Computer Answers update
29-Apr-96
The Cyrix/IBM 686 Processor
What is this new CPU and is it all it's cracked up to be?
Basically the 686 is a Pentium clone with a few differences. It is claimed to be faster than an Intel Pentium, but then they would say that, wouldn't they. It is true - it is faster that a Pentium if you pick the right Pentium to compare it with. It's faster than a slow Pentium, but slower than a fast one.
Before getting to the bottom line (it is better value for money) I'll just ramble on (as I do) about why it is different.
As far as I can see (and I've got the technical manuals), it has the following major differences:
- The Pentiums code and data caches have been replaced by one unified cache. This means it can vary the proportions of data and code which are cached on-chip. Is this useful? Perhaps.
- There is a seven stage pipeline on the 686, only five-stage on the Pentium. This should make a difference.
- The floating point instructions on the Pentium look like they'll be a lot faster than the 686 versions. Whether this makes a difference depends a lot on how much floating point you use. People either use very little or lots.
- The closest Pentium in price to the 686-100 is the P-100 (strange that). The 686 costs around £200, whereas the P-100 is about £180 (today, that is). However, the P-100 does this by multiplying its external clock frequency of 66MHz by 1.5 whereas the 686 doubles an external clock of 50MHz. What this boils down to is that the Pentium is working its external bus around 30% faster. The difference this will make to performance depends on how much is held in the on-chip cache, but in theory the Pentium wins.
You will be able to get 66MHz 686 chips in the future, which will run at 133MHz inside. There is also an option to triple the internal speed of the 686 (by raising one of its pins during a reset sequence), pushing it to 200MHz in the end - which is where the current plans for the Pentium stop. This isn't here yet.
To compare a 686 properly with a Pentium you would need to run a P-120 (with an external clock of 60MHz) at 50MHz. Then it would be doing the same job as a 686 at 100MHz. I haven't been able to try this yet, but if I can get the bits together I'll let you know.
The real question is: Will it deliver more performance for the same money? At the moment this looks unlikely to me. It costs more than the P-100 but has a slower external bus. I don't think the pipelining improvements will compensate for this and the FP performance (if it is as I suspect it to be) could spoil it for some too.
At the moment, my money goes on a Pentium-75. I never liked the P-60/66, which ran a 5V rather than 3.3V and got rather too warm. But a P-75 only costs about £80 and is really cool (in all senses).
The best bit is that the current chips will run quite happily at 100MHz if you jumper their external clock from 50-66MHz. .
Intel obviously don't guarantee a chip sold as a P-75 will run at 100MHz , but plenty of people have done this for a long time now without problems and saved about £100 each.
It may reduce the life of the CPU so doing so is entirely at your own risk, but who wants to keep the same processor forever? If it burns out after a year then buy another one - they'll probably cost about £25 by then. Maybe the 686 will have moved from .5 to .35 micron by then and be a lot faster too.
AMD are about to release something too.
.
© 1996 Frank J Leonhardt
Computer Answers Index
Later News
Earlier News