Talk:Am5x86

Tip: Anchors are case-sensitive in most browsers.

This article links to one or more target anchors that no longer exist.

[[CPU cache#Details of operation|write-back]] The anchor (#Details of operation) has been deleted by other users before.
[[CPU cache#Details of operation|write-through]] The anchor (#Details of operation) has been deleted by other users before.

Please help fix the broken anchors. You can remove this template after fixing the problems. | Reporting errors

It's great to see the lesser-known processors being filled out at last. Two things:

I removed "it went into very few new computer systems" because this is untrue: it was a major part of the entry-level market for quite some time.
On cache, the much earlier Cyrix 486SLC was the first 486 CPU to have had write-back cache, and most of the later 486 designs did too. The common Cyrix 486DX chips were all write-back (only the older DLC was write-through), one of the two AMD DX/4s was write-back, and the later Intel DX/4 had 16k write-back. Tannin

I'll take your word for it, but I almost never saw a 5x86 in a new desktop system, just laptops. I guess it depended on where you were. I saw it in plenty of laptops, though. I had a few systems running these as recently as the turn of the century, they were as good as AMD said they were. I even got one going at 200 MHz, so I know it was possible. John

Computing trends can be quite different in the different parts of the world alright. VIA's C3 has had a reasonable slice of the CPU market share in Asia, but is largely unheard of & unavailable elsewhere.

Cool to see people running these all at the higher speeds... Years ago I got a 133 one going at 160mhz after some messing about with my old 486 board (inherited from the family machine - I was still at school and too poor to afford any *real* upgrades, or a pentium..) and thought myself a hardware god :D hadn't seen anyone else do this til now. Was quite satisfied with the performance - a good step above the 25s, 33s and 66s that were most people's 486 experience, and far more cost effective than the slightly laughable Intel Overdrives - and eventually sold it to an even more cash strapped friend as an entry level multimedia (!) pc (a "P-90" I think I talked it up as) when i finally reached the heady heights of a Cyrix PR-166 (not really very much more swift, I think most of the benefit came from the motherboard & memory being 66mhz/64bit vs 40mhz/32bit rather than the CPU being faster!). They went all the way up to 200? Eeeeeexcellent! - tahrey 20/1/07

I read somewhere that AMD produced a 200MHz Am5x86 especially for embedded systems. Can anyone confirm this?

They still do, but it's a special-order part and not something they advertise. The 5x86 is still a very popular embedded CPU, but is usually sold as a regular 133 MHz chip even today. Jsc1973 (talk) 06:06, 13 August 2008 (UTC)[reply]

The 5x86 die has been in production up to 2007 if I am right though not as a stand alone 486 cpu. In the end it was/is an embedded cpu running at nearly 500Mhz. Not that anyone ever used those in a "real PC" though. Even today its core is part of some embedded designs 01:01, 25 July 2011 (UTC) — Preceding unsigned comment added by Crass Spektakel (talk • contribs)

ADZ vs ADW

There were two major versions of the PGA version of this CPU, ADZ and ADW. According to the data sheet, the ADW chip ran cooler than the ADZ, and in my personal experience that was true. The ADW was the preferred chip for overclocking to 160Mhz and could tolerate running at a full 5 volts where the ADZ ran hotter and while 5 volts wouldn't immediately fry it, it probably could not tolerate it for long without extreme cooling.

I ran an ADW version on 5 volts for nearly a year with a cheap 486 CPU cooler, and it ran cooler than the SMT 486 SX33 (soldered to a PGA conversion board) it replaced.

There are two pins on these CPUs that can be shorted together to force it to run at 4x on boards with no clock multiplier support, or to force it to run at 4x instead of 3x or 2x on 40 or 50 Mhz bus speed. (I'd have to dig up the info, been a long time since I did that!)

The problem with running it at 4x50Mhz wasn't so much finding a VESA Local Bus videocard that could handle it, it was finding a motherboard with an L2 SRAM cache that could run at 50Mhz. I never found a board that wouldn't automatically disable the L2 cache at 50Mhz bus, which drastically hurt performance, so 4x40Mhz worked well enough. Perhaps replacing the typical 10ns cache chips with 8ns ones may have worked, but at the prices those chips sold for back then, ouch!

The ADW on 3.3 or 5 volts and the ADZ on 3.3 volts both ran cooler than the Cyrix competitor. I got a bad second degree burn and blister on a finger once when I accidentially brushed a fingertip against the heatsink on a Cyrix 5x86 in a PC that had just been turned off. Real world performance of the Cyrix VS the AMD on business apps (in an actual office I was the IT person in) was quite poor. The Cyrix PC was the first to be replaced by a shiny new Pentium 120.

There was no difference in runnign cooler or getting hotter between these CPU - it was just the ADZ specified to be able to work under more temperature-intensive circumstances. AFAIK ADW was specified to 60 or 65 degrees celsius whil the ADZ up to 85 degrees. --Denniss 12:52, 28 March 2006 (UTC)[reply]

Which should make the ADZ the more overclockable chip. It would be similar to the Athlon XP mobiles that were very popular a year or two ago. The mobile variants are capable of remaining stable with far higher temperatures, meaning that they ought to be able to maintain a higher clock speed than a regular desktop chip in the same environment. I've used several ADZs in the past and I don't remember 160 being a problem. I'm not positive anymore, but I think I did have trouble with an ADW. --Swaaye 22:09, 28 March 2006 (UTC)[reply]

What's VLB got to do with it?

I can credit the poor availability of 50mhz-bus 486 boards, as they were rare beasts indeed... but VESA Local Bus was dead in the water by the time even 100mhz 486s were common, let alone overclocking parts to try and keep up with Pentiums. So why is it given as a reason for the impracticality of 200mhz 5x86s? I mean, if you're still rocking a VLB graphics card and bus, you've got bigger performance concerns to worry about than raw CPU speed. Probably only got the ability to chuck 16mb RAM in, limited L2 cache levels and poor hard disk performance for a start. And a slow original CPU - getting up to 133mhz would be a perfectly adequate speed boost over your original DX-33 or DX2/50.

(I'm the guy with the 160mhz one waaaay up above - when I did that, I was maxing a by-then quite old 486 board from 1994... which had a maximum of 40mhz bus... plus three PCI sockets, 256kb cache, PIO4 hard drive access and some vague stab at DMA/bus mastering, and 64mb RAM capability. VLB was on the way out even when we bought it as part of our main PC, in an era when the P75 had JUST arrived and was very expensive; and the 5x86 overdrive wasn't bought until '96-97)

I may remove it, if I can figure a way of patching up the hole it'll leave. 193.63.174.10 (talk) 16:36, 11 February 2010 (UTC)[reply]

PCI is only specified for 16, 25 and 33Mhz and the 5x86 allows only 3x and 4x multiplier so with PCI your maximum is 4x33Mhz while with two VLB cards it is 4x40Mhz and with one 4x50Mhz. Overclocking PCI is NOT a good idea because its design parameters are a lot narrower. Even 37,5Mhz often where unreliable. About overclocking, mine ran at 3x50Mhz, 4x40Mhz, 4x44Mhz and 4x50Mhz but the 4x50Mhz where quite unreliable and bus speeds over 40Mhz required more waitstates in cache and memory, resulting in an overall slower performance in some cases. At 4x40 it outperformed PCI based systems in nearly every aspect. Having a quite fast and ultrarare VLB-VGA-S3Virge, a fast VLB-SCSI-Adaptec2840 and a ISA-Ethernet-100MBit "Corkscrew" 3c515 helped also. Crass Spektakel (talk) 01:19, 25 July 2011 (UTC)[reply]