Monday, August 18, 2014

GAW Miners Fury Mods - Voltage [MOD]


How many hashes per second does the GAW Miners Fury leave on the table in the compromise between performance, power usage, and stability? We've looked at replacing the Fury's power supply and improving the Fury's cooling system. Now we want to make it go faster.

The GAW Miners Fury is getting a little long in the tooth. It's about three months old. That's a lifetime in the crypto coin mining world these days. Squeaking a little bit more out of the Fury might put some life back into the old boy.

There's really only one way to get more out of a computing device and that's overclocking. The Fury's I've seen can all be overclocked a little, but somewhere between a clock rate of  350 and 354 the Fury's ASICs start starving for voltage. To overclock past that, and of course we want to go past that, we need to provide the ASICs with more juice.

If you're an electrical engineer or good with electronics you might be able to look at the Fury's circuit board and figure some stuff out. The Fury uses a Texas Instruments TPS40193 buck controller to provide power for the ASICs. A look at the TPS40193 datasheet will show you how it works and how to tune the output. It sounds simple in theory, but it's a little more difficult in practice.

Luckily, there's the Internet. Part of what makes the Internet great is that some people like to share the hacks and stuff they do so that other's can follow in their footsteps. Some of those people share what they've done, and so on, and so on. Bitcointalk.org is where all that happens for crypto currency and nst6563's thread, Fury/Blizzard tuning and mods, has everything we need to know about volt modding the Fury. The volt mod below is based on information in that thread. I recommend reading the original Bitcointalk thread and sending some love nst6563's way if you find it helpful.

Volt modding the Fury almost certainly requires upgrading the cooling system and power supply. See GAW Miners Fury Mods - Power Supply and GAW Miners Fury Mods - Cooling for more information on those mods.

Fury Volt Mod

Below is picture the Fury's circuit board with a green box representing the area we're concerned with.


Here's a closeup of the area of interest with a green box around the resistor that needs to be replaced. It's a 10.5k, 0603, surface mount resistor with the number 912 on it. It's annotated as R10 on the Zeus schematic which can be found in Fury/Blizzard tuning and mods thread.


I used a hot air rework station, similar to this one at Adafruit, to remove R10 and replace it with a 7.5k resistor.

More Power

Raising the ASIC voltage allows the clock rate to be set up to 382. Unfortunately, that's as high as the Fury can go. There's something in the mining software or the Fury's controller that doesn't allow clock rates higher than 382. I really want to see how far a Fury can be pushed but no one has figured out how to remove the limit.

Stock voltage Fury with modded cooling and ATX PSU

Volt modded Fury
The results show that a stock voltage Fury starts accumulating more hardware errors and doesn't consume any more power somewhere between a 350 and 354 clock rate. The same Fury with the volt mod can go all the way up to 382 clock without a significant rise in hardware errors.

Pool hash rates can be all over the place. The hash rates in the table above are averaged from at least 3 samples after hashing for several hours to allow the numbers to stabilize. With a clock rate of 382, 1.6 mh/s hash rates are not uncommon. Best case, the volt modded Fury does around 200 kh/s more than the stock Fury with a clock of 340. That's about a 14% increase. Not too bad.

No Free Lunch

Of course, you can't make more hashes on the same chips without making more heat and, worst of all, using more power. A volt modded Fury, hashing as fast as it can, consumes about 70 watts. It costs 23% more to run a volt modded Fury and you only get, at best, a 14% higher hash rate. This would probably be acceptable in the early days of the Fury when the cost to run it wasn't such a big percent of what it produces. Today it's probably not worth the effort. If it is, it wont be for much longer.







Fury Test Bench
Criteria: Watts, amps, volts, hash rates, rejects, and hardware errors are an average of several samples. Data was recorded after running for several hours to allow for the numbers to stabilize. Measurements were made with the following tools: Kill A Watt Electricity Usage Monitor, Fluke 62 MAX Plus Infrared Thermometer, Fluke 87-V Digital Multimeter, Extech EX330 Autoranging Multimeter.

1 comment:

  1. OK we need more hash power less voltage! Go for it!

    Almost kidding :)

    ReplyDelete