by Andrey Kuznetcov
03/03/2004 | 11:27 PM
USB 2.0 interface has become an inalienable part of any contemporary computer, but the device connected to this port will not always perform at its maximum. The reason for this phenomenon is most frequently the use of various USB 2.0 controllers designed by different manufacturers, which actually demonstrate different efficiency in real life. They are exactly the ones to blame in most cases, when your super-fast external storage device, for instance, doesn’t show the desires speed.
In this situation it is necessary to make sure that the USB 2.0 controller in your system will not act as a limiting factor for the performance of your external device. Since the manufacturers roll out new USB 2.0 controllers every now and then, it makes sense to compare the performance of the newcomers with the already selling models. This is exactly what we are going to do today: we will check out the performance and efficiency of the new USB 2.0 solution from VIA.
Not so long ago there appeared a new USB 2.0 controller from the Taiwanese VIA Company. It was VT6212. This solution is intended to replace its predecessor known as VT6202. We have already devoted a lot of time to the performance investigations of USB 2.0 controllers by different manufacturers and of course, we couldn’t leave out the new solution from VIA. Especially, since VT6202 appeared a little bit below our expectations in the previous test session.
So, what is so special about VT6212. In our case it was a PCI controller card. In order to give it a more detailed description I would like to draw a few parallels with the older product, VT6202. Just like its predecessor, VT6212 supports 4 ports (downstream). Here there is nothing unexpected. And regarding all the rest, I would like to point out a few positive changes. The manufacturer claims that the new controller consumes less power, boasts higher actual interface bandwidth and supports PCMCIA Cardbus, which makes it suitable for mobile and low-profile computer systems.
The controller owes its higher performance to four DMA Engines with pipelined control. Lower power consumption and lower heat dissipation are ensured by the support of Dynamic Clock Stop Control. The 32bit CardBus mode allows using PCMCIA interface as an additional USB 2.0 concentrator, increasing the number of devices to be connected to the controller.
The controller meets USB 2.0 specifications, supports EHCI (Enhanced Host Controller Interface), UHCI (Universal Host Controller Interface) and features Serial EEPROM. The microchip is manufactured in two types of packaging: 128-pin PQFP (VT6212) and 128-pin LQFP (VT6212L). VT6212 is manufactured with 0.22micron CMOS technology. The chip is designed to support 2.5V of power.
In order to figure out the actual efficiency of the new VIA VT6212 controller we decided to compare it with the predecessor, VT6202 (implemented as a PCI solution). As a reference we will also use the results shown by the USB 2.0 controller integrated into the ICH5 South Bridge from Intel, as this one proved to be the fastest of all USB 2.0 controllers we have tested so far.
We estimated the controllers performance by testing the fastness of the external Maxtor OneTouch HDD with 250GB storage capacity, which supports USB 2.0 interface. This drive features pretty high performance characteristics, which are definitely far beyond the USB 1.1 capabilities. Therefore, we get a beautiful opportunity to see what the controllers are capable of.
We used the following benchmarks for our today’s test session:
Our testbed was configured as follows:
We tested the controllers with the drivers provided in the operation system. Only for VIA controllers we installed “four-in-one” driver version 4.51. The drive was formatted in FAT32 and NTFS as a single partition with a default cluster size. In some cases described below we used 32GB logical partitions also formatted in FAT32 and NTFS with a default cluster. For Maxtor drive we wrote the data all over the work surface first to ensure correct results of ongoing tests.
We will start with the controllers performance in the DataBase pattern. With the help of this pattern we checked the HDD performance during the processing of requests with different queue depth and varying writes share.
In order to illustrate the controllers performance we built three graphs demonstrating the controllers efficiency under three types of workload (three types of queue depths).
The first graph demonstrates the controllers performance under linear workload laid onto the HDD (queue=1). There is hardly any significant difference between the three controllers here. Intel ICH5 falls a little bit behind VIA solutions.
The next graph shows how efficient the controllers are in case of queue depth=16. The situation hardly differs from the previous one. Both VIA controllers are almost equally fast and Intel ICH5 falls a little bit behind them.
The last graph shows the controllers performance with queue=256 requests. No comment is necessary since the picture is exactly the same as in the previous two cases.
During the performance tests in these patterns we check how well the HDD can cope with ordered requests with linearly growing address. The requests queue depth starts at 4 requests and then once per minute the size of the requested data block increases. The highest results in the tables below are highlighted with the blue color, the lowest results – with red color.
The Sequential Read graph shows very clearly that the controller integrated into Intel ICH5 performs best of all. VIA VT6212 outperforms it just slightly on small data blocks, but loses to the rival as soon as the data block size increases. At the same time, VIA VT6212 controller seems to be more efficient than its predecessor – VT6202.
The Sequential Write graph is very similar to the previous one. The general picture remained unchanged, but VIA VT6212 got much closer to Intel ICH5. At the same time, the gap between VIA products has grown significantly bigger as we shifted to large data blocks.
With the help of this pattern the benchmark imitates the workload typical of a regular workstation, which is limited at 32 requests queue. The results were taken for the drives formatted as a single partition, and as a 32GB partition.
This graph shows the controller efficiency during the HDD requests processing. The results dispersion is not that big, though the advantage of VIA VT6212 over the opponents is evident. VT6202 and Intel ICH5 show about the same performance, so it is pretty hard to name the winner.
The performance index diagram clears the situation. The best is VIA VT6212, then comes VIA VT6202 and the last one of the three is Intel ICH5.
The results obtained in a 32GB partition of the drive appear generally higher for all controller tested. Although the testing participants are still positioned in the same way. The most preferable one of the three is still the newcomer – VIA VT6212. Intel ICH5 is a little bit slower. The poorest performance of the three in this test belongs to VIA VT6202.
The performance index diagram shows perfectly the detailed results. The performance difference between the three controllers is negligible, and the positioning of the controllers tested remains the same as in the previous case.
With the help of a well-known WinBench99 test package we tested the performance of our controllers when the HDD was formatted as a single partition and as a 32GB logical partition. We ran the tests for both: FAT2 and NTFS file systems. In the tables below the best results are highlighted with blue and the worst results – with red colors.
Let’s start with the situation when we involve the entire hard disk drive:
In FAT32 file system we see an indisputable advantage of Intel ICH5. VIA VT6212 falls behind the leader but is significantly ahead of the predecessor.
The graph illustrating the controllers performance in NTFS file system show that the situation in High-End Disk WinMark remained unchanged compared with the previous case. Intel ICH5 is still the best, then comes VIA VT6212 outperforming its predecessor, VT6202. However, if we look at the results obtained in Business Disk WinMark, we will see that the controllers lined up in the completely opposite order.
Now let’s pass over to the case when we have only a 32GB partition involved in the test process.
On the graph for FAT32 you can see that the overall performance of our testing participants has grown bigger, but the positioning of the rivals remained the same as in the pervious cases: VIA VT6212 is behind the leader integrated into Intel ICH5 South Bridge, but is still much faster than the predecessor.
The use of NTFS file system doesn’t change anything. Intel ICH5 is an indisputable leader. Then comes VIA VT6212, and the very last one of the three appears the good old VIA VT6202.
Now it’s high time we tested the efficiency of our controllers in FC-Test, which differs from the other benchmarks by the realism of the testing environment. We ran the tests for NTFS and FAT32 file systems.
We will start the discussion with FAT32:
The results of file creation show that VIA VT6212 catches up with Intel ICH5. The gap between them is really small. However, the third competitor, VT6202, is much farther behind the leaders.
File reading diagram demonstrates that Intel ICH5 is considerably ahead of the rivals. Our today’s main hero, VIA VT6212 controller falls behind the leader but retains its advantage over the VT6202 predecessor.
File copy within a single partition again gives the laurels to Intel ICH5. VIA VT6212 is unable to catch up with the winner, but is considerably ahead of the VT6202.
In case of file copying from one partition to another VT6212 manages to reduce the gap to the leader, Intel ICH5. at the same time, VIA VT6202 is as usual the slowest.
Now let’s check the performance of our USB 2.0 controllers in NTFS file system:
The graph illustrating the controllers performance during file creation shows very clearly that VIA VT6212 runs almost neck and neck with Intel ICH5, but lacks just a little bit to catch up with it completely. VT6202 is considerably far behind the successor.
Reading the files from the hard disk drive when the data transfer rate usually increases pushes Intel ICH5 far ahead of the VIA solutions. At the same time VT6212 is much faster than VT6202.
File copy within one and the same partition doesn’t add anything to the situation. All the testing participants remained where they were. VIA VT6212 still can’t catch up with Intel ICH5, although the gap is not that huge at all. VIA VT6202 is still behind the two leaders.
Copying files from one partition to another hardly changes anything: Intel ICH5 remains the leader, then comes VIA VT6212, and the last racer is VT6202.
Of course, we should admit that VIA made a significant step forward having launched its new VT6212 USB 2.0 controller. Our performance investigation clearly proves this point. The advantage of VIA VT6212 over the predecessor is more than convincing in most tests. The real progress the new USB 2.0 controller will make compared with the predecessor is simply indisputable. It is especially evident in the benchmarks, which are close to real-life applications, such as WinBench99 and FC-Test.
Unfortunately, one of the tasks VIA engineers might have also set, namely to catch up in performance with solutions from Intel, hasn’t been fulfilled. Even though the new VIA VT6212 controller performed really close to the USB 2.0 controller integrated into Intel ICH5, the gap between the two is still there. Although, their actual efficiency is not that very different. Taking into account traditionally very flexible VIA’s pricing policy and a few specific peculiarities of the new solution, as get every right to guarantee that VIA VT6212 will be highly demanded by both: manufacturers and us, users.