For BIOS modding, this site is a very informative read http://www.rebelshavenforum.com/sis-bin/ultim … ?ubb=forum;f=52
😉

@gerwin: that P3 750 slot1, was it easy to find ?

And, any luck making it process a 2x multiplier properly ?
(say, 2x 66mhz = 133mhz and not 267mhz?)

I got the P3 750 E.S. from German ebay. It was there on an Abit mainboard with a minimum bid of Euro 100. Nobody put any bid on it till the end. I contacted the seller and negotiated a price (about half).
Anyway I could give you a hint on searching, but I don't want to type it here.

I used all 16 possible multiplier request signals/settings that a P-II/III can receive. And none of these got it below 3.0x.

My retro build now looks like this:
-P-III-E 150MHz / 50MHz-FSB (no fan, underclocked 750MHz Engineering Sample)
-Gigabyte GA-6BXC rev2.0 i440BX (AGP, 4xPCI, 3xISA)
-Voodoo3-2000 16MB 128-bit AGP (underclocked from 143 to 100MHz, new heatsink)
-128MB RAM
-Sound Blaster 16 CT2800 with genuine OPL3 (and genuine hanging notes).
-Roland MPU-401AT, Midi signal connected to the Midi DB.
-Terratec Dream 4MB Midi Daughterboard, on homemade riser with stereo reverser (no it is not connected to that PCI slot Wink )
-1GB IDE Flash module (Need a bigger one)

I can pull the FSB to 133 with SoftFSB and get a reliable 400MHz CPU. The AGP card then goes goes from 33MHz to 89MHz, as the AGP divider is kept on 2/3. I ditched the PCI voodoo 3. To much heat/power usage, I don't like it. Throttle now works to slow down as low as 18,7 MHz. I had to Disable ACPI at Windows 98 setup in order to use throttle under Windows.

gerwin wrote:

Anyway I could give you a hint on searching, but I don't want to type it here.

Something with sspec? 😜

Edit:I say let the Vogoners know how to find these chips, P3 Engineering Samples are quite easy to find these days, there are tons out there so let everyone here have a shot at obtaining one 😉

At cpu-world they are for sale on a regular basis, though I can't tell which ones are actually unlocked 😉

@Tetrium: that 1400mhz Celeron (QKE0), how low would it go, I wonder ?
Still no 2.0x multi usable I guess, or ?
(assuming your mobo has those)

Or, this beeing a Qualification sample, was it perhaps locked?
(just so I know before I somehow purchase my own)

bestemor wrote:

@Tetrium: that 1400mhz Celeron (QKE0), how low would it go, I wonder ? Still no 2.0x multi usable I guess, or ? (assuming your m […]
Show full quote

@Tetrium: that 1400mhz Celeron (QKE0), how low would it go, I wonder ?
Still no 2.0x multi usable I guess, or ?
(assuming your mobo has those)

Or, this beeing a Qualification sample, was it perhaps locked?
(just so I know before I somehow purchase my own)

I don't know. I do have 1 or 2 ES/QS laying around but I haven't gotten around to testing them yet.
I'd reckon their minimum multi is around 3x or so?

One last note:Not all of these QS or ES have a free multiplier!

One last note:Not all of these QS or ES have a free multiplier!

Jepp, kinda figured that, after some heavy reading/googling... 😒
Still, hoping at least the same Qspec (and/or stepping) would share multiplier properties (locked/not), hence comparable.

From what I read, some might have it locked upwards, while allowing just a few settings lower - and some opens all lower but still with the original multi as a max upwards. And that the QS ones would be more likely to be locked in general(??).

Oh well, whenever you get around to testing yours, could you give us the lowdown (heh) on the results ? 😁

Surely 😉

I know I should have at least 1 QS/ES. I 'might' have another one, but I can't remember.

Frankly I'm not even sure if the chip is functional, 🤣

But indeed, ES's seem to have a higher probability of being unlocked then the QS's

I've done a couple tests.

Summary:

Good news:The chip works and appears to have selectable multipliers ranging from 7.5x to 16x
Bad news:The results were quite flaky

I used a TUSL2-C (unknown BIOS version) with 64MB PC-133 and the QS/ES. The chip is obviously a Tualatin core with 256kb cache.

I first tried setting the busses to their lowest, which is 66/100/66 iirc. FSB was 66 to prevent accidental damage in case of an overclock
I first tried 16x which resulted in the board booting at 266Mhz, multi was 4x
I then tried 14x (which is the maximum normal multi for Tualatin) and it ran at 933Mhz with 66Mhz FSB, 🤣!
Next was 15x, result was 266Mhz again
7.5x worked normally

Then I upped the FSB to 133Mhz and 7.5x multi, but the board took much longer to post and ended up posting at 500Mhz with FSB 133
Then I tried 100Mhz FSB and 7.5x multi twice but I got loooong single repeating beeps only.

Full chip details
Written on the chip is:

INTEL (bla bla) 01 MALAY
Q204A515-0010 QID2QS

(Here comes the heatspreader)

RK80530RY017256

Apparently RK80530RY017256 stands for a Celeron 1400, which was also what the board identified the chip with.

Hope this helps 😉

Many thanks for taking the time to test 😎 !
Not to mention how quick the response!!

Good news:The chip works and appears to have selectable multipliers ranging from 7.5x to 16x

erm... well, I have to interpret that as 'bad' news then, minimum 7.5x sound very high indeed... 😢
(or the 'random' 4.0x for that matter)

But at least it helps somwhat vs my quest for finding the fastest pentium that will do 133mhz ...
So, seems tualatins(even as celerons) are out then, probably better start looking for the special coppermines - if there are ANY of those that will do 2.0x that is...

Well, at least this one has selectable multipliers. For all I know it could've been fully locked.

If anyone else here also has an ES/QS, feel free to chip in 😉

Tetrium,
when trying multipliers on a Tualatin, or on any Pentium for that matter, It is actually better to just talk about the 4 request bits then about BIOS multiplier setting. Like my gigabyte board has just 4 jumpers, giving you 16 request signals. It gives you full control.

My other boards have a few BIOS multiplier settings, never up to 16 choices, and not very consistent for different Pentium II / III / Tualatins. The latter boards just give you a few of the "4 jumper" settings, hidden behind multiplier numbers that may not work as such.

So for testing multipliers, a jumpered board is far superior: No limit on settings and less confusion.

gerwin:
I have a softbios, with multis ranging from 2.0x all the way to 12.5x(in 0.5x steps) - but I take it the 16 signals possible will limit which ones that actually work as the 'face value'.
So, would that mean a jumpered board would hinder the 2.0x from becoming a 3.0/4.0x on a coppermine or tualatin?

Btw, what would be the highest multi you can effectively do with those 16 settings on your particular jumpered board.
10x 66mhz ? 11x ?

bestemor wrote:

Btw, what would be the highest multi you can effectively do with those 16 settings on your particular jumpered board. 10x 66mhz ? 11x ?

I guess you did not get the idea. It all depends on how the CPU interprets the request signals. With four jumpers one can send every possible request signal combination, from the highest to the lowest that the CPU can accept and understand.

Gerwin wrote:

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.

See my list on page 2 (above the photo) for an example. These 16 4-bit settings that I tried there are all that is possible with any PII/PIII. The resulting observed multipliers are all that is possible with the tested CPU.

You see the highest that particular P-III could do is 8.0x. So when Intel releases CPU with a 9.0x Default multiplier, they will have to make room for that setting, thus sacrifice another (lower) setting on that CPU. After all they have only 16 settings max.

gerwin:
I only did a quick test at bestemor's request, it's a "take it as it is" review 😉
I tested the chip with the 1 Tualatin board I happened to pick from top of my "Tualatin board pile" and did it only to see if it had a free multiplier at all 😉
Btw, the BIOS interfaced the multiplier with a multiplier option, not with the 4 mentioned pins 🤣

If you have any tips regarding a board or a feature that's more flexible then TUSL2-C's crummy BIOS, by all means, enlighten me 😉

Wyi, I forgot to type all the available settings I had with this particular combo:
7.5x all the way to 12x in 0.5x intervals. After that it was 13x to 16x in 1.0x intervals. Total number of choices given was thus 14.

gerwin wrote:

So for testing multipliers, a jumpered board is far superior: No limit on settings and less confusion.

Cheers mate 😉

Next time I'll see if any of my Tualatin boards can use multiplier jumpers 😉
I only used the BIOS for convenience

No problem, whenever you have time.
For the Tualatin you could use the same board as me, but with a powerleap adapter? Unlike 440BX's I don't know too much about native Tualatin boards.

Which board did you use? Good chance I have one laying around anyway, 🤣. Just not sure if I have the correct Slot 1 adapter for it though and using a linlin and slotket is not something I can be bothered with at this time (I'm in the middle of a major overhaul of my attic). I'd rather just use a Tualatin board

You see the highest that particular P-III could do is 8.0x. So when Intel releases CPU with a 9.0x Default multiplier, they will have to make room for that setting, thus sacrifice another (lower) setting on that CPU. After all they have only 16 settings max.

That's what I was afraid of.... 😵

And after re-reading page 2 closely, I think I'm starting to get it now, maybe... 🤣
Sorry for the 'confusion'!

bestemor wrote:

That's what I was afraid of.... :dead: […]
Show full quote

You see the highest that particular P-III could do is 8.0x. So when Intel releases CPU with a 9.0x Default multiplier, they will have to make room for that setting, thus sacrifice another (lower) setting on that CPU. After all they have only 16 settings max.

That's what I was afraid of.... 😵

And after re-reading page 2 closely, I think I'm starting to get it now, maybe... 🤣
Sorry for the 'confusion'!

I wonder...does anyone happen to have a PPGA Celeron ES?
One of those is definitely on my wanted list!

Finally obtained a Klamath core Pentium II.
It is a 233MHz part, SL2HD, Malay, Produced week 08 1998.
Tested this processor on an i440BX board, the Asus P2B rev 1.04. With its VRM chip replaced and an AGP divider jumper added.

I found multiplier jumper BF3 is ignored, jumpers BF0/BF1/BF2 work like this:

1BF0 BF1 BF2
2O - X - O -> No Post
3O - O - O -> 133/66MHz (2.0x),    No L2, Speedsys 154,80
4X - X - X -> 133/66MHz (2.0x),    No L2, Speedsys 154,80
5X - O - O -> 133/66MHz (2.0x), 512kB L2, Speedsys 154,80  <--- !!!
6O - X - X -> 166/66MHz (2.5x), 512kB L2, Speedsys 193,52
7X - O - X -> 200/66MHz (3.0x), 512kB L2, Speedsys 232,26
8O - O - X -> 233/66MHz (3.5x), 512kB L2, Speedsys 270,98
9X - X - O -> 266/66MHz (4.0x), 512kB L2, Speedsys 309,70
10
11(O means 1-2 is jumpered. X means 2-3 is jumpered.)

Now other people have done this before, but there were some doubts about wheter L2 cache remains functional at 2.0x on an i440BX.
As to the upper limit: This CPU runs 300MHz @ 100FSB and 266MHz @ 133FSB, but does not boot at 350MHz @ 100FSB or more.

For changing the FSB speed on several i440BX mainboards there is also a DOS program, which also works in Windows 98: SMB by Rayer. Combined with a Dos Batch selection menu it works like SoftFSB did. Tested OK with PLL chips of the type ICS9148xx-26 and ICS9150xx-08.

Edit: corrected processor speed
.