tehsiggi wrote on 2025-07-01, 07:47:

Perhaps I didn't read properly, does the issue also occur if you just idle around in BIOS etc?
Some cards with memory issues that I had in the past worked mostly "okay" until the driver has been loaded.

It depends. When card is "cold" there is no artifacting at all 'til windows booting - probably after loading driver the glitching begins. Sometimes when the card is already "worm" and I make a reboot the artifacts occur even in text mode - on POST screen, in BIOS settings menus, in DOS prompt. That's why I'm thinking now that maybe there's a memory problem in addition to instability issues.

tehsiggi wrote on 2025-07-01, 07:47:

Regarding mapping your memory to the result of the tool: […]
Show full quote

Regarding mapping your memory to the result of the tool:

You'll need a schematic of a 9550 or 9600, often the memory alignment (data bits to memory chips) is the same even if the rest of the layout is different.
There you'll see which data bit goes to which IC. They often also describe where the memory is located on the board.
The RV350/RV360 have two memory channels, A and B, each 64bits => 4 x 16 Bit ICs if TSOP-66 memory is used like on your card.

Fortunately board partners often just copied big chunks of the reference design, like the memory part. So you'll find the exact same memory arrangement and identifiers even if other parts are different. I attached the schematic of a Radeon 9600 from gigabyte for you. It's not exactly the same in terms of the power section, but memory looks good.

Ok, I'll figure this out later as obviously I'll have to dig into the card's datasheet. For now I just figured out the most important part of the log-messages from R3MEMID in your card repairs examples:

1R3MEMID version 1.07, (c) Copyright ATI Technologies Inc, 2003
2Log file generation enabled to .\R3MEMID.LOG ...
3Reference data file (RDF) loading disabled ...
4[1  ] Fill                                                : FAIL
5    Error ID 0VB001
6  1024 x  768 - 32 bpp ( 60 Hz): TEST FAILURE
7    failing bit : MDD0 32 33 34 35 36 37 38 39 ...

MDD0 32 33 34 35 36 37 38 39 ... - "D channel", failing bits 32 - 63

1R3MEMID version 1.07, (c) Copyright ATI Technologies Inc, 2003
2Log file generation enabled to .\R3MEMID.LOG ...
3Reference data file (RDF) loading disabled ...
4[1  ] Fill                                                : FAIL
5    Error ID 0VB001
6  1024 x  768 - 32 bpp ( 60 Hz): TEST FAILURE
7    failing bit : MDC0 0 1 2 3 4 5 6 7 ...

MDC0 0 1 2 3 4 5 6 7 ... - "C channel", failing bits 0 - 31

Yep, you're reading the log output correctly.

Hi,
I don't want to straddle this discussion but I'd like to ask if maybe one of the experts here can have a look at my thread here about a X700 AGP:
Are X700 Radeons prone to die?

I have mostly given up on the card because I don't know what else could be tried to bring it back to life but maybe I'm missing something that someone else sees...
If there's anything to measure I have a multimeter as well as a basic scope. What I don't have are values to compare to.
And if you think there's nothing to do about the card's current state this would just give me more reason to let it go.
Thank you!

Just removed one of the two electrolytic capacitors at the output of switching voltage regulator - PCE10 (Hitano 1000 uF/6.3 V). The chineese component tester mostly doesn't recognize the damn thing as capacitor at all 😁 The multimeter also shows no capacitance ("- - - -") and internal (almost) dead short ~ 0 Ohms. Also I measured the second cap in parallel (PCE11) left on the PCB and the short is still there, so I expect exactly same results for no capacitor at all.

My second theory confirmed - these Hitano caps are among most cr*ppiest electrolytic capacitors ever: no bulging, no leaking, dead short and zero capacitance left! I prefer to deal with leaking and/or bulging junky capacitors - at least they show their problems at first glance.

I hope not to get to use the DOS memory diagnostic software for this card after a full recap.

Well, while leaking is obvious, it can also kill traces. So I'd rather have a non leaking cap.

However good to see that theses caps appear to be at the end of their life. You should do well for a couple of decades if you use at least some decent electrolytics.

analog_programmer wrote on 2025-07-01, 09:40:

My second theory confirmed - these Hitano caps are among most cr*ppiest electrolytic capacitors ever: no bulging, no leaking, dead short and zero capacitance left! I prefer to deal with leaking and/or bulging junky capacitors - at least they show their problems at first glance.

Many people prefer mistresses younger than this video card 😀

tehsiggi wrote on 2025-07-01, 10:10:

Well, while leaking is obvious, it can also kill traces. So I'd rather have a non leaking cap.

However good to see that theses caps appear to be at the end of their life. You should do well for a couple of decades if you use at least some decent electrolytics.

Of course they can, if left for years untouched after pouring out their guts. But after leaking starts they're already so degraded, so the problems begin to appear and there's the reason to immediately look at them and see the obvious 😉

At my nearest electronic components shop I just found a polymer replacements for the three (including the two fully dead) 1000 uF / 6.3 V caps and for the rest (six 100 uF / 16 V and two 1000 uF / 10 V) - Samxon low ESR. I think these will be good to go at least for decade.

shevalier wrote on 2025-07-01, 10:20:

Many people prefer mistresses younger than this video card 😀

Probably, If I was 30 again. I can't understand the kids from zoomer and alpha generations at all. So milllennial ladies are still good and "young enough" for me 😀

analog_programmer wrote on 2025-07-01, 11:01:

At my nearest electronic components shop I just found a polymer replacements for the three (including the two fully dead) 1000 uF / 6.3 V caps and for the rest (six 100 uF / 16 V and two 1000 uF / 10 V) - Samxon low ESR. I think these will be good to go at least for decade.

Don't get carried away with the voltage reserve, it's useless.
For electrolytic capacitors - the higher the voltage (and the larger the case), the higher the permissible current.
For polymer ones - the currents in any case start from 3A.
Even super_cheap
http://www.kosmodrom.com.ua/pdf/capxon-ps.pdf
Therefore, for 5/3,3V - 6.3V is enough.
For the core and memory - 4V

shevalier wrote on 2025-07-01, 11:20:

Don't get carried away with the voltage reserve, it's useless. For electrolytic capacitors - the higher the voltage (and the lar […]
Show full quote

Don't get carried away with the voltage reserve, it's useless.
For electrolytic capacitors - the higher the voltage (and the larger the case), the higher the permissible current.
For polymer ones - the currents in any case start from 3A.
Therefore, for 5/3,3V - 6.3V is enough.
For the core and memory - 4V

The main limit here is cap's terminals raster (spacing). And I'm not 100% sure if I can use 1000 uF / 6.3 V polymers for the two 1000 uF / 10 V (3.5 mm raster).

For the six 100 uF / 16 V there are no available polymers with 2.5 mm raster in this shop, so I'll use some OK-quality Samxon low ESR 100 uF / 25 V electrolytic caps (there are no 16 V or lower values low ESR caps with 2.5 mm raster and suitable case dimensions).

in that case - an excellent choice

As I started to remove all the remaining Hitano junk from the videocard's PCB I couldn't resist to measure the other two 1000 uF / 6.3 V caps. It's funny how these two have also failed in all the possible ways. PCE11 (the one in parallel with the first PCE10 I've removed) shows some "ghost" capacitance left in nF - about 80 nF . And the third one (PCE2 - near the memory linear VRM) shows 1400+ uF (too much for 1000 uF cap) which is also symptom of certain failing.

I've seen many chineese electrolytic junk caps "brands", but these Hitano c(r)aps are my new "favorites" among all the junk caps 😁

And then I'll remember a cap from an old oscilloscope i had. 33uF at 100V. It failed after 31(!) years of constant use.

tehsiggi wrote on 2025-07-01, 16:49:

And then I'll remember a cap from an old oscilloscope i had. 33uF at 100V. It failed after 31(!) years of constant use.

This seems like normal life time period for a quality electrolytic capacitor, if not overheated.

I tried the card with half removed Hitano c(r)aps and it showed absolute artifacted garbage since boot. So there is a hope, that full recap will solve the problems. The reason to try this was to measure the voltages on the two 1000 uF / 10 V and these will be also replaced with 1000 uF / 6.3 V polymers as it turned out that 10 V is really an overkill and 6.3 V (even 4 V) will be more than sufficient.

As I saw, that you're very much into the old videocard repairs, I have a question to you: This card lacks any 5 V fan terminals, do you have a suggestion where I can safely solder fan's wires if I install an active cooler (probably some of those "universal" chineese coolers)?

Above the cooler there ia JU2 with two unpopulated footprints for, i guess, 0402 resistors. These are meant to be bridges with 0 ohms. You can select voltage by populating one of them. Usually one goes to the 12V rail and the other one goes to the 5V rail.

JU2 would be the spot for your FAN header.
Most replacement fans are 12V.

A quick continuity check with your multi meter should reveal which one goes to the 12V and which one goes to 5V.

tehsiggi wrote on 2025-07-02, 04:14:

Above the cooler there ia JU2 with two unpopulated footprints for, i guess, 0402 resistors. These are meant to be bridges with 0 […]
Show full quote

Above the cooler there ia JU2 with two unpopulated footprints for, i guess, 0402 resistors. These are meant to be bridges with 0 ohms. You can select voltage by populating one of them. Usually one goes to the 12V rail and the other one goes to the 5V rail.

JU2 would be the spot for your FAN header.
Most replacement fans are 12V.

A quick continuity check with your multi meter should reveal which one goes to the 12V and which one goes to 5V.

Nice catch, thank you! I didn't noticed these pads since not lo0ked fan connector so close to the original heatsink. Or maybe I need glasses 😁

I may try those chineese "universal VGS coolers" if one fits - they come with 12 V fans.

The mounting hole distance on your card appears to be standard, meaning 55mm. Most of the universal 3rd party coolers will fit just nicely.

I'm often using these silver ones, but they're more out there on the market. I'd not really recommend them for anything above 9600, though they'll get the job done as a temporary solution for a 9700 as well.

For what it's worth. The same design has been the stock cooler on Sapphire Radeon X700 cards, so it's not too bad. Plenty for a 9550 I'd say.

Yeah, I just took some measurements and one of these 55 mm chineese coolers will fit perfectly, and the orange one will be a perfect match for the orange PCB 😀

I already used one of these (silver aluminium) on passively cooled FX5200 and it does better job than the original bare heatsink.

I looked at my Gigabyte R9550 GV-R955128D which is also with pathetic passive heatsink cooler, but here I really can not see any pads or holes for fan connector 🙁 At least it has pretty good Nichicon electrolytic caps and after more than 20 years - no picture issues.

The picture is from i-net, but my Gigabyte R9550 is exactly the same:

I just checked the schematic for your Gigabyte card. They did not plan for any active cooling for that one. There's no FAN header in the schematic, nor on the board. It's been designed for passive cooling. So either you steal voltage somewhere on the board or have to power it externally.

tehsiggi wrote on 2025-07-02, 05:25:

I just checked the schematic for your Gigabyte card. They did not plan for any active cooling for that one. There's no FAN header in the schematic, nor on the board. It's been designed for passive cooling. So either you steal voltage somewhere on the board or have to power it externally.

Yep, Gigabyte's R9550 is completely missing the fan header 🙁 Gigabyte GV-R955128D is better looking card, though it comes with smaller amount (128 MB) of slower VRAM than the Asus card, so I think I'll just keep it stock.

And since we opened the topic about the active cooling for these R9550/9600 videocards, do you think it's worth replacing the stock cooler on my GeCube R9550 (its stock frequencies are higher than the standard R9600NP) with these 55 mm chineese coolers? The stock cooler looks like this:

P.S. From where did you got the schematics/circuit diagrams for these videocards?

I don't think you'll gain any significant cooling performance improvement over the stock fan. Having both here, they appear to be pretty on par.

There are pages like vinafix on the web that have a ton of schematics available. Electrotanya is also always worth a shot.