vanderlinde wrote on 2026-05-30, 19:43:

It seems that ChatGPT is able to craft, design me a fully working slotket for Slot A that you could use, more future CPU's (Sock […]
Show full quote

It seems that ChatGPT is able to craft, design me a fully working slotket for Slot A that you could use, more future CPU's (Socket A) into your K7 / Slot A Motherboard.

The attachment d21d0ba0-9492-4994-bcc4-0039c6aaf7ff.png is no longer available

But, since the project requires lots of work, i'm stepping away from it. It's likely the other way around as well, where you can insert K6 based CPU's into a K7 based Board.

Note: a slotket for a Slot A has never been made. After Slot A it was S462 or Socket A. AI Summerize:

Copy code to clipboard

1Concept Project: Active Slot A → Socket A Adapter / “AMD Slotket”
2
3I’m exploring the design of an AMD Slot A to Socket A / Socket 462 adapter, basically an AMD version of an Intel slocket, but much more complex.
4
5The goal would be to allow selected Socket A Athlon/Duron CPUs to run on older Slot A motherboards, such as AMD 751/Irongate or VIA KX133-based boards. This would not be a simple passive pin adapter. It would likely need to be an active interposer board with startup-configuration logic, voltage handling, multiplier/VID controls, and possibly BIOS support.
6
7Basic idea
8
9The adapter would plug into the original Slot A edge connector on the motherboard and provide a Socket A / Socket 462 ZIF socket on top.
10
11The concept board would include:
12
13Slot A gold-finger edge connector
14Socket A / Socket 462 ZIF socket
15CPLD/FPGA for startup/ROMSIP patching
16VID/FID jumper blocks for voltage and multiplier control
17Voltage-regulation and filtering components
18Debug/test headers
19Mechanical support bracket for the Socket A cooler
20Optional adapter ID or configuration EEPROM/registers
21
22I made some concept drawings showing:
23
24The adapter by itself
25The adapter installed on a Slot A motherboard
26The adapter installed with a Socket A heatsink/fan cooler
27
28These are only concept renders for discussion, not final CAD.
29
30Why this is difficult
31
32Slot A and Socket A are both AMD K7-family platforms, but they are not directly interchangeable.
33
34The biggest issue is not just pin mapping. The hard part appears to be the bus startup configuration, especially the difference between older Slot A-style signaling and later Socket A behavior. Some Socket A CPUs may need push-pull bus driver configuration, while older Slot A boards expect different startup behavior.
35
36Because of that, a passive adapter would probably not work reliably. The design would likely need a small CPLD or FPGA to intercept or patch the CPU startup/ROMSIP configuration before the CPU begins normal bus communication.
37
38First realistic target
39
40The first realistic test target would be conservative:
41
42Slot A board: AMD 751/Irongate or VIA KX133 board
43CPU: early Socket A Thunderbird or Duron
44FSB: 100 MHz / 200 MT/s effective
45Goal: POST, enter BIOS, boot DOS/Linux, then stress-test
46
47Later CPUs such as Palomino, Thoroughbred, or Barton might be possible later, but they add more problems: BIOS support, multiplier mapping, power draw, and FSB compatibility.
48
49BIOS / firmware idea
50
51The first step should probably be patched stock BIOS, not a full open-source BIOS.
52
53The original BIOS already knows how to initialize the motherboard chipset, SDRAM, Super I/O, PCI routing, AGP, and onboard devices. Replacing all of that from day one would make the project much harder.
54
55Suggested firmware path:
56
57Build Rev-A hardware prototype
58Use a patched stock Award/AMI BIOS to accept the new CPU
59Add or modify CPU ID / microcode-style handling where possible
60Adjust ROMSIP/startup tables if needed

…Show last 58 lines

61Once hardware is proven, investigate a coreboot + SeaBIOS port
62
63A future open-source firmware path could be:
64
65coreboot + SeaBIOS + slotket-specific hardware detection + CPLD startup patching
66
67However, coreboot cannot fix the problem if the CPU never reaches executable code, so the adapter hardware still needs to handle the early startup configuration.
68
69Rough cost estimate
70
71This would not be a cheap adapter at prototype stage.
72
73Estimated hobby prototype cost:
74
75First bare prototype PCBs: around $150–$700
76Parts per adapter: around $100–$400
77Assembly/rework/tools/test hardware: $300–$1,500+
78First serious prototype effort: roughly $800–$2,500 if tools are already available
79More realistic full bench/debug budget: $2,000–$5,000+
80
81If the design were proven and produced in a small batch, finished units might still cost around $250–$500 each, depending on sockets, CPLD programming, brackets, testing, and documentation.
82
83Main risks
84Incorrect Vcore/VID could kill CPUs or motherboards
85Wrong bus mode could cause electrical contention
86EV6 signal routing may be sensitive to trace length and impedance
87Slot A boards may not provide enough current for later Socket A CPUs
88BIOS may fail before useful debugging is possible
89Socket A cooler load needs mechanical support to avoid damaging the slot or CPU die
90Current project status
91
92At this stage this is a concept/proposal, not a working product.
93
94The next useful steps would be:
95
96Choose one target motherboard family
97Build a full Slot A ↔ Socket A pin and signal map
98Identify all required power, clock, reset, VID, FID, and ROMSIP lines
99Design a minimal Rev-A test adapter
100Add CPLD/FPGA logic for startup configuration experiments
101Patch stock BIOS for basic CPU recognition
102Test with cheap/sacrificial Slot A boards and early Socket A CPUs
103Help wanted / discussion points
104
105I’m especially interested in input from people with experience in:
106
107AMD Slot A electrical design
108Socket A / Socket 462 pinout and signal behavior
109AMD 751/Irongate and VIA KX133 chipsets
110ROMSIP / SIP tables
111Award/AMI BIOS modification
112coreboot on old K7-class hardware
113High-speed PCB layout for EV6 bus signals
114CPLD/FPGA interposer logic
115Vintage AMD motherboard repair/testing
116
117The goal is to find out whether a practical AMD Slot A → Socket A active slotket can be built, and what the cleanest path would be for hardware and BIOS support.

What are you expecting now?

Nothing,

I just figured, maybe there's someone out there who's able to design a fully working Slotket device for Slot A. Intel had it and supported various chips. I don't have the equipment let alone patience or time to actually start doing it.

At lost it would only support Thunderbird CPUs on the few Slot A boards that support Thunderbird Slot A CPUs unless somebody wants to modify BIOSes to support even newer CPUs.

Then you are stuck with a slow-ish 200 (100 ddr) fsb and slow RAM.

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

Almost all Slot A boards were immensely more flakey and finicky than even the earliest Socket A boards.

I think the slotket could be redesigned to support, older CPU's as well, as weird as it may sound. It requires a lot of work, not just pin wise but also controller, signal, BIOS or chipset wise.

But it's possible.

vanderlinde wrote on 2026-05-30, 19:43:

It seems that ChatGPT is able to craft, design me a fully working slotket for Slot A that you could use, more future CPU's (Sock […]
Show full quote

It seems that ChatGPT is able to craft, design me a fully working slotket for Slot A that you could use, more future CPU's (Socket A) into your K7 / Slot A Motherboard.

The attachment d21d0ba0-9492-4994-bcc4-0039c6aaf7ff.png is no longer available

But, since the project requires lots of work, i'm stepping away from it. It's likely the other way around as well, where you can insert K6 based CPU's into a K7 based Board.

Note: a slotket for a Slot A has never been made. After Slot A it was S462 or Socket A. AI Summerize:

Copy code to clipboard

1Concept Project: Active Slot A → Socket A Adapter / “AMD Slotket”
2
3I’m exploring the design of an AMD Slot A to Socket A / Socket 462 adapter, basically an AMD version of an Intel slocket, but much more complex.
4
5The goal would be to allow selected Socket A Athlon/Duron CPUs to run on older Slot A motherboards, such as AMD 751/Irongate or VIA KX133-based boards. This would not be a simple passive pin adapter. It would likely need to be an active interposer board with startup-configuration logic, voltage handling, multiplier/VID controls, and possibly BIOS support.
6
7Basic idea
8
9The adapter would plug into the original Slot A edge connector on the motherboard and provide a Socket A / Socket 462 ZIF socket on top.
10
11The concept board would include:
12
13Slot A gold-finger edge connector
14Socket A / Socket 462 ZIF socket
15CPLD/FPGA for startup/ROMSIP patching
16VID/FID jumper blocks for voltage and multiplier control
17Voltage-regulation and filtering components
18Debug/test headers
19Mechanical support bracket for the Socket A cooler
20Optional adapter ID or configuration EEPROM/registers
21
22I made some concept drawings showing:
23
24The adapter by itself
25The adapter installed on a Slot A motherboard
26The adapter installed with a Socket A heatsink/fan cooler
27
28These are only concept renders for discussion, not final CAD.
29
30Why this is difficult
31
32Slot A and Socket A are both AMD K7-family platforms, but they are not directly interchangeable.
33
34The biggest issue is not just pin mapping. The hard part appears to be the bus startup configuration, especially the difference between older Slot A-style signaling and later Socket A behavior. Some Socket A CPUs may need push-pull bus driver configuration, while older Slot A boards expect different startup behavior.
35
36Because of that, a passive adapter would probably not work reliably. The design would likely need a small CPLD or FPGA to intercept or patch the CPU startup/ROMSIP configuration before the CPU begins normal bus communication.
37
38First realistic target
39
40The first realistic test target would be conservative:
41
42Slot A board: AMD 751/Irongate or VIA KX133 board
43CPU: early Socket A Thunderbird or Duron
44FSB: 100 MHz / 200 MT/s effective
45Goal: POST, enter BIOS, boot DOS/Linux, then stress-test
46
47Later CPUs such as Palomino, Thoroughbred, or Barton might be possible later, but they add more problems: BIOS support, multiplier mapping, power draw, and FSB compatibility.
48
49BIOS / firmware idea
50
51The first step should probably be patched stock BIOS, not a full open-source BIOS.
52
53The original BIOS already knows how to initialize the motherboard chipset, SDRAM, Super I/O, PCI routing, AGP, and onboard devices. Replacing all of that from day one would make the project much harder.
54
55Suggested firmware path:
56
57Build Rev-A hardware prototype
58Use a patched stock Award/AMI BIOS to accept the new CPU
59Add or modify CPU ID / microcode-style handling where possible
60Adjust ROMSIP/startup tables if needed

…Show last 58 lines

61Once hardware is proven, investigate a coreboot + SeaBIOS port
62
63A future open-source firmware path could be:
64
65coreboot + SeaBIOS + slotket-specific hardware detection + CPLD startup patching
66
67However, coreboot cannot fix the problem if the CPU never reaches executable code, so the adapter hardware still needs to handle the early startup configuration.
68
69Rough cost estimate
70
71This would not be a cheap adapter at prototype stage.
72
73Estimated hobby prototype cost:
74
75First bare prototype PCBs: around $150–$700
76Parts per adapter: around $100–$400
77Assembly/rework/tools/test hardware: $300–$1,500+
78First serious prototype effort: roughly $800–$2,500 if tools are already available
79More realistic full bench/debug budget: $2,000–$5,000+
80
81If the design were proven and produced in a small batch, finished units might still cost around $250–$500 each, depending on sockets, CPLD programming, brackets, testing, and documentation.
82
83Main risks
84Incorrect Vcore/VID could kill CPUs or motherboards
85Wrong bus mode could cause electrical contention
86EV6 signal routing may be sensitive to trace length and impedance
87Slot A boards may not provide enough current for later Socket A CPUs
88BIOS may fail before useful debugging is possible
89Socket A cooler load needs mechanical support to avoid damaging the slot or CPU die
90Current project status
91
92At this stage this is a concept/proposal, not a working product.
93
94The next useful steps would be:
95
96Choose one target motherboard family
97Build a full Slot A ↔ Socket A pin and signal map
98Identify all required power, clock, reset, VID, FID, and ROMSIP lines
99Design a minimal Rev-A test adapter
100Add CPLD/FPGA logic for startup configuration experiments
101Patch stock BIOS for basic CPU recognition
102Test with cheap/sacrificial Slot A boards and early Socket A CPUs
103Help wanted / discussion points
104
105I’m especially interested in input from people with experience in:
106
107AMD Slot A electrical design
108Socket A / Socket 462 pinout and signal behavior
109AMD 751/Irongate and VIA KX133 chipsets
110ROMSIP / SIP tables
111Award/AMI BIOS modification
112coreboot on old K7-class hardware
113High-speed PCB layout for EV6 bus signals
114CPLD/FPGA interposer logic
115Vintage AMD motherboard repair/testing
116
117The goal is to find out whether a practical AMD Slot A → Socket A active slotket can be built, and what the cleanest path would be for hardware and BIOS support.

ChatGPT can say all kinds of things, but just take a look at that pic - that doesn't remotely look like either a Slot A edge connector or a Socket A socket. Do you really think whatever generated that mess would make something technically functional?

vanderlinde wrote on 2026-05-30, 20:09:

I just figured, maybe there's someone out there who's able to design a fully working Slotket device for Slot A.

But you just said ChatGPT was capable of that, so why not make it do it?

I never expected slop reaching this niche 😀

There is zero information in the AI summary or image to provide anything useful.

While a sloket type device would be theoretically possible for Thunderbird socket A CPUs, nothing else would be realistic or make any sense.
The EV6 CPU bus is quite different from that used by any previous x86 CPU. It is actually based on an early DEC ALPHA CPU bus.

rasz_pl wrote on 2026-05-30, 22:36:

vanderlinde wrote on 2026-05-30, 20:09:

I just figured, maybe there's someone out there who's able to design a fully working Slotket device for Slot A.

But you just said ChatGPT was capable of that, so why not make it do it?

I never expected slop reaching this niche 😀

It did research, the whole thing, from socket layout to CPU pin layout, the components required and everything with that. I mean if you run ChatGPT pro you be surprised how well it is crafting these things.

Why i'm not doing it is because the price cost and expertise required to even perform the above,

1JLCPCB advertises 6-layer PCB prototypes starting very cheaply, but that headline price is for simple cases, not a large controlled-impedance card with gold fingers, beveling, and unusual mechanical requirements. PCB vendors treat gold fingers as a specific edge-connector manufacturing option, and guides emphasize plating, bevel, and design-rule constraints for reliable insertion.
2
3So I would not expect the real prototype board to be a $10 item.
4
5First working prototype budget
6
7For me, the honest first-build budget would be:
8
9Item	Estimate
105 prototype PCBs	$500–$1,500
11Components for 5 units	$500–$1,500
12Socket sourcing / donor boards	$100–$500
13Assembly/rework supplies	$200–$800
14Programming cable / adapters	$20–$250
15Sacrificial Slot A board + CPUs + coolers	$150–$600
16Misc. mechanical brackets	$150–$600
17Practical first prototype spend	$1,620–$5,750
18
19That assumes you already have lab gear. If you need to buy a decent oscilloscope, logic analyzer, bench supply, hot-air station, microscope, and POST/debug tools, add $1,000–$5,000+.

So don't think you will be generating a first design that runs flawless. You'll likely blow up some components along the way.

akimmet wrote on 2026-05-30, 23:32:

There is zero information in the AI summary or image to provide anything useful.

While a sloket type device would be theoretically possible for Thunderbird socket A CPUs, nothing else would be realistic or make any sense.
The EV6 CPU bus is quite different from that used by any previous x86 CPU. It is actually based on an early DEC ALPHA CPU bus.

Yes and this idea randomly popped into my head; why is there no slotket for a Slot A system? So i figured it is technically possible; but it requires a certain skill to it. It would be amazing to use a Socket A thunderbird chip on a older Slot A board 🤣. Pretty sure you will be making headlines with that compared to unlocking a K6-III with additional 128KB cache after 20 years.

Technically, a dual slot A board would be possible too, but that requires way much experise then the above.

There's no practical use for such adapter. That's why they were never made back when Slot A was relevant.

dionb wrote on 2026-05-30, 21:37:

ChatGPT can say all kinds of things, but just take a look at that pic - that doesn't remotely look like either a Slot A edge connector or a Socket A socket. Do you really think whatever generated that mess would make something technically functional?

This.

ChatGPT would not be much help with something like this once you figure in having to diagnose and correct all the errors it would make.

I recently asked an LLM (can't remember which) to help me figure out key scan codes for a computer that was giving me input issues. It listed them in binary so I could compare keys that worked vs ones that didn't. The list looked something like this (where _ is a key):

_ = 110010
_ = 101101
_ = 111001
_ = 011005

... yes. Five.

I had to say "Why does that one have a five in it?" . It then apologized for the incredibly obvious mistake and changed the answer. All I asked it to do was to list key scan codes for roughly 10 keys. Not exactly difficult or complex.

LLMs are helpful for looking at large amounts of data and trying to make sense of it, but they are not good at properly completing complex projects. Especially niche ones like this.

The reason Slot 1 to Socket 370 slotkets existed was the provide support for common (simpler, cheaper, easier to produce) Celeron CPUs in common motherboards that were produced for 4-5 years. Enthusiasts also embraced them because they offered a lot of flexibility. Decades later, retro enthusiasts liked them because the chipsets used on Slot 1 systems tend to be great and offer at least a couple of ISA slots. Excellent slot 1 boards with ISA slots are far more common than even decent Socket 370 boards with ISA slots.

Slot A and Socket A are basically the opposite. Slot A boards are quite uncommon because they were only used for a short time (maybe a year?) . Also, they really offer nothing over early Socket A boards... other than it being more common to have one ISA slot on Slot A. Even with that though, as others have said, just getting a Socket A board with an ISA slot (or a blank spot to solder one) makes a lot more sense than trying to use a Socket A chip in a Slot A board.

Also, Socket A required more cooling and the heatsinks had significantly higher clamping force on the CPU die (hence the pads in the corners for protection) compared to Socket 370 chips. Both of these would be problematic for slotket use. Not impossible... just... more downsides with few or no upsides.

It did recommend only the use of "early" athlons Socket A, but not the thunderbird's and beyond, as exactly as you said, more power, heat etc. I mean ive done some work for a while with the pro version and it's, getting better. In regards of comparison it used many sources such as spec and datasheets, and figured out all by itself that certain signals have to be, translated or emulated in order to make it work. I'm pretty sure that if you bash it long enough you'll get a working unit here or there, but the road to it is far from easy. I don't have all the equipment - not interested in spending money into it either. The only reason why Slot A existed is to stick with current industry standards if i was not mistaken (same components as Intel) and later on to a socket. Let me rephrase what it wrote, and you be the expert on the subject or matter.

1Yes — but not as a simple passive pin adapter. A realistic AMD Slot A → Socket A/462 “slotket” would need to be an active interposer with very careful power, signal-integrity, BIOS, and startup-configuration handling.
2
3Also, one correction: it was basically never a mainstream retail product, but it has been seriously attempted. Tweakers reported that early 2000 announcements never turned into a practical market product because it was too complex/economically unattractive, and in 2023–2026 a hobbyist project by auto660 reached prototype/revised-design territory.
4
5The feasible target
6
7This would only be for Socket A / Socket 462 K7 CPUs: Thunderbird, Duron, maybe Palomino/Thoroughbred/Barton with extra work. It would not adapt later socket CPUs such as Socket 754/939/AM2, because those moved to very different platform assumptions.
8
9Slot A and Socket A are at least in the same K7 family: Slot A uses a 242-pin EV6 interface, while Socket A is 462-pin and also EV6-based; Socket A later supports higher effective FSB speeds than most Slot A boards. AMD’s EV6 bus was 100 MHz double-pumped on early Athlons, giving 200 MHz effective transfers, so routing and timing matter a lot even though the raw clock sounds low today.
10
11Why a passive slotket probably fails
12
13The obvious work is mapping 462 socket pins to 242 Slot A edge fingers, handling VID/FID pins, and feeding the right voltages. The deeper problem is the startup driver mode.
14
15Classic Slot A Athlons used an open-drain-style bus environment with pullups on the CPU module. Socket A Thunderbird-era CPUs introduced/used push-pull signaling for higher speed and fewer external parts. The community project found that Socket A CPUs would acknowledge startup configuration but then would not start normal data-bus communication when initialized in the old open-drain mode. The revised 2026 approach is to intercept and modify the ROMSIP/SIP startup stream so the CPU is initialized for push-pull mode, using an FPGA/CPLD-class device rather than a passive trace adapter.
16
17That means the slotket should not just be “wires and jumpers.” It should be designed more like a small high-speed motherboard daughtercard.
18
19The design I’d use
20
21I’d design it as a 6- or 8-layer controlled-impedance PCB with these blocks:
22
231. Slot A edge-card interface
24Gold-finger 242-contact Slot A edge connector geometry, with mechanical retention matching Slot A cartridge guides where possible. The adapter must not rely on the motherboard slot alone to hold a heatsinked Socket A CPU.
25
262. Socket A ZIF footprint
27A real Socket 462 ZIF socket, ideally oriented so the cooler load can be braced to the chassis or to a dedicated bracket. Socket A dies are exposed and fragile, so cooler mounting is a major part of the design, not an afterthought.
28
293. EV6 signal interposer
30Direct point-to-point routing for clocks, command/address/data/control, reset, power-good, and initialization lines. This needs short, length-controlled traces, uninterrupted reference planes, dense decoupling, and minimal stubs. I would include configurable series damping footprints and optional pullup/pulldown networks for board-family experiments.
31
324. ROMSIP / startup patch logic
33This is the make-or-break part. I would add a small nonvolatile FPGA/CPLD, for example a Lattice MachXO2-class device, clocked from the CPU/platform clock domain. Its job would be to pass the startup configuration stream through with minimal delay, but flip the driver-mode bit needed to initialize the Socket A CPU in push-pull mode. That mirrors the direction auto660 described in 2026 after the open-drain approach stalled.
34
355. Voltage and VID handling
36Slot A boards were generally in the right voltage era, but the adapter must protect against bad VID interpretation. CPUShack lists Slot A around 1.3–2.05 V VID VRM and Socket A around 1.1–2.05 V, so a lot of CPUs overlap, but late Socket A parts can draw much more current than early Slot A boards expected. I’d include DIP/jumper VID override, Vcore sense points, and ideally an optional onboard buck regulator mode for boards with weak or incompatible VRMs.
37
386. FID/BP_FID/multiplier control
39Many older Slot A boards will not know what to do with later Socket A multiplier encodings. The adapter should expose multiplier controls via DIP switches or solder jumpers. For Athlon XP/Barton, this becomes especially important because a 133/166/200 MHz FSB CPU may need to run at a 100 MHz physical FSB on a Slot A board.
40
417. BIOS compatibility plan
42The hardware alone is not enough. Early Slot A BIOSes may need patching for CPU identification, cache initialization, mobile/FID behavior, or just to avoid halting on an unknown K7. The 2023 project also identified BIOS/ROMSIP support as a central hurdle.
43
448. Thermal and protection layer
45I’d add thermistor/diode readout pads, Vcore test pads, POST/debug header, current sense option, and a hard “safe mode” jumper that forces conservative VID/FID defaults.
46
47Compatibility expectations
48
49The most realistic first success target would be:
50
51Board: AMD 751 “Irongate” Slot A board or selected VIA KX133 board with known Thunderbird tolerance.
52CPU: Socket A Thunderbird or early Duron at 100 MHz FSB / 200 MT/s effective bus.
53Goal: POST reliably, enter BIOS, boot DOS/Linux, then stress-test memory and cache.
54
55Second-stage targets would be Palomino/Thoroughbred/Barton, but I would treat those as experimental. They may physically work, but BIOS support, multiplier mapping, VRM current, and FSB assumptions become progressively nastier.
56
57The biggest risks
58
59The design could damage hardware if the CPU and chipset ever fight each other electrically — for example if push-pull drivers meet the wrong pullup/open-drain environment. It could also kill a CPU through wrong Vcore, crack a Socket A die through poor cooler mechanics, or destabilize the EV6 bus through trace-length/reflection issues.
60

61So the first prototype should be tested with a sacrificial Slot A board, sacrificial Socket A CPU, current-limited bench supply where possible, POST card, high-bandwidth oscilloscope, and thermal monitoring.
62
63Bottom line
64
65Yes, I’m able to design the architecture for one, and the viable version would be:
66
67Slot A edge card + Socket A ZIF + controlled-impedance EV6 routing + VID/FID controls + ROMSIP patch FPGA + strong mechanical cooler support.
68
69The critical insight is that the adapter is less like an Intel Slot 1 slocket and more like a startup-protocol and signal-mode translator. Without the ROMSIP/push-pull handling, it is very likely to initialize partially and then fail before real bus communication starts.

https://www.hardwareluxx.de/community/threads … 8#post-29852384

Just noticed, through the source of ChatGPT there was someone else who made a prototype of a Slot A > Socket 462 Converter.

https://www.hardwareluxx.de/community/s ... 660&o=date

And his last update seems to be around Jan 2026 this year, stumbling on the same issues as ChatGPT has predicted (signal, timings etc)

Extraordinary claims require extraordinary evidence. So far we have a summary that is entirely devoid of any technical information, and a laughably wrong concept rendering.

The socket appears too small.
An entirely useless support bracket.
No way to mount a heatsink.
Only a single row card edge.

vanderlinde wrote on Yesterday, 00:49:

https://www.hardwareluxx.de/community/threads … 8#post-29852384

Just noticed, through the source of ChatGPT there was someone else who made a prototype of a Slot A > Socket 462 Converter.

Wait wait wait.

You said this...
"Let me rephrase what it wrote, and you be the expert on the subject or matter."

Followed by a huge wall of text that others are expected to read.

But what you meant was, "let me paste what ChatGPT wrote and I will read it afterward" ?

ChatGPT is, at best adding on to what auto660 has already accomplished after years of work... at worst it is giving you a sloppy or incorrect approximation of what he did while only giving a vague reference to the fact that he has used a significant amount of his own time and energy to figure all of this out.

If someone wants to help to design and test a Slot A to Socket A Slotket, contact the person who has been working on one for over three years: auto660 on hardwareluxx.de

I'm sorry, but I believe this thread is exactly why rules like this exist on this forum:

Snover wrote on 2026-05-02, 18:19:

#2: You won’t post topics or replies that are authored by AI.

VOGONS is a community for sentient beings, not LLMs. People don’t like talking to other people’s chatbots, and anyone can go ask Gemini themselves if they want a confidently wrong hallucination. AI-authored posts are uninteresting and they damage trust and connection within the community. They are now prohibited.

Note that this is not a blanket prohibition on all uses and discussions of AI. It does not apply to posts that are authored by non-English speakers and passed through an AI translator, so long as the translator isn’t adding its own ideas that weren’t in the input. It also does not prohibit posts about AI, or posts about projects that use AI (but please disclose if a project uses AI).

⚠️ Terms of use update: Two new forum rules

vanderlinde wrote on Yesterday, 00:53:

https://www.hardwareluxx.de/community/s ... 660&o=date

And his last update seems to be around Jan 2026 this year, stumbling on the same issues as ChatGPT has predicted (signal, timings etc)

ChatGPT very likely did not "predict" these things. There is a good chance that it is basing a good portion of this information on his posts and the posts of others who have already worked on this, combined with some loose references from technical documents or other sources.

You can ask various LLMs for information about very niche topics that people on forums like this have spent hours\days\months\years working on, and it will mash up all of their research into whatever can be efficiently compiled in a few milliseconds while citing some of the source posts. Very rarely does an LLM actually have any deep understanding of any of these kinds of topics without just reading human made posts like these in the first place.

The AI should have immediately pointed you to the source it found. Instead of vague posting issues to consider, that would be obvious to anyone aware of high speed digital logic signaling and layout.

Look, this is cool, and when LLMs mature in years to come, they absolutely will become capable of such exciting projects.

But right now, ChatGPT thinks it can do all kinds of things it's not capable of; it blunders just making a custom autoexec.bat for a DOS boot menu.

In the past it's told me it can modify vintage games and save files, it never works. As somebody said above, any technical plans that ChatGPT gives you for that design at this stage will absolutely be full of errors.

I feel like this entire thread is AI generated, joined today and immediately starts regurgitating low effort slop

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?