PD2JK wrote on 2023-12-11, 13:08:

Kahenraz wrote on 2023-12-11, 12:38:

PD2JK wrote on 2023-12-11, 09:30:

A few days back I thought I blew an AT PSU, so I tried another PSU. Thankfully the short circuit protection kicked in. So after some inspection of the board, I saw this.

Why are there so many clock chips on that motherboard? I count no less than four from just that one angle.

I don't know. Maybe it's a (WD) Paradise thing, one some cards here I count just as much.

They even explain it at the bottom of the page. Very cool.

Paradise/Western Digital Crystals Most Paradise and Western Digital SVGA graphics cards come with 3 or 4 crystal oscillators, wh […]
Show full quote

Paradise/Western Digital Crystals
Most Paradise and Western Digital SVGA graphics cards come with 3 or 4 crystal oscillators, which are defined as follows:

MCLK is the memory clock and determines the DRAM read/write access timing as well as the system CPU I/O and memory timing. MCLK should be 36 MHz if 120ns DRAMs are installed, and 42 MHz if 100ns or faster DRAMs are installed. This crystal should be of a higher value than VCLK0 and VCLK1.
VCLK0 is the video display clock (aka, the dot clock) for both text and graphics modes. At 25.175 MHz, this displays 640 pixels per horizontal display line.
VCLK1 can either be a second video display clock input or an output to an external clock selection module. If used as an input, 28.322 MHz is used to display 720 pixels per horizontal display line.
VCLK2 (if a fourth crystal is present) can either be a third video display clock input, or an output to an external clock selection module. If present it is usually 42.000 MHz or 44.900 MHz - don't confuse VCLK2 with MCLK. If the card also has a 36 MHz crystal elsewhere, that one is likely to be your MCLK crystal, and the 42 or 44.9 MHz one would be VCLK2. Your card might have a jumper to enable this fourth crystal, i.e. VCLK2.

For 320 or 360 pixel modes, VCLK0, VCLK1 or VCLK2 is divided by 2 by the chipset to derive the pixel clock (PCLK).

Woo, the e-waste Voodoo 3 that I thought could not be fixed is now fixed, which earns it a real heatsink. This time not one that screws down, just pushpins:

Now it can actually be used in a computer instead of sitting in the parts bin.

What was wrong with it?

Kahenraz wrote on 2023-12-11, 18:21:

What was wrong with it?

I broke a BGA solder ball at the corner of the chip when screwing down a heatsink, so one of the 128-bits of memory was broken: Re: Accurately troubleshooting video memory faults with VMTCE

PcBytes wrote on 2023-12-09, 09:25:

Finished another PS3, on which someone thought it was a good idea to jam in some lytic grenades. […]
Show full quote

Finished another PS3, on which someone thought it was a good idea to jam in some lytic grenades.

I would've normally went full speed ahead and removed the CELL TOKINs and go full tantalum on that side of the board, but my nasty cold didn't allow that. So, I just put in 8x 330uF tantalums in place of the lytics stuck there (on the RSX side), and it booted fine in Recovery Mode.

I had to restore the firmware (4.81 OFW) as the previous owner somehow semi-bricked it. From there, checked syscon log and got a bunch of 1002s (which explains the grenades), then went on with exploiting it so I could upgrade to 4.90 CFW and retrieve the Blu-Ray key (eid_root_key).

After all was done, I painted the bottom case white, added a custom theme and coldboot (custom logo and 98SE startup sound), but not before replacing the 160GB Seagate the previous owner left there with a 200GB Fujitsu.

Testing concluded. Console runs silent and pretty cool for a 90nm RSX unit. Only game that heats up the RSX quite a bit is NFSMW 2012 (continuous 67, peak 71), the rest seem to keep it rather cool. (continuous 60-62, peak 65-66)
Pretty happy how this turned out after proper mending. Left to do is drop a working BD drive (probably going for a BD-410 drive off anything from CECHHxx and newer) and it's officially done.

Played around with my Intel SE440BX-2 and the 400 MHz P-II it came with. First and foremost I wanted to test a couple of P-!!! before passing them on, but it needed a newer BIOS, and I don't have a running floppy drive in another rig with an OS, so to be able to make a boot floppy for BIOS update, I found myself installing 98se from the dish image on the secondary partition, and after the BIOS update, I just tested out some 256 MiB sticks of SD-RAM which was supposed to be BX compatible. 3 found good, the last 2 needs a little more investigation - and I really need to dig out an optical drive to make a memtest86 cd to thoroughly test it ...
So tomorrow I can test some slots 🤣

Thermalwrong wrote on 2023-11-29, 15:39:

Check under the palm rest, you can unscrew that just by lifting up the keyboard tray. There's 4 screws underneath from what I re […]
Show full quote

CrazyCatman wrote on 2023-11-29, 06:44:

Spent time dismantling an IBM ThinkPad 760 XL which no longer starts up (before dismantling), but just have a speaker symbol flashing. I read that it could be just the switch for hibernation, but that didn't seem to be the case. I also read that it could be most likely the main system board or the power board.
I tested the PSU and it gives 16.4 V which is close to the 16 V DC it says.

Check under the palm rest, you can unscrew that just by lifting up the keyboard tray. There's 4 screws underneath from what I recall and then the plastic can be rotated upwards to remove it.
There's a standby battery inside that area which is unfortunately a NiMH cell, which sadly afflicts all Thinkpad 760 series - my 760LD needed some trace repair in there after cleaning up all the battery corrosion.
The PCB in there is fairly simple though with maybe only 2 layers, it hooks up some RTC stuff and the keyboard from what I recall.

Ideally it would've been best to check that battery first before pulling the rest of the laptop apart. The board sandwich inside is less likely to have faults.
Additionally - the Thinkpads support POST code reading through the parallel port so you can get a standard parallel POST code reader to check what stage of the boot process it's getting to.

There didn't seem to be any damages when I first looked at the batteries - however, I will check again.
There were a slight corrotion spot on one of the component boards (I don't remember right now which one it was, but I believe it was the power board), however it did seem to get power in and out.

Going to check the board at the palm rest again - might as well measure both batteries while in there again, however I believe that they are both dead.

How many runs to say the memory is good? (at 100 MHz at least)

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

I usually figure 10, but if a single pass takes 12+ hours, I figure on how much time I want to give it. Minimum of 5 even on a very slow machine.

debs3759 wrote on 2023-12-12, 20:44:

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

I usually figure 10, but if a single pass takes 12+ hours, I figure on how much time I want to give it. Minimum of 5 even on a very slow machine.

So I'll let it run over night. Should give me 5-7 runs ...

Thermalwrong wrote on 2023-12-11, 17:16:

Woo, the e-waste Voodoo 3 that I thought could not be fixed is now fixed, which earns it a real heatsink. This time not one that screws down, just pushpins:
IMG_2717 (Custom).JPG

Now it can actually be used in a computer instead of sitting in the parts bin.

I think it's finally time for me to put away the hot air station - I've now broken the Elsa Victory II enough that one of the memory data lines is just broken. I mean I was fixing corrosion but with hindsight I can say that drilling through the PCB to run new traces through was a terrible idea. I didn't notice until after I'd drilled those holes that the PCB is 6-layer and some of the data traces route internally, or they did until I started on it. Maybe I'll fix that in the future but it's going away for a while now - repairing a Banshee card is tricky business since there isn't the plethora of information like there is for the Voodoo 1 and Voodoo 3 cards.

I got one card to actually post once after reflowing it so thought I'd have a go at reflowing or removing the GPU of this PCI Voodoo Banshee - I had thought its GPU was fried since there were missing pads and traces on the PCI slot which I repaired but it would never run, just gave the no video card detected beeps.
Trying to get some practise with the USB hotplate which works for these few 3dfx cards since they have completely flat areas on the back of the video card:

For the CT6760 banshee card it actually started working 100% after doing this, I truly thought the GPU was fried and was going to try swapping the AGP Banshee's chip onto it as a replacement. Even the ram just works 100% and this card didn't come from e-waste so it's in pretty good shape apart from the PCI pin damage.

Removing the thermal epoxy on the GPU was pretty scary, jammed a craft knife blade in an edge I could see light through and put the card in the freezer for an hour, pulled it out and pushed the blade in a little further, then squished a second blade in until it went pop.

Use a piece of plastic like a credit card or gift card between the PCB and the heatsink to give yourself some leverage without gouging the surface.

Yes, BGA reflowing is very tricky business. I've done board repair for years and I'm just not confident yet to tackle this.

I hope you manage to get it working again. Keep us updated.

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

As a general rule of thumb I do 3 100% runs..

appiah4 wrote on 2023-12-13, 05:56:

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

As a general rule of thumb I do 3 100% runs..

Great, that's what a complete night of testing gave 🤣

Problem is that this Intel SE440BX-2 has no settings for memory timings, so I can't even try to push the envelope 😉

Playing a little more around with the SE440BX-2 today. Upgrading to a 600 Katmai, experienced that the board won't play with Coppermine - it POSTs, but halts with a message that the voltage isn't supported. Tough luck - but the board's serial number pretty much implied it.
After a bit of 3DMark and some Ultim@te Racing on my Matrox M3D, I decided to test a couple of extra 256 Meg RAM sticks besides the 3, that had been running memtest86 all night. But no luck. And all of the sudden the good sticks also failed within 20 or so minutes? WT actual F? Ok, thinking a bit, I'd run them with the 400MHz Deschutes before, could the cpu actually be the culprit? The temperature of the heatsink made me think that. So now, after placing an extra 120mm fan blowing across the CPU's HSF, it seems better again. But, otoh, 600 was the top speed released for Katmai, so it might be close calls. Especially since there's no thermal grease - only that rubbery stuff, which is probably pretty dry now. But swaps the sink around, and tbh, I hate that mess 🤣

Futzing around with an absolutely-cursed AST Advantage 824 still (yes, the one I was curious about having a VRM modded into). Turns out the thing is FAR more cursed than I originally imagined. Stock CD drive was going bad (gave corrupted directory reads, since the ONLY DOS driver it worked with was theTorisan driver on the recovery disk set). ATI graphics are somehow LESS compatible t than a dedicated Mach64 PCI card (PacPC2 and Ms. PacPC both hard-crash the machine if ran in full-screen mode, on top of the usual DOS incompatibilities). Using more than two ISA cards makes the system unbootable (half the time the system halts on POST with a "DMA Controller Error"), and my sound cards (both the built-in CS4232 with QDSP wavetable and my original GUS PnP Pro) both fail to be detected by any game/demo if using the Intel Configuration Utility for PnP cards (hell, on a CF card with a WfW install, the DOS setup program for the GUS just hard-locks the computer every time, making installing the Windows drivers impossible).

Kind of a shame, really, as I was hoping I could get this machine up and running with its 28.8 modem installed (once I can get a long-enough phone cable to route from behind the couch to the cable modem, I wanna try dialing into some BBSes), 3Com ISA ethernet, GUS, Voodoo1, and POD-MMX. Especially with a tri-boot setup of DOS 6.22/WfW3.11, Win95, and OS/2 Warp 4. Just strange that in Win95, both serial and parallel ports still show up as active despite being disabled in BIOS.

Moar PS3 restoration going on, and this isn't even the beginning 🤣

Up today, a japanese CECHB00 with fully black trim has gotten a new life - a fresh COK-002 board from a recent european C04 purchase- someone clearly maintained the C04 very well. Those temps on the RSX are absolutely delightful. 19 blade fan was very clean, and I suppose the thermal paste used by the seller I got the C04 from is likely Thermal Grizzly Kryonaut, at the very least.

Already loaded it with a few PSX and PS2 titles, as well as a few PS3 faves (and a free GT5 copy I had received with the C04!) and it's been running absolutely sweet.

Got a 9700 (Pro?) from a colleague, replaced the cooling paste and is now sitting in the Athlon 1200 testing rig.

R3x0 cards tend to die on me, hopefully this one doesn't. Fingers crossed, so far so good.

Update: 7155 marks (3DMark2001), all default XP SP2 drivers, no artifacts!
George Peppard time.

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

One pass for all tests, then test #8 (random numbers) in loop for 24 hrs at least - this will put pressure on CPU, mem controller and RAM.
I have 775i65g setup e5800 overclocked high (3.78GHz, ddr1 400 @470 3-3-3-8 ) which gives me an error only in #8 after 10-12 hours.
If something unstable #8 start throwing errors almost immediately.

vsharun wrote on 2023-12-15, 13:27:

One pass for all tests, then test #8 (random numbers) in loop for 24 hrs at least - this will put pressure on CPU, mem controll […]
Show full quote

H3nrik V! wrote on 2023-12-12, 20:08:

How many runs to say the memory is good? (at 100 MHz at least)

One pass for all tests, then test #8 (random numbers) in loop for 24 hrs at least - this will put pressure on CPU, mem controller and RAM.
I have 775i65g setup e5800 overclocked high (3.78GHz, ddr1 400 @470 3-3-3-8 ) which gives me an error only in #8 after 10-12 hours.
If something unstable #8 start throwing errors almost immediately.

Funny, I had a feeling that the one with random numbers could be a joker, since you'll never know when it's been through all possible combinations ...