PS3 factoring?

[Ken_g6]

Originally Posted by xilman

Lenstra's group has implemented a rather cute way of parallelizing the group multiplication for elliptic curves over small base rings. That algorithm works well on machines with only a small number of parallel processors but doesn't scale as well to architectures with hundreds.

Well, I'm not sure about AMD cards, but nVIDIA cards act like a number of either 32-core (for recent cards) or 8-core (for GTX200-series and older cards) machines. A GTX260, for instance, could act like a cluster of 96 8-core machines (192 cores * 4 cycles/core minimum/8-core blocks). Though their clocks sort of have to be synchronized (all starting and stopping operations when a kernel starts or stops), and the 8 or 32 cores have to perform the same operations together, each group can perform completely different operations from every other group. Given that Lenstra has a cluster of PS3s, I assume a cluster formation like this might be useful to him?

[Christenson]

More evidence that Sony is probably in fortress mode:
http://www.wired.com/threatlevel/2011/06/sony-lulzsec/ (Sony hacked again)

[diep]

Originally Posted by xilman

You really don't get it, do you?

Please re-read my post again. If you pay attention you will see that you are attacking a straw man.

Of course the EPFL group would not have built a PS3 cluster in 2009 to perform the computations reported in a 2010 paper. The cluster was commissioned years earlier to address an entirely different research question. As I wrote, but with emphasis this time:

The cluster is still producing good results not because it is state of the art but because it has been fully paid for, it has achieved all that it was intended to do and because it still works. For the moment, the only cost to use it is the electrical power and a relatively small amount of attention.

Paul

I get it exactly, the average guys still are believing, because of their 2010 publications that the CELL still is a capable chip against todays GPU hardware.

As simple as that.

Nothing else as that. Whatever you write down in some subsection paragraph doesn't matter if you claim you achieve XYZ out of the CELL meanwhile not achieving that out of a GPU.

What happens in the backgrounds people are simply not aware of. For example that most HPC commissions do not believe in long term in the CELL architecture they told me, yet never publicly.

That's how this world works man.

So if you then publish a paper with respect to what you achieved on the CELL, then that's basically the opposite of what they tell you when you meet 'em.

There is a huge difference between the perception of the general public, caused by what press releases you do, and the subsectioned paragraphs that you and i can read.

In this case i publicly write down that the CELL is an outdated processor, and you attack me for that.

Meanwhile you show 0 evidence of the opposite, namely that "lenstra knows in private that gpu's are very fast".

His paper doesn't say that. The paper says he achieved XYZ on the CELL and nothing on a gpu.

Get the picture do you?

What you say in public matters and the rest doesn't.

Regards,
Vincent

p.s. and in this case the publication is what we call in Dutch: "broodje aap verhaal", if i try to translate that,
it would be something like: "a desinformation story".

[ldesnogu]

Vincent, I'm afraid you are the one who doesn't get it. They don't claim the Cell is the best processor; they had some computations to do, they had a PS3 cluster at disposal, they developed some specific algorithms and programs for that cluster and they published an article presenting what they did.

[jasonp]

Maybe diep should come to the workshop in Amsterdam this December and complain to them personally.

[xilman]

Originally Posted by jasonp

Maybe diep should come to the workshop in Amsterdam this December and complain to them personally.

A suggestion I whole-heartedly recommend!

Paul

[xilman]

Originally Posted by diep

Meanwhile you show 0 evidence of the opposite, namely that "lenstra knows in private that gpu's are very fast".

I could provide evidence but it is contained within private email from him and I don't release such material without permission.

Why don't you mail him and ask his opinion directly? His email address is easy enough to find. Ask me by email if you really can't find it for yourself.

Paul

[Christenson]

Originally Posted by xilman

I could provide evidence but it is contained within private email from him and I don't release such material without permission.

Why don't you mail him and ask his opinion directly? His email address is easy enough to find. Ask me by email if you really can't find it for yourself.

Paul

You know, Mr Lenstra put together a PS/2 cluster...and had lots of friends warn him when Sony tried to patch him out of it...and he's in the literature, and going to meet half the people on this forum that contribute math in December in person...the conclusion that he's smart, is very connected, and very aware of GPUs follows immediately, though RDS would remind me this isn't quite a proof. It would be quite reasonable that he's looking at CUDA and OpenCL and six other tools I can't name and programming and profiling for his own purposes as we speak! (Watch out, or we'll all be running his code soon on our GPUs....)

Me, I think a third silicon player (IBM) in the GPU space would be a good thing....but I'm not holding my breath.

[Christenson]

On the continuing Sony story:
http://arstechnica.com/gaming/news/2...m-hardware.ars

Interesting explanation for WHY the PS3 didn't take over the world.