So I can't cite a particular example since this is something I saw it 26 years ago on a hardware review site that would show leaks from Akihabara but I understand there were a few made and sold but they were massively unstable and weren't common for a reason. Slot-A Thunderbirds did exist but were only really viable with AMD Irongate, they had issues with signaling with KX-133 boards similar to BCLOCK OC on either with Orion and Pluto cores was super limited compared to later Socket A chipsets.

From what I remember, Socket A slotkets weren't really a thing because they introduced too much frequency noise to the point that the early K7 boards couldn't tolerate it at the time.

RetroLizard wrote on Yesterday, 02:31:

cyclone3d wrote on 2026-05-30, 20:16:

Going socket A with ISA is a much better way and they are way more available than good Slot A boards.

There are Socket A boards with ISA?

A few, just off the top of my head the Abit KT7/KT7A had 1 standard.

Yes, almost all that have the kt133 or kt133a chipsets have ISA.

Then there is a single Biostat board that combined the AMD northbridge with a VIA South bridge to allow an ISA slot on a board with DDR RAM.

SteveoPDX wrote on Yesterday, 02:32:

So I can't cite a particular example since this is something I saw it 26 years ago on a hardware review site that would show leaks from Akihabara but I understand there were a few made and sold but they were massively unstable and weren't common for a reason. Slot-A Thunderbirds did exist but were only really viable with AMD Irongate, they had issues with signaling with KX-133 boards similar to BCLOCK OC on either with Orion and Pluto cores was super limited compared to later Socket A chipsets.

From what I remember, Socket A slotkets weren't really a thing because they introduced too much frequency noise to the point that the early K7 boards couldn't tolerate it at the time.

I think Anand mentioned seeing it at a trade show or something. But since Anandtech pulled all it's articles, it's going to be a bugger to dig out of waybackmachine.

I can't send out likes to posts here, but for those who took the effort, thank you for the understanding.

vanderlinde wrote on Yesterday, 00:31:

It did recommend only the use of "early" athlons Socket A, but not the thunderbird's and beyond

Exactly what '"early" athlons Socket A, but not the thunderbird's' do you have in mind?

The very first socket A CPUs used the Thunderbird core. There are no earlier.

RetroLizard wrote on Yesterday, 02:31:

[...]

There are Socket A boards with ISA?

Lots. The majority of early KT133(A) and AMD750 based board had at least one ISA slot (take Abit's KT7A as an example; Gigabyte's GA-7IXE4 even had two). After that ISA starts getting pretty rare, due to lack of southbridge support, although there's at least one (Biostar) AMD760 board with one.

BitWrangler wrote on Yesterday, 03:46:

I think Anand mentioned seeing it at a trade show or something. But since Anandtech pulled all it's articles, it's going to be a bugger to dig out of waybackmachine.

According to these Tweaker.nl posts there was actually one produced in Germany 1999 via an edition of 250.
https://tweakers.net/nieuws/13104/socket-a-gr … signaleerd.html
https://tweakers.net/nieuws/11711/slot-a-slot … et-adapter.html

But never, a released or working version. It seems that particular board only was a pin converter, but no actual VRM or any other logic like that.

I guess there was a market for it, for people who owned a Slot A board but did not want to have a upgrade path of replacing their complete motherboard + RAM.

vanderlinde wrote on Yesterday, 11:32:

But never, a released or working version. It seems that particular board only was a pin converter, but no actual VRM or any other logic like that.

According to the text there was an edition of 250 pieces produced -and- marketed.

Slot 1 slotket models with on-board VRM are very rare.
Powerleap and Kingston produced slotkets with VRM as well as Buffalo and I-O Data in Japan - but that is it as far as I know.

No one made a buffered slot 1 sloket either, it would have cost too much to be a viable product. Many cheap motherboard and sloket combinations have stability issues at 100MHz and 133MHz bus speed. Signal buffering could have possibly solved those issues.

dionb wrote on Yesterday, 09:48:

Lots. The majority of early KT133(A) and AMD750 based board had at least one ISA slot (take Abit's KT7A as an example; Gigabyte's GA-7IXE4 even had two). After that ISA starts getting pretty rare, due to lack of southbridge support, although there's at least one (Biostar) AMD760 board with one.

Yes, but if they work well is another story and depends on what you want to do. I've got DMA Clicking on all my SB16 with a KT133 Board, only my CT3670 works flawlessly. Don't know if it's just my QDI Board, but I've got other funny issues like slowdowns on other Socket A Boards too. So even with my warm feelings for the Socket A platform (I had 2 boards back then, one was the 7IXE4), I would recommend an Intel Chipset if ISA is such an important thing.

auto660 is a master of his craft. I believe that if anyone is capable of building this adapter, it is him.
However, solving the core problem requires much more than just designing a PCB. All Slot A CPUs utilize an open-drain design, whereas all Socket A CPUs use a push-pull configuration. This is probably one of the reasons why AMD switched from the slot to the socket. During system boot, the CPU requires instructions/parameter detailing its operating parameters. These are known as ROMSIP [SIP = Serial Initialization Packet]. These parameter are sent by the chipset, and the SIP values are hardwired by the chipset. Crucially, the information specifying whether to use the open-drain or push-pull design is embedded within these values. This is why all Slot A chipsets can only send opendrain SIP parameter. To compatible-fit a Socket A CPU, these SIP values must be changed via an external device, which significantly increases the complexity of the project. Nevertheless, a precedent exists: Gigabyte successfully achieved the reverse implementation with their 7IXE4 motherboard, pairing a Slot A chipset with a Socket CPU. This proves the concept is theoretically viable, and this is precisely what auto660 discovered.

dionb wrote on Yesterday, 09:48:

Lots. The majority of early KT133(A) and AMD750 based board had at least one ISA slot (take Abit's KT7A as an example; Gigabyte's GA-7IXE4 even had two). After that ISA starts getting pretty rare, due to lack of southbridge support, although there's at least one (Biostar) AMD760 board with one.

Yeah KT133 boards are about as fast as you can go using easily-available consumer grade hardware.

Socket 370 has better compatibility but is also slower and processors are rarer.

I didn't know about the Biostar motherboard. There's even an appnote explaining how to do it - https://theretroweb.com/chip/documentation/ds … fe715814471.pdf - China are you listening? 😀