Seller Note “The fan spins as it should. However, the screen comes on but then freezes, or goes blank. I have tried in two machines, and with different drivers but the same result.”
Summary
- Nice cosmetically, warranty sticker, boxed.
- Resistances
- VCore – 0.9
- VMem – 93
- PEX – 178
- 5V -509
- 1.8V – 775
- 3.3 – 4.6K
- 12V PCI – K++
- Loads drivers, but has stability issues (blank screen, crashing windows in the end)
- Think the HDMI port may be damaged, the display port and DVI both work
- Small black square artefacts sometimes appear and go, unusual artefacts
- Under light load (GPU-Z stress test)
- TPD shows as 0
- Vrel limit is present, even though the GPU load is only ~60%
- Possible VRM issue?
- Initially, I had suspected a possible memory issue with this card. However, MATS passes, MODS 400.250 says the card is unsupported for test 178? The same blanking of the screen can also be noticed. I am now thinking there is more likely a power delivery issue, hopefully not a core issue.
- Since it runs to an extent, I should probably investigate the VRM phases with an oscilloscope
Investigating the VRM phases
Before windows, both the Vcore VRMs and VMem VRM look clean (likely boot mode). When Windows is reached and the driver is loaded, the phase nearest the PCI slot remains mostly clean (some minor instability). However, the other VCore VRM appear rough.
Taking a closer look at the VRM design we have a seemingly simple setup:
high-side SM4377 N-Channel MOSFET https://datasheetspdf.com/pdf/1484615/Sinopower/SM4377/1
low-side On-semi 4503 N-Channel MOSFET https://www.onsemi.com/pdf/datasheet/ntms4503n-d.pdf
3/2/1-Phase Synchronous-Rectified Buck Controller https://datasheetspdf.com/pdf/1347333/uPISemiconductor/uP9509/1
Phase 1 (near PCI) – H/S Gate resistance ~30K, L/S Gate resistance ~10K (both)
Phase 2 (problem phase) = H/S Gate resistance ~30K, L/S Gate resistance ~3M (both)
Not sure why the low-side gates are so high on the problem phase? Looks suspicious, could be a resistor somewhere that has gone high? Will have a dig…
Actuallly, comparing to other similar GTX 1050 VRMs, these readings appear in range. My next idea will be to try replacing the high and low side MOSFETs, but I need to order them first.
Update 19/12/2022 – More testing notes
I got the high-side MOSFETs, so decided to retest before replacing the phase that appeared rough before. However, I don’t see a noticeable difference between either VCore phase at rest or under load, under load, both appear a little rough.
Other observations
- The card loads drivers and can display as a primary card (display port), but the display soon becomes laggy, occasion ‘black square’ artefacts appear and the PC will eventually crash.
- As a secondary, the card seems stable and will tolerate some load (GPU-Z render test). However, TDP shows quite low and PCI slot voltage/power and board power draw all read 0 all the time.
- Voltages appear normal, no obvious dips like with some of my other 1050 TIs.
- HDMI port seems to be faulty, Display and DVI work.
Here you can see the behaviour as a second card (no display lead connected). The middle part of the graph are the stats when the GPU-Z render test is running. The core and memory even appear to run at their proper higher clock speeds.
I have a few GTX 1050 TIs now that all seem to have power-reporting anomalies (some even unstable voltages) and are OK, until a display is connected. I cannot be sure yet, but am concerned these could all simply be core issues.