Classic Computer Magazine Archive COMPUTE! ISSUE 151 / APRIL 1993 / PAGE 114

Diamond Technologies 486DLC-40, MicroExpress ME 486DLC/40. (microcomputers) (Hardware Review) (Evaluation)
by Richard C. Leinecker

At the heart of these two powerhouses is one of the latest Intel-compatible chips to hit the scene, the Cyrix 486DLC microprocessor. Operating at 40 MHz with 4MB of memory, a 120MB hard drive, and Super VGA graphics, these two units will satisfy the requirements of practically any software on the market.

I was relieved when neither one came shoehorned into a low footprint case. The MicroExpress minitower case and the Diamond Technologies desktop case let me have a field day installing and swapping cards.

There was a noticeable difference in the feel of the two. In general, the MicroExpress felt better. Its keyboard has a superb touch, the included Z-NIX mouse is sleek and comfortable, and the buttons on the case feel solid and durable. In fairness to the folks at Diamond Technologies, theirs costs approximately $400 less.

Both come with video systems well above basic Super VGA. The Diamond Technologies is equipped with a Diamond Stealth VRAM.

These cards are competitive with every nonaccelerator Super VGA card. The MicroExpress comes with an AVIEW Super VGA card with GUI hardware graphics accelerator capabilities.

The 486DLC chip is functionally compatible with Intel's 486SX. Both of these processor types lack the built-in math coprocessor of the Intel 486DX. Cyrix provides a math coprocessor add-on, which, when installed, provides full compatibility with 486DX chips.

You have to ask yourself what advantage you'd gain by buying a 486DLC when 486DX compatibility is so important. One answer is price. The Cyrix alternative is less expensive, even with the math add-on. And many people don't need the math coprocessor functions for the applications they normally run.

There are some other factors besides the obvious. The 486DLC has a smaller processor cache. Intel 486s all have 8K processor caches. The cache stores processor instructions. Keeping 8K of instructions within the processor saves time because the processor doesn't have to go directly to RAM to get the next instruction; it's already within its cache area.

The 486DLC has a 2K cache, and performance is reduced as a result. Fortunately, the performance isn't reduced by a proportional amount. That's because instructions often cause a jump or call to a location so far away that the cache is invalidated anyway.

There is an inherent advantage built into the 486DLC: a faster integer multiplication command. This command is frequently used at the machine level and can dramatically affect performance. The best use is in the area of graphics programming. That's a good area to improve in light of Windows and the proliferation of other graphics-intensive applications.

The best design doesn't mean much if the performance isn't there. But I found both units were up to par. I routinely ran demanding applications from Windows and was satisfied. Mathbased programs like Mathematica performed at the same level as on my 486DX.

The MicroExpress has a 256K hard disk cache, while the Diamond Technologies unit has a 64K cache. Most of the performance differences between the two units resulted from this. I compiled identical programs on my own 486DX and these two units, and their performance times were within a few seconds of each other.

I write entertainment software using Super VGA graphics. Manipulating Super VGA graphics makes more demands on microprocessors than practically any other application type. Both of these computers performed at least as well as any computer in my lab. Part of that may be the fine video systems, but a large part of it is a result of the performance of the 486DLC.

Oddly enough, running Norton SYSIN-FO left me somewhat confused. All of the 33-MHz 486DXs where I work produced a Norton Index of between 70 and 74. The Diamond Technologies computer produced a Norton Index of 65, not too surprising in light of the smaller processor cache size. But the MicroExpress produced a Norton Index of 36. The MicroExpress's performance was good and not at all indicative of the low Norton Index.

Resorting to a benchmark test of my own making, I tested out the claim that the integer multiplication instruction was faster than that of the Intel chips. The 486DXs did 40 percent as many multiply instructions as add instructions. That's just about right, according to the Intel manual. The Diamond Technologies unit did 86 percent as many multiplies as adds, more than twice as good as the 486DXs. The MicroExpress did 100 percent as many multiplies as adds, 2 1/2 times better than the 486DXs.

A technician at Cyrix explained that memory systems and motherboards need to be optimized for the 486DLC chip; otherwise, they perform about as well as 386DX chips. The difference between the two 486DLCs is probably due to system design considerations. If you're considering a 486DLC, you'd be well advised to ask the vendor if the system was designed around the chip.

I continued with a benchmark test put out by Chips and Technologies. It measured the number of MIPS (Million Instructions Per Second) for different instruction categories. The results substantiated the low Norton Index I found for the MicroExpress. For almost every category the MicroExpress scored about the same as a 33-MHz 486DX, but lower than the Diamond Technologies.

I ran the PC Benchmark program. It goes through a battery of tests that time the instruction set. The results rated the Diamond Technologies best, the 486DX second, and the MicroExpress third.

The last test I ran was a 3-D benchmark program but out by Virtual Technologies. It goes through a series of 3-D graphics image manipulations. Bearing in mind that the video system has a lot to do with the results, I got the same rating order as with the PC Benchmark test.

Overall, the Diamond Technologies computer outperformed the MicroExpress and a 33-MHz 486DX. The MicroExpress didn't do as well as the 33-MHz 486DX. Given the same processor type, memory speed, and clock speed, that leads me to believe that the MicroExpress wasn't designed as well as it could've been. But since most of my usage involves disk access, the large disk cache made up for the difference. In practical terms, these three computers were neck and neck for real-life applications.

Your decision about whether to buy a 486DLC-based computer will depend on your needs. If price is important and getting every ounce of performance isn't, this might be an alternative to the more expensive Intel 486DX-based machines. After using these for six weeks, I'd recommend them as viable alternatives.