use PCIe GPU hft private equitries funds can frontrun

requires only a single six-pin PCIe cable to feed its modest 120W TDP, 

AMD gpu is not have best track record in power use

TECHNICAL SPECIFICATIONS

GPU :

Hawaii (GCN)

Process :

28 nm

Shader Units :

5632 (2 x 2816)

Texture Units :

352 (2 x 176)

ROPs :

128 (2 x 64)

Core Clock :

up to 1018 MHz

Memory Clock :

1,250 MHz GDDR5

Memory Bus :

512-bit

Memory Bandwidth :

320 GB/s

Memory Capacity :

8 GB (2 x 4 GB)

DirectX, Shader, OpenGL :

11/5.0/4.2

Max. TDP :

500 W

Aux. Power Connector(s) :

2 x 8-pin PCIe

Min. Power Supply :

50A combined current to the 8-pin PCIe power connectors

gpu and frontrun

"Front running" may refer to the practice of high-frequency private equity traders (HFT), as a form of "legalized" front running,^[4] where they connect to various exchanges with direct peering fiber networks and try to detect orders as they propagate from a broker's order router and then use fast transit paths to beat those orders at reaching their destination. HFT private equity fund traders use expensive equipment, especially GPUs and FPGA technology to place many small orders at very high speeds that indicate buying/selling pressure and attract HFT momentum. Those with the most expensive equipment to shorten lags in reaching other exchanges then anticipate more orders and so to place orders on those exchanges to catch the rest of the order, at a more advantageous price.^[5] According to an estimate from Frederi Viens of Purdue University, profits from HFT in the U.S. has been declining from an estimated peak of $5bn in 2009, to about $1.25bn in 2012.^[6] Also, according to Harvard Political Review writer Austin Tymins, HFT private equity fund Citadel LLC made billions of dollars front-running the trades of large institutional investors, many of which are investing on behalf of middle-class clients.^[7] These estimates show that Citadel LLC makes several times of all of the HFT private equity funds profits singlehandedly, using such strategies. John Dizard of Financial Times argues that this behavior takes advantage of political reforms in fragmented market structure and the lack of liquidity in the National Market System.^[8] Using HFT strategies, large limit orders can be "front-run" by "order matching" or "penny jumping". For example if a buy limit order for 100,000 shares for $1.00 is announced to the market, many traders may seek to buy for $1.01. If the market price increases after their purchases, they will get the full amount of the price increase. However, if the market price decreases, they will likely be able to sell to the limit order trader, for only a one cent loss. This type of trading is probably not illegal, and in any case, a law against it would be very difficult to enforce because of Regulation NMS.^[3]

two sigma investments is uses gpu too

have many refs say two sigma investments is uses gpu too

https://people.maths.ox.ac.uk/gilesm/talks/pp10_key.pdf