Cyrix 6x86 200 shows in CheckIt as running at 166mhz

[Taffer]

I recently acquired a small tower with a Cyrix 6x86 marked as 200mhz.

CheckIt, installed on Win98 on the HDD, says it's only running at 166mhz. My understanding of this chip is that the 200mhz is a "comparison Mhz" to a Pentium, and the CPU is actually running at some speed slower than 200mhz.

I found this chart which helps support this from https://www.cpushack.com/CyrixID.html#3


Can I just boost the clock multiplier up to 3x and enjoy 180, or are there multiple versions of the PR200 chip that become unstable at 180mhz? I'm a bit rusty on this era of computers but I remember once discovering a friends computer running his new processor well under what it was capable of.
Thanks.

 

[slewlimit]

As you can see from the "Bus" column, it depends on the FSB speed of your motherboard. CPUs in this era derived their clock speed from the FSB with a multiplier. At the common (at the time) speed of 66MHz and a 2.5x multiplier, you get 166MHz. The only way to see if it's stable is probably to try it and see what happens. Your motherboard or your CPU could turn out to be unstable with the higher multiplier.

 

[GiGaBiTe]

Yes, Cyrix used a "PR" scheme, which they later renamed to "GP" with the Cyrix MII. The PR and GP numbers were intended to be a representation of relative performance to Intel CPUs of the specified clock speed. Early on, those numbers were in most cases true in regards to integer performance (Cyrix had a much more efficient integer core), but it was not at all representative of floating point performance. Cyrix's FPU in their 6x86 and MII parts was largely unchanged from their 486 era chips, they had made a fatal error in judgement that most workloads in computers was going to continue to be integer workloads. They were blindsided by the rapid advance of 3D games like Quake.

But in addition to the weak FPU, Cyrix's chips tended to draw more power and produce more heat than competing chips from AMD and Intel, and tended to have stability issues. This was exacerbated by their penchant for weird bus speeds like 75 and 83 MHz, which caused other devices in the system to run alarmingly out of spec due to the nature of cascaded bus clocks used on motherboards at the time. An example would be that the PCI bus would be run at 41.5 MHz on a 1/2 divider with a bus speed of 83 MHz, and PC-66 memory would be run way out of spec, or significantly underclocked if the motherboard offered a 1/2 or 2/3 divider.

If you choose to run a Cyrix, I would recommend getting as close to a 66 MHz bus speed as you can, unless you have a later Super 7 board that can tolerate the weird bus speeds better.

I would also recommend keeping an eye on the CPU power regulators. Motherboards at the time often used really crap linear regulators that ran smoking hot and suffer thermal death, or early buck regulation circuitry that ran in the linear region and was equally as hot.

 

[Timo W.]

If you want a faster CPU, get a faster CPU. There is no sense in killing retro hardware early by overclocking it only to gain a fraction more speed.

 

[GiGaBiTe]

It's not really overclocking anything. Cyrix released multiple chips with the same PR number that ran at different bus speeds and core clocks.

 

[Taffer]

If you want a faster CPU, get a faster CPU. There is no sense in killing retro hardware early by overclocking it only to gain a fraction more speed.

I'm not looking to overclock or for a faster computer. I have approx 100 computers from the 1979 to 2001 era. Just wanted to learn about this Cyrix processor as I haven't run across a system fitted with one before. All my machines of that era are AMD or Pentium which have a single speed rating.

It's not really overclocking anything. Cyrix released multiple chips with the same PR number that ran at different bus speeds and core clocks.

Looks like I'll have to look at mine more closely and see if there is a way to ascertain which mine is. I'll see if there's a code on the bottom of the chip and pull the mobo model.

 

[Taffer]

In my quest to make this run at the correct speed, I may have found a clue. The processor is marked 75mhz bus, 2.9v. Mobo is A(r) PM8600. Will research.

 

[Plasma]

Your CPU is already configured correctly...

 

[Eudimorphodon]

I had a Cyrix machine back in the day, and while sure, the confusing “PR” system lets you make the CPU run “about as fast as” a 200 mhz Pentium in a bunch of different ways and you might *think* the combination that result in the highest core speed would be the best, well… it’s not. For most motherboards this 166mhz (66mhz X 2.5) setting is going to give you the best overall performance without introducing potential instability problems.

To make it short, for all the bus speeds slower than 66mhz you’re trading bus speed for a relatively low bump in core speed which, as that chart demonstrates, results in basically a wash. (They benchmarked it as “PR200” in all those combinations.) And a knock on effect of the slower CPU bus is going to be a slower effective memory and PCI (and AGP, if so equipped) clock, meaning that even if 180mhz with 60mhz x 3 gave you a small bump in benchmarks that fit in the CPU cache you’re going to pay for it with slower speeds for video and memory-bound tasks.

Conversely, the problem with the 75mhz bus 150mhz speed is motherboards that allow you to use 75mhz for the CPU bus typically either:

A: overclock the PCI bus to 37.5mhz, which can cause reliability issues, or
B: use a /3 instead of a /2 divider and run PCI at 25mhz instead of 33mhz.

“A” might give you the fastest possible overall system, but degrade reliability, while “B” might result in even slower video and I/O speeds than the slower-than-66mhz options. (Memory speed should at least be better.)

Feel free to try different options, but to really rate them you need to use a “whole system” benchmark, not something that just does math on the CPU. I think in the end you’ll probably find what you already have is the best compromise.

 

[Taffer]

Conversely, the problem with the 75mhz bus 150mhz speed is motherboards that allow you to use 75mhz for the CPU bus typically either:

A: overclock the PCI bus to 37.5mhz, which can cause reliability issues, or
B: use a /3 instead of a /2 divider and run PCI at 25mhz instead of 33mhz.

Thank you! I think that was the part I was not understanding. Yes, makes sense to keep the bus at 66mhz then, which leaves you with the decision to run a multiplier of 2.5x for 166mhz as configured now.

If you choose to run a Cyrix, I would recommend getting as close to a 66 MHz bus speed as you can,

Thanks to you for your writeup as well, it took me a bit to get it.

I do find it interesting that they label it 75mhz bus 2x right on the chip. Were they doing that knowing that despite either instability or underclocking the PCI bus that it would improve integer performance just to make the chip look better in comparison?

ended to have stability issues. This was exacerbated by their penchant for weird bus speeds like 75 and 83 MHz, which caused other devices in the system to run alarmingly out of spec due to the nature of cascaded bus clocks used on motherboards at the time. An example would be that the PCI bus would be run at 41.5 MHz on a 1/2 divider with a bus speed of 83 MHz, and PC-66 memory would be run way out of spec, or significantly underclocked if the motherboard offered a 1/2 or 2/3 divider.

Seems like it.

What a terrible long term strategy!

 

[bakemono]

The original 6x86 (non-L) series weren't great at achieving high clock rates, which is why they maxed out at 150MHz while marketing it as PR200, and likely why they had so many bins. (BTW, I'm not too sure about that PR100 in the chart, wasn't it supposed to be 45*2 ?) If the chips could have reliably hit 166MHz, they would have released a 166MHz part. 75MHz * 2 was the best option during that brief window of time and delivered some benchmark wins against Intel. Sure, they were asking for a weird bus speed, but it was their highest performing part, so there was some incentive for the market to deal with that.

It wasn't until later, after a die shrink or two, that they ended up with different bus/multiplier combinations selling with the same PR speed. By then plenty of motherboard chipsets could handle different bus speeds without stability problems. 37MHz PCI was rarely an issue to begin with, but eg. VIA chipsets could run the PCI asynchronously at 33MHz regardless of bus speed making it moot.