PS3 Frankenstein PHAT PS3: CECHA with 40nm RSX

Octal450 said:
This is spot on. I have used this exactly to successfully diagnose the issues with very high reliability (95%+).
Kind Regards,
Josh
Click to expand...

With an xbox 360, which has an error asociated solely with bump failures. So in your case it was possable to distinguish between them.

If we apply pacos reasoning, 'just because it was the bumps in the 360 GPU or other Nvidia GPUs doesn't mean the same defect applies to the RSX.' Likewise, just because the bolt mod / pressure test worked to distinguish between bumps and BGA in the 360, doesn't mean it would for the PS3.

The PS3 design is unique. I have said before I wonder if there ia tortional strain placed on the bumps and interposer do the RSX die being place at a 45-degree angle to the rest of the chip. Most chips don't do that. 360 didnt.

So it may or may not be fair to compare them. We don't know.

Pacorretaco said:
I did not ask for "proof". I just asked for some "evidence" AT LEAST, which I would still like to see... Otherwise are just incomplete stories, sorry...
Click to expand...
Why don't you accept the myrads of examples of "evidence" I have already supplied?

No, you do keep asking for "proof." Not proof in the strict scientific sense that I keep saying is a dirty word in science (because it's a burdon most science cannot live up to). No, you keep pushing infomercial science. "Tests" that wouldn't control for confounding varibles. Confounding varibles that would prevent you from drawing valid conclusions.

If you purpose a valid test, I'm willing to listen to the results. But that burdon is on you. You can buy the consoles and do the testing, at your expense. Or if I come across one I'll do it. But I'm not going to risk it if I don't believe it would be usefull.

If and when you do perform a test and draw your conclusions, I will be happy to peer review it.

My issue with that test is that you say that 'it cannot be a BGA because it was reballed,' but that's not a safe assumption. You are also implying that the bump defect would not be temporarily reconnected by the change in pressure, or tin melting, like I was talking about before, or other varibles I'm not thinking of.

In your case, the boltmod pressure would also reconnect the BGA. So if the BGA failed again, the same issue with this test applies. Without a 3D xray, there's no way to know for sure.

In other words, we don't get the benefit of knowing how genuine bump faiures behave. Whether or not the heat test, or pressure test is suficient to overcome the underfill support. Whether Octal's experiance applies to the PS3 like it does for the 360, or not.
PS3 doesn't have an error code like the 360 has to distinguish between.

We don't have a way to know one way over the other.

All we do have is Nvidia's track record during the time the RSX was made. And that was not good. Does it apply specifically to the the RSX? No. But that's still good circumstatial evidence! And that's enough to build a case for bump defects. There is no need to provide more "evidence" of them, especially when those "test" would not yeild useable results.

Now it's time to build a case refuting the bumps. And all those tests have the same problem. They would not yeild useable results.

So from where I stand, we're at an impass. The the only "evidence" we have is circumstantial. That is enought to make the defacto theory ATM that bumps are a contributing factor. And we should believe it until there is a competing theory. Not a bunch of hypotheticals. Or purposed tests that no one has actually tried, and wouldn't mean anything if they did.
 
@RIP-Felix Absolutely Not, Decom pressing the Bumps or Slight pressure while delidding would Not Make the Bumps Cracks to be fixed for a time as you said, thats is absolutely ridiculous, youre kind of BELIEVING in the BUMPS Failure Theory, Decompressing an Eletrical connection will nearly always Decrease the Path of Eletrical Energy on it.
 
Pacorretaco said:
Felix I get it, nobody has time to argue.
But maybe our case is a bit different? After all it was you who made long videos and all, which I like too.

I did not ask for "proof". I just asked for some "evidence" AT LEAST, which I would still like to see... Otherwise are just incomplete stories, sorry...

And maybe the test that @Victor Hugo Alvarez suggested was incomplete because it was missing a step but...
There was another proposed test that did not involve expensive X-rays. Should not that difficult, especially according to the stories.
That was my original question which I will quote again:
Click to expand...

Yes Pacorretaco, that test of yours should be great to diagnos an Bump failure and my test should work too if we use líquid metal on die.
 
I'll play devils advocate a bit. Let's say for a moment Josh is correct and the following applies to the PS3 as well...

I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.

I listened to this song while typing this message. It seemed fitting.
 
Last edited:
Geez, we're revisiting the TOKIN thread from 2 years ago when everyone refused to believe me when I said just taking the console apart and putting it back together was enough get it flexed a micron or two and working.

I haven't been keeping detailed logs long enough to run in to any of the "tap 'em on the top and they'll artifact" ones since, get back to me in a few years and I might have an opinion with data to back it up. My current opinion (computer engineer that fixes consoles for a living) is that all signs and evidence point to an out of the ordinary amount of both bump failures on 90nm due to poor manufacturing practices and BGA failures on all nm due to newly lead free processes. The larger processes also drink a whole heap of juice, so even if they're kept cool, with lead balls, and have good bumps, they're still eating themselves alive from electromigration.

I honestly think it's nuts that we're debating that flexing things around in any direction could or could not have an impact on mechanical reconnections, no matter the scale. Things get warped and flexed in all three wavy dimensions, so changing any amount of pressure will add or relieve stress "randomly" in any direction. I'm doing a little carpentry right now, and I'm ripping some 2x4's lengthwise that are perfectly straight. Once I rip them, a lot of them end up warped and twisted like crazy immediately after because I've relieved internal stresses. Sure, these things seem unlikely, but they happen nonetheless.

Anyway, someone mentioned the PS2 - those chips last forever because they pretty much don't have BGA failures since they had shiny lead balls, and they don't get hot enough or sink enough juice to eat themselves alive. I used to do a lot of those, and I'd frequently run in to dead / fully clogged fans in Slims and they would chug along just fine. I think the heatsink in the last 90000 models was like 1x1x1cm. Honey badger don't give a F%$#. edit: and they were made without bumps.. those would have been wire bonded?

Though I just don't care in the end. None of us have the equipment to prove anything, and it doesn't change my workflow/methods/warranty - I don't care how the chip died, I just care that it's dead. I don't care how it got fixed, as long as it is reliable. At this point, I'd say move the argument to an entirely different thread so this can just be about Frank.
 
Last edited:
RIP-Felix said:
I'll play devils advocate a bit. Let's say for a moment Josh is correct and the following applies to the PS3 as well...

I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.

I listened to this song while typing this message. It seemed fitting.
Click to expand...

Great music, Funny song nane.
 
RIP-Felix said:
I'll play devils advocate a bit. Let's say for a moment Josh is correct and the following applies to the PS3 as well...

I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.

I listened to this song while typing this message. It seemed fitting.
Click to expand...

I told Ya Felix, also Felix for what I know when I tryed to rip off that glue in lots of silicons I dyscover that thing is Titanium Solid so I think pretty much is nearly impossible to have variation of Pressure inside that bumps by just Moving and Puting a Little pressure in the board as the Glue in the Silicon is Extremely Solid, i mean i try remove it with sharp Blades, try remove it with Real 500c and that Thing wont come off, im talking about the underfill glue as you said Felix.
 
RIP-Felix said:
I'll play devils advocate a bit. Let's say for a moment Josh is correct and the following applies to the PS3 as well...

I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.

I listened to this song while typing this message. It seemed fitting.
Click to expand...
I appreciate that friend, but theres no devil around here, I hope.
So you can put the devil aside and put the scientist goggles back on.

The concept applies to anything that has balls attached to a board. If light pressure/flexing change the behaviour... It points to a precarious BGA that should be reballed. But it does not say anything about the "bumps", as they are not equally susceptible to these tests. (Because of geometry and because of their underfill) . Thats the difference and why tests like that could be useful.
And yes whe have seen that in PS3 too. Some were caught on video as well, not just 360, and not rare at all.
For example people often "fix" their problems by simply disassembling and assembling, then blame the "tokins" or the "thermal paste" or whatever else they think they fixed... Etc.
But it points to the BGA.

RIP-Felix said:
I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.
Click to expand...
Again please put the devil aside because thats not what the test says.
The light pressure test can only speak about the BGA. Thats its blessing and its curse.
If 4 of them responded and 3 didnt, it doesnt say much about the last 3...
What it does point is that the first 4 have a legitimate BGA problem (not the bumps) and need reballing.

Because as you rightfully say, the test doesnt "rule out" BGA failures. It can only "confirm" them.
The thing that can "rule out" BGA failures is not that. It is the reballing.
(Could be dodgy/broken pads... Debris, who knows, or the break is too big or old that was too oxidized over time)

About the last three... Well you can make guesses as always but if you want to "confirm" the "broken bumps", the process is different, and was the thing that I still didnt see in these machines, and would like to see at least once.
"Rule out" the BGA by reballing, then do a very different kind of pressure test (strong, compressive bolt mod) that now could affect the bumps, and only the bumps. (Because the BGA was already ruled out by reballing)
I hope it makes some sense
 
@Pacorretaco my first frankie that been reballed by dead end. had gone through multiple reflows reballs because he didnt have a proper frame and at the end balls have merged 2 or 3 times. my original problem was ylod with syscon error 3034 and after 2 months after geting it back with 40nm it was glod and syscon error 2031 (thermal censor) the system then has been sent to victor and in the end rsx had damage because of the multiple reflows-reballs and the motherboard pads on the rsx are suffer from delamination. that wasnt 40nm problem it was a bad service. but lets go back to 90nm or 40 nm or reball the 90 nm. i know all the argues are about a permant fix. but i still insist on replace with a working one or newer one in design as 40 nm is
 
Hello friends,
I made a short clip of my Frankenstein PS3:

As for reball vs replace, my operation has always been the same for years. I conduct some pressure tests and the results can tell you with 90-95%+ accuracy whether the chip needs a reball or a replacement. Unless I see obvious evidence a reball is needed, I replace the chip. Saves me time and money from having to do it again in the future, and if it is for a customer, much much much less risk of a failure due to failure months or 1 year later.

Kind Regards,
Josh
 
RIP-Felix said:
I'll play devils advocate a bit. Let's say for a moment Josh is correct and the following applies to the PS3 as well...

I just went over my spreadsheet of SYSCON errorlogs and found 7 consoles that had 3034 in which people applied a pressure test to them, as Josh did above. 4 of them booted. 3 of them did not respond at all. If he is correct, then those 3 consoles had Flip-Chip failures (such as bumps).

No-one contests the BGA can fail. But using this simple test's logic and taking it at face value, then Bumps are part of the story too. But as I said before, I don't believe the pressure test rules out BGA failures.

I listened to this song while typing this message. It seemed fitting.
Click to expand...

Again I just want to apreciate you Felix, Thank you so much for your hard work on bringing the codes and all data and all job to make 40nm RSX swap possible on ps3 fats and slim, is Really Amazing and you super amazing!!!

I cant wait to install one 40nm Rsx on my Fat ps3 Cok-002.

I'm just arguing so much on swaping the 90nm because I dont want the 90nm to be trash cause maybe theres chance it could still be alive, maybe bumps are 50% and Bga 50% maybe something Like that just like ~ the Xbox 360 fat Xenom.
 
Xbox 360 Fat Very rarely have BGA issues. Almost all of them are defective chip - even Microsoft admitted this was the cause and this why they didn't just reball them to fix them, they replaced the chips with re-engineered ones from a different fab until the main TSMC supply was sorted out.

You really can't compare X360 to PS3 is my point.

Kind Regards,
Josh
 
well, and even if the problem were the bumps, is there even a way to reliably replace the parts? If they're dead, you might as well replace the whole chip, and at that point you might as well put a 40nm in. I think the difference is mostly for science than for any practical use.
 
marciolsf said:
If they're dead,
Click to expand...
Well, that was the question all along.
Are they dead? How do you know? And how do you know it is or is not "the bumps"? Not everybody understoood it.
We know they are not always dead, and is still hard to say how often they are or not and why.

One of the things we can do, that came from all the previous noise is the light pressure test. If the machine respond to light pressure, flexing of the board, disassembly/reassembly or other manual forces or tamperings... It almost certainly is not dead. It points to the big old BGA that can be legitimally solved by reballing. This is often the case and is not "the bumps".

If it doesnt respond, well... Things get a lot more uncertain. It doesnt mean much and it still doesnt say anything about "the bumps", as they are not affected equally by the same forces. Also doesnt "rule out" the BGA... Depending on your priorities you could still reball for more certainty, but the test vaguely points away from the BGA.
Now... I was hesitant to bring this up because it can be confusing and get misinterpreted. But there is a next voodoo step that we can perform.
Now involving some tricky heat. Not too much, the famous "hairdryer trick" and all that...
Often it works. But why?
The heat can affect many things...
It will expand and then contract the materials, also warp the board if the heat is uneven. All this could again affect the big old BGA... but how likely is this?
At this point we have to make some shaky assumptions.
If we did the manual pressure/flex test and it made no change, why would the thermal expansion/warping of the board make the change now? Yes it is possible, but other things are more possible too.
Maybe the precarious connection affected by the heat was actually inside the chip, and thats why the manual pressure did not work.
The famous "broken bumps" as in the 360. Possible explanation but still not confirmed... The "broken bump" still hasnt been found, hasnt been isolated. Only assumed as always.
Because even then, the heat could have done other things as well...
Most notably internal shorts have been found (the opposite of "broken bumps") that open again with some heat. Often in the data lines, between each other or to ground.
Electromigration is another thing that affect all chips in the world, but I am not too sure how this can behave with heat.
The VRAM could possibly be affected by the heat as well, or could be something related to moisture, who knows...

So now the reballing could work, but these tests would be pointing that it was probably "not" the big old BGA and chances are, it wouldnt be solving the real underlying problem then.
Now just need to figure out what we are trying to do...
If just trying to repair the thing, decide what is more practical and be happy that we can do the replacement at all.
But if trying to "blame the bumps"... Make generalizing theories, etc... Need to do some more steps because as far as I'm aware, "the broken bump" still hasnt been found even once, like it was found in the 360. Let alone know how common it actually may be. Currently is mostly guesses and speculation.

(Edit: used spoiler to make it easier for those that dont read all, to not read all)
 
Last edited:
marciolsf said:
well, and even if the problem were the bumps, is there even a way to reliably replace the parts? If they're dead, you might as well replace the whole chip, and at that point you might as well put a 40nm in. I think the difference is mostly for science than for any practical use.
Click to expand...
Yes, you have to replace the part! Impossible in any sensical way to repair the flip chip bond.

Kind Regards,
Josh
 
lol, should have known I'd kick this storm up again :).

yeah, the big if is that we just don't know for sure one way or another, I agree with you there. We (as a community) don't have the tooling to know for sure, nor do we have the tooling to replace it. I'm all for figuring this out, but from a repair perspective, replacing the whole chip is the way to go.
 
You must log in or register to reply here.

Similar threads

Back
Top