Topic: Soldering vacant power circuits on gpu (6850/6870)? (Read 2140 times)
Sick
By the way, if you really want to try this then check out EVGA's ePower board -> http://www.evga.com/products/moreinfo.asp?pn=100-ut-0400-br&pwindow=specs&family=accessories%20-%20hardware&sw=4
To see an example of what I was talking about above, TiN has a great guide on unleashing the full potential of the GTX 680(his mods will get you 1.7 GHz to 2 GHz under LN2 with the right core). http://kingpincooling.com/forum/showthread.php?t=1682
To see an example of what I was talking about above, TiN has a great guide on unleashing the full potential of the GTX 680(his mods will get you 1.7 GHz to 2 GHz under LN2 with the right core). http://kingpincooling.com/forum/showthread.php?t=1682
You'll see that very often on production boards. The PCB may be over engineered and a phase on the VRM is removed as the lesser design is proven to be satisfactory with sufficient overhead to maintain proper voltages and current supply for the lifetime of the card.
There are also cases where the same ASIC on the same PCB is used except the ASICs have various portions of the chip disabled. This is the case of the 5870/5850/5830. In theory they could all run on the same PCB but the 5830 wouldn't have the current requirements of the 5870 and the components on the 5830 PCB would be "wasted".
The only time you could realistically need another phase is if you are hitting the overcurrent protection on the GPU or suffering major voltage droop. Back when I was overclocking GPUs under liquid nitrogen I would frequently override the OCP values in the VRM controller. This trick worked by decreasing the signal voltage such that when the GPU was pulling 150 amps the VRM might only sense 100 amps passing and therefore not shut down the VRM due to hitting the designed OCP. Tricking the VRM is much easier than augmenting it.
That all being said I would never advise doing this to a card to be used for mining. You put yourself in a very unsafe scenario where it is possible for the VRM to roast itself and potentially damage your board, other GPUs and PSU. Adding another phase to a card that has the available pads may or may not work depending on the VRM controller and how it is configured. Frequently the controllers support a number of phases(2/3/4/6/8/etc) but require pins to be pulled high/low to indicate how many phases to operate with. In this case, even if you solder everthing up perfectly your additional phase will not receive signals to the low and high side fets to trigger and therefore won't do anything.
tl;dr - don't bother.
There are also cases where the same ASIC on the same PCB is used except the ASICs have various portions of the chip disabled. This is the case of the 5870/5850/5830. In theory they could all run on the same PCB but the 5830 wouldn't have the current requirements of the 5870 and the components on the 5830 PCB would be "wasted".
The only time you could realistically need another phase is if you are hitting the overcurrent protection on the GPU or suffering major voltage droop. Back when I was overclocking GPUs under liquid nitrogen I would frequently override the OCP values in the VRM controller. This trick worked by decreasing the signal voltage such that when the GPU was pulling 150 amps the VRM might only sense 100 amps passing and therefore not shut down the VRM due to hitting the designed OCP. Tricking the VRM is much easier than augmenting it.
That all being said I would never advise doing this to a card to be used for mining. You put yourself in a very unsafe scenario where it is possible for the VRM to roast itself and potentially damage your board, other GPUs and PSU. Adding another phase to a card that has the available pads may or may not work depending on the VRM controller and how it is configured. Frequently the controllers support a number of phases(2/3/4/6/8/etc) but require pins to be pulled high/low to indicate how many phases to operate with. In this case, even if you solder everthing up perfectly your additional phase will not receive signals to the low and high side fets to trigger and therefore won't do anything.
tl;dr - don't bother.
I have a Sapphire 6850 lying around, and noticed solder terminals to attach a secondary 6 pin PCIE power cable. It seems the solder points are connected to to an incomplete power circuit. Has anyone reconstructed such a circuit on a graphics card? A 6850 flashed to a 6870 remains a 6850 functionally; would an additional power circuit alter the functionality or permit performance as a Sapphire Toxic (OC) card?
Example pic:
Example pic:
Jump to: