- VCF South West - June 14 - 16, Davidson-Gundy Alumni Center at University of Texas at Dallas
- VCF West - Aug 2 - 3, Computer History Museum, Mountain View, CA
- VCF Midwest - Sept 7 - 8 2024, Schaumburg, IL
- VCF SoCal - Mid February 2025, Location TBD, Southern CA
- VCF East - April 2025, Infoage Museum, Wall NJ
-
Please review our updated Terms and Rules here
P66, POD83, and AM5x86 133 comparison.
- Thread starter Moogle!
- Start date
Binarydemon
Member
- Joined
- Jul 14, 2018
- Messages
- 10
Seems to be about what I would expect - Didn’t AMD basically market the 586-133 as a Pentium 75 equivalent?
Moogle!
Experienced Member
Thanks for the response. Glad to know someone appreciates that. Supposedly they're supposed to be a Pentium 75 equivelent, at leats as integer performance is concerned. (They are marked P75+). I'm mildly curious how a Pentium 75 actually stacks up against a Pentium 66, as the former runs a 1.5 times a 50Mhz Bus while the latter runs at a full 66Mhz.
Also, there is an error in that chart. Both boards were run with 32MB total, and were tested separately, with the same two sticks of 16MB ram.
Also, there is an error in that chart. Both boards were run with 32MB total, and were tested separately, with the same two sticks of 16MB ram.
dan951
Experienced Member
That's pretty cool. If you can get your hands on one of the Cyrix 5x86 CPUs, I think you will be surprised how well they perform. I have a AMD 5x86 at 160MHz and the Cyrix 5x86 at 100MHz completely blows the doors off of it when it comes to gaming.
GiGaBiTe
Veteran Member
The reason the Cyrix 5x86 beats the AMD 5x86 is because they're two different class processors. The Cyrix 5x86 is a cut down Cyrix 686MX with a 486 bus interface (much like the Intel POD63/83), while the AMD 5x86 is just a fast 486.
The Cyrix 5x86 isn't all roses and unicorns though, it has compatibility issues. It doesn't implement the full i486 instruction set, and has very limited support of Pentium class instructions. It also would have been a much faster part if branch prediction was enabled on final silicon, which caused a bit of controversy back in the day (Cyrix advertised benchmarks with BP turned on.) There is a tool to turn it on manually, but it usually results in an unstable system.
wesleyfurr
Veteran Member
Wasn't there also something where you ran a utility to enable the L2 cache on the 5x85 or something odd like that? I recall running something on mine back in the day that would speed it up...
Wesley
Wesley
Agent Orange
Veteran Member
The Cyrix 386 FPU has a small executable to enable/disable its cache. Don't know about the Cyrix CPU.
wesleyfurr
Veteran Member
With a little digging I found this - https://web.archive.org/web/2003102....nifty.com/sandy55/PowerBoard/PowerBoard.html
"Cyrix5x86
Just put it on the ZIF socket directly and set J15 "speed" jumper unplugged. If you run
DOS, Win3.1, Win9x or NT4, you'd better use appropriate cache programs to activate L1
cache W/B feature.
I've been using ET586.exe by Evergreen which is contained in et9603.exe"
Of course that may or may not have been what I'm remembering, but that sounds right, and I do recall it helping the performance.
Wesley
GiGaBiTe
Veteran Member
There were several utilities that had varying capability to enable/disable 5x86 CPU features. Vogons has an extensive post covering it:
https://www.vogons.org/viewtopic.php?t=30607
Branch prediction has the most speedup potential, but is also the most unstable unless you have a late revision part.
Considering the fact that Pentium is just a fast 486 which in turn is just a fast 386 I would strongly disagree with the above statement. Which instructions are missing??? Don't tell me about CMPXCHG or those manipulating some specific MSRs. They were avoided back in time in order to maintain software compatibility with all CPUs starting 386. The latter introduced the IA32 architecture.
I strongly disagree with your reasoning. The new instructions added to the i486 are more than just cute pages in the datasheet, and the main reason that the i486 can still be supported by modern operating systems (minus some bitrot), while the i386 has been simply dropped many years ago.
You might not care about RDTSC, but many applications did. In fact, so many that Cyrix disabled CPUID by default, to prevent applications from seeing a "family 5" processor and dare to treat it like one!
GiGaBiTe
Veteran Member
If the Pentium was "just" a fast 486, nobody would have bought them because the 486 arch went all the way up to 200 MHz. This wasn't the case at all though, the Pentium is a completely different design that was much more efficient at a lower clock speed and the chart in the OP shows.
The PC industry in the 90s was moving at a blistering pace, nobody was interested in maintaining backwards compatibility with a 386 CPU from the mid 1980s, especially games like Quake which made heavy use of the pipelined FPU of the Pentium.
The 486 went up to 133 at the tail of of their desktop use. Some people like me overclocked a 133 AMD to 160 but that's about it.
Nobody in gaming cared about the FPU on any of the chips until Quake came along and actually used it. I feel bad for Cyrix coming out with a Pentium killer only to get kneecapped when the FPU was a big deal and they were still using the outstanding (for 386 era) FPUs they had designed but were outdated compared to Intel and to a lesser degree AMD.
Nobody in gaming cared about the FPU on any of the chips until Quake came along and actually used it. I feel bad for Cyrix coming out with a Pentium killer only to get kneecapped when the FPU was a big deal and they were still using the outstanding (for 386 era) FPUs they had designed but were outdated compared to Intel and to a lesser degree AMD.
GiGaBiTe
Veteran Member
AMD did release 150 and 160 MHz 486 (5x86) parts, but they're pretty rare.
http://www.chipdb.org/cat-150-23.htm
http://www.chipdb.org/cat-160-24.htm
200 MHz was possible on boards which could handle a 50 MHz FSB. I remember back when I was in college, we had a 486 system in the lab with a Kingston Turbochip that ran at 200 MHz.
http://www.chipdb.org/cat-150-23.htm
http://www.chipdb.org/cat-160-24.htm
200 MHz was possible on boards which could handle a 50 MHz FSB. I remember back when I was in college, we had a 486 system in the lab with a Kingston Turbochip that ran at 200 MHz.