News
« Today's the day! | Main | One week countdown »
Thursday
May102012

News of the month

Welcome to our more-or-less regular new post. Seen something awesome? Get in touch!

Convention time!

Next week is Canada's annual petroleum geoscience party, the CSPGCSEGCWLS GeoConvention. Thousands of applied geoscientists will descend on Calgary's downtown Telus Convention Centre to hear about the latest science and technology in the oilfield, and catch up with old friends. We're sad to be missing out this year — we hope someone out there will be blogging!

GeoConvention highlights

There are more than fifty technical sessions at the conference this year. For what it's worth, these are the presentations we'd be sitting in the front row for if we were going:

Now run to the train and get to the ERCB Core Research Centre for...

Guided fault interpretation

We've seen automated fault interpretation before, and now Transform have an offering too. A strongly tech-focused company, they have a decent shot at making it work in ordinary seismic data — the demo shows a textbook example:

GPU processing on the desktop

On Monday Paradigm announced their adoption of NVIDIA's Maximus technology into their desktop applications. Getting all gooey over graphics cards seems very 2002, but this time it's not about graphics — it's about speed. Reserving a Quadro processor for graphics, Paradigm is computing seismic attributes on a dedicated Tesla graphics processing unit, or GPU, rather than on the central processing unit (CPU). This is cool because GPUs are massively parallel and are much, much faster at certain kinds of computation because they don't have the process management, I/O, and other overheads that CPUs have. This is why seismic processing companies like CGGVeritas are adopting them for imaging. Cutting edge stuff!

In other news...

This regular news feature is for information only. We aren't connected with any of these organizations, and don't necessarily endorse their products or services. 

PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (6)

> BP have released an amazing-sounding 300GB deepwater Gulf of Mexico seismic dataset to academia, according to their in-house magazine (via OilIT).

I can't find it on OilIT. The BP magazine says it's donated to 14 US universities. Would be great for the OpendTect OSR. Anyone for a URL with this data?


Bert

May 11, 2012 | Unregistered CommenterBert Bril

@Bert: It was just a passing mention in the April issue of OilIT, and really only said what I wrote. The BP magazine article seems to be all there is to go on. I know one person who is 'in the queue' for the data, but it seems as if it's academia only and certainly not open — I'm sure there's a long, complicated, scary license agreement.

Aside: In my only run-in with academic data, I wanted to use two lines of ION's very nice Beaufort SPAN data for teaching undergraduates. They told me the university already had access to a similar dataset and just to ask around and find the data. But the researchers with the data claimed the license agreement prohibited their sharing it and wouldn't even give me rasters of the data.

Yet another reason why the OSR is so awesome.

May 11, 2012 | Registered CommenterMatt Hall

Ahh Paradigm....

All Marketing and no bite.

http://www.youtube.com/watch?feature=player_embedded&v=g1l9b9He-7E#!

May 11, 2012 | Unregistered CommenterToastar

@Toastar: Oookay, that's different. I suppose there's nothing wrong with aspiration! I wonder if any of their marketing people have ever actually used one of their products in battle...

That hallucinogenic video reminded me of a great parody that someone at Landmark (I think) made about volume viz and interpretation. It was awesome, featuring a cardboard box,... you can probably imagine the rest. I just scoured YouTube but can't find it anywhere. Shame — it was the perfect antidote!

All that said, their adoption of NVIDIA Maximus is pretty cool, even if it doesn't quite look like Minority Report.

May 11, 2012 | Registered CommenterMatt Hall

IDK it's just the concept of GPU's in interpretation space doesn't really excite me. The Thing with GPU's as math accelerators is they really only do well with embarrassingly parallel problems. The big thing to come to mind with this is migration, And Acceleware has been offering production level GPU RTM since like 2009.

Post Stack I think the only real big gains you might see are from Certain Complex Trace Math, I'm think Spec Decomp. With this I'm thinking you could really get the max potential (10x speed up). With stuff like Coherency Your going to get bottlenecked at the bus. The Big Limitations of The Tesla is the PCI-E Bus is limited to about 8 GB/S Whereas an i7 has a memory bandwidth of something like 20-35 GB/S per chip depending on configuration.

So yeah I'd love to see some benchmarks showing how long that coherency actually is sped up by adding a Tesla.


I think I may just be bitter from actually trying to use PDGM. >.>

May 11, 2012 | Unregistered CommenterToastar

@Toastar: It's a good point: I haven't seen any benchmarking, and now that you mention it, benchmarks are rare in our business. There are reasons for that, I realize, but it would still be fun to see how long various operations take in different software and on different machines. Hmm, an Olympiad at SEG...?

May 12, 2012 | Registered CommenterMatt Hall

PostPost a New Comment

Enter your information below to add a new comment.
Author Email (optional):
Author URL (optional):
Post:
 
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>