Quick update: 120 CPUs tabulated to date. S7 and SS7's complete. Slot 1, Cyrix MediaGX, VIA <=600, and Athlon <=600 remaining. PPro outsourced.

Re-requesting volunteers for P60/66, Xeon <=600, and NexGen.

I noticed you don't have a Pentium 120 in your list -- I have one of these. It's too big to ship (full tower) but if you'd like to include it in testing, I'm willing to help out.

Thank you for your interest, however the Pentium 120 has been tested already. I did not include a list for all CPU's tested for this comparison. The list kept growing. Every socket 7 and super socket 7 CPU has already been tested. I'll move onto the slot 1's <=600 soon.

I'm only looking for someone to test P60/66, slot 2 Xeon <=600, and NexGen's.

I have the following Slot-2 Xeons:

P2 Xeon 450/2MB Engineering Sample
P2 Xeon 450/512Kb
Two P3 Xeon 550/1MB
Two P3 Xeon 700/1MB
Two P3 Xeon 1Ghz/256kb

I could test for 400-550mhz, but as I'm in the process of moving it'll be a while until I can actually get around to doing them.

I also have a Matrox Millennium G200 AGP, will that be suitable for the tests?

Also, is there a thread or links to grab the benchmarking utilities that will be used?

Great and thank you for volunteering! There is no rush for completing this.

It looks like you will be able to test the P2 Xeon 400-2MB, P2 Xeon 450-2MB, P2 Xeon 450-512KB, and P3 Xeon 550-1MB.

A Matrox Millennium G200 PCI-16 MB is preferred for consistency, but if you only have the AGP version then that is better than not having the Xeon tabulated at all. If you have the 8 MB expansion module for the G200 that is best.

m1919 wrote:

I have the following Slot-2 Xeons: […]
Show full quote

I have the following Slot-2 Xeons:

P2 Xeon 450/2MB Engineering Sample
P2 Xeon 450/512Kb
Two P3 Xeon 550/1MB
Two P3 Xeon 700/1MB
Two P3 Xeon 1Ghz/256kb

I could test for 400-550mhz, but as I'm in the process of moving it'll be a while until I can actually get around to doing them.

I also have a Matrox Millennium G200 AGP, will that be suitable for the tests?

Also, is there a thread or links to grab the benchmarking utilities that will be used?

If your willing to lend me your chip, I can do the testing in my Asus board. I have all the benchmarking setup. I should have the "extra" pentium pro speeds done today. I was messing with my board and found a 50mhz fsb setting. I could set my 200mhz chip to run at 100mhz. Unless feipoa wants it, I'm going to pass on that bus speed.

I can't see much interest for running a 50 MHz FSB aside from a P75 and maybe one or two Cyrix 6x86's.

if anyone has a 512kb pentium pro, let me know. i was biding on one and it went for $50 to some scrapper.

Feipoa - I picked up a cheap 440LX motherboard that I'll try for 133 MHz with. I will let you know if it works.

Swaaye, this will help to shed a little light on the 2x dependency -- if it is BIOS, chipset, or CPU related. It is unfortunate you do not have an ASUS slot 1 board or a PII-266 with a newer CPUID.

Sadly it appears that the Abit LX6 (440LX) also disables the L2 cache at 2x multiplier. I wonder if this is done by the BIOS because of CPU/cache errata at that speed. I tried two PII CPUs, a 233 and the 266 that worked with cache at 2x on the 440FX board.

Interesting. I wonder what mechanism the 440FX board places over the PII-133 which the other motherboards cannot. It is surprising that 3 chipsets, other than the 440FX, were unable to properly enable the L2 cache of the PII at 2x/66. I suppose such a CPU configuration would not have been a priority of any motherboard maker.

It's also possible that the cache is intentionally disabled. Maybe there is some sort of stability issue at that clock rate for example. The 440FX board may have been behaving in a manner not recommended. I tried reading through the PII errata specifications but nothing stuck out.

What happens if you try running the 440LX board and PII-266 at 5.0 x 66? Prior to wrapping up Klamath testing, I discovered that running the PII-266 on a VIA Apollo Pro 133 (VIA 694/596) with a 5.0 x 66 setting allows for 133 MHz with L2 cache enabled. Chkcpu claims it is running at 5.0 x 26.7, however PCI performance is identical to a 33 MHz PCI bus at 66 x 2.5, for example. The memory throughput also seems relatively agreeable with a 66 MHz FSB. Cachechk memory read for a Pentium II at (MHz)

133 is 92.7 MB/s
166 is 101 MB/s
200 is 111 MB/s
233 is 118 MB/s
266 is 124 MB/s

I have completed all benchmarks in DOS and Windows and everything checked out OK.

For historical value, I'm using an ASUS P3V4X. Tests in Chkcpu and Speedsys reveal that,

2x = 24.33 x 5.5 (L2 disabled) - 133 MHz
2.5x
3.0x
3.5x
4.0x
4.5x = Screen will not turn on
5.0x = 26.7 x 5.0 (L2 enabled) - 133 MHz
5.5x = 24.3 x 5.5 (L2 disabled) - 133 MHz
6.0x = 66.6 x 2.0 (L2 disabled) - 133 MHz
6.5x = 2.5x
7.0x = 3.0x
7.5x = 3.5x
8.0x = 4x

Due to these results, it is likely that the L2-disable phenomenon is due to a sensor on the multiplier setting rather than on the raw CPU frequency.

5x does appear to result in 133MHz CPU speed with L2. 5.5x and 6x result in 133MHz but no L2.

Ok, so these may be CPU-specific symptoms. What happens at 5x on a 440BX board?

440BX behaves the same at 5x. It seems to be perhaps entirely a function of the Klamath PII.

Does Chkcpu report 26.7 x 5.0 on your boards? I wonder if there is some asynchronous magic going on...

I have seen all kinds of strange multiplier and FSB combinations from diagnostic software. It seems such software has good methods of determining the total CPU speed, but not so much for determining the FSB speed.

@swaaye, I understand you have a Klamath Pentium II running 2x66MHz=133MHz with the L2 cache functional on a 440BX mainboard. Neat! What it is the sspec of the CPU?

gerwin wrote:

@swaaye, I understand you have a Klamath Pentium II running 2x66MHz=133MHz with the L2 cache functional on a 440BX mainboard. Neat! What it is the sspec of the CPU?

No not at 2x multiplier. It runs 133 MHz with L2 using Feipoa's 5x multiplier setting which he discovered results in 133 MHz at an unknown FSB rate. He believes that it is actually 66x2.

With the normal 2x multipler setting, the L2 is disabled and this seems to be related to motherboard/chipset. This PII is SL28L which worked at 2x66=133 w/L2 on the 440FX board that I once had. 440LX results in no L2 @ 2x66.