JPMorgan Rolls Out (Another) FPGA Supercomputer

An anonymous reader writes "JP Morgan is expanding its use of dataflow supercomputers to speed up more of its fixed income trading operations. Earlier this year, the bank revealed how it reduced the time it took to run an end-of-day risk calculation from eight hours down to just 238 seconds. The new dataflow supercomputer, where the computer chips are tailored to perform specific, bespoke tasks (as explained in this Wall Street Journal article) — will be equivalent to more than 12,000 conventional x86 cores, providing 128 Teraflops of performance."

Comments

All this..

[ackthpt]

So they can project how much money to borrow from the Federal Government the next time they have lent beyond sane limits to property speculators or invested in schemes even Mandelbrot wouldn't be able to simulate.

Re:All this..

[fsckmnky]

Let's hope the federal government regulators are paying attention this time.

It would suck for them to be confused by the cool new computer and unable to seperate systemic or institutional risk from faster calculating devices.

Re:All this..

[timeOday]

Look at it the other way, by rapidly exploring lots of different risk profiles, they're trying to avoid getting into trouble in the first place.

Besides, the technology to rapidly reconfigure FPGAs for specific tasks could have a lot of applications. I guess the obvious question is whether the FPGA approach can win over mass-produced general purpose CPUs (the article says yes), and also over GPUs, which are increasingly general-purpose SIMD units.

Re:All this..

[ackthpt]

Let's hope the federal government regulators are paying attention this time.

It would suck for them to be confused by the cool new computer and unable to seperate systemic or institutional risk from faster calculating devices.

Wishful thinking. Wall Street moves at the Speed of Light with all these computer trades now. Federal regulators need a super computer to keep an eye on JPMorgan, et al.

"You were insolvent 23 times today, for a total of 3.77 seconds. Federal guidelines mandate not being insolvent more than 15 times per day, over 1.78 seconds."

Re:All this..

[Omega Hacker]

they're trying to avoid getting into trouble in the first place.

No, they're looking to avoid getting into the kind of trouble they a) comprehend and b) actually care about. It's all the other kinds of trouble that are wreaking havoc with this country and planet right now.

Re:All this..

[fsckmnky]

I'm all over reconfigurable computing as a technology. I didn't intend to imply the technology isn't cool, or even that JP morgan is evil. I've been personally advocating the performance enhancements and power improvements attainable via reconfigurable fabrics or hardware assisted acceleration for awhile now. Most people start lecturing me about bus contention when I mention it, which makes me cry. ;(

Re:All this..

[Samantha Wright]

I think you mean "So American Slashdotters can ignore the geek appeal of the topic and continue bitching about how destructive their culture of competition is."

That's really what you meant to say, right? Right?

Re:All this..

[iluvcapra]

"A sound banker, alas, is not one who foresees danger and avoids it, but one who, when he is ruined, is ruined in a conventional and orthodox way with his fellows, so that no-one can really blame him."

--- John Maynard Keynes

Re:All this..

[Vaphell]

1. regulators are bought and paid for, besides they are not the brightest of the pack - if they were they would be farming gold on wallstreet themselves
2. regulations are always looking backwards at the last crisis, never predict origins of the next one
3. when everything is leveraged 30-50x there is nothing you can do to provide stability that is not make-believe
4. you don't need fancy regulation to crack down on good ol' fraud, you just make it harder for small players to comply

Re:All this..

[fsckmnky]

3. when everything is leveraged 30-50x there is nothing you can do to provide stability that is not make-believe

Except of course, limit the amount of leverage allowed for any specific market or player, as they already do.

Whether the limits are being sanely applied or set at appropriate levels is another issue.

Re:All this..

[jo42]

Look at it the other way, by rapidly exploring lots of different risk profiles, they're trying to avoid getting into trouble in the first place.

Oh my goodness, how naive you are. To 'make' money, all the financial industry needs to do is to add extra zeros before the decimal point to their account balances. Problem is they can't do that so easily -- this is something they still can't get away with. So they have come up with all of these 'financial instruments' and 'trading systems' where they shuffle shit around between themselves and their accounts, producing no true value or real wealth, but somehow still end up with extra zeros before the decimal point in those accounts.

Yes Timmy, the financial industry is one huge friggin' gigantic Ponzi scheme.

Re:All this..

[JoeMerchant]

So they can project how much money to borrow from the Federal Government the next time they have lent beyond sane limits to property speculators or invested in schemes even Mandelbrot wouldn't be able to simulate.

No, it's just fun building excessively expensive crap to justify your trading profits and look like you're "cutting edge." The only edge they are cutting is any shred of "all men are created equal" left in this country.

Re:All this..

[Anonymous Coward]

This is not a rapidly reconfigurable FPGA application. It is unrolling computation loops ideal for sequential processors into long compute pipelines implemented directly as hard logic. However from one calculation type to the next, the FPGAs are likely re-configured.

The algorithms are directly compiled into the FPGA fabric with the framework of the pipeline dataflow pre-defined. Schedulers also exist to coordinate the data flow to/from the pipelines and to manage data flow dependencies if the data needs to re-enter the pipeline.

The speed up occurs due to calculation latency overlap in the pipeline and parallelism with multiple simultaneous running pipelines. There is no overhead for data movement as there are in GPU's and CPU's. The data flows through the compute stages unhindered with data storage access occurring at speed during pipeline fills and pipeline empties.

Yes this technology showcases how FPGA's can compete with GPU's and multiple core CPU's. As an example, OpenCL is coming to one vendors' FPGAs and very soon at that. People who understand HPC and it's challenges understand how FPGA's offer them a more efficient solution that allows customizability. The tools to do this are in the early stages but the efforts to improve the ability to create HPC structures is in full earnest.

I work in FPGA's so that's how I know this. I can't really say much more . . .

Re:All this..

[qualityassurancedept]

I would guess they have a stake in some FPGA manufacturing process and this is just a way of bringing attention to it to trick investors into handing over their money... and anyway, since the easiest way to make money on wall street is to take it from other people who already have it, I would imagine that their calculations really just show them where other people are exposed to risk so that they can bet the other way or something. Whatever they are doing, it isn't fair, it probably isn't even legal and eventually the government is going to have to give them 1 trillion dollars to save us all from some apocalyptic starvation scheme based on the devaluation of the Euro or some ludicrous arcana in the mathematics of finance.

Re:All this..

[timeOday]

So, do you have an opinion on why Transmeta didn't pan out, or a reason to consider it irrelevant?

Re:All this..

[Anonymous Coward]

Yes, now instead of failing to think beyond the next quarter, Wall Street can fail to think beyond the next millisecond.

Cue apologist explanations of how these leeches are creating value and liquidity by sitting in the middle of every transaction.

Re:All this..

[fsckmnky]

I'm not really familiar with enough details of the Transmeta saga to really have an opinion on why they didn't take over the world.

The only thing that comes to mind, would be that they created an "x86 compatible" and that, when people were tasked with deciding what to standardize upon, Intel the real deal, or AMD the accepted competitor, Transmeta seemed like a bad choice. Anyone willing to deal with non-intel might have found ARM much more attractive. I dunno.

It's purely a guess on my part, and I'm sure I'll get blamed for being totally wrong. That said, I think I recall them selling off their technology, so it's not like their efforts were totally in vane. I would bet their ideas and IP have been absorbed by someone, somewhere, for something.

Re:All this..

[saleenS281]

Read another way: It's not murder if they can't find a body.

Re:All this..

[Beeftopia]

1. regulators are bought and paid for, besides they are not the brightest of the pack - if they were they would be farming gold on wallstreet themselves 2. regulations are always looking backwards at the last crisis, never predict origins of the next one 3. when everything is leveraged 30-50x there is nothing you can do to provide stability that is not make-believe 4. you don't need fancy regulation to crack down on good ol' fraud, you just make it harder for small players to comply

When things are going great, no one wants to change ANYTHING, no matter how outrageous, for fear of upsetting the apple cart and ending the party.

When things go bad, only then are people willing to change. Except of course, those still engaging in outrageous practices, as they are still making money.

As far as regulators being dim, sometimes that's true, sometimes not. IMHO, the far greater problem is the muzzling and influencing of regulators by the industries they are tasked to regulate via the politicians owned by those industries.

Re:All this..

[slimjim8094]

Hey, I know it's fun to hate on banks, and with good reason, but JPM was on the right, or at least neutral, side of this mess. They didn't do subprime loans, and they only took federal money by request (so taking federal money wouldn't become a scarlet letter) and paid it back as soon as they were allowed.

Let's hate on the irresponsible folks. It's not like there's a shortage.

Re:All this..

[danhaas]

If you can't run a sanity check over what your computers are doing, you aren't an engineer or administrator. You are a message boy, slave to the computer and to who really understands what's going on.

Re:All this..

[dave562]

They did not do the actual loans. They just packaged them up into mortgage backed securities that they (presumably) knew they were going to fail.

Re:All this..

[microbox]

Wrong, wrong, wrong.

We used to believe in regulations, which is why the meat in your freezer is edible. See, it does work.

But then the neoliberals moved in, and the Republicans were overtaken with faith-based economics, and now, regulations are the bane of everything good.

Brooksley Borne was just one of many people who predicted the financial meltdown. She was shut down by people who really believed that regulations are bad.

Regulatory capture is only part of the problem.

Re:All this..

[The Finn]

Guess who Dave Ditzel is working for these days...

Re:All this..

[definate]

if they were they would be farming gold on wallstreet themselves

Why would the regulators do the job someone in China could do for a lot less? I mean, we don't even know if they like playing World of WarCraft, let alone whether they want to grind for gold.

Re:All this..

[definate]

That's a terrible analogy, doesn't make sense at all.

Now, if it was they can't find the body, because of all the other bodies, due to everybody murdering, then it would make sense, but loses its punch.

Re:All this..

[ADRA]

"4. you don't need fancy regulation to crack down on good ol' fraud, you just make it harder for small players to comply"

It wouldn't help anyways if federal prosecutors don't actually do anything about it. The number of Sarbox related lawsuits as a result of the latest crisisis? I think its bordering on 0.

Re:All this..

[fgodfrey]

The real problem with reconfigurable computing isn't bus contention (that can be solved easily with a good system architecture). The problem with it is programmability. Let's say I want to sum an array in C. All I do is write a for() loop and add up the array. 2 lines of code.

Now, to do it on an FPGA in Verilog, I need: a functional block that's an adder, registers to store the sum, a RAM with addressing and output enables (etc) to store the array, a clock (or maybe more than one if the RAM timing is different) and a finite state machine to tie it all together. Clearly, there are tools that make these things easier, but not as easy as C (for instance, I need to know home long that adder takes to work so I can choose my clock speed).

Until I can take my C code (or, more likely, Matlab or FORTRAN) and compile it into an FPGA and not have to figure out whether I meet timing and provide hints to the place and route tools to get it all on the FPGA, reconfigurable computing is never going to catch on. Furthermore, it will ned integration in some way with a conventional CPU to move the results around. A number of companies, including one founded by Seymour Cray called SRC have promised that they would release a "C to FPGA Compiler". Thus far, it hasn't happened in the general sense.

At the moment, it really only makes sense in cases like this where they probably have a very specific application that doesn't change much and they can afford to hire a small army of electrical and computer engineers to do the FPGA development.

Re:All this..

[Anonymous Coward]

Yes, they need quite a bit of computing power to do that. But I suspect the thing that needs even more computing power in risk management is running simulations to know how to better avoid regulations.

Re:All this..

[Anne Thwacks]

You don't know how to do use FPGAs for this type of application. Its not like what you are doing. It needs a totally different approach. (I could tell you, but then I would have to kill you).

If you are using Verilog, you are not doing anything right.

You can take C code and compile it into an FPGA, and have been able to do it for more than 7 years. There are at least two perfectly good C comilers available for Xilinx.

These bits you got right: they probably have a very specific application that doesn't change much. You need a conventional CPU for management.

Re:All this..

[Anne Thwacks]

Mod parent up.

Re:All this..

[SuricouRaven]

This computer isn't for high-speed trading. While it is true that some high-speed trading technology can get rather silly, that is a different subject entirely.

Re:All this..

[Anonymous Coward]

C-to-Gate compilers don't give you the performance you need.

Maxeler, which is the consultant that JPMorgan used, has a completely different approach using data-flow computing. They also have their own java based tools for describing these data flow machines. it's actually quite neat.

You are correct in that if you're using Verilog, you are not doing anything right.

Re:All this..

[zippthorne]

The regulators weren't put in place to save you from the meat industry's shady practices. They were put in place to protect the meat industry from the public's fear and outrage....

Re:All this..

[Anonymous Coward]

"Brooksley Borne was just one [pbs.org] of many people who predicted the financial meltdown. She was shut down by people who really believed that regulations are bad."
Including our current Treasury Secretary (when he was head of the NY Fed)

Re:All this..

[Anonymous Coward]

Ah ha! Cognitive dissonance! Read some history, if you dare!

Re:All this..

[tehcyder]

The regulators weren't put in place to save you from the meat industry's shady practices. They were put in place to protect the meat industry from the public's fear and outrage....

mod parent -2 doubleplusungood.

Re:All this..

[LoRdTAW]

First off HDL vs. C or other HHL's is an apples to oranges comparison. There are ways to program an FPGA using C, but I believe it involves using soft CPU cores as a library of sort. You don't compile the C directly into an FPGA, but rather wrap a soft CPU core around it and then generate the HDL code. That or you just upload a soft cpu core and program that. I may be wrong but that is how it works from my understanding.

Second, most users are not going to write an entire configurable computing HDL design from the ground up. Most likely you are going to have an FPGA on the CPU, motherboard or expansion card along with drivers and API's that are invisible to the user. There are already PowerPC cores on FPGA chips and Xilinx just launched Zynq, an FPGA with a dual core ARM and tons of I/O. Intel is also going to offer an Atom CPU with an FPGA in the same package.

FPGA's can contain multiple "cores" that can be anything as long as they fit within the logic cell count of the FPGA. The architecture I envision is an endpoint block that interfaces the FPGA to the computer (PCI, Hyper transport, QPI, etc.) And a driver that talks to the device allowing for a common API, or at the very least raw communication. Then a configuration application allows you to download cores and then program the FPGA with the cores. Its all transparent to the user. Applications could take advantage of the FPGA directly by using a configuration API that automatically loads its own core(s) for data processing. So a video editing suite can load video codecs for en/transcoding, math programs could compile user algorithms directly into an array of parallel cores for fast processing (MatLAB already does this), audio programs can apply real-time effects through DSP cores in addition to encoding.

FPGA's are logic chips that can do anything. They can Interface to just about any bus/device and tie them together. There are already plenty of PCIe cards with FPGA's, memory, I/O and even high speed multicore DSP's. People have added compact flash cards to Apple II's, hacked the original Xbox by using an fpga as a bus sniffer. Students have implemented the NES entirely in an FPGA, there is also an FPGA arcade emulator that emulates the actual arcade hardware instead of using software emulation.
There are plenty of companies that offer INSANE FPGA platforms to build super computers, DSP farms and even emulation of complex hardware designs using FPGA farms. This tech is only going to get better and better.

FPGA Porn:
http://www.picocomputing.com/
http://www.hitechglobal.com/
http://nanobiowave.com/ATCA_FPGA_FARMS.aspx
http://www.edaptability.com/home.htm
http://enterpoint.co.uk/

I have a few FPGA boards myself. A small Cyclone board from Knjn and a 1.2M gate Spartan 3E board from Digilent. Knjn also has what I believe is the cheapest PCI(e) cards, there is one based on a cheaper Lattice chip that is 99 bucks but Its tied to a crippling and expensive annual license. These vendors have great starter kits:
http://www.digilentinc.com/ (sweet student discounts and great tools that are available for Linux. Their boards are targeted at education and research, lots of I/O interfaces and add on's called "pmods")
http://www.knjn.com/ (Cheap easy to program boards but I think their documentation is lacking. They are however, very helpful when you email them. I believe its a relatively small mom and pop company)

The Papilio is the Arduino of FPGA boards:
http://papilio.cc/index.php?n=Papilio.Papilio

Happy hardware hacking.

Re:All this..

[stdarg]

We used to believe in regulations, which is why the meat in your freezer is edible. See, it does work.

It probably has more to do with advances in technology like freezer cars for transportation. What evidence is there that without regulation meat businesses wouldn't still adopt advances in technology that make their customers happier and their products safer?

Brooksley Borne was just one [pbs.org] of many people who predicted the financial meltdown. She was shut down by people who really believed that regulations are bad.

There are thousands of people who predicted the financial meltdown. Many of them predict financial meltdowns every year. People who can predict ups and downs seem more worth listening to... it's just that there aren't any of them. Henry Paulson was famed for "predicting" the housing downturn and making a lot of money off it. Now he lost a lot of money with his next prediction. Doesn't that failure make you think maybe his first prediction was just luck?

Maybe Brooksley Borne is right and more regulation would have prevented things from going so sour. Maybe she's wrong, and unexpected factors like 9/11, the commoditization of technology, the increasing acceptance and dependence on debt by American consumers, massive price instability in food and gas, huge economic growth in developing countries, change in philosophy about debt obligations, increased life spans, emerging flaws in the social security and medicare programs, rising cost of education, or something else had a bigger effect than what regulation could have countered. Because I can tell you that all of those things have affected the financial industry.. I just don't know how much. And neither do you. And neither does Brooksley Borne.

Re:All this..

[Anonymous Coward]

Many companies use FPGA's not just for the speed, but it's easer to lock down a proprietary algorithm inside a FPGA/device driver installed inside a secure PC than it is in a blob of software on that same PC. Applications like CFD, financial analysis and encryption come to mind.

Re:All this..

[microbox]

It probably has more to do with advances in technology like freezer cars for transportation. What evidence is there that without regulation meat businesses wouldn't still adopt advances in technology that make their customers happier and their products safer?

History. Read about it.

Maybe Brooksley Borne is right and more regulation would have prevented things from going so sour.

She specifically attempted to investigate the derivatives market for fraud, and was shot down by those who believed that even investigating fraud would introduce inefficiency into the market. (Greenspan, Rubin and Summers specifically, with the help of a willing congress.)

It is amazing that people thing businesses will just behave because it is in their best interests -- /however/, we need police because private individuals cannot be trusted to behave in their best interest.

Re:All this..

[Uber Banker]

The parent could be referring to Montagu Norman. Not sure. Keynes had a great dislike for him, sort of, some admiration also. Read 'Lords of Finance' by Liaquat Ahamed, probably the finest text on the financial world that existed in the lead-up to Keynes ever written or to be written.

Re:All this..

[Anonymous Coward]

We need a tax on financial transactions. That will stop high-frequency trading.

http://www.thedailybell.com/3180/Fascism-Now-US-Congress-Europe-and-the-G20-Seek-Transaction-Tax-and-Backdoor-VAT-on-US

Re:All this..

[Beeftopia]

Brooksley Borne was just one [pbs.org] of many people who predicted the financial meltdown. She was shut down by people who really believed that regulations are bad.

There are thousands of people who predicted the financial meltdown. Many of them predict financial meltdowns every year. People who can predict ups and downs seem more worth listening to... it's just that there aren't any of them. Henry Paulson was famed for "predicting" the housing downturn and making a lot of money off it. Now he lost a lot of money with his next prediction. Doesn't that failure make you think maybe his first prediction was just luck?

You know what this entire financial crisis, US and European, is caused by? Loans going bad. Loans being defaulted on.

You have to ask yourself one question - "Why would a lender make a loan that he doesn't care about having repaid?"

Answer: Because finally, since the late 70s, lenders figured out how to separate themselves from repayment risk. They got better and better at it. Nowadays, a bank has a loan on its books for 60 days, then it's sold off. So they don't give a rip whether it gets paid off or not. It's a great system for loan originators. Write a number on a piece of paper, fill out some pro forma paperwork, sell it, and get a commission that's a percentage of the number you wrote down. It's the Dutch Tulip Bulb mania on steroids. The debt markets are completely flawed.

THAT'S the root of the problem - lenders not having repayment risk. It used to be that lenders cared about being paid back. So, very, very few bad loans.

However, there are a lot of big companies who pay a lot of money to politicians who want to keep this system going. And a lot of politicians who get that money who want to keep this system going. Because bribery is de facto legal in the US government ("The US government is the best money can buy"), the current system keeps going. This system involves loan originators making loans and then the US government, either through Fannie and Freddie, or through the Federal Reserve, taking the bad loans off the lenders/originator's hand, and giving them either face value for the loan, or just flat out buying it and generating the commission.

The Economist magazine was talking about all of this in 2003. I remember reading a cover story on the global real estate bubble in the early mid 2000s in The Economist. So much of this was foreseen by disinterested observers. But, as I've noted previously, when things are going great, no one wants to change anything because no one wants to stop the party. Especially those who are making money off of it. And it especially doesn't change when those benefiting from it are big companies and the politicians they own.

Re:All this..

[Beeftopia]

It is amazing that people thing businesses will just behave because it is in their best interests -- /however/, we need police because private individuals cannot be trusted to behave in /[their best interest]/the best interest of others.

FTFY.

Paid for with Corzine's customers funds

[Anonymous Coward]

Great! Maybe they can use their new supercomputer to find all the money that their UK branch helped steal in the MF Global collapse. Maybe they can even figure out the greatest mysteries of life, like whether there's any collateral at the bottom of the endless chain of rehypothecation that is the City.

Re:Paid for with Corzine's customers funds

[ka9dgx]

Amen... now that I understand rehypothecation, (having cake, borrowing against same cake, borrowing (again) against same cake... an unlimited number of times, then eating cake, is all legal thanks to "the city of london")

Any company with a division or branch in England can use this trick to hide/steal spectacular amounts of money. Because of the interlinked nature of the markets, Nothing is safe from this, anything in an "account" of any sort (stock, bank, savings, checking, etc) is only as good as the personal bond between yourself and the person who actively manages it. Since none of us have that connection to anything stock related... it's all a big Ponzi scheme, ready to implode.

Unfortunately, it may take all of civilization with it.

Re:Paid for with Corzine's customers funds

[dave562]

It will be unwound, messily, but it will unwind and society will go on. Do not buy into all of the hype about civilization coming to an end. There will always be people with excess capital that they want to invest. There will always be people with good ideas who need funding to get them off of the ground.

Re:Paid for with Corzine's customers funds

[ka9dgx]

All of our supply lines have had the slack engineered out of them in the name of "efficiency" in the last 20+ years. The US doesn't make everything it needs, let alone any given state... without international credit, next day air, and a continuous flow of liquid energy, it's all going to crash.

Eventually we'll recover... but the days of ever faster extraction of resources to be "consumed" then buried in a landfill, is rapidly approaching it's end. Unfortunately, we don't know how to work any other way.

238 seconds is about 4 minutes

[DRAGONWEEZEL]

for all you people who don't really think in seconds when seconds is > 60.

Re:238 seconds is about 4 minutes

[ackthpt]

for all you people who don't really think in seconds when seconds is > 60.

Put that sucker in the Crimson Assurance and it would go like the Space Shuttle!

right through the main feature and a dozen other screens in the cineplex as well!

Re:238 seconds is about 4 minutes

[confused one]

I write controls and test software. I think in multiples or fractions of 3600 seconds (one hour for the mathematically challenged).

Re:238 seconds is about 4 minutes

[Anonymous Coward]

ahem - it's the crimson PERMANENT assurance to you buddy

lol

Re:238 seconds is about 4 minutes

[Anonymous Coward]

I'm guessing you have 86400 memorized.

Re:238 seconds is about 4 minutes

[rubycodez]

lucky you, at least your sums don't change. I've done payroll software, varying amounts such as average workdays per month in a leap year and non-leap years, working hours per year (leap and non) are funny numbers most don't recognize. A year doesn't have exactly 52 weeks, either.

Re:238 seconds is about 4 minutes

[Anonymous Coward]

That's why I do all my currency calculations in floating-point under -ffast-math.

My what a big computer you have!

[Anonymous Coward]

All the better to fleece you with...

risk vs. electricity

[flying squirrels]

How much money are they spending in manpower, electricity and consumables by calculating risk? how about make a super computer to figure out how to solve the world debt. then we can all have thirteen billion terflops of power to torrent pr0n with.

Re:risk vs. electricity

[viperidaenz]

using FPGA's instead of x86 would probably consume a significant amount less electricity. using manpower is good in terms of the many men being paid for their efforts instead or a few ceos just pocketing the money as extra bonuses

Re:risk vs. electricity

[Anonymous Coward]

The electricity bill is no issue for these companies.
The sole purpose of adopting these technologies is to be able to evaluate assets realtime and to be able to make decisions or trade as fast as possible.

Re:risk vs. electricity

[NFN_NLN]

How much money are they spending in manpower, electricity and consumables by calculating risk? how about make a super computer to figure out how to solve the world debt.

Everyone knows the answer to that question already. Learn to live with less resources for each person, or figure out how to have less people.

Re:risk vs. electricity

[Anonymous Coward]

World debt? Debt is sort of a zero-sum game in this scenario. Unless the world is indebted to ET somewhere? Is the sun going to send a bill for all the rays it has been supplying us?

Re:risk vs. electricity

[benjamindees]

Actually global exports are over $300 billion more than global imports. So we are running a trade surplus with *someone*.

Re:risk vs. electricity

[tombeard]

More importantly, it makes it impossible for anyone to dispute their results. After all, no one else has exactly the same system so no one is better qualified to evaluate their conclusions. "No, we made the best possible choice at the time. You would know that if you had OUR analytic engine, but since you don't your speculations are baseless."

Re:risk vs. electricity

[Anonymous Coward]

/facepalm

Yet another misinformed soul parroting whatever it is he heard or read. Overpopulation is not the real problem, it's the fact that we're living in an outdated society with outdated structures (poor space management). Really, a lot of things in our society are inefficient and there's a lot of waste. We produce so much waste it makes me physically ill to think about, especially when you consider the problem could be mostly avoided with better management of our resources.

As for solving the world debt... that's easy. Abolish currency. Money is debt. Think about this for a second... if all of the world's debt could somehow be suddenly paid off, where is the money to pay the interest on that debt? Simple; it doesn't exist. It was never meant to exist. The whole system has become a device used to enslave you with debt. This isn't some crazy conspiracy theory. In fact, it's actually pretty obvious and it blows my mind when people get defensive about this and try to start defending the big red, white, and blue dick that's being jammed in their ass.

The problem here in America is that we are essentially letting a bunch of rich assholes, who do not care about the average Joe in any way, shape, or form, run our country and make all the important decisions for us. Hell, the internet might be censored by the time I post this. Are you people getting it yet? These people do NOT care about us! We could all be living in a much better environment if we would just quit with the distractions for a minute and deal with these assholes that are trying to fuck us over.

I know, I know... you feel "safe" with your $60k/year and full benefits, no sense in changing something that isn't broken (for you!). Well what about your children? What kind of world will they be living in? What about your grandchildren? Do you like the direction our country is headed in? Do you really think it's going to get better? It's only gonna get worse, folks... unless we do something. If I may, I'd like to suggest a possible route we could go with our society: http://www.thevenusproject.com/. Before you go off dismissing this and pointing out flaws/etc... again, don't be afraid of change and just ask yourself, "Would this be better than what we have today?" I think the answer is pretty obvious. Or you could bury your head in the sand trying to chase the ever-elusive "American Dream", but I'd prefer you enlighten yourself and help the rest of us out!

I think George Carlin said it best, "It's called the American Dream 'cuz you have to be asleep to believe it." In fact, the whole bit that's from is spot on: http://www.youtube.com/watch?v=acLW1vFO-2Q/.

Re:risk vs. electricity

[NFN_NLN]

/facepalm

Yet another misinformed soul parroting whatever it is he heard or read. Overpopulation is not the real problem, it's the fact that we're living in an outdated society with outdated structures (poor space management). Really, a lot of things in our society are inefficient and there's a lot of waste. We produce so much waste it makes me physically ill to think about, especially when you consider the problem could be mostly avoided with better management of our resources.

Inefficiencies like eating meat when insects are a far better at turning grains into protein?

The "Food and Agriculture Organization" (FAO) are making recommendations for the reduction of traditional meat and instead advocate eating insects. Why? To combat the growing demand for protein in world with too many people, which is only growing. This is only one example of one resource limitation... and they are all limited.

I don't know why you think we have a shortage of people or why the existing people need to reduce their quality of life to accommodate them.

And since you like George so much, here is a quote from your hero asshole: "It's irresponsible to have more than one child. Have one. Have one child, replacement value for yourself, that's all." - George Carlin

http://www.guardian.co.uk/environment/2010/aug/01/insects-food-emissions

Just FPGAs

[Anonymous Coward]

So, what this is saying is that lots of money and FPGA's can outperform conventional microprocessors when given a specific task. This is not really news.

Re:Just FPGAs

[fsckmnky]

This is not really news.

Nope. OpenCL 1.2 was released recently. That's news.

But it's certainly popular to blame banks with computers for everything. It's been happening for as long as I can remember.

As a boy being forced to attend church, I remember sermons on the evil computer somewhere in a room, nicknamed "The Beast," calculating everything everyone did. Boy were those people wrong. Turns out the computer was named google. ;)

too bad

[Khashishi]

These banks aren't just siphoning money, they are also siphoning talent away from more important projects. The people working on these things could be brilliant physicists or engineers, if they weren't sucked into the dark side.

Re:too bad

[fsckmnky]

The people working on these things could be brilliant physicists or engineers

You think the people implementing fast calculations for financial firms, can't and aren't using the knowledge and source code for projects in other areas ?

Re:too bad

[ackthpt]

These banks aren't just siphoning money, they are also siphoning talent away from more important projects. The people working on these things could be brilliant physicists or engineers, if they weren't sucked into the dark side.

It's not even dollars and cents. Hasn't been for years. It's points. It's all virtual and simply points. Just like whacking a goblin in some game, they're keeping score, not minding money.

Re:too bad

[Anonymous Coward]

In this day and age, it's either smart talent sucked into financial dark side or web-happy / mobile-app happy "yay yet another laundry app / angry bird / stupid web site"

Re:too bad

[SiMac]

You think the people implementing fast calculations for financial firms, can't and aren't using the knowledge and source code for projects in other areas ?

Correct. Do you think they are? That would almost certainly be in violation of their contracts...

Re:too bad

[fsckmnky]

Ah, so JP Morgan is labotamizing them upon termination. Now that is evil !

Re:too bad

[Anonymous Coward]

It's an interesting computing system and software project with a good financial backing. What's not to like? It's almost as if the spending of the military-industrial complex wouldn't be for the betterment of man kind (rolls eyes). ;)

Re:too bad

[Anonymous Coward]

These game companies aren't just siphoning money, they are also siphoning talent away from more important projects. The people working on these things could be brilliant physicists or engineers, if they weren't sucked into the dark side.

Re:too bad

[Anonymous Coward]

Finance is a simple game with simple human rules. Nature is far more complex.

The greatest thing a casino sells is myth, and one of its better myths is that "bankers" have some special talent. In reality the quant is a drop-out who lacks the imagination, tenacity and integrity to produce original scholarship.

Re:too bad

[Anonymous Coward]

I know plenty on unemployed science/engineering graduates, myself included.

Having a real job solving interesting problems, even if it's not at NASA, would be such a blessing.

It's not like people just magically find employment. And it takes a lot of capitol to do projects on the forefront of science; very few people can go it alone.

Re:too bad

[tukang]

If these algorithms are profitable they ultimately buy low and sell high, which is good for society. When you buy low, you're providing cash to those who need it most (therefore the willingness to sell low) and when you sell high, you provide the asset when it's needed most. Essentially, a profitable trader (computer or not) helps moderate markets by preventing them from going too low or too high, and pricing things correctly is important to society because resources are allocated by price. I know it's trendy to bash finance but it has an important function in society.

Re:too bad

[Anonymous Coward]

I find that interesting, as the few quants I know very well became bored with the lie which is American academia and decided to do something better with their lives then to fight bitterly over the small amount of money in their fields of fundamental physics research and now are happy, successful people contributing to the growth of mathematics and computing instead of lecturing other people who want to live the myth of a physicist. I instead became a pilot; I was smart enough not to get a PhD in physics.

Re:too bad

[JoeMerchant]

These banks aren't just siphoning money, they are also siphoning talent away from more important projects. The people working on these things could be brilliant physicists or engineers, if they weren't sucked into the dark side.

But, the dark side is filled with shiny baubles - why take a conventional job for $85K when you can start with a financial firm at $135K? That's a Porsche in the garage on day one, upgraded to a Ferrari after the 2 year lease is up. With conspicuous consumption like that, the vacuous model types start paying attention to you - what 20-something genius doesn't want that?

Re:too bad

[JoeMerchant]

Ah, so JP Morgan is labotamizing them upon termination. Now that is evil !

I've seen some contracts to that effect (not work for any competing firm or use, in any way, technology developed at the company for a period of 10 years after termination)... they told me it was standard, I told them it was my standard policy to not sign idiotic crap like that, they hired me anyway.

Re:too bad

[fsckmnky]

The article already pointed out, the hardware developer is a sub-contractor, and they also have customers in the oil industry. The rant I was responding to concerning "brain drain" I've seen on wired.com in FUD articles before.

I mean, the microprocessor itself started out as military technology, and it made its way into damn near every aspect of life. It's crazy to think wall street is sucking up all the science brainpower.

Re:too bad

[Nursie]

You think people that work on this stuff have anything in mind beyond personal enrichment?

Believe me, I know a guy that does some of this stuff. His opinions are that corporate morals are unnecessary, that we can't and shouldn't seek to blame or look negatively on companies for seeking profit without regard to the social, environmental or other costs, and that open source is basically hippie communism.

We used to argue about that sort of stuff quite a lot until I stopped speaking to him.

Re:too bad

[fsckmnky]

Well, open source is basically hippie communism. That doesn't mean it's not useful, or it should be banned.

A creator of a work should have the right to release it for all to use for free, or choose to recoup their investment via sales. It's their product, and their business, and their finances, who am I to mind their business for them ?

I'm sorry that guy was so selfish. But painting everyone in finance as a selfish greedy money lover who sells childrens organs in exchange for a new set of spinners on their Hummer, is a bit much. Alot like claiming everyone who uses open source, or is an open source enthusiast, is a hippie communist who eats dead skin from their big toe.

Perhaps it might renew your faith a little to know I work in Financial IT, and I use and support open source, and I'm not personally greedy.

It would be nice to live on a planet where money wasn't necessary. At this time however, that's not Earth.

Re:too bad

[fsckmnky]

When I said I support 'open source', let me give you a recent example.

The kid that works at the local gas-n-go said he was in college for computers, and that he would like to learn linux, but didn't have access to a computer with linux, because the only computer he had at home was his parents which had windows on it.

I returned an hour later, with a netbook I had lying around but wasn't using, that had a fresh install of linux on it, and gave it to him for free.

I think that was a non-greedy, socially responsible, resource re-distribution in order to avoid waste, open source supportive thing to do.

Re:too bad

[SiMac]

Your original post was:

You think the people implementing fast calculations for financial firms, can't and aren't using the knowledge and source code for projects in other areas ?

Using the source code is almost certainly against your contract, since it almost certainly becomes property of your employer. It doesn't take a lobotomy...

In any case, I think relatively few people go into science from Wall Street. Once you're fed up with it, it's often too late to go into grad school. Instead, they probably get a job at a technology company...No one in my PhD program was a stockbroker...

Re:too bad

[tombeard]

Economics Ph's gotta eat too y'a know.

Re:too bad

[Anonymous Coward]

JPMC only hires H1B Indians for developers anyway, so no big loss.

Re:too bad

[fsckmnky]

Using the source code is almost certainly against your contract, since it almost certainly becomes property of your employer. It doesn't take a lobotomy...

You are confusing copying the source code from the financial firm that you work for, and selling/giving/using it elsewhere, with ... developing source code that has uses in multiple industries, of which the financial industry is just one of. The article even points out how the hardware developer also has oil industry customers. I would be willing to bet, the hardware comes with some software, an application framework, compilers, tools, OpenCL kernels, that are in fact shared by the financial industry, as well as, in this example, the oil industry. I know for a fact that the general field of "Signal Processing" is applicable in nearly every sector you care to name, which results in a production set of IP that does get utilized across multiple industries. In your words, "It doesn't take a lobotomy ... "

In any case, I think relatively few people go into science from Wall Street. Once you're fed up with it, it's often too late to go into grad school. Instead, they probably get a job at a technology company

Technology companies don't use science ?

...No one in my PhD program was a stockbroker...

In case you aren't aware, a "stockbroker" is but one occupation in the financial industry. It's also one of the lowest status, high churn positions in the financial industry. Stockbroker is essentialy one step up from used car salesman, as that is what a stockbroker is, a salesman. There probably aren't alot of used car salesmen with PhDs either.

If you ask me, this attitude you have, is a direct result of the fact that you indeed were in a PhD program. Most people I've known in PhD programs look down on everyone who isn't also a PhD candidate, or doesn't already have a PhD, so I say unto you brother, check thy own head.

Re:too bad

[benjamindees]

Yeah, they buy low, right at the point when yet another pointless generation of nitwits is killed off by their insurance companies and healthcare providers, clutching a lifetime collection of assets to be liquidated at the bottom of the market. Then they sell high, when the government is going into debt buying up everything in sight prosecuting some otherwise needless war to kill off the excess of the latest generation of pointless nitwits' impoverished progeny. This is known as the "business cycle", since it keeps them all in business even though they are clearly just a bunch of bankrupt degenerate morons who couldn't turn a legitimate profit if their lives depended upon it.

Re:too bad

[The Askylist]

The strange thing about your characterisation of the "business cycle" is the reliance on governments influencing the markets. This seems to me to be one of the best arguments for small-government liberal capitalism that I have heard - doubtless the opposite of what you intended.

The fact remains that in any system, capital has to be found to allow business to flourish, and the more technologically advanced the society, the more capital is required.

The ability to process complex risk calculations very quickly, which of course allows more complex calculations to be considered, is a good thing - it ensures that the bank is less exposed to unexpected risk than it otherwise might be, and thus less likely to either go bust (in a liberal capitalist world) or to have its mistakes socialised via bail-outs (in the big-government, small mind world).

If anything, this sort of development is good because it will allow banks to make a reasonable return even with tighter capital requirements, and those banks which do not have this sort of capability will soon either acquire it or have to cut their trading activity.

Re:too bad

[Laser Dan]

If these algorithms are profitable they ultimately buy low and sell high, which is good for society. When you buy low, you're providing cash to those who need it most (therefore the willingness to sell low) and when you sell high, you provide the asset when it's needed most. Essentially, a profitable trader (computer or not) helps moderate markets by preventing them from going too low or too high, and pricing things correctly is important to society because resources are allocated by price. I know it's trendy to bash finance but it has an important function in society.

What you described is how trading SHOULD work, where a trader holds a certain stock for at least hours or days.
You are correct that this is good and stabilises things.

How it actually seems to happen these days is computers buy stocks on minor fluctuations, then sell them on another fluctuation a couple of microseconds later.
That doesn't seem to provide any useful function to me.

Re:too bad

[Nutria]

I mean, the microprocessor itself started out as military technology,

Only indirectly, since the first CPUs (from Intel, TI and General Instruments) all powered calculators.

Re:too bad

[Nutria]

My bad...

s/first CPUs/first microprocessor CPUs/

Re:too bad

[serviscope_minor]

Well, open source is basically hippie communism.

No, not really. Or yes, if you like. It can be considered from a whole variety of points of view from capitilistic such as loss leader (BSD license) or payment in kind (GLP)) to hippy communism to apolitical self interest. Redhat's $903 million revenue seems far divorced from hippy communism to me. RMS is certainly political but more in a revolutionary way to overturn the status quo than peace and love hippyism.

The thing is that for may problems going open source is the best solution. This is more of a statement about the practicalities of the suituation than about politics. This is why open source doesn't really fit political explanations.

Re:too bad

[Anonymous Coward]

Well, I work for a bank. Let's say I could work for a different corporation instead- I still need to pay my bills. I consider pretty much all of them evil. Every single corporation has responsibility to maximize shareholder value with no regard to society, environment, morals, law (if they can get away with it) or anything else. So I might as well sell my time to the highest bidder and retire early (if I manage that). I consider my time spent working to earn money as time wasted anyway, and time is money, so the more I earn, the less time I waste.

Please enlighten me if there is a more efficient way to live your life, I'm all ears.

Re:too bad

[Anonymous Coward]

The ones that know someone else would be the biological father and that they would just be bill paying suckers if they married the vacuous model types.

Re:too bad

[janzen]

Well, the argument is that it does add liquidity to the markets, and tightens bid-ask spreads, which are useful functions for investors and traders both large and small: your orders are more likely to get filled quickly, and at a better price.

However, high-frequency trading (HFT) does lead to some very interesting, um, emergent phenomena. Check out the beautiful graphics produced by Nanex, dissecting the Flash Crash and other assorted HFT weirdness.

Re:too bad

[JoeMerchant]

It's crazy to think wall street is sucking up all the science brainpower.

Agreed, but I think it's less crazy to think that wall street is sucking up all the economic "free will" of the government/99% of people and leaving space exploration and other "fun" endeavors in the hands of the 0.001%

Re:too bad

[SleezyG]

The thing about brilliant physicists and engineers is that they're really good at math. Of course they're going to take the higher paying job! If there are other "more important projects," then perhaps these projects need pay rates that are competitive with Wall Street.

Re:too bad

[Anonymous Coward]

These banks aren't just siphoning money, they are also siphoning talent away from more important projects. The people working on these things could be brilliant physicists or engineers, if they weren't sucked into the dark side.

Indeed! When I was RIF'd from my position as Principal Engineer for a top 50 software firm (with patents and major system innovations to my credit), I got a job 3 months later writing risk analysis software for the options/derivatives industry, at a serious salary+bonus uptick... I left that job a year and a half later because the hours (and commute) sucked! Goes to show, there are some things (like health and sanity) that are more important than $$! :-)

Re:too bad

[stdarg]

No matter how fast you buy and sell you're going to be fulfilling those functions. There are just more steps in between the start and the finish. The exception is when a group of algorithms start competing against each other and explore each other's unexpected corner cases. One bank will make a profit at the other bank's expense. There's no useful function to you, but also there's no harm.

Even when it devolves into a "flash crash" -- the long-term buyers and sellers will hardly be affected by volatility over the course of a few hours.

Re:too bad

[fsckmnky]

Yes, my mistake as well, I should have said "integrated circuit" of which a microprocessor is an integrated circuit.

The idea of the integrated circuit was conceived by a radar scientist working for the Royal Radar Establishment of the British Ministry of Defence, Geoffrey W.A. Dummer (1909–2002). Dummer presented the idea to the public at the Symposium on Progress in Quality Electronic Components in Washington, D.C. on May 7, 1952.[4] He gave many symposia publicly to propagate his ideas, and unsuccessfully attempted to build such a circuit in 1956.

Re:too bad

[Anonymous Coward]

Maxeler is the FPGA development consultancy company that was contracted to do this development work. They develop supercomputing solutions to a wide variety of fields including CFD, signal processing, image processing, video codecs, just about anything that requires custom ASIC's. That talent isn't being wasted.

Re:too bad

[jader3rd]

His opinions are that corporate morals are unnecessary, that we can't and shouldn't seek to blame or look negatively on companies for seeking profit without regard to the social, environmental or other costs

I suggest reading Robert Reich's "Supercapatalism". He was the Secretary of Labor for Bill Clinton, and he argues the same thing. As a society we created corporations to create goods, revenue and jobs. Expecting them to do something else is wrong. Corporations and neither good nor evil. They have no morals. All they can do is follow the law. Trying to pressure corporations into being moral is wasted effort, because if you succeed their competitors will overtake them, by doing what you were guilt tripping them into not doing. If you don't like the way things are going, change the law, and make sure there are resources to enforce the law. I'm not saying that corporations should be set free to do whatever they want, they should be given a framework and boundaries on how to operate in society which will be beneficial to all of us. But the same rules need to apply to all players in the market. Corporations have they morals, they aren't people.

Re:too bad

[Khashishi]

What would be even better is if the person who needed cash the most sold directly to the person who needed the asset the most.
Middlemen (retailers) are useful for tangible things like staplers because they increase efficiency in transporting goods and provide customers with a large variety of products at a single location. But the finance market is just a bunch of digital debts being shuffled around, never to be consumed. There is no value to finance hedging.

Re:too bad

[Anonymous Coward]

doubtless the opposite of what you intended

Don't quit your day job ;)

So... what?

[nicholas22]

You can get petaflops with a fraction of 12,000's x86 price, just use GPUs...

Re:So... what?

[ackthpt]

You can get petaflops with a fraction of 12,000's x86 price, just use GPUs...

Mrs. Schwartz, we've .. eh heh .. there's been some problems with the computers lately .. eh heh .. but rest assured, Mrs. Schwartz, we will have it all sorted out very soon and not a penny will be out of place .. eh heh .. just as soon as we figure how your checking account came to have a balance of $7,404.06 and .. eh heh .. oh my .. and 12 Klingons. What a dreadful day this has been. Oh, and in case you were considering transfering to another institution .. eh heh .. Bank of America's journals are infested with Romulans and Chase are having a devil of a time with Ferengi. Oh, I do so much need some aspirin. Oh my."

Re:So... what?

[JoeMerchant]

What they are after is minimal lag in computation - at some point, the FPGAs will win this over standard CPUs or GPUs, all you need is cubic dollars to fund them, and, if you haven't noticed, cubic dollars is their business.

Re:So... what?

[SteeldrivingJon]

GPUs may not be appropriate for the kind of processing that is required. Off the top of my head, perhaps GPUs aren't ideal for operations on decimal monetary values?

Presumably, they would have gone with GPUs if it was more advantageous. If they're spending more, on FPGAs, then chances are it's because FPGAs are better for the task.

Debug?

[Anonymous Coward]

Debugging a program on clusters is bad enough, but how do you debug your FPGA program?

I guess you can run it 120 more times in 8 hours but can you printf?

Re:Debug?

[The Finn]

you could build in a serial interface or shared memory interface for printing status or debugging, or debug over JTAG by tapping the signal flow inside the FPGA with signal tap (altera) or chipscope (xilinx).

Re:Debug?

[fgodfrey]

Typically, you write a testbench that can, in fact, printf() (sort of). Basically, you end up running a timing-level simulation of the FPGA, or sub-blocks thereof. You're really not developing a piece of software, you're developing a small ASIC. In any event, after you run timing simulations through your testbench where you put in known inputs and verify that you get the expected outputs, you're ready for anywhere from a few minutes to a few days (depending on the size of the FPGA) of compilation to get your code turned into a bitstream to program the device. Then you run the same inputs from the testbench and see if you get the same output. At this point, you hope that you remembered to build various debugging registers into your design so you can have a prayer of finding problems. You can also stop the clock and scan out the value of every register on the chip through something called JTAG. You can then import that back into your simulator to try to figure out what has gone wrong. Then restart at the "few days of compilation" stage.

This is why so few people do reconfigurable computing....

Re:Debug?

[Alioth]

You can run chipscope (Xilinx) to see what's going on inside. You can also run simulations before you instantiate your hardware in the FPGA. In a simulation, you can indeed do the equivalent of "printf".

More Speeds Please

[fsckmnky]

equivalent to more than 12,000 conventional x86 cores, providing 128 Teraflops of performance.

128 Tflops / 5 Tflops = 25.6 AMD dual gpu video cards.

25.6 x $700+/- = $17,920

I wonder what they paid for it.

Re:More Speeds Please

[zill]

GPU ALU have a limited set of operations. Their calculation probably uses operations outside of that set.

Re:More Speeds Please

[Haven]

ALU * clock is meaningless measure. Hardware scales easily, code does not.

Re:More Speeds Please

[ackthpt]

GPU ALU have a limited set of operations. Their calculation probably uses operations outside of that set.

That's OK, because with the volume of dollars now on the move it's no longer GAAP, but quantum.

Re:More Speeds Please

[bas_vijfwinkel]

>>I wonder what they paid for it.
That is just like saying 'A Ferrari 360 consists 1100Kg of metal so I wonder what they paid for that 1000Kg of metal'.
FPGAs and GPUs can't be compared in that way.
FPGAs are slower in flops performance but that can do everything in a massive parallel setup if that is needed for the data processing. If a certain processing step can be done with 2000 independent multiplications then you can just create such logic in FPGAs and perform that step in just a few clock cycles while the GPU must revert back to ordinary loops because no GPU has that many ALU's.
And don't have too much admiration of financial companies and their contractors.
They are generally not into this stuff. A lot of senior developers have tons of certificates but can't handle new technologies because they are so caught up in their fixed paradigms. One of the so-called senior architect engineers at a financial software company told that they turned of a brand new GPU cluster because no-one could get it to work.... Well, I guess that was because there was no Java/Microsoft certificate for GPU supercomputing yet...

Re:More Speeds Please

[fsckmnky]

I work with this stuff on a daily basis.

Re:More Speeds Please

[fsckmnky]

ALU * clock is meaningless measure.

Nowhere did I mention clock speeds. I mentioned TFlops. Any TFlop numbers from the manufacturers, are most likely, best case theoretical hardware performance numbers, based on ALU * clock.

As for whether they are meaningful or meaningless, that depends on your point of view.

If you are a kid playing video games that only cares about frame rates, then perhaps it is meaningless to you.

If you are a developer looking at hardware, who can't yet benchmark the actual application, then it is as meaningful a measure as can be estimated.

I don't know any competent developer or systems integrator who thinks they will get 100% of the manufacturers advertised performance out of anything once they execute their own software. Please feel free to downgrade the manufacturers stated performance to suit your personal situation.

Re:More Speeds Please

[Anonymous Coward]

lol

Re:More Speeds Please

[beltsbear]

Hmmm. My bitcoin cluster has 19 gpu's so far. I am 2/5's of the way there!

Re:More Speeds Please

[fsckmnky]

Schweet.

Re:More Speeds Please

[Anonymous Coward]

The WSJ article listed in the summary states

Not that J.P. Morgan is limiting its experiments to FPGAs. The bank has also been experimenting with the use of chips called GPUs

However it is likely that since Xilinx can connect to QPI (153.6 Gbit/s) that the data bandwidth is higher than the PCIe 2.1 (64 Gbit/s) of an AMD HD6990 or AMD Firestream. The competition on the AMD side would be FPGAs that can connect to Hypertransport, but a quick googling lists and HTX3 Altera card that caps out around HTX3 8 lanes (~83Gbit/s).

Re:More Speeds Please

[Anonymous Coward]

You say in another post that you work with this stuff on a daily basis. Therefore you must realize:

1. You can't divide actual performance of A with theoretical peak performance of B and get a meaningful figure
2. A GPU can only perform certain types of computation at close to its peak performance; for other computations the GPU may perform extremely poorly; FPGA designs on the other hand can be optimized directly to the algorithm.
3. Latency
4. Locality of reference

I suspect that your implied "stupid JP Morgan spending all this money when you can do the same thing on GPUs" is a little naive. JP Morgan have experts too.

Re:More Speeds Please

[fsckmnky]

I suspect that your implied "stupid JP Morgan spending all this money when you can do the same thing on GPUs" is a little naive. JP Morgan have experts too.

I am sorry that you have incorrectly inferred anything I may have implied.

My "back of the napkin calculations" were just that, calculations. They speak for themselves. Anything you choose to add to them, is your own creation.

Have a great day.

Re:More Speeds Please

[SteeldrivingJon]

Do you really want to use floating point to work on decimal values?

Re:More Speeds Please

[fsckmnky]

I'd prefer continuous calculation via an analog based computer myself, but I'm not certain I could afford the water bill.

http://www.youtube.com/watch?v=rVOhYROKeu4

FPGAs as coprocessors?

[Targen]

This story got me thinking that many of the tasks routinely executed on personal computers (perhaps cryptography, video decoding, and such) may benefit from including a FPGA in PCs to serve as a programmable coprocessor. Much like graphics-intensive software can come with shader code to offload processing to the GPU, couldn't a video codec or an implementation of SSL or whatever come with code that would allow an FPGA to do part of the work?

I googled around and found that at least CERN has done something of the sort, but that was over seven years ago. There was a story on Slashdot about something of this sort, but it's even older than the CERN publication. Is anyone working on this sort of idea? If not, why? Is it simply a matter of cost, or is there some other issue that makes this impractical?

Maybe I just suck at googling...

Re:FPGAs as coprocessors?

[Samantha Wright]

I think you probably do [suck at Googling]. I'm almost certain there was a story about this exact thing within the past year or two. At the very least, the Google search "FPGA co-processing" returns a lot more results than "FPGA co-processor".

Re:FPGAs as coprocessors?

[fsckmnky]

There are 2 companies that I am aware of, who's name temporarily escapes me ( perhaps I'll search and find them ) who are using FPGAs with hardwired hypertransport interfaces that plug directly into an Opteron mobo socket. Some of Crays recent models use the same approach.

This gives the co-processor direct access to the entire system bus, memory, and the Opteron CPU installed in the primary socket.

Last I read they were about $5k each. Perhaps the price has come down since.

Re:FPGAs as coprocessors?

[fsckmnky]

DRC is one. http://www.drccomputer.com/index.html

Re:FPGAs as coprocessors?

[PaulBu]

I think that the problem with FPGAs is how many times can they be reprogrammed. For this (banking) application probably not that many, but if you have to re-program your in-PC FPGA every time you decide to check your encrypted e-mail while watching streaming video, you will run out of programming cycles pretty quickly.

Besides, if task is well-defined (as in, has a standard, like MPEG, or SSL), if it probably more cost-effective to have a special chip doing it (you can definitely get cheap stand-alone chips for tasks like this), and just add it in the corner of motherboard.

Paul B.

Re:FPGAs as coprocessors?

[zill]

It's been envisioned, implemented, and commerialized a long time ago. You can buy a duel socket AMD board, and put in an AMD CPU in one socket and a FPGA co-processor in the other socket. That was 5 years ago.

Quite a few supercomputers on top500 have the above mentioned configuration. JPMorgan is very late to the party.

You're probably wondering why every desktop, laptop, and smartphone doesn't come with this wonderful technology already, and there are many many reasons for that:
- FPGA programming is difficult, and it's a much rarer talent than software programming
- The FPGA industry is currently a duopoly and combined that with the small market of FPGAs means that the price is too high for consumer electronics
- Specialized functionality can always be more cheaply implemented in ASICs (cryptographic co-processors, new instructions in CPUs, H264 decoding ASICs)
- The chicken and egg problem. Developers won't start hiring FPGA programmers en masse until there are enough machines out there with FPGA co-processors installed. And people won't start buying FPGA co-processors until their favorite program supported co-processor acceleration.

Re:FPGAs as coprocessors?

[bas_vijfwinkel]

They already exist but they are pretty expensive. A single Intel's XD2000 inSocket FPGA accelerator costs more than a really high end user workstation. I worked with one of such devices for a while and it is amazing what you can do (besides bitcoin mining *laugh*). As long as what you want to calculate or process can be done in parallel, the speed ups are phenomenal. With devices like the XD2000, the limit was the size of the FPGA but with the next generation FPGA's from Xilinx or Altera, we will definitely see a big leap forward in certain areas. When you can let go of the old paradigms then the sky is the limit. More interesting however are hybrid devices like the new Zynq which is basically a (dualcore) CPU plus FPGA. In this case you never have to leave the silicon to use customizable logic. Some mainframes back 30 years ago already had reprogrammable CPU's where custom logic could be added but not to the extend of these device.

Re:FPGAs as coprocessors?

[Anonymous Coward]

> Quite a few supercomputers on top500 have the above mentioned configuration. JPMorgan is very late to the party.

Can you name one?

Re:FPGAs as coprocessors?

[Anonymous Coward]

Actually, almost all FPGAs are SRAM-based and reconfigure (program) themselves every time they power on. The biggest problem is the compiler tools themselves (the Xilinx tools are pretty horrid all things considered).

Will be interesting to see how long it takes them to move to GPUs. They aren't as flexible but the raw amount of computing power is much cheaper. Then again, we're talking about banks, heh

Re:FPGAs as coprocessors?

[zill]

My bad. I must've confused FPGA with GPGPU. My apologies.

JPMorgan is indeed one of the first places to mass deploy clusters with FPGA co-processors. The technology isn't new per se, but they are still pioneers in terms of scale.

Re:FPGAs as coprocessors?

[Anonymous Coward]

> You can buy a duel socket AMD board,...

I've been buying motherboards for more than twenty years to build systems, and I've never seen a model with sockets designed to fight each other. Can you provide a link to one? A search on Newegg for "duel socket" returns zero results.

Re:FPGAs as coprocessors?

[petermgreen]

In electronics a large part of the cost is upfront so niche products are FAR more expensive than common products of similar complexity. This applies to some extent to assemblies (compare the cost of a PXIe controller module to the cost of a laptop for example) and applies to an even greater extent to semiconductor products.

Powerful FPGAs are a niche product and hence expensive. GPUs are a mass market product and hence cheap. So if a GPU can do your calculations reasonablly efficiently it's probablly the cheaper option.

Re:FPGAs as coprocessors?

[JoeMerchant]

We use FPGAs to do custom video pipelines alongside our processor cores, we don't, as a general rule, reconfigure them very often, because that takes lots and lots of testing to make sure you haven't screwed up, but the flexibility does allow us to move forward without fully specifying the hardware at an early stage.

Re:FPGAs as coprocessors?

[wirehead_rick]

http://www.xtremedata.com/technology/parallel-architecture

Re:FPGAs as coprocessors?

[wirehead_rick]

The move to GPU's won't happen in HPC circles. However the move of OpenCL to FPGAs is happening:

http://www.altera.com/corporate/news_room/releases/2011/products/nr-opencl.html

Re:FPGAs as coprocessors?

[Anonymous Coward]

If you don't mind using PCI Express, there are plenty of off-the-shelf FPGA development boards that will slot into a PC.

At a very high level, tools like Xilinx System Generator allow you to develop DSP systems using Matlab and Simulink, which can be migrated/offloaded to the FPGA when the workload is suitably large or parallel enough that it would be faster than the CPU.

Re:FPGAs as coprocessors?

[Anonymous Coward]

Is this a joke? The move to GPU's in HPC circles has already happened. GPU-related presentations are commonplace at HPC conferences (both academic and commercial), GPU-related publications are commonplace in HPC and scientific computing journals, anyone building an HPC cluster is going to at least give serious consideration to including GPUs.

Re:FPGAs as coprocessors?

[cachimaster]

Much like graphics-intensive software can come with shader code to offload processing to the GPU, couldn't a video codec or an implementation of SSL or whatever come with code that would allow an FPGA to do part of the work?

Yes but decent-sized FPGAs are crazy expensive.

It is false that FPGAs consume less power, they are power hogs and are slow (max 200-300 mhz). Neither are there because of their reconfiguration powers.
They are used because of the bandwidth. Specifically, memory bandwidth. CPUs are just big state-machines that can read from DRAM 64 bits at the time. Not many algorithms are optimized for that. But you can optimally implement every algorithm on a FPGA, and gains are usually tremendous, even with a slow, hot and expensive chip.

Re:FPGAs as coprocessors?

[Laser Dan]

This story got me thinking that many of the tasks routinely executed on personal computers (perhaps cryptography, video decoding, and such) may benefit from including a FPGA in PCs to serve as a programmable coprocessor. Much like graphics-intensive software can come with shader code to offload processing to the GPU, couldn't a video codec or an implementation of SSL or whatever come with code that would allow an FPGA to do part of the work?

Xilinx (one of the two big FPGA companies) very recently released their "Zynq" family of combined CPU+FPGAs. They contain a dual core ARM Cortex-A9 running at 800 MHz and a pretty decent amount of 28nm FPGA logic, with interconnects between the two. That is basically what you are describing I think.
http://www.xilinx.com/products/silicon-devices/epp/zynq-7000/index.htm

Re:FPGAs as coprocessors?

[The Finn]

Intel has supported socket-connected FPGAs for years now. A few vendors (including xtremedata and nallatech) offer(ed) FSB-attached FPGAs. Pactron (with Altera FPGAs) and Xilinx are offering QPI-attached FPGAs on the Nehalem/Westmere -EX platform and have announced support for Sandy Bridge.

I work for Intel on these technologies.

Re:FPGAs as coprocessors?

[YoopDaDum]

- Specialized functionality can always be more cheaply implemented in ASICs (cryptographic co-processors, new instructions in CPUs, H264 decoding ASICs)

Yes, but only for high volume use-cases where the cost of doing an ASIC (millions of $) is not an issue as it's spread over a large number of chips, and when you don't need the programability an FPGA offer as the specialized function is very well defined and stable. But for small scale applications FPGA rules from a cost point of view. Volume will tell you whether you need to go ASIC or FPGA, assuming other considerations like absolute single chip performance (better with an ASIC), power consumption (lower with an ASIC) or programability (FPGA only) are not factors in the choice. In any case, there's room for both.

But in the case discussed here it's both low volume and flexibility is required, so FPGA make a lot of sense. And even if people able to program FPGAs are harder to find than software developers, well JPM has the financial resources to attract them.

Re:FPGAs as coprocessors?

[zill]

why every desktop, laptop, and smartphone doesn't come with this wonderful technology already

I was talking about consumer products.

Re:FPGAs as coprocessors?

[Anonymous Coward]

There is an intel atom processor with integrated altera FPGA available now:

http://www.altera.com/devices/processor/intel/e6xx/proc-e6x5c.html

Re:FPGAs as coprocessors?

[Anonymous Coward]

VHDL is not hard, it is very like ADA for a reason.

A lot of it also uses pre-written libraries glued together.

Re:FPGAs as coprocessors?

[Anonymous Coward]

FPGA's are heavily used in physics, when it comes to read out electronics and triggering (as in making sure you don't get overlapping events, not to mention syncronizing 10 000 metric tonnes of detector in 25 nano seconds....). There are plenty of physicists that do FPGA design!
CERN has even given industry awards to FGPA manufacturers.

Kudos to the JPMC engineers!

[msobkow]

I spent two great years working for J. P. Morgan Chase, starting in 1999, followed by a year with the merged JPMC, so I have some knowledge of how this new system will be used and how it fits into the business process of running a bank. I can't discuss details about that, but I just wanted to share my congratulations with the JPMC team for tackling that thorny issue.

You have to understand that investments can't be made until those risk analyses are done, so cutting 7-8 hours off the run time will earn the company millions over the course of a year. We're talking about the kind of investment loans where even a 4-5 hour overnight "float" of capital to help someone seal a bigger deal can be worth a significant amount of interest and profit.

Remember: the big investment banks are dealing with numbers that cause spreadsheets to overflow. You can't even visualize the data with standard desktop tools. You wouldn't believe the totals I saw come out of some reports, and I wish I could forget them. Such numbers are not meant for the grasp of mere humans living on a working wage.

Re:Kudos to the JPMC engineers!

[Turbine2k5]

You wouldn't believe the totals I saw come out of some reports, and I wish I could forget them. Such numbers are not meant for the grasp of mere humans living on a working wage.

...so what you're saying is that the numbers you saw are the CEO's salary?

Re:Kudos to the JPMC engineers!

[ackthpt]

I spent two great years working for J. P. Morgan Chase, starting in 1999, followed by a year with the merged JPMC, so I have some knowledge of how this new system will be used and how it fits into the business process of running a bank. I can't discuss details about that, but I just wanted to share my congratulations with the JPMC team for tackling that thorny issue.

You have to understand that investments can't be made until those risk analyses are done, so cutting 7-8 hours off the run time will earn the company millions over the course of a year. We're talking about the kind of investment loans where even a 4-5 hour overnight "float" of capital to help someone seal a bigger deal can be worth a significant amount of interest and profit.

Remember: the big investment banks are dealing with numbers that cause spreadsheets to overflow. You can't even visualize the data with standard desktop tools. You wouldn't believe the totals I saw come out of some reports, and I wish I could forget them. Such numbers are not meant for the grasp of mere humans living on a working wage.

And just think, replying to that programming school on the matchbook cover and adding a little clause of code to catch the fractions beyond cents could have you living on Long Island in a pleasant estate in about half a second of activity. -- Superman movie reference.

Re:Kudos to the JPMC engineers!

[zill]

Remember: the big investment banks are dealing with numbers that cause spreadsheets to overflow.

Wow! They're using more than 65536 rows? Impressive!

Re:Kudos to the JPMC engineers!

[confused one]

the big investment banks are dealing with numbers that cause spreadsheets to overflow

so last century. Excel 2007 and 2010 have no such limitations. which is a good thing because I sometimes generate datasets too big for the older versions to handle.

Re:Kudos to the JPMC engineers!

[Anonymous Coward]

Or is it: "such numbers are not meant to exist in the control of a single bank?"

Re:Kudos to the JPMC engineers!

[seven of five]

Such numbers are not meant for the grasp of mere humans living on a working wage.

Screw you, your uppity attitude, and the horse you rode in on.

Re:Kudos to the JPMC engineers!

[EdgeCreeper]

1048576 rows now.

Re:Kudos to the JPMC engineers!

[Anonymous Coward]

They max at a little over 1 million rows now.

Re:Kudos to the JPMC engineers!

[msobkow]

Buddy, if you can look at numbers that exceed trillions and claim they honestly seem like reasonable and rational values compared to a sub-six-figure income, you're either a mathematical genius or a liar.

As to the idiot who claimed I was "one of those responsible", is there anyone who actually thinks the peon hordes who work at banks or big corporations are collecting the fat-cat bonuses? We took pay cuts in 2000 after the big push for Y2K was over; we did not get bonuses for helping the bank ride out the paranoia.

And the systems programmers are not responsible for designing risk analysis algorithms, only implementing them.

I'm proud of the work I did for JPMC. We delivered a critical system in record time after the project was put at risk by two years of fiddling around accomplishing nothing. We improved it's reliability and speed over it's predecessor. And we did get bonuses for that one, just not for the Y2K work.

We deployed a new version of a third-party product, and came up with migration scripts to get the data over from the old mainframe version. The vendor had contracted to do the same with Citibank's copy of the software, spent two years of vendor consulting time and a 20 person team, and failed.

Our competent team of four not only did the job the product vendor themselves could not, but we automated the process, and we did so in under a year. It just goes to show what happens when you get four good programmers working on something instead of a horde of average incompetents paid for by a virtually unlimited budget.

So, was I "one of those responsible"? Not in the sense of making decisions, but in providing successful deployments of software supporting the decisions of management. I feel no shame for being good at my job, and rather pissed off that management obviously ignored what was reported by the systems my cohorts worked on.

Re:Kudos to the JPMC engineers!

[seven of five]

I wrote that comment too hastily. I find it hard to be cheery when a bank does yet more esoteric, inscrutable things to make a buck.

BTW, the GDP of the entire world is 60 trillion, so numbers "that exceed trillions" do not seem based in reality, unless you're using the world's assets leveraged 1000 to 1 to buy out Rigel 7.

Re:Kudos to the JPMC engineers!

[msobkow]

Poor phrasing on my part. The big numbers don't actually leap into the next level beyond trillions. But they still are mind boggling. They're the kind of numbers that get hockey stick business planners thinking "if I can only get a fraction of a percentage point of that..."

Re:Kudos to the JPMC engineers!

[Anonymous Coward]

If your "investment" really depends on the events of a single 8 hour work day then you're not investing, you're gambling, plain and simple. Dress it up all you like but the same investment industry that tells us to "invest for the long term" does no such thing once we actually let them use our money and this is just further indication that they have no intention to ever do so.

PDF on JP Morgan's Website about this stuff

[Anonymous Coward]

Talks about this stuff and some low latency stuff too...

http://techcareers.jpmorgan.com/downloads/Field%20Programme%20Gate%20Array%20factsheet.pdf

Re:PDF on JP Morgan's Website about this stuff

[mako1138]

Migrating algorithms from C++ to FPGAs involves doing a Fourier Transform from time domain
execution to spatial domain execution in order to maximize computational throughput.

What the hell is this supposed to mean? The document seems like the standard FPGA boilerplate otherwise, great for showing to management.

Re:PDF on JP Morgan's Website about this stuff

[Laser Dan]

Translation (probably):
"We make things run in parallel".

Doesn't sound so impressive though does it!

Let the looting begin!

[Maltheus]

Now JP Morgan can raid future MF Globals all that much faster, while hiding their shenanigans at the COMEX.

Accelerated Economic Evolution

[Anonymous Coward]

Is it just me or do others see things like this as dangerous just from an evolutionary perspective? If I recall, a species that accelerated it's evolution is doomed to burn itself out.

If the bankers had a clue...

[Anonymous Coward]

this would be scary(er)

Re:If the bankers had a clue...

[cashman73]

If bankers had a clue they would use these supercomputers to generate bitcoins!

Be First

[Erastus]

In the movie, Margin Call (2011), the investment bank that saw the danger of their risk exposure first was prompted to sell and start a massive panic that sunk a competitor. The movie is loosely based on the Lehman crisis. "John Tuld: There are three ways to make a living in this business: be first, be smarter, or cheat. "

Re:Be First

[fsckmnky]

Thanks for the IMDB summary.

I know fpga programming

[Ryanrule]

Sounds like I should get a job in the financial industry. /subscribes to "Giant Boats"

They need "FLASH GORDON" Technology... apk

[Anonymous Coward]

For I/O intensive work & as far as "super-computing" goes, it has tremendous IOPS possibles...

* RamDisks/RamDrives, RULE!

They absolutely rule where I-O's to files matter, & especially for that area (& that system below's got a lot of that going for it - databases, websites, terminal servers, & FAR more gain hugely using them (what's below's just another example thereof))...

APK

P.S.=> Think I'm kidding? These folks @ the San Diego SuperComputing Center aren't, quoted from this below, verbatim:

"SDSC thinks the 36 million IOPS number makes it the fastest supercomputer in the world in terms of I/O operations." FROM -> http://www.theregister.co.uk/2011/12/15/sdsc_gordon/

... apk

Hi gang

[Anonymous Coward]

The malcontent WoW players haunting Slashdot haven't the first clue what this story is about.

Wow, they spent all that money...

[Anonymous Coward]

...and they could've just built a beowulf cluster.

Scary

[Metricmouse]

That's 38 more teraflops than Skynet, and I thought I was afraid of big banks before...

Re:Scary

[Anonymous Coward]

Yeah, while this is true, they'll be running Windows 9 beta which will bring it to XP levels...

Meanwhile America is dying

[Anonymous Coward]

Fuck JP Morgan. Rape and kill their families.

FPGAs aren't power efficient nor cost-efficient

[Anonymous Coward]

They are generally greatly overrated. For most tasks a CPU would beat them, a GPU too.

We use them at work, for $200,000 in FPGAs we can do what a $10 integrated circuit chip can do, only at 1/400th the speed and about 100x the power.

Their only virtual is maximum reprogrammability and they just aren't good enough at that to make sense for many things. Look at CPUs, GPUs and DSPs instead. These can utilize logic blocks in different configurations/sequences far better than FPGAs.

Re:FPGAs aren't power efficient nor cost-efficient

[Anonymous Coward]

They are generally greatly overrated. For most tasks a CPU would beat them, a GPU too.

We use them at work, for $200,000 in FPGAs we can do what a $10 integrated circuit chip can do, only at 1/400th the speed and about 100x the power.

I work on emulating those $10 ASICs with FPGA systems for pre-silicon validation of the ASIC design. This is common practice in the ASIC design business since turning on a FPGA emulation platform for a few minutes runs more clock cycles than you'll probably ever be able to simulate. (it doesn't replace simulation, sim is better at some things, FPGA emulation is better at others.)

Your ratios are way off the mark. Yes, FPGA logic is slower and uses more power than ASIC, but I know for sure that it isn't 400x slower. It's generally not difficult to get ASIC RTL logic to close timing at 1/10th the ASIC clock rate in FPGA, and sometimes much better ratios than that are possible.

The only way you end up at 400x slower is if your ASIC design is too large to fit in a single FPGA (which is usually the case, since FPGA logic is much less area efficient than ASIC), and, when you partition the ASIC logic across multiple FPGAs, you end up with too many nets between FPGAs. The largest FPGAs tend to have around 1000 user IO pins, so if you need more, you have to multiplex them. This in turn artificially restricts the core clock rate to some fraction of the maximum serial bit rate you can run a FPGA-to-FPGA connection at.

Even so, I've managed better than 1/10th clock rates despite 8:1 pin multiplexing. 1/400th does not reflect the capabilities of the FPGA logic fabric. If you really can't do better, there's something else going on.

Same comment re: power, though at least there you aren't quite so wildly off the mark.

Their only virtual is maximum reprogrammability and they just aren't good enough at that to make sense for many things. Look at CPUs, GPUs and DSPs instead. These can utilize logic blocks in different configurations/sequences far better than FPGAs.

This also makes me think you don't have as much of a clue as you think you do. Modern FPGAs have hundreds or thousands of DSP blocks which can run at several hundred MHz, and you can easily construct a flexible internal datapath linking them together using the FPGA's logic fabric.

1999, before the first Synthetic CDO was sold?

[decora]

before credit default swaps grew to dozens-of-trillions of dollars business?

before Commodity Index Funds?

before the stock exchanges and commodities exchanges got rid of their open outcry trading pits and went electronic?

before JP Morgan bought Bear Stearns (with a Fed loan)?

before JP Morgan got bailed out by taxpayers?

no offense dude... but what on earth does 1999 have to do with 2011?

So JP Morgan can calculate VaR faster. so what? they didnt calculate it fast enough in 2008 - they took that big fat payout like everyone else, they were up to the arms in CDOs and subprime like everyone else, they had their fingers in every pie. They arent "earning" anything, they are "stealing" it from the taxpayer, with an enforced monopoly based on corruption in the federal government. JP Morgan should not exist.

I do not buy your thing about spreadsheets, considering how CDOs were often calculated by people called 'F9 monkeys'. and one CDO deal could be worth a billion dollars.

Re:1999, before the first Synthetic CDO was sold?

[msobkow]

*sigh*

The problems the risk analysis team faced even in the 2000 era was such a tough nut to crack that they had to limit the complexity of the algorithms they used just because there wasn't hardware powerful enough.

All the things you mentioned have only added to that complexity, making the calculations that much worse and that much more expensive.

So instead of making me change my mind, you just made me realize how much more impressive their achievement was than I first thought.

Spreadsheets back then did not have arbitrary precision decimal or integer values -- they used floating point. I have no idea whether newer spreadsheets have shifted to arbitrary precision values or not, but if they haven't the spreadsheets blow up.

Remember: little companies like GE, GM, and Exxon are the kind of customers who have deposits with an investment bank. As a result, numbers like total deposits held blow floating point values out of the water by a significant margin. The amount of money floating around the world really does generate some stunning sequences of digits, they're almost magically long numbers like the nth digit of Pi. They just don't register as "billions" or "trillions" automatically, you have to count the digits and think a moment about what that number is supposed to be called. :D

Re:1999, before the first Synthetic CDO was sold?

[Viol8]

The complexity of these algorithms and products is what got us into this mess int the first place FFS! The last thing we need is banks doing even more complex calculations that even fewer people understand.

Re:1999, before the first Synthetic CDO was sold?

[ShakaUVM]

>>The problems the risk analysis team faced even in the 2000 era was such a tough nut to crack that they had to limit the complexity of the algorithms they used just because there wasn't hardware powerful enough.

There was also a very fundamental problem with the assumptions made, such as mortgage defaults are independent.

Because, after all, if one person defaults on their mortgage it has no effect on another person defaulting, right? :p

Re:1999, before the first Synthetic CDO was sold?

[Rich0]

The problems the risk analysis team faced even in the 2000 era was such a tough nut to crack that they had to limit the complexity of the algorithms they used just because there wasn't hardware powerful enough.

Look, the problem here is the black swan. You can't model a black swan unless you can simulate the entire world economy down to the last neuron in some farmer's brain in a rural Chinese village. Right now we can't model a single human brain let alone all of them.

The world economy didn't melt down because some spreadsheet only calculated 12 decimal places when it should have calculated 325. It melted down because everybody decided to leverage themselves 100x on the bet that housing prices wouldn't ever go down, and they did. Now the world governments are starting to leverage themselves in small multiples on the bet that nobody would ever stop buying their bonds, mostly to bail out the bankers who bet on housing prices. I don't need arbitrary precision arithmetic to tell you where that is going to end up if it doesn't change FAST.

Re:1999, before the first Synthetic CDO was sold?

[msobkow]

Exactly. It's management and the risk analysis team that decides what gets considered by the algorithms, not the programmers who have to implement the algorithms and systems.

Do you think a properly implemented set of risk analysis points would have allowed loans to sub-prime mortgage holders in the first place? It wasn't risk analysis that did that; it would have been overriding management-imposed company policy. It didn't take a genius to realize that people who couldn't come up with their downpayments in the first place probably wouldn't be able to afford the interest increases after the initial term.

The people who signed those mortgages should have thought about whether they could afford market rate mortgages, not whether they could afford the "low, low introductory price" of a few year's subprime mortgage to start. Sure the banks raped them by bumping interest rates, but only a complete moron would have expected the banks to continue to leverage loans at a loss!

Dirty pool by the banks? Yes. But no more so than when anyone else uses a "loss leader" product to "buy" the high-value add-on sales that follow.

Re:1999, before the first Synthetic CDO was sold?

[msobkow]

I'm not speaking to JPMC specifically because I had no contact with the Citibank branch of the merger that had the consumer loans and mortgages, but to banking in general. It's not the risk analysis engines that produced the bad loans, but the individual loan managers who had the option of ignoring the risk reports and apparently did so.

Re:1999, before the first Synthetic CDO was sold?

[Rich0]

only a complete moron would have expected the banks to continue to leverage loans at a loss!

News flash - the world is full of complete morons. However, they still need someplace to live. By all means restrict their reproduction privileges, but unless we want them starving on the streets at some point we need to look out for them...

No matter how powerful this FPGA super computer is

[Anonymous Coward]

It is just garbage in, garbage out.

first as a panelist on the Gong Show...

[Anonymous Coward]

...now putting out high tech product! Is there NOTHING this woman can't do?

Re:Best use for this computer

[A12m0v]

It is a FPGA supercomputer, don't you need to reconfigure it first?

Bitcoin Miner!

[madhi19]

Let face it JP is doing some Bitcoin Mining on the side with this hardware! loll Off course that would be less of a scam that what they really do with it!

A relevant opensource financial project is Bitcoin

[Anonymous Coward]

Bitcoin community is using GPUs and FPGAs for mining bitcons (mining is a process of finding a number satisfying a criteria, in a huge sample space
) The main algorithm in a nutshell is SHA256 done twice, and checking if the output of the hash produces the given number of zeros.

The community uses primarily AMD GPUs and Xilinx Spartan 6 FPGAs.
https://en.bitcoin.it/wiki/Mining_hardware_comparison

The combined processing power of the bitcoin network is 101.15 Petaflops/s according to http://bitcoinwatch.com/

Garbage In, Garbage Out

[slamb]

What a waste. These banks can build the most impressive hardware in the world, perform calculations in the petaflops, and still have absolutely no clue what risk is involved in the business they're doing because their assumptions and data are all wrong. If by some freak accident they were to get the right answer, they would conceal it from their clients and investors anyway because their incentive is to take big risks - they get enormous rewards if they are right and lose little if they are wrong. They are incompetent and amoral, which is simply not a technical problem.

That and Washington is too cozy with Wall Street

[Shivetya]

and they do their damn best to keep it quiet. Look at how quickly stories about Congress and their staffs using insider information was quashed. It went from being front page to gone in days, as if it didn't exist. Similar to how Fast and Furious vanished.

We need an OKS, Occupy K-Street, Wall Street is fully enabled by Washington. They just pay their dues and Washington insiders reap the real rewards.

Let's be clear about what they're getting

[WOOFYGOOFY]

What they're NOT getting is the ability to more closely model the reality of the world. I say this because it appears from the article that they think they are (from the article)

There (is) new evidence of a payoff for certain kinds of customers from J.P. Morgan, which has been working since 2008 to adopt Maxelerâ(TM)s hardware and software to help assess its trading risks. The companies say the approach has allowed the financial-services company to quickly examine tens of thousands of possible scenarios for how its investments might be affected by events in financial markets, reducing the time for running certain scenarios from hours to a few minutes....quickly examine tens of thousands of possible scenarios for how its investments might be affected by events in financial markets,

Yeah sure, they examined them, chop chop, and what did this tell them? Nothing, unless all this refers to is guessing what other hedge funds will do in response to each other in a time frame that is not longer than a few milliseconds. .

Beyond that narrow (and getting narrower all the time) interpretation, there is no "prediction" about the market to be had, since the market is determined by what happens in the real world which is pathologically and permanently unpredictable.

Oskar Mencer, Maxelerâ(TM)s chief executive and founder, says the technology can prove its worth to J.P. Morgan in situations like handicapping whether debtor countries might default on obligationsâ"allowing the bank to make hedging transactions to reduce the financial impact. âoeThey can ran different default scenarios two days ahead,â Mencer says. âoeThey can hedge against those scenarios very effectively.â

So you get what they're trying to do, right? They're trying to use historical data - about say, defaults, or just generally about what real world actors do under circumstance X when conditions Y and Z hold, and assign probabilities to them so they can make a decision and "beat the market".

The problems here are manifold.

One is, a perfect record of all the financial events in the world, however you want to define them, is not enough to form a statistical basis for analysis. Why? Because there's no reason to believe that such events are the product of Random (not "random") variables, which is a axiomatic of all statistical analysis.

Two, the events they're concerned with interact with each other and random (not "Random") world events in ways which are so complex they are fundamentally unpredictable. Only systems whose events of interest can be causally isolated from other events, however broadly you have to define events to make that happen, are amenable to statistical analysis.

In the real world, where literally every social and financial event could and does effect all others, you have the impossible task of having to model everything in order to predict anything accurately and consistently, or even on a, you know, regular basis.

The real world of human and financial events is not amenable to analytic predictive methods . Full stop.

This is not to say there won't be winner and losers where winner's predictions come true and losers predictions go bust. This is to say they don't know who they are in advance.

The unfolding of any given course of financial events is a one-time historical event happening forwards in time and is ultimately infinitely sensitive to any other real world event or set of conditions, either knowable or unknowable. There is no basis to form a probability distribution of likely futures because there is no clearly definable event or circumstance- every event and circumstance is one-off and sui genris.

But this s not REALLY what they're trying to achieve, at least, we can hope not.

What they're trying to achieve is to guess how OTHER institutions who understand the world as they themselves do are likely to act, and beat them to it or o

Re:Let's be clear about what they're getting

[WOOFYGOOFY]

There. Fixed that.

I forgot to add. What this seems to be is a huge public expenditure on Voodoo which will then used to separate people from their money, and charge them a fee for doing so.

THAT is how hedge funds make money when considered as a group. They con(vince) people to give them large sums of money. They then charge them for this "serrvice". They then go bust or make bank using their customers' money at exactly the rate and to exactly the degree that chance would predict. Within this framework, there are winners and losers. The winners then heavily advertise (tot he right set of people) their record and proceed to charge even MORE money for their services. The losers no one hears from again.

See The Ascent Of Money by Niall Ferguson for the numbers behind this.

Ditto with banks.

If banks could really predict the market etc etc why would they zealously lobby Congress to defend the outrageous fees they saddle their unremarkable customers with?

Why wouldn't they gratefully accept without provocation any money anyone wanted to give them whatever the amount and quickly invest it in their sure-money machine?

Because they of all entities knows the long term record of that "sure money machine" and by comparison those $20.00 fees are, well, money in the bank.

So they've spent a lot of real money and used the services of serious researchers and technicians to build a... glossy brochure sure to convince the gullible-but-rich people to hand over some substantial part of their fortunes.

For a fee.

can't wait for this to appear on eBay ...

[Lazy Jones]

... when JPMoran is broke due to "fat finger" bugs in the trading software...

Story misses details on Maxeler hardware

[gentryx]

Sadly, both stories lack details on how the FPGAs are used in the computing architecture. Instead the spend great lengths on listing telephone number like, meaningless speedup comparisons with conventional hardware. A typical drawback of FPGAs is that they cannot accommodate as many floating point units (FPUs) per chip as current GPUs and that FPGAs run at about 10x lower clock speeds. Their advantage however, is that the internal chip architecture can be reconfigured to match the algorithm, so that all FPUs run at maximum efficiency. At the end of the day, it really depends on the algorithm, whether it's run best on FPGAs, GPUs or standard CPUs. This is also the reason why one cannot say that an FPGA is X times faster than a GPU: it really depends on the algorithm.

Maxeler, the manufacturer of the machine, had a booth at SC11. The basic component is the MAX3 card, a PCIe 2.0 8x card with up to 96 GB of DRAM on board. The boards are optimized for data stream processing. This is not unlike how GPUs are architectured.

Up to 4 of those boards are located in a MaxNode, which can then be networked via 10Gbit Ethernet or InfiniBand. Multiple MaxNodes can be put into a MaxRack, which can also be seen in the WSJ article. The MAX3 boards can be connected via a custom MaxRing network, which provides a bandwidth of 8 GB/s.

So You're One of Those Responsible For...

[Anonymous Coward]

the current economic malaise?

Nice of you to come out from under that rock and tell us how nifty your former employer's toys were.