[UPDATE] [SOLVED] Need help identifying this chip (on a Radeon HD Graphics Card)

[flubba86]

Hi all,

I was swapping around my graphics cards a few weeks ago and somehow accidentally knocked an IC off the board. I tried to solder it back on, but two of the pins had pulled out of the chip package, it was destroyed.

The Chip:

(Picture is a cropped portion of a fullsize scan of the same area on a similar card)

The chip in question id designated U100 on my card too. The two pads immediately above the marked area are for an unpopulated resistor R104, not labelled in this picture. I have no idea what the chip is. It is seriously tiny, about 2/3 the size of a SOT-23. I can't work out whether it is 3 pin or 5 pin. There are certainly 3 pins on the lower side, the two pins above it are for the unpopulated resistor, but there are two pads on the board which would sit under the chip, maybe for grounding or maybe for heatsinking?

The card is a Gigabyte Radeon HD6870. All backside images of 6870 boards from all manufacturers that I have found on google image search show this same chip in the same location.
This is the full backside scan of an AMD Reference 6870... for reference http://www.ixbt.com/video3/images/barts-2/hd6870-scan-back.jpg

I would appreciate if anyone who has experience with these boards could help me try to identify the chip. I have electronics technician experience, and hot-air rework equipment available. I can certainly repair the board if I can only source a replacement chip.

It would be great if someone who has a 6870 could take a super-high-res image with a macro-lens or through a microscope.

Cheers,

[flubba86]

Ive done some more research, looking at high resolution scans of cards on google image search. It looks like this tiny chip is on every ATI/AMD card from the X300SE era (can get them for less than $10 now) right up to modern top end cards. It is a 5 pin package, possibly SOT353, and you can almost get a code from the chip in this image:
http://www.vgamuseum.info/images/stories/zaatharen/ati/radeonX300SE_asus_fhq.jpg
It is designated U102 on that card.
Edit *nevermind, the chip on the X300 is different*

[flubba86]

Ive found it!
I probed around on the board and garnered an approximate pinout. It has a VCC, a GND (the cap above the chip is across these two) and it has a direct connection to the PERST# pin on the PCI-E connector.
Im 90% sure its an NL17SZ04 - A Low-power high speed inverter. Comes in SOT353 and the pinout matches perfectly.
http://pdf.datasheetarchive.com/datasheetsmain/Datasheets-23/DSA-452473.pdf

Im guessing the ATI Radeon Chips use an inverted logic for their RESET circuit, and this little IC inverts the PCI-E PERST# reference RESET line.

Now I have three ways forward on this issue. I can
#1 - Try to source a NL17SZ04 in single or low QTY from somewhere.
#2 - Work out which RESET logic the Radeon chip needs and tie it ON with a 1K resistor
#3 - Buy a cheap 2nd hand old Radeon card from ebay and salvage the inverter from that.

[cjo20]

#1 - Try to source a NL17SZ04 in single or low QTY from somewhere.

http://www.digikey.co.uk/product-search/en?x=0&y=0&lang=en&site=uk&KeyWords=NL17SZ04

[flubba86]

#1 - Try to source a NL17SZ04 in single or low QTY from somewhere.

http://www.digikey.co.uk/product-search/en?x=0&y=0&lang=en&site=uk&KeyWords=NL17SZ04

Thanks for the link. I had already found the part on Digikey, its just the $32 postage that is the issue.

It looks like I can order samples from ON Semi for free, just $20 postage. I might end up going with that. I will find some other stuff to sample too.

[amyk]

It may not be an inverter, it could just as well be a buffer. And then you have the question of whether it has a Schmitt input or not.

In the X300SE picture pin 1 looks like it's connected to something (the trace goes into the via below it), whereas that is NC on the NL17SZ04. What would be the input appears to be connected to +3.3V. Does the package have any marking?

[flubba86]

Thanks for the input, amyk.
Now that you point it out, I think the chip on the X300 is different. On that image it does indeed look like the pin 2 is connected to VCC.

On my board pin 2 is connected directly to the PERST# pin on the PCI-E connector, pin 3 is ground, pin 4 is output and pin 5 is VCC. Pin 1 does not appear to be connected to anything, but it may be connected on an internal layer.

You are right, it could certainly be a non-inverting buffer, such as this one:
http://www.onsemi.com/pub_link/Collateral/MC74VHC1G125-D.PDF

The chip which came off the board was pretty damaged and was lost shortly after, so I didnt manage to get any markings from it.

There is another clue though, there is an unpopulated resistor location (R104) which is underneath the footprint for the chip (U100). The resistor pads essentially connect to U100 pins 2 and 4.
I had originally thought this meant that they populate the resistor if they want non-inverted RESET connection, or the inverter if they want inverted RESET connection, though it could be also true that they populate a buffer if they want buffered RESET, or the resistor if they want to save a few cents and have unbuffered connection.

[flubba86]

Sorry for digging up an old thread.

I received a PM today from someone asking if I ever sourced this chip or found the solution to my problem here.

I just realized I never posted a follow up with the resolution. And yes it does have a resolution.

Firstly, to fix my Radeon HD6870, I salvaged the same IC from my older working Radeon HD4850. This was still a good card in working condition that I wanted to sell, thats why I initially didn't consider taking the IC from it, and never mentioned it in my previous posts (it was my only working video card, and was in my main desktop while I was researching these chips, and posting on forums for a solution). But curiosity got the better of me, and I wanted to know if my HD6870 was even repairable at all, so I made the swap. Damn thing is so tiny! Even with a hot-air rework station, tiny soldering iron tip, tweezers, and a magnifier it was very difficult. But I got the chip swapped over, and it did fix my HD6870, but left my HD4850 missing the IC.

After removing the tiny 5-pin package from the HD4850, I saw that it too had an unpopulated tiny 0201 resistor footprint (R104) under it and within the SOT353 package footprint. I probed out the R104 pads, and saw that it is indeed connected across pins 2 and 4 (input and output) of the SOT353 footprint, so it does essentially bypass that chip. I put a tiny blob of solder across the R104 pads, as I postulated in my previous post. And it works!

So that chip must be a non-inverting buffer as @amyk suggested, most likely the MC74VHC1G125-DF or a jellybean part very similar to that. And bridging the R104 footprint pads just eliminates that chip from the circuit so it isn't needed.

My old HD6870 and HD4850 are both still working today, though I do have a newer card in my main desktop system.