Talk:Cyrix Cx5x86

From Wikipedia, the free encyclopedia

Where are the benchmarks that show the AMD 5x86 to be faster then the Cyrix 5x86 while the Cyrix part is running default with features disabled? The article states that is the reason the AMD part is faster, I'd like to see a benchmark proving that otherwise I believe the article should refer to the Cyrix 5x86 as the fastest socket 3 processor. Even links listed in this article suggest that the Cyrix part was faster.

http://www.pcguide.com/ref/cpu/fam/g4.htm
http://www.pcguide.com/ref/cpu/fam/g4C5x86-c.html
http://www.interwb.com/486bench.txt The hardcore icon the sandman 09:35, 29 July 2006 (UTC)

The 5x86 M1SC is not the fastet Sock 3 CPU, the 133 MHz was hardly available (if ever, have no yet seen anyone having it). You also have to compare them to the 150 MHz AMD 5x86 P75+ (it was at least available as the 120 MHz Cyrix was) and the 160 MHz AMD 5x86 P90 (at least announced, maybe not released or in very limited numbers). I still have one or two M1SC and I really loved them apart from their unwillingness to overclock to 50x2 (Board was capable of doing this) or 120 MHz.--Denniss 11:42, 29 July 2006 (UTC)

Benchmarks showing an Am5x86 @ 150 or 160mhz as being faster then a Cx5x86 @ 133mhz would be nice to see. It doesn't matter how rare a CPU was or wasn't, fastest for a particular socket is still fastest for a particular socket. The hardcore icon the sandman 21:02, 29 July 2006 (UTC)

It's pretty tough to overclock Cx5x86. The fastest official speed is 120 MHz. That would be a 40 MHz bus. The CPU itself has little extra clock headroom, and I've never had luck with a 50 MHz bus. The cache RAM just can't do it usually, and neither can the expansion cards or system RAM. There's a reason the 486DX50 never went very far.

I've actually heard of people getting Am5x86 up to 200 MHz (50 x 4) but I've tried 3 of them and none will even POST at that speed. That would be an amazing 486 though!

I did run Sandra benchmarks when I had the machine running. I tested all 3 CPUs in the same board with all of the same components. The Am5x86 160 was on top with the 120 Mhz Cx5x86 next and then the PODP5V 100 MHz at the bottom (for integer). For FPU the Pentium was well at the top. The other two were pretty close to each other. I don't have the numbers unfortunately. And there was my semi-subjective testing of Jedi Knight + Voodoo3, where the Pentium felt the fastest. Doing Windows stuff (browsing with IE) the Am5x86 160 definitely felt to be the snappier of the bunch.

I've always been a bit fascinated by the Pentium Overdrive. They really reworked that CPU to run on a 486. It has 2x the L1 cache of a normal P54 (32K instead of 16K) and has that built-in heatsink-fan combo. They stuck voltage regulation right on the chip (weird!) and the thing has half the data bus of a normal Pentium cuz it's on a 486 board. Too bad it has quirks with memory performance or it would be the CPU to go with I think. Here's a link to a forum post I made way back when I was messing with this hardware [1]. --Swaaye 21:50, 29 July 2006 (UTC)

I messed around with a PODP5V, Cx5x86, and Am5x86 about a year ago. The AMD chip comes out on top, especially at 160 MHz, because it's just the easiest of the bunch to get speed out of. It doesn't have very many special requirements to maximize its speed. In fact, the only hitch is whether your board can run a writeback L1 or not cuz writethru is a bit slower. The Cyrix chip has a ton of disabled features that can theoretically bring its speed up, but they work on an extremely small number of boards if at all. It's famed branch predictor, for example, is actually totally broken on most chips excepting the latest revisions. You still need a board that will work with it enabled even if the chip works with it on.

The Pentium OD is just a mess. It absolutely rocks for FPU speed, easily outpacing the other two chips no matter what clock you run, but its integer performance is very bad because it doesn't get along with most boards' L2 caches. On my board board, a late model PCI 486 mobo that supports every CPU under the sun, it would secretly disable the L2 entirely (shown within cachechk). It ends up with very bad memory access times and thus in Windows it feels a lot less snappy, and scores lower on integer than a Cx at 100 MHz and well below the 160MHz AMD, even at the 100 MHz I got it running at.

I actually stuck a Voodoo3 2000 PCI in the 486 and tested all three CPUs with Jedi Knight. They were all fairly close but I believe the Pentium was fastest with its powerful FPU. In Windows the AMD chip felt the fastest, with the Pentium just feeling noticeably more sluggish (even at 40/100MHz).--Swaaye 16:29, 29 July 2006 (UTC)

I never had problems with enabling the M1SC features but I only tested late-generation PCI board with UMC 8881 chipset (Shuttle HOT-433 rev 1.x and 3.x; Biostar 8433UUD, the Biostar without 50 MHz FSB suport). I may need to revive one or two of these boards and check my M1SC for stepping. --Denniss 17:39, 29 July 2006 (UTC)