PS3 Frankenstein PHAT PS3: CECHA with 40nm RSX

[DeadEnd]

Deadend Bro I think he @alty75 pretty much shared his professional analisys, if he said he reball lots of fats with 90nm Rsx and they work for long time with no deffect then he just show us his experience and he May need to bring some scientific text just to Fully prove as an fact that the 90nm Rsx rarely has bump failure and only reballing can solve the problem.

But Still @alty75 prove his point with his logical experience fixing bgas on Fat ps3s.

Bro, what professional analysis? There's no analysis without data, he only said he's got lots of experience and that many of the playstations he fixed worked after reball. I simply asked for some numbers and more details. So what is the problem? And no offense, but I was not asking you. I requested this from him. So you can relax. I am sure he could answer for himself.

 

[Victor Hugo Alvarez]

yes i use webman and observe the temps and found as i said. althouh the other facts about cell may play a role here. but lets go on another famous to fail console the xbox 360 xenon and zephyr boards are 90nm cpu-gpu again gpu issues. although on this console gpu always run hotter, but on 2009-10 there was a 80nm elpis gpu that didnt suffer from defects. also there is a video from the manufacters that shows the issus is the underfill. and again same thing like ps3 reball or reflow was belived to be the cure. hell i have booted a ylod and rrod consoles by simply overheat them. the reball or reflow simply close the gap of the crached solderballs between die and substrate after all the crack is nanometers. now when thay will fail again is a mater how well the crack is cured. and how much the console will be used.

Well I never HEARD of an 80 nm model of 360 that did not Fail!!!!! I know the Jaspers model that use 65nm processors And Sram, and thats the Kronos gpu and I have that Board and I absolutely 100% think that gpu dont has problem cause its an 65nm model but with Standart Heatsink of 90nm models so thats why 65nm model must Not Even touch 80c on its gpu or Bga and for that not Cracking its Bga, so obvious as the Slims launch in 2011 using 45nm and because they have small heatsink and fan runs low they run ultra Hot nearly melting the Plastic and they came back with Bgas Problem in Fact Xbox 360 Slim is one of the most reballed console ever, so theres no Such thing as saying newer gpus will have no bump issue as you sugest.

 

[Victor Hugo Alvarez]

Bro, what professional analysis? There's no analysis without data, he only said he's got lots of experience and that many of the playstations he fixed worked after reball. I simply asked for some numbers and more details. So what is the problem? And no offense, but I was not asking you. I requested this from him. So you can relax. I am sure he could answer for himself.

Ok then Bro but you shall give some credits to somebody like him Who has more than a decade doyng reballing and fine we can ask him to bring some scientific Data....

 

[truemaster]

well ive read and see a lot about bga failures. also i have to add that they are cases that the solder crack between substrait and motherboard and on those cases the reflow-reball will do fine. but the reality is the solder job wasnt perfect (cold). so with the flex of the motherboard solder gaps appear and the rest is known but those cases are the minority. just watch some @Octal450 videos and you will undrestand. he have done a xenon gpu swap with 80nm elpis and he sayed that elpis is free from bga defects

 

[DeadEnd]

Ok then Bro but you shall give some credits to somebody like him Who has more than a decade doyng reballing and fine we can ask him to bring some scientific Data....

Man, I don't take orders from you lmao.. It's especially hilarious because I never accused Alty of anything in this thread. Since beginning I've only been asking for more data and details. Chill the hell out, jesus christ.

 

[Victor Hugo Alvarez]

Man, I don't take orders from you lmao.. It's especially hilarious because I never accused Alty of anything in this thread. Since beginning I've only been asking for more data and details. Chill the hell out, jesus christ.

Chill out? When I order you to do anything? I just said you Should maybe Consider him, Whatever do as you please

 

[alty75]

Bro, what professional analysis? There's no analysis without data, he only said he's got lots of experience and that many of the playstations he fixed worked after reball. I simply asked for some numbers and more details. So what is the problem? And no offense, but I was not asking you. I requested this from him. So you can relax. I am sure he could answer for himself.

There is no scientific data in what I do,It is qty based data only. I am no scientist so dare not apply some guess at a scientific fact lol.
From my history at least 50-80% of reballs work and last.
I have had units come through 1-8 years after a reball with a different issue,such as fuses,IC's caps etc so all in all from my history,i wont use the swear word experience! is that reballs work and last.
I don't throw ridiculous never ending "data" regarding any one fix/repair like a lot of users do with the bump theory that has no data at all!

 

[DeadEnd]

From my history at least 50-80% of reballs work and last.
I have had units come through 1-8 years after a reball with a different issue,such as fuses,IC's caps etc so all in all from my history,i wont use the swear word experience! is that reballs work and last.

And yet again evaded the question on what models of consoles you are talking about and what type of RSX did they have? Here the argument is only for 90nm and BC models specifically. Also 50 to 80 % is very ambiguous data. So could it be closer to 50% or closer to 80 % in the end?

 

[alty75]

And yet again evaded the question on what models of consoles you are talking about and what type of RSX did they have? Here the argument is only for 90nm and BC models specifically. Also 50 to 80 % is very ambiguous data. So could it be closer to 50% or closer to 80 % in the end?

The majority 60% at the least and again BC units as the majority,80% of

 

[DeadEnd]

The majority 60% at the least and again BC units as the majority,80% of

Even with that number that that would still put 40 % that didn't hold up. So yeah.

 

[squeept]

@DeadEnd @alty75 We actually are working on some science, if you want to contribute. I think @RIP-Felix is still working on data collection. Download the spreadsheet from my signature and post all your results on the syscon thread. And if you feel like the sheet is missing anything important that can't be shoehorned in under an already existing heading, just let me know and I'll see if I can work it in.

I've got like forty thousand consoles and handhelds under my belt, several thousand reballs of various devices, maybe six or seven hundred backwards compatible PS3? I don't know, I been drinkin', and my memory ain't so good no more. Results are skewed from recency bias, but my guess these days is among 90nm RSX problems, it's a third BGA, a third bumps, and a third otherwise dead chip (electromigration or some random fluke). It's important to remember that bumps can accidentally (and reliably) be fixed when reballing. The success rate of my reballs is definitely up since I upped my limits on ohms testing.

Bumpgate is real, I don't know what to say there, I'll start keeping a personal, accurate tally on that as soon as I have an extra fifty grand burning a hole in my pocket to get an x-ray machine. The waters are also muddied here for us because I think there are a LOT of BGA issues that are mechanically connected so they don't present as a problem, but when you reball, there's evidence of oxidized pads, so it can lead to misunderstandings of the initial problem.

@Pacorretaco the 65nm return is sitting with Felix, and it's looking increasingly likely that I misdiagnosed a bad CPU BGA.

 

[Pacorretaco]

Ok, finally we are getting something thats not just theory.
Excellent.

I've got like forty thousand consoles and handhelds under my belt, several thousand reballs of various devices, maybe six or seven hundred backwards compatible PS3? I don't know, I been drinkin', and my memory ain't so good no more. Results are skewed from recency bias, but my guess these days is among 90nm RSX problems, it's a third BGA, a third bumps, and a third otherwise dead chip (electromigration or some random fluke). It's important to remember that bumps can accidentally (and reliably) be fixed when reballing. The success rate of my reballs is definitely up since I upped my limits on ohms testing.

So the current guess is that "when there are RSX related problems"...
(Worth remembering that this is only a subset of the 90nm machines that failed, and are no longer working. Many work, and not all of those that did fail, are related to RSX. I understand we are leaving out all the rest of possible problems that are nothing to do with RSX)

So when it is related to the RSX,
Roughly you expect it to be about
33% the big old BGA (reballing would legitimately work and solve the problem)
33% "the bumps" (reballing could "appear to work") (In theory, this is the 33% that the 65/40nm RSX fundamentally dont have. Thats it)
33% something else (reballing would simply not work) I guess it includes stuff like VRAM... Etc

If I am understanding it right,
Seems that the estimate changed slightly since last time, but still sounds reasonable to me. Now this raises some more questions:
How do each of those failures look like?
Are they the same or different symptoms, or slightly different behavior?
Most importantly I see you are making a distinction between "the bumps" and "something else". As if "the bumps" are something that you distinctivly recognize.
How?

As I understand it, In case of xbox 360 for example... the "broken bumps" can be identified because they can be "fixed" or temporarily reconnected by applying extreme pressure to the chips. The famous "bolt mods" and all that voodoo. (Notice I mean extreme pressure. Not really possible with hands or pushing, flexing etc.. Needs to be something like nuts and bolts)
Example: a 360 was successfully reballed but it just failed again with similar problem.
However, placing extreme pressure like a dodgy bolt mod, makes the thing work! It cant be the big old BGA because it was already reballed. So it must be the broken bumps. No longer in quotation marks. The chip needs to be replaced.

Now, do we see this in PS3? Notice how I am not trying to deny "bumpgate" or saying it cant be real in general or anything. Also I cant.
The only thing that I am asking is:
How does this translate to our PS3 in particular?
Because I dont think it is precisely the same. Thankfully we dont see PS3 bolt mods for example. But maybe is because they would not work? It would only work if the failure actually were "broken bumps".

So what is your way to know when you find that 33% of "broken bumps"?

Because I'm not so sure they are actually breaking... As some people mentioned before, often in the cases when there is the funny internal damage, it is an internal short between some of the data lines. Not really what you would expect from "broken bumps"... If anything, it kinda sounds like the opposite.

And thank you for the answer

 

[DeadEnd]

33% the big old BGA (reballing would legitimately work)
33% "the bumps" (reballing could "appear to work") (In theory, this is the 33% that the 65/40nm RSX fundamentally dont have. Thats it)
33% something else (reballing would simply not work) I guess it includes stuff like VRAM... Etc

This sounds reasonable indeed. Especially with those numbers I don't understand why are you trying to promote the idea that they are most likely not dead and reballing should be done first in hopes that it maybe fell into that 33%, while there is a 66% chance that it actually is internally damaged. I mean, there is clearly no point in doing that anymore when a better alternative is available. The only reason would be to save money.

 

[Pacorretaco]

I mean, there is clearly no point in doing that

Yeah yeah, Ive heard that before but you are missing the point again
You dont want to do it? Great man dont do it. Is up to you.
But to arrive at those numbers he had to reball, reball a lot. And wait, did he say somewhere that he stopped reballing because it is pointless? No, thats only you.

The success rate of my reballs is definitely up since I upped my limits on ohms testing.

What he actually said is that his reballs are most successful than ever thanks to a better understanding of the situation. "That" was the point all along. Just trying to understand better what is going on, so that everybody can make better decisions. All win.

Saying that is always "the bumps" and never the balls doesnt help. Saying that is always the balls and never the bumps also doesnt help.
Reballing is still not a myth. In fact in his own %, both the "bumps" and the big old balls are given an equal chance.

But the discussion hasnt ended. There is still more room and things that can be understood better than they are now.

 

[alty75]

Even with that number that that would still put 40 % that didn't hold up. So yeah.

Oh you've misunderstood me! So if I say 60% were repaired with a reball and the other 40% was repaired with a different issue. This would be over a qty of about 500 BC units at least. It is most likely much more, i've never kept a tally but i do know over the years the average was 2-3 a week times x 670 weeks = ?
There would be an average of about 5% that were not repairable.
As for longevity of the repair. ‍ I cant speculate but at a guess the majority lasted at least 90 days as that is and was the ebay/paypal recoup your money period. So to have very little claims i would guess it pinpoints they worked fir at least that period of time and or longer and to this day. Ive seen many 5/8/10 year reballs still going string and had customer feedback years after the fact,when they return for some other console relair

 

[DeadEnd]

About the supposed "40nm/65nm superiority"...
Yeah maybe you could make a prediction that a 40nm/65nm chip would probably last longer than a 90nm.
But is not so easy to predict which "board" would last longer overall.
The new chip may be a couple years newer, but the boards are still 16 years old and have many other components.
You dont know what will fail next, or if the "underfill" will have a chance to actually matter or not.

I just think that the truth must be somewhere in the middle...
Again I could be wrong like any human being. But I know I am right in asking.

Here you agreed that 40nm can and most likely will outlast 90nm, yet you immediately tried to manipulate the point into the fact that "the board itself may not last long". Completely irrelevant statement because the question was not about the board. The points I made concerned only RSX's lifespan and reliability. The board faults are a whole other discussion and if some of them would need more maintenance, they will need them regardless of what RSX is installed. That was manipulative of you to attempt to distract from the main point .

Yeah yeah, Ive heard that before but you are missing the point again
You dont want to do it? Great man dont do it. Is up to you.
But to arrive at those numbers he had to reball, reball a lot. And wait, did he say somewhere that he stopped reballing because it is pointless? No, thats only you.

What he actually said is that his reballs are most successful than ever thanks to a better understanding of the situation. "That" was the point all along. Just trying to understand better what is going on, so that everybody can make better decisions. All win.

Saying that is always "the bumps" and never the balls doesnt help. Saying that is always the balls and never the bumps also doesnt help.
Reballing is still not a myth. In fact in his own %, both the "bumps" and the big old balls are given an equal chance.

But the discussion hasnt ended. There is still more room and things that can be understood better than they are now.

I'm not missing anything and you're twisting my words again. I never said he shouldn't have reballed. I said it doing it now for 90nm chips when a better alternative is available is not the most logical choice anymore. Promoting such idea with a small chance that it's alive is also not reasonable anymore unless you're trying to save money. It's simple common sense. You even contradict yourself without noticing. You just said that according to squeept there is more evidence that those chips are dead rather than they are alive. Contrary to what you've been claiming before. Now you basically agreed that it's the opposite and what do you do? Well, once again you are attempting to shift attention from the main argument by focusing on what squeept said about reballing... Except that's actually not what he said. He never said it's "most successful than ever". He wrote that the success rate is up because he started doing ohms testing to eliminate the low ohm chips. From my understanding he has become more selective in his reball choices, which does make sense. But I still don't think he would be promoting to keep trying to save 90nm by any means necessary. Again, unless money is in question, you would normally want the customer to choose the better option, to install parts with most reliability.

And Paco, I think I get it. You just need to be right even if you're wrong. Because as you sort of admit that you may be wrong , you are still trying to spin the points in your favour by any means necessary. And I think alty75 is doing something similar by trying to shift focus on his experience and other claims, which are mostly beside the point, while deliberately avoiding my questions and providing very vague and inconclusive data . See below

Oh you've misunderstood me! So if I say 60% were repaired with a reball and the other 40% was repaired with a different issue. This would be over a qty of about 500 BC units at least. It is most likely much more, i've never kept a tally but i do know over the years the average was 2-3 a week times x 670 weeks = ?
There would be an average of about 5% that were not repairable.

Well, if you didn't keep a tally on it so the data would be highly approximate and not very conclusive then. Out of 60% that you reballed, how many can you say with absolute certainty had a genuine BGA fault? Considering there is at least 33% chance (this is an assumption but let's play with that) some of them could have appeared to work by reball. It's not easy to know.

As for longevity of the repair. ‍ I cant speculate but at a guess the majority lasted at least 90 days as that is and was the ebay/paypal recoup your money period. So to have very little claims i would guess it pinpoints they worked fir at least that period of time and or longer and to this day.

Yea, but that is the issue. 90 days is not enough to draw conclusions. Some reballs can reactivate internal damage and will work for several months. Like you say, at least for a period of time. After that, all bets are off. Why the customer didn't return? Plenty of reasons and therefore it's difficult to make any conclusions...

Ive seen many 5/8/10 year reballs still going string and had customer feedback years after the fact,when they return for some other console relair

How many though? Again, not enough conclusive data. But of course , some of them will last. We were just saying about approximate 1/3 of all...So it would fit in line within that theory.

 

[Victor Hugo Alvarez]

This sounds reasonable indeed. Especially with those numbers I don't understand why are you trying to promote the idea that they are most likely not dead and reballing should be done first in hopes that it maybe fell into that 33%, while there is a 66% chance that it actually is internally damaged. I mean, there is clearly no point in doing that anymore when a better alternative is available. The only reason would be to save money.

Hey well a point to do that, to try save some 90nm is that you Will be saving 1 console more cause if the 90nm is trashed then you lost its gpu and then you Will have to "Damage" an slim by removing its Rsx, of Course slim can have another gpu installed But soon as the gpus will be over as people is Maybe Teoretically Trashing and good chip that would be the 90nm.

 

[squeept]

Again, we don't have a way to clear this up here without x-ray. What we have are educated guesses based off of the data available, and we have proven science in failure analysis from other devices and chips that we can apply to our assumptions. There's a lot of guesswork, sure.

In super general terms, I call low chip ohms a bad die, I assume things that die again during stress testing or early warranty period a bad bump (dry reflow of lead free isn't an ideal solder joint), and obvious oxidized / damaged pads, good ohms, and a successful lasting reball is BGA.

My data set is small so far with Frankenstein, I've been slow on this front for a year or so, but again, post your spreadsheets in the syscon thread.

I'd take the smaller process size all day every day. Lower TDP -> slower electromigration -> longer time before inevitable die failure. No bumpgate, lower potential running temperature -> far less bump failures. Lower potential running temperature -> slower / less BGA failures. And frankly, even though they may only be a few years newer, I'm still counting those couple of years as a big check in the positive column.

 

[squeept]

Hey well a point to do that, to try save some 90nm is that you Will be saving 1 console more cause if the 90nm is trashed then you lost its gpu and then you Will have to "Damage" an slim by removing its Rsx, of Course slim can have another gpu installed But soon as the gpus will be over as people is Maybe Teoretically Trashing and good chip that would be the 90nm.

Since I only work on backwards compatible models, that doesn't bother me at all to sacrifice a slim to save a CECHA01 system.

edit: and by upping the standards for selecting 90nm chips to reball, I'm putting less rework cycles on backwards compatible boards, so it's saving a few extra of those from the dumpster.

 

[DeadEnd]

I'd take the smaller process size all day every day. Lower TDP -> slower electromigration -> longer time before inevitable die failure. No bumpgate, lower potential running temperature -> far less bump failures. Lower potential running temperature -> slower / less BGA failures. And frankly, even though they may only be a few years newer, I'm still counting those couple of years as a big check in the positive column.

That's what I've been saying as well. But according to Paco and the rest, it's not logical because 90nm can still be alive. So you must hang on to it like your life depends on it lol.