In early June of this year, I got tired of my wife's suffering with her now-ancient 120 MHz. Pentium system. After considering my options, I decided to build myself a new machine and hand down to her my own 450 MHz. Pentium II system. That would provide a far more usable system for her and provide me with a fair amount of new horsepower as well. As a developer of software and a video game enthusiast, I'll take all the speed I can get (grin).
Selecting the components was relatively simple. I read quite a few on-line reviews and settled finally upon an ASUS P3V4X motherboard for the new system with a 933 MHz. genuine Intel Pentium III CPU. In retrospect, I should probably have gone with an Athlon system, but there remained open some serious questions regarding the compatibility of the Athlon line and the latest and greatest NVIDIA GeForce video card on the market. All of the hardware sites I frequent had nothing but glowing comments about the ASUS motherboard, and I found virtually nothing negative about it elsewhere.
Thus, I ordered the various components and assembled them with nary a hitch into a functional system. The very first time I powered it up, it came on-line and just worked. I installed Windows 98 and got all my devices working, then spent several hours applying all the various patches and so forth available from Microsoft, et al. After installing some of my software, I decided to start running benchmarks to make sure that everything was working smoothly.
Naturally, it wasn't. The numbers I was seeing with my benchmarks were consistently dramatically lower than what the various hardware review sites were reporting, and I couldn't understand why. For example, I was using the Quake III Demo from Id Software to gauge the gaming performance, and while Anandtech, for example, was reporting 119 frames per second (FPS) with a 733 MHz. PIII CPU and a lesser video card, I was seeing a maximum of around 80 FPS with my 933 MHz. PIII CPU and GeForce2 video card! Great. I paid a premium price for the 27% faster CPU and other cutting-edge components to get 67% of the performance of what should be a slower system. Just great. I wish I could say this was atypical.
So, I did what any gamer would do: I started tweaking my system like a mad fiend. I spent at least a full 24-hour day trying different combinations of BIOS settings and did manage to squeeze another few FPS from my hardware. By tweaking the memory timings, bus timing parameters, caching and so forth I was able to up my results with Quake III Arena to around 90 FPS. Obviously, something still wasn't right. Along the way, I also discovered an interesting incompatibility that I'll discuss in more detail later.
For the moment, I began to wonder if I'd somehow been the recipient of bad components (deliberately or otherwise). I went to my local Fry's Electronics and purchased another motherboard and CPU. I replaced components one at a time until I was satisfied that my original components were good. Fortunately, Fry's has a generous return policy, and I spend enough money there as it is that I don't feel bad at all about making use of that policy when in the process of tracking such a problem. For what it's worth, I'm grateful every time I get into one of these messes that I can always count on Fry's.
Eventually I got so frustrated that I sent email to some of the hardware review sites I had trusted, asking if they could provide more details about the configuration of their testing setup. The responses got me thinking along lines that led, ultimately, to the real performance-stealing culprits in my system.
I got so frustrated that I eventually pulled all of the cards out of my system, save for the video card, and re-ran the benchmarks as I replaced each card one at a time. What I found surprised me greatly: my Creative Labs SB Live! sound card was stealing roughly 20% of my system's performance. I discovered this when I pulled that card out of my system and my frame rate shot up to 109 FPS. I realize that a sound card is going to need some of the system's power, but that seemed ridiculous to me, so I called Creative Labs.
After speaking with representatives Terry, Tyler, Mike and James, the situation was unchanged. They all assured me initially that my suspicions were correct; there was no way the sound card should be eating that much of my system's performance. And yet one after another they had to agree that my system was configured correctly and operating as well as could be expected. Ultimately, Creative Labs support could not explain what I was seeing, and they concluded that, in fact, that 20% must be the price of running one of their sound cards.
I must observe, however, that Creative Labs technical support put forth a great effort. Despite the fact that the problems might have been with my motherboard, the chipset on my motherboard or other components in the system, they didn't try to pass the buck even once. Rather, they treated me as a customer who was having a problem with one of their products. That may not seem like much, but anyone who spends a lot of time with technical support will likely realize how rare that approach really is.
Despite my long history of using products made by Creative Labs, I decided to vote with my dollars and find an alternative. After some reading, I settled on acquiring an Aureal SQ2500 sound card, which is based on the A3D Vortex2 chipset. While I've usually bought sound cards from Creative Labs, I have always admired the A3D approach—particularly the A3D v2.0 API for positional audio which is, in my estimation, technically superior to Creative's own EAX specification. I bought the card over the Internet on the cheap (about $45 at the time) and hoped it would solve my performance problems when it arrived.
It didn't. In fact, it wouldn't even work with my system. To be more specific, while the Aureal SQ2500 card was installed in my system, I experienced random system freezes quite frequently. The system would not run for more than twenty minutes without freezing completely, generally while a sound was playing. Unfortunately, I had picked precisely the right time to buy an Aureal card; i.e., the company went out of business just about the very day I received it. Naturally. That's my luck: lots of it and all bad. Needless to say, technical support was nonexistent. After doing a great deal of digging on my own, I came across the following:
For audio cards that use A3D audio chip there appears to be an incompatibility that causes the KX133 system to deadlock when system traffic is heavy. The reason for this is that the A3D chip issues one memory read request cycle to the North Bridge and expects a grant within 4 mico-sec of latency time. If PCI TRDY# is not asserted after 4 mico-sec, the A3D hardware will de-assert the FRAME# and REQ# to terminate the operation. However, the A3D audio software driver cannot acknowledge the termination and continues waiting for hardware response. Hence the system hangs. This issue will only occur during heavy system traffic when the North Bridge cannot grant access to A3D audio chip within the 4 mico-sec latency time. There is no known work-around for this issue, and the principal advice is not to use A3D-based sound cards with the ABIT KA7 or other KX133 based motherboards. Source: VIA Application note Serial #AN137 Date 07/05/00. This information found at BrokenPixel.
When I called Creative Labs about my discovery, the support engineers with whom I spoke said that this sounded like precisely what must be happening in my case. I don't know whether they said this because they believed it, or because they simply wanted to get me out of their hair and blame the problem on my motherboard chipset. Given their performance to date, however, I will charitably assume the former. With all the problems I'd had, I sent the Aureal card back to the vendor and bought another SB Live! card. Oh, and I made sure to purchase it from Fry's Electronics—see, they do get their money!
I actually wasn't surprised to see the name 'VIA' turn up because I'd seen it already once before. As I mentioned earlier, I ran across some interesting incompatibilities while initially diagnosing my system. All of them were related to the VIA chipset on my motherboard. Since the first day I put that motherboard in my case, I was unable to set my video card to 4x AGP. Or to be more precise, I could set it to 4x AGP, but this would result in an immediate system freeze whenever I tried to run a 3D application. Upon further checking, I discovered that I wasn't alone. In fact, entire on-line forums exist for people to discuss the problems they have with their ASUS P3V4X motherboard, and all of them that I visited featured numerous complaints about incompatibilities with the VIA chipset, not the least of which being unstable 4x AGP support.
I figured I would throw my own hat into the ring and go straight to the horse's mouth: I contacted ASUS. Not surprisingly, their "support" consisted of denying any problems and claiming that insofar as any problems may exist, ASUS is surely not responsible for them insofar as they stem from the VIA chipset This strikes me as completely ridiculous. To illustrate, let me ask this: how many of you are willing to accept a car dealer's disavowal of any responsibility for a car you've bought from them being incapable of using the top gear in its transmission because said transmission was assembled by another company? Obviously, no one would be willing to accept this. So why should I accept it in the computer industry?
To make a long story short, ASUS would do absolutely nothing for me. They suggested that I contact VIA directly. This was quite a fun chore in itself as the only telephone numbers I could find initially were for Taiwan, and all of my email messages to VIA support went unanswered. Eventually I managed to track down a US telephone number and gave it a try. Not surprisingly, I got a similar load of crap from VIA. They claimed that the fault was either with ASUS or with Microsoft and would do nothing for me. Big surprise. It didn't matter to them that I could get 4x AGP working with my system if I was willing to use an extremely old set of their chipset drivers, sacrificing about 50% of my system performance in the process. Nope. That must not matter. It just can't be the fault of their chipset or drivers. Oh, and go away, kid, you bother me.
Eventually, I gave up trying to optimize the system. I never did get the kind of performance from it that I should have. That system is presently functioning as a server for which it is more than adequate. It's a good machine, but it isn't what it should have been. And this is due largely to the culture of denial that exists among hardware and software vendors. "It isn't our fault" is more than just a motto for them, it's practically a religion. And until customers have genuine alternatives and vote with their money, nothing is going to change I'm afraid.
06/30/2000