kool kitty89 wrote:Hmm, I don't see the AMx5 at 200 MHz (or 180 MHz) listed in the 6x86 comparison at all, only the 133 (33 FSB) and 160, and the 160 seems to run Quake 1 faster than the 5x86 120 (both in the 686 tests), in spite of the latter running with a 60 MHz FSB.
I am not sure what conclusion you are drawing from this; everything seems to be in agreement to me. The AMD X5-160 is just barely faster (0.1 fps) than an IBM 5x86-120 in Quake at 640x480 for the case of the 686 benchmark comparison. This is in good agreement with the results from the 486 benchmark comparison. The limitation with running at 60 MHz is that you need to run the cache at 3-2-2 and with 1 WS for memory read. The L2 cache speed at 60 MHz ends up being the same as at 33 MHz, but the memory speed jumps ahead. The real advantage for L2 speed comes from a 66 MHz bus, which will still run fine at 3-2-2 and 1 WS for memory read. Unfortunately, the motherboard won't function properly if we leave all the timings on the fastest settings for higher and higher FSBs. Another disadvantage of running with a 60 MHz FSB is that the PCI bus will be at 30 MHz instead of 33 MHz. 66 MHz is really the optimal FSB for these fast 486's.
kool kitty89 wrote:Also, you mention the x5 200 being at 4x50 MHz, but did you ever get it working at 3x66 or 2x66? That would be nice to compare with the Cyrix 5x86 on 66 MHz FSB.
If you look in more detail at the Ultimate 486 Benchmark Comparison, you will the answer to your question. I noted in the text, as well as the charts, that the AMD X5-200 is the only CPU I did not test for this comparison. Of the several AMD X5's I had, neither would run at 200 MHz and 5 V. I also could not get them running at 180 MHz. It was noted that vogons user Retro Games 100 (rg100) tested this CPU for the comparison. I have only ever heard of 2 AMD X5's running at 200 MHz, however the long term stability of this speed is doubtful. One user had to run his AMD X5-180 outside in the winter cold for operation.
I know that the AMD X5-160 and IBM 5x86c-133 are generally stable and readily obtainable. Unfortunately, rg100 is not available to carry out tests for the Ultimate 686 Benchmark Comparison. I have included a few more tests in this comparison which were not on the 486 comparison, so I cannot copy over the results. I do have an AMD X5-133 with a mid-1998 datecode, however it also did not operate at 200 MHz. Although the X5 is 5 V I/O tolerant, the Am5x86 specification sheet lists Vcc at 3.3 V +-0.3 V. Running the chip at 200 MHz, let alone 5 V, is probably way out of thermal specification. Oddly enough, rg100 found that a ADW chip is the one which would run at 200 MHz, however the one other guy who got his chip running at 200 MHz, it was an ADZ version.
For those late 1998/1999 AMD X5 chips, the markings are a little different. Some are marked as:
Am486 DX5-133W16BGC
Am486 DX5-133V16BGC
Anyone know the difference between the W and the V?
sliderider wrote:It's not easy to get a AMD 5x86 to run at 200mhz. A lot of the time either the chip itself or the motherboard becomes unstable at that speed. You have to find a magic CPU and a magic motherboard to get it to work.
There is certainly some truth to that. While rg100 and I both have the same v3.x motherboard, his would operate at 50 MHz with a 2-1-1 cache setting, whereas mine would not. I even find that some MB8433 boards won't run at 40 MHz with 2-1-1. Surprisingly, the M919 boards seem to do quite well with fast cache settings and high FSBs.
Plan your life wisely, you'll be dead before you know it.