-   EVGA Overclocking (
-   -   EVGA EPower V Living Project Log (

MLLRKLLR88 09-21-2017 03:39 PM

EVGA EPower V Living Project Log
This is my first time posting here with my new account, I am excited to be back!

Vince and TiN saw that I was playing around with the older generation EPower’s and they sent me a sample of their newest creation to play with. Firstly, I am incredibly grateful for the sample, thank you!

It’s not just beautiful, it’s packed with features. There is no need to reinvent the wheel and list all the features and capabilities, just read them for yourself in this incredible guide the masters made for us.

My Plan
I acquired a nicely binned GTX480, so I wanted to make that my first project. My goal for this GTX480 was to try and eliminate the interconnection bottleneck. To do this I chose an incredibly thick piece of copper. I sized the copper so that the epower would be as close as possible to the core plane on the GTX480.

Just for fun I decided to calculate the maximum theoretical current throughput of my copper piece. Obviously I won’t be able to attain that throughput, not even half of it, but it was a fun exercise knowing the max potential.

Here is the math.

The final design changed a bit after I did those calculations since I decided to notch the copper for better alignment. 
Soldering down the copper bar

I filed the enormous copper bar at an angle so it would mount on the GTX480 VCore plane nice and flush for maximum throughput.

The Card Preparation
I wanted to try memory EPower because it is available to me. I have seen in the past that most memory does not scale a whole lot with voltage and that memory phases are usually adequate to handle the load. I knew there would be a large gap between the memory planes, but I assumed copper plate could bridge the gap without issues.

Here you can see I removed the surface mount electrolytic caps on the core to open up the power can ground planes a bit more. My plan was to relocate them but I ended up choosing better caps and piggybacking them on the back.

Here is the GPU plane perpetration

Here you can see a “token” rear ground piece of copper plate that will connect the rear EPower ground to the top of the card under the power connectors

EPower Mount Rev01
I did my best to maximize the available real-estate and keep all he copper as short as possible. I decided to go ahead and do remote voltage sense straight away. Remote sense is by far one of the best features of EPower and some competitor power boards don't offer this truly game-changing feature.

EPower Mount Rev02
I was shocked at just how bad the memory performed from the initial EPower mount. I spent a while forming the copper just right and making it as flush and large as possible. However, given the HUGE distance it needed to travel I understand and accept the failure.

In the end I think it would have been wise to keep the stock memory VRM intact. However, before I try and restore the stock memory VRM, I will see if I can help the memory with more copper!

EPower Mount Rev03
After testing the Rev02 additional memory copper it became clear I would not be able to achieve pre-EPower results. In this revision I restored the stock memory phases as well as added input and output caps for the memory. While I was at it I decided to mod the PLL and add some voltage displays for memory and PLL.

I was a little bummed to remove the EPower memory but I need to go where the results take me. It was a good learning experience and I certainly added to my knowledge base. The core EPower is so amazing that it makes up for the memory not working well :)

Test Results
I saw a huge gain in passable core clocks running 3DMark Vantage given the same cooling and same voltage. I expected to see some gain, but it was certainly more than I imagined. I tested voltage scalability with a modest +15mv above stock and its scaling nicely!! It looks like the card is doing exceptionally well with EPower core, so the prospects are exciting for LN2!

What I didn’t expect is the memory to be so terribly bad after EPower. I knew going in that there would be little to no scaling, but I expected to be close to stock performance. In fact, the memory would not OC at all with 1.52v on the EPower.

TiN 09-21-2017 07:02 PM

You can have remote sense for memory too, using two resistor pads same way as the core. Resistors are MFP/MFN on the EPOWER back.

But it's not uncommon to have bad memory OC with external power source, as memory power delivery need to be clean, and having connections not helping that.

MLLRKLLR88 09-22-2017 11:09 AM

I see the MFP/MFN resistors on the back, they are conveniently located so one can easily get to them after the EPower has been mounted. Its so nice that you made remote sense so easily available for memory and core, instead of manually finding the sense lines on the PWM.

This EPower is AMAZING!!! :D

In this particular case however I don't think remote sense for memory will help. There is NO memory VDroop during load, so I don't see a reason to compensate for it. Ultimately adding more copper didn't help much either. Its not the EPower's fault, or the mount quality, I think it all comes down to the physical distance from plane to plane.

For Rev03 I am going to revert to stock memory.

TiN 09-25-2017 01:13 AM

You must use scope to see realtime drop performance and peak-peak noise, which is really important for memory. DMM will show nothing.

nrpeyton 11-11-2017 06:38 PM


Where does one go; to learn this stuff?

(I.E. to be able to reach the ability level required to actually work with for example; an epower board and an old nvidia GPU)?

Online courses? College? Or just self taught amongst friends?

All times are GMT -5. The time now is 09:13 PM.

Powered by vBulletin® Version 3.8.3 Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.