j^aws wrote:

You can see above that NSSI reports for an Ezra-T that both processor and co-processor (FPU) are running at the same speed of 10 […]
Show full quote

matze79 wrote:

C3's FPU works only with half CPU Clock.

DSC_0325.JPG

DSC_0326.JPG

You can see above that NSSI reports for an Ezra-T that both processor and co-processor (FPU) are running at the same speed of 1072 MHz (this is slightly off anyway, but the point being the same speed). The later Nehemiah should also be running the same. Even then, these FPUs are still fairly weak. Nehemiah being stronger than Ezra-T.

I doubt the accuracy of NSSI on this matter. I checked the datasheets of Ezra-T and EBGA Ezra-T and they both confirm that the FPU runs at 50% CPU speed.
Nehemiah runs at full speed.

gdjacobs wrote:

Okay, so looking at the 440bx specification, DRAM timing data is on register 58h. The way I read it, I would OR the contents of […]
Show full quote

Okay, so looking at the 440bx specification, DRAM timing data is on register 58h. The way I read it, I would OR the contents of that register with 3h or 0h and write to the register to add or remove additional CAS and RAS wait states. Sound correct?

Edit:
Wait states are on by default (3h is set). Looking at register 77:76, I'm hoping to change from CAS 2 to CAS 3 timings. I try altering bit 2 by inputing

Copy code to clipboard

1pciset 7190 8086 76 00000100 00000000

The response, in German, is:

Copy code to clipboard

1before :   76h: 000.0111

and the computer hard locks.

I've tried specifying hex for the mask and value, but that doesn't seem to be accepted by the utility as nothing is changed.

Known issue? Any workarounds?

Unfortunately I don't think CAS latency can be changed after boot (I also had a hard lock when I tried it now), but you can change the other timings. It looks like you got the PCISET utility right.

Alright. I'll try flipping the other two bits and see how it works. Hopefully I can scare up a CL3 PC100 DIMM as well. To me, top end performance isn't as useful (rare?) as having a low end tuning range. DeathTrack is hellish when you've got 0.5s of missile warning!

lazibayer wrote:

j^aws wrote:

You can see above that NSSI reports for an Ezra-T that both processor and co-processor (FPU) are running at the same speed of 10 […]
Show full quote

matze79 wrote:

C3's FPU works only with half CPU Clock.

DSC_0325.JPG

DSC_0326.JPG

You can see above that NSSI reports for an Ezra-T that both processor and co-processor (FPU) are running at the same speed of 1072 MHz (this is slightly off anyway, but the point being the same speed). The later Nehemiah should also be running the same. Even then, these FPUs are still fairly weak. Nehemiah being stronger than Ezra-T.

I doubt the accuracy of NSSI on this matter. I checked the datasheets of Ezra-T and EBGA Ezra-T and they both confirm that the FPU runs at 50% CPU speed.
Nehemiah runs at full speed.

I'd be more inclined to believe the datasheets than NSSI. However, it's been a while, but I can't remember if NSSI was picking up half-speed FPU for a normal Ezra (non-T) and older cores. Also, the datasheet refers upto 933MHz cores, and I don't know if there were any changes past this, as the core I tested was a 1GHz one.

Reduced DRAM RAS and precharge timings with PCISET for a further 2% drop. I'm clocking at approximately 10.8 minimum on Speedsys, which is about a 386DX-25 to 386DX-33 territory. Aside from locating some RAM with CAS 3 SPD timings, does anyone have any other ideas to cut speed further?

Maybe the Throttle utility could be the solution for you as it leaves no TSR. Using a setting of 4 about halves the performance of my Ezra as seen in this spreadsheet under SE440BX-2 results (shortened the name of the CPU to C3E).

Perhaps the slowest possible solution without it would be with a LX based moteherboard running 3x50 with EDO RAM,. I consider buying an EDO DIMM just to try it out myself.

Yes, although I've run into two other issues which may send me looking for another Slot 1 motherboard.
1) The SE440BX-2 only supports one floppy drive. I'd prefer to run my 3.5" and 5.25" drives on the same rig.
2) A 50mhz FSB would really be preferable for even deeper slowdown capability. 50mhz + Throttle is even better than 66mhz + Throttle.

Any ideas for an early BX board which fits this description and won't require organ sales to purchase?

gdjacobs wrote:

Yes, although I've run into two other issues which may send me looking for another Slot 1 motherboard. 1) The SE440BX-2 only sup […]
Show full quote

Yes, although I've run into two other issues which may send me looking for another Slot 1 motherboard.
1) The SE440BX-2 only supports one floppy drive. I'd prefer to run my 3.5" and 5.25" drives on the same rig.
2) A 50mhz FSB would really be preferable for even deeper slowdown capability. 50mhz + Throttle is even better than 66mhz + Throttle.

Any ideas for an early BX board which fits this description and won't require organ sales to purchase?

Throttle hooks onto ACPI to slowdown and has issues with smoothness and compatibility; side-by-side comparisons with speed-sensitive demoscene programmes resulted in corrupted graphics.

For 50 MHz BX boards, the Gigabyte 6BXC rev 2.0 is good, and I picked one up for very little, although I've seen prices greatly fluctuate.

Certain VIA Apollo-based boards, even with SB-Link, can go slower. These can have 66 MHz FSB, and still achieve slower speeds by having more tweakable BIOS options, especially handy forcing -33MHz lower memory speeds. However, most have gimped speeds in DOS. I've only come across one board I like, forget its name, but it's S370 from DFI.

Another board with SIS chipset can go slower, too. But these are even less common...

j^aws wrote:

For 50 MHz BX boards, the Gigabyte 6BXC rev 2.0 is good, and I picked one up for very little, although I've seen prices greatly fluctuate.

I just bought one on e-bay -- but unfortunately you're right ...they're quite expensive now. I have been told the capacitors are not so good on these boards? I wonder if I should have them replaced... even though they are in good condition for now.

I may have given you a false alarm on the caps. I just partly recapped mine with new 330uF ones. I still had an occasional refusal to bootup, and then found that the board boots again when changing that red jumper J15 near the battery to the closed position. open= Soft Off (default) / close=Full on. It is OK now. The only remaining annoyance is that the SD card (on adapter) is only detected after a cold boot.

infiniteclouds wrote:

j^aws wrote:

For 50 MHz BX boards, the Gigabyte 6BXC rev 2.0 is good, and I picked one up for very little, although I've seen prices greatly fluctuate.

I just bought one on e-bay -- but unfortunately you're right ...they're quite expensive now. I have been told the capacitors are not so good on these boards? I wonder if I should have them ... even though they are in good condition for now.

I've seen caps bulge around the AGP slot for these motherboards. For old hardware, failing caps is just an acceptance. Remember to always use quality PSUs.

I have seen that too. But that may be because of a lazy/uninformed former user. These boards have the 'voodoo jumpers' which need to be set for power hungry video cards, including Voodoo 3 and Geforce. If not then the VRM near the AGP slot will get very hot and the capacitors near it will suffer.

gdjacobs wrote:

Yes, although I've run into two other issues which may send me looking for another Slot 1 motherboard. 1) The SE440BX-2 only sup […]
Show full quote

Yes, although I've run into two other issues which may send me looking for another Slot 1 motherboard.
1) The SE440BX-2 only supports one floppy drive. I'd prefer to run my 3.5" and 5.25" drives on the same rig.
2) A 50mhz FSB would really be preferable for even deeper slowdown capability. 50mhz + Throttle is even better than 66mhz + Throttle.

Any ideas for an early BX board which fits this description and won't require organ sales to purchase?

As mentioned in another thread some revisions of the Asus P2B and P2B-D does 50MHz. Perhaps if your soldering and electronic skills are good you can make a homemade Turbo PLL to slow things down even further.

Hmm... maybe make a PLL daughterboard so you're not limited by pin form factor or BIOS compatibility.

Anyway, time to hunt for another motherboard. Anyone familiar with the Acorp 6BX? Will it melt things?

Just thought I'd share my experiences with one of the modified III-s Tualatin's that are sold on Ebay when used with my MS6905 and GA-6BXC motherboard. With any of the Tualatin supported BIOS I can get a POST and boot only if the FSB of the board is set to 100Mhz (meaning 1050Mhz max on the Tualatin). While the board has no problem with 133Mhz FSB on the VIA C3 (133x9) it will either not POST or immediately freeze on the Pentium III. I have been in touch with the seller and am told that boards that don't natively support 133Mhz FSB aren't guaranteed to work with the chip but that if I send my motherboard he can fix it..?

I'm not sure what the modification would be but the postage probably wouldn't be cheap (they're based in South Korea). There's somewhat of a language barrier but he seems very knowledgeable.

Maybe some bad capasitors on your motherboard or in your PSU causes this, a Pentium III-S 1400 pulls much more power than any VIA C3.

The powersupply is brand new - http://www.microcenter.com/product/457384/Bas … TX_Power_Supply. While there are no caps that are bulging at all I can't rule it out because I know some other users here have had bad capacitors with this board.

I ran these speeds succesfully on a GA-6BXC rev 2.0 motherboard:
Re: Tualatin Celeron vs Williamette Celeron
Note1: This was before I partially recapped the motherboard.
Note2: The FSB speeds mentioned do not necessarily mean that that was the bootup speed. Since one can change FSB speed on the fly through software.

gerwin wrote:

I ran these speeds succesfully on a GA-6BXC rev 2.0 motherboard: Re: Tualatin Celeron vs Williamette Celeron Note1: This was be […]
Show full quote

I ran these speeds succesfully on a GA-6BXC rev 2.0 motherboard:
Re: Tualatin Celeron vs Williamette Celeron
Note1: This was before I partially recapped the motherboard.
Note2: The FSB speeds mentioned do not necessarily mean that that was the bootup speed. Since one can change FSB speed on the fly through software.

Ahh, the Tualatin you were using is 133x9.5 and the III-s they modify is a 1.4Ghz 133x10.5.

Even if I boot at 100 FSB and try to change in SMB it will immediately freeze at 133Mhz. The seller offered to modify the board for free (not including shipping) and offered to do my SEBX440 as well.

He had this to say about the 6BXC:

All basic function will not be changed. Modification key point is signal quality adjustment. It is not adding or removing IC par […]
Show full quote

All basic function will not be changed.
Modification key point is signal quality adjustment.
It is not adding or removing IC part.
After measurement of signal quality, adjusting value of resistor or capacitor.

I don't have a Pentium III-S 1.4Ghz 133x10.5 at hand, but suppose I could run the Tualeron at that speed. But you can pretty much assume my setup will run that fine. There is just something with your setup that is giving trouble... Does 112 MHz FSB work? Are you sure the AGP divider jumper is set correctly (jumper no.4 of the FSB jumper block)?

It is interesting what the seller is suggesting. Would be nice to have all the details of his modification, but I understand if he wants to keep some tricks to himself.

For years I ran a Pentium III-S below max speed, around 1 GHz, on a system that was still in daily use. Since I could not get a Geforce MX440 to cooperate with 89MHz AGP, and it was fast enough already for my use case.