VOGONS


First post, by watson

User metadata
Rank Member
Rank
Member

Hi, this is my first post on VOGONS and already quite a downer, but here goes.
Last week on a local Craigslist equivalent I stumbled upon a GeForce 256 DDR from the title for around €7. I bought it together with a GeForce 6600 AGP from the same seller for a combined €15 with shipping.
Both GPUs were listed as working. However, the idiot packed them in one sheet of paper stacked right on top of each other and put it all in a regular envelope.
When the packaged arrived and I saw what it was, I was fuming. Of course, inside the envelope were the two cards and about two and a half SMD capacitors rattling around.
Upon further inspection, it became clear that someone had already tampered with the Erazor. Fan wires were spliced together and the N-channel MOSFET that controls the fan had two of its pads ripped off and it was soldered back completely wrong (there are also pads for a through-hole version of the MOSFET, and both gate and drain were soldered to the SAME pad).
Therefore, the fan wasn't spinning at all. Thankfully, it was possible to resolder the MOSFET correctly and that (combined with a bit of oil) was enough to bring the fan back to life.
However, that was not the cause of the main issue. The symptoms are those of heavy memory corruption, photos of which can be seen here: https://imgur.com/a/GJ0Sx
There was a missing SMD electrolytic on the front which I replaced, but it didn't fix the issue.
There was also a missing SMD ceramic on the back near one of the RAM ICs (visible in the photos). Of course, I didn't know the value so I desoldered a cap near another RAM chip which (I guessed) had the same functionality.
It turned out to be a 1nF cap, so I replaced the broken one with the same value. Nothing changed. These are both probably decoupling caps (because they are connected to ground on one side) so I presume their presence doesn't make much of a difference anyway.
The last thing I tried to do was reheating the solder joints on the all the RAM ICs on the back. This was hard to do with a €10 LIDL iron and took quite a while, but it is possible (with the help of some solder wick here and there). Once again, nothing changed.
I did manage to boot into Windows XP once and installed the driver, but device manager just showed a Code 10 (This device cannot start), which is to be expected. After that, it just BSODs at the XP loading bar.
At this point you might be thinking that I don't have a clue what I'm doing and that I'm going to eventually kill the card and I mostly agree with you.
It is possible that the RAM is simply bad. It is also possible that the GPU is dead from overheating beacuse it presumably ran without a fan for a long time (and it gets hot very quickly without it).
I would like to know if there is anything I could possibly do to bring this card back to life because it is a piece of history and I don't think I'll ever have to opportunity to buy another.
I'm thinking of just throwing it in the trash and forgetting it ever happened. This has pissed me off so much that I think I won't be buying "retro" hardware (i.e. junk) ever again.

BTW, the 6600 was missing 2 caps near the core and 1 near the RAM. Since there is no possible way to know the values (unless I had another exact same card, good luck with that) I randomly replaced them with 100nF ceramics. There is just a black screen on boot, nothing happens.

Reply 2 of 7, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Overall condition seemed to be fair, but after that RAM pin resolder, card looks like !@#$. Tip for the future - use sharp needle (really sharp) to test every pin by applying slight pressure on each pin, you'll instantly see which pins are detached this way.

Judging from screen shots, it's definitely RAM related issue. I recently encountered same corrupted screen pattern after replacing all RAM ICs on GF2 and the problem was... A missing connection on a SINGLE ADDRESS PIN of RAM IC. Out of all 216 combined RAM IC pins only 1 was not soldered properly and that caused such corruption (it was Address 6 pin on a 8Mx16 SDRAM).

It could be anything in your case: a cut trace, debris shorting something under GPU/RAM, missing/cracked smd resistors, bad value bypass caps (most of the times it's not an issue, but when it comes to 1nf values, it means that switching speeds must be insane). Once i had a bypass cap broken off a PC-133 SDRAM, could never get it working stably at that frequency till i got caps value right.

Also, those long through-hole ceramic cap legs pretty much nulifies its existance due to parasitic inductance, replace them with SMDs.

Reply 3 of 7, by watson

User metadata
Rank Member
Rank
Member

Thank you very much for your input, SSTV2.

Since I have no access to a hot air rework station and only a very crappy iron, I decided to put it in the oven before throwing it away (preheated to 200°C, 10 minutes).
It ended just as one would imagine. I lost yet another SMD cap and something caused a short on the output of the voltage regulator for the RAM (upper right corner on the front of the card).
Due to the fact that I can't find the point of failure (in fact, the card looks better than before my hackjob), I've finally decided to give up on this thing. I've wasted enough of my time as it is.
I know you can't win them all, but if only I could win at least one...

Reply 4 of 7, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Totally been there watson. Some times you just have to let em go. I've spent way too much time on some projects that didn't work out... and I've spent a bunch of time on some projects lately that I'm hoping might still eventually work out. 🤣

You could always keep it anyway... having a non working Erazor X2 is still kinda cool. 😉

Now for some blitting from the back buffer.

Reply 5 of 7, by watson

User metadata
Rank Member
Rank
Member
Ozzuneoj wrote:

You could always keep it anyway... having a non working Erazor X2 is still kinda cool. 😉

I know, but it will haunt me forever. You probably know the feeling: "Maybe I could still try this..." And then it doesn't work.
I guess I'll just put it away somewhere. Bury it deep.
What a shame.

Reply 6 of 7, by SSTV2

User metadata
Rank Oldbie
Rank
Oldbie

Haste makes waste, by the way, did you remove GPU heatsink before reflowing whole thing? It looks like that card had cracked solder blobs under GPU that could have fused together (hence shorting RAM VRM). As long as PCB is good, it's restorable, I'm in worse situation with my Voodoo2 right now, but I did not lost hope in bringing it back yet, because card works with texturing disabled, here it is - http://i.imgur.com/dUkk3aSh.jpg?1, http://i.imgur.com/s7e7TkKh.jpg?1. Two EDO RAM ICS are ripped from main TMU with pads alltogether. It's restorable, but I can't diagnose exact fault, could be anything in this case, because I don't know its story. Full thread: Repairing defective Voodoo2.

Reply 7 of 7, by watson

User metadata
Rank Member
Rank
Member
SSTV2 wrote:

Haste makes waste

Very true.

SSTV2 wrote:

by the way, did you remove GPU heatsink before reflowing whole thing?

Unfortunately, I didn't, so that's probably what killed it. I baked a few cards in the past and they all had removable heatsinks, so I would always remove them (I only ever managed to get an X700 Pro PCI-E back from the dead).
Erazor's heatsink (as on a lot of older cards) was glued down with thermal glue and I had no thermal glue to put it back on afterwards so I thought I could get away with leaving the heatsink on the GPU. Big mistake.

SSTV2 wrote:

As long as PCB is good, it's restorable, I'm in worse situation with my Voodoo2 right now, but I did not lost hope in bringing it back yet, because card works with texturing disabled

I read your Voodoo2 thread and I'll just say three things:
1. Wow, your level of dedication is amazing. I would have given up a long time ago. In fact, I gave up after three days with my card.
2. You have way more knowledge of electronics than me. I try to do what I can with my limited skills and tools and usually it doesn't end very well (as can be seen here).
3. Any Voodoo will always be more valuable than any GeForce because it's a Glide card (and therefore worth saving). The GeForce 256 is just a DX7 card and there will always be millions of DX7 cards around. In that sense, I guess I can be happy that I haven't killed a Voodoo, just a GeForce. Heck, I haven't even seen a Voodoo in my life (and it's probably for the better).