What is the most realistic power cap for Extreme 1650 LP overclocking shunt mod?
What is the most realistic power cap for Extreme 1650 LP overclocking shunt mod?
I have managed to install a shunt mod in a PCIe slot for a powered GTX 1650 Zotac low profile model card.
The standard wattage for a normal GPU is around 66-68W, but I've found the highest tested so far is 75W.
For me, going up to 80W or 90W would be reasonable, though 90W would add 15W over the 75W specification.
My friend has achieved similar results with a shunt mod on an RTX 3090 drawing about 100-105W through the PCIe slot.
I wouldn't exceed the recommended wattage, particularly considering the specific motherboard. Some boards include 6 pin PCIe ports to aid in power delivery, but these are built for reliability rather than constant operation.
is there any documented safe tolerance range around 25 to 50% higher than the rated value? most parts are typically rated 25 to even 50% above what they need for reliability. it seems like a concerning design choice if a pcie slot would fail under a 76w instead of 75w requirement.
75w is considered the maximum expected wattage. It represents the capacity that sockets, traces, and power delivery systems are designed to manage. Some time ago, Rx480 chips were tested to handle up to 94w from the PCIe slot. This issue was quickly resolved through microcode updates in driver updates and firmware changes in later cards, which capped performance at 75w.
The explanation is straightforward. Motherboard designs vary slightly. Z boards feature wider traces, use higher quality copper, and have more powerful VRM units. B and H boards don’t require such high power because they lack overclocking support. These specifications can differ by vendor, board model, and generation. Therefore, one B board might handle up to 80w, another B could go up to 85w, a Z up to 95w, while another manufacturer might allow only 5w less for more balanced boards. It’s impossible to predict precisely.
However, all boards should generally support 75w, except for some unusual HP models that are limited to 25w.
You’re trying to get the most from a small amount of effort. It won’t work perfectly, and pushing it will likely lead to one of two outcomes: either damaging the card or destroying the motherboard.
I included a 6-pin connector to the graphics card, which should theoretically divert the current away from the PCIe slot and straight through the 6-pin PCI cable directly to the power supply. There are some pictures available. Should I need to fully isolate the PCIe slot and +12V pins to stop any current from passing through them? Or will electricity simply follow the path of least resistance via the soldered 6-pin connector wires?
Your theory is incorrect. If the card extracts 50w from the slot, you're actually offering a parallel source, meaning the slot will continue to supply around 25w±, just like the cable. You'd have to remove the power and ground pins on the card to separate it from the slot or any connected components.
It would be quite challenging to separate all the ground pins! I don't think it's even practical. There are so many of them.
Isolating just the +12v pins would be a real hassle.
If power is shared as you mentioned, as long as it draws under 75w through the PCIe slot.
Drawing at least half through the 6-pin connector should work fine.
With the 15mohm resistor shunt modifications, it shouldn't exceed 90-95w maximum.
I got it to work. it works like I set up. didn't even need to do anything with the pcie slot pins or anything either.
it just "works"
However the ground wires aren't sinking as much current as expected. 1-2 amps and it fluctuates somewhat erratically. kind of concerning.
though the +12v positive wires of the 6-pin do draw the correct current. seen up to 84W or about 7 amps of current draw. MSI Afterburner predicting about 84-90w at the same time.
So not a big deal? There are 68 ground pins in the PCIe slot! That's much more than the total number of +12V pins, which were only 5. After adding the 6-pin connector, it seems the sheer number of ground pins available in the slot might be more important than the wires I soldered from the connector to the capacitor grounds. Or maybe I placed the grounds in a less ideal spot to reduce current draw through the PCIe slot.