Bus Mastering Explainedby Anand Lal Shimpi on March 15, 1998 7:44 PM EST
- Posted in
The people at Intel aren't idiots, simply put, they aren't...and they are smart enough to realize that by releasing Busmastering Drivers that would falsely inflate benchmark scores they would be asking their competition to do the same...so why bother? If you take someone's wallet knowing that they will take yours in return, why would you instigate the situation by making the first move? All rational thought leads us to believe that Intel had some ulterior motive in doing so, did the industry giant actually want the competition to blindly follow their lead in releasing benchmark enhancing Bus Mastering drivers? If so...for what purpose? All of these questions will be pondered and hopefully answered later on in this article, but first let's discuss Intel's first experience with chipsets, from a marketing standpoint that is.
Say you own a decently sized microprocessor manufacturing company, for the sake of argument lets call this company Intel ;), now you and your company have great potential...who knows, you may someday become the world's largest microprocessor manufacturer. In realizing your potential, you also realize that you must work to fulfill that potential, but if you have THAT much potential why stop at microprocessors? Wouldn't it be more profitable to have ever single computer running on an Intel based processor on an Intel motherboard featuring an Intel chipset and even furthermore using an Intel PCI ISA IDE eXcellerator (PIIX for short)? Definitely. Well, consider that Intel's dream...much like Microsoft's "a computer in every household running Microsoft applications" dream, Intel wanted to dominate the hardware world, not only with the processors but with their motherboards, chipsets, and their now popular PIIX chips.
While this is a difficult task to achieve, Intel decided to try their hands at it, they produced fairly successful motherboards, widely used chipsets, and even managed to manufacture one of those PCI ISA IDE eXcellerator chips that quickly caught the eyes of the industry. This was all taking place during the days of the early Intel 430FX or Triton I chipset, at that time Intel had little or no competition in the chipset market and common sense tells us that if we have no one competing with our products whatever we manufacture will be purchased by a buyer with no other choice than to purchase our products. Early on Intel had very little if any competition in this market, it wasn't really until Intel began producing chipset solutions that didn't fulfill all the needs of the buyers that companies like VIA Technologies and Silicon Integrated Systems (SiS) started stealing the lime light from the tyrannical powers of Intel. History has shown us, that when threatened, even the strongest powers stoop to new levels in order to eliminate the competition, and we are finally beginning to see the results of some of those efforts today with Intel's Busmastering Drivers.
Let's first discuss the installation of the Bus Mastering drivers provided by the various manufacturers, particularly the differences between the installation procedure of Intel's v3.01 drivers and those from SiS and VIA. In order to install Intel's Busmastering Drivers all the user is required to do is run the setup program, click that he/she wants to install the drivers, answer yes to a few questions and reboot the system, its as easy as that. When the system boots up for the first time (if it boots up that is) after installing the drivers Windows 95 will report finding new hardware, specifically a Primary and Secondary Busmaster IDE Controller. In the event Windows asks you for the location of ideatapi.mpd you need to provide it with the path C:\Windows\System\Iosubsys (where c:\windows is your Windows 95 directory). One more reboot is required for the newly installed Busmastering Drivers to take full effect, but other than that you are finished. Notice anything peculiar? Well, the purpose of these Busmastering drivers is to accomplish one major thing, and that is enable software support for DMA Transfers...however did you ever actually chose to enable this? No, of course not. All you did was install the drivers, and you never told your system that you wished to enable DMA Transfers, so in this case your system was told by a third party (Intel's drivers) to automatically enable these transfers. But who is to say that the drivers did just that and nothing more?
Keep that in mind while we move over to the SiS and VIA Installation methods for their Busmastering Drivers. Both the SiS and VIA Busmastering Drivers are installed the same way as Intel's however they do give the user the choice of enabling DMA Transfers during/after their installation. A point in case would be the installation of VIA's Busmastering Drivers, VIA's chipsets are very popular in FIC motherboards however they have also been exposed to Shuttle with their elusive HOT-603 and EPoX with their AGP P55-VP3. The installation process for the VIA Busmastering Drivers is much like that of Intel's drivers, however it contains one extra step. After running the setup software provided by VIA or your motherboard manufacturer and rebooting for the last time the drivers have been theoretically installed. As mentioned above, VIA's setup utility allows you to enable/disable DMA transfers on particular drives/devices attached to your system, and by default DMA transfers are enabled on ALL channels. But is this what really gives you the performance boost under Winstone/Winbench? Officially, yes, in reality...well...see for yourself.
Win95 Performance - AMD K6/233
|32MB RAM||64MB RAM|
|DMA Transfers Enabled||48.3||55.9|
|DMA Transfers Disabled||48.8||55.2|
Well...a 0.5 point improvement with 32MB of RAM, and a 0.7 point improvement with 64MB of RAM installed when DMA Transfers were enabled...not a huge performance increase as one would expect, right? Now let's take a look at those scores in comparison to the same system using the Busmastering Drivers included with Windows 95 Service Release 2.
Win95 Performance - AMD K6/233
|32MB RAM||64MB RAM|
|VIA Drivers - DMA Transfers Enabled||48.3||55.9|
|VIA Drivers - DMA Transfers Disabled||48.8||55.2|
Notice something odd? Well, with 32MB of RAM, the Microsoft drivers perform just about on par with the VIA drivers (with DMA Transfers Enabled or Disabled) however with 64MB of RAM the gap between Winstone scores grows to a whopping 6.2 points. These scores can be put up to your own interpretation, however here is the obvious question: Why is the performance of the test system ALREADY inflated when using the VIA BM Drivers, even when DMA Transfers are disabled? The obvious answer to that obvious question? VIA's drivers are falsely inflating your benchmark scores, REGARDLESS of whether or not you choose to enable DMA Transfers. By giving you the option to enable DMA Transfers VIA is in effect saying that they are in no way aware of the possibility that their drivers can inflate benchmarks, they are even willing to allow you to disable the DMA Transfers to show you that they are sincere in their claims.
How does this all relate to Intel and their drivers? Intel doesn't provide you with the option to enable (or disable) DMA/PIO Transfers, was it that Intel didn't realize that the competition would also manufacture their drivers to falsely inflate benchmarks...or was it that Intel wanted the competition to follow their lead blindly? Consider this a game of chess, Intel has positioned their pieces in such a manner, that the competition only has one option but to make a particular move, in this case, follow Intel's lead in order to keep that competitive edge. If a buyer goes out to purchase two different motherboards he/she will most likely place a large part of his/her decision on the benchmarks the two boards managed to achieve, regardless of whether or not one motherboard was tested using drivers that didn't falsely inflate the scores. For that reason VIA (and the competition) had no choice but to follow Intel's lead.
The question as to why Intel would do this has yet to be answered, however something such as this cannot be approached by looking at the immediate effects. Intel causes the competing chipset manufacturers to release Bus Mastering drivers that produce the same inflating effect, so what? We'll have to closely watch what develops with the next generation of benchmarking utilities, Winstone/Winbench 98...how do the Bus Mastering Drivers affect them? The fine line between Intel's drivers and those of the competition may be accented or smeared in the months to come, time can only tell what motivated the industry giant to make such a move...that time will come very soon.
Post Your CommentPlease log in or sign up to comment.
View All Comments