Followup: IBM 5170 "extreme" overclocking (w/ associated BIOS patch)
![Adrian's Digital Basement ][](http://yt3.ggpht.com/ytc/AIdro_k7CrQRe71qN6c0Ua7A1P_I8KkxjqLa70u0zYeCvdvnKw=s48-c-k-c0x00ffffff-no-rj)
HTML-код
- Опубликовано: 9 авг 2021
- Turns out the IBM 5170 Type-3 BIOS has code in it to prevent overclocking! I already showed you could run the 6mhz Type-1 board at 8mhz in a previous video -- but how far can I actually push it? I also show exactly what was wrong with my XT-IDE card to prevent it from working.
-- Video Links
IBM 5170 BIOS Patching:
• Fixing and improving t...
Mystery Box Video: (where I first overclock the 5170 motherboard)
• Mystery Box: Testing, ...
Patched BIOS download and information:
archive.org/details/ibm-5170-...
Support the channel on Patreon:
/ adriansdigitalbasement
Adrian's Digital Basement (Main Channel)
/ @adriansdigitalbasement
-- Tools
Deoxit D5:
amzn.to/2VvOKy1
store.caig.com/s.nl/it.A/id.16...
O-Ring Pick Set: (I use these to lift chips off boards)
amzn.to/3a9x54J
Elenco Electronics LP-560 Logic Probe:
amzn.to/2VrT5lW
Hakko FR301 Desoldering Iron:
amzn.to/2ye6xC0
Rigol DS1054Z Four Channel Oscilloscope:
www.rigolna.com/products/digi...
Head Worn Magnifying Goggles / Dual Lens Flip-In Head Magnifier:
amzn.to/3adRbuy
TL866II Plus Chip Tester and EPROM programmer: (The MiniPro)
amzn.to/2wG4tlP
www.aliexpress.com/item/33000...
TS100 Soldering Iron:
amzn.to/2K36dJ5
www.ebay.com/itm/TS100-65W-MI...
EEVBlog 121GW Multimeter:
www.eevblog.com/product/121gw/
DSLogic Basic Logic Analyzer:
amzn.to/2RDSDQw
www.ebay.com/itm/USB-Logic-DS...
Magnetic Screw Holder:
amzn.to/3b8LOhG
www.harborfreight.com/4-inch-...
Universal ZIP sockets: (clones, used on my ZIF-64 test machine)
www.ebay.com/itm/14-16-18-20-...
RetroTink 2X Upconverter: (to hook up something like a C64 to HDMI)
www.retrotink.com/
Plato (Clone) Side Cutters: (order five)
www.ebay.com/itm/1-2-5-10PCS-...
Heat Sinks:
www.aliexpress.com/item/32537...
Little squeezy bottles: (available elsewhere too)
amzn.to/3b8LOOI
--- Links
My GitHub repository:
github.com/misterblack1?tab=r...
Commodore Computer Club / Vancouver, WA - Portland, OR - PDX Commodore Users Group
www.commodorecomputerclub.com/
--- Instructional videos
My video on damage-free chip removal:
• How to remove chips wi...
--- Music
Intro music and other tracks by:
Nathan Divino
@itsnathandivino 
Наука
I just spent 5 minutes watching an ad about an overclocking meter without realizing it wasn’t the video
Well done! IBM rules! I would like to see you build a KILLER 286 at 20MHz and put OS/2 1.1 on it!
I felt excited to see someone rewriting a BIOS and have it applied and working, with all the patches. I'm a software developer, and I LOVE being able to write applications to do my bidding. I've never gone as far as to build a BIOS for any machine, so the process in this and previous videos have been wicked awesome. LOVE it. Thanks for the update!
IBM also did give the full source code for the BIOS, a far cry from the modern offerings from them, or from almost every other manufacturer now.
From a firmware perspective at least, these boards should be unusually "easy" to write a BIOS for, since they were made of largely commodity chips that were/are well documented- it's akin to a more comprehensive DOS extender or other DOS program (the biggest differences are that the DOS/program relationship is reversed, and that you aim for "extender-esque" behavior). The biggest weirdness is probably that you need to setup stuff like DRAM refresh (early PCs used one of the DMA channels for it), which in basically _every_ other piece of IBM PC code is unneeded because it's already configured.
A good question is: did the person who modified the BIOS have the source code, or did he have to disassemble the machine code?
@MisterMac56 with the knock on effects of utterly bringing IBM to the brink…..
I read and article about how it was a common upgrade back in the day to upgrade the clock chip. When IBM got word of it. They put int he bios lock out. Didn't stop places from offering a kit with the clock and bios in a package.
I would probably put the Harris 80c286 back, if anything just to reduce the temperature inside the case. Oh BTW, you can get 25 MHz versions of that chip on eBay for like 5.50 bucks with free shipping.
Next step, liquid nitrogen overclock? ;D
Its not cool enough until it hits absolute Zero!
A collaboration with GamersNexus!
@@DankNoodles420 might be hard to pull off
133mhz ibm 5170
Sometimes, overclocks may cause strange hung up's.. I did it even on my P1 machine, back in day's..
In this case - check out what exact version of programmable interrupt controllers (two of them 8259) main board has installed and some other bus parts.. Even one slower part on bus can cause similar issues, because, didn't release bus in time..
Anyway, great video, as always! 👍😉
Excellent job. Very satisfying to watch it.
I love your both main and secondary channels Adrian, cheers from South America, keep it up!
when you think about how 2Mhz overclock was huuuge increase in speed back in the XT era
I *love* that DIP chip remover you used, very slick. I'll have to pick one of those up one of these days!
IIRC, the Tera Drive was a joint venture between Sega and IBM, so it makes sense it'd have an IBM ROM.
great video,, thanks for sharing.. cant tell you how much I enjoyed all the info ..
I found it interesting that Adrain dropped into the 286 running at 10MHz - the 286-based PS/2s (Model 25 286, Model 30 286, Model 50, Model 50Z, Model 60, and 7552 Industrial "Gearbox") all were 10MHz systems. He placed heatsinks on motherboard chips, but one socket is unpopulated that would definitely need one for the transistor count: The optional 80287 FPU.
Only the microchannel systems of the 50, 50Z, and 60 ran the 80287 FPU at the same speed as the 286 CPU ('CKM', pin 39, "Clock Mode" tied high) - the ISA models and 7552 (where the FPU was included standard) ran the 80287 at 6.66MHz ('CKM' tied low to ground, so two-thirds of the 287 speed). It was primarily that someone could use a lower-rated (and lower cost) 80287. The later i287XL didn't use the 'CKM' pin in the same manner to run it at the higher frequencies it was capable of.
The instability might be due to the cpu heat. Give it it's own fan/heatsink combo to fit.
I think this was worthy of a first channel video
This is a really good ii channel video
Another great video, always informative. I did have a 20Mhz 286 machine as my first PC and I think I did run in to problems with the 10Mhz ISA bus speed but that was only with one card so don't overlook that if you continue on with the overclock but on the whole my 10Mhz bus was fine, the stepper motor hard drive from Kaloc? that was another matter, it worked when it was cold or when it was warm but never both not matter when it was low level formatted, sounded wonderfully clunky though.
2:16 I don't think you are framing it the historically correct way. I'm very confident this wasn't meant as a dedicated check against overclocking, but instead at those times the Power On Self Test actually did what the name promised - i.e. it made sure that the system indeed behaves up to specification, and prevents booting if something is out-of-spec. This also includes verifying that the different clocks run at design speed.
Of course, checking that a system behaves "in-spec" has the collateral effect of detecting overclocking as "out-of-spec", but the purpose of this check is more likely to check that the 14.318MHz clock operates correctly (by comparing it to the 6MHz processor clock) than to check whether someone "illegally" runs the processor too fast. The fact that it checks for both overclock and underclock supports the idea that this is not a dedicated "overclock" check, but meant merely as "clock plausibility check".
Unlimited Power!!!
On XTs, PCs, and PCjrs, you'd need a PC-SPRINT modification. The clock system is set up to allow for a separate bus clock chip, so you can run the CPU at a different speed than the bus. It may still be possible to use that mod on an AT.
This is great! 🙂
Wasn't expecting any mention, but certainly appreciate it. Likewise, I enjoy your videos and often have them on in the background when working on things. The Tera Drive was indeed built by IBM, even uses their weird edge connector hard drives, and it wouldn't be their only venture into the world of games console related hardware, as I at least think they built the Atari Jaguar too.
The XT-IDE thing sure is curious, even though changing that chip seems to have solved it for you. On my off-brand 286-6, installing an XT-IDE BIOS (after butchering a regular IDE card to sit on IRQ 5 and, I think, port 320h) causes the system to weird out and run the CPU from double the FPU's clock instead (20/2, where the FPU is 20/4). Have no idea what it does or how on earth the board is wired to be tapping a clock from the middle of a completely separate 7474. It's slower than your machine at that speed, though, as well as extremely unstable. Luckily it's back on MFM now and doesn't exhibit this issue.
If you change to a CMOS version of the CPU, you probably won't need a heatsink on it. Maybe if you have one that was made later, or especially any that came from laptops, it could be worth a look. These usually, but not always, have a model number like "80C286" as opposed to just "80286" stamped on them. There sure is just something about feeling that radiant heat from a bunch of NMOS logic, though.
My IBM AT had a Kingston 386sx/25 SX Now! Upgrade fitted, this was the most reliable system I ever had. Everex RAM3000 and Intel ABOVE BOARD AT making 5mb ram, and an IDE interface. Later fitted a K6-2 500 and wide scsi drives to make the fastest AT ever.
Very very interesting!
Fascinating stuff, I'm about to embark on my own 5162 project, which is a 6MHz machine with the same BIOS lockout for overclocking. I have a 12MHz 80286 and a few different crystals ready to drop in - although I think the 5162 BIOS is unique to that machine so I'm not sure your patched version would work. Either way, I'll definitely be referring back to your videos (and of course giving you a shoutout) when it does happen!
The 108 error symptoms sound similar to reset issues. Once reset is released somebody is starting up more quickly at the higher clock rate, which is violating some other timing in the system. You could add a cap on the reset line and lower the slew rate, giving everybody more time in reset. But that's a 50/50 chance of making it better or worse, plus some components are sensitive to super low reset slew.
Sad to hear IBM even managed to lock users out who try to change the system. I played with a lot of clone systems in the day and overclocking was easy with no real problems other than instability when going really too far. Heck even my Atari pc3 even runs at 10Mhz at the moment (and for more than 20 years now) just by changing the 24Mhz crystal to a 30Mhz one.
In those days IBM would have been guarding against shady resellers rather than the end-user.
Yeah, I don’t see IBM as being the type to prohibit users from doing whatever they wanted. Too much of what they delivered to the customer encouraged opening the case and poking around.
I do wonder if maybe some of those speed checks were there to catch abnormalities - or, as RetroSwim said, to keep the retailers honest.
11:52 to some modern gamers, that's just normal everyday overclocking :p
as in, turn it up until it bluescreens and then back it off until it becomes stable enough to work
And what some people consider "stable" isn't always "stable". I mean if you can get several hours of gaming or production out of an overclock before a crash, thats good for some people.
I've tried overclocking on my PC but the computer crashes with WHEA errors on large memory hogging games (or just games taht eat up a lot of memory if you push it)
Hah, I was going to suggest the DS1077 clock synthesis chip but then you mentioned it. Definitely would be interesting! You could also probably use our good friend the ESP32 in a pinch as well, the RMT peripheral can give you a carrier wave up to 80MHz, and you can set it to divide it down with a lot of resolution as well.
Adrian 👍
Hi Adrian,
To answer your question about the ISA bus speed on the 5170: It's a synchronous design, meaning no clock dividers were used to drive the expansion slots so when you overclock the CPU to 10MHz the ISA bus speeds will also be 10MHz.
If you're looking for more stability when running at 10MHz: Swap out the 82284 clock generator and the 82888 bus controller for 10 MHz rated parts. In stock form both of these chips are 6 MHz rated parts and cannot keep up at 10MHz (60% overclock) ;-) RAM should be 150ns and should be fine at 10MHz.. Some of the earlier machines had 200ns rated memory though so I'm not entirely sure they will be as comfy at 10MHz as their 150ns counterparts.. (not sure about the chips on your board as well)
Hightreason is still one of my favorite youtubers
take a look at the original clock crystal. Solid construction, precise engineering, low impedance leads which were probably plated with silver or gold when it was new. Overall, a superb crystal, custom made with strict capacitance and inductance properties, to function in parameters for many years. Also it fits its ceramic socket perfectly.
now look at that sorry ass cheap crystals which you can find today. small case, low Q, thin legs.
and you wonder why this motherboard hangs sometimes with those crystals? wonder no more.
Put it on HWbot, good score remember to add photos !
Tha's realy neat.
Interesting that the XT IDE interface card seems to be using tantalum capacitors for IC decoupling. You would usually use a ceramic capacitor, due to their better high frequency response.
Great video Adrian. I love the idea of getting a clock synthesizer chip as a replacement to the crystal oscillator, it would be possible to change the speed on the fly that way? That would be awesome.
Computers were always meant to be the electronics equivalent of the hot rod.
Problem with hct chips is driving power and with buses of length need power. On boards I built back in the mid to late 80s I used hct on my boards but had to use Als to buffer but these als buffer chips cost as much as all the other chips on the board. Now one cpld is far less than all the ic I used to fit to one board.
The BIOS over/underclock check might have been a quality control thing. Users would rather deal with a computer that won't boot due to a clock defect than deal with an unstable one that could possibly corrupt data or provide an excessively subpar experience.
Also, agreed. You're probobly rarely jumping an initial condition. A clock-control chip might be able to push further. Boot at 8Mhz, then gradually ramp up. Might even get as high as 14Mhz on that thinking about the parts,
he gets it..
Don't be greedy 🙂 Just leave it at 8MHz
200ns RAM is spec'd to run at 5MHz... 1/200e-9=5e6 and that's without taking into account RAS/CAS setup times, etc.
Adrian need to invest in liquid cooling now...
Or, put it in the freezer and see how far he can push it...
'Push it real good'
Very cool. Did someone overclock a 5150 board? What would happen with an overclocked 5150 to its tape port, would it still work at the same speed? Questions over questions... I love this old stuff...
Are you happy with that little EYOYO monitor? Thinking of getting one of the 1024x768 4:3 ones.
I know this is old but if you pick up a 220v muffin fan and run it at 110, it still blows a nice stream of air, but is virtually silent.
And/or how hard would it be to get(purchase?) or rig a larger heatsink and fan to the cpu?
My money is on the slower rams, I'm much younger but from my athlon XP days, overclocking with linked clocks always tended in really unstable machines. You have ONLY linked clocks in that machine so getting memory access violations/bad access with slower rams is probably a common thing with your setup.
I wonder if adding heat sinks to the ram would help?
My first overclocks were then too. Overclocking the FSB (AKA EVERYTHING!) just made stuff so unstable.
@@ericpaul4575 : Short answer... No. Long answer (still abridged)... in order for a memory cell being read, some prerequisites are needed. First the memory controller needs the address, it then decodes this address to a row and column, then said row needs to be powered up, said column needs to be selected (ras to cas, blah blah blah). All of these actions take time and only have a limited time frame to complete their task, if you have slow physical memory that in a very short amount of time simple can't do what the memory controller orders it to do, you either get incomplete reads or incomplete writes... both is a catastrophic failure mode. Think of a grandpa at the trainstation, train arrives, grandpa finally realizes his train has arrived, finally has all his sh*t together and begins crawling the train, train already closes its doors without grandpa bording. Gone.
That CPU needs a better cooling, what are you planning to put on it? Something bigger but passive stuff or a little dedicated fan? Maybe some special passive cooling?
Simplest to just put a better thermal solution on there. Easiest is to take a Intel copper core heatsink and use a hammer to slide the copper core out 5mm from the fin area, and use that copper slug and some good epoxy to attach it to the CPU. Will keep it cold, I have done that to a lot of 266/386 CPU's, using old south and north bridge heatsink units, and a small amount of 2 part epoxy, typically Pratley quickset steel epoxy or Araldite rapid, a thin coat applied to a clean, degreased and lightly sanded surface both sides, and firmly pressed down to get a thin even coat. Basically a blob the size of 2 match heads applied to the surface and then squeezed out to cover the mating surface.
@@SeanBZA Do you need a special 2 part epoxy for cooling fans or any 2 epoxy will do it? The epoxy in old C64 PSU is meant to dissipate heat but looks more like a heat trap. I'm not familiar with heat conducting properties of different types of epoxy glues.
@@Dukefazon Yes you do get thermal epoxies, but a thin film of the metal loaded 2 part epoxy will work almost as well, and in most cases you can not tell the difference. I used them a lot with hot running IC chips where you can not use any other mounting method, and the thin film, along with smooth flat surfaces that are cleaned well to remove all grease, you will be hard pressed to tell a difference.
I use it often to attach LED units on star aluminium core boards to a larger heat spreader, and you have to solder the wires on first, as afterwards you will never be able to get the solder to melt, the heat is removed so well. The 2 part steel epoxy is also a lot cheaper than the special thermal ones, so there is a benefit to it as well.
Regular Epoxy would not be good, as it is actually an insulator. There is actual thermal adhesive used for such an application. But I would not recommend it as it is a permanent bond. I would rather recommend thermal adhesive tape, which is removable.
I wonder, many errors and instabilities I got in early computers where due to data read errors of media either by controller or reception errors due to chip bus speeds, recheck the hdd controlleer and or try using a ramdisk instead.
The RAM likely is fine, as when the manufacturers got the process dialled in they would find all the RAM that passed basic function test would run at pretty much the fastest speeds fine, so to do the low cost versions they would mark them with the slow speed, as they do meet that speed, while selling the tested high speed devices for a premium, and using the bulk sales to cover the run, the performance parts would be the profit. You probably though are running the memory right at the edge at 18MHz, as the long traces, and not matched traces, are likely leading to the last chips on the board barely able to meet spec for data read and write, and to latch address bits into the chips. Run the board hotter and it likely will not fail, but with the 17MHz clock crystal simply freeze the RAM chips furtherest from the CPU, and it is likely you will get a reliable crash as they get cold.
Funny, the math suggests even what he's doing isn't going to be possible much less reliable. That said, you're absolutely right and his experience bears it out. Pushing the limits of the design and material science of the period means it's going to halt under certain circumstances due to data corruption, bus collisions, wait times, etc.
Fun nonetheless. I discovered in my own experience doubling clock speeds eventually led to instability and uselessness that heat and power management couldn't help.
As someone with a newly acquired 5170, I was wondering if you have tried a 16bit ide controller to drive your compact flash. I’m going to try it when I get back from a trip. I guess you might still need the XTIde bios if you are using the ibm bios - maybe not if you are using the quadtel ?
See my 3 part series on the 5170 -- I address this with a BIOS patch.
You should revisit your Composite EGA card, but make something that will actually _convert_ an EGA video signal into Composite Color. That way you can still play with Artifact Colors and stuff. I'd really love to see the 8088 MPH demo running on overpowered hardware.
8088 MPH is very timing sensitive, it doesn't run correctly on anything other than a 4.77MHz 8088. It also doesn't run on EGA, since many of the effects exploit quirks of the 6845 CRTC on the IBM CGA.
@@RetroSwim The CPU timing can be probably be patched, and the various graphics quirks can probably be replicated with a custom adapter.
@@AmaroqStarwind That's not how either of those things work, I'm afraid.
I actually own an EGA that has composite color output from the factory. It's awesome and makes EGA games look like arcade games 👌
Oh and 8088 only runs correctly on XT class machines with CGA. Even a NEC V20 will make it less than 100% compatible.. V20's mess up the CGA colorburst signal so even when these will run it will be in monochrome.. It really has to be a 4.77MHz 8088 or the timing will be messed up. Same thing with Area 5150 which can only be run correctly on an actual 4.77MHz 8088..
1mhz has a period of 1000 nanoseconds.
100 nanosecond response time memory is good to about 10mz before it runs out of time to respond to an operation.
200 nanoseconds equates to 5mhz theoretical max clock speed.
Memory makers are going to be conservative to ensure stability and longevity. Obviously things are more complicated and reality often defies theory, but it's more like a guideline.
Faster memory would allow you to push the clock further to a point.
Yeah this board runs at 1 wait state for the RAM -- so that allows the 200ns RAM to work up to 10mhz I guess.
HI Adrian. When programs like Landmark (or the BIOS for that matter) do a speed test what do they use for a time base--i.e. what do they compare the CPU clock to? Do they get it from the Real Time Clock?
if i recall correctly, before HPET was introduced (~2005) PCs had an internal clock running at 32.768 kHz from the 16-bit era onwards, maybe even the eight bit computers, but that seems peculiar because 32768 is exactly 2^15, and i have no clue as to how an 8bit cpu would process that.
tldr: not sure about the 8bit computers, but the 16bit machines (and onwards) had an oscillator that was independent from the system bus for RTC
edit: im probably going to take a look at the source code of the ibm bios, when i get the time
@@nikgolinar4378 all x86 cpus had 16bit registers
@@MonochromeWench fair point - so 16 bits of data would just have to get there in two cpu cycles
@@nikgolinar4378 The programmable interval timer is fed a clock signal of around 1.2MHz (originally the 5150's master clock divided by 12) and this is usually used for timing. In normal use it overflows around 18 times a second and triggers IRQ0. Most timing functions in the PC, such as the BIOS clock, are based off of this.
Even with faster CPU clocks the PIT is still driven at the same frequency for compatibility.
Is it possible to implement a clock divider function for the ISA bus on the bios?
Later PC's used this -- but surely that wouldn't be easy without major rework on this board.
See if you can find better heatsinks. Those types are pretty bad for larger chips. I have some blue ones where the fins spread out over the top and allow more airflow between. You should also look for larger heatsinks in regards to CPUs. That little chunk of aluminum there wont do much, especially since its not covering the entire thing. Shop around for some more sizes and types. :)
I suspect the lock up issue is the ctrl-alt-delete ISR has a reset for the PIT/PIC chips, which makes sense in a soft reset.... (this would reset timers, pc speaker output etc)
I bet the IO being written to these chips may be getting missed in some way, so maybe programming the PIC is failing?
There were other thing you mentioned that make me suspect the pic/pit being unable to handle the speed...
When the computer was "locked up" at the end, I wonder if the keyboard is actually what stopped working, not an actual lock up? It's hard to say really BUT!
I have a challenge, go way over the top to cool the PIC/PIT chips and see if the instabilites go away?
Alternatively, perhaps if there is some resets inside the ISR for ctrl-alt-delete, you could double up the writes to those io ports, inserting a delay into the system .
Close but I don't think the PIT/PIC are the actual problem here as they're both 10MHz rated parts on these machines.. Even the DMACs have a lot of reserves as they run at 1/2 duty cycle and are 5MHz rated parts but run at a measly 3MHz when running using a 12MHz stock crystal (6MHz clock speed). The real problem actually lies within the clock generator and bus controller.
Both the 82284 clock generator IC on the 5170 (which also handles resets) and the 82888 bus controller are 6MHz rated parts on type 1 AT's. A 60% overclock (6 to 10MHz) is just way too much onions for those old parts ;-) It's surprising the machine even POSTs at that speed.. The Schmitt trigger that is used for the reset circuit will most likely be out of whack and will not be able to respond to synchronous inputs which is why the machine will not be able to handle resets/warm boots..
Replacing these parts with 10MHz variants will make an early 5170 run at 10MHz no problem.. Any higher than that an the RAM will become an issue as IBM used 150ns DRAM on type 1 5170s (it's still a '84 machine ;-) ).. On the type 2 machines the on board DRAM is easily replaced by faster parts.. On a type 1 not so much as it uses 128K piggyback DRAM chips and as far as I know they're always 150ns or slower as those are 1984 parts..
Would such overclocking effect playup frequensy on a soundcard or is it just the communication with some chip/DSP that is effected?
What is the actual soundchip on an AdLib card it can't be the OPL chip or?
Almost surely the Sound Blaster card would have no issue with this. It's designed to work at 4.77mhz to 8mhz (the XT and Type-3 5170 ISA speed) so it likely wouldn't have any issue running at 10mhz either.
Sound cards all use either an on board master reference oscillator, or use the 14MHz clock from the motherboard, which is not generated from CPU clock.
@@SeanBZA Maybe it was primitive games using pc speaker that used cpu clock that i remember.
@@JmanNo42 Even those used the counter timer chip, which was fed with a 4.77MHz clock, often derived from another crystal on the board, or derived from the 14MHz clock source, so not linked to CPU clock. PC speaker sound generation that involved turning the output bit on and off using CPU control though did need to have the routines first use a speed checking bit of code, that then was used to vary all the loops used to make the sounds, unless it was single tones made using the timer chip that originally made the PC speaker beep, which could generate tones all by itself of fixed frequency and duration, only needing an interrupt to signal end of tone to the CPU.
The colour direcory listing you're using, would that be "Paul's Directory" by any chance ?
Not sure - it's called "hdir.exe" and I found it ages ago one some random website
Probably hotdir? I used to use that back in the 80s.
Are the BIOS chips keeping up? Had some weird issues with e/eeproms when run at faster than spec. Their delay spec is highly conservative but you will eventually reach their hard limit where they just can't latch data as fast as you're reading it. Had one particularly sinister encounter where it would work perfectly fine if you read it linearly or in very small increments but it would crap out if you jumped from one location near the start of it to another near its end. Had a really, not so, fun time debugging that one :)
With modern flash not an issue even going to 20MHZ bus speeds, but older Eprom and EEROM yes you want the fastest speeds around, but modern ones easily will do 30ns timing no problem, and even with the IBM PC the manufacturer put in a single wait state, so they could use cheap 120ns Eprom with no problems, or cheap 120ns mask ROM, instead of going to the much more expensive at the time 80ns parts.
I'm curious what the power consumption differences are between stock and the different overclocks
Voltage has not been increased, so only more current might be going through.
@@markae0 power is voltage times current (v*i), so in theory it could get more power hungry.. depends on if the chips need more current to run at faster speeds or not though
@@YoureUsingWordsIncorrectly Just use a multimeter on amps on the 5v rail.
in this case not much.
@@YoureUsingWordsIncorrectly For testing power draw at the wall (AC), I would use a P3 Kill-A-Watt meter (or one of its clones).
Or, for a more accurate reading, a DC power meter of some sort inline with the 5V rails like @Richmondo mentioned.
1/200 ns = 5 MHz. I guess it's running with 1 wait state to keep in spec? If that's the case, that RAM should be good up to 10 MHz.
It's a bit more complex than that. Access time is not the same as cycle time for DRAM, and the 286 still used more than one cycle to access memory (the 8086/88 used four cycles).
It would be interesting to measure the power consumption.
DMA speeds may cause some problems, but I'd guess not by default. The DMA controllers were never designed for the x86 line, and instead were a cheaper device (as compared to the Intel option- Intel's apparently was a byte-code microprocessor, much more expensive, and used by almost nothing!) that IBM licensed from AMD, which had a semi-weak coupling to the bus- my general impression is that if you don't try to program it too fast, then it'll usually work fine... and at about the same speed, _regardless_ of your bus clock speed (though I may be misunderstanding something about that).
It's interesting is I had a bad 286 motherboard that ran at 12mhz which seems to use the exact same chipset as the 5170 Type-1 board. So perhaps those are later controllers capable of more speed?
You're right about the x86 DMAc's being weak links in general. I mean they were obsolete off the shelf parts and were actually originally meant for the (8-bit!) i8085 CPUs from the seventies so even in '81 when the 5150 launched they were already outdated, which is why the two (problematic) 5MHz DMACs on the 5170 ran at 1/2 duty cycle, which is 3MHz. They shouldn't cause any additional problems on a 10MHz "insanely overclocked" 5170 other than being the usual pain in the @$$ ;-)
It's the 82284 clock generator and the 82888 bus controller (both 6 MHz rated parts) that are causing issues when overclocked to 10MHz...
Adrian, alot of your links above for the tools you use are broken. I was interested in the post code card you use.
Unfortunately yeah keeping those links working is really hard and I've had that card for several years now. Just look for ISA post card and you'll see several inexpensive options.
Adrian, how about placing a fan directly on top of the CPU like in a modern machine? I've read that there are fires now in the region of your city (Portland isn't it?)
Yeah, wild fires usually caused by overclocked CPUs without proper cooling! :)
I don't think they make fans that small, the chip is just slightly bigger than a thumb.
For the CPU that is «hot », maybe the running voltage is too high, 286’s had two possible voltages :5V (most common) and 3.3V (less common / mobile segment) the model Number of the CPU may say that is the 3.3V variant
Negative. All 286 CPUs are 5V parts.. Most HMOS incarnations of this CPU ran hot as hell. Most of them having a max. rating of 85 degrees centrigrade so is normal for them to be nice and toasty.. If you want a cool running 286 look out for a mobile variant, a 80c286 of which most later ones were of the CMOS variety but also ran at 5V.. The Compaq SLT 286 was one of those machines that had a Harris 80c286 on board.
I think that the size of the 286 CPU in this motherboard is similar enough to the area covered by some Raspberry Pi HSF kits. It might be worth running one of those in order to allow this overclock to be more stable or even taking it up to higher speeds.
200ns = 5 MHz; so I am guessing this memory is setup with 1 wait state. 200ns with 1 WS would be good up to 10 MHz. So you might be hitting the memory limit. The wait states are probably not configurable, but adding one would give you more overhead.
Yeah from my understanding all 5170 boards have 1 wait state on the RAM -- likely exactly for the reason you said, tops out at 5mhz which is faster than even the slow 6mhz speed. The 5162 286 board uses SIMMs for RAM and it is 0 wait state apparently. I wonder if there is a way to disable the wait state? Clearly all that RAM would need to be swapped out.
See if you can get a 19.5 MHz xtal. That might get things a little less quirky.
With that RAM being stacked like that, is the buried set getting hotter than the upper due to being sandwiched?
No he clearly stated the lower section was bank 1 ( Bank 0) and the upper section was bank 2 (bank 1) most of the data be accessed is from the lower bank (First 256K)
@@CotyRiddle No sorry, I didn't mean the upper and lower banks, I meant the lower chips in each bank (the RAM is made up of piggy-backed chips, so there are 2 layers in each bank) - I was merely curious if the chips at the bottom of the stack got warmer than the ones piggybacking on them.
@@geezerdiamond Oh that I am not sure about. And thank you for not getting all offensive like some people do on here.
the bios is the OS that adjusts the system .. learned this threw graphics cards.. just by putting a little heatsink on your bios chips you can increase the performance of your GPU.
LOL! Overclocking so we can run Planet X3!
Sega Tera Drive just seemed like a horrible idea. Integrating a Mega Drive console into a 286 PC when 486 was available. Why not just go with that?! :-(
Try the same setup with another CPU. May be that the CPU is not correctly working at 10 MHz and that is weird for a 12 MHz CPU. Also, being hot is very weird. May it be that the CPU is somehow damaged? Or is there some another clock signal that makes it behave so weird? (I think at the ISA clock)
I did the original testing with the Harris 20mhz CPU
@@adriansdigitalbasement2 Thanks for your answer.
And what is your opinion? Is the 12MHz CPU faulty or the motherboard makes it becoming so hot?
I had few 286 systems and I overclocked some of them but a 286 CPU to run so hot to not be able to touch it I never saw. Definitely is something strange.
All this pathing sounds like making bios crack 😀
Hopefully you've had your anti-virus "patch" installed also 😁
Come on man, you need to put that Harris 20Mhz part back on that board, get a fish tank, and do a full chilled mineral oil submersion of the board. 20Mhz full board clock would make it extreme.
(Reminds me of a friend who insisted on getting a 486DX50 instead of a DX266, because: "VLB @ 50Mhz == best VLB".)
If this led to selecting expansion cards that can survive the bus clock, then it offers a lot of opportunity. For example, drop in an AMD DX4-120 and instantly run at 150. But it's a pain in the ass to get working in the first place, with cards that don't like being run out of spec. I had a CGA card actually spark and pop and blow off traces from dropping it in a 286-16 with no divider on the ISA bus.
@@mal2ksc LOL... 16Mhz ISA bus - that's outstanding, and not surprising that a CGA card would be less than happy with that situation.
Long term you will boil that MB.
200ns ram is, in theory, fine up to about 14Mhz. Reality, probobly close to 10Mhz.
Heres a suggestion for custom clock generation tool by Atheatos : ruclips.net/video/odcH80wiiPs/видео.html
Those little hestsinks are probably worth more than that board.
Just swap cpu that is known to oc 16mhz too see if its the board or that hot 286 causing instability
Not the processor, they normally ran toasty hot even at 6MHz.
The most important question is will it run gorilla from the dos disk…
Think it was on dos 5.0…
5.0-6.22 on the supplemental disk
I think you need a liquid cooler on that cpu.
I think now that this machine is overclocked, you should try to find EMU386 (a 386 emulator that runs on a 286) to see what the penalty is for trading increased processor speed for some facsimile of real mode. LOL
Now water cool it ^_^
Computer is sketching out because it's overheating
you are way better off using a non ibm bios. i have been running my type 3 mobo with a 10mhz 286 and a 20mhz crystal for 3 years now and never had a problem. I use the AMI bios and yes I do have the 80ns ram on the motherboard too and that made a big difference. I tried the 24mhz crystal back 3 years ago too and had the same issues. I would get garbled text on the screen etc...
Can it run Crysis?
えよよ
you call it "extreme" OC when nitrogen is involved ;-)
The talking hands talk again.........