PS3 (Research/Experimental) - NEC/TOKIN Capacitors Replacement - YLOD

I had some trouble with mine having HDMI handshake problems a couple of times during testing. Wouldn't show up when I set it to HDMI, but would work with AV to my CRT. Either my TV or my HDMI switch were remembering the PS3's previous settings and were expecting that, but it wasn't working. I had to change inputs (use a different port) and use a different HDMI switch to get it to refresh the setting. I have no Idea what HDMI non-sense is going on behind the scenes, but once I changed the HDMI environment sufficiently it worked. Ridiculous!

I knocked 3 of the tiny bypass caps off of my CPU and it still worked fine. I can't find the schematic for your board revision, but it may not be crucial. You might try cleaning your HDMI port and inspecting the pins on the motherboard to be sure they are making contact. Sometimes times the strain of plugging in the cable over and over again can brake the connection. Some flux and an iron can fix that so long as the pads aren't destroyed.
 
HDMI output stopped working. I can only output image by AV cable

It will almost certainly run fine with a few of those little guys knocked off, I wouldn't worry about that.

Try resetting your video a few times. Then check the HDMI chokes. The four little black components with 4 legs each between the Panasonic MN864709 and the HDMI port. Each pin should have continuity from the Panasonic side to the connector side, and be open to the adjacent pin. If that's not it, there's a small chance the Panasonic chip is dead, and a small chance you have a GPU BGA defect, but it's more than likely that your GPU is dying.
 
I had some trouble with mine having HDMI handshake problems a couple of times during testing. Wouldn't show up when I set it to HDMI, but would work with AV to my CRT. Either my TV or my HDMI switch were remembering the PS3's previous settings and were expecting that, but it wasn't working. I had to change inputs (use a different port) and use a different HDMI switch to get it to refresh the setting. I have no Idea what HDMI non-sense is going on behind the scenes, but once I changed the HDMI environment sufficiently it worked. Ridiculous!

I knocked 3 of the tiny bypass caps off of my CPU and it still worked fine. I can't find the schematic for your board revision, but it may not be crucial. You might try cleaning your HDMI port and inspecting the pins on the motherboard to be sure they are making contact. Sometimes times the strain of plugging in the cable over and over again can brake the connection. Some flux and an iron can fix that so long as the pads aren't destroyed.


Well, you were correct :O

I've switched the hdmi port and voila it worked :O But I can only output 1080i, 1080p doesn't work don't know why. Maybe it is a limitation from the second HDMI port since it is named as HDMI2/DVI. HDMI1 worked in 1080p. Tried to went back to HDMI1 and didn't work.

Thank you very much for the reply and the tip :)

Only the CPU temps are worrying me now. It goes hot very fast.
 
Does your fan ramp up to a loud level within a few minutes of booting? If so, you need to delid and replace TC as you CPU is overheating and will certainly hasten a YLOD if you don't.
 
Does your fan ramp up to a loud level within a few minutes of booting? If so, you need to delid and replace TC as you CPU is overheating and will certainly hasten a YLOD if you don't.

More and less. I've replaced the thermal compound just today when I changed the caps. After some minutes of gameplay, I launched control fan unit and CPU temps were at 71ºC and RSX 52ºC
 
My scope is also on the way (it is some noname model but it should be good for this diagnosis). too bad I already desoldered all the NECs before I could take a before shot of the power rails with the scope. Glad to see @squeept already did that so I have something to compare against.

Would it also be possible for someone to take a measurement on the in and out legs on one of the NECs and subtract the out from the in so we could clearly see how a good and a bad filtering can look like? And to do an FFT on the in and out legs?
 
Last edited:
You weren't able to boot into the PS3 menu to get those voltage readings, right? You mean that during the 1-2s it takes for the PS3 to discover the error and shut down (YLOD) you saw voltage spikes on your...what...multimeter?

No menu, as it YLODs on me very soon, about 2 seconds after power on. The measurement was done with a simple multimeter, still waiting for my scope to arrive, but the fact that it blew a cap that was rated at 6.5 volts, and there was definitely no soldering mistake, led me to believe it was over voltage because of the unconnected feedback leg.
 
More and less. I've replaced the thermal compound just today when I changed the caps. After some minutes of gameplay, I launched control fan unit and CPU temps were at 71ºC and RSX 52ºC
The answers to these questions were not clear form your response:
  1. Did you Remove the Internal Heat Spreader on the CPU? Just replacing the TC between the heatsink and IHS is not enough!
  2. Are you using a Fan controller and thermal probes or webMAN mod on a jailbroken PS3? If you're using thermal probes, they can be inaccurate, but usually on the low side. If you're using webMAN what is the setpoint? For example, mine is at 65C and a lowest fan speed of 25%, assuming if it can maintain that temp without having to ramp up the fan - which it usually does. If fact to maintain 65C the fan usually ramps between 28% and 35%. If you leave it on SYSCON, the fan wont even kick in until it gets to 75C, so the temps you're seeing are "normal."
 
PS3 #2 - Part 3: I've been Hoodwinked!
(...continued from Part 2 here.)

Yesterday I removed the rest of the NEC/TOKINs and installed 18x 270uF, 2.5v, 6mOhm ESR TaPol caps. Total capacitance = 4860uF matching SONY's specifications and the ESR is 0.333, which is better than the NEC/TOKIN array. I used 3x 20AWG solid core conductors for the + rail bridge this time (larger than PS3#1). While the soldering is a bit sloppy, it's electrically sound. The resistance after installation was 2.8 - 2.9 Ohms +/GND, indicating they are installed correctly without shorts. Pictures:
A9ORZpd.jpg
GfXoIh5.jpg
g53WCWt.jpg
Yi5qZRS.jpg
Vf09Y17.jpg
Result = YLOD, no change.

Discussion:
Originally I replaced only one NEC/TOKIN on the bottom side of the board (for both the CPU and RSX):
fgJGRtR.jpg
That worked great and was super stable during the first few days of testing. I was able to jailbreak the console and install webMAN mod to get fan control away from the SYSCON. I tested NBA Live 2010 for 1 hour. No problems. Then I download/installed all the updates to Gran Turismo 6 which took all night (it's a seriously HUGE DL). The console was on for about 18hrs strait. The next day I played GT6 for a few hours, temps were completly under control (<70C). I concluded that the console was fixed and shut it down.

Over the next few weeks I may have turned it on a couple of times, but it mostly sat unused. Then I picked up a PS2 title cheap from a local thrift shop. I thought, "I haven't tested a PS2 title yet." So I poped it in and within 2 minutes the console YLOD!

So that brings me this weekend. If the BGA was fine and the NEC/TOKINs were the problem, then replacing them should have fixed the PS3. But it didn't! The EE research I've done suggests that 4800uF is important for the best performance of the circuit SONY engineered (COK-001 motherboard). Theoretically, I should not need more capacitance. @squeept's oscilloscope measurements suggest this as well. Although his probing technique could have been better, they do show more noise above and below 4800uF. That seems to confirm the theory. So I highly doubt that adding more capacitance at this point would make a difference.

The only way forward for me now is to reflow/reball the RSX. This makes two consoles that have failed shortly after trying the "Tantalum Fix." While yes, I confirmed the tantalum "fix" temporarily made the YLOD disappear, so far in my experience it has not lasted more than a few power cycles. This may be discouraging news, but my piddly two experiences, @squeept's many experiences, and the evidence we've gathered so far is consistent with the YLOD being primarily a BGA problem. The idea that, "we have all been assuming the BGA was at fault, when all along the capacitors were the culprit" is a tantalizing narrative, but that doesn't make it true. As a scientist, I'm obligated to suspect it's a red herring.

Continued on Part 4 here...
 
Last edited:
it blew a cap that was rated at 6.5 volts, and there was definitely no soldering mistake, led me to believe it was over voltage because of the unconnected feedback leg.

Aside from the electrical characteristics, ESR is very important in this application because these high frequency circuits can cause the capacitor to generate a great deal of heat. If the ESR is too high, it will absolutely burn itself up. If you got your caps from a good source, and you got a low enough ESR, then I don't know. But, if you got them from AliExpress or a similar shop, it's highly likely that the specs are incorrect, either from poor quality control or outright intentional fraud.
 
The answers to these questions were not clear form your response:
  1. Did you Remove the Internal Heat Spreader on the CPU? Just replacing the TC between the heatsink and IHS is not enough!
  2. Are you using a Fan controller and thermal probes or webMAN mod on a jailbroken PS3? If you're using thermal probes, they can be inaccurate, but usually on the low side. If you're using webMAN what is the setpoint? For example, mine is at 65C and a lowest fan speed of 25%, assuming if it can maintain that temp without having to ramp up the fan - which it usually does. If fact to maintain 65C the fan usually ramps between 28% and 35%. If you leave it on SYSCON, the fan wont even kick in until it gets to 75C, so the temps you're seeing are "normal."

1. I only changed the thermal compound between heatsink and heat spreader. To be honest I have some fears removing that since it is kinda glued and I can easily damage the console. The same applies to the RSX side but like I said here it is better the temps.
2. I use the Control Fan Utility Homebrew. Dont know if it is the same version but here is a video example (not mine)
 
Too low ESR can be an issue as well in some cases, especially in applications like buck converters (and thus VRMs) or switching power supplies. So the goal should be to actually exactly nail sonys original specifications, including ESR.

About the BGA cracks: In many cases, it's actually not the BGA which is at fault, but the connection between die and substrate. These TINY flip-chip solder balls are much more prone to breaking than the BGA balls, and this was a huge problem during that era (since having so much TDP on such a tiny area was a pretty new thing around back then and manufacturing wasn't where it is today), not just the PS3 and the 360. Since this type of defect also responds to the pressure test and might temporarily disappear after heating (even if it's lower than the melting point of the solder due to different thermal expansion followed by some tiny deformations), we can not really know. But this was a huge issue, especially with NVIDIA and AMD GPUs at that time, including PCs and notebooks.

I also doubt BGA cracking was the actual reason for the 360 failing (the official statements only talked about solder cracks, but not about where they happened), since in that case they might simply have changed the solder and added some supports behind thr PCB or changed the PCB material to avoid deforming, which they never did (it would have saved them MILLIONS if not billions). Instead, the following revisions had the exact same issues which only appeard later (that is, after the warranty) due to improved cooling, and ROD was only permanently fixed with the all-65nm consoles (Jasper and Jasper v2/Kronos) which produced much less heat.

For us, this doesn't really change that much, though, no matter if it's the BGA balls or the flip-chip balls.
 

This reminds me that I need to save up for a nice inspection prism.

Every 90nm PS3 GPU I've lifted for a few years now has visual indicators of BGA defects, usually in the same corner every time. I've rotated the board every which way, tried different heat profiles, and watched on thermal camera to make sure it's not anything in my method that could be causing it. To me, it's a given that there are BGA defects, even on a perfectly functional console. So the real question when one stops working is "what ELSE is wrong?"

I dump bump failures under the general heading of "dead GPU" because that's just beyond both my skills and the machinery I can afford. I watched a lady grind off a die with CNC and replace it before, and I was just in awe. I'm sure I've accidentally fixed plenty while reballing, but I'm very generous with my warranty to compensate, and they rarely end up coming back.

I think the PS3 has so many BGA failures because the chip is just so big (it's almost 45mm across while the 360 is closer to 30mm), so it's prone to have more uneven heating across the substrate, so thermal stress is exacerbated. If I had to guess, the bumps fail less often because the die is smaller and likely heats more evenly. Combined with the underfill, they may not flex very much at all. Whereas the substrate is super hot near the die, but the edges and corners are relatively cool.

I don't do nearly as many 360 since they aren't worth very much, but there is a far lower incidence of visual evidence of BGA defects. I do remember in the early days of the BGAmods forum, there was a fellow there with access to an x-ray machine. He threw a few boards in and did find BGA failures, not bump failures, then he disappeared. It was a small sample size, and the forum is long gone, so I may be misremembering. But, from what my own two eyes have seen, the 360 GPU is more likely to be toast.
 
1. I only changed the thermal compound between heatsink and heat spreader. To be honest I have some fears removing that since it is kinda glued and I can easily damage the console. The same applies to the RSX side but like I said here it is better the temps.
2. I use the Control Fan Utility Homebrew. Dont know if it is the same version but here is a video example (not mine)
That explains your temperatures. The Themal compound on your CPU is exhaused, dried up, gone, dead...not working! This is causing your CPU to overheat, increasing the temperature differential on the BGA and die bumps, which will quicken a YLOD. YOU NEED TO DELID. Make a CPU delidding tool and it will be easy to get off:
eGXWbUm.jpg

pfWkQW9.png


Here is a video of NSC doing it. Skip to 1:03:46 for the relevant bit. I don't like to use his method to remove the RSX, I prefer to heat it and remove the normal way. But to each their own.
You need to grind, file, or sand the thickness down. The tip MUST be very thin. There is very little space for you to get the tool under the edge of the IHS. A razor blade is too thick. A painters knife is too thick. Also very important is that you sand the bottom edge to "blunt" it. This is the edge that will be in contact with the substrate and could dig into the solder mask and destroy traces. If it's blunt or rounded and you come in with a flat angle, you won't have to worry about the traces. The sharp edge will cut the silicone along the IHS and force the blunt edge up onto a cushion of silicone. With braced steady pressure, you can easily cut through the silicone and remove the IHS in under a minute. Making the tool is the hardest part.
 
Last edited:
A painters knife is too thick. Also very important is that you sand the bottom edge to "blunt" it.

I like the painter's knives :( I need the nice big handle to avoid slipping. Half of them go straight in to the trash, though, since like you said they're too sharp or too thick. I don't like trying to sand them, once that crappy metal starts to get roughed up, it seems to try to resharpen itself constantly. I use one of the long thin ones to cut the CPU silicone, then I use one of the short fat ones to protect the substrate when I jam a screwdriver in to pry off the GPU IHS.

A quick test for whatever tool you go with is to drag it forwards and backwards on a sheet of paper on a flat surface. If it doesn't glide totally smooth, it's no good.
 
That explains your temperatures. The Themal compound on your CPU is exhaused, dried up, gone, dead...not working! This is causing your CPU to overheat, increasing the temperature differential on the BGA and die bumps, which will quicken a YLOD. YOU NEED TO DELID. Make a CPU delidding tool and it will be easy to get off:
eGXWbUm.jpg

pfWkQW9.png


Here is a video of NSC doing it. Skip to 1:03:46 for the relevant bit. I don't like to use his method to remove the RSX, I prefer to heat it and remove the normal way. But to each their own.
You need to grind, file, or sand the thickness down. The tip MUST be very thin. There is very little space for you to get the tool under the edge of the IHS. A razor blade is too thick. A painters knife is too thick. Also very important is that you sand the bottom edge to "blunt" it. This is the edge that will be in contact with the substrate and could dig into the solder mask and destroy traces. If it's blunt or rounded and you come in with a flat angle, you won't have to worry about the traces. The sharp edge will cut the silicone along the IHS and force the blunt edge up onto a cushion of silicone. With braced steady pressure, you can easily cut through the silicone and remove the IHS in under a minute. Making the tool is the hardest part.

I guess I need to do this then to increase life span of my PS3. But strangely why it only happens at CPU? I never changed the RSX.
 
I guess I need to do this then to increase life span of my PS3. But strangely why it only happens at CPU? I never changed the RSX.

The RSX heatspreader is manufactured and mounted in a completely different way, and it's also a matter of luck. There are also consoles where RSX is running extremely hot while the CPU is fine (without ever delidding it).
 
The RSX heatspreader is manufactured and mounted in a completely different way, and it's also a matter of luck. There are also consoles where RSX is running extremely hot while the CPU is fine (without ever delidding it).

So my question is: should I make for both or only for CPU since this is worse?
 
So my question is: should I make for both or only for CPU since this is worse?
Do both when you're in there. Be sure to watch videos on getting the RSX off. Please note, the thermal adhesive on the RSX is too hard at room temp. A hair drier for 30s or so will soften it and it'll come away with much less prying force. Replace with high quality TC, I like MX-4. Once done you won't need to do it again for many years, but if your console outlives the new TC (4-5 years for MX-4), it will be easy to re-apply since you wont have to delid ever again.

SONY should never have used Internal Heat Spreaders. They should have done like they are with the PS5 and gone strait heatsink to Die. Speaking of the PS5, I don't think the liquid metal is a good idea. Unless they've solved the problem of it "drying out", then it'll stop working within 1.5 - 2 years. I had this happen to my PC after using liquid metal. After about 1.5years the stuff was hard and not liquid.

About liquid metal. RANT WARNING:
I've read that it forms an alloy with copper, slowly soaking into and discoloring it. It doesn't destroy it like it it does Aluminum, but soaking into the metal slowly does present a problem. Unless it's saturated, or they use a special metal that doesn't form an alloy, it'll "dry" out. Also, all liquids evaporate well below their boiling point! Even metals. Mercury evaporates into the air at room temp, for example. So, even if they prevent it from forming any alloy it'll still slowly evaporate. At least it's easy to replace, or so they made it seem in the tear down.
Yasuhiro Ootori didn't really address this except to say that the 'decision was necessary cool the APU, given it's high thermal density' and that they spent "over 2 years adotping this liquid metal cooling mechanism". He did use marketing buzz words like "long-term, stable, high cooling performance." However, unless they tested the APU with liquid metal for 2 years under load, which they couldn't do because the technology hasn't existed that long, alpha/beta hardware design changes in the manufacturing process prevent you from having finalized hardware for long term testing, and other logistical constraints. In short, they don't know for sure! These claims are always based on simulations, computer and real world test benches. An accelerated quality assurance load wouldn't work, as time is an inextricable factor in the evaporation rate of Liquid metal. So, there's no way they can properly simulate this. Only time will tell. Just like the PS3 and PS4.

What really has me worried is that the temperature density of the APU "required" them to use liquid metal! That means that this thing will place a serious stress on the FCBGA and bumps of the SoC. If and when the TIC becomes less effective, or the heatsink gets clogged with dust, BGA defects can occur even quicker than they did for PS3/PS4. FCBGA is just bad technology and needs to be abandoned! Hot chips in video game consoles is inherently going to limit their lifespan. Unless console manufacturers maintain backwards compatibility or forward port their game libraries, the modern era of gaming is doomed to never become retro. That's because the consoles that can play these games will not last long enough for the games to become retro. Unless new hardware is made to play theses games, or emulation can keep up then all the hardware capable of playing these games will be e-waste.

No problem, you say? "The games can be downloaded! So when the PS6 come out, you can just DL your PS5 games." You're assuming the PS6 will be backwards compatible, that SONY doesn't de-list your games, and that there will even be a PS6 at all. Maybe they go all subscription, like Valve. Playstation Plus and XBOX Game Pass is all you need on whatever device you have that's fast enough to work. SONY and Microsoft will be out of the hardware business and in the subscription business. Then you games rely on you paying in perpetuity to play. You own nothing, just how they like it. That's the future of gaming!

Meanwhile NINTENDO is sitting back with it's cool running low TDP consoles that don't generally have overheating problems; they're laughing all the way to the bank while MS and SONY beat each other to death and loose consumer confidence in their overpriced hardware. "If I want PC gaming performance, I'd upgrade my PC." For the price they charge you can get a graphic card that'd beat a PS5 anyway. The days of consoles outperforming PC's in purpose built applications is long gone. The SSD tech isn't that impressive and PC's will not lag behind it long enough to pay the early adopters tax.

Nintendo took a different strategy altogether and I'm inclined to agree that staying away from high TDP at the cost of performance is the wiser move. Portability, strong characters, and great games is sufficient to weather the storm. Once MS and SONY have converted all their followers to the subscription service fever dream, Nintendo if free to continue gouging it's fan base with locked down hardware, overpriced games, toys, collectables, and anti consumer tactics. Namely, artificial scarcity to justify never lowering prices. Business as usual, but without competition in the console market.

Rant over.
 
Back
Top