PowerBook G4 Titanium (400 MHz) cache chip replacement

[Ivan Holmes]

Hi all,

I have a PowerBook G4 Titanium 400 MHz with failed L2 cache. When I try to boot into Mac OS 9, I get an error message to this effect. I can't remember the exact message, and I have taken the machine apart now so I can't easily find out, but I don't think it's important in itself. Apple System Profiler shows the machine as not having L2 cache (it should show as 1 MB).

On trying to install Mac OS X Tiger, the installation fails due to some cache-related error, so I am pretty sure that the problem is to do with the cache. The only other thing I could think of was that the cache interface on the 7410 CPU has failed, but I feel that this is less likely.

Examining the logic board I can see that the cache is implemented by two Samsung K7A403600M-QC20 chips. It runs at 200 MHz which makes sense for the speed grade used here. I have included a datasheet for these chips (although it only mentions older, slower speed grades). The IS61LPS12836A-200TQLI, which is currently available new from Mouser at a reasonable price, seems to be totally compatible. I have also included its datasheet in case anyone else wants to take a look.

I have been considering replacing the cache chips to see if this will fix the issue. I appreciate that the 400 MHz model is the worst of the bunch, but getting hold of any TiBook at all in the UK is difficult/expensive now.

Before I take the risk of moving forward with this project (I am pretty confident with hot air rework but any project of this nature has risks!), I have a few questions:

1. Will the machine care if the SRAM chips are of a different brand (I assume not)?
2. Is it possible/likely that the SRAM chips would fail?
3. Is there any possible advantage to installing the 250 MHz version of the ISSI chips?

Please let me know if you need more information or have any thoughts...

Ivan

 

[herd]

I think the problem is more likely to be a bad solder joint. If you can get the board set up in a way that you can test it on the bench, a common diagnostic procedure is to press on the suspect chip or lightly flex the PCB. The SRAM pins are exposed and you may even be able to see a bad joint or a pin that moves when poked gently. If pressing down on the BGA CPU fixes it then I would suspect a bad joint there.

If you do decide to replace the SRAM chips there are several options, including an upgrade to 2MB.

Good luck with the project and let us know how it goes.

 

[Ivan Holmes]

Thanks for the ideas. It is quite difficult to run the TiBook board outside the case because the power button is attached to the case. It may be removable but my case is already very weak so I'm worried about damaging it.

For this reason I can't really press on the CPU or cache chips. The CPU is pulled towards the heatsink by two bolts through the board so in a way that's pressing on it already. I managed to press on one of the cache chips, the other is obscured below the heatsink so I can't reach it. I also tried flexing the board. Unfortunately none of this lead to any change.

I have just finished reflowing the chips with a hot air gun. I could see the solder on the cache chips melting so I'm pretty sure that worked but I don't think I got the G4 hot enough. I was using air that the soldering station claims is at 380 ºC - I'm a bit wary of going hotter. After all this there is still no change! I think I can rule out a bad solder joint on one of the cache chips now.

Just for the sake of it, I tried to enable the cache using Powerlogix G3/G4 Profiler and it caused the system to freeze instantly - which is no surprise I guess. I have attached some screenshots of how this shows up in the OS and G3/G4 Profiler, just because this issue hasn't been documented much online.

I'm thinking I have to heat up the CPU more to make sure the solder has definitely melted. I guess that it being in a ceramic package might help it to take a bit of extra heat. At work there is a preheater I can use, which I think will help to get the solder balls to melt.

The prospect of getting 2 MB cache is interesting... I wonder how much difference it would make. EDIT: Just seen this: https://68kmla.org/bb/index.php?threads/experiments-with-g4-external-backside-cache.47085/

 

[Ivan Holmes]

I think I know where the problem is. After cleaning the CPU and inspecting it closely, I noticed a discoloured patch on the die (see image, I know it looks like it could be thermal paste but it definitely isn't - I cleaned it with isopropyl alcohol and then acetone just in case). This area of the die corresponds to the part of the chip where the cache connections are, according to the MPC7410 datasheet. Based on some other threads I've read (can't remember exactly which), this makes me think that the die is damaged and that's why the chip can't talk to the cache.

It seems that my only option at this point is to replace the CPU. There are some (500 MHz, so an upgrade) on eBay from China for about £50. They look alright from the photos and the price seems realistic. Are there any other sources for this kind of thing?

 

[obsolete]

Hi @Ivan Holmes, I think I actually have what you need! I bought a 500MHz TiBook logic board on eBay a while ago to harvest its Uni-North and KeyLargo chips. I also removed the CPU and cache chips, but I don't need them. I can't guarantee that the parts will be functional since I bought the board as-is and never tested it myself, but I assure you the price will be low. Send me a private message if you're interested. I could also probably reball the CPU for you.