Tomb Raider (1996) better with write-back cache and why is my CMOS battery discharging? (Part 2)
HTML-код
- Опубликовано: 28 июн 2024
- Part 2 of a quest to assemble a socket 3 system that will be able to run Tomb Raider at an acceptable frame rate. Today, we are looking at another 486 DX2-66, but this time, the level 1 cache is capable to operate in write-back mode. This cache strategy is supposed to be superior compared to write-through cache, but in some of my tests, I got results I did not expect...
Oh, and I am fixing an issue on my Soyo motherboard: The CMOS battery is flat after a few weeks - let's fix this along the way too!
▬▬▬▬ PCBWay ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Sponsored by PCBWay: 🔸pcbway.com/g/lF253I
PCBWay provides full feature custom PCB prototyping services, 3D printing, and much more. Check them out using my referral link above and get a 5 USD welcome bonus when you sign up as a new customer!
My contribution to the Community Project Space at PCBWay
🔸www.pcbway.com/project/member...
▬▬▬▬ Affiliate Links ▬▬▬▬▬▬▬▬▬▬▬▬▬
Titanium Alloy Tweezers Professional Repair
🔸s.click.aliexpress.com/e/_Ddj...
7-45X Continuous Zoom Stereo Trinocular Microscope
🔸www.aliexpress.com/item/32880...
AliExpress: Mechanic Solder Mask
🔸s.click.aliexpress.com/e/_Ddy... (this is the exact item I am using)
Solder Mask from Amazon
US:
🔸amzn.to/43K8KOy (6pcs)
🔸amzn.to/470ISRw (4pcs + UV light)
DE 6pcs:
🔸amzn.to/3Dr49WU
🔸amzn.to/43Av6lC
🔸amzn.to/44XIbGB
FNIRSI HS-01 Soldering Iron
🔸s.click.aliexpress.com/e/_DE4...
Flux for easy soldering
🔸s.click.aliexpress.com/e/_DFk...
InfiRay P2 Pro Thermal Camera
🔸xinfraredx.com/products/infir...
▬▬▬▬ Support Bits und Bolts ▬▬▬▬▬▬▬▬▬▬
Patreon:🔸 / bitsundbolts
Bitcoin: bc1q2dvf9flf78lam33qjj9w9xgvfxdgre7g9x3rmw
Twitter/X: / bitsundbolts
Facebook: / bitsundbolts
▬▬▬▬ Timestamps ▬▬▬▬▬▬▬▬▬▬▬▬▬▬
00:00 Intro
00:49 Intel 486 DX2-66 WB
01:46 Disaster strikes
02:19 PCBWay
02:53 Undo disaster
04:46 Drop-in-socket test
05:00 Configure CPU and initial tests
05:51 Soyo's discharging battery
07:12 Diagnosis
11:06 Tomb Raider
11:56 Benchmarks
14:06 Conclusion 
Развлечения
Great video and repair. Sorry for slightly off topic, but I have also problem with CMOS memory on one of my newer retro board and don't know where is fault.
My Asus P4P800 Deluxe randomly loosing CMOS settings even with new battery. Weird is, when this happens, also clock in BIOS not running.
Any tips what can be wrong here? 32768 kHz crystal, some component in CMOS circuitry or fault in chipset? Thanks for help.
I think Necroware had a video where the time is not working properly or stops after some time. I don't remember if it was related to a clock generator. Your problem seems to be something I have no advise for at the moment. But I'd start with testing the crystal and maybe replace the BIOS chip. Is it always happening or is it a random occurance?
@@bitsundbolts It happens randomly. Currently it last whole week without problem . I thought it fixed magically itself after start using it again (it was long in storage without battery) and build new retro PC with this board and now it happens again.
I don't know where to start debugging your issue, but I pinned this comment. Maybe someone else will be able to help. If I come across a similar issue, I hope that I remember your comment
@@tomashasek7588 If there's an intermittent problem it's usually going to be a bad connection. It's strange that the clock doesn't advance when in BIOS, the clock portion of the ICH must not be running. Did you make sure the CMOS clear jumper is in the normal position? If missing it won't work reliably.
@@bitsundbolts I also had the problem with the crystal, it had a bit of corrosion from battery leakage from another board that was sitting on top, one of the crystal legs was gone and the system wouldn't boot to MSDOS, I could only go to the BIOS settings and saw the clock time go crazy fast.
For framerate analysis off the video capture, "trdrop" should do. It can actually recognise partial screen updates/tearing.
I'll definitely check it out! I have seen other channels use this tool. Very nice graphs! Once I get the additional CPUs fixed up, I'll run the benchmarks and try to use the tool to analyze the footage! Thanks!
I'd be interested to see how it handles footage captured from an analogue source like VGA since I've only seen it used on direct digital captures over DVI/HDMI before
Great video and excellent root cause analysis of the defective capacitor! Thanks also for testing those two DX2, something I wanted to do for long time already 😊.
Thanks Peter! Hope you're doing well. I miss your videos!
Yea, WB will become more beneficial as the multiplier increases. The 2x on 66MHz one is not enough to make WB a clear winner. The on-board cache is still relatively fast enough.
With 3x (100MHz) or 4x (133MHz) the difference should be bigger as the difference between L1 and L2 cache speed is bigger.
I was surprised to see it perform lower in Doon compared to write-through. I don't think it's due to my testing method because there wasn't much to it. I'll definitely revisit this topic in a dedicated video.
Good job soldering that pin back on! My hands would definitely be too shaky to get it anywhere near straight, so I would have to use a helping hands or something, and I'm not sure how well it would hold such a tiny pin.
As I'm quite a Cyrix fanboy, I'm already looking forward to your next video!
I'm working on it. I already started straightening the pins of those two CPUs... Unfortunately, they are not as nice to work on compared to Intel and AMD.
You can test the current draw on the battery with a probe on top of the battery and the other pushing the side clip away so it's not touching the battery, ie, the current is now going through the DMM. Modern MBs are about 4-6uA.
Yeah, that is a way to verify that your board draws too much current from the battery. You still need to find out why though.
@@bitsundboltslet's just use a 6v motorcycle battery and call it a day. 😅
Nice repair. It is very satisfying to me to see scrapped processors (and other hardware) working again.
Measuring the in circuit resistance of capacitors is a valid diagnosis technique. However knowledge of the circuit being measured is needed. For instance capacitors on the supply rail for the CPU can measure very low resistances in circuit because the CPU presents a low resistance load. Thankfully with a socketed CPU its easy to remove and measure without the CPU in circuit.
Yes, I have seen many other RUclipsrs talking about low resistance on silicon chips (e.g. VGA core / CPU core). Specifically important when one would do voltage injection since a short may just be the CPU at low resistance. I still have a lot to learn, but I am getting there slowly!😃
Interesting performance results!
Great repairs on the CPU and motherboard.
Thanks!
This is what I found regarding caps: "Values under 1nF generally are C0G dielectric... Color is generally greyish for C0G and brownish for other dielectrics".
Your previous cap was light brown, so it was larger than grey one.
Thanks for sharing. So, I guess something in the range of 10nf? I was not aware of the color coding of caps, but good that you mention it - thanks!
@@bitsundbolts yes, I think 10nF should be good, but it is hard to say for sure. Probably making schematic for that part of PCB can help to better understand what this capacitor is used for and, consequently, which values are more suitable for such task.
You have an amazing YT channel, thank you.
Thank you very much!
They also sell ml2032 battery that is rechargeable used in dreamcast I use that sometimes to avoid messing with battery circut.
Nice catch on the faulty MLCC! My 386 board also drains its battery quite quickly. I figured that's why they used to have rechargeable batteries, but maybe it suffers from a similar issue. I'll check for it next time I get the opportunity to dig it up and play around with it. It is surprising to me, how unreliable MLCCs are. I also had one on a modern laptop not too long ago (that I was able to pinpoint with a borrowed thermal camera, before I won the InfiRay P2 Pro in your giveaway, woo hoo! 🕺). One would think that as "solid state" devices, they would last indefinitely. I suppose they're just too brittle for that though.
In anticipation of this video, I watched your pin soldering video (which previously somehow eluded me), and the sound effects in both of them gave me a good chuckle. 😂 Reminded me of the groan Peter from CPU Galaxy let out when he dropped that CPU on the board. 🤣
Regarding the performance comparison, I knew that the difference between WT and WB would be minuscule, but I'm already a little surprised, that WT is faster in some cases. I suppose that you could get the best of both worlds with the WB model though by reconfiguring the mainboard for WT. Might be a fun little project to add a button on a case that rearranges the jumpers accordingly. I definitely would have done that for an additional 1,4 FPS in Doom back in the day. 😄 That is, if I am right in assuming that both CPUs behave exactly the same when the board is configured for WT?
How about both with passthroughs
I heard that Faber Castell Finelead Pencils (These pencils with removable tips) should work good for straigthening bend pins without breaking them, as you apply the pressure more evenly
excelent work
Thank you!
I enjoy your fixing videos the most ^-^
Glad you like them!
@@bitsundbolts it might just be me but repairing these maschines and giving them a second chance at life instead of them landing on some landfill is 10000 times more rewarding to me then accually using and playing the maschines themselfs haha. I am more of a console repair guy myself but I still love to see the PCs get saved from an early grave
I have a similar feeling towards those old motherboards, CPUs, and what not. Last time I defended a few old boards from scrap metal hunters (looking for valuable metals in those boards and CPUs). They just wanted to get the gold from this old hardware - luckily I got there before them!
@@bitsundbolts Yeah having seen so many pentium pros beeing melted down on youtube breaks my heart x.x Yes recovering the rare metals is great but making these CPUs rare to find and super expensive in the process c.c
good stuff keep it
Thanks, will do!
What a fantastic job, what a great video. I have thoughtlessly disposed of all CPUs with broken pins so far, what a fool I am.
Thank you! Ah well, you're definitely not the only one who was careless with old hardware. I wish I kept some of my stuff - even the broken stuff... 😔
I never de-soldered a pin from a CPU but it might take quite a lot of heat - and pre-heat - if the pin is thermally connected with something. Did you try soldering iron AND hot air? Great video and amazing that you found the bad capacitor! Nasty little one, only partly faulty!
Looking forward to the next chapter!
Yes, I did try to get the pin off with a soldering iron - no luck. Either those ceramic CPUs can take a lot of heat away so the solder doesn't melt, or they are just differently attached compared to Pentium IIIs. I am sure I would damage the CPU die before any pin would come off.
With the 486 build I did last year, I had playback issues with digital sound samples in some games. A strange crackling sound. You could easily hear it during the digital sound test in de Descent game setup program. Turns out the WB cache on my DX2-66 CPU was the issue. Setting it to WT in the BIOS didn't change anything. I had to replace the CPU with a actual WT cache only version. Then the issue was solved. I don't know if the WB cache was bad or something on the other CPU but as actual improvement in games and benchmarks was minimal or non existent with WB cache, I just use the WT cache version CPU permanently and no more sound issues 😅 PS: It took me days to finaly find the WB cache was the rootcause and no sound card, motherboard, RAM or any other type of hardware problem 😅 Perhaps you try and test it using the digital sound test in the Descent setup? 😉
Wow, that must have taken a lot of time to figure out that it was related to the cache! That would be at the very bottom of the list to test - if it's even on the list! I guess it was more like a lucky coincidence that you found out your sound issues were related to the cache!
Yeah, I have two 486 boards that do the same thing, I installed the Varta replacement from Scrap Computing with the CR2032 holder and Schottky diode, the battery gets drained in a couple of days. Need to take some spare time to check the charging circuit of these boards.
Hope you'll be able to find out what is causing the drained batteries - well, hopefully it's similar to what was the issue in my board. Good luck!
@@bitsundbolts well, it seems that I found the problems. One board had a bad capacitor like yours, installing a new capacitor solved the problem. The other board was different, it lost the BIOS settings after some minutes but the battery drained really slow, for my surprise the negative pad on the motherboard for the battery was not connected to the rest of the board ground plane, maybe because of the previous battery leakage that I cleaned but the damage was already inside the layers. Anyway, I soldered a wire from that pad to the keyboard connector ground pad and now it keeps the BIOS settings. Every old PC board is a new adventure 🤣🤣
Great to hear that you were able to fix both boards!
That are interesting results of the different cache strategies. Perhaps Intel did some internal magic to match the performance of these two different cache controllers and if there could be any undocumented control registers to unelash a better performance in write back mode.
Also from my daily experience, I would recommend to order some 100nF and 1uF caps, because these are the most common values used. The 100nF is in my opinion the best compromise for voltage stabilisation and filtering. 😉
It may depend on the situation. Some tests performed better with write-back, but I expected games to be slower on write-through models. Well, Doom disagrees. Maybe it is not as noticeable on those lower clocked models. Thanks for the capacitor value suggestion. I will replace the cap soon with a value in the range of 10-100nF.
Interesting ... it's one of the issues I've encounter with my 4SA2 as well, but never cared enough to fully investigate. I'll make sure to check it some time in the near future and report back.
I'd be interested to know if this is issue is spread across more boards.
@@bitsundbolts Seems to be the case by my observations, as C5 also had a 2 figure kiloohm reading. Original value seems to be along the lines of 4.7nF (measured 4.4nF out of circuit). Sadly my board stopped working recently (it's either a busted RAM slot or battery damage taking it's toll and I can't be bothered to fix it at the moment).
Hm, there are definitely more caps failing in this spot than usual. Hope you can fix your board some day!
@@bitsundbolts Turns out not only CPU pins can bend, but also whole boards leading to contact issues with the RAM slots, got it fixed for the moment and will check back in 3-4 weeks if the board still remembers the settings. Installed a fresh battery 3 weeks ago and I'm already down to 2.8v
A leaky MLCC SMT cap. That’s a new one for me. Good find! Maybe it was cracked.
And by some responses, I am not the only one where the same cap has issues. Maybe some design flaw or just wear and tear.
If would be cool to see how hot the chip is running while running the game. I don’t think i486 had temp sensors but you could run a probe between the chip and heatsink
These boards have some sort of SMD capacitor plague. I have this board as well and the broken one was the capacitor next tot the PWRGD line of the 496 resulting in it not booting. Removing the capacitor and glitching the line, to make sure the 496 starts after the 497, resulted in a boot. Replacing it with a 10nF cap resulted in a proper boot sequence.
Interesting. I had a feeling that there were more caps bad in this are. If you look closely, I also replace another cap in the bottom right (second row) corner. This one also seemed to have some leakage current - but by far less than the one I mention in the video.
😳😯
Original value is 4.7nF on a 4SA2
Cool! Thanks for letting us know!
What kind of wooden stick are you using to straighten the pins?
Toothpicks and cuticle pushers
Want to see tomb raider on 486 33-40 MHz + 3dfx. I dont know why but i need it ))
Haha, I also have an SX-25 if you want to go down further! I may try that just for "science" 😉
I struggle to find a use for a 486 platform though. What can it possibly do that a pentium or something newer can't do though? It's too fast for 80's dos games too
It all depends. Sometimes, it is just the connection one has to this platform. A 486 was my first PC - and I skipped the Pentium until a PII-350.
I am curious about the BIOS SCREEN MESSAGE "REV WA0911 128GB BETA" Is this a modified BIOS version for use with 128 gb hard disks?
Yes, this is a modified BIOS I got from "The Retro Web" site.
I guess that the capacitor was not really needed, you may have it removed without changing the funcionality of the CMOS. Remember that the original circuit had a rechargeable battery and when powered on, there was the main 5V with its ripple and all the spikes generated by the digital ICs. It is normal to use a filtering capacitor there (my guess is that the value was from 100nF to 1uF) but when you removed the 5V for charging, there was no need for that capacitor anymore. Maybe I am wrong, please correct me, because I do not know for sure how is the circuit further down from the point between R4 and C8 (maybe it goes to a battery voltage sensing circuit and that means a comparator, a 339 maybe, as I saw them on some older motherboards).
The board worked well without the capacitor - you are right. I just wanted to show how the values are when there is a properly working component on the board. It shows that there shouldn't be a such a big voltage difference across the resistor. It's all about discovering the issue and it's good to know what one should expect. Then it is easier to solve the problem. Regarding the value of the capacitor - I have no idea. I took another capacitor off which also seemed bad, but I didn't mention it in the video. This capacitor had a value of 500pf - which could also be wrong. As you said, those caps may no longer be needed since the charging circuit is now disabled.
@@bitsundbolts You are right, due to the age of the components, they tend to change values and this not only for capacitor but some resistors I found that can change value more than 20%. Your troubleshooting was impecable and understanding the basics of electronics (voltage divider, Ohm law, etc) is a must.
I hope that this video will help to keep working many motherboards.
Thanks! I really hope that I explained it well enough. This is one of the rare cases where I successfully troubleshoot a circuit. I want to say that I am getting better at it. I am also still learning and such little successes are vital to stay motivated!
at least it's broken off in the corner. Makes it easier to replace.....not that it's easy to replace. lol
Odd to see a failed bypass cap.
Yes, that is weird. But it goes towards becoming a dead short I guess. Assuming that me taking it off reduced the resistance even more, I could imagine it to become shorted some time soon. This is then when nothing works, or your battery / powersupply causes issues due to overheating or protection triggering.
I wonder how an optimized OpenLara would run on these...
I wonder if those will run at all. Aren't those ports meant to be working on newer hardware?
@@bitsundbolts They're a reverse engineering of the original codebase, so I'd imagine you could adapt it back to run on DOS again. Probably compile it with DJGPP and go from there...
I mean it _is_ running on Sega 32X and the GameBoy Advance, so there is some hope.
It sounds interesting, but I'm not sure I'll find time to dig into this material. Maybe someone else is going to do this some day
You R "The Classic CPU Whisperer"....
🥲 Hehe, that's pretty cool!
odd I saw a video from jayztwocents. He tried to repair a modern Ryen CPU with a couple missing pins. I don't recall if he got it or not, but he found that removing pins was easy. Putting them back on wasn't too difficult. He didn't have to break any off. Just add heat and they pop right off. It's odd you couldn't do that. The pins on the ryzen look similar to this CPU. They kind of look like tiny nails.
I really tried - they definitely do not come off. I even burnt the warranty sticker on the other side - so, there was definitely enough heat to melt solder.
@@bitsundbolts I wonder why the newer CPU is more repairable. lol
It’s OK to drain batteries for Intel CPUs :D
Well... The capacitor was the culprit... After removing it the battery stopped discharging and the bios settings were saved! This was a thorn on the side for many months! I haven't replaced the capacitor as i understand it was a part of the charging circuit and therefor not necessary.
I have tried a wide range of processors with this motherboard, from 486dx 50 to amd P5 133. I have to say that I never had me a cyrix to try it, so looking forward for the test with the cyrix chip.
Another strange thing I noticed is that I used 2 simm's of 70ns edo memory and the board worked! The thing is that the board only worked with those 2 and not with any other edo I tried. I am 100% certain that the simm's are edo.
The board is definitely not a performance monster. My shuttle hot 433 runs circles around it. But the 3 slot types spanning all the phases of 486 evolution are there.
Got me thinking, though, what can be done with unlocking bios features... Would it be terrible for me to let you bite the bullet first...? 🤣
Maybe stick a VLB graphics card in there and see the difference, I always wondered what the results would be by using cards with similar chips.
Than you for the solution and can't wait to see if I am missing by not having a Cyrix.
Wow - I am really surprised that you seem to have had the same issue! Happy to hear that this video could help you!
I just finished the pins on the Cyrix CPUs and both seem to work! Now I need to find out how they work and how to enable all their special features...
Regarding the memory issues you have - I did run into trouble which was documented in the Pentium OverDrive video. My Soyo board does not boot with EDO memory (at least the ones I have). It actually boots into the Boot Block Recovery BIOS. But nothing seems to work properly and I wasted a few hours on this one! But, I want to modify a 60ns EDO memory and turn it into FPM compatible RAM. I only have one stick of 32 MB FPM - I think this is a good reason to try if that works - similar to the 30-pin SIMM modules.
Ohh, and the BIOS - well, I also have something planned regarding this one... Found some details on a German forum where they talk about swapping the BIOS of this Soyo board with another board's BIOS that is way more capable of setting timings. So, that is coming soon too.
And the VLB card is still pending. I need to go over it and check if I can see any obvious damage. Anything I pull out of those bags at the scrapyard needs to be inspected properly to minimize preventable disasters!
You are very welcome - and I hope the next video will answer your question of missing out or not on Cyrix CPUs! Have a good one!
Why are RUclipsrs always begging for likes or subscription? 🤷♂️
Well, there is a lot of work going into these videos (which creators provide for free). A growing number of subscribers and likes are metrics that encourage people to create content to their best ability. Without growth or an audience that engages and shows their appreciation, creators would be less interested to put up with the effort required. Reminding viewers that they are able to show their appreciation is one way to build a larger community and therefore encourage creators to make great content.