JDR Computer Products and Electronic Components

Compendium

Processor Upgrades

 

  1. OverDrive Overload
  2. 486 "Vacancy" Slot
  3. Improved Math Co-Processor Performance
  4. Math Co-Processors

<Memory :Table of Contents : Motherboards>


OverDrive Overload
From Catalog 27, page 59

I'm confused, and I'm supposed to be right on top of the situation. It doesn't help that several competitors are rushing competitive products to market with great fanfare and flourish.

OK, I got that out of my system! Now let me see if I can simplify the OverDrive questions.

First, all OverDrive processors run at twice the speed internally of the chip they replace.

Second, there are OverDrive processors for DX systems and OverDrive processors for SX systems.

Third, if a motherboard has a socket for the OverDrive processor, you can just install it. If there is no OverDrive socket, you must remove the current chip first.

And last, most OverDrives have a math co-processor, which makes them the functional equivalent of a 486DX2.

Time for an update....

The 486DX2 processor runs at twice the speed of the external oscillator, and the 486DX4 runs three times as fast as the oscillator. This means that a motherboard with a 33MHz oscillator should be able to use a 486DX-33 or a 486DX2-66 or a 486DX4-100 (really 99).

However, before you run out and try to buy a 486DX4 for your current machine please be aware of the following.

NOTE: The 486DX4 processor requires, demands, and won't work without a 3.3 volt supply voltage. Since nearly all motherboards built for the 486 supply 5 volts to the processor, the 486DX4 will burn up if you install it on one of these boards.

Back to the Top


486 "Vacancy" Slot
From Catalog 20, page 81

Ever since Intel began running ads showing a "vacancy" on 80486SX motherboards, people have been asking what's missing. Why would a smart company like Intel promote a piece of missing silicon? Is this another piece of vaporware or what?

Well, much as I dislike the approach, I have to admit that they have a good idea. From both a marketing and end user view, it makes good sense. Here's what they are doing:

As a proportion of total PC users, the number who need or use a math co-processor is fairly small. Recognizing that, Intel has chosen to offer a 80486SX without the math co-processor that is standard in the original 80486.

Initial reaction to that move was generally negative because the first 486SX's were just crippled 486's. Naturally people felt it was wasteful and somewhat deceptive. In reality, it was just a transitional stage until Intel could begin to deliver the real 80486SX.

Manufacturing the 80486SX is less costly than a standard "full feature" 486 because less circuitry and silicon are involved. In a business where one error on a chip makes it unrepairable and useless, lower fallout can be expected from a simpler layout. And therefore, prices for a 486SX can be lower.

Over time, many PC users who don't currently require a math co-processor may find it to their advantage to add one to their machine. Many Graphical User Interfaces (GUI's) require considerable computation to place displays on the screen, and other new uses for PCs will generate their own demand for a co-processor.

The "Vacancy" is there in case you later need a math co-processor, or in case you want to speed up your already fast computer.

Soon, Intel will begin to deliver a 486 chip for that vacant socket that will incorporate a "clock doubler." This means that a 486SX running from a 33MHz clock now, may later accept a 486 running at an internal speed of 66MHz.

Externally, the clock doubling will still operate all of the CPU's peripherals at normal speed, so no other board changes will be necessary. The net improvement to be gained by filling the "Vacant" socket should be on the order of 70%.

As for prices for CPUs continuing to fall, mostly due to increased competition and somewhat better yields, the cost to fill that vacant socket will continue downward. If you don't need a co-processor today, consider the buy now, wait for a better price later approach. It makes sense to me.

As I reread that last paragraph, all I can think is, "Ain't that the truth".

And, prices will continue to decline as newer, better, faster chips come on line. Just don't be like the guy I was talking to several months ago. He is still using an original 4.77 MHz IBM PC with two full size low density floppy drives and a monochrome display.

He can afford the absolute best that the industry has to offer, but he is waiting. Every time he thinks about buying a new PC, he peeks into the future and decides he couldn't be happy with today's best, he wants what is coming next. As a result, he has never run Windows, OS/2 or even a WYSIWYG word processor. When he does buy, if he ever does, he will find himself far behind the curve, and he may never catch up.

P.S. I didn't make this guy up. He really exists, and at one time he was working with leading edge equipment. Back then, 13 years ago, he helped me lead a large group of people who were just learning about PC's. Today, any of that group probably knows more about getting the best from a PC than he does.

Back to the Top


Improved Math Co-Processor Performance
From Catalog 15, page 81

If you haven't already guessed, I depend on my computers a great deal. They remember everybody's phone number and dial it without error, keep track of my bills, check my poor spelling, and provide my livelihood in numerous ways. In addition to writing these columns, I also do a great deal of CAD work while designing PC cards.

In my experience, the fastest computer available is never fast enough. I always want something faster. With all the recent competition in the math co-processor market, it was time for my 10 MHz 286 machine to get a boost.

My problem (indeed most everyone's), was that a 20 MHz co-processor would still only run at 6.66 MHz in that computer. The math co-processor usually runs at 2/3 the speed of the 80286. The solution: provide the conditions necessary to run the 20 MHz math chip at its rated 20 MHz, without affecting the rest of the system and causing errors.

The Second Stage accelerator is a result of that quest for speed. By placing it in the math chip's socket, and putting the co-processor on top, you can run your math functions at full speed.

Times do change things. The Second Stage has been discontinued. 8088 and 80286 motherboards are on the "rare and endangered" list. Single floppy drives went the way of the Dodo bird, and many new programs won't run on a monochrome, color, or EGA display.

Did you get caught with obsolete equipment you couldn't sell in time? My advice, hang on to it, it may become an antique in a few decades!

Back to the Top


Math Co-Processors
From Catalog 10A, page 4

"Do I need a math co-processor?" and "How much will a math co-processor speed up my work?" are questions we're frequently asked here at JDR. Since they are expensive options, you need to know how you will benefit from their addition to your computer.

If you are running programs that specifically state support for a co-processor, you will realize significant improvements in speed from its addition. However, if the program doesn't support a co-processor, then no matter what its speed, the program won't run any faster.

Some programs benefit more than others; in particular, those that make heavy use of floating point, trigonometric, logarithmic, and exponential calculations show the most radical improvement. Therefore, CAD, spreadsheets, some databases, and Mandelbrot programs are frequently written to use a co-processor if it's available.

How much your application will speed up depends on the ratio of time spent on math calculations versus other operations. A 3 to 10 times improvement is not uncommon, and certain specific operations are even faster.

The co-processor you need is determined by the type of processor and the speed of the co-processor clock source. For 8088/86 and 80386-based machines, the general rule is to use an 8087 or 80387 of the same speed as the processor, i.e. an 80386-25 requires an 80387-25. 80286-based machines frequently use a co-processor running at two-thirds the processor speed, i.e. an 80286-12 requires an 80287-8. When in doubt, consult your manual or call our Technical Support Department.

Back to the Top