VOGONS


First post, by RetroSpector78

User metadata
Rank Member
Rank
Member

I'm playing around with my IBM 6x86 PR200 and noticed that it initially came out with a core frequency of 150Mhz (75mhz fsb x2).

A couple of months later it came out with a core frequency of 166Mhz (66mhz fsb x2.5) under the exact same PR200 branding.
The latter was about 10% faster (due to its higher core frequency).

But why did the 150mhz came first and with a higher FSB ?

Was it the manufacturing process that forced them to stick to 150mhz instead of 166mhz ?
Why did they revert back to a 66mhz FSB and did not stick with the 75mhz FSB (is it simply because 166mhz can only be achieved with 66*2.5 and the next bump on 75 would be 187.5) ?
Is there any benefit over having a faster FSB to achieve a slower overal core frequency (75*2 vs 66*2.5)
Also would the 75mhz version be a better overclocker (due to an already higher FSB) ?

Thx

Reply 1 of 8, by quicknick

User metadata
Rank Oldbie
Rank
Oldbie

I think that happens with most (if not all) PR ratings. Was putting some order in my vast collection of retro-related photos and stumbled upon this one that a seller sent me a couple of years ago. I've chosen the 66MHz FSB one because it was less scratched, but the one on the bottom (or right) could've been a better overclocker (whatever that means with Cyrix 😁 )

3a32e247-5da2-43d5-a3c5-54db90fdc992.jpg
Filename
3a32e247-5da2-43d5-a3c5-54db90fdc992.jpg
File size
414 KiB
Views
501 views
File license
Fair use/fair dealing exception

Reply 2 of 8, by BSA Starfire

User metadata
Rank Oldbie
Rank
Oldbie

As far as I'm aware Cyrix had to make a special effort to have the 200+(original 75Mhz part) made, the yields weren't particularly high and also motherboards that could take the 75 MHz bus speed weren't in high availability either. It was however the fastest x86 part on the market at the time. The other thing was as cooling systems were pretty scare that could keep the chip at acceptable temperature levels(a normal sink for a pentium was not up to the job) and even VRM's on the board need better cooling than the Pentium or the K5 did. The later "L" models of the 200+ kept the same speed but needed both less cooling or power. The 66 MHz part as far as I'm aware were the later MX parts, a redesigned core with MMX and much lower requirement for both power and cooling at similar ratings. Cyrix always tended to push the FSB boundaries with both 75 MHz and later 83 MHz parts on socket 7, finally 100 MHz right near the end, the board makers were a little slower to react to this but SiS is your best bet for a good compatible motherboard chipset(they also supported the Cyrix Linear Burst mode) for the 6x86 and a cooler for a socket 370 machine is the simplest solution these days.
The Cyrix is VERY fast for integer, but fell behind the Pentium for FPU performance, Cyxis's FPU was based on their 386 FPU, by far the best of the time by a country mile, but guess they didn't see Quake coming or know the significance that would have for future games. But outside of Quake it's a great chip and a real piece of innovative history and will run all your games just great, even Quake is OK until you start to compare it to the Pentium(Quake used hand coded intel P5 FPU optimizations).
As to overclocking, forget it to be honest, the 200+ was at it's limits at launch, some of the later Cyrix MII chips can manage a bit however. It is however worth getting the Cyrix tweak utility to enable some registers on the chip for various functions, this can help a whole lot.

286 20MHz,1MB RAM,Trident 8900B 1MB, Conner CFA-170A.SB 1350B
386SX 33MHz,ULSI 387,4MB Ram,OAK OTI077 1MB. Seagate ST1144A, MS WSS audio
Amstrad PC 9486i, DX/2 66, 16 MB RAM, Cirrus SVGA,Win 95,SB 16
Cyrix MII 333,128MB,SiS 6326 H0 rev,ESS 1869,Win ME

Reply 3 of 8, by RetroSpector78

User metadata
Rank Member
Rank
Member
quicknick wrote on 2020-01-02, 13:57:

I think that happens with most (if not all) PR ratings. Was putting some order in my vast collection of retro-related photos and stumbled upon this one that a seller sent me a couple of years ago. I've chosen the 66MHz FSB one because it was less scratched, but the one on the bottom (or right) could've been a better overclocker (whatever that means with Cyrix 😁 )
3a32e247-5da2-43d5-a3c5-54db90fdc992.jpg

Is that because it has a lower FSB meaning there's more room to increase it ?
I assumed the FSB was determined during the manufacturing process of the CPU, so that it wouldn't really matter.

For my PR200 (75mhz FSB, running at 150Mhz) I was thinking about the following options :

Overclock 1 : 75 * 2.5 = 187.5
Overclock 2 : 83 * 2 = 207.5

I was able to run it at 187.5Mhz without any issues (win98se install went fine, played some games, ran some benchmarks, did the prime95 torture test for 30+ minutes). Didn't notice any errors.
Had to up the voltage to 3.2v (from stock 2.9v), cpu got a little bit warmer but it "felt" ok. (unfortunately it is difficult to get good temperature readings from these things).

Reply 4 of 8, by derSammler

User metadata
Rank l33t
Rank
l33t
RetroSpector78 wrote on 2020-01-02, 18:38:

Is that because it has a lower FSB meaning there's more room to increase it ?

That, and because the multiplier is higher, so you can reach higher clock rates at the same FSB than with a version which uses a higher FSB from the outset and hence a lower multiplier.

Reply 5 of 8, by RetroSpector78

User metadata
Rank Member
Rank
Member
BSA Starfire wrote on 2020-01-02, 14:14:

Cyrix always tended to push the FSB boundaries with both 75 MHz and later 83 MHz parts on socket 7, finally 100 MHz right near the end, the board makers were a little slower to react to this but SiS is your best bet for a good compatible motherboard chipset(they also supported the Cyrix Linear Burst mode) for the 6x86 and a cooler for a socket 370 machine is the simplest solution these days.

In that respect it is still not clear to me why they initially went from 150Mhz (75*2) to 166Mhz (66*2.5) in the 6x86MX PR200 evolution. Was it simply a matter of opting for a higher base speed and sacrificing FSB performance and being ok with that ?

BSA Starfire wrote on 2020-01-02, 14:14:

The Cyrix is VERY fast for integer, but fell behind the Pentium for FPU performance, Cyxis's FPU was based on their 386 FPU, by far the best of the time by a country mile, but guess they didn't see Quake coming or know the significance that would have for future games. But outside of Quake it's a great chip and a real piece of innovative history and will run all your games just great, even Quake is OK until you start to compare it to the Pentium(Quake used hand coded intel P5 FPU optimizations).

I ran a couple of games on my PR200, and with some 3DFX help a lot of games are definitely playable. In software-rendering it does struggle a bit and there is a big difference between the Cyrix and Intel MMX, but 3DFX takes a lot of the load from the CPU. The difference is still there, just not mega obvious.

BSA Starfire wrote on 2020-01-02, 14:14:

As to overclocking, forget it to be honest, the 200+ was at it's limits at launch, some of the later Cyrix MII chips can manage a bit however. It is however worth getting the Cyrix tweak utility to enable some registers on the chip for various functions, this can help a whole lot.

I've been running my 6x86MX PR200 (75x2) stable at 187.5mhz (75*2.5 @ 3.2v). Survived win98se install / prime95 torture test / various benchmarks + games. To my surprise it seemed to be running fine. Heard others were even able to manage 207.5Mhz (83*2.5) but my current motherboard maxes out at 75mhz FSB.

Reply 6 of 8, by RetroSpector78

User metadata
Rank Member
Rank
Member
derSammler wrote on 2020-01-02, 18:46:

That, and because the multiplier is higher, so you can reach higher clock rates at the same FSB than with a version which uses a higher FSB from the outset and hence a lower multiplier.

But both CPUs are multiplier unlocked, so you can increase it just as easily as you would the FSB.

I'm missing the different between increasing FSB Vs increasing the multiplier as both seem to be possible. (the only thing I can think of is that using the FSB allows for smaller OC increments, as opposed to the multiplier that will immediately bump the frequency by a minimum of 33mhz).

Stock 6x86MX PR200 75*2 = 150Mhz (OC option1 = 75*2.5 = 187.5Mhz , OC option2 = 83*2.5 = 207.5Mhz)
Stock 6x86MX PR200 66*2.5 = 166Mhz (I'm guessing this one has the same OC options .... why would this 66Mhz FSB version be any better than the 75Mhz FSB version for over-clocking).

Reply 7 of 8, by quicknick

User metadata
Rank Oldbie
Rank
Oldbie
RetroSpector78 wrote on 2020-01-02, 18:38:

Is that because it has a lower FSB meaning there's more room to increase it ?
I assumed the FSB was determined during the manufacturing process of the CPU, so that it wouldn't really matter.

Well, derSammler answered that. One version is 66x2 (=133), the other is 60x2.5 (=150). Higher frequency to compensate for the lower FSB.

The hypothetical better overclock comes from that very fact. Since the Cyrix CPUs were sold almost at their limits, the 133MHz chip would probably not reach 166 whereas the second might have a chance. If performance is really better at 166 than 150, it depends on a lot of other things I guess.

In my (limited) experience FSB overclocking is mostly a motherboard matter and not a CPU one. I would expect the 60x2.5 CPU to work all the way to 100x1.5 (if it supports such multiplier; I believe it does). I really have no idea how Cyrix determined the FSB/multiplier for their chips, and it annoys me greatly that I cannot remember what I had back in the day, I'm quite sure it was an 6x86MX PR200 or 233, maybe spec'd at 75x2.5. I remember trying lots of FSB/multiplier combos in the search for ultimate performance. And a utility called 6x86opt or something close.