Standard Def Steve wrote on 2023-08-23, 16:21:

dm- wrote on 2023-08-23, 04:28:

Screenshot 2023-08-23 at 08.20.05.png […]
Show full quote

Screenshot 2023-08-23 at 08.20.05.png

I did the mod for generic slotket and BX440 mobo GA-6BXC and have some stats:

1166 PIII-S does not start on 133mhz, 112Mhz stable.
1266/1400 PII-S starting, working, gets stuck under heavy load. 112Mhz stable.

Cel 1100, 1200 @ 112Mhz (1.34ghz) works super stable.

i tried 3 generic slotkets and all of them won't work @ 133Mhz. maybe it's just my mobo issues. But when i use P-III Coppermine (650/100/256 Slot1) @ 133Mhz - it is also super stable...

I don't know what it is about PIII-S CPUs, but they generally seem to be harder to drive for older motherboards. I've seen it a couple of times now: your FSB133-ready Slot 1 or S370 motherboard happily runs Coppermine and adapted Tualatin-256 CPUs at crazy, overclocked bus speeds. But drop in a PIII-S, and if you're lucky the board will let you run it at stock bus speeds. There's a good chance, though, that you'll have to underclock the bus a bit (or a lot; perhaps all the way down to 100 MHz) to stabilize it.

Perhaps it's because PIII-S is intended for servers, but something is slightly different about the way the CPU operates. It seems to requires higher drive strength than Tualatin-256, which not all Coppermine-native boards can provide.

The only difference is the 512k cache ..not sure how much harder it is to drive 512k over 256 however but perhaps its the slight bump of Vcore voltage required to drive that extra cache, pretty sure SMP and ECC are available on all Tualatin CPUs so we can rule them out.

Trashbytes wrote on 2023-08-23, 16:29:

Standard Def Steve wrote on 2023-08-23, 16:21:

dm- wrote on 2023-08-23, 04:28:

Screenshot 2023-08-23 at 08.20.05.png […]
Show full quote

Screenshot 2023-08-23 at 08.20.05.png

I did the mod for generic slotket and BX440 mobo GA-6BXC and have some stats:

1166 PIII-S does not start on 133mhz, 112Mhz stable.
1266/1400 PII-S starting, working, gets stuck under heavy load. 112Mhz stable.

Cel 1100, 1200 @ 112Mhz (1.34ghz) works super stable.

i tried 3 generic slotkets and all of them won't work @ 133Mhz. maybe it's just my mobo issues. But when i use P-III Coppermine (650/100/256 Slot1) @ 133Mhz - it is also super stable...

I don't know what it is about PIII-S CPUs, but they generally seem to be harder to drive for older motherboards. I've seen it a couple of times now: your FSB133-ready Slot 1 or S370 motherboard happily runs Coppermine and adapted Tualatin-256 CPUs at crazy, overclocked bus speeds. But drop in a PIII-S, and if you're lucky the board will let you run it at stock bus speeds. There's a good chance, though, that you'll have to underclock the bus a bit (or a lot; perhaps all the way down to 100 MHz) to stabilize it.

Perhaps it's because PIII-S is intended for servers, but something is slightly different about the way the CPU operates. It seems to requires higher drive strength than Tualatin-256, which not all Coppermine-native boards can provide.

The only difference is the 512k cache ..not sure how much harder it is to drive 512k over 256 however but perhaps its the slight bump of Vcore voltage required to drive that extra cache, pretty sure SMP and ECC are available on all Tualatin CPUs so we can rule them out.

I'm thinking it has more to do with AGTL signaling strength than Vcore/VRM. Like, since PIII-S is meant for servers, it needs a slightly cleaner bus signal than some older consumer boards can provide.

That's totally just a guess though.

As mmx_91 already stated lower voltage is a bad thing, means more amps - same logic as with EVs.
All it will take to answer is a person with scope and one of the failure cases to probe at. I would start with monitoring VTT.

i controlling voltage, some of cpu's (i have many of p-3s from scrap yard) doing 1.67, so i have to not ground VID2 to get 1.45 from mobo. some cpu's set voltage correctly and i have 1.5 from mobo.

the thing is, all of 1133Mhz P3-s won't start @133 in SlotKet! all of them. some won't post at all, some can run and post for few seconds... Same for Celerons 1100/1200. (O/C@133)
and of course all cpu's work fine on VIA s370 modded mobo @ 133. Celerons also work fine in slotket @112 (O/C from 100). I pretty sure this is slotket or mobo or slotket/mobo compatibility issues.

have to find one more suitable slot1 mobo on my scrap yard and i'll find out more.

Always said, no slotket - no problem.
Aopen Mx3s - 150FSB with P3-s 1113.
Cubx-L - 133FSB with the same P3-s.
Except perhaps not with 1GB of RAM, because I can’t find last stick of RAM PC133 with CL2.

mmx_91 wrote on 2023-08-23, 11:59:

Mmm that would be one of the differences between working Coppermine and non working Tualatins. […]
Show full quote

H3nrik V! wrote on 2023-08-23, 11:25:

shevalier wrote on 2023-08-23, 11:14:

I don't think.
High-frequency Coppermine also consumes a lot of energy.

And probably more than the Tualatin cores?

Mmm that would be one of the differences between working Coppermine and non working Tualatins.

An average around 1Ghz Coppermine pulls around 30w at 1.75v = 17.14A. For a 30wTualatin working at 1.45A, it will pull near 21A.

Checking documentation I found for example an i810 official Intel board manual that says don't use cpus that pull more than 22A, but I suppose it just depends on the board design!

Talking about my Tekram board, the 1266 P3-S I'm refering to is a modded one by the Korean guy. It works perfectly fine at 1266Mhz in other S370 boards so the issue here might be either the slotket or the board that refuses to work at above 1050Mhz or so using 133 FSB.

Good point there, however, I was under the impression that the Tualatins had lower power consumption as well, and not only same power at lower voltage ...

shevalier wrote on 2023-08-23, 03:55:

As with any popular (albeit long-standing) topic, a more of urban legends was born. 1. Check connection G35 to G37 at socket. At […]
Show full quote

As with any popular (albeit long-standing) topic, a more of urban legends was born.
1. Check connection G35 to G37 at socket. At any copermine-ready system its must be connected by default. But there are exceptions, so you need to make sure. All Vtt does not need to be checked, it is enough that it enters several pins.
2. Isolated AJ3, AK4.
- AJ3 - its RESET and must be float,
- AK4 (CPU side) is Vttpwrgd and must be pulled to any ~1.5V source. Vtt OR Vcc. In general, it doesn’t matter, as long as there is a logical 1. Easiest way - AJ5 or AM4. It is easier to wind the wire on AM4.
3.1. If your adapter has voltage jumpers (Or you will wind the wires on the VID to get the desired voltage - like AM34-AK36-AL37) , you don't need to isolate AH3. The Vid must be in Hi-Z state, otherwise, from noise, the input will randomly switch and change the state of the Vids.
You set any voltage (1.45 or 1.5V) and that's it.
Just make sure Vid4 is properly grounded. Or get 2+ Volts on the CPU.
3.2. If you mod the socket itself so that any Tualatins can be inserted, then VID4(the same Vid25mv) in the socket is removed and the hole in the board itself is grounded.
On Dyn-OE AN3 at CPU side (connects to Vtt, for example, to AN11).
Then the processor itself will set the voltage.

Thats all.

Thanks for all the input!

Regarding 1. - I checked and there was no continuity between G35 and G37. soldered a small jumper wire to the back of the slotket and now I have continuity between those pins in the socket 😀

Regarding 2. - Shouldn't I also isolate AN3(I don't know why, just that everyone seems to have done that 😁 )?

Regarding 3.1 - My slotket does not have a voltage jumpers so gonna have to solder the VIDpins. You mention to bridge AM34-AK36-AL37, that will result in 1.5v if I understand it right. But if I also bridge AM36 so it reads like: AM34-AK36-AM36-AL35, then I'll get 1.45v right? Which is the default for SL657.
But you also mention that there is no need to isolate AH3, there doesn't seem to be such a pin, did you mean AN3?

Do you know why people seems to have success with some Coppermine compatible slotket models , while other models doesn't seem to work. Could it be down to wrongly made modification, incompatible motherboards or simply that they are not equally "made"?

Trashbytes wrote on 2023-08-23, 11:56:

H3nrik V! wrote on 2023-08-23, 11:25:

shevalier wrote on 2023-08-23, 11:14:

I don't think.
High-frequency Coppermine also consumes a lot of energy.

And probably more than the Tualatin cores?

IIRC copper mine is 1.6 - 1.8v while Tualatins are 1.45 - 1.5v so Tualatins should be easier on the VRM, though this may be offset a bit due to Tualatin generally running faster clocks with a 133FSB. The biggest issue with Tualatins and Slotkets is the Motherboard being able to supply the 1.45 volts, its not a normal voltage for Slot 1 so many VRMs on slot 1 boards dont natively support it.

I wouldn't assume Tualatins are easy on VRMs. Sure, you're running a lower voltage, but you're pulling 30+ watts. You're probably drawing a lot of current.

rasz_pl wrote on 2023-08-24, 01:28:

As mmx_91 already stated lower voltage is a bad thing, means more amps - same logic as with EVs.
All it will take to answer is a person with scope and one of the failure cases to probe at. I would start with monitoring VTT.

smtkr wrote on 2023-08-24, 23:10:

Trashbytes wrote on 2023-08-23, 11:56:

H3nrik V! wrote on 2023-08-23, 11:25:

And probably more than the Tualatin cores?

IIRC copper mine is 1.6 - 1.8v while Tualatins are 1.45 - 1.5v so Tualatins should be easier on the VRM, though this may be offset a bit due to Tualatin generally running faster clocks with a 133FSB. The biggest issue with Tualatins and Slotkets is the Motherboard being able to supply the 1.45 volts, its not a normal voltage for Slot 1 so many VRMs on slot 1 boards dont natively support it.

I wouldn't assume Tualatins are easy on VRMs. Sure, you're running a lower voltage, but you're pulling 30+ watts. You're probably drawing a lot of current.

The thing is, you can overvolt PIII-S to 1.5v (which is what the generally problem-free Tualatin-256 runs at), or undervolt Tualatin-256 to 1.45v. It doesn't change the outcome if you happen to own one of the "problem" boards. In fact, I run a 1.4Ghz Celeron at 1.3v in my Dell XPS-T machine and it's perfectly stable; however, the same machine usually doesn't make it past POST or Win98 splash with a PIII-S (at 1050/100/1.45v). That's why I'm around 60% sure that the issue has more to do with differing AGTL signal quality tolerances between desktop Tualatins and their server bound counterparts.

And that's probably why some manufacturers kept the PIII-S off of their Tualatin desktop board CPU support documentation. Although, tbh every Tualatin-ready desktop board I've owned has had absolutely no problem running (and even overclocking) PIII-S CPUs.

nizce wrote on 2023-08-24, 19:33:

Regarding 2. - Shouldn't I also isolate AN3(I don't know why, just that everyone seems to have done that 😁 )?

Everyone hopes that the float pin will pull the charge and there will be a logical "1".
With all my tualatins, "1" is very erratic. After a couple of minutes, it often drops to "0".
If the supply voltage is set by external wires, then it makes no sense to isolate it.

Regarding 3.1 - AM34-AK36-AM36-AL35, then I'll get 1.45v right?

Yes, you can set 1.45V if you want. But there is a subtlety.
If the motherboard is not capable of voltages less than 1.8V, then any combination of less than 1.8V will lead to no voltage.

I would recommend inserting an adapter without a processor into the board slot and turning it on.
And measure the voltage. If 1.45 (or 1.5), then everything is fine.
If not, then you need to figure it out

are not equally "made"?

this.

shevalier wrote on 2023-08-25, 06:02:

Everyone hopes that the float pin will pull the charge and there will be a logical "1". With all my tualatins, "1" is very errat […]
Show full quote

nizce wrote on 2023-08-24, 19:33:

Regarding 2. - Shouldn't I also isolate AN3(I don't know why, just that everyone seems to have done that 😁 )?

Everyone hopes that the float pin will pull the charge and there will be a logical "1".
With all my tualatins, "1" is very erratic. After a couple of minutes, it often drops to "0".
If the supply voltage is set by external wires, then it makes no sense to isolate it.

Regarding 3.1 - AM34-AK36-AM36-AL35, then I'll get 1.45v right?

Yes, you can set 1.45V if you want. But there is a subtlety.
If the motherboard is not capable of voltages less than 1.8V, then any combination of less than 1.8V will lead to no voltage.

I would recommend inserting an adapter without a processor into the board slot and turning it on.
And measure the voltage. If 1.45 (or 1.5), then everything is fine.
If not, then you need to figure it out

are not equally "made"?

this.

Thank you, i’ll be sure to measure the voltage without a CPU in the slotket before actually trying 😀

By the way, how its do ECS in 100 000 series of matherboard.

Note that this was the Dodgy Cheap Slocket that was at computer shows until the end of (their) time that didn't meet specs. If you can get a better one, get one.

Managed to get my hands on a MS-6905 Master v2.
So if the soon modded super slocket iii doesn’t work, I’ll try that one instead. Seems that more people have reported success with that one.

It also supports setting the voltage via jumpers, so perhaps only isolating the pins is needed(and the AK4 bridge I guess)

nizce wrote on 2023-08-26, 09:51:

and the AK4 bridge I guess

Rumor has it that there are Tualatins who work without that.
But personally, I have not met.

PS. The BIOS should already be Tualatin ready by the time the processor is inserted.
It is better to start with Coppermine and update the BIOS.

shevalier wrote on 2023-08-25, 06:02:

Everyone hopes that the float pin will pull the charge and there will be a logical "1". With all my tualatins, "1" is very errat […]
Show full quote

nizce wrote on 2023-08-24, 19:33:

Regarding 2. - Shouldn't I also isolate AN3(I don't know why, just that everyone seems to have done that 😁 )?

Everyone hopes that the float pin will pull the charge and there will be a logical "1".
With all my tualatins, "1" is very erratic. After a couple of minutes, it often drops to "0".
If the supply voltage is set by external wires, then it makes no sense to isolate it.

Regarding 3.1 - AM34-AK36-AM36-AL35, then I'll get 1.45v right?

Yes, you can set 1.45V if you want. But there is a subtlety.
If the motherboard is not capable of voltages less than 1.8V, then any combination of less than 1.8V will lead to no voltage.

I would recommend inserting an adapter without a processor into the board slot and turning it on.
And measure the voltage. If 1.45 (or 1.5), then everything is fine.
If not, then you need to figure it out

are not equally "made"?

this.

Sorry for another stupid question 😀
How do I easiest and safest measure the voltage before inserting the CPU in the slotket? Is it between a random Vcc pin and a random Vss pin on the socket on the slotket?

Then again, if I have no CPU installed when testing and since I can’t set the voltage on the current slotket I guess the VRM on the motherboard doesn’t know what voltage to send to the CPU/slotket?
I have a P2B-F(with custom bios installed which should include microcodes for tualatin)where I cannot set the voltage in Bios either, but the voltage regulator chip is a HIP6019BCB. So if I understand it correctly is should be able
To handle down to 1.3v. But I guess I need a CPU installed with the VID mod set to 1.45 for it to actually supply that?

dm- does your bios have cpu microcode for tualatin? maybe it needs one for cpu to fly at 133MHz

nizce wrote on 2023-08-27, 11:47:

Is it between a random Vcc pin and a random Vss pin on the socket on the slotket?

For the CPU, it doesn't matter if it's 1.45V or 1.42V.
Therefore, it is worth measuring behind the inductance on the main board.
As a minus, you can use any connector shield, for example, the USB.

PS. Yes, it is the correct voltage measurement that is between the pins on the socket. But this is not important in this case, the main thing is to determine that there is not 2.3V or 1.8V.

doesn’t know what voltage to send to the CPU/slotket?

If the pins on the adapter are grounded, then the ungrounded pins will be a logical "1".
Thus, regardless of whether the processor is inserted, the voltage will be the same.
If you have already started soldering the adapter, then solder wire to VIDs =1.45V on the pins to duplicate the CPU voltage state.
This will not affect jumperfree mode

I manager to get ahold of a MS6905 Master rev.2.

Tried without the VID mod since it has jumpers to set voltage on the slotket.
So just isolated the 3 PINs(also bridging AK4-AN11.

Managed to get it to boot the first time at 100FsB. But it froze when identifying the disks. And the subsequent boot tries nothing happened(black screen).

The first time when it did boot, the first thing I did was to go into BIoS and looked at the vcore voltage and it reported 1.4v.

Don’t know if the CPU is broken now somehow, but the motherboard still works with another CPU so that’s good 😀

nizce wrote on 2023-08-30, 15:49:

The first time when it did boot, the first thing I did was to go into BIoS and looked at the vcore voltage and it reported 1.4v.
Don’t know if the CPU is broken now somehow, but the motherboard still works with another CPU so that’s good 😀

Congratulations, the case has moved forward.
Check the connection between pins G35 and G37 on this adapter as well.
Well, add the supply voltage to 1.5V. And you should set the bus frequency with jumpers forcibly.
download/file.php?id=96958