And this is why I sent you that CPU. :) I knew that I wouldn't be able to fix it myself, and I should leave it to the expert. Like I wrote before, it was exactly like that when I got it, including the missing pins. I didn't attempt to straighten any myself because they're so weak. I'm happy that it lives again.
@@dim0n1 It helps caching a larger amount of main memory, which is/was at the time primarily useful in servers. In a day-to-day retro-gaming-PC scenario, likely slim benefits to none.
@@tomteiter7192absolutely, released in end of 1995! The central Unix system shared by lots of students had 64Mb of RAM and was considered huge. Regular PCs would've had on the order of 4 to 8Mb I'd think.
@@madmax2069 It often amazes me how back in the day people had such an imbalanced setups. I had K6-2 350@500Mhz and Banshee, and that was bad enough, but voodoo3 needed at least decently clocked Pentium 3 or early Athlon to stretch its legs. The poor thing must have been so bored on that MMX 200.
While it was an amazing feat of engineering, the 1M parts were obsolete on launch. They were no more than 10% faster than the older 256K and 5% faster than the 512K variants. They were also released so late that the Pentium II was already in the mid 300 MHz range, and were much faster for half the cost. Unless you were already invested in Socket 8 hardware, it made no sense to buy them.
@bitsundbolts A request: *please* use Windows NT for Pentium Pro benchmarking. PPro is pretty bad at 16-bit code, so it runs poorly under DOS or Win98.
You'd be disappointed, they're no more than 10% faster than the original 256K variants. The Pentium Pro wasn't an architecture that heavily relied on cache for performance, unlike the much later Pentium 4. I have both 256K and 1M versions of the Pentium Pro 200 and did basic benchmarking on them. They're also VERY bad in 16 bit code and mixed 16/32 bit code because of a bug in the core design. Whenever the Pentium Pro wants to execute a 16 bit instruction, it has to flush the entire instruction pipeline every time. Back in the day, the only people that these 1M parts would appeal to are people that are already heavily invested in Socket 8 hardware. These chips were released when the Pentium II was in the mid 300 MHz range, and those chips were half the cost and netted far better performance. It made no economic sense to buy these, they were obsolete when they were launched.
Straightening tiny pins like this can be done using a mechanical/propelling pencil tip/tube if you remove the pencil lead. Gives support all along the pin and doesn't need tiny pliers.
I have tried this method and I agree that it works. Unfortunately, my hands and the pencil will block the view when I work on the pins. I also believe that all the pins that broke off would also break with a mechanical pencil. Those pins are the most fragile pins I've worked on so far.
Having done a few, they're done for when bent at the base. The material stress is too high. A mechanical pencil works only so well. Physics has claimed those pins, time for donors.
@@bitsundbolts Yeah if you are trying to film under a microscope not ideal for sure and there's no saving pins that are visibly damaged at the base are done for regardless of method as you say.
Half a month ago, I saw a small laser device for repairing gold necklaces at an exhibition in Shenzhen. My first reaction was that if it could be used to repair CPUs and sockets,
@@weedmanwestvancouverbc9266 That machine is laser welding, the welding material is gold wire, there is a laser indicator positioning, there is an LCD display to view the welding process, the welding speed is about 0.5 seconds, I don't know if it can be used to repair the motherboard CPU socket
@@weedmanwestvancouverbc9266 I know that argon arc welding and laser can weld smaller things. The manufacturer demonstrated on-site welding to repair the smallest chain of a gold necklace.
I had a couple of dual PPro 200 512kb systems I built for where I worked at back in the day, but never saw a 1mb cache version. One of them was used until around 2005 due to lack of a budget to replace them, and at 233mhz and 2 cpus it was still decent for office productivity at the time.
Glad to see you got it running again. I had a similar experience when reviving one of my 1M PPros, they have significantly weaker pins than the ceramic 256k/512k models. I think these are machined from a rod of brass, while the others are gold plated copper.
Absolutely agree! Those pins on the black-top model are very fragile. And seems like those pins are one-way bend only. You can't bend most of them back - they'll all break.
I remember I was ham fisted in getting a CPU socket, and I used one of those mechanical pencils that use the bare legs that were like 5 mm and they were a perfect fit to straighten the pins
An empty mechanical pencil tip will do wonders to straighten these CPU pins, you should try it. Put the pin inside up to the first bend then straighten it enough to push it in to the next bend and straighten again, repeat until you get the whole pin inside and can straighten it all.
Excellent repair on that CPU! The Pentium Pro has a special place in my heart as it was a CPU I had always wanted to own but never did back then due to cost. I did have a friend who had a Pentium Pro in a Gateway 2000 PC that had that same mainboard in it. He had acquired that machine through less than savory ways though and I lost touch with him back in 1999 after he moved out to the west coast of the US.
Sweet. You would not believe how many of these things labored away in quad or dual CPU boards, well into the P3 era. These were certainly Server CPUs. The replacement for these was the SLOT 2 Pentium II Xeon.
I always use a mechnical pencil (the ones you use in school for geometric drawing, i live in austria so please don't ask what they're called in english, we call them Druckbleistift), the tip of those pens fit perfectly onto the pins and the pencil gives you very good yet controlled leverage on the pins (it also fits perfectly between them so you don't nudge another pin accidentially). The pencil method never failed me on bent pins.
Great job on the repair and BIOS upgrade! I was surprised how weak those pins were, but I'm glad you could get this CPU working again. I had a Pentium Pro for a while, it was a 180MHz model (I believe 256K cache), which I immediately overclocked to 200MHz since it came with a big heatsink (much larger than the one on yours). I even bodged a larger fan on, which dropped the temperature a little, but it ran stably for a few years, I never had any problems. Looking back, I'm not sure how much of a real upgrade it was, since I came from a Pentium MMX 166MHz. It's possible those MMX instructions may have made it faster than the Pentium Pro in some games.
Those heatsinks came with those boards. I'm not sure what they are, but they seem to be just good enough to cool the CPU. A better cooler would be great, but I've heard those coolers are rare these days.
Hearing your vocal expressions of joy upon success never gets old. 😄 No idea why you were apprehensive of this repair, it literally went exactly as I expected. But then again, I'm an optimist, haha. Those pins cannot be pure gold, they behave almost like an aluminium alloy. Looks like Intel was trying to reduce costs on those chips, as the huge cache probably made them expensive enough already. On a side note, Knipex are of course the best (not just the flat jaw ones) and these pins were probably all beyond saving so it didn't make a difference here, but generally speaking I believe the toothpick method may be more gentle and yield better results in some cases (esp. pins on pads), as it allows you to apply counter leverage and thereby better focus on the area that force needs to be applied to. Anyway, it's always heartwarming to see another historic piece of computing tech back in action, I'm really looking forward to seeing more content about it in the future! On another tangent, I'd like to rant about the fact that BIOS flashing can STILL be a super irritating process, specifically if it is necessary to flash a board so that it actually supports the CPU that you bought for it (e.g. a B650 board so that it supports a Ryzen 9000 series CPU). At least these days, you can usually flash it without obtaining and installing a supported older CPU first, but the process is usually not sufficiently well documented and often fails without apparent reason. I pity the person that doesn't have a bag of old USB Sticks to try until one is found that works. End of rant! 😉
Good point regarding cost cutting from Intel. These pins definitely are a lot weaker than on any other CPU I worked on. Even other PPGA models like Pentium MMX and Celeron Mendocino have better pins even though they have the same type of body.
I did not expect the detached pins to flow that well! I expected that the tweezer would steal just enough of the heat to make it incredibly challenging, obviously this was not the case 😅 well done!
The PPro 1M are also stable at 3.5 x 66MHz = 233MHz (if your board supports 3.5x). I have two of these PPro 1M running on a Tyan S1662 board at 233MHz. :)
Perfekt! Geduld zahlt sich aus! Leider habe ich keinen 1MB Ppro in meiner Sammlung aber immerhin einen Slot2 Xeon. Ebenso fehlt ein PentiumII-Overdrive für Sockel8. Wenn die Pins so dünn und empfindlich sind, probier doch mal die Dinger punktuell anzuwärmen, damit sich das Metall etwas entspannt und nicht bricht. Jetzt auf jeden Fall Benches mit Pentium 233MMX gegen PPro in allen Cachestufen und P2-233.
Ja, da muss auf jeden Fall sowas wie ein Benchmark Marathon kommen. Ich habe ja alleine schon 3 verschiedene Cache Ausbaustufen: 256, 512, 1024 beim PPro. Und dann vielleicht Pentium und Pentium MMX.
Fantastic job! I'd love to see you repair pins with a mechanical pencil (it must have a metal tip). I've been fixing pins that way since high school 25 years ago.
I always wanted a dual CPU computer in the day, finally got one in the socket1366. One motherboard CPU combo I would love to see is a dual socket 7 with a pare of K6-3 CPU.
Impressive stuff. I basically just write-off CPUs with any kind of deranged pins, the amount of frustration from such delicate work isn't worth it. Really cool to see what successful CPU repair _can_ look like, lol. I'll never experience the joy myself, but can appreciate seeing others triumph.
Back in the day when worked in the computer shop customers would do this time to time when upgrading their cpus. As a last ditch effort to fix it for them we’d stick a appropriate gauge wire into the CPU socket holes where there were missing pins and added a small curl to the top of the wire to give it a bit of spring tension. If you do it right it works about 80% of the time as long as the original pad is left on the CPU. Have you tried watch makers tools? They are kinda a godsend for this kind of work.
I used to work for a Geek Squad warranty fulfillment center during the days of the Athlon 64. Our parts department was so careless, they would send CPUs to the repair line in loose plastic bags with no protective foam and no anti-static bag. They were probably stored that way in a bin I guess. By the time they made it to the repair line, the pins were all bent to hell and I really had no option except to waste 20 minutes straightening them out with a pocket knife. And I had to replace a lot of those Athlons, as I recall they had some problem with the onboard memory controller. Lucky I never had to re-attach any pins like this!
Really awesome. I always loved the Pentium Pro, I found it such an awesome CPU especially using multiple sockets. truth be told I have never seen the 1MB version IRL.
Ha, that would be a nice find at the scrapyard - a dual PPro board! I do have two matching ceramic PPros that would probably work as well! This black-top 1MB model is really nice. More videos to come in the future 😄
36:58 It is highly likely that the CPU threw an exception, without there being a handler for it. I would run it with EMM386 or something similar loaded into memory. My guess is that DOS will then bail out with exception 6 or something similar, since it is, technically, trying to execute an invalid instruction.
An invalid opcode 06, or general protection 0D, run these in real mode. And patch in an exception handler. Lucky you still have interrupt enabled STI then youll need cold boot if interrupt off CLI
I use syringe needles for straightening pins - it is much more convenient than a toothpick: - find the needle of the right size (one that just slips over the pin); - clip off the tip (do not mash it); - put the needle over the pin and bend it back. For thick pins of older CPUs a mechanical pencil can be used instead of needle.
In the past, I've found the best tool for bent pins is a mechanical pencil, and to make sure the pins have been warmed up first to minimize cracking. Not too hot to melt the mechanical pencil, but like, hair dryer warm.
I'd be interested to see performance comparisons between the PPro and a Pentium(/MMX?) PPro was garbage at running 16-bit code, just how much of a difference is actually there? Running a pure 32-bit environment (say, Unreal on WinNT), what performance difference is there?
You might know this, but the 1mb PPro was not flat and needed a special heatsink to match. I sanded one down years ago but didn't have a way at the time to check any difference in temperature.
Do you ever "load setup defaults" after flashing a new BIOS? I remember that being a recommended procedure due to possible differences in the layout of saved user settings between BIOS revisions.
I did indeed once faced a lot of issues and loading Setup Default fixed the issue. I did not do it for this video I believe. But yes, this should be done based on my experience.
Thank you! Yes, that is what I heard as well. Socket 8 coolers seem to be rare. I was lucky that I found three boards with coolers - all the same as shown in this video. Even though they're no high performance models, they get the job done. The best option is probably a custom made solution. Maybe something exists already
Thank you for saving this historic CPU. One reason there are so few Pentium Pros left is because each one contains about 0.3 grams, or $25 worth of gold!
Great save! The 1MB are totally different beast compared to the ceramic variants, not sure why they changed the package but they look a lot cooler =). And the PPro overdrives are worse still when it comes to the pins, very strong and rock hard but breaks even easier then the black tops. Regarding to the heat output from these "beasts" I suspect Intel was "creative" with the power numbers cause there's no way the 1MB chips are just 9w more then the 256Kb variants. I reckon the 1MB pulls close to double the power in real use! Also I'd recommend using a thin thermal pad (I use 0.5mm) instead of paste, the black aluminium top might look cool but its certainly not flat enough for paste to work properly. And beware the heatsink in the CPU power supply on the board, it gets incredibly hot when running theese chips. Looking forward to some benches, don't forget to enable write combine! =)
This is probably the most iconic and interesting CPU I have now. Followed by an IBM blue lightning 386/486 75 MHz (most likely) and a 66 MHz 486 SX CPU 🤣
You could also the mechanical pencil trick, just remove the graphite and slide the pencil over the pin and you can easily use it to straighten the pins. You managed it though, wonderful work.
@@bitsundbolts I've a question. I got an Asus P2B motherboard for retro gaming. I went to turn it on by shorting the pins. BUT. By accident, I hit a 3rd pin. Got a spark & the board got messed up. I let it sit for a bit & when I tried to turn it back on. I get power to the board, Hard drive & CD Rom. But, I have no video (AGP card) & the cpu fan isn't working. Is there a chance to fix it or is it a lost cause ? Im so upset with my self, Wanted this board for ever. It was rock solid untill I messed it up.
I tried this on 486 CPUs. Unfortunately, heating pins doesn't help much - if at all. Those metal pins would require a lot more heat which most likely will damage the plastic body of the CPU.
That P3 is actually quite likely to still work even with the damage, as this is a single layer chip (like nearly all modern silicon, as far as I know multi-layer is still very experimental), so only the bottom few nanometers are actually relevant, the rest is just a piece of glass. So if you can fix those pins (maybe with the help of uv solder mask to prevent solder from connecting to the wrong places), there is a good chance it still lives. I have actually seen worse chips still work, including some fairly modern gpu (3060 I think), which is still completely fine despite the damage.
I tested all those chips before I put them in a box for scrap pins. I'm afraid, this chip is dead. Some have visible cracks through the entire die. But you're right, from chipped corners, you can't tell that a CPU is dead. I have many CPUs with chipped corners that work perfectly.
@@bitsundbolts That's unfortunate, but it is vintage hardware, so not entirely surprising. Interestingly, I remember the P3s to be fairly reliable (unlike xp-era AMDs and celerons, and more recently the ryzen 3600), but shorted voltage regulators and bad ATX supplies weren't rare back then, so that might have been it. But whatever broke it, at least its parts help you save other chips, so I guess that's the best possible ending for a dead CPU.
ppro in ASCI Red was first supercomputer to hit over 1TFLOP. it was originally 200MHz 256k but later upgraded to specially packaged ppro II overdrive 333MHz 512k. now some pocket computers can do over 1TFLOP running from batteries.
The toothpick method puts all the force on a single weak point which is why it's so destructive. If you can get the pliers to grip the pin it can provide some bracing and support as well as spreading the load.
Quick suggestion that worked for me: when the crack is just between the pin and the base (like two of yours), I put some solder on it FIRST to create a solid base and then straight it up in second place, so the weak base do not break when you straighten the pin, and it will be already fixed. It's just a suggestion that worked for me...
Interesting. That would definitely work if the pin isn't too far bent so it would bridge with the solder. I might have been able to save some of the original pins that way. But I'm still happy that the PIII pins are a perfect match as well.
@@bitsundbolts You are correct. Happy that you were able to save the processor !!! Those used to cost an arm and a leg in 1997-98. Never seen one in person. Hope to see it back in another video ;-)
5:50 On that 850, the pins have small heads so you should be able to solder them back on, it's just that you'll have to carefully soldermask that top copper layer. See if that top layer is connected to a GND pin, or just scratch off some soldermask, then test continuity to it after soldering. Maybe even sand it back further for more isolation/clearance.
I am sure it's possible to add the pins, but chances are low that the CPU works. I find these CPUs a lot at the scrapyard. It would be nothing more than a challenge to fix this CPU.
I think this is always the case. Reading seems to be limited by the memory speed. There is nothing available in the cache. Writing and moving seems to benefit from the cache
0xC3 will only work under the C runtime environment, for a DOS COM executable you need to use: B8 00 4C CD 21 To return to DOS. The entire program should therefor be: F0 0F C7 C8 B8 00 4C CD 21 Which means: lock cmpxchg8b eax mov ax, 0x4C00 int 0x21 0x4C00 being the exit program with return code function and int 0x21 to call the given DOS function.
at first i really liked that heatsink+fan combo but after you showed the thermal camera footage it's obvious that it kinda sucks. it seems like the airflow doesn't get passed over the edge fins.
i've had good luck fixing pins with a mechanical pencil and using it as a pin straightener. Going slowly and gently bending pins and straightening them slowly.
Yeah I'm not sure. It's been forever since I coded in x86 assembly for DOS, but you might need to do some kind of call to return control to DOS. It's clearly not hung if you can Ctrl+Alt+Del.
Just an idea - wouldn't it be easier if you heat up the pin before attempting to bend it? I think this way the material will be softer and more flexible and perhaps not break? Otherwise amazing video as always!!!! :-)
Unfortunately, extra heat doesn't help. Those pins would require a lot more heat to get softer. I tried that on a 486 CPU, but that might also be a different material there.
Now that's what I call a save! I think the BIOS could have been flashed normally by booting from the floppy. The jumper being called BIOS Recovery leads me to believe it's only meant as a safeguard in case the board doesn't boot normally due to BIOS corruption. In most benchmarks, the Pro 233 completely crushes the 233 MMX and it's very close in performance to the Pentium ii 233. If only I had a Pentium Pro 200 instead of a sad Pentium 166 without MMX when I was young...
I'll try to flash my other two boards with the jumper in place. I think I read somewhere that I should put the BIOS in recovery mode. Maybe it was written in the manual.
8:46 (ish) - I remember reading on some forum somewhere, someone saying they had very good luck un-bending CPU pins by using a mechanical pencil with the lead removed. I suspect that might be too small for pins this size, though.
An interesting note about the pentium pro line of CPU's is that they are true 32 bit but with a 16 bit translation layer in a time where consumer cpus and windows itself were 16 bit. This CPU should work better with Windows NT or Unix/Linux.
I always thought the best tool for this job was a mechanical pencil with no pencil lead in it - it's a tube that supprts the pin fairly well as you bend it back. Also I'd think getting the pins hot would help them bend more than break?
Yes, a mechanical pencil does work. Unfortunately, it would block the view white a bit when I'm working under the microscope. That is why I decided to use the pliers. They also work very well! I tried heating pins on a 486, unfortunately, I never noticed a difference when heating up those pins. It just makes it harder to work on the CPU because it also gets hot.
tl;dr - I think the jumper you moved to flash was actually for BIOS recovery mode - no harm done but it skips any verification and is why you didn't see anything during the flash. If you had just normally booted from the floppy w/o moving that jumper, I think you'd get a more normal experience of it loading a flash util and you see progress on screen? Assuming this board is like how many later Intel reference boards are the jumper you moved to flash was actually the "bios recovery mode" which I think was intended to be used if you accidentally bricked the mobo (assuming the BIOS boot block still lived). I know of it because with at least with Intel socket 7 and later Slot 1 boards you could also use that to "cross-flash" between various OEM "flavors" of firmware - for instance if you wanted to change from the Intel branded to Gateway branded bios. Sometimes OEM versions (like Gateway) would fix things like sleep or resume that the Intel "original" didn't work right. Note that since it seems to skip verification yes you can also blind flash totally wrong firmware too which can be entertaining - I think as long as the boot block still lives you can recover from that also, but obviously not recommending you test that unless you have an eeprom programmer handy =)
Hm, interesting. I didn't try to flash the BIOS without changing the jumpers. I think I read this procedure in the manual. I really don't want to experiment with those BIOSes - they're those flat BIOS chips that would be a pain to remove and flash. Good that it worked like this too.
Pentium Pro, the little architecture that could. It may have been criticised at the time it was first released (mostly because poeple didn't understand the target audience of the cpu) but all modern Intel CPUs still fit comfortably in the p6 Pentium Pro family. 30 years for a cpu architecture is very impressive. I do wonder, if AMD gets a few more Punches in, will Intel have to try something new again. I wouldn't trust soldering broken pins like that if the socket wasn't zif. The force needed to insert and remove a cpu from a non zif socket would surely break the solder. The really bad thing about the f00f bug is that it could take down protected mode operating systems by completely crashing the cpu and it couldn't respond to interrupts like the Ctrl-Alt-Del interrupt from the keyboard controller. On the PPro the instruction sequence no longer stops processing of interrupts. It just generaates an ivalid instruction execption as intended and crash the thread of execution. In Dos it just hung because dos doesn't have a useful default exception handler for the crash foof causes. Ctrl-c might even get back to the dos prompt from this hung state
I think (just a guess) that silver-containing solder wire (Sn62Pb36Ag2) is slightly better in such situations. low soldering temperature, greater strength. Anyway, excellent work!
Don't let those chipped corners fool you. I have at least one AMD chip from the Athlon XP days that has a corner chipped way worse, but still works perfectly fine. I don't think they put transistors all the way to the edge of the die. If only to allow for some tolerances when cutting the wafer.
Yes, I have plenty of chipped silicone dies as well that work perfectly. I test each CPU before it goes for a donor piece. Unfortunately, this one really didn't work. These days, I only take CPUs that are still installed in boards with a cooler. Those always worked for me and are in very good to perfect condition
I just had an idea. I wonder if soldering tip made out aluminium would help with repairs like this. It would be shaped like a normal conical tip but with the tip shaved off and hole for the pin drilled. It should also have very litle thermal mass. It would be used by placing replacement pin inside the hole and adding solder paste to the CPU. After putting both parts together the paste should melt and create a joint. Then the iron should be cooled down as fast as possible (to avoid CPU damage) while still holding parts together. The fact that the tip is made out of aluminium should make it impossible for solder to wet the tip and this way making it possible to be removed while cold. Small thermal mass would help with cooling it down quickly with freeze spray. If the joint is not satisfactory it could be then reflowed with extra flux with regular iron or hot air. I need to get my hands on some scrap PGA chip and try it.
It sounds like an interesting project to try! Please let me know if you get to it. I am just a bit worried that somehow the solder would still climb on the pin up where the soldering iron is. Capillary effect it is called I believe.
If the CPU does not have MMX at all, then the F00F instruction may generate an "undefined opcode" exception that might very well be the softlock you experience here. I guess you can try to safeguard even more about this by adding more code to catch that exception, but it's kind of a corner case.
I think you recommended to add the return instruction after the fourth byte. It makes sense, but I was still surprised that the CPU got stuck - not in a deadlock though as with the foof bug on the P5 architecture. Thanks for the help btw!
![[LIVE] : ONE ลุมพินี 88 | คู่เอก "ป้อมเพชร vs อัสลามจอน"](http://i.ytimg.com/vi/1ZqTVVbMgbo/mqdefault.jpg)
And this is why I sent you that CPU. :) I knew that I wouldn't be able to fix it myself, and I should leave it to the expert. Like I wrote before, it was exactly like that when I got it, including the missing pins. I didn't attempt to straighten any myself because they're so weak. I'm happy that it lives again.
Sorry that it took so long to make that video. Thank you for the CPU - I am sure it will appear in more videos!
as bits und bolts sad at the end of video, is there any benefit to have a 1m cache version or its just a collection thing please?
@@dim0n1 It helps caching a larger amount of main memory, which is/was at the time primarily useful in servers. In a day-to-day retro-gaming-PC scenario, likely slim benefits to none.
@@dim0n1 useful or not 1M Cache "on package" was insane at the time
@@tomteiter7192absolutely, released in end of 1995! The central Unix system shared by lots of students had 64Mb of RAM and was considered huge.
Regular PCs would've had on the order of 4 to 8Mb I'd think.
I ran a PPro 200Mhz for WAY longer than I ever should have up until 2003 on a windows 2000 box w/ 3DFx Voodoo 2. This video warmed my heart.
That's gangster man.
That's about how long I ran my P200 mmx (non pro) with a voodoo 3 2000, but I never went higher than 98SE.
@@madmax2069 I had 486/160 running NT4.0 :-)
@@madmax2069 It often amazes me how back in the day people had such an imbalanced setups. I had K6-2 350@500Mhz and Banshee, and that was bad enough, but voodoo3 needed at least decently clocked Pentium 3 or early Athlon to stretch its legs. The poor thing must have been so bored on that MMX 200.
Well, it was an insane monster setup at the time. Hard to let it go, I imagine :D
Now that is a CPU.
1MB was an insane amount of on package cache for the time as well, so expensive.
While it was an amazing feat of engineering, the 1M parts were obsolete on launch. They were no more than 10% faster than the older 256K and 5% faster than the 512K variants. They were also released so late that the Pentium II was already in the mid 300 MHz range, and were much faster for half the cost. Unless you were already invested in Socket 8 hardware, it made no sense to buy them.
It will be interesting to see some benchmarks with this CPU :)
@bitsundbolts A request: *please* use Windows NT for Pentium Pro benchmarking. PPro is pretty bad at 16-bit code, so it runs poorly under DOS or Win98.
it would be interesting comparing it to a regular Pentium (P54) without 3D acceleration in Quake (software rendering)
Very
You'd be disappointed, they're no more than 10% faster than the original 256K variants. The Pentium Pro wasn't an architecture that heavily relied on cache for performance, unlike the much later Pentium 4. I have both 256K and 1M versions of the Pentium Pro 200 and did basic benchmarking on them.
They're also VERY bad in 16 bit code and mixed 16/32 bit code because of a bug in the core design. Whenever the Pentium Pro wants to execute a 16 bit instruction, it has to flush the entire instruction pipeline every time.
Back in the day, the only people that these 1M parts would appeal to are people that are already heavily invested in Socket 8 hardware. These chips were released when the Pentium II was in the mid 300 MHz range, and those chips were half the cost and netted far better performance. It made no economic sense to buy these, they were obsolete when they were launched.
What a ride. Thanks for making this video and having the courage to fix those pins. Your patience is legendary!
Thanks!
Straightening tiny pins like this can be done using a mechanical/propelling pencil tip/tube if you remove the pencil lead. Gives support all along the pin and doesn't need tiny pliers.
.7mm .5mm or .3mm lead size?
I have tried this method and I agree that it works. Unfortunately, my hands and the pencil will block the view when I work on the pins. I also believe that all the pins that broke off would also break with a mechanical pencil. Those pins are the most fragile pins I've worked on so far.
Having done a few, they're done for when bent at the base. The material stress is too high. A mechanical pencil works only so well. Physics has claimed those pins, time for donors.
@@bitsundbolts Yeah if you are trying to film under a microscope not ideal for sure and there's no saving pins that are visibly damaged at the base are done for regardless of method as you say.
Nothing better than starting my Friday morning with some BuB CPU repair action!
Wow! Getting the solder to just fill in the crack seemed amazing enough, actually transplanting full pins was awesome!
Thanks! I was surprised that it worked so well!
Definitely above my pay grade.
love the music during that one pin bending attempt
Half a month ago, I saw a small laser device for repairing gold necklaces at an exhibition in Shenzhen. My first reaction was that if it could be used to repair CPUs and sockets,
You can get incredibly small tig welders meant for Jewelers to tax stuff together so that they can flow gold or silver solder on them
@@weedmanwestvancouverbc9266 That machine is laser welding, the welding material is gold wire, there is a laser indicator positioning, there is an LCD display to view the welding process, the welding speed is about 0.5 seconds, I don't know if it can be used to repair the motherboard CPU socket
@@weedmanwestvancouverbc9266 I know that argon arc welding and laser can weld smaller things. The manufacturer demonstrated on-site welding to repair the smallest chain of a gold necklace.
I had a couple of dual PPro 200 512kb systems I built for where I worked at back in the day, but never saw a 1mb cache version. One of them was used until around 2005 due to lack of a budget to replace them, and at 233mhz and 2 cpus it was still decent for office productivity at the time.
Glad to see you got it running again. I had a similar experience when reviving one of my 1M PPros, they have significantly weaker pins than the ceramic 256k/512k models. I think these are machined from a rod of brass, while the others are gold plated copper.
Absolutely agree! Those pins on the black-top model are very fragile. And seems like those pins are one-way bend only. You can't bend most of them back - they'll all break.
I remember I was ham fisted in getting a CPU socket, and I used one of those mechanical pencils that use the bare legs that were like 5 mm and they were a perfect fit to straighten the pins
I've had a few 1M Pentium Pro's, always ran them on dual socket boards. Fun stuff back in the day.
Happy days! Nice to see it running again.
An empty mechanical pencil tip will do wonders to straighten these CPU pins, you should try it. Put the pin inside up to the first bend then straighten it enough to push it in to the next bend and straighten again, repeat until you get the whole pin inside and can straighten it all.
Excellent repair on that CPU! The Pentium Pro has a special place in my heart as it was a CPU I had always wanted to own but never did back then due to cost.
I did have a friend who had a Pentium Pro in a Gateway 2000 PC that had that same mainboard in it. He had acquired that machine through less than savory ways though and I lost touch with him back in 1999 after he moved out to the west coast of the US.
Sweet. You would not believe how many of these things labored away in quad or dual CPU boards, well into the P3 era. These were certainly Server CPUs. The replacement for these was the SLOT 2 Pentium II Xeon.
at first look i also thought the pins werent too bad. great work, easy for you. bios flash was a little scary.
Yes, the pins are quite easy to fix. There's a lot of space between them. Indeed, that BIOS flash was scary! Good that it worked.
Congratulations. Really good job. Also cool from you for using a switch on the ATX front panel connector.
I always use a mechnical pencil (the ones you use in school for geometric drawing, i live in austria so please don't ask what they're called in english, we call them Druckbleistift), the tip of those pens fit perfectly onto the pins and the pencil gives you very good yet controlled leverage on the pins (it also fits perfectly between them so you don't nudge another pin accidentially). The pencil method never failed me on bent pins.
Nice work, thanks for sharing. Happy to see this recovery completed. Hopefully I will be able to flex some similar skills like this in the future.
I'm sure you can! Good luck!
Wow. Amazing job with the soldering.
Thanks
That bios update: Yikes!!!
Yeah, that was quite scary... I'm glad it worked!
Great job on the repair and BIOS upgrade! I was surprised how weak those pins were, but I'm glad you could get this CPU working again.
I had a Pentium Pro for a while, it was a 180MHz model (I believe 256K cache), which I immediately overclocked to 200MHz since it came with a big heatsink (much larger than the one on yours). I even bodged a larger fan on, which dropped the temperature a little, but it ran stably for a few years, I never had any problems. Looking back, I'm not sure how much of a real upgrade it was, since I came from a Pentium MMX 166MHz. It's possible those MMX instructions may have made it faster than the Pentium Pro in some games.
Those heatsinks came with those boards. I'm not sure what they are, but they seem to be just good enough to cool the CPU. A better cooler would be great, but I've heard those coolers are rare these days.
Hearing your vocal expressions of joy upon success never gets old. 😄 No idea why you were apprehensive of this repair, it literally went exactly as I expected. But then again, I'm an optimist, haha. Those pins cannot be pure gold, they behave almost like an aluminium alloy. Looks like Intel was trying to reduce costs on those chips, as the huge cache probably made them expensive enough already. On a side note, Knipex are of course the best (not just the flat jaw ones) and these pins were probably all beyond saving so it didn't make a difference here, but generally speaking I believe the toothpick method may be more gentle and yield better results in some cases (esp. pins on pads), as it allows you to apply counter leverage and thereby better focus on the area that force needs to be applied to.
Anyway, it's always heartwarming to see another historic piece of computing tech back in action, I'm really looking forward to seeing more content about it in the future!
On another tangent, I'd like to rant about the fact that BIOS flashing can STILL be a super irritating process, specifically if it is necessary to flash a board so that it actually supports the CPU that you bought for it (e.g. a B650 board so that it supports a Ryzen 9000 series CPU). At least these days, you can usually flash it without obtaining and installing a supported older CPU first, but the process is usually not sufficiently well documented and often fails without apparent reason. I pity the person that doesn't have a bag of old USB Sticks to try until one is found that works. End of rant! 😉
Good point regarding cost cutting from Intel. These pins definitely are a lot weaker than on any other CPU I worked on. Even other PPGA models like Pentium MMX and Celeron Mendocino have better pins even though they have the same type of body.
I did not expect the detached pins to flow that well! I expected that the tweezer would steal just enough of the heat to make it incredibly challenging, obviously this was not the case 😅 well done!
I was surprised as well. Seems like the base had enough thermal capacity to draw the solder nicely around the pins and the base.
The PPro 1M are also stable at 3.5 x 66MHz = 233MHz (if your board supports 3.5x). I have two of these PPro 1M running on a Tyan S1662 board at 233MHz. :)
I have a couple 180MHz overclocked to 210MHz. If I ever get 200MHz 1M, I will definitely try to get 233MHz!
Perfekt! Geduld zahlt sich aus! Leider habe ich keinen 1MB Ppro in meiner Sammlung aber immerhin einen Slot2 Xeon. Ebenso fehlt ein PentiumII-Overdrive für Sockel8. Wenn die Pins so dünn und empfindlich sind, probier doch mal die Dinger punktuell anzuwärmen, damit sich das Metall etwas entspannt und nicht bricht. Jetzt auf jeden Fall Benches mit Pentium 233MMX gegen PPro in allen Cachestufen und P2-233.
Ja, da muss auf jeden Fall sowas wie ein Benchmark Marathon kommen. Ich habe ja alleine schon 3 verschiedene Cache Ausbaustufen: 256, 512, 1024 beim PPro. Und dann vielleicht Pentium und Pentium MMX.
Fantastic job! I'd love to see you repair pins with a mechanical pencil (it must have a metal tip). I've been fixing pins that way since high school 25 years ago.
This is the consent in the comments. I'll get one of those one day. For the next project where I need to bend pins again.
I always wanted a dual CPU computer in the day, finally got one in the socket1366. One motherboard CPU combo I would love to see is a dual socket 7 with a pare of K6-3 CPU.
Impressive stuff. I basically just write-off CPUs with any kind of deranged pins, the amount of frustration from such delicate work isn't worth it. Really cool to see what successful CPU repair _can_ look like, lol. I'll never experience the joy myself, but can appreciate seeing others triumph.
It is especially rewarding when it is a CPU like this one! But practicing on other, less rare, CPUs definitely helps!
Back in the day when worked in the computer shop customers would do this time to time when upgrading their cpus. As a last ditch effort to fix it for them we’d stick a appropriate gauge wire into the CPU socket holes where there were missing pins and added a small curl to the top of the wire to give it a bit of spring tension. If you do it right it works about 80% of the time as long as the original pad is left on the CPU. Have you tried watch makers tools? They are kinda a godsend for this kind of work.
I used to work for a Geek Squad warranty fulfillment center during the days of the Athlon 64. Our parts department was so careless, they would send CPUs to the repair line in loose plastic bags with no protective foam and no anti-static bag. They were probably stored that way in a bin I guess. By the time they made it to the repair line, the pins were all bent to hell and I really had no option except to waste 20 minutes straightening them out with a pocket knife. And I had to replace a lot of those Athlons, as I recall they had some problem with the onboard memory controller. Lucky I never had to re-attach any pins like this!
Wow you fixed that CPU. Loved watching you revive this old cpu
Really awesome. I always loved the Pentium Pro, I found it such an awesome CPU especially using multiple sockets. truth be told I have never seen the 1MB version IRL.
Ha, that would be a nice find at the scrapyard - a dual PPro board! I do have two matching ceramic PPros that would probably work as well! This black-top 1MB model is really nice. More videos to come in the future 😄
36:58 It is highly likely that the CPU threw an exception, without there being a handler for it. I would run it with EMM386 or something similar loaded into memory. My guess is that DOS will then bail out with exception 6 or something similar, since it is, technically, trying to execute an invalid instruction.
An invalid opcode 06, or general protection 0D, run these in real mode. And patch in an exception handler.
Lucky you still have interrupt enabled STI then youll need cold boot if interrupt off CLI
Excellent. You're a CPU surgeon. 👍🙂
I use syringe needles for straightening pins - it is much more convenient than a toothpick:
- find the needle of the right size (one that just slips over the pin);
- clip off the tip (do not mash it);
- put the needle over the pin and bend it back.
For thick pins of older CPUs a mechanical pencil can be used instead of needle.
Amazing video love your pin repair techniques
Thanks!
In the past, I've found the best tool for bent pins is a mechanical pencil, and to make sure the pins have been warmed up first to minimize cracking. Not too hot to melt the mechanical pencil, but like, hair dryer warm.
I'd be interested to see performance comparisons between the PPro and a Pentium(/MMX?) PPro was garbage at running 16-bit code, just how much of a difference is actually there? Running a pure 32-bit environment (say, Unreal on WinNT), what performance difference is there?
Good idea. I'm curious too. Sounds like a future project!
You might know this, but the 1mb PPro was not flat and needed a special heatsink to match. I sanded one down years ago but didn't have a way at the time to check any difference in temperature.
I am not aware of this. However, I used thermal paste between the CPU and the cooler. It looked pretty flat to be honest.
Do you ever "load setup defaults" after flashing a new BIOS? I remember that being a recommended procedure due to possible differences in the layout of saved user settings between BIOS revisions.
I did indeed once faced a lot of issues and loading Setup Default fixed the issue. I did not do it for this video I believe. But yes, this should be done based on my experience.
11:40 thriller background music... 🤣
Amazing work and video! One question: Pentium Pro heatsink and cooler seems impossible to find...what alternatives/compatible we have for this issue?
Thank you!
Yes, that is what I heard as well. Socket 8 coolers seem to be rare. I was lucky that I found three boards with coolers - all the same as shown in this video. Even though they're no high performance models, they get the job done.
The best option is probably a custom made solution. Maybe something exists already
Thank you for saving this historic CPU. One reason there are so few Pentium Pros left is because each one contains about 0.3 grams, or $25 worth of gold!
Amazing repair thank you for sharing 👍
Beautiful job. Nice to see this great CPU being rescued!
Parabéns!!! 👏🏼👏🏼👏🏼Você é muito Bom!!! 👍🏼👍🏼👍🏼Salvou a história dos computadores vintage!!! 💛⌨️💛🖱💛🖥💛🖨💛
Great save!
The 1MB are totally different beast compared to the ceramic variants, not sure why they changed the package but they look a lot cooler =).
And the PPro overdrives are worse still when it comes to the pins, very strong and rock hard but breaks even easier then the black tops.
Regarding to the heat output from these "beasts" I suspect Intel was "creative" with the power numbers cause there's no way the 1MB chips are just 9w more then the 256Kb variants.
I reckon the 1MB pulls close to double the power in real use!
Also I'd recommend using a thin thermal pad (I use 0.5mm) instead of paste, the black aluminium top might look cool but its certainly not flat enough for paste to work properly.
And beware the heatsink in the CPU power supply on the board, it gets incredibly hot when running theese chips.
Looking forward to some benches, don't forget to enable write combine! =)
This is a wonderful job done. And a great and rare CPU!
This is probably the most iconic and interesting CPU I have now. Followed by an IBM blue lightning 386/486 75 MHz (most likely) and a 66 MHz 486 SX CPU 🤣
You could also the mechanical pencil trick, just remove the graphite and slide the pencil over the pin and you can easily use it to straighten the pins. You managed it though, wonderful work.
I have never seen a black Pentium Pro.
And your soldering skills are very impressive. Great job.
Thanks!
@@bitsundbolts I've a question.
I got an Asus P2B motherboard for retro gaming. I went to turn it on by shorting the pins. BUT. By accident, I hit a 3rd pin. Got a spark & the board got messed up. I let it sit for a bit & when I tried to turn it back on. I get power to the board, Hard drive & CD Rom. But, I have no video (AGP card) & the cpu fan isn't working. Is there a chance to fix it or is it a lost cause ?
Im so upset with my self, Wanted this board for ever. It was rock solid untill I messed it up.
awesome job. thank you for the great video. way to save these old chips. so what are the plans for this little gem?
Most requests are to run some benchmarks on Windows NT/95/98. It may show the strength and weakness of this massive CPU.
@@bitsundboltsit might perform good for some cpu demanding games, like pod?
Yes. Some games make use of MMX (POD is one of them) - the PPro will not do well - I believe.
Use a 0.7mm empty lead pencil (like Ballograf Rondo) to straighten pins. It's magnificent for that task.
Have you tried to heat up bent pins with soldering iron and then try to unbend them? Maybe heat will help with brittleness...
I tried this on 486 CPUs. Unfortunately, heating pins doesn't help much - if at all. Those metal pins would require a lot more heat which most likely will damage the plastic body of the CPU.
Nice job again! Love it!
Thanks!
That P3 is actually quite likely to still work even with the damage, as this is a single layer chip (like nearly all modern silicon, as far as I know multi-layer is still very experimental), so only the bottom few nanometers are actually relevant, the rest is just a piece of glass. So if you can fix those pins (maybe with the help of uv solder mask to prevent solder from connecting to the wrong places), there is a good chance it still lives. I have actually seen worse chips still work, including some fairly modern gpu (3060 I think), which is still completely fine despite the damage.
I tested all those chips before I put them in a box for scrap pins. I'm afraid, this chip is dead. Some have visible cracks through the entire die. But you're right, from chipped corners, you can't tell that a CPU is dead. I have many CPUs with chipped corners that work perfectly.
@@bitsundbolts That's unfortunate, but it is vintage hardware, so not entirely surprising. Interestingly, I remember the P3s to be fairly reliable (unlike xp-era AMDs and celerons, and more recently the ryzen 3600), but shorted voltage regulators and bad ATX supplies weren't rare back then, so that might have been it. But whatever broke it, at least its parts help you save other chips, so I guess that's the best possible ending for a dead CPU.
Awesome rescue.
Thank you!
great job awesome love these old computer hard ware
ppro in ASCI Red was first supercomputer to hit over 1TFLOP. it was originally 200MHz 256k but later upgraded to specially packaged ppro II overdrive 333MHz 512k. now some pocket computers can do over 1TFLOP running from batteries.
The toothpick method puts all the force on a single weak point which is why it's so destructive.
If you can get the pliers to grip the pin it can provide some bracing and support as well as spreading the load.
The problem is that the metal already is work-hardened by the initial bend. The bending back will fracture it more, no matter the method, I think.
I tried this same method on a ceramic PPro - no issues there. It must be the material used for those pins that make them so weak.
Those older cpu's was nice. Big fat pins to solder broken one's back on.
I can suggest heating up the pins before you try to make them straight, it will make the metal softer and the pins will not break :) this work for me
Quick suggestion that worked for me: when the crack is just between the pin and the base (like two of yours), I put some solder on it FIRST to create a solid base and then straight it up in second place, so the weak base do not break when you straighten the pin, and it will be already fixed. It's just a suggestion that worked for me...
Interesting. That would definitely work if the pin isn't too far bent so it would bridge with the solder. I might have been able to save some of the original pins that way. But I'm still happy that the PIII pins are a perfect match as well.
@@bitsundbolts You are correct. Happy that you were able to save the processor !!! Those used to cost an arm and a leg in 1997-98. Never seen one in person. Hope to see it back in another video ;-)
Brilliant repair!!!
Thanks!
I really did love building servers with the Pentium Pro 200, but I did prefer the ceramic one over the black board ones.
5:50 On that 850, the pins have small heads so you should be able to solder them back on, it's just that you'll have to carefully soldermask that top copper layer. See if that top layer is connected to a GND pin, or just scratch off some soldermask, then test continuity to it after soldering. Maybe even sand it back further for more isolation/clearance.
I am sure it's possible to add the pins, but chances are low that the CPU works. I find these CPUs a lot at the scrapyard. It would be nothing more than a challenge to fix this CPU.
I remember using a fine size mechanical pencil (without the lead obviously) to easily straighten the pins on chips like these.
Why is the L1cache not showing up on READ - the line is flat up to 1024 ?
I think this is always the case. Reading seems to be limited by the memory speed. There is nothing available in the cache. Writing and moving seems to benefit from the cache
@@bitsundbolts I seem to remember most other CPUs showing the L1 cache on that line too
0xC3 will only work under the C runtime environment, for a DOS COM executable you need to use:
B8 00 4C CD 21
To return to DOS. The entire program should therefor be:
F0 0F C7 C8 B8 00 4C CD 21
Which means:
lock cmpxchg8b eax
mov ax, 0x4C00
int 0x21
0x4C00 being the exit program with return code function and int 0x21 to call the given DOS function.
Wow! Nice and thank you for sharing this. I'll try that once I get that system back up for some comparisons.
Wow great fix for a great CPU! Am I the only one that got dizzy with the fast movements under the microscope?
Nice save on that classic CPU.
at first i really liked that heatsink+fan combo but after you showed the thermal camera footage it's obvious that it kinda sucks. it seems like the airflow doesn't get passed over the edge fins.
Yes, the performance of this heatsink/fan combo is not good. Airflow is minimal. Maybe I need to experiment and put a Delta fan on that heatsink 🤣.
i've had good luck fixing pins with a mechanical pencil and using it as a pin straightener. Going slowly and gently bending pins and straightening them slowly.
The master of pin-fixing strikes again. :)
Haha, thanks 👍
Yeah I'm not sure. It's been forever since I coded in x86 assembly for DOS, but you might need to do some kind of call to return control to DOS. It's clearly not hung if you can Ctrl+Alt+Del.
I thought I did add the fifth byte to the code that should return, but it was just weird that it didn't exit properly. It works under Windows though.
If returning to DOS you have to call INT 21 exit to dos function. Really old dos let you exit with INT 20. Windows likely handles bad exits silently.
I had 12 Pentium Pro (200MHz) 512KB, each one was $1250 early 1996. I was able to overclock them all a year or two after, to 220 if I recall.
Jeez Sata3 speed L1 cache, we take a few things for granted nowadays…
Just an idea - wouldn't it be easier if you heat up the pin before attempting to bend it? I think this way the material will be softer and more flexible and perhaps not break? Otherwise amazing video as always!!!! :-)
Unfortunately, extra heat doesn't help. Those pins would require a lot more heat to get softer. I tried that on a 486 CPU, but that might also be a different material there.
Those old pins were fully formed, not just a cut and polished wire.
Awesome cpu repair video. As always good to watch but just where is background music??? 😢
Eh, I need to put some smoothing jazz again. I'll keep that in mind for the videos after the next one.
Now that's what I call a save!
I think the BIOS could have been flashed normally by booting from the floppy. The jumper being called BIOS Recovery leads me to believe it's only meant as a safeguard in case the board doesn't boot normally due to BIOS corruption.
In most benchmarks, the Pro 233 completely crushes the 233 MMX and it's very close in performance to the Pentium ii 233. If only I had a Pentium Pro 200 instead of a sad Pentium 166 without MMX when I was young...
I'll try to flash my other two boards with the jumper in place. I think I read somewhere that I should put the BIOS in recovery mode. Maybe it was written in the manual.
you made that look easy pal good job
I guess I had some good practice before! This was quite easy. Thanks 😊
@@bitsundbolts i bow to your skill my good man
Hi! Use pen refill for straightening bent pins and you'll notice how much better it is than pliers
8:46 (ish) - I remember reading on some forum somewhere, someone saying they had very good luck un-bending CPU pins by using a mechanical pencil with the lead removed. I suspect that might be too small for pins this size, though.
An interesting note about the pentium pro line of CPU's is that they are true 32 bit but with a 16 bit translation layer in a time where consumer cpus and windows itself were 16 bit.
This CPU should work better with Windows NT or Unix/Linux.
Good evening... could you straighten the pins using a 0.5 mechanical pencil?
Yes, a mechanical pencil might work as well.
I always thought the best tool for this job was a mechanical pencil with no pencil lead in it - it's a tube that supprts the pin fairly well as you bend it back. Also I'd think getting the pins hot would help them bend more than break?
Yes, a mechanical pencil does work. Unfortunately, it would block the view white a bit when I'm working under the microscope. That is why I decided to use the pliers. They also work very well!
I tried heating pins on a 486, unfortunately, I never noticed a difference when heating up those pins. It just makes it harder to work on the CPU because it also gets hot.
tl;dr - I think the jumper you moved to flash was actually for BIOS recovery mode - no harm done but it skips any verification and is why you didn't see anything during the flash.
If you had just normally booted from the floppy w/o moving that jumper, I think you'd get a more normal experience of it loading a flash util and you see progress on screen?
Assuming this board is like how many later Intel reference boards are the jumper you moved to flash was actually the "bios recovery mode" which I think was intended to be used if you accidentally bricked the mobo (assuming the BIOS boot block still lived). I know of it because with at least with Intel socket 7 and later Slot 1 boards you could also use that to "cross-flash" between various OEM "flavors" of firmware - for instance if you wanted to change from the Intel branded to Gateway branded bios. Sometimes OEM versions (like Gateway) would fix things like sleep or resume that the Intel "original" didn't work right.
Note that since it seems to skip verification yes you can also blind flash totally wrong firmware too which can be entertaining - I think as long as the boot block still lives you can recover from that also, but obviously not recommending you test that unless you have an eeprom programmer handy =)
Hm, interesting. I didn't try to flash the BIOS without changing the jumpers. I think I read this procedure in the manual.
I really don't want to experiment with those BIOSes - they're those flat BIOS chips that would be a pain to remove and flash. Good that it worked like this too.
Pentium Pro, the little architecture that could. It may have been criticised at the time it was first released (mostly because poeple didn't understand the target audience of the cpu) but all modern Intel CPUs still fit comfortably in the p6 Pentium Pro family. 30 years for a cpu architecture is very impressive. I do wonder, if AMD gets a few more Punches in, will Intel have to try something new again.
I wouldn't trust soldering broken pins like that if the socket wasn't zif. The force needed to insert and remove a cpu from a non zif socket would surely break the solder.
The really bad thing about the f00f bug is that it could take down protected mode operating systems by completely crashing the cpu and it couldn't respond to interrupts like the Ctrl-Alt-Del interrupt from the keyboard controller. On the PPro the instruction sequence no longer stops processing of interrupts. It just generaates an ivalid instruction execption as intended and crash the thread of execution. In Dos it just hung because dos doesn't have a useful default exception handler for the crash foof causes. Ctrl-c might even get back to the dos prompt from this hung state
I think (just a guess) that silver-containing solder wire (Sn62Pb36Ag2) is slightly better in such situations. low soldering temperature, greater strength. Anyway, excellent work!
Interesting. I use my regular solder wire and it worked white well. Maybe some day I'll be able to test something like that.
That was incredible
I love BIOS write protect, please manufacturers bring it back!
Good job!
Don't let those chipped corners fool you.
I have at least one AMD chip from the Athlon XP days that has a corner chipped way worse, but still works perfectly fine. I don't think they put transistors all the way to the edge of the die. If only to allow for some tolerances when cutting the wafer.
Yes, I have plenty of chipped silicone dies as well that work perfectly. I test each CPU before it goes for a donor piece. Unfortunately, this one really didn't work. These days, I only take CPUs that are still installed in boards with a cooler. Those always worked for me and are in very good to perfect condition
I saw a life hack from a blogger where he used a shortened syringe needle to align the processor pins.
13:33 Dude the whole base of that pin is moving inside the CPU structure!! I wonder why? is the pin like recessed into the CPU maybe 🤔
No idea, but you're right! The entire base moves. Well, since I'm not planning on digging those pins out, I guess we'll never know.
I just had an idea. I wonder if soldering tip made out aluminium would help with repairs like this. It would be shaped like a normal conical tip but with the tip shaved off and hole for the pin drilled. It should also have very litle thermal mass. It would be used by placing replacement pin inside the hole and adding solder paste to the CPU. After putting both parts together the paste should melt and create a joint. Then the iron should be cooled down as fast as possible (to avoid CPU damage) while still holding parts together. The fact that the tip is made out of aluminium should make it impossible for solder to wet the tip and this way making it possible to be removed while cold. Small thermal mass would help with cooling it down quickly with freeze spray. If the joint is not satisfactory it could be then reflowed with extra flux with regular iron or hot air.
I need to get my hands on some scrap PGA chip and try it.
It sounds like an interesting project to try! Please let me know if you get to it. I am just a bit worried that somehow the solder would still climb on the pin up where the soldering iron is. Capillary effect it is called I believe.
@@bitsundbolts That could be prevented by using tiny amount of high temperature grease. Most likely some kind of silicone.
If the CPU does not have MMX at all, then the F00F instruction may generate an "undefined opcode" exception that might very well be the softlock you experience here. I guess you can try to safeguard even more about this by adding more code to catch that exception, but it's kind of a corner case.
I think you recommended to add the return instruction after the fourth byte. It makes sense, but I was still surprised that the CPU got stuck - not in a deadlock though as with the foof bug on the P5 architecture. Thanks for the help btw!