VOGONS


First post, by lowlytech

User metadata
Rank Member
Rank
Member

I am trying to nail down a socket 7 pentium build and so far I am having trouble. I have 3 boards and each has a flaw that's a dealbreaker. Out of the 3, I am back to the Asus vx97 board that fails memory tests only when L2 cache is enabled. If i run cachechk it will show 512k working cache and has no issues during execution when L2 is enabled, but goldmemory and microsoft memory diagnostic have tons of failures. The system runs great with L2 disabled. Any ideas on what to try to fix this or what causes this? Is it indeed defective cache chips? I have swapped power supplies, tried 5 or 6 different pairs of memory, changed the refresh values in CMOS, swapped processors to change the MHz host from 50 to 60, tried flashing BIOS to latest version I could find, all with the same results. I even pulled 4 capacitors by the cache memory area suspecting them to be out of value and they were in spec, but changed out nonetheless.

Anything I am missing on the troubleshooting side of things?

Thanks...

Reply 1 of 5, by lowlytech

User metadata
Rank Member
Rank
Member

I went ahead and cleaned all the chipset and cache chips with deoxit and scrubbed them down with a toothbrush and I still get memory errors with L2 enabled. The only thing I know of now is to maybe go over the area with a heat gun and flux to maybe reflow solder around those areas but I am not too hopeful this will change anything.

I just got an additional socket 7 board from ebay so hopefully 4th time is a charm with this and I can finally get the pentium build tentatively finished from a hardware standpoint. Learned a few things in the process, don't blindly buy a board just cause it looks to have an intel PCIset, and coin battery. The last board I got from ebay had no brand name but was advertised as soyo, had a nonflashable bios, but I guess that didn't matter cause I couldn't find any documentation for what bios version it even used. It would only detect hard drives up to 2.1gb, no cdrom boot option, and fake cache soldered on the board. I thought all that nonsense stopped with the 486. Of course that board worked like a champ with no glitches, other than my hard drive i want to use with this build is a 6.4GB.

Reply 3 of 5, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

That's the most likely reason. If you have a spare chip, you can swap it one by one, and find out which one is faulty.

YouTube, Facebook, Website

Reply 4 of 5, by lowlytech

User metadata
Rank Member
Rank
Member

Thanks for the confirmation. I unfortunately don't have any extra cache chips, but if I did, I am sure I could make a mess of things pretty quick if I attempted to solder these things as I only have a pencil iron.

Reply 5 of 5, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
lowlytech wrote:

Thanks for the confirmation. I unfortunately don't have any extra cache chips, but if I did, I am sure I could make a mess of things pretty quick if I attempted to solder these things as I only have a pencil iron.

You need SMD paste and a hot air SMD rework station to correctly solder leads that small. It's very easy to do if you have the right equipment... I managed to re-solder a TMU on a voodoo card first try, and those are pretty hard to do since the pins are extremely flimsy and there's a buttload of them. The trick is to carefully remove old chips using heat and a lot of patience (I use a hot air gun for that since it's quicker then the SMD rework station and heats up more of the board), then clean up the contacts with some solder wick. You can then wipe with a cloth with flux on it, put a tiny bit of SMD paste on, very carefully allign the chip to the pads. Hand solder one or two edge pins in place with a thin soldering iron (so the chip won't blow off the board when you apply hot air), then finally use the air blower from the SMD station to finish the job.