Delidding the RSX
(CTE and the IHS Matters)
A mismatches in CTE can lead to stresses that can warp the RSX package and transfer strain to the BGA/Bumps. Eventually these thermal cycles cause a crack to form and propagate. As it does, it increases the impedance of the solder joint (the resistance actually increases). This can manifest in strange console behavior, like freezing and artifacting.
I think it has to do with the FlexIO calibration stage (BitTraining), where impedance is calibrated. The process attempts to center the "eye patten," but with dodgy solder balls the impedance can change with temperature as the package warps. So yeah, you could get weird function, artifacts, freezing, livelock detection, deadlocks, checkstop errors, etc.
We've seen consoles with 3034 error (confirmed BGA defect) preceded by 1001 errors (CPU VRM). That's where this hypothesis stems from. I'm not sure it rises to the level of Theory yet, but it explains this observation.
The point is that CTE's of different materials lead to thermomechanical sheering forces that will eventually break the solder joints. The higher the change in temperature (Delta-T), the faster that happens. Eventually the crack propagates fully and the BGA breaks completely (instant YLOD). You can see the steps the entire industry has taken to mitigate this. I went into more
detail in this post.
We already know this. That's why we delid, replace thermal paste, adjust fan curves, etc. My point is that we should be gluing the IHS back on to stiffen the RSX package and reduce the warpage as much as possible, to keep the RSX lasting as long as possible. The question I want to address today is what adhesive should we be using?
- CELL = High temp Silicone Gasket maker
- RSX = I'm not sure there's a good alternative to the one Nvidia used. There are thermally conductive 2-part epoxies like this one, but it may be too strong to ever get off after it cures. So this is a one time deal. They do make a fast cure version of that, but it hardens in 15 minutes, which may not be long enough to apply and assemble the console. I suggest the slow cure one so you can assemble the console and have contact pressure while it sets. That should harden in the correct position to minimize strain. But I have not used this adhesive! I have delidded before but only recently recognized the need to glue it back on. So I'm looking for a good adhesive too. I remember @squeept say what he uses, but for the life of me I can't find the post. EDIT Found this though...
Sounds like he uses a silicone based thermal plaster,
like this. But I agree with his assessment that it's not the same stuff as SONY used and isn't ideal. Not sure what is. I did find
Heatsink Plaster, but none of the comments say what the consistency is after curing (hard like the SONY plaster, or rubbery like the more common silicone based stuff). Maybe I should get a tube and find out.
For this reason I don't recommend delidding the RSX! It almost never overheats anyway. We've never actually had a console show an RSX overheat error code (1201).
The 2-part epoxy I linked in that reply (8329TFS) cures at room temp in 4 hours. That's good, but it's glass transition temperature is only 9C. Meaning it'll get rubbery above that temperature and not be solid. The biggest issue I see is the CTE of 164 (ppm/deg-C). What you want is for all the materials in your RSX sandwich (package) to have the same exact CTE (Coefficient of Thermal Expansion). This means there will be no difference in how much they expand and contract, so no warping stresses and no cracked solder balls or bumps. But that's not how real world materials work.
So what we want is a glue with a high glass transition temp and CTE closest to 20...
9460TC from the above list is the closest. The only issue I see with it is the material has to be heat cured. It takes 2 hours at 80C. I don't think that is an absolute minimum tho. It should cure on it's own over the span of a few weeks of normal use. I actually like this idea, because it allows the thermal paste to burn in, the IHS to settle in place with the Heatsink, and the package to shift/adjust into a natural position with a few thermal cycles. It probably will prevent the IHS from locking in strain.
But the console will need to be warmed up and used during that time. Otherwise the MB will need to placed on a jig (to keep it strait) and baked with contact pressure on the RSX IHS to cure the adhesive. That's if the delid is done by the technician for a customer. It really should be cured before the console is shipped.
https://www.dropbox.com/s/8tcefhfuaffplgs/IMG_4683.jpg?dl=0 on the board but the positive sides are still intact, when i added the tantalums the PS3 would show green and immediatly red then blinking red here are some pictures
