Some follow-up notes: As pointed out be several patrons, the connector on that video card is the DMS-59 en.wikipedia.org/wiki/DMS-59 and luckily the adapter (splitters) are readily available. Also, on the SmartDisk ][ I've been talking to Bradley about the non functional Disk ][ (physical drive) support and it's a known issue with now. He mentioned it's worse than I may have noticed -- reading from the physical disk can actually cause the disk to get corrupted! Luckily when I tested it I had my floppy write protected, which physically keeps it from being written to. So for now, if you make one, it's best to use a separate Disk ][ card for normal floppy drive operations.
I've updated the README now, it looks like the corruption issue is easily avoided. There might be another reason some devices don't work in floppy mode, but I'm sure we can get those resolved.
That NVIDIA card was very common with dell back in the early 2000’s. I don’t know how common as most systems I built used onboard graphics. But nevertheless it was common for the idea that it was accelerated and could support multiple displays. That card was far more common in the small form factor cases as in it was usually the only option to upgrade to. So the more rare thing would be the fact the card was using a full sized slot.
The Slot 1 and Slot A designs were used as a stop gap solution as both Intel & AMD realized that the L2 cache needed to be closer to the CPU or in the CPU die itself in order to keep the CPU fed with data to maximize performance. Having the L2 on the mainboard meant that it could not be clocked high enough to keep the CPU fed with data as CPU speeds increased. Intel had tried to integrate the L2 into the same package as the CPU with the Pentium Pro, but the yields were very low with the Pentium Pro as the entire processor package had to be discarded if there were any issues with the CPU itself or the L2 cache. Each part had a separate die but could not be tested until they were both placed in the physical CPU package. Once that was done if either part had an issue the entire CPU package had to be discarded. Placing the L2 onto the same card as the CPU (but not on the mainboard itself) allowed for the CPU die & L2 to be tested individually which greatly increased yields. Having the L2 cache physically located right next to the CPU die allowed it to be run at a much faster speed than if it was on the mainboard. However, once both Intel & AMD had figured out how to integrate the L2 into the processor die itself (Intel - Mendocino & Coppermine cores, AMD Thunderbird & Duron cores) they both went back to having a CPU in a socket again as the slot configuration was no longer needed. AMD had also integrated the L2 into the socket 7 based K6-II+ and K6-III CPUs but the chips that had that were released around the same time as the Athlon/Duron CPUs that had integrated L2 as there was some overlap in the late K6 chips and the earlier Athlon CPUs. Intel's Slot 1 used the AGTL+ FSB protocol. AMD's Slot A used the EV6 FSB protocol which was licensed from DEC and also used by the Alpha 21264 processors. Physically the slots are the same but reversed and use different FSB protocols.
AMD decided to save few cents by using same physical dimensions like Intel. So there was no need to pay for new design of plastic socket and PC case air flow
I like your assessment of retro. It’s totally fair to say that it’s in the eye of the beholder and then share your personal taste. I happen to agree with it.
Nice one! Thanks for featuring my emulator in your video Adrian... It sold out on eBay, but will try to re-stock it in the coming week. A few comments, although the emulator uses 128k SRAM, it can only do up to 64k EPROMs - so 2716 to 27512 (initially the design was using 2x 32 SRAMs chips, but those days it is cheaper and faster to solder, to use 1x 128k SRAM chip but only use 64k of it). The SPI EEPROM onboard is to "save" the image so it can be recalled using the red button or recalled automatically when the emulator is powered up (with the -auto option).
It's a great device I think! It's actually given me a little inspiration for a project I've been thinking about, I shall do some more thunking and possibly get in touch! Either way, thank you - it's awesome!
The Slot-1 form factor was actually quite reasonable. It saved space on die by having the bulky L2 cache on a dedicated board that wasn't user modifiable, and at the time lithography was still huge. It improved yield by separating cache and CPU, it improved profitability by binning CPU and L2 separately and then matching them. And of course Intel wanted their new cool CPU to look cool. It simplified potential upgrades. The Pentium chips need more cooling and by angling the CPU out at 90 degrees (with hefty brackets) cooling was made easier. And the edge connector was a great answer to all those bent pins on 486's, if anyone ever bother to upgrade their Slot-1 systems. There are a bunch of good and not so good reasons.
I came to say much the same thing. I agree completely. It was also a very good thing from the standpoints of signal integrity and ease of motherboard layout.
@@Rorschach1024, yup. Dell used DMS-59 on a lot of their OEM video cards in their business desktops, even after they switched to PCIe. Such PCs came with Y cables that provided for dual monitors: dual VGA, dual DVI or one of each. DMS-59 carried two distinct video channels no matter what monitor interfaces were used, so cards like these provided dual monitor functionality before dual integrated video became common on motherboards. I haven't seen DMS-59 on any computers other than Dell, though.
Yup, endless 'workstation' class desktop machines had these for dual-screen. We used to hate it when clients would send in the cable only to find we didn't send it back!!!
DMS-59 cables are what you need. They were really common with Dell systems and tons of companies who bought/leased those systems have boxes of those adapters laying about. I bet if you put an ask out you would get an avalanche of those adapter cables.
Right, I can vouch for having boxes of those dual monitor adaptors at work back then. He was so close. Was right there on the Wiki page under "See also".
The big advantage of AGP is that the AGP slot was connected to the northbridge, whereas PCI slots were connected to the southbridge. That reduced the latency between the CPU and the graphics card.
This is not universally true. Some earlier chipsets (for example, the i440BX) handles AGP and PCI at the same time. However, all devices on PCI share the same bus (thus bandwidth) while AGP allows one device only and it doesn't share bandwidth with PCI.
Depending on the chipsets, actually. Later ones moved PCI away from the Northbridge chipset, with slower peripheral moved to the Southbridge chipsets (nowadays, they're effectively combined into one in recent CPUs, with an optional Southbridge chipset connected to PCIe bus), to better prioritize the messages.
@Dr_Mario2007 Yeah, I remember the history of how they split the north and south bridge and eventually the south bridge was moved into the CPU. There was some commotion over that move because they didn't want to add heat to the CPU and everyone knew that the north bridge got hot at the time. I think in the end it was all "hot" air though because the miniaturisation on the CPU die was such a smaller process node it was more efficient and cooler in the end.
@@lexluthermiester Well on some modern CPU, the high speed lanes can work as TB / USB3 / SATA / PCIe and they also have video output, so one could argue part of the south bridge (which typically handles IO like drives and USB among other) got moved to the CPU too ...
@AdriansDigitalBasement][ 17:25 I did! The slot design had a number of advantages over Pin Grid designs, the primary being: No bent pins, ever! Another benefit was better cooling possibilities. There were a number of other benefits that became less relevant a few years later and completely irrelevant a few years later. At it's time, the slot designs(Slot1, Slot2 for the original Xeons & SlotA) were great ideas! EDIT: I mean think about it, we had and still have expansion card interfaces, RAM, ISA, PCI, AGP, PCIe and so on. Why not do the same with a CPU? It made logical sense, at that time.
The big thing was that slot 1 and slot A allowed you to run the cache at a higher bus speed than the motherboard bus. Initially the L2 cache speed was 50% of the CPU speed and later 100%. Motherboard bus speed during the initial Pentium 3 era was 100 or 133 Mhz versus the CPU speed of 450 Mhz or higher. The later socket 370 Pentium IIIs had L2 cache embedded in the chip die, running at the full core speed. Slot 1 was pretty much a stop gap to squeeze out extra performance and that cache made a huge difference.
13:00 Yup, on the Amiga, normally when you move a window it just shows an outline. However, there were public domain & shareware utilities like MCP, MultiCX, PowerWindows, and Opaque that allowed you to see a window's contents as the windows were being moved on the Workbench screen, with varying degrees of slowness depending on your hardware specs. This was one thing I loved about the Amiga PD/Shareware scene, since you could usually find a utility to do many different things that weren't built into the OS, but were features of the other popular OS's at the time.
About the 486 MB. First, visual inspection for damaged traces. Second, that PLL chip should be outputting "doubled" frequency, i.e. 66Mhz (at oscilloscope) for your DX33, not 33Mhz / 25MHz. Third, you should definitely check for correct supply voltages in the cpu socket (without the cpu in it).
Not sure if anyone else has mentioned it yet, so I'll go ahead. Slot A is flipped 180 degrees from Slot 1. So you were right, they're not electrically compatible, and it *can* be confusing, but since the key on the slot only allows the heatsink to go on in one direction, generally speaking it gives you a good idea if it's a Slot 1 or Slot A motherboard.
Just to yes and this, the reason Intel and AMD went with Slot 1/A was that integrating L2 cache on die (or on package) wasn't economical yet, so both companies put it on a daughter board with high speed interconnects to the CPU. Once foundry technology sufficiently improved, on die L2 cache become economical and the P3 Coppermine / Athlon Thunderbird were the result.
Even though the PCs of the nineties were fairly similar, one can really get into the variety of graphics cards of this era. Especially the dawn of 3D. Very retro.
One thing you have to understand is that AMD at the time was so small compared to Intel. I mean they're small today compared to Intel but back then they were even smaller. Up until this point they had been socket compatible with Intel. The slot generation really signaled the end to single socket, multiple vendor chips. All AMD could do was to tell the motherboard manufacturers "look man you got all the same parts, you just got to arrange them slightly different and you can use our chips." Luckily the athlons were so good that people wanted them because otherwise AMD would have failed to get any motherboard manufacturer to support them or in the future any of their custom sockets. It wasn't until they were socket adapters that you saw Cyrix or other competitors chips in these machines.
AMD made cheaper Intel clones until K5, that's when they started to do their own thing and improved over Intel. The run K6 - Athlon 64 was legendary for AMD.
That Nvidea AGP card uses a DMS-59 (Dual Monitor Solution w/ 59 pins) port for connecting digital and analog displays to. The DMS-59 port provides two DVI or VGA outputs in a single connector. You'll need adaptor dongles to use it with a standard DVI, DVI to HDMI, or VGA cable. There are Y-style breakout cable adaptors for DMS-59 ports that allow dual monitors to be connected to such cards. Note: If you look at 15:12 in the video, you will notice a link to Wikipedia's article on the DMS-59 connector.
what is nice now for slot 1 is that if you have a board like the 440bx, you can really easily swap between pentium 2s and pentium 3s all the way from 233MHz p2 to 1Ghz p3 (and 1.4GHz p3s with right board, a bit of hackery and a slotket adapter)
I just got home from work and got a zillion RU-vid notifications. As a looked down the list I realized Adrian's channels are part of a very small subgroup where I don't even read what it says, I just click. 👍
About the card with the non-standard DVI port, babk around 2008-2011 or so the Dell and HP workstations we purchased came with video cards with those adapters. They would split out into 2 standard DVI ports and you could drive 2 monitors, and included the analog signals.
The early Athlon's were also on a card for the same reason as Pentium 2 the cache was external to die. It was possible to resolder the dividers resistors to change the divider to make overclocking easier. I resoldered my A600 cache divider from 1/2 to 1/3 so CPU could be overclocked to 850Mhz. I still have the CPU in my Win98 games PC still has the divider at 1/3 running stock 600Mhz clock speed. The advantage was the cache could have its own interface to the CPU independant of the front side bus clock. The reason they used the same slot an packaging so motherboard and cooling manufactures could use the same mechanical connection to save retooling.
That Slot 1 Pentium II was made in the factory I used to work in back in Ireland in the late 90's. I moved to another part of the company just before they began making those CPUs there, but I did get to visit the floor a few times.
I am a few months away from becoming 24, and I generally respect and share your opinion on what we choose to call retro-computers. I honestly hold no nostalgic thoughts for the fairly standartized beige boxes of the 90s, nor for the bland black boxes of 2005 to 2013-ish. But those late 70s and 80s machines, that are experimental and sometimes one-off computers with mediocre capabilities (compared to today, ofcourse), which came with adorable form factor cases are totally different, and thus, of interest to me. And I find it especially interesting that you can diagnose and talk about circuitry in detail,- which I, as someone who doesnt get how these things work exactly, find it very useful to know about. (Though I havent found any real-life appliance of this wisdom yet.)
On the Promise card, the Non-RAID and RAID cards were almost identical. Just two wiring changes and then reflashing the BIOS. I have a few hacked cards around. Worked on both ATA66 and ATA133 versions of the card. I think I also have the original RAID card too.
Slot 1 and Slot A used the same connector, just rotated the other way and the vertical posts were keyed differently. The busses operated very differently. Intel's bus was basically the same from Socket 8 (Pentium Pro) to Slot 1 to Socket 370, and there were adapters for both sockets to Slot 1, with the only real differences being frequency and voltage. AMD did move the Athlon L2 cache off-die for the same reasons as Intel, but licensed Alpha EV6 bus protocol from DEC. That got the Athlon over the 1GHz line before Intel.
I have recently been thinking about what makes a computer "retro" for me and I have realized it is not purely chronological. My dream computer has a one-piece integrated design, brings the user to a text console immediately upon power-on, and is incompatible with every other model of computer software. That is the world of my childhood. Paradise.
It's interesting to watch your journey with the M601 board. I have the same board (perhaps a different version number), and I went through the same process of discovery with the different websites to determine the clock settings for the PLL52C05. It's a shame you weren't able to discover why it's not posting.
Adrian, I had a dead 486 motherboard with similar issue. I checked the CLK pin on CPU socket and there was no clock. The clock was coming from a 74 series buffer which was connected to chipset. I took the buffer chip out and tested it on TL866 and it was fine. Next step I checked the chipset, it was receiving the clock from crystal but not outputting to the buffer chip. Replacing the chipset fixed my issue. I think checking CPU Socket pins will give you a better idea what the culprit could be.
Speaking of getting mixed up by L1, L2, and L3 cache... You might try out one of the rare systems with an L4 cache. The easy one would be a Broadwell based Core-i7 5775C, or i5 5675C. They give quite a bump up for anybody still running an old Haswell Socket 1150 board... At least in apps that can take advantage of the extra cache. It's also possible to run an intel Optane accelerator as cache on these... Technically designating it as an L5.
Anything is retro that is able to make you tell a passionate story about something benign, that makes your eyes sparkle, that makes you pause and dream about something, that was unobtainable for you at the time, or is from a time long gone by before you started.
Slot 1 was developed to solve a production problem. To putthe cache in the same package as the CPU required them to be bonded together before they could be tested, so any flaws in either resulted in having to discard the whole assembly, which made yields low and production costs high. Slot 1 allowed the CPU and external cache to be placed on q small daughterboard, allowing them to be tested before installation, making yields higher and production costs lower. And yes, Slot A and Slot 1 used the same connector with a different pinout, but they also were oriented differently on the motherboard, being backwards to each other.
Slot1 and SlotA both used the same SC242 edge as their physical interface. They were not electrically compatible. They were at least designed to have clearance (for the heatsink, etc.) on opposite sides relative to the key, so MOST motherboards had a physical obstruction to keep you from sticking incompatible parts together. Slot processors were mostly about making room for on-package but off-die caches to improve yields (smaller processor die + SRAMs instead of one large die with both). In recent years we're seeing an explosion various fancier multi-chip modules, chiplets, stacked dies, etc. which are kind of the same idea. Ed: I see that you recalled later in the video.
During the 90's I built about 2 PCs per year starting with a 386sx 16mhz. Then I slowed down in 2000s. I still like playing around with early to mid 2000s tech. I call them the XP class PCs. If it has a floppy controller and IDE controller then I consider them retro.
I would have to check, back at home... I think I've got a card like the "weird DVI" one, with two adapter cables, one 2xVGA and one 2xDVI... you might have more use for it than myself, hehe
The reason for the slot1 (and AMD's slotA) was the cache. By placing that on a daughtercard with the CPU it allowed faster access to the cache bypassing the system bus. Of course that became obsolete once they figured out how to place the L2 cache onto the CPU die, but they continued to make Slot1/SlotA CPUs in that form factor for a while for backwards compatibility reasons. And that's also the reason why Apple used CPU daughtercards in the PPC era, many of those also had external cache on the daughterboard.
There were some slot 1 gems. Like the Celeron 300A with on die cache that would almost always overclock to 450MHz. They sold for $180 and were just as fast as the $500 Pentium II 450
As I recall, the Pentium Pro had yield issues with the L2 cache. The "Slot 1" Pentium 2 & Katmai Pentium 3 was Intel's solution to have 512K L2 on a separate manufacturing process running at half CPU speed versus the much slower motherboard bus speed cache used by older CPUs. Later Coppermine (and Toiletin) Pentium 3 had 256K on-die L2 running at full CPU speed. After Intel solved the problem of acceptable yields of on-die L2, they went to Socket 370. Your Celeron 300A had 128K on-die L2. The original non-A Celeron 300 had no L2 cache at all. There were of course other speeds for both core types of Celeron. As for AMD's reason to follow with a slot formfactor, I suspect cache yields was their reason as well, because Slot A eventually got replaced with Socket A. Also, there was a way to tap the Slot A PCB with a device called a "Golden Finger" for overclocking.
I was building pcs in the slot era! If I recall Intel’s on die cache yields were awful so they came up with Slot 1 to carry the cache. Leaving it up to the board manufacturers would end up with no cache in budget machines and the P2 needed the cache to have decent performance. AMD were just playing follow intel. The slots the same connector but rotated 180° on the board because they’re not electrically compatible, don’t want people frying their CPU because you copied intel!
Amd did not have a choice much like Intel, cache memory at the time wasn't able to reach the speeds cpus were operating at. so the best solution they had until tech caught up was the processor on a card. They could have gone with L2 on the motherboard again but that would have been even slower due to the distance from the cpu. memories a bit hazy but I believe amd had the performance lead during this era over the pentium 3.
@@whelmy if I remember that was the era pentium3 went super-scalar and had a long pipeline, AMD scaled up vertically by having less latency, which wasn't good because they also had less L1-cache on die, so they ended up also going to the insane slot thing
Rumor has it that AMD used the same connector as Intel for the K7 (rotated 180 degrees) to save mainboard manufactures from issues with different manufacturing processes. Intel did not license Slot 1 so AMD created Slot A... but yeah, it was a stop-gap/bodge. 😅
The phrase 'gan canny marra' Lawrence used in his letter is local dialect used in the North East in the UK. Like Newcastle / Sunderland or South Shields area. Funny to hear that spoken on your channel here!. I'm a Sunderland lad, and it basically means 'go well sir' a kind of 'take care'.. where 'marra' is a man.
That 486 motherboard looks very similar to my second motherboard that I had long time ago. It used to have the same problem. Sometimes, bending a little bit the board can fix it for a while. Probably is not a bad chip. It's more probably a broken circuit trace near BIOS, near processor or near power.
For the defective motherboard i experienced several times already that pins of the chipset can get lose somehow over the time or if the board was bend once too strong. Checking with high magnification if you have a lose pin. I was able to fix most of my declared dead boards after I got this hint.
Slot-1/Slot-A processors were not a mistake, and nobody thought the slot was better than a more standard PGA/LGA chip. But they didn't have much of a choice at the time. At that time, we didn't have these really tiny nm-scale die processes. It was no longer practical to put cache memory on the motherboard, but cache chips were too big to put on the CPU's die. So they made a small circuit board that could use separate (off-die) cache chips tightly coupled with the CPU core. That board connected to the motherboard with the slot. It may be worth noting the huge size of contemporary socket-based CPUs like the Pentium Pro and the Pentium II Overdrive designed to fit into a PPro socket. Those were huge ceramic packages and with the process technology of the time, advances were going to make the package too large for use in PCs. Hence the slots. But when process technologies improved so you could put the cache on the die and put the results into a normal size package (Pentium 4, Athlon 64, etc.), the companies immediately abandoned slots, because they were no longer necessary.
The slot 1 CPU design was a necessary evil for reasons of cache. They went back to a single chip design after they made a jump forward with later pentium chips.
That is a DMS59 connector. It's commonly available once you know the right search terms, ha. You are correct in the fact that it is used to send dual DVI connections out from a single port.
That graphics interface is a DMS-59 which as you guessed, provides two signals from one port with a Y-cable. You can output either two DVI signals or two VGA signals, or one of each with the appropriate Y-adapter cable. This standard is NOT proprietary and was only deprecated in 2020. The cables should be readily available with very little premium pricing, if any.
The EPROM sized SRAMs are also used in the Timex Sinclair. I am actually planning to do a RAM upgrade on a couple Timex Sinclair 1000 systems that I purchased at an estate sale last summer. I also need to fix the keyboard flex cable. I've only opened one, but assume they are both damaged.
That Athlon may have looked a lot like a P2/P3, but it used a vastly different bus. IIRC they basically used a DEC Alpha style bus & double-pumped it to boot. And it was an absolute floating point monster in its day. I don't think AMD _wanted_ to create a slotted CPU like Intel. It's just that, like Intel, AMD needed their CPU performance to scale with clock speed as they raced to a GHz, which required something significantly faster than motherboard cache. All at a time when it was still a little too pricey to put a meaningful amount of the stuff on-die. Even PowerPC chips in Macs were on daughter cards along with the L2 at around that time.
Intel were basically forced into putting the Pentium II on a card. With the Pentium Pro, the L2 cache and the CPU dies were connected on chip using glue logic and they had no way of testing the two dies separately before committing them to the chip, and it ended up with disastrously low yields, hence why the Pentium Pro was so expensive. So with the PII, they put the cache in separate chips that went onto the card in Slot 1 so they didn't have to just throw out good CPU dies with bad cache or vice versa. Anyway, by the time the second iteration of Pentium III came out, they'd shrunk the process down enough that they could fit the L2 cache in the same die as the CPU. Nostalgia Nerd did a video on it if you are interested in hearing more.
That Nvidia connector is a dual monitor cable and was used often in Dell Optiplex (P4) machines from the mid-00s. I think I might still have one rolling around.
Slot A isn't a copy of Slot 1 per-se, the pinout is completely different, they flipped the socket around to take advantage of economies of scale (meant motherboard manufacturers only needed to buy one type of connector)
The AMD Slot A was also flipped compared to the intel. The heatsink would usually face the opposite direction if you tried to put it in a slot 1 motherboard and not fit as it would interfere with the IO ports.
Retro SOFTWARE on things like a tricked-out Slot 1 machine would be great. Like, make a period-correct Windows NT 4 computer and run Exchange on it :) See to what extent you could get IIS to talk to the modern internet!
Hard Hat Mack was one of my favorites on my Apple //e (my first computer). I also loved Gemstone Warrior, and Conan the Barbarian. Plus Loderunner and Choplifter, of course. Oh and Ultima II (never played any past II).
Intel used the slot design in the days before they were able to package memory and cpu chips on a single substrate. They also were able to include a heat sink onto the package. The 8251 was used in embedded processors. This was Intel's serial chip that would do both synchronous and asynchronous protocols. It had a few bugs that had to be programed around, but it did work. It was limited in it's max communication speed. Motorola had the 6850, MOS had the 6550. However the NSC 8250 family became the standard for the PC. AMD introduced a similar part for their second source of the Zilog Z8000 family.
There was a LOT of PC junk in the 90s. Still, that PIII I remember when my buddy got one, a 500mhz but man I was on a 486dx4100 which was really not that great either... but that was really the time of the big change.. win98 came out around that time.. then it was like a jet pack got put on technology and it went crazy. AMD a few years later came out with their 64 Athlon and things never went backwards from there.. That pIII is probably right at the change from vintage to 'junk' to modern times.. it was necessary though.. we needed that technology and those things were NOT cheap either!
E1 is the last "test" in the Power On Self Test (POST) that completed successfully. It's pretty late on the way to FF, which could be a clue. (FF = all tests completed successfully) Search for a POST code list according to BIOS maker (especially) and version. The first test in the POST *after* E1 is what failed. It could be E2, but not necessarily, as not all POST tests/codes are used with a particular board/BIOS. Then there's the subtle problem of people assuming the code displayed is the actual problem (instead of it actually displaying a success). So, a user made list could be off by one line in the test order. I presume you know all this, but someone who buys a POST card with no instructions might need this info.
I think the slot 1 approach stopped everyone from wrecking the pins on their CPU. It's been a while, but I suspect those were before the pinless Intel processors. Before that, even with zero insertion force sockets, many people would always damage their CPU pins. I never had problems with the pins, but I was careful :P
Possibly already said, but 9:09 DMS59 connector. Usually required a breakout connector for two DVI connectors, two VGA connectors or a combo. Usually used on low profile cards
I love that EPROM emulator! It would be cool if there were a version that supported multiple EPROMS in a board like in the case of arcade boards with an array of 27C64s
if I recall correct. I once did brake of the flat part of a spare dvi cabel. to use in a simular gfx. and I could use it normal with a single screen :)
Slot 1 brought in an era of the first consumer dual-cpu boards. Before that you had to buy expensive server boards. I had a Tyan Tiger 100 with Dual PII 400's and dual 15" CRTs. OH BABY. Slot 1 also had the infamous Celeron C300A. You couldn't run them in a dual config though.
So apple's slot, one design during the Mac OS8 era was pretty bad a** you could take 1 of their towers stick 2 of their slot, 1 processors into the machine. Max out the ram at 128 gigs and the machine went from slow to speed demon from hell
Slot 1 didn't begin with the Celeron 300A, but with the Celeron 266 (that I had at the time, to kick some K6-233 asses ;)). C266 had no cache, so less performing that the C300A (besides frequency), however the CPU cache was a real blocker when overclocking. The C266 had some HUGE overclocking capabilities, more that C300A in terms of %age, you could set it up at more than 400Mhz easily. If I remember well, you had to mask (or tape) the B21 pin of the Slot 1 to overclock ;) You missed something big, as you couldn't overclock your Pentium 233MMX neither your Athlon. Overclocking a Celeron recalls the 8088 overclockings ;)
The infamous PC Chips 486 motherboards with the fake cache chips and hacked bios. I bet the different bytes were the ones hacked into Amibios from the factory.
So at the time the slot processors came out the idea if I remember correctly was that the processor manufacturer (not the main board manufacturer) had control over the distance the cache was from the processor. That way they could run it faster. When the P3 and newer AMD chips started putting the cache on die there was no need for the slot cards any more. The reality was I think it introduced more speed issues then it solved so....
@Adrian, the connectior on the gforce fx5200 AGP card is called DMS59 and you have analog and digital breakout cables.. you had it mentioned in your wikipedia articles and just didnt catch it :D DMS59 - "a single DVI sized connector providing two single link DVI or VGA channels" Anyways, i have few of such cards, and yes, you need a breakout cable, but it has 2xVGA or 2xDVI ports on other side..
While the P3 holds a special place in my heart, I can agree with your view on where it sits for retro, for me retro ends at the end of the 90s, so tail end of the p3 market, before NT based operating systems were standard, and 3DFX graphics cards were king.
The slot processor was introduced by Intel primarily to protect IP and keep competitors from taking a public specifications for electrical connectivity of pins, and making a plug in replacement. AMD seriously made bank with the Am386 through the socket 7 series, and so did Cyrix, with drop in upgrades. When Intel introduced Slot 1 their customers and 3rd party board manufacturers had to sign an NDA and license not to reveal electrical connectivity to anyone. The reason why AMD had a slot processor was to sell to customers that had purchased slot "pick and place" machines. Since consumers were still in the middle of the megaherz wars, CPU heat generate rapidly outstripped what the "slot" could dissapate, and so PC companies went back to regular horizontal sockets/PGAs, but never again have the particulars of connections been public or available in a non-licensed book or spec.
I have been developing an “EPROM” emulator, it’s in quotes because it does other stuff as the “personality” is managed by another board. It has high speed USB and uses CPLD’s to handle a lot of stuff like bus arbitration and also handle anything else that is required for the target, it also has wider bus width than a target may use, so the CPLD handles any address translation that needs to occur so that the target reads the correct address, the reason for this is so I can write to multiple RAM chips with as shorter time as possible, this dramatically speeds up uploads to the RAM from the PC, I see 8 megabyte per second transfers because of this architecture and my design uses high speed USB. I created it so I can run development code on real retro console hardware so it includes emulation of things like memory mappers for the targets I’m interested in, it also provides the ability to trap and send debug messages from the code running on the target. Running on real hardware is so useful for testing. ru-vid.comS-7JU_MLERM?si=0vnMecBBgzLIXxF-
![Adrian's Digital Basement ][](http://yt3.ggpht.com/ytc/AIdro_k7CrQRe71qN6c0Ua7A1P_I8KkxjqLa70u0zYeCvdvnKw=s48-c-k-c0x00ffffff-no-rj)