Thanks a lot. If you use a "special" P2 CPU, and attempt to run it at the undocumented 50 FSB speed setting, with a CPU multiplier setting of 2x, will the CPU's L2 cache behave "oddly", or will it behave normally? And if you run Speedsys with this "2x & 50FSB = 100MHz" configuration, will the overall speed score be significantly higher than a regular classic Pentium 100MHz CPU, or about the same?

Well, I don't have that board yet... You do. But I expect the L2 cache problem at multiplier 2x and 2.5x to be there, and the speedsys score should be about (155/66*50)=117 Pts.
That is if there are no surprises.

gerwin wrote:

.....a multiplier limited P-II.

I've given this some extra thought.
Given the range of free multipliers, ranging from 2x to 5.5x, this gives 8 possible multipliers. The earlier Pentiums had 1.5x to 3x or 2.0x to 3.5x, giving 4 selectable multipliers.
Some Pentiums had their higher multipliers disabled (I tested this a while ago).
The K6-2's (Chompers core) had multipliers ranging from 2.5x to 6x, giving 8 multipliers to choose from.
Many 486 class CPU's had only 2 multipliers to choose from, or none at all

All of these processors have a number of multipliers equal to a power of 2 (4 for the Pentiums, 8 for the K6-2's and 8 for some of the Deschutes).
I think the Deschutes 400Mhz isn't multiplier limited at all, I think it's fully unlocked and won't go over 5.5x simply because 5.5x is the highest multiplier Deschutes was made to understand.
The Deschutes cores which are limited to a max multiplier of 5x only have their multipliers locked upward from their default multiplier.

But my Pentium II does not do 5.5x.
There was this observation that the FSB speed influenced the maximum multiplier on a pentium II. but I have to review that idea.

This is what I noted:
For the Pentium II/III they use four channels [on/off] to select the multiplier at CPU reset:
LINT[1], LINT[0], A20M#, IGNNE#.
So in any case the maximum is 16 different multiplier request signals.

And what I read is that Intel did not define more multipliers then were needed for the CPU range on sale at that moment. At least not in the available documents. Also it seems that the Coppermine redefined a few, and therafter the Tualatin line redefined many multiplier request codes: They interpret the codes differently. The lowest a P-II will understand is 2.0x, an unlocked coppermine 3.0x and an unlocked Tualatin 4.0x.

Thanks for the info, it's very interesting!

Afaik all unlocked P2-400's will accept a 5.5x multiplier, so far all unlocked 333's work with 5x and lower. I can't remember if there were any Descutes 300's with an unlocked multiplier.
So far the Deschutes which had a free multiplier only had their upward multi's locked.

I don't think FSB influences the maximum multiplier, but I sure can't tell that's true for certain 😉

It is common for CPU manufacturers to grab more available multi's from te bottom and remap those multi 'pins' at the top.

I wonder what the minimum multi for Katmai would be btw, but I don't have an a Katmai engineering sample to test it.

Just got in a Gigabyte GA-6BXC mainboard. But there is one capacitor near the AGP slot that has lifted itself up... I hope the other caps are still OK.

gerwin wrote:

Just got in a Gigabyte GA-6BXC mainboard. But there is one capacitor near the AGP slot that has lifted itself up... I hope the other caps are still OK.

If you see something bulging out of the bottom of the cap, then it's a burst cap.
If not, then it should probably be fine, provided the 2 metal thingies are still soldered to the board.

There is goo below that capacitor. But all the other caps seem to be in shape from the looks of it. Also the mainboard functions. phew...
Now I am gonna play around with the CPU multiplier jumpers and the 50MHz FSB.
This Mainboard has two jumpers that should be closed in the specific case of using a voodoo 3. Funny.

I put everything together with the new mainboard GA-6BXC.
I tested things with a SL2QF Intel Pentium II 333 MHz first, then with a QH31ES Intel Pentium III 750 MHz (Unlocked)
I set the FSB to the nifty 50MHz setting (DIP OOOO) and tried all 16 possible Multiplier combinations. These are my observations.

1DIP  PII          / PIII 
2OOOO x2 (no L2)   / x4
3XXXX x2 (no L2)   / x4
4XOOO x2.5 (no L2) / x8
5OXOO x3           / x3
6XXOO x3.5         / x3.5
7OOXO x4           / x4
8XOXO x4.5         / x4.5
9OXXO x5           / x5
10XXXO no post      / x5.5
11OOOX no post      / x6
12XOOX no post      / x6.5
13OXOX no post      / x7
14XXOX no post      / x7.5
15OOXX no post      / x8
16XOXX no post      / x3
17OXXX no post      / x4
18[O=on, X=off]

Speedsys results of the slowest setting with each CPU are attached. No surprises, everything works as I hoped. I tried several games and they ran fine. I will keep the PIII installed and configured for 150 or 200 MHz, with SoftFSB for windows this can be raised to 400 MHz quite easily.

Now before I put everything in a case I have to fix that bad capacitor.

gerwin wrote:

There is goo below that capacitor.

If it looks like one of the caps on one of these pictures http://www.badcaps.net/pages.php?vid=5 then it's a leaking cap.
Expect trouble 🙁

Yes I have seen them before, but then it was usually seen in multiple caps, and usually the top was bulged. This is a picture of the problem. Will it be severe enough to ask the ebay seller for warrenty, or is that useless?

what on earth is that metal clip doing on the AGP slot..

Dunno what the metal clip thing is. It's not one of those wacky Dell things, right?

I have seen multiple boards which had many caps and had only 1 bulging one.
I can't see good enough on your second pic and the 1st pic I can't see if it's something that came from your cap or if it's just something stuck under it, a bug or something 😜

It kinda looks more like melted plastic or something, usually that goo is more brown colored.

man i also had a celeron950ES running at 50*4 on a 815e before, took around 8:35 to run superpi. overclocking to 150*8 takes around 2:00. 😎

A 815 @50Mhz FSB? 😳

Tetrium wrote:

A 815 @50Mhz FSB? 😳

yeah it involves making use of a slight bug
the board has jumpers to choose between clock ratios:
fsb:mem:agp:pci ratio available fsb range in bios
2:3:2:1 66~83(default66)
3:3:2:1 75~124(default100)
4:4:2:1 100~240(default133)
4:3:2:1 100~240(default133)
as you can see, there is no way you can enter 50fsb here.
however, if you set 3:3:2:1 jumpers and enter 75fsb in bios, then power off and change jumpers to 2:3:2:1, power on again and it would be 50fsb. 😎

I would say that would depend on the PLL chip used on a particular mainboard, there are hundreds of types. And the ones supporting 50MHz FSB got out of fashion when the BX chipset arrived. The PLL chip datasheets are very specific about which frequencies they support, as that is all they do.
So it is a strange bug, I cannot really imagine how it happened with that 810.

Indeed, it depends on the PLL chip, of which different versions were used even on boards that themselves are the exact same revision.

Still, would be interesting to post somekind of list in case one of us ever finds a PLL chip on a BX and post a pic+motherboard model number and such?

okay, just looked at my board and the PLL is cypress w305bh. 😎
and even if a frequency is supported by PLL, it could be purposefully restrained by bios.

True. It might even be there in the BIOS, but hidden. I've once found a forum about modding BIOS flash files with a LOT of info about this, very interesting read if you're into that stuff 😉

From the Cypress W305B datasheet:

Programmable clock output frequency with less than 1 MHz increment.... W305B supports programmable CPU frequency ranging from 50 MHz to 248 MHz

So yes, and with 1 MHz increment. That is something you won't find on a BX mainboard.