maxtherabbit wrote on Yesterday, 12:49:

Would you say the same thing about the ECS K7S5A? https://theretroweb.com/motherboard/image/1ds … b9950295964.jpg It has a visua […]
Show full quote

shevalier wrote on Yesterday, 09:45:

it's a piece of the not_cake, sorry. https://theretroweb.com/motherboard/image/dsc-8840-647-6889454f92782530065577.jpg This moth […]
Show full quote

ChrisK wrote on 2026-05-11, 15:34:

I have a K7SEM board which

it's a piece of the not_cake, sorry.

This motherboard has a single-phase VRM with diodes in the lower branch instead of MOSFETs, and a PWM controller from an ATX power supply unit (KA7500b).
And the Barton is rated for around 60 watts of power consumption.
You’re a very risk-taking sort of person.

Would you say the same thing about the ECS K7S5A?
https://theretroweb.com/motherboard/image/1ds … b9950295964.jpg
It has a visually similar VRM layout but tbh I don't really understand power circuit design.

Same same - KA7500. Here three parallel FET (incl. push/pull transistor driver) + diode + inductor stages - all driven by the KA7500 which is a single phase switching regulator controller.

ChrisK wrote on Yesterday, 12:25:

But while speaking about that: What would be a decent board or at least a recommendable chipset?

Something that supports DDR400 RAM, such as Via KT600 chipset. But I believe max DDR400 is 2GB using 2 slots total.

I had an Asus A7V600 for years, caps never went bad, but on another Gigabyte board from the same year those did. ECS has great caps.

ChrisK wrote on Yesterday, 12:19:

While Palomino and Thoroughbred work, underclocked Barton does not.
Shouldn't Barton take less power than the other two when underclocked, even at stock Vcore?

https://theretroweb.com/misc/documentation/26 … 27846461744.pdf
page 33 Figure 10. VCC_CORE Voltage Waveform

This is because the core voltage is lower when the ripple is the same or greater.
In the Barton series, AMD had already made it a strict requirement that VCC_CORE_AC_MIN (Maximum excursion below VCC_CORE_NOM) be taken into account.

A single-phase DC/DC converter operates without phase shifting, meaning the capacitors do not operate at a higher frequency.
The currents flowing through them and the voltage ripple across them reach their maximum values.

Put simply, VRM does not smooth out the voltage dips.
An oscilloscope would reveal everything; most likely, the ripple there exceeds 0.2+ V peak-to-peak.
Perhaps with new high-capacity polymer capacitors (around 2000+ µF each) it would work without any problems, but no one can guarantee that.

SHUTTLE MV43 v3.1

They were up to the same old nonsense with the motherboards for the Pentium 4 as well.
I wouldn’t be surprised if this Shuttle was an OEM product from ECS.

butjer1010 wrote on Yesterday, 07:28:

Get a decent Motherboard for Barton. They are pretty cheap. Don't even waste Your time with this board 😉

Well, of course. But it's just a testing platform where I wouldn't take a valuable board for.
Performance is second and running a Barton would just be nice to have but is no strict requirement.

In case it wasn't clear enough: This is more of a research to gain knowledge and I'd like to know the different requirements between Barton and other Socket A CPUs in terms of motherboard compatibility.

But while speaking about that: What would be a decent board or at least a recommendable chipset?

rmay635703 wrote on Yesterday, 07:38:

Abit NF7 is the best Socket A mbo in my subjective opinion!!!
I did overclocked my mobile Barton XP to 2,6GHz (from 1,8) if i remember god (but of course, with massive cooler and 12cm fan), but after a while, i did reduce it to 2,4GHz, and it still working on this freq...

That reply doesn’t sound like something I would write, 🤣

rmay635703 wrote on Yesterday, 16:31:

That reply doesn’t sound like something I would write, 🤣

Hehe, i got a little bit lost when deleting whole thread 😀

shevalier wrote on Yesterday, 13:47:

SHUTTLE MV43 v3.1 https://i.ebayimg.com/images/g/uVsAAOSwIzBbOMe1/s-l1600.jpg They were up to the same old nonsense with the mot […]
Show full quote

SHUTTLE MV43 v3.1

They were up to the same old nonsense with the motherboards for the Pentium 4 as well.
I wouldn’t be surprised if this Shuttle was an OEM product from ECS.

A lot of Shuttle boards from Athlon and P4 were rebranded ECS boards. Only few of them are actual Shuttle designs, excluding their XPC boards.

ABIT also used ECS for some of their boards.

shevalier wrote on Yesterday, 13:35:

https://theretroweb.com/misc/documentation/26 … 27846461744.pdf page 33 Figure 10. VCC_CORE Voltage Waveform […]
Show full quote

ChrisK wrote on Yesterday, 12:19:

While Palomino and Thoroughbred work, underclocked Barton does not.
Shouldn't Barton take less power than the other two when underclocked, even at stock Vcore?

https://theretroweb.com/misc/documentation/26 … 27846461744.pdf
page 33 Figure 10. VCC_CORE Voltage Waveform

This is because the core voltage is lower when the ripple is the same or greater.
In the Barton series, AMD had already made it a strict requirement that VCC_CORE_AC_MIN (Maximum excursion below VCC_CORE_NOM) be taken into account.

A single-phase DC/DC converter operates without phase shifting, meaning the capacitors do not operate at a higher frequency.
The currents flowing through them and the voltage ripple across them reach their maximum values.

Put simply, VRM does not smooth out the voltage dips.
An oscilloscope would reveal everything; most likely, the ripple there exceeds 0.2+ V peak-to-peak.
Perhaps with new high-capacity polymer capacitors (around 2000+ µF each) it would work without any problems, but no one can guarantee that.

So you're saying that even if a 45W TDP thunderbird was working properly in one of these boards you would expect a 45W TDP Barton-M to struggle just because the Vcore is 0.05 less?

I successfully got a Barton mobile cpu running on my ABIT KT7-RAID with bios mods and socket mod for higher multipliers, but it's a waste of time
if You want the extra performance as the cpu is strangled by 133mhz sdram.

maxtherabbit wrote on Yesterday, 23:15:

shevalier wrote on Yesterday, 13:35:

https://theretroweb.com/misc/documentation/26 … 27846461744.pdf page 33 Figure 10. VCC_CORE Voltage Waveform […]
Show full quote

ChrisK wrote on Yesterday, 12:19:

While Palomino and Thoroughbred work, underclocked Barton does not.
Shouldn't Barton take less power than the other two when underclocked, even at stock Vcore?

https://theretroweb.com/misc/documentation/26 … 27846461744.pdf
page 33 Figure 10. VCC_CORE Voltage Waveform

This is because the core voltage is lower when the ripple is the same or greater.
In the Barton series, AMD had already made it a strict requirement that VCC_CORE_AC_MIN (Maximum excursion below VCC_CORE_NOM) be taken into account.

A single-phase DC/DC converter operates without phase shifting, meaning the capacitors do not operate at a higher frequency.
The currents flowing through them and the voltage ripple across them reach their maximum values.

Put simply, VRM does not smooth out the voltage dips.
An oscilloscope would reveal everything; most likely, the ripple there exceeds 0.2+ V peak-to-peak.
Perhaps with new high-capacity polymer capacitors (around 2000+ µF each) it would work without any problems, but no one can guarantee that.

So you're saying that even if a 45W TDP thunderbird was working properly in one of these boards you would expect a 45W TDP Barton-M to struggle just because the Vcore is 0.05 less?

So you’re saying that the VRM has no other parameters apart from the TDP?
Barton https://theretroweb.com/misc/documentation/26 … 27846461744.pdf
Thoroughbred https://theretroweb.com/misc/documentation/25 … 97669335761.pdf
Thunderbird https://theretroweb.com/misc/documentation/23 … 5d545947274.pdf

P.S. If the motherboard and BIOS recognise that this is a "full mobile" processor and attempt to dynamically adjust the clock speed and supply voltage, as well as working control the StopGrant BUS Disconnect, the power consumption and voltage will fluctuate wildly within a very short space of time.
Not every motherboard—even a decent one—can cope with this.
However,, desktop motherboards with mobile processors only allow you to adjust the multiplier.
And even then, only on VIA chipsets.

Not true. You can change the multiplier (othefly) on many socket a chipsets. I tested it with AMD761 (+VIA southbridge); SIS746/748, VIA KT600 / KT133A. The Nforce and KT133 (southbridge is too old) chipsets doesn't allow the multi change.

In theory, the Barton CPU should work on every socket A board. I did manage to run a Barton CPU on a KT133 (non A) board. The BIOS won't recognize it, but it will work. The BIOS patcher can help at this point.
A SD-RAM + Barton combo doesn't make any sense imho.

Imperious wrote on Today, 00:23:

I successfully got a Barton mobile cpu running on my ABIT KT7-RAID with bios mods and socket mod for higher multipliers, but it's a waste of time
if You want the extra performance as the cpu is strangled by 133mhz sdram.

_digitalbath wrote on Today, 05:46:

Not true. You can change the multiplier (othefly) on many socket a chipsets. I tested it with AMD761 (+VIA southbridge); SIS746/748, VIA KT600 / KT133A. The Nforce and KT133 (southbridge is too old) chipsets doesn't allow the multi change.

In theory, the Barton CPU should work on every socket A board. I did manage to run a Barton CPU on a KT133 (non A) board. The BIOS won't recognize it, but it will work. The BIOS patcher can help at this point.
A SD-RAM + Barton combo doesn't make any sense imho.

If you want a motherboard with an ISA slot, DDR isn't an option. Barton's cache should be even more important with SDRAM as it somewhat mitigates low memory bandwidth.

Btw, what's that BIOS patcher you mentioned?

Mike_ wrote on Today, 09:09:

Btw, what's that BIOS patcher you mentioned?

This https://www.rom.by/articles/BP/index_english.htm

Socket A boards with ISA slot and SD-RAM also stuck at FSB 133MHz. In most of the times only a rare XP-M Barton CPU will be a solution then. There could be some obstacles with the miltipliers depending on BIOS and board.
I would also think about a "simple" 2400+ Thoroughbred CPU. Less problems and fast enough in my eyes.

BIOS patcher is a tool made by applerom from rom.by. This tool is able to patch BIOSes to fix some issues like harddrive limitation or CPU Support (i.e. AMD K6-2/3 ). There is good YT content on this topic.

_digitalbath wrote on Today, 05:46:

Not true. You can change the multiplier (othefly) on many socket a chipsets.

'on the fly' does not mean 'on demand'
https://support.ts.fujitsu.com/Dokumente/FAQ/ … 73-httpimg1.pdf
Full AMD PowerNow! functionality (with frequency and voltage adjustment based on the operating system’s requests) rarely works on desktop motherboards, if at all.
The relevant ACPI tables should be present in the BIOS.

In theory, the Barton CPU should work on every socket A board.

In theory, yes, but you haven’t actually compared the datasheets from the links I provided.
For its 130nm CPUs, AMD has introduced limits on supply voltage ripple, unlike its 180nm CPUs.
Most likely, even an ECS like this, with a more ‘conventional’ VRM, would be capable of it.
https://theretroweb.com/motherboards/s/ecs-741gx-m-rev-1-0
But with the motherboard in question, everything has been designed in the strangest possible way.

shevalier wrote on Today, 10:03:

'on the fly' does not mean 'on demand' https://support.ts.fujitsu.com/Dokumente/FAQ/ … 73-httpimg1.pdf Full AMD PowerNow! functi […]
Show full quote

'on the fly' does not mean 'on demand'
https://support.ts.fujitsu.com/Dokumente/FAQ/ … 73-httpimg1.pdf
Full AMD PowerNow! functionality (with frequency and voltage adjustment based on the operating system’s requests) rarely works on desktop motherboards, if at all.
The relevant ACPI tables should be present in the BIOS.

Desktop boards use only a part of the PowerNow feature. It was never designed to work in a desktop system. So, yes you're right.
Nevertheless it works. It is just changing some MSR register and a restart of the SIP-stream.

shevalier wrote on Today, 10:03:

In theory, yes, but you haven’t actually compared the datasheets from the links I provided. For its 130nm CPUs, AMD has introduc […]
Show full quote

In theory, yes, but you haven’t actually compared the datasheets from the links I provided.
For its 130nm CPUs, AMD has introduced limits on supply voltage ripple, unlike its 180nm CPUs.
Most likely, even an ECS like this, with a more ‘conventional’ VRM, would be capable of it.
https://theretroweb.com/motherboards/s/ecs-741gx-m-rev-1-0
But with the motherboard in question, everything has been designed in the strangest possible way.

I saw your link to the datasheets. I didn't deny it. Without measuring the ripple it is just guessing. As you already wrote, caps are crucial in that case.
Maybe you're right and the VRM design is too bad for a Barton CPU. The fact is, it was never build to support it.
I guess it will work without errors in most cases. Even with the higher ripple.

_digitalbath wrote on Today, 10:23:

Maybe you're right and the VRM design is too bad for a Barton CPU. The fact is, it was never build to support it.

In those days, engineers entertained themselves as best they could.

For example, in the AMD SLOT A (by Gigabyte, but with a design very close to the reference), the VRM contains no electrolytic capacitors at all, thanks to its genuine 4-phase DC/DC.

_digitalbath wrote on Today, 10:02:

Socket A boards with ISA slot and SD-RAM also stuck at FSB 133MHz. In most of the times only a rare XP-M Barton CPU will be a solution then. There could be some obstacles with the miltipliers depending on BIOS and board.
I would also think about a "simple" 2400+ Thoroughbred CPU. Less problems and fast enough in my eyes.

BIOS patcher is a tool made by applerom from rom.by. This tool is able to patch BIOSes to fix some issues like harddrive limitation or CPU Support (i.e. AMD K6-2/3 ). There is good YT content on this topic.

What problems are you referring to, shouldn't 166MHz FSB bartons work as well, even though at lower clock speed? For example, 3000+ would work at 1,7GHz. You could also mod multiplier using those bridges on CPU.