Kfa2 GeForce GTX 750 Ti 2GB GDDR5

Seller Note “When installed in pc the fan spins but with no output to the screen unfortunately otherwise in great condition”

Pre Checks / Multimeter Only

Test / ChecksResultNotes
1. Visual Inspection Eye / Magnifying GlassOKNo missing components, but a bit dirty.
2. Visual Inspection Microscope
3. Check for shorts on 12v PCI, 12v PCI-E, 3.3vOKCouldn’t find any obvious shorts.
4a. Resistance VCore (~0.5)Ω
4b. Resistance VMem (~60Ω) Ω
4c. Resistance 3.3v Rail (~960Ω+) Ω
4d. Resistance 5v Rail (~5KΩ)K Ω
4e. Resistance 1v PEX Rail (~123Ω+)Ω
4f. Resistance 12v BUS (14.7kΩ )
4g. Resistance 8.8v Rail (18.7KΩ )

Power On / Test Rig Checks

Tests / ChecksResultNotes
1. Observations (e.g. PCB heat noticed,
blank screen, power light on monitor,
bios screen, o/s loaded, artefacts, )
No display (no light on monitor)
Fan spins, GPU never gets warm (no power). Some warmth in the VRM area.
No power to CPU.
2. Voltage VCore (~0.88v)0.914vOK
3. Voltage VMem (~1.4v)~0v (4.3mV)!!! Need to investigate further
4. Voltage PEX Rail (~1v)?!!!
5. Voltage 3.3 BUS3.3vpresent
6. Voltage 12 Ext12vpresent
6. Voltage 5v Rail5vpresent
7. Voltage 8.8v Rail

That memory area needs further investigation, access is a bit hard with the cooler in place. The board is quite small and the heatsink is quite large and gets in the way of diagnostics. I will see if it can be run with a make-shift heat sink. Alternatively, I will try and rig up a lab power supply to a riser.

Next Steps

Actually, there is no need for a heatsink, as the GPU remains cold. There is some warmth towards the IO shield / VRM area, but hard to pinpoint. Still cannot find any trace of the 1v PEX rail. I am guessing VMem is missing perhaps because PEX is also missing. Hopefully, there is a chance that the core and memory are ok and there is some failed power switching/delivery component. Interesting!

Deep Dive

Used my cheaper microscope to take a closer look at the PCB, in particular, components relating to possible minor voltage production (VMem 1.5v, PEX 1v). Found two Low Dropout Regulars to investigate (refer to Fig 1 & 2 in photo above for positions) :

Fig1 U503
Fig2 U501

I struggled to find a datasheet for these exact components, but I did find this seemingly very helpful image from a Russian site:

Taken from https://cxem.net/sprav/sprav139.php (worth a look for the other information there).

Also here is a link for a similar chip http://images.100y.com.tw/pdf_file/38-RISHTEK-RT9018.pdf

Time to check voltages on these chips:

PINExpected VActual V
1. PGOOD (Dot)From vCore?0v Missing
2. EN2v+ ? 3.07v OK?
3. VIN3.3v0v Missing
4. VDD5v5v OK
6. VOUT1v0v Missing
Fig 1 U503
PINExpected VActual V
1. PGOOD (Dot)From vCore?0v
2. EN2v+ ? 5v OK?
3. VIN3.3v0v Missing
4. VDD5v5v OK
6. VOUT1v0v Missing
Fig 2 U501

Seems we could have a problem with the 3.3v rail.

TODO Test resistor R649

Want to understand more about the Vmem rail.

RT8120A http://www.chipset-ic.com/datasheet/RT8120A.pdf (single-phase synchronous buck PWM DC/DC controller designed to drive two N-MOSFET) I am guessing this drives the two MOSFETs in the memory VRM.

Also, there is a chip marked 0T=1C 622 which I think is a PWM used to drive the Vcore VRM http://www.smdcode.net/uploads/Files/Datasheet/0T/RT8811A.pdf

Update: Unfortunately, I accidentally shorted to of the Vcore mosfet pins whilst measuring with my new meeter, smoke seen and now no VCore – dam it! The driver appears to have a burn mark on it, would imagine at least the driver and the mosfet need replacing. This is now long fix territory, as I need to find replacements and hope for the best in terms of no further collateral damage.

Update 29/07/2022 – Have a working identical card to validate most of the above

I now have a working identical GTX 750 ti and many months more experience, so thought it could be a good time to revisit and correct some of my initial findings.

Working card resistances

  • VCore – 9.5Ω (This card 10.5Ω)
  • VMem – 43.9Ω (This card 41.6Ω)
  • PEX – 319Ω (This card 329Ω)
  • 5V – 508Ω (This card 509Ω)
  • 3.3V – 7.5KΩ (seems relatively high, but similar on this card 7.6KΩ)
  • 12V – 637Ω (This card 440Ω)

So all the resistances look fine.

Working card voltages

  • VCore – 0.97V (This 0V card)
  • Vmem – 1.52V (This 0V card)
  • PEX – 1.05V (This 0V card)
  • 5V – 5.15V (This 5.11V card)
  • 3.3V – ok (This card – ok)
  • 12V – ok (This card -ok)

Well, I know I fried the VCore PWM. Originally there was VCore, but I am not sure if PEX was ever there (how I missed the marked measure points shows my complete lack of experience then). PEX and VMem may be dependent on VCore, probably the way forward is to consider:

  • Buying another RT8811A VCore PWM and replace it
  • Buying another RT8120A, not proved, could be why VMem was never present
  • To be on the safe side, buy some uP0104S (can see now that u501 is PEX, might just need enable)
  • And possibly some MDU1514, MDU1516 MOSFETs
  • Probably not worth the investment in all those parts I might not need, but I would really like to fix this card as it is one of my first projects.
  • Will also take a look in my donor card collection. Ah ha!, I already bought some RT8811A for a GTX780! Will give that a try 🙂

OK, to the VCore PWM replacement appears to be successful:

KFA2 GTX 750 TI – RT8811A 

It appears to be working as we have 0.9V on VCore, as we did before I shorted the controller. Very little heat on the core though. Now we need to work out what happened to VMem and PEX.

U501 (uP0104S) – PEX

PINWorking CardThis Card
1. pgood?(Dot)0.95V?0V
2. ?5V5V
3. ?1.5V0.01
4. VDD?5V5V
5. GND?0V0
6. FB?0.8V0
7. VOUT1.04V0
8. GND0V0

U503 (uP0104S) – I don’t know what this rail is?

PINWorking CardThis Card
1. PGOOD (Dot)0.1V?0v
2. ?3V3.09
3. ?1.5V Vmem?0v
4. VDD5V5V
5.0.01V0v
6. 1.06V peX?0v
7.VOUT0.8V0.01V
8. GND0V0v

U504 (RT8120A) – VMem Rail

PINWorking CardThis Card
1. BOOT13V?1.36
2. UGATEOscilloscope: nothing on this
3. GND
4. LGATEOscilloscope: Has waveform
5. VCC12V12V
6. FB0.6V0.01
7. COMP/EN0.97V1.13
8. PHASE

OK, this was quite painful to measure without any cooler, the core gets hots very quickly. Since there is VCC and EN are present, I would suspect the U504 Vmem PWM controller is faulty. A further indication of this is that the UGATE pin shows no signal and the LGATE has a signal, the net result would be 0v.

So, I think I need to either buy or find a RT8120A. Then, hopefully its a case of:

5V -> VCore (resolved) -> VMem (needs fix) -> PEX

Ah ha! I have found two donor cards with a RT8120A 🙂

TODO Add picture of replaced VMem PWM

Well, there’s some improvement in that there is some VMem (~0.6V) and some PEX (~0.4V), but those aren’t high enough. Strange to have partial voltage. Need to dig deeper, maybe the feedback curuit isn’t working right? Strange PEX Is also low..

I have to imagine there is something disfunctional about the VMem VRM, the PWM gets pretty hot (it was from a very old donor board). I would also say the single VRM capacitor seems hotter than I would expect. Maybe it is faulty and that could have lead to the original issue? I should next try replacing the capacitor.

I have replaced the capacitor, but things have not improved. Possibly that memory VRM is faulty the donor board it came from was in quite bad shape.