Man, I still remember bringing my Pentium II (I think it was 233 Mhz) cpu to some computer shop, exchanging it for a Celeron 300A while getting some cash back and then overclocking that from 300 to 450Mhz, ending up with a much faster PC *and* *getting* *money* *for* *it* . Those were crazy times!
Hahaha. Yeah I had a 300a @ 464. It could have done 504, I just didn't know SHIT about things back then and didn't even use thermal paste. Just the pad that came with the cooler. Made a Celery Sandwich with it, without thermal pads. Lol. I also had a PII 300 SL2W8 that was a 450 in disguise. Did 450mhz easily. One day my Mom came home (It was her PC) and I told her the CPU was at 450mhz now. Lol, great times.
Thank you Adrian! I'm very glad to know, that despite of the amazing amount of (great) work you are doing, you still have some spare time to watch others videos ;)
@@necro_ware Is this overclocking technique possible on a 486 as well? Once you've maxed out the FSB speed and CPU multiplier there's nothing more you can do on the software side of things, but swapping out the crystal oscillator/clock generator sounds like it could work. I'm asking because I recently found a Socket 3 board, the Shuttle HOT-433, that can do 83MHz on the FSB, and I wanted to know if it would be feasible to push it even further than that.
@@FordGTmaniac Some early 486 also used such oscillators, but later version like HOT-433 used clock generator IC like MX8315 (see between PCI slots on your board). This clock generator can generate up to 80MHz. I personally don't think, that this board or any 486 CPU can stand 83MHz FSB anyway. At least not for longer period of time.
@@necro_ware Oh, ok. That particular board uses the UMC 8886AF / 8881F chipset which was also used for Pentiums, or so I've read. That would explain why the FSB can be made to run so fast despite it being a 486 board. 83MHz isn't an officially listed setting even though it is fully capable of doing so, probably to prevent people from accidentally frying their CPU's.
I remember changing jumpers took a Pentium 166mhz to 233mhz for massive gains. Opening internet explorer 5 while playing music on Winamp with no stutter was amazing
I had a P3 500, 100mhz bus that OC'd to 733/133. Also had a Celeron 300/100 overclock to a 450/133 I think. Good old days. Then again my i7 7700k OCs from 4.2~4.5 to 5Ghz.
This reminds me of the old TurboPLL project, which replaced the clockgenerator on the motherboard. You could then increase/decrease clocks on the fly. This project improved overclocks in the early 2000s.
I have to say, this was my favorite time during computer history. I loved the 8088 days too... but the 386 SX time, I don't know, it felt like computers were so much better put together.
Just watched your video on overcloking the 386SX, and I'm seriously blown away by the quality of your content and explanations! Your work is fantastic, and I'm itching to try it on my own 386SX 40 MHz. Thanks a ton for the top-notch content and insights. Truly impressive!
Thank you! It took some time of training, but at least I can do it, if needed. However, I only do it, if I need to. I still prefer through hole soldering for tinkering whenever possible.
I have never seen a 386 above 40MHz that was factory boosted. Yes, you had to be at 40 to play DOOM FS, but you got like 15-20 FPS. My friend would stomp me in DOOM but I trounced him in Descent (similar performance, I was just better in 3D space.) My friend had a 486 DX and I had a 386/40 (it was a 25, the 40 was turbo.) You remember your first computer... This was not my first computer, that was a Compaq 251 Deskpro, an 8086 CPU.
I will not start to argue, because everybody has own expectations on, what is playable and what's not. It's in the end up to you, but I think Doom is not playable on an any 386DX, not to mention "smoothly". At least here you have some numbers and they are completely in-line with my own measurements: ruclips.net/video/qQEHHc1q06c/видео.html
TI 486dlc that fit in a 386 socket @ 40mhz with a 33mhz or 40mhz 80387 type *ulsi or iit were my goto*. Consistently got near 486dx-40. That combo probably only worked out well because of the board i was using. Good times.
@@Tris289 I have this combination on a very fast 386 mobo. 486DLC + IIT4C87 both @40MHz (not overclocked). It is from the performance between 486DX-25 and 33. More like 25 in the most benchmarks.
I did overclocking on a 286 by changing the crystal oscillator from 16MHz to a 20 or 25Mhz.proved by PC Doctor.very nice game.thank you for remind me about this :)
When I see people torturing old hardware, I do think that, if done properly, it can be a valuable lesson for all of us, and save other boards from such a fate. This was a fine masterclass in the limitations and capabilities of old hardware. Thanks you.
I actually also don't like to "torture" this grandpas of computer history, but curiosity won and many people seem to be interested as well. Now they don't need to to put their hardware in danger :D
Back in 1993, my first PC was a clone of IBM AT 12 Mhz. It did have a CPU in a ceramic housing and quite hefty heatsink attached to it. It was just too slow for everything i was trying to accomplish. So i've had this idea of overclocking. Went to a local parts store and bought every ceramic resonator between 12 and 22 mhz they had on offer. Removed the original one and soldered a socket in its place to aid for easy swap. Then, gradually inserted faster and faster crystals and running the PC for couple of hours to see if it malfunctioned. Found out that i could safely run it at 19 Mhz which was a huge improvement. But that was not the end. The machine did not have math coprocessor and i was torturing it with CAD tasks so i found out that maybe it would be good idea to populate that empty Coprocessor socket on the board. Scavenged a 80287 MPU and when casually going through its datasheet, noticed that it actually runs totally asynchronous to the main CPU. Rigged up a sort of adapter that went in between the board and MPU that contained its own resonator (it was a proper, powered generator supplied with 5V while board used traditional 2-pin crystal) that ran@20 Mhz. The result was a genuinely fast computer that ran blazingly fast 3D CAD calculations. Chipset didn't seem to be the limit but i do not recall its brand or IC manufacturer. Most likely, it was an older "hybrid" board that integrated some of the functions (like MMU, interrupts...) into a sort of multifunction chip, but it stil contained large number of individual Intel circuits, like UARTs, keyboard etc.
Thank you for sharing your experience! If you tinkered with 286 back in the days, may be you are also interested in my various 286 restoration videos ;)
My first overclock was made on 80286 CPU with use of the TURBO button :) With later CPUs I used many different methods for overclock (switching jumpers, replacing oscilators, cuting traces or isolating CPU pads/pins). I miss the good old times when oveclocking gave fun and was worth it (like overclocking Celeron 266MHz/66MHz FSB to 400MHz/100MHz FSB or more).
@@gagarin777 True. The turbo function was underclocking, but sounded fancier as "turbo". Many people thought they were saving the system's stability and reliability by having it run off the turbo mode as much as possible.
Absolutely amazing journey you took me on here. Thank you, for your many different approaches on overclocking this CPU/Mainboard combo, highly appreciated
I ended up damaging my K6-200 by going from 66mhz to 83mhz on the bus. 75 however worked fine, but going to 83 made it so the CPU was no longer stable even at the stock settings anymore.
Hi! What a fantastic video! I must say that you have the nuts in place :D Dessoldering the CPU from a good working Mobo was very rad and risky. I love to see the charts and the performance increase. My first PC was an 386SX25, one week later, in the same store, I saw an 386DX33 at same price. :(
Thank you Jorge! Yes, desoldering the CPU was kind of crazy idea, but I couldn't stand the curiosity :) Unfortunately I couldn't reach 33MHz with this mainboard, but at least it was an interesting experiment. And since I could restore everything to the original state, I guess it went quite good in the end. My father bought me my first PC, it meant to be a 386SX-25 either, at least, that was, what I thought. Years later, as I understood a little bit more about it, I realized, that my father was scammed and the CPU was actually 20MHz, but sold with 50MHz crystal oscillator. So my first PC was overclocked from the beginning by an ugly scammer :) I still have that old overclocked hardware under a glass cover in my working room. It is still completely working!
@@necro_ware wow, who sold him overclocked CPU? But in that time, it was matter of few hundreds $, so it was tempting to do by sellers. Was it regular store, or he bought used computer from someone?
@@warrax111 It was a "regular" store, but I probably have to add, that it was in Moscow :) It were 90's, a dark age, no rules, no law,, no government. Scammers on every corner, so no wonder, that even a usual store scammed on their customers. As you can imagine, the price difference was quite a lot between real 20 and 25MHz machines. Anyway, I find it quite funny today. It's a part of my history now :D
@@necro_ware Yep, I am also from Soviet Union, I know what you are speaking about. :) Nice story though, it was just like that, when people started to grow up, and know something about computers, it was too late for doing anything, the sellers used it in early 90's.
What a walk down memory lane such nostalgia I have been trying to get a running 286/386/486 motherboard I have a 486dx2 Pentium overdrive CPU but finding a working or even dead motherboard is a very expensive task even if dead I would repair it but ebay sellers are not the most reliable and our local e waste disposal depot wont sell any of the good components I have seen there yet they charge an arm and a leg for people to dispose of them I do however have managed to salvage 2 8086 processors and a 8088 but motherboards for those in the wild are rare..... Thank you for such a nostalgic memory
I had a 486 back in the mid nineties, started life as a DX-33, but I managed to get an AMD DX4-120 CPU. Problem was my motherboard could only do 33Mhz bus, not 40, so I couldn't reach 120Mhz, "only" 100. I replaced the crystal on the board to get 40Mhz bus and it worked flawlessly. The PC still "thought" it was running at 33Mhz since I had changed the reference frequency (I think?), but benchmarks and performance clearly showed the improvement. The computer lived on for a couple of more years running Windows 95. Even battled with my friends' Pentium 60 and fared pretty well actually, which made him pretty pissed since he spend a lot of money on his new shiny Pentium :P I don't necessarily miss those days, things were so complicated, but I do miss the hardware and the "feel" of successfully getting the PC to run fast and stable. Which is why I build retro PCs, collect hardware and help arrange retro LAN parties and play lots of retro PC games these days. Just keeping those old feels alive :)
Hehe, you probably terribly overvolted your CPU back then. If you have had a mainboard, where you could replace a crystal, then it definitely didn't have a 3.3V support. All the mainboards with 3.3V CPU support never had a crystal, they already were all with jumpers and variable clock generator IC. So you probably powered your CPU with 5V instead. You were lucky, that it survived so long. And regarding the DX series, this was the first time, where such a thing like Front Side Bus (FSB) was born. DX, DX2, DX4 was nothing else, but a multiplier 1x, 2x and 3x. The mainboards usually don't care which one of the said CPU is inserted. The main difference was between DX and SX (no FPU). DX2 and DX4 were technically still the same as DX, the multiplier made only difference internally inside of the CPU and was completely transparent for the mainboard. I also would assume, that your mainboard could actually do 40MHz FSB as well, but probably this frequency at 5V was too much for the CPU. Back in the days you had problems with FSB of 40MHz only on mainboards with VLB support, but most of them had additional wait state jumper, in case FSB was set to over 33MHz. Furthermore, I suppose, your mainboard was ISA only, because as VLB was introduced, they didn't use plain crystals anymore.
I feel kinda young. Our first family computer was the cyrix 486dx2, which sadly I gave away. I've kept a complete 386sx low profile board for a wet day project. Really enjoyed your video.
I remember a 386sx 20 machine, doing a RAM upgrade on it and spotted the CPU was actually a 25 - and the oscillator was socketed - hmm, some free performance to be had? - couldn't find a cheap 50MHz, but picked up a 48 from a radio rally... Stuck it in, and a perfect 24MHz was achieved - and the ISA ran at a nice 1/3 for 8MHz, instead of an evil 6.66MHz or an overclocked 10
celeron coppermine cd0 stepping 566mhz is an overclocking monster 66mhz bus and high multiplier allow this chip to run at 1133mhz by removing bsel pins on the cpu (forcing bus selection to be recognised as 133 by default) and even overclock from there since via chipsets for coppermine/tualatin cores can't aply fsb/pcie dividers correctly this was also the only way to get stable overclocking for tualatin celerons over 133mhz bus i had 2 cd0 coppermines 566 and both of them run crazy stable on stock 1,5v and 133 bus, what more that barely ever got warm i love your experiments. they really shows how much potential washidden inside old computer parts , artifically segmented and downclocked to create "less expensive" options for the market
First exposure to the channel; I'd forgotten what a nerd I am at my core. Also forgot how much fun it was to play with these components as a kid! As well as being thankful we had an OEM 486SX/25 build that ran Doom quite well!
Our memory cheats on us :D I also always remember, that some game ran well on some of my old machines, until I try it once again. In full screen Doom ran on 486sx-25 at around 15-16 FPS, if I remember right. Let's say it was kind of playable :D
My first overclock in ninetities was 486SX/25 @40MHz just by swapping oscillator. It was already in socket. Later got 486DX2/66 and that went up to 90MHz. It was sick fast with VLB display card. Chipset on the mobo was from UMC by the way.
My 1st easy overclock was with classic Pentium 150 Mhz with 60 to 66 Mhz FSB jumper which resulted in a Pentium 166Mhz. MMX is limited to integer formats that recycled the FPU registers. Switch between X87 and MMX resulted in pipeline frush. SSE's support for FP32 and separate register storage fixes MMX's problems.
I was concerned about breadboard as well and tested the signals with oscilloscope before I started to do anything. The signals were quite clean and reliable. The temperature was also completely ok. If I remember right, I even show it in the video. Anyway the problem with the chipset was already confirmed by multiple people. It doesn't stand frequencies above 25 MHz.
Found this video in my feed probably because of being subscribed to Adrian Black aka Adrians Digital Basement. Liked the video. You’ve got a new subscriber. Cudos my friend.
Thank you and welcome on my channel. I really have no clue why, but this 386 overclocking video got a lot of attention in the last days, despite, that it is actually quite old. Well, that makes hope, that people will take a look on the other videos I made in the last time as well ;D
Having the idea of switching the CPU is one thing. Actually doing it is another. Crazy. Fun to watch, though! I overclocked my DX4/100 to 120MHz, which was a big financial saving in those days. A friend of mine bought a PC that was advertised as 120MHz, but was actually an OC 100MHz model and kept crashing. They very quickly replaced the CPU when we took it back!
This is how I got started with overclocking -- soldering in new oscillators! EDIT: wow - i would have bet money it was the motherboard or ram holding back that clock speed, not the cpu itself! I'm curious what manufacturing process each of the SX's were from.
And you would be right. As I told in the video, I think, it's the chipset, which can't handle anything above 24MHz. So far, I didn't see this mainboard with other CPUs, than 16-20MHz 386sx. So I guess, it was just too bad for 25MHz or even 33MHz, which I tried.
Back in 1993 I overclocked my 486 SX 25MHz to 33Mhz by soldering in a new crystal oscillator on the motherboard. Insane how much hassle that was compared to overclocking today but very effective and cheap to do!
The problem in USSR was not even, that they were so expensive. The problem was, that you couldn't buy the at all. All you could get more or less in regards of x86 were some soviet 8086/8088 clones. Actually, in the mid/end of 80' USSR was economically ruined and you couldn't get even simple things, which you needed day in day out, not talking about computers and stuff...
@@necro_ware My brother built a 286 in the 80's out of fried parts smuggled from Western Europe, which were worth a fortune. There is no way in the world anyone could afford new parts, when a 1.44 floppy drive was worth 3 monthly salaries on the black market. There was no legal way to obtain parts otherwise. My school only had domestic computers. Clones of various foreign systems that by then were already 10-20 years old and falling apart. Having an IBM-compatible at home was an absolute miracle.
Yes, the x86 compatibles were inaccessible. As I wrote, you could still buy XT clones. They were also very expensive, but a t least, you could get them legally (in theory). And the school computers were not as old, as you think. They just looked old, but just as you said, they were clones of various western personal computers. For example of ZX Spectrum, which was introduced in 1982. It was cloned in the USSR very much, but they needed around two years to get there first. So such computers could only be 5-6 years old as USSR fell apart. If they would be 10-20 years old in the year 1991, that would mean, that USSR was ahead of the western personal computer technology. I have my doubts about that :) However, USSR was really not bad at supercomputers at some point. For example, take a look at Elbrus-I and -II. Intel was so keen about that technology, that they snatched all the engineers from the project right after the USSR was history. This very same engineers developed Pentium III later, which technology is used until today in all the Core CPUs.
My first desktop was a 33MHz 386 clocked at 40MHz. When I discovered that I felt a little betrayed (this is not a genuine 40MHz cpu!) and that was the last pre-build I ever bought. Always sourced my own components ever since and typically slightly underclock/undervolt my latest hardware to not be blown away by the cooling. If you grew up with 33MHz, 3.8 or 4.0GHz both are exceptionally fast.
@11:35 Pro (?) tip for soldering large SMDs: use a strip of kapton tape to hold the component in place, it makes it much easier to align the pins. Once the component is properly aligned, tack down (solder) a couple of pins on opposite corners, then gently remove the kapton tape and solder the rest of the pins.
Thanks for the tip. I usually just hold it in place using fingers, it's not too tricky, if you don't have to film the procedure. But I'll keep your tip in mind should I have problems with my usual approach.
Back in the day when these weren't old, I did have a 386sx system I overclocked from 16MHz to 33MHz successfully. Since the motherboard is made for 286s and the role of the 386sx was to be a drop-in replacement of the 286, the chipsets were often the limitation and that's also why yours is specced at 24MHz.
But 24 Mhz wasn't exactly a standard speed for 286es either, although 20 was. So I bet they designed for 20 MHz top speed, then validated and accepted whatever they got as it was better than 20, and this turned out to be 24. It's probably worth trying to find a 48 MHz crystal.
What a trip down memory lane, especially with Doom... I had a 40MHz 386DX board, and Doom (under OS/2) just ran fine. It got a bit laggy when a lot of enemies got into the same room as you, and a lot of shooting and shouting ensued.
I feel the same, that's why I'm doing this videos in the first place :D Glad you liked it. And in regards of Doom, I think our memory cheats on us way too often. Sure everybody has his own view on where a game is considered to be playable or not. At least here you can find some real numbers and measurements made by Phil are in-line with my own experience: ruclips.net/video/qQEHHc1q06c/видео.html
Remembers me on my first PC - Pentium 200 MMX. I had that system running stable with 250MHz, a speed increase that was indeed noticeable in games back then. And for a long time the only system I could overclock for 25% it's original speed... After a long time I had a 3GHz cpu that I could only overclock about 200MHz, I chose not to do it because the speed increase was not significantly enough. Many years later I had a 3,4GHz cpu that could be overclocked automatically to 4,3GHz, that was worth it and it ran absolutely fine.
11:07 - hah I had a DX-40 MHz board that would take an external CPU, and I wanted to put in an SXL (TI) - I played with all the jumpers, it would just not run with any other CPU I had in the PGA socket. Finally I took a heatgun to the CPU on board (I needed one for my display collection anyhow) and it came right off. I didn't take any special precautions, luckily it was in the very corner of the board. I cleaned up any pads that may have cross-soldered in the removal, and plopped in the TI "486" CPU. Booted right up. :D
This was the exact mainboard i had on my first PC, i replaced the crystal with a 40hz one and added the math co-processor. Doom ran quite a bit better after that. Excellent video!
@@pauldame9925 Time often plays tricks on our memory. I think you mean Wolfenstein 3D, that was kind of playable at 20MHz 386sx, but for Doom you really need at least a 386DX-40 to run it in a very small window in low details and to have it playable you need at least a 486DX-40, or even better a 486DX2-66
Brings back memories of a 386sx my brother got me from work a clear out in the 90's, many nights were spent overclocking it just so i could play DooM on it. I think it was a 33mhz cpu, but there were dip switches and jumpers on the board to adjust clock speed. it just about manged to run it at low graphic setting full screen and high settings at half screen size. Wish i still had it. thats where my love for overclocking and using old tech comes from to this day. would like to see how far you could push a 486, or if one could run on the 386 board ? its great seeing the real time effects of the clock increase on games like DooM... from slide show to playable.
Yeah, in the time of 386, Wolfenstein 3D was more a game to play. First 386DX-40 could run Doom at low settings. Unfortunately even that was too much. As soon as more enemies came to get you, your FPS dropped dramatically to unplayable. It was first more or less playable with 486@40MHz, because of 8K (!) L1 cache :) Usual 486 CPU was not compatible to 386 mainboards, but Cyrix did some magic and brought own 386 CPU with 1KB cache and 486 instructions. It was still slower, than a real 486, but could help to push your 386 a bit. Good news, I have all parts for this recipe and wanted to do a video about it soon :)
@@necro_ware does having the Math Co-processor help much with these 386 - 486 cpus? Would there be any benefit in overclocking those as well ? Look forward to seeing the Cyrix! I'm sure I had a Cyrix "586" at one stage in the 2000s that was fun to mess with. Subbed! Glad you popped up in my recomended
@@austinmaxi Glad you've subbed :) Regarding math co-processors, it depends on what you are interested in. They didn't help in that days for games, because math co-processors were for acceleration of floating point arithmetic and most games didn't use floating points at all. Only professional applications, like simulation or CAD did benefit of it. The floating point arithmetic came first with 3D polygonal games and one of the first famous games, which started to rely on a math co-processor was Quake. But it is the era of Pentium CPUs already and 386/486 are sitting nervous in the corner :) It is also not easy to benchmark games from the time before Pentium, because all the integrated benchmarks in the games started to appear first as overclocking became a "sport" or "challenge", as I told in the video. Unfortunately, with the old games you can just say: "Yes, it feels faster", not very exciting, yes? Well, this is one of the reasons, why I just used some synthetic tests in this video, at least you get some numbers to compare. However, I'll try to do better next time :)
Like people already commented, when you overclock 386, you also overclock its FSB (1x20MHz, 1x25MHz, etc), so the limit is always the core logic and other not-so-OC-friendly motherboard components. Also, I've seen 48.000 oscillators on some period ISA video cards at least, so there is a real chance that some of these boards were indeed 24MHz (O_o), with downclocked 386SX-25 (?). Overall great video! Really enjoyed your soldering and test results! I'm playing with the later SX-40 now, and honestly, Doom is still unplayable on it... But.. there's always DX-40.. ;) Keep those videos coming! ;)
Yes, everything's true, what you are saying. One thing is, that Doom is not playable on 386DX-40 as well :D Only in small window with low quality you will get near some playable values. This game was just made for 486. It's like Wolfenstein3D is for 386, Doom is for 486 and Quake is for Pentium.
Aah yes, my first encounter was accidently putting the jumper on my Pentium 133Mhz to 150Mhz... I was amazed it worked 😁 Thumbs up for your soldering. I also once got a by-fault-factory-overclocked 3dfx card that kicked ass but overheated... being an idiot at the time I had it replaced under warranty... should have just added some cooling and let it kick ass. The new card I got naturally didn't perform as good.
Whoaaa I remember replacing those oscillators myself in the 90s. I was a teenager and those 386 SX boards were cheap to get, yet they were able to run 32 bit software. Because of that, they take a special place in my heart. I did all kinds of experiments. Would like to somehow see Doom running playable on a SX, which is a really hard task I guess.
Yes, those 386sx were more or less 286 CPUs with 386 features. Internally they were 32 bit, but used 16 bit memory bus. That's why, 286 mainboards could be used with slight modifications by the manufacturers. This made such machines very cheap... and absolutely not capable running doom. If you are happy with 12 FPS in a window as big as a post stamp, then you need at least 386DX-40. Really playable is the game first on 486DX2.
My first computer? IBM Turbo XT @10mhz. Did the 286,386sx and dx builds, and then special ordered a 486 dx 2/66 board, cpu, and all the goodies. Conner 170mb IDE HD ("you'll never fill that in your lifetime") Got a wild hair, later and bought a 486dx 4/100 CPU, put in 100meg of ram and shut off virtual memory in windows 3.1... That thing was a screamer! ***sigh*** My brain was a sponge for all things electronic back then. Now I'm lucky if I can shut off the alarm on my smartphone.
That was awesome! I'm currently trying to get a Octek Jaguar II 386 board back to life! It currently has a 25Mhz Intel processor in it but according to the manual I can also upgrade it to 33Mhz. However, it has a 50Mhz oscillator crystal soldered onto the motherboard. I'm curious to see if it would still give me that 33Mhz that they claim. Would be interesting! But I may need to replace that crystal with a 66Mhz. Aside from that, the reason it doesn't seem to be working is memory issues. There are two bios chips of AMI that came with the board, one shows all 8mb of RAM but crashes when it's time to boot. The other only seems to see 1mb of the 8mb installed and that's it but it does work stable. I've taken a look and it seems the SIMMs that are installed in the board are 60ns while the manual indicates that for a stable working system it would need 70ns or 80ns and then also setting up the proper wait-states. Eitherway, looking forward to sinking my teeth into this motherboard and if I can get it to work, hopefully find a nice 386-era case to put it in! :D
Hi, I don't know if I got you wrong, but as I read it, you complain about 60ns memory. Well actually, in this case less is better, 70ns is better than 80ns and 60ns is better than 70ns. And 60ns is the fastest SIMM memory you can get as far as I know. Could be, that you have problems to get it booting with 8MB, because your memory is bad. Octek Jaguar II 386 is a 386DX mainboard, it has 8 memory slots and you have to use it in packs of four sticks. If you have 8 memory sticks with 1MB each, you can try to remove 4 and try to boot with only 4MB, if that doesn't work replace all the sticks by another four. If that also doesn't work, take two from the first pack and two from the second. If you go through all of the memory sticks in this way and only up to 4 of them are bad, you will end up with 4MB of working RAM. For this machine it is mostly enough anyway.
We did once weird stuff with an AMD 386DX-40. Friend had not enough money for an 486 but the 386DX40 were insanely cheap at those days. We found out, his sample was totally happy running with 80mhz.
Yes, in very rare cases it is possible, but you heavily overclock the ISA bus and I can imagine, that many expansion cards would get instable resulting in broken audio, video etc...
@@necro_ware the beauty of that board was: you could lower the speed of that ISA bus by a factor tied to the CPU. It took my friend days to find stable settings, where everything was performing real fast, without crashing. He exchanged even his sound card for that one of a friend, because his original soundblaster had problems with 12mhz ISA-Bus speed and the compatible soundcard was rocksolid at that speed. That dude ran a VLB setup and saw no need for overclocking his ISA-Bus. But with an ISA bus, you want to run your Tseng4000 as fast as it goes without crashing.
@@ciddax754 Yes, some rare boards allow to set CLK/6 divider for ISA BUS. This means about 13MHz, when CPU is @80MHz. Unfortunately most of the 386 boards doesn't allow to set divider at all and the ones, which do, they provide CLK/5 at best. Again, on @80 MHz CPU, means 16MHz. That would require a lot of fine tuning and hardware selection. Not even every Tseng ET4000 would run at that frequencies stable. I have one here at hand, which says goodbye already at 12 MHz ISA. It's not always an easy task to find ISA hardware which runs above the standard 8MHz and if it does, usually not for long. I overclocked one 386 a year ago to 50MHz it was stable for about 3 Months, then the graphics card got broken first. I replaced it and a month later the chipset on the board died as well.... RIP
I know this is an older video. Anyhow - nice job! What is that metal you use to desolder? I can ´t find it anywhere. The board you have is more or less "286ish" chipset which was to save costs back then compared to 386DX. Even running a 286 at 25Mhz is a challange with most of the chipsets and unfortunately it can even kill the chipset later on. Adding pasive cooling will help. Those old UMC guys are nice but not the powerhouses 😁. For those experiments try to get later single (double) ICs integrated chipsets. Old NEAT and HEAT as far as I know don ´t OC well. You are my favorit tech RUclipsr and I learned a lot from you!!
I still have my aunt's old Pentium MMX 166. It came with a fake badge on it saying it was a 200 and ran clocked at 200Mhz for its entire life, kinda want to try and get a new mainboard PSU etc for it and build a DOS gaming PC because as far as I know it should still work fine.
how did you do surface mounted soldering so cleany!? i tried it and messed up. may be u should do a video on how to do it properly? i have a 286 M216 mobo with bad soldering on the chipset and it needs to be reflowed i want to know how this is done well
You are finding the DMA controller is running away from the clock speed. Probably the CPU will overclock, but is held back by the DMA, causing memory allocation errors, as it's unable to clear it's instruction code intime for the following set. I had this and got a error: Divide by zero error.
Yes, I suspect something like that as well and on this board the DMA controller is integrated into the chipset, so I can't do anything about it. Thanks for sharing your idea.
I remember the days of overclocking with a pencil... My first entry into overclocking was with an AMD K6-2 500 with 3 fans tie-wrapped to the heatsink and 60mhz later with 2.81v. To this day I still overclock every system, card, and chip hah
Yeah, I used the pencil trick on my Athlon as well back then. AMD cut some wires to prevent people to setup the multiplier, but you could simply draw over the wire with a pencil and it was good to go ;D
That wasn't my first try, but I also remember trying to squeeze more speed out of a K6/2-500 and not getting very far. I figured I could at least get to 550 since there were K6/2 chips rated to that speed, but no. The other hardware on the board wasn't happy at anything above standard ISA and VLB clocks, so I had to run 5.5x and less than 100 MHz FSB, finally topping out around 530 MHz, and even that required 0.2V overvoltage and got a lot hotter. The loss of FSB, thus drive throughput, video throughput, memory throughput, etc. meant on balance it wasn't worth it.
I was under the impression the DX chips were much more overclock friendly. The SX versions needed the math co-processor for any realistic speed boost. And well, it was a long time ago. So I'm probably forgetting everything. Lol. Fun video to watch, thanks.
Well, it may vary from CPU to CPU, but in general DX are not really more overclock friendly, but they are a lot faster to begin with. Compared to SX the DX had 32-bit RAM bus instead of 16-bit and ability to have up to 128KB cache, which made DX by far a more superior CPU. SX on the other hand was just a 286 on steroids, equipped with internal 32-bit capability and 386 instructions set. It was made as a cheap solution, so the mainboard manufacturers could continue to use their slightly modified 286 mainboards with 386sx. And in regards of math co-processor, it is absolutely useless on a 386 (either DX or SX). I know only two games, which made use of it - SimCity and Falcon 3.0. All the other games didn't make any use of it. You could get some slight performance gain in Word and Excel under Windows and in CAD software though. So, if you didn't use that programs, you wouldn't see any change with or without math co-processor.
@@necro_ware thanks much for the refresher. I didn't think my household expected to run anything outside of Wordperfect and Lotus back in those days. We had a 386sx with the math co-processor, and NU's benchmarks always showed the increase. This was a great video to watch though and your information was top notch, thank you.
At a Hamfest in 2005, I witnessed an Intel 286 overclocked at 1.2Ghz. It functioned, but needed to be cooled with liquid Nitrogen to keep it from melting.
I barely can imagine that :D It is physically not possible. The technology used in 286 was not capable to run at such frequencies, doesn't matter which voltage or cooling you used. May be you mixed something? I mean 2005 is long time ago.... ;)
I had a Pentium 60 with the floating point division bug. I swapped the crystal to put it up to 66MHz and took it into Computer City to have them replace the CPU with a non-bugged version. Free Pentium 66 upgrade.
yeah 25% overclocking might be too much. You also have to watch the ISA Bus clock doing so. Some mainboards can't adapt the frequency and go to far over 8MHz, which makes ISA cards unstable...
I have a long history in overclocking. My first overclocking was Pentium 120 running at 133 MHz with cache maxed out from 256 to 512kb on board. I also had the best results with a Pentium 233MMX, running at 250MHz with 100 MHz FSB on a FIC VIA motherboard with 1MB cache, completely smoking Pentium II 233 of the time. Now I don't overclock, I prefer to buy second hand Xeon to push my motherboards to their upper limits
My original pc had a 386, but i managed to get a 486dx, or a cyrix 586 running on it. The system had jumpers for bus multiplier and voltage changes. More voltage was needed for the better performance. It was a few years ago so i cant remember the exact details i used.
When I was a kid, I remember we had a Intel Celeron @ 366MHz and I couldn't play any game at decent framerate, it was software rendering too. So I tried overclocking it as much as I could, increased FSB frequency multiplyer and so on. Needless to say, computer went dead. My dad took it to a repair shop, I was terrified and totally anxious if they will tell him what I did, I will surely get some ass beating :) But they reset BIOS, replaced CPU and RAM, and my dad never found out what I did :) Either they didn't know that overclocking destroyed it, or they realized that it's propbably too hard to explain. Either way I've learned my lesson even without beating of my ass, the fear of it was enough :)
I remember I had a PC with a Pentium at 90mhz as a kid one day I was playing around with it and noticed a switch on the motherboard to change the frequency from 90 and 120, easy +30mhz.
Thanks, Fnirsi-5012H. For simple measurements it's quite good, because it's portable and delivers a decent job. But it is advertised for up to 100MHz and it works reliably only for up to 35-40MHz. From there it starts to show wrong voltage levels, but if you just want to see some frequency or if some activity happens on a pin, it's still sufficient...
Performance increase of 386 isn't any more linear than later CPU's. If you did tests in Doom you would see that you are limited by the speed of video card (though from numbers in Landmark I see your VGA is pretty awesome so it might not be as big of an issue, but still...). The best overclock I had in my life was cheap E2160 CPU that went from 1.6GHz to 3.2GHz and that would be comparable to if you could overclock 16MHz to 32MHz. This was through FSB overclock so raw CPU performance including its ability to talk to memory linearly increased by 100% which was reflected in pure CPU and CPU+memory benchmarks. Of course this affected memory timings and clock divider so memory had to have bigger timings but this is not CPU related and actually identical to 386... and by the way, did you try to increase memory timings after OC? There were in most BIOSes setting like turbo timigns or slow timings. Tinkering with these might improve stability with higher clocks
I used the fastest graphics card I had. The problem with Doom and ISA is the limited bandwidth, but 386 is not yet limited by ISA bandwidth. You run against this limitation first with 486DX-33 and faster CPUs. And in regards of the overclocking, I had some talk here and there and it seems to be the chipset indeed. All the mainboards I found were running CPUs with 20MHz max. I didn't find one board which would go higher then that. So the conclusion is so far really the chipset. It probably can't stand higher frequencies, than 25MHz at all. I also played with different memory and timings. No way to convince it :D
i started overclocking with a Spitfire Duron 700 to almost 1,1 GHz with a homemade air-duct-cooling contraption which sucked cold air from outside in the winter too cool the CPU, i was very proud of myself - but later i was pretty sad as i discovered that there were folks out there that cranked up the clock to 1,5 GHz for those CPUs :)
One of my buddies discovered these coolers (beer type cooler) that you could plug into your lighter socket and it would get cold or warm depending on how the power was hooked up. They used a peltier chip that was about the same size as a 486, so several of us used them for CPU cooling. We were mainly messing around with Cyrix chips, which got really hot and flaky when overclocked, and it worked like a charm.
Man, I still remember bringing my Pentium II (I think it was 233 Mhz) cpu to some computer shop, exchanging it for a Celeron 300A while getting some cash back and then overclocking that from 300 to 450Mhz, ending up with a much faster PC *and* *getting* *money* *for* *it* . Those were crazy times!
The noise of the CPU fan though...
Hahaha.
Yeah I had a 300a @ 464.
It could have done 504, I just didn't know SHIT about things back then and didn't even use thermal paste. Just the pad that came with the cooler.
Made a Celery Sandwich with it, without thermal pads. Lol.
I also had a PII 300 SL2W8 that was a 450 in disguise. Did 450mhz easily.
One day my Mom came home (It was her PC) and I told her the CPU was at 450mhz now. Lol, great times.
Excellent video! I've run into the same issue when doing this on a 286. The chipset was the limitation.
Thank you Adrian! I'm very glad to know, that despite of the amazing amount of (great) work you are doing, you still have some spare time to watch others videos ;)
geez, spoilers! ;)
@@necro_ware Is this overclocking technique possible on a 486 as well? Once you've maxed out the FSB speed and CPU multiplier there's nothing more you can do on the software side of things, but swapping out the crystal oscillator/clock generator sounds like it could work. I'm asking because I recently found a Socket 3 board, the Shuttle HOT-433, that can do 83MHz on the FSB, and I wanted to know if it would be feasible to push it even further than that.
@@FordGTmaniac Some early 486 also used such oscillators, but later version like HOT-433 used clock generator IC like MX8315 (see between PCI slots on your board). This clock generator can generate up to 80MHz. I personally don't think, that this board or any 486 CPU can stand 83MHz FSB anyway. At least not for longer period of time.
@@necro_ware Oh, ok. That particular board uses the UMC 8886AF / 8881F chipset which was also used for Pentiums, or so I've read. That would explain why the FSB can be made to run so fast despite it being a 486 board. 83MHz isn't an officially listed setting even though it is fully capable of doing so, probably to prevent people from accidentally frying their CPU's.
Great video and amazing solder skills. 👍🏻 Greetings from CPUGALAXY
Thank you very much!
I remember changing jumpers took a Pentium 166mhz to 233mhz for massive gains. Opening internet explorer 5 while playing music on Winamp with no stutter was amazing
Ahhh... the memories huh? Hadn't heard the word Winamp in years.
@@garyr7027 It really whips the llamas ass.
I had a P3 500, 100mhz bus that OC'd to 733/133. Also had a Celeron 300/100 overclock to a 450/133 I think. Good old days. Then again my i7 7700k OCs from 4.2~4.5 to 5Ghz.
This reminds me of the old TurboPLL project, which replaced the clockgenerator on the motherboard. You could then increase/decrease clocks on the fly. This project improved overclocks in the early 2000s.
I still remember my dual celeron 366 overclocked to to 550mhz. Man it was the SHIZ as a 12 year old
Yup, had two Abit BP6 myself. Too bad Intel got greedy and removed SMP support... good for AMD Thunderbird and Tyan boards. ;)
@@JohnnyCacheX same here! BP6 w/dual 550MHz celey life was so sweet in its day.
How did we ever survive with single CPU cores? How did we ever think that was the way it _should_ be?
I have to say, this was my favorite time during computer history. I loved the 8088 days too... but the 386 SX time, I don't know, it felt like computers were so much better put together.
Just watched your video on overcloking the 386SX, and I'm seriously blown away by the quality of your content and explanations! Your work is fantastic, and I'm itching to try it on my own 386SX 40 MHz. Thanks a ton for the top-notch content and insights. Truly impressive!
Never stop making videos, you are fun, interesting, your projects are awesome and you are very carismatic.
Great video, nice to see someone with SMD reworking skills!
Thank you! It took some time of training, but at least I can do it, if needed. However, I only do it, if I need to. I still prefer through hole soldering for tinkering whenever possible.
Most enjoyable video. If I recall 40MHz 386 DX is about the minimum for full screen Doom to run smoothly
I have never seen a 386 above 40MHz that was factory boosted. Yes, you had to be at 40 to play DOOM FS, but you got like 15-20 FPS. My friend would stomp me in DOOM but I trounced him in Descent (similar performance, I was just better in 3D space.) My friend had a 486 DX and I had a 386/40 (it was a 25, the 40 was turbo.) You remember your first computer... This was not my first computer, that was a Compaq 251 Deskpro, an 8086 CPU.
I will not start to argue, because everybody has own expectations on, what is playable and what's not. It's in the end up to you, but I think Doom is not playable on an any 386DX, not to mention "smoothly". At least here you have some numbers and they are completely in-line with my own measurements: ruclips.net/video/qQEHHc1q06c/видео.html
I remember playing Doom on 386DX/40. It was an AMD chip. Intel only went up to 33.
TI 486dlc that fit in a 386 socket @ 40mhz with a 33mhz or 40mhz 80387 type *ulsi or iit were my goto*. Consistently got near 486dx-40. That combo probably only worked out well because of the board i was using. Good times.
@@Tris289 I have this combination on a very fast 386 mobo. 486DLC + IIT4C87 both @40MHz (not overclocked). It is from the performance between 486DX-25 and 33. More like 25 in the most benchmarks.
You are a soldering wizard, sir. Respect!
I did overclocking on a 286 by changing the crystal oscillator from 16MHz to a 20 or 25Mhz.proved by PC Doctor.very nice game.thank you for remind me about this :)
When I see people torturing old hardware, I do think that, if done properly, it can be a valuable lesson for all of us, and save other boards from such a fate. This was a fine masterclass in the limitations and capabilities of old hardware. Thanks you.
I actually also don't like to "torture" this grandpas of computer history, but curiosity won and many people seem to be interested as well. Now they don't need to to put their hardware in danger :D
you made that look like a piece of cake. great history lesson.
Back in 1993, my first PC was a clone of IBM AT 12 Mhz. It did have a CPU in a ceramic housing and quite hefty heatsink
attached to it. It was just too slow for everything i was trying to accomplish. So i've had this idea of overclocking. Went to a local parts store and bought every ceramic resonator between 12 and 22 mhz they had on offer. Removed the original one and soldered a socket in its place to aid for easy swap. Then, gradually inserted faster and faster crystals and running the PC for couple of hours to see if it malfunctioned. Found out that i could safely run it at 19 Mhz which was a huge improvement. But that was not the end. The machine did not have math coprocessor and i was torturing it with CAD tasks so i found out that maybe it would be good idea to populate that empty Coprocessor socket on the board. Scavenged a 80287 MPU and when casually going through its datasheet, noticed that it actually runs totally asynchronous to the main CPU. Rigged up a sort of adapter that went in between the board and MPU that contained its own resonator (it was a proper, powered generator supplied with 5V while board used traditional 2-pin crystal) that ran@20 Mhz. The result was a genuinely fast computer that ran blazingly fast 3D CAD calculations. Chipset didn't seem to be the limit but i do not recall its brand or IC manufacturer. Most likely, it was an older "hybrid" board that integrated some of the functions (like MMU, interrupts...) into a sort of multifunction chip, but it stil contained large number of individual Intel circuits, like UARTs, keyboard etc.
Thank you for sharing your experience! If you tinkered with 286 back in the days, may be you are also interested in my various 286 restoration videos ;)
My first overclock was made on 80286 CPU with use of the TURBO button :)
With later CPUs I used many different methods for overclock (switching jumpers, replacing oscilators, cuting traces or isolating CPU pads/pins). I miss the good old times when oveclocking gave fun and was worth it (like overclocking Celeron 266MHz/66MHz FSB to 400MHz/100MHz FSB or more).
Wasn't Turbo to actually slow down the cpu so DOS application/games wouldn't run at abnormal speed ?
@@gagarin777 True. The turbo function was underclocking, but sounded fancier as "turbo". Many people thought they were saving the system's stability and reliability by having it run off the turbo mode as much as possible.
I really wanted to do this as a kid building my first computer, an SX20.. But I was too scared. What an excellent video!
I admire your patience.
I admire YOUR patience, you watched the video :D
WOW. Soldering that by hand! WOOOOOOOOOOOOOOOOOOOOOOOW! I would NEVER even consider that... 👍👍👍👍👍👍
Absolutely amazing journey you took me on here.
Thank you, for your many different approaches on overclocking this CPU/Mainboard combo, highly appreciated
Landmark Speedtest!? Thanks for making me feel old! 🤣
You are welcome! :)
I remember over clocking my amdK6-2 to 600mhz in 2003 to play city of hero's. Good times. (And no it didn't play well at all. But it did play)
Been there. Just did whatever we could back then for 3 more fps
I ended up damaging my K6-200 by going from 66mhz to 83mhz on the bus. 75 however worked fine, but going to 83 made it so the CPU was no longer stable even at the stock settings anymore.
Did the same to my athlon. Scorched the bottom of the cpu.... Let the magic smoke out. 1.2GHZ to 2.1.... it ran for 3 months :)
Great video insight to clock speeds and frequencies when experimenting with overclocking.👍
That was very interesting indeed. I have a 386SX myself that I plan on working on, I will be taking your findings here on board.
You are welcome!
Hi! What a fantastic video! I must say that you have the nuts in place :D Dessoldering the CPU from a good working Mobo was very rad and risky. I love to see the charts and the performance increase. My first PC was an 386SX25, one week later, in the same store, I saw an 386DX33 at same price. :(
Thank you Jorge! Yes, desoldering the CPU was kind of crazy idea, but I couldn't stand the curiosity :) Unfortunately I couldn't reach 33MHz with this mainboard, but at least it was an interesting experiment. And since I could restore everything to the original state, I guess it went quite good in the end.
My father bought me my first PC, it meant to be a 386SX-25 either, at least, that was, what I thought. Years later, as I understood a little bit more about it, I realized, that my father was scammed and the CPU was actually 20MHz, but sold with 50MHz crystal oscillator. So my first PC was overclocked from the beginning by an ugly scammer :)
I still have that old overclocked hardware under a glass cover in my working room. It is still completely working!
@@necro_ware wow, who sold him overclocked CPU? But in that time, it was matter of few hundreds $, so it was tempting to do by sellers. Was it regular store, or he bought used computer from someone?
@@warrax111 It was a "regular" store, but I probably have to add, that it was in Moscow :) It were 90's, a dark age, no rules, no law,, no government. Scammers on every corner, so no wonder, that even a usual store scammed on their customers. As you can imagine, the price difference was quite a lot between real 20 and 25MHz machines. Anyway, I find it quite funny today. It's a part of my history now :D
@@necro_ware Yep, I am also from Soviet Union, I know what you are speaking about. :) Nice story though, it was just like that, when people started to grow up, and know something about computers, it was too late for doing anything, the sellers used it in early 90's.
@@warrax111 Hehe, the world is a village. Then you know, what I'm talking about :)
What a walk down memory lane such nostalgia I have been trying to get a running 286/386/486 motherboard I have a 486dx2 Pentium overdrive CPU but finding a working or even dead motherboard is a very expensive task even if dead I would repair it but ebay sellers are not the most reliable and our local e waste disposal depot wont sell any of the good components I have seen there yet they charge an arm and a leg for people to dispose of them I do however have managed to salvage 2 8086 processors and a 8088 but motherboards for those in the wild are rare..... Thank you for such a nostalgic memory
Just overclocked my 16mhz to 20mhz thanks to this video
I had a 486 back in the mid nineties, started life as a DX-33, but I managed to get an AMD DX4-120 CPU. Problem was my motherboard could only do 33Mhz bus, not 40, so I couldn't reach 120Mhz, "only" 100. I replaced the crystal on the board to get 40Mhz bus and it worked flawlessly. The PC still "thought" it was running at 33Mhz since I had changed the reference frequency (I think?), but benchmarks and performance clearly showed the improvement. The computer lived on for a couple of more years running Windows 95. Even battled with my friends' Pentium 60 and fared pretty well actually, which made him pretty pissed since he spend a lot of money on his new shiny Pentium :P
I don't necessarily miss those days, things were so complicated, but I do miss the hardware and the "feel" of successfully getting the PC to run fast and stable. Which is why I build retro PCs, collect hardware and help arrange retro LAN parties and play lots of retro PC games these days. Just keeping those old feels alive :)
Hehe, you probably terribly overvolted your CPU back then. If you have had a mainboard, where you could replace a crystal, then it definitely didn't have a 3.3V support. All the mainboards with 3.3V CPU support never had a crystal, they already were all with jumpers and variable clock generator IC. So you probably powered your CPU with 5V instead. You were lucky, that it survived so long. And regarding the DX series, this was the first time, where such a thing like Front Side Bus (FSB) was born. DX, DX2, DX4 was nothing else, but a multiplier 1x, 2x and 3x. The mainboards usually don't care which one of the said CPU is inserted. The main difference was between DX and SX (no FPU). DX2 and DX4 were technically still the same as DX, the multiplier made only difference internally inside of the CPU and was completely transparent for the mainboard.
I also would assume, that your mainboard could actually do 40MHz FSB as well, but probably this frequency at 5V was too much for the CPU. Back in the days you had problems with FSB of 40MHz only on mainboards with VLB support, but most of them had additional wait state jumper, in case FSB was set to over 33MHz. Furthermore, I suppose, your mainboard was ISA only, because as VLB was introduced, they didn't use plain crystals anymore.
This was really cool. Thank you for the work you put in.
protip: dead sound / MDAish video cards often supplied with many onboard XTALs at different clock rate, to support different refresh rates.
They're also still used to tune old ham radio gear.
I feel kinda young. Our first family computer was the cyrix 486dx2, which sadly I gave away. I've kept a complete 386sx low profile board for a wet day project. Really enjoyed your video.
This brings back memories. The.very first personal computer I built was a 386sx. I think it had like 1 Megabyte of Ram. Lol
I remember a 386sx 20 machine, doing a RAM upgrade on it and spotted the CPU was actually a 25 - and the oscillator was socketed - hmm, some free performance to be had? - couldn't find a cheap 50MHz, but picked up a 48 from a radio rally...
Stuck it in, and a perfect 24MHz was achieved - and the ISA ran at a nice 1/3 for 8MHz, instead of an evil 6.66MHz or an overclocked 10
celeron coppermine cd0 stepping 566mhz is an overclocking monster
66mhz bus and high multiplier allow this chip to run at 1133mhz by removing bsel pins on the cpu (forcing bus selection to be recognised as 133 by default) and even overclock from there
since via chipsets for coppermine/tualatin cores can't aply fsb/pcie dividers correctly this was also the only way to get stable overclocking for tualatin celerons over 133mhz bus
i had 2 cd0 coppermines 566 and both of them run crazy stable on stock 1,5v and 133 bus, what more that barely ever got warm
i love your experiments. they really shows how much potential washidden inside old computer parts , artifically segmented and downclocked to create "less expensive" options for the market
First exposure to the channel; I'd forgotten what a nerd I am at my core. Also forgot how much fun it was to play with these components as a kid! As well as being thankful we had an OEM 486SX/25 build that ran Doom quite well!
Our memory cheats on us :D I also always remember, that some game ran well on some of my old machines, until I try it once again. In full screen Doom ran on 486sx-25 at around 15-16 FPS, if I remember right. Let's say it was kind of playable :D
@@necro_ware let me have this (idealized/flawed) memory dammit 🤣🤣🤣🤘🤙
My first overclock in ninetities was 486SX/25 @40MHz just by swapping oscillator. It was already in socket. Later got 486DX2/66 and that went up to 90MHz. It was sick fast with VLB display card. Chipset on the mobo was from UMC by the way.
i used to overclock dx4 100 to 166 mhz it was really a performance boost back in the day to have almost pentium performance
wow, this was a gamble: VLB cards didn't like overclocking
@@DolganoFF Card was S3 Vision864 VLB.
My 1st easy overclock was with classic Pentium 150 Mhz with 60 to 66 Mhz FSB jumper which resulted in a Pentium 166Mhz.
MMX is limited to integer formats that recycled the FPU registers. Switch between X87 and MMX resulted in pipeline frush. SSE's support for FP32 and separate register storage fixes MMX's problems.
What a wonderful video with nice and professional explanations! I enjoyed it a lot. Thank you so much😊
Your asking a lot from your breadboard! Also It's also a good idea to add heat sinks . I remember clocking an old 286 by replacing the crystal.
I was concerned about breadboard as well and tested the signals with oscilloscope before I started to do anything. The signals were quite clean and reliable. The temperature was also completely ok. If I remember right, I even show it in the video. Anyway the problem with the chipset was already confirmed by multiple people. It doesn't stand frequencies above 25 MHz.
Found this video in my feed probably because of being subscribed to Adrian Black aka Adrians Digital Basement. Liked the video. You’ve got a new subscriber. Cudos my friend.
Thank you and welcome on my channel. I really have no clue why, but this 386 overclocking video got a lot of attention in the last days, despite, that it is actually quite old. Well, that makes hope, that people will take a look on the other videos I made in the last time as well ;D
Having the idea of switching the CPU is one thing. Actually doing it is another. Crazy. Fun to watch, though!
I overclocked my DX4/100 to 120MHz, which was a big financial saving in those days. A friend of mine bought a PC that was advertised as 120MHz, but was actually an OC 100MHz model and kept crashing. They very quickly replaced the CPU when we took it back!
This is how I got started with overclocking -- soldering in new oscillators! EDIT: wow - i would have bet money it was the motherboard or ram holding back that clock speed, not the cpu itself! I'm curious what manufacturing process each of the SX's were from.
And you would be right. As I told in the video, I think, it's the chipset, which can't handle anything above 24MHz. So far, I didn't see this mainboard with other CPUs, than 16-20MHz 386sx. So I guess, it was just too bad for 25MHz or even 33MHz, which I tried.
Back in 1993 I overclocked my 486 SX 25MHz to 33Mhz by soldering in a new crystal oscillator on the motherboard. Insane how much hassle that was compared to overclocking today but very effective and cheap to do!
Especially compared to what the PCs did cost back in the days.
In USSR a 286 was worth as much as a house, even while in the West they already had 486's.
The problem in USSR was not even, that they were so expensive. The problem was, that you couldn't buy the at all. All you could get more or less in regards of x86 were some soviet 8086/8088 clones. Actually, in the mid/end of 80' USSR was economically ruined and you couldn't get even simple things, which you needed day in day out, not talking about computers and stuff...
@@necro_ware My brother built a 286 in the 80's out of fried parts smuggled from Western Europe, which were worth a fortune. There is no way in the world anyone could afford new parts, when a 1.44 floppy drive was worth 3 monthly salaries on the black market. There was no legal way to obtain parts otherwise. My school only had domestic computers. Clones of various foreign systems that by then were already 10-20 years old and falling apart. Having an IBM-compatible at home was an absolute miracle.
Yes, the x86 compatibles were inaccessible. As I wrote, you could still buy XT clones. They were also very expensive, but a t least, you could get them legally (in theory). And the school computers were not as old, as you think. They just looked old, but just as you said, they were clones of various western personal computers. For example of ZX Spectrum, which was introduced in 1982. It was cloned in the USSR very much, but they needed around two years to get there first. So such computers could only be 5-6 years old as USSR fell apart. If they would be 10-20 years old in the year 1991, that would mean, that USSR was ahead of the western personal computer technology. I have my doubts about that :) However, USSR was really not bad at supercomputers at some point. For example, take a look at Elbrus-I and -II. Intel was so keen about that technology, that they snatched all the engineers from the project right after the USSR was history. This very same engineers developed Pentium III later, which technology is used until today in all the Core CPUs.
My first desktop was a 33MHz 386 clocked at 40MHz. When I discovered that I felt a little betrayed (this is not a genuine 40MHz cpu!) and that was the last pre-build I ever bought.
Always sourced my own components ever since and typically slightly underclock/undervolt my latest hardware to not be blown away by the cooling. If you grew up with 33MHz, 3.8 or 4.0GHz both are exceptionally fast.
My story was very similar.
Maybe adding a small heatsink to the CPU or chipset might help abit when overclocking?
@11:35 Pro (?) tip for soldering large SMDs: use a strip of kapton tape to hold the component in place, it makes it much easier to align the pins. Once the component is properly aligned, tack down (solder) a couple of pins on opposite corners, then gently remove the kapton tape and solder the rest of the pins.
Thanks for the tip. I usually just hold it in place using fingers, it's not too tricky, if you don't have to film the procedure. But I'll keep your tip in mind should I have problems with my usual approach.
Back in the day when these weren't old, I did have a 386sx system I overclocked from 16MHz to 33MHz successfully. Since the motherboard is made for 286s and the role of the 386sx was to be a drop-in replacement of the 286, the chipsets were often the limitation and that's also why yours is specced at 24MHz.
Yes, this is my guess as well. Thank you for sharing your opinion.
But 24 Mhz wasn't exactly a standard speed for 286es either, although 20 was. So I bet they designed for 20 MHz top speed, then validated and accepted whatever they got as it was better than 20, and this turned out to be 24. It's probably worth trying to find a 48 MHz crystal.
Rose Gold/Low Melt Solder is a blessing.
Ngl, This was refreshing compared to how easy it is these days
What a trip down memory lane, especially with Doom...
I had a 40MHz 386DX board, and Doom (under OS/2) just ran fine. It got a bit laggy when a lot of enemies got into the same room as you, and a lot of shooting and shouting ensued.
I feel the same, that's why I'm doing this videos in the first place :D Glad you liked it. And in regards of Doom, I think our memory cheats on us way too often. Sure everybody has his own view on where a game is considered to be playable or not. At least here you can find some real numbers and measurements made by Phil are in-line with my own experience: ruclips.net/video/qQEHHc1q06c/видео.html
i once overclocked a 25 mhz 486 sx to 50 mhz using a cooling fan and heat sink
I have one just like it, but it's an AMD DX-40.
Remembers me on my first PC - Pentium 200 MMX.
I had that system running stable with 250MHz, a speed increase that was indeed noticeable in games back then.
And for a long time the only system I could overclock for 25% it's original speed...
After a long time I had a 3GHz cpu that I could only overclock about 200MHz, I chose not to do it because the speed increase was not significantly enough.
Many years later I had a 3,4GHz cpu that could be overclocked automatically to 4,3GHz, that was worth it and it ran absolutely fine.
I had a 386DX-40Mhz - wish I had never sold it - worth a fortune now!
I still partially have my very first PC. At least the internals are near my desk under a glass dome.
11:07 - hah I had a DX-40 MHz board that would take an external CPU, and I wanted to put in an SXL (TI) - I played with all the jumpers, it would just not run with any other CPU I had in the PGA socket.
Finally I took a heatgun to the CPU on board (I needed one for my display collection anyhow) and it came right off. I didn't take any special precautions, luckily it was in the very corner of the board. I cleaned up any pads that may have cross-soldered in the removal, and plopped in the TI "486" CPU.
Booted right up. :D
This was the exact mainboard i had on my first PC, i replaced the crystal with a 40hz one and added the math co-processor. Doom ran quite a bit better after that. Excellent video!
You mean you had 1,1 FPS in Doom instead of 0,9? :D
@@necro_ware i want to say it was at least playable but that was a long time ago. love your channel.
@@pauldame9925 Time often plays tricks on our memory. I think you mean Wolfenstein 3D, that was kind of playable at 20MHz 386sx, but for Doom you really need at least a 386DX-40 to run it in a very small window in low details and to have it playable you need at least a 486DX-40, or even better a 486DX2-66
@@necro_ware That could be...time has taken it's toll
That's one fast board right there, be really good for the latest titles.
Brings back memories of a 386sx my brother got me from work a clear out in the 90's, many nights were spent overclocking it just so i could play DooM on it. I think it was a 33mhz cpu, but there were dip switches and jumpers on the board to adjust clock speed. it just about manged to run it at low graphic setting full screen and high settings at half screen size. Wish i still had it. thats where my love for overclocking and using old tech comes from to this day.
would like to see how far you could push a 486, or if one could run on the 386 board ? its great seeing the real time effects of the clock increase on games like DooM... from slide show to playable.
Yeah, in the time of 386, Wolfenstein 3D was more a game to play. First 386DX-40 could run Doom at low settings. Unfortunately even that was too much. As soon as more enemies came to get you, your FPS dropped dramatically to unplayable. It was first more or less playable with 486@40MHz, because of 8K (!) L1 cache :)
Usual 486 CPU was not compatible to 386 mainboards, but Cyrix did some magic and brought own 386 CPU with 1KB cache and 486 instructions. It was still slower, than a real 486, but could help to push your 386 a bit. Good news, I have all parts for this recipe and wanted to do a video about it soon :)
@@necro_ware does having the Math Co-processor help much with these 386 - 486 cpus? Would there be any benefit in overclocking those as well ? Look forward to seeing the Cyrix! I'm sure I had a Cyrix "586" at one stage in the 2000s that was fun to mess with. Subbed! Glad you popped up in my recomended
@@austinmaxi Glad you've subbed :) Regarding math co-processors, it depends on what you are interested in. They didn't help in that days for games, because math co-processors were for acceleration of floating point arithmetic and most games didn't use floating points at all. Only professional applications, like simulation or CAD did benefit of it. The floating point arithmetic came first with 3D polygonal games and one of the first famous games, which started to rely on a math co-processor was Quake. But it is the era of Pentium CPUs already and 386/486 are sitting nervous in the corner :)
It is also not easy to benchmark games from the time before Pentium, because all the integrated benchmarks in the games started to appear first as overclocking became a "sport" or "challenge", as I told in the video. Unfortunately, with the old games you can just say: "Yes, it feels faster", not very exciting, yes? Well, this is one of the reasons, why I just used some synthetic tests in this video, at least you get some numbers to compare. However, I'll try to do better next time :)
Like people already commented, when you overclock 386, you also overclock its FSB (1x20MHz, 1x25MHz, etc), so the limit is always the core logic and other not-so-OC-friendly motherboard components. Also, I've seen 48.000 oscillators on some period ISA video cards at least, so there is a real chance that some of these boards were indeed 24MHz (O_o), with downclocked 386SX-25 (?). Overall great video! Really enjoyed your soldering and test results! I'm playing with the later SX-40 now, and honestly, Doom is still unplayable on it... But.. there's always DX-40.. ;) Keep those videos coming! ;)
Yes, everything's true, what you are saying. One thing is, that Doom is not playable on 386DX-40 as well :D Only in small window with low quality you will get near some playable values. This game was just made for 486. It's like Wolfenstein3D is for 386, Doom is for 486 and Quake is for Pentium.
Aah yes, my first encounter was accidently putting the jumper on my Pentium 133Mhz to 150Mhz... I was amazed it worked 😁
Thumbs up for your soldering.
I also once got a by-fault-factory-overclocked 3dfx card that kicked ass but overheated... being an idiot at the time I had it replaced under warranty... should have just added some cooling and let it kick ass. The new card I got naturally didn't perform as good.
Whoaaa I remember replacing those oscillators myself in the 90s. I was a teenager and those 386 SX boards were cheap to get, yet they were able to run 32 bit software. Because of that, they take a special place in my heart. I did all kinds of experiments. Would like to somehow see Doom running playable on a SX, which is a really hard task I guess.
Yes, those 386sx were more or less 286 CPUs with 386 features. Internally they were 32 bit, but used 16 bit memory bus. That's why, 286 mainboards could be used with slight modifications by the manufacturers. This made such machines very cheap... and absolutely not capable running doom. If you are happy with 12 FPS in a window as big as a post stamp, then you need at least 386DX-40. Really playable is the game first on 486DX2.
My first computer? IBM Turbo XT @10mhz. Did the 286,386sx and dx builds, and then special ordered a 486 dx 2/66 board, cpu, and all the goodies. Conner 170mb IDE HD ("you'll never fill that in your lifetime") Got a wild hair, later and bought a 486dx 4/100 CPU, put in 100meg of ram and shut off virtual memory in windows 3.1... That thing was a screamer! ***sigh*** My brain was a sponge for all things electronic back then. Now I'm lucky if I can shut off the alarm on my smartphone.
Good work, comrade!
That was awesome! I'm currently trying to get a Octek Jaguar II 386 board back to life! It currently has a 25Mhz Intel processor in it but according to the manual I can also upgrade it to 33Mhz. However, it has a 50Mhz oscillator crystal soldered onto the motherboard. I'm curious to see if it would still give me that 33Mhz that they claim. Would be interesting! But I may need to replace that crystal with a 66Mhz.
Aside from that, the reason it doesn't seem to be working is memory issues. There are two bios chips of AMI that came with the board, one shows all 8mb of RAM but crashes when it's time to boot. The other only seems to see 1mb of the 8mb installed and that's it but it does work stable. I've taken a look and it seems the SIMMs that are installed in the board are 60ns while the manual indicates that for a stable working system it would need 70ns or 80ns and then also setting up the proper wait-states. Eitherway, looking forward to sinking my teeth into this motherboard and if I can get it to work, hopefully find a nice 386-era case to put it in! :D
Hi, I don't know if I got you wrong, but as I read it, you complain about 60ns memory. Well actually, in this case less is better, 70ns is better than 80ns and 60ns is better than 70ns. And 60ns is the fastest SIMM memory you can get as far as I know. Could be, that you have problems to get it booting with 8MB, because your memory is bad. Octek Jaguar II 386 is a 386DX mainboard, it has 8 memory slots and you have to use it in packs of four sticks. If you have 8 memory sticks with 1MB each, you can try to remove 4 and try to boot with only 4MB, if that doesn't work replace all the sticks by another four. If that also doesn't work, take two from the first pack and two from the second. If you go through all of the memory sticks in this way and only up to 4 of them are bad, you will end up with 4MB of working RAM. For this machine it is mostly enough anyway.
you deserve more subscribers m8, very interesting video
Thanks to you the number got already better :)
A perfect example of an underrated channel. 😞
It's getting slowly better ;)
We did once weird stuff with an AMD 386DX-40. Friend had not enough money for an 486 but the 386DX40 were insanely cheap at those days. We found out, his sample was totally happy running with 80mhz.
Yes, in very rare cases it is possible, but you heavily overclock the ISA bus and I can imagine, that many expansion cards would get instable resulting in broken audio, video etc...
@@necro_ware the beauty of that board was: you could lower the speed of that ISA bus by a factor tied to the CPU. It took my friend days to find stable settings, where everything was performing real fast, without crashing. He exchanged even his sound card for that one of a friend, because his original soundblaster had problems with 12mhz ISA-Bus speed and the compatible soundcard was rocksolid at that speed. That dude ran a VLB setup and saw no need for overclocking his ISA-Bus. But with an ISA bus, you want to run your Tseng4000 as fast as it goes without crashing.
@@ciddax754 Yes, some rare boards allow to set CLK/6 divider for ISA BUS. This means about 13MHz, when CPU is @80MHz. Unfortunately most of the 386 boards doesn't allow to set divider at all and the ones, which do, they provide CLK/5 at best. Again, on @80 MHz CPU, means 16MHz. That would require a lot of fine tuning and hardware selection. Not even every Tseng ET4000 would run at that frequencies stable. I have one here at hand, which says goodbye already at 12 MHz ISA. It's not always an easy task to find ISA hardware which runs above the standard 8MHz and if it does, usually not for long. I overclocked one 386 a year ago to 50MHz it was stable for about 3 Months, then the graphics card got broken first. I replaced it and a month later the chipset on the board died as well.... RIP
this reminds me of clocking my 386sx25Mhz up to 40Mhz... only needed a jumper reconfigure, and a heatsink😀 good times
I know this is an older video. Anyhow - nice job! What is that metal you use to desolder? I can ´t find it anywhere. The board you have is more or less "286ish" chipset which was to save costs back then compared to 386DX. Even running a 286 at 25Mhz is a challange with most of the chipsets and unfortunately it can even kill the chipset later on. Adding pasive cooling will help. Those old UMC guys are nice but not the powerhouses 😁. For those experiments try to get later single (double) ICs integrated chipsets. Old NEAT and HEAT as far as I know don ´t OC well. You are my favorit tech RUclipsr and I learned a lot from you!!
Thank you! The metal you search is named rose's metal, or rose's alloy.
@@necro_ware 👍
I still have my aunt's old Pentium MMX 166. It came with a fake badge on it saying it was a 200 and ran clocked at 200Mhz for its entire life, kinda want to try and get a new mainboard PSU etc for it and build a DOS gaming PC because as far as I know it should still work fine.
how did you do surface mounted soldering so cleany!? i tried it and messed up. may be u should do a video on how to do it properly? i have a 286 M216 mobo with bad soldering on the chipset and it needs to be reflowed i want to know how this is done well
You are finding the DMA controller is running away from the clock speed. Probably the CPU will overclock, but is held back by the DMA, causing memory allocation errors, as it's unable to clear it's instruction code intime for the following set. I had this and got a error: Divide by zero error.
Yes, I suspect something like that as well and on this board the DMA controller is integrated into the chipset, so I can't do anything about it. Thanks for sharing your idea.
Very cool! New subscriber! I have a 386sx16 I want to do this to also! Thank you for showing the way.
really interesting experiment
I remember the days of overclocking with a pencil... My first entry into overclocking was with an AMD K6-2 500 with 3 fans tie-wrapped to the heatsink and 60mhz later with 2.81v. To this day I still overclock every system, card, and chip hah
Yeah, I used the pencil trick on my Athlon as well back then. AMD cut some wires to prevent people to setup the multiplier, but you could simply draw over the wire with a pencil and it was good to go ;D
That wasn't my first try, but I also remember trying to squeeze more speed out of a K6/2-500 and not getting very far. I figured I could at least get to 550 since there were K6/2 chips rated to that speed, but no. The other hardware on the board wasn't happy at anything above standard ISA and VLB clocks, so I had to run 5.5x and less than 100 MHz FSB, finally topping out around 530 MHz, and even that required 0.2V overvoltage and got a lot hotter. The loss of FSB, thus drive throughput, video throughput, memory throughput, etc. meant on balance it wasn't worth it.
I was under the impression the DX chips were much more overclock friendly. The SX versions needed the math co-processor for any realistic speed boost. And well, it was a long time ago. So I'm probably forgetting everything. Lol. Fun video to watch, thanks.
Well, it may vary from CPU to CPU, but in general DX are not really more overclock friendly, but they are a lot faster to begin with. Compared to SX the DX had 32-bit RAM bus instead of 16-bit and ability to have up to 128KB cache, which made DX by far a more superior CPU. SX on the other hand was just a 286 on steroids, equipped with internal 32-bit capability and 386 instructions set. It was made as a cheap solution, so the mainboard manufacturers could continue to use their slightly modified 286 mainboards with 386sx. And in regards of math co-processor, it is absolutely useless on a 386 (either DX or SX). I know only two games, which made use of it - SimCity and Falcon 3.0. All the other games didn't make any use of it. You could get some slight performance gain in Word and Excel under Windows and in CAD software though. So, if you didn't use that programs, you wouldn't see any change with or without math co-processor.
@@necro_ware thanks much for the refresher. I didn't think my household expected to run anything outside of Wordperfect and Lotus back in those days. We had a 386sx with the math co-processor, and NU's benchmarks always showed the increase. This was a great video to watch though and your information was top notch, thank you.
@@jeffchisamore1556 Thank you too ;)
At a Hamfest in 2005, I witnessed an Intel 286 overclocked at 1.2Ghz. It functioned, but needed to be cooled with liquid Nitrogen to keep it from melting.
I barely can imagine that :D It is physically not possible. The technology used in 286 was not capable to run at such frequencies, doesn't matter which voltage or cooling you used. May be you mixed something? I mean 2005 is long time ago.... ;)
@@necro_ware even with liquid helium a 286 would never reach 1.2ghz 🤣
possibly the 30pin ram limiting it. high end 30pin 486 era modules could do 66 busclock but those can be hard to find.
I thought my initial K6-2 OC memory is already old enough, couldn't even imagine a world OC is done by replacing the oscillator...
Ah, yes. Let's do this. I will use a 386sx in my next build. Sure.
Excellent video! I really liked it
man you have some serious skills!!!!
I had a Pentium 60 with the floating point division bug. I swapped the crystal to put it up to 66MHz and took it into Computer City to have them replace the CPU with a non-bugged version. Free Pentium 66 upgrade.
i attempted to overclock an amd 386 40mhz to 50mhz, it would post but not boot.
yeah 25% overclocking might be too much. You also have to watch the ISA Bus clock doing so. Some mainboards can't adapt the frequency and go to far over 8MHz, which makes ISA cards unstable...
I have a long history in overclocking. My first overclocking was Pentium 120 running at 133 MHz with cache maxed out from 256 to 512kb on board. I also had the best results with a Pentium 233MMX, running at 250MHz with 100 MHz FSB on a FIC VIA motherboard with 1MB cache, completely smoking Pentium II 233 of the time. Now I don't overclock, I prefer to buy second hand Xeon to push my motherboards to their upper limits
My original pc had a 386, but i managed to get a 486dx, or a cyrix 586 running on it. The system had jumpers for bus multiplier and voltage changes. More voltage was needed for the better performance. It was a few years ago so i cant remember the exact details i used.
I think you memory tricks on you a little bit :)
@@necro_ware Oh definitely. But I believe it was a PS2 70 designed for a 486 that had been downgraded to a 386 before I got my hands on it.
My Core I3: “Dad, Do You OK?”
When I was a kid, I remember we had a Intel Celeron @ 366MHz and I couldn't play any game at decent framerate, it was software rendering too. So I tried overclocking it as much as I could, increased FSB frequency multiplyer and so on. Needless to say, computer went dead. My dad took it to a repair shop, I was terrified and totally anxious if they will tell him what I did, I will surely get some ass beating :) But they reset BIOS, replaced CPU and RAM, and my dad never found out what I did :)
Either they didn't know that overclocking destroyed it, or they realized that it's propbably too hard to explain. Either way I've learned my lesson even without beating of my ass, the fear of it was enough :)
Not, that your dad will read it here now :D I'm really afraid of your ass now....
@@necro_ware Hopefully my dad does not know my YT username :D
But he is on YT, just not interested in tech, so my ass should be safe :D
I remember I had a PC with a Pentium at 90mhz as a kid one day I was playing around with it and noticed a switch on the motherboard to change the frequency from 90 and 120, easy +30mhz.
This is an amazing content for a noobie like me.
Well done, excellent video! Model of the portable oscilloscope used?
Thanks, Fnirsi-5012H. For simple measurements it's quite good, because it's portable and delivers a decent job. But it is advertised for up to 100MHz and it works reliably only for up to 35-40MHz. From there it starts to show wrong voltage levels, but if you just want to see some frequency or if some activity happens on a pin, it's still sufficient...
Still running a Q6600 (2.4GHz) overclocked to 3.4GHz as my HTPC. The Celeron 300A - 450MHZ was also a legend
Performance increase of 386 isn't any more linear than later CPU's. If you did tests in Doom you would see that you are limited by the speed of video card (though from numbers in Landmark I see your VGA is pretty awesome so it might not be as big of an issue, but still...). The best overclock I had in my life was cheap E2160 CPU that went from 1.6GHz to 3.2GHz and that would be comparable to if you could overclock 16MHz to 32MHz. This was through FSB overclock so raw CPU performance including its ability to talk to memory linearly increased by 100% which was reflected in pure CPU and CPU+memory benchmarks. Of course this affected memory timings and clock divider so memory had to have bigger timings but this is not CPU related and actually identical to 386... and by the way, did you try to increase memory timings after OC? There were in most BIOSes setting like turbo timigns or slow timings. Tinkering with these might improve stability with higher clocks
I used the fastest graphics card I had. The problem with Doom and ISA is the limited bandwidth, but 386 is not yet limited by ISA bandwidth. You run against this limitation first with 486DX-33 and faster CPUs. And in regards of the overclocking, I had some talk here and there and it seems to be the chipset indeed. All the mainboards I found were running CPUs with 20MHz max. I didn't find one board which would go higher then that. So the conclusion is so far really the chipset. It probably can't stand higher frequencies, than 25MHz at all. I also played with different memory and timings. No way to convince it :D
I would like to know model numbers of these main boards. Tks.
I am trying to imagine my parents reaction if I would have told them in 1993 I am planning to solder on my PC motherboard :D
i started overclocking with a Spitfire Duron 700 to almost 1,1 GHz with a homemade air-duct-cooling contraption which sucked cold air from outside in the winter too cool the CPU, i was very proud of myself - but later i was pretty sad as i discovered that there were folks out there that cranked up the clock to 1,5 GHz for those CPUs :)
One of my buddies discovered these coolers (beer type cooler) that you could plug into your lighter socket and it would get cold or warm depending on how the power was hooked up. They used a peltier chip that was about the same size as a 486, so several of us used them for CPU cooling. We were mainly messing around with Cyrix chips, which got really hot and flaky when overclocked, and it worked like a charm.
An arbitrary waveform generator should be the right tool to generate whatever frequency you want just like the modern PC
Not every arbitary waveform generator can generate 66 MHz or more :D The clock generator as I used in the video is a good and very cheap way to go.
Nice video! :)