The x86 CPU landscape of the 1980s and 1990s was competitive in a way that probably seems rather alien to anyone used to the duopoly that exists today between AMD and Intel. At one point in time, Cyrix was a major player, who mostly sought to provide a good deal that would undercut Intel. One such attempt was the Cx486DLC and the related Tx486DLC by Texas Instruments. These are interesting because they fit in a standard 386DX mainboard, are faster than a 386 CPU and add i486 instructions. Check your mainboard though, as these parts require a mainboard that supports them.
This is something that [Bits und Bolts] over at YouTube discovered as well when poking at a TX486DLC (TI486DLC) CPU. The Ti version of the Cyrix Cx486DLC CPU increases the 1 kB L1 cache to 8 kB but is otherwise essentially the same. He found the CPU and the mainboard in the trash and decided to adopt it. After removing the very dead battery from the Jamicon KMC-40A Baby AT mainboard, the mainboard was found to be in good working order. The system fired right up with the Ti CPU, some RAM, and a video card installed.
That’s when the excitement began, as although the mainboard is ‘Cyrix-aware’, its BIOS support is somewhat buggy. Although it technically will beat the living tar out of a 386, the Speedsys benchmark utility was crashing because the internal L1 cache wasn’t being enabled properly. Fixing the problem required an external Cyrix utility application. These issues were why so few people were interested in bolt-on solutions. The sentiment extended to Intel with their ill-fated Pentium OverDrive products.
For enthusiasts looking for a good deal, they were an exciting option, but Intel took offense to Cyrix barging into the x86 CPU market without a negotiated license. Cyrix instead utilized reverse engineering to make their own x86-compatible designs. This included their later 5×86 and 6×86 CPUs. After a series of lawsuits Cyrix was merged into National Semiconductor and later sold to AMD, who sold Cyrix’s latest designs under the Geode name before discontinuing it in 2019.
AMD didn’t purchase Cyrix. VIA Technologies picked up Cyrix, and Centaur with their WinChips within in the same year.
They both did: Cyrix was bought by Nat Semi, who then sold off most of it to VIA, but not the MediaGX, which stayed with Nat Semi, and eventually to AMD.
VIA never really used the Cyrix x86 designs, other than the early bits of the Cyrix III: everything later was Centaur-derived (who eventually got sold off to Intel).
So the “last descendant” of the Cyrix designs was the Geode by AMD.
Hence Centaur Hauls if you ever look into those processors. And the Centaur team was just let go recently by Intel or maybe VIA prior to the sale if I recall.
I had a Cyrix 486 DX2-66 for a gaming rig. It was pretty kick-ass for a teenagers budget.
https://www.x86-guide.net/en/cpu/Cyrix-486DX2-66-(Write-Back)-cpu-no1462.html
“Fixing the problem required an external Cyrix utility application. These issues were why so few people were interested in bolt-on solutions. The sentiment extended to Intel with their ill-fated Pentium OverDrive products.”
I had one of those Cyrix processors and was also underwhelmed by its performance. Then I found out that it was supposed to come with a floppy that contained a driver, or had to be used in a mainboard whose BIOS could enable it. But mine didn’t come with the tool and my BIOS was old and I couldn’t find any upgrade.
So, I believe with some pointers of the German C’t magazine, I wrote two little .com applications for MS-DOS through which I could enable or disable it. After that, my system zoomed. I uploaded my programs to FIDONET at the time. Long lost, I’m sure. I basically created them using DEBUG.EXE as assembler.
But those programs were very simple, just switched on or off the cache. With the official tool, you could also specify non-cacheable areas, etc.
Cyrix processors were garbage.
There’s a reason the company o longer exists.
As a tech I saw them in whiteboard workstations and heaven forbid whiteboard servers. HUGE mistake.
“One such attempt was the Cx486DLC and the related Tx486DLC by Texas Instruments.
These are interesting because they fit in a standard 386DX mainboard, are faster than a 386 CPU and add i486 instructions. ”
I think that describes it very well.
Not seldomly 80286/386SX and 80386 motherboards had interesting hardware features such as EMS.
They could provide EMS or UMBs without requiring V86 mode (used by EMM386, QEMM, 386Max etc). Real-Mode DOS caused less headaches, after all.
Having an upgrade processor that provides 486 instructions on such older, more intelligent motherboards was very appealing in the 90s, thus.
It wasn’t as good performing as a real 486 system, but still an improvement.
The little L1 cache helped to boost performance on motherboards with little L2 cache or no cache at all.
Yes, such boards had existed, too. 386SX or 286 chipsets often had no cache.
On such boards, a 486SLC could really keep these old chipsets alive for a few more years to come.
But even without improved performance, the 80486 was what had defined the DOS era and early multimedia era of the early-mid 90s.
I’m thinking of 486DX-25, 486DX-33 and 486DX2-66 here.
The the availability of the i486 instruction set was important in network appliances and on *nix systems of the time.
On plain DOS, though, a 80286 or 80386 was still usable, I admit.
Especially the 386DX-40 was popular among small BabyAT motherboards.
It provided a solid platform that didn’t cause much trouble.
No excessive power draw, no overheathing issues, ordinary +-12v, +-5v AT power rails without need for 3.3v.
An 486DLC or 486SXL was a simple upgrade to do.
It could provide performance similar to a 486DX-25.
Another advantage was that most 386 chipsets had “native” ISA support.
No Plug&Pray, no APIC or ACPI. No VLB or PCI bus. Just plain AT architecture.
Sure, EISA and MCA and OPTi boards had existed, too, but I’m thinking of the small, yet reliable BabyAT boards. :)
At that time I had a motherboard with a 386DX-33. I replaced the CPU with a Cyrix 486DLC, and also added a 387DX coprocessor. I was in University at the time.
My friend had the budget for an actual 486DX-33, but I didn’t.
We did comparisons, and of course his 486DX-33 was faster than my 486DLC/387DX combination. But all software that needed a 486DX with an FPU worked fine on my computer as well. And that was the goal.
A year later or so, I sold my computer and bought a computer with a 486DX2-66 P24D core, and zoomed by my friend’s computer. And a year or so after that, he bought a Pentium 60 system. And so we kept alternating. ;)
“Cyrix processors were garbage.”
I was raising a wife and child in the 90’s and was very “poor”.
I had a few Cyrix CPU’s and was happy I did.
I didn’t have daddy to buy me the best.
I remember putting a 486DX chip into a 386 PC back in the mid-1990s. But as it was a second hand chip, that may explain the frequent enough crashes.
Ah yes, there had been 486 mainboards with both an on-board 386 (say am386DX-40) and a free 486 socket.
They were affordable complete systems with an upgrade path.
Some may even had the 386 CPU socketed, so it could be removed and the board could work with either an 486DLC and an 486DX! :D
That’s a bit like 486DX boards with the OverDrive socket or 486SX boards with an i487 socket (the i487 basically was an full 486DX with x87 FPU).
Speaking of FPUs, some 386/486 maonboards had a Weitek socket.
The 386 boards usually had it as part of their 387 socket.