(under development, will update as I go)
Seller Note “Dead”
Summary
- The Titan Black could be seen as the king of 7-series graphics cards, with the core of a GTX 780 TI, but with double precision arithmetic enabled and a much more practical 6Gb. Despite this, versions of the GTX 780 TI could probably still slightly outperform it. However, these days 6Gb or VRM is much more useful that 3Gb for HD gaming, which is the most any 7 series could reasonably be expected to do. I would think a strong customisation would be adding some kind of double-sided waterblock cooling, as in my experience, even the GTX 780 will naturally boost much higher with strong cooling and a duel sided cooler would help protect the VRAM chips on the back of the card – I have a suitable cooler so might try this 😉
- I bought two of these very cheap (£8 each) and didn’t expect much, both missing fans. When I received them, I measured them both to have only 1.2Ω on their VMem rails and I jumped to the conclusion they both probably were dead and have faulty cores. So I put them to one side, almost directly in the scrap area. However… I thought I should at least give them a proper look. Injecting 1v into the VMem rails revealed that each in fact has 1 shorted VRAM bank, which can be fixable! So, game on after all! 🙂
- Resistances
- Vcore – 3.7Ω
- Vmem ~ 60Ω
- PEX – 196Ω
- 5V – 1.37KΩ
- 3.3V – 578Ω
- 12V Slot 3KΩ 6pin 4KΩ++ 8pin 4KΩ++
These cards have VRAM chips on both sides of the card, making a very useful 6Gb of VRAM.
One has a damaged component on the back:
I need to try removing the shorted VRAM chip from each card and see if the VMem rail resistance rises.
Card A
Starting with the one with no ripped component. Lets test by injecting <1v into the VMem rail and see if anything heats up.. and yes, one of the chips does:
Quick update, I removed the seemingly short VRAM chip.
But the short remains, possibly I accidentally messed up some nearby components.
No nearby components seem impacted and injection <1v again reveals the adjacent VRAM chip on the same side is also shorted.
TODO Add thermal images.
Removing the 2nd shorted chip restores the VRAM chip to a much more healthy 60Ω. Whilst this seems like progress, it could be a bad sign that 2 chips have shorted.. The only way I know how to find out is to replace the 2 chips with new VRAMs that I have in stock – Hynix H5GQ2H24AFR-R2C
After removing the other VRAM chip with a preheater and 450 deg hot air, I cleaned up the old solder by first applying a little leaded solder and then wicking the pads.
Then I applied a light coating of flux gel.
Then, before using the preheater and hot air to resolder the 2 new chips, I very slightly misaligned the chips so that when the solder balls melt, the chips can be seen to visibly shift into place so I know when to remove the heat.
After the board cooled down, I remeasured the Vmem resistance and it is still close to 60Ω (say 50Ω and climbing due to the board still being warm). Time for a test, and …. nice! Great to see the card working against the odds, now I need to find a decent cooler! 🙂
TODO picture of the card working and more stress testing (limited without a fan on the remains of the cooler it came with.
Card B
Lets hope we get lucky with this one too! 🙂
Well, hopefully a good start, after quite a lot of heat (450 deg C, fast airflow) and 100 (deg C preheatering from underneath), the shorted VRAM chip was removed and the VMem rail is now back to ~60Ω like it should be. I’m still not sure why card A had two shorted chips…
Having replaced the faulty chip (following the same steps as shown above), the card does not post and the core remains cold… time to check voltages… it seems the the 5V rail is too low at 0.8V, so no others like VCore, VMem and PEX can start.
The buck converter responsible is a APW 7142 https://www.alldatasheet.es/datasheet-pdf/view/316999/ANPEC/APW7142.html
It has adjustable output between 08V and VIN
Need to check:
- VIN (should be 12V) – 12V
- EN -12V
- COMP -0.8V
- FB – 0.8V
It’s quite possible that the buck converter is faulty. I will look for a replacement to try from a donor card. As an aside, I am not sure why there was a shorted VRAM module and the 5V rail is missing? Quite possibly other problems will emerge…