computer, and we've presented the most important results below. As we kept the hard disc under test and the host computer constant, the performance of the interface will be apparent. Incidentally, the hard disc used was a 60GB Hitachi Deskstar 160GXP.
Firstly, a few notes on what each test does and what the outcome represents:
- Sequential byte access
- There are various approaches used by software when saving and loading your information to and from storage media. The information can be saved as a stream of write operations, by saving one byte (or character) after another in a sequential manner. It's not terribly efficient, but it can be used when saving information that is being constantly updated or is generated 'on the fly'. The sequential byte access test investigates the interface's performance in this area and reports how much time each process took in centiseconds - where one centisecond is a hundredth of a second.
- Block access
- Another method of saving and loading information is to transfer the data in a block. This is more efficient, but requires the application to have all the information together beforehand. This is the preferred method when saving and loading large files like images and documents, so it gives a good indication of loading and saving speeds in the real world. The block access test investigates the interface's performance in this area and reports the transfer speed in kilobytes per second.
The HDspeed testing was first undertaken on the RiscPC with the hard disc attached to the UniPod IDE interface. Then for reference, HDspeed was run again on the RiscPC with the hard disc connected to the internal Acorn IDE interface and again with the hard disc connected to an APDL Blitz IDE card. Next, a Dell Optiplex GX260 1.8GHz PC, with a 100/1000MBit network card, was connected via a cross-over cable to the RiscPC's 100MBit network interface (NIC) on the UniPod. HDspeed then tested transfers to and from a LANMan mount to briefly investigate real world network transfer speeds.
Finally, before we even begin, one of the most limiting factors is going to be the RiscPC's legacy podule bus, which by design can achieve a maximum data throughput of approximately 6100KB per second.
|UniPod IDE||Internal IDE||Blitz IDE||UniPod NIC|
|Sequential byte access||(centiseconds)|
Essentially, the UniPod IDE interface sustains around 3.5MB/s for saving files in blocks, and 3.9MB/s for loading files in blocks. The network interface was shown to achieve 1.3MB/s when saving a file via LanMan and 1.7MB/s when loading a file via LanMan. Other UniPod users have confirmed that they also get around 3.5MB/s when load and saving files to disc, although STD told us that "the environment in which a card runs is obviously very significant for speed tests".
During the course of our testing, our results were initially much lower than anticipated as STD expected us to get at least 4MB/s when reading and writing to disc. As a result of our findings and the following investigation, STD discovered a 'safety mechanism' in the UniPod system that forces attached hard discs to run in a low PIO mode, which limits the data transfer speed - we're told that the UniPod documentation is in the process of being updated on how to ensure that your hard disc transfers data at an optimum speed. Plus, a newer version of the not entirely bug free UniPod IDE hard disc formatter and configuration tool, IDETool, and new USB firmware should also be available soon.
The UniPod card tested, which incidentally is a developer model, also includes an internal USB header which allows you to use internal USB devices, or to connect USB ports to other parts of the computer; although the device supports a maximum of two USB devices. The IDE connector is also keyed, which means that 80-pin cables that have solid pins to ensure correct keying, can be used without resorting to manual drilling.
Pushing to one side the arguments against why people should indeed bother to upgrade legacy computers like ten year old RiscPCs, the UniPod does offer IDE access that outperforms the internal IDE interface, whilst providing two other important interfaces: USB and fast ethernet.
Update at 15:56 3/6/2004
It should, of course, be noted that the RiscPC's Kinetic card disables DMA operation for the podule cards shown above. Therefore, if you're a Kinetic RiscPC user, the results above should be what you'll get. If you're a non-Kinetic RiscPC user, then you'll see much faster speeds, in the order of 25 to 50 percent faster. When Andrew gets his StrongARM card, we'll post DMA enabled results.
Previous: Iyonix mobo spotted in the wild
Next: RISC OS Adjust Review
DiscussionViewing threaded comments | View comments unthreaded, listed by date | Skip to the end
Please login before posting a comment. Use the form on the right to do so or create a free account.
Search the archives
Today's featured article
Spam fighting apps reviewed
20 comments, latest by semore439 on 18/4/04 1:56PM. Published: 11 Jul 2003
Dynamic recompiler started for RPCEmu
Feedback is appreciated, says coder Tom Walker
Discuss this. Published: 27 Mar 2007
News and media:
RISCOS Ltd •
RISC OS Open •
MW Software •
Advantage Six •
CJE Micros •
Liquid Silicon •
Chris Why's Acorn/RISC OS collection •
The Register •
The Inquirer •
Apple Insider •
BBC News •
Sky News •
Google News •