Discrete Opamp Open Design - Page 249 - diyAudio
Go Back   Home > Forums > Source & Line > Analog Line Level

Analog Line Level Preamplifiers , Passive Pre-amps, Crossovers, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 1st January 2013, 09:39 PM   #2481
diyAudio Member
 
scott wurcer's Avatar
 
Join Date: Jan 2004
Location: cambridge ma
Quote:
Originally Posted by RNMarsh View Post
So the foke-lore on the street says. But, the laptop cpu architecture isnt optimized for number crunching and the word size is a lot smaller etc etc. Larger word size means more can be done in each cycle. Number crunching is what the super's do best as they have a lot of it to do. Also, the use of memory is much faster and a lot more of it. etc etc etc. [ I was at LLNL when the first Cray went in and subsequent ones etc.] Thx-RNM
The CRAY-2 added up to 2G bytes considered excessive. Can you quote some computational comparisons? The IC's were not up to it there was no solid state memory in 1985 that even remotely matches what we have now i.e. the CRAY had an 80MHz or so clock and 50nsec RAM.

I made my first prediction of 2013 and was right! I went to walk my dog at 5:15 and said when I get back Wisconsin will already be down 14-0.
__________________
"The question of who is right and who is wrong has seemed to me always too small to be worth a moment's thought, while the question of what is right and what is wrong has seemed all-important."
  Reply With Quote
Old 1st January 2013, 11:21 PM   #2482
diyAudio Member
 
scott wurcer's Avatar
 
Join Date: Jan 2004
Location: cambridge ma
I remembered later, when researching the literature for an graduate course project in 1998 I ran across a paper on phase retrieval in image processing from the 80's. They quoted massive numbers of 128X128 2D FFT's off loaded to a CRAY and the numbers are a joke by today's standards. FFT throughput is one of the bread and butter apps for number crunchers.

EDIT - This is really fun stuff in the nerdy sense, the paper I mentioned is somewhere in the appendix... http://digitus.itk.ppke.hu/~matyi/op...omparison1.pdf
__________________
"The question of who is right and who is wrong has seemed to me always too small to be worth a moment's thought, while the question of what is right and what is wrong has seemed all-important."

Last edited by scott wurcer; 1st January 2013 at 11:36 PM.
  Reply With Quote
Old 2nd January 2013, 06:06 AM   #2483
RNMarsh is offline RNMarsh  United States
diyAudio Member
 
RNMarsh's Avatar
 
Join Date: May 2012
Location: 2457 Cascade Trail; Cool, CA. 95614
Supercomputers are designed formost to be efficient; Computationally efficient is the metric of interest. This is where the men get seperated from the boys.

FFT are not a bread and butter app for these supercomputers. PC's can do that fast enough... even without being computationally efficient. Besides, who cares what a 30+ year computer can do compared to a PC today? The best PC today isnt a supercomputer. The IBM Seqouia is a brillient and highly efficient number cruncher... ten years in the design with IBM and LLNL computer scientists and others.

Update! Cray XK7 is now the worlds fastest supercomputer (Nov 12, 2012)! IBM Seqouia is second.

-Thx RNMarsh

Last edited by RNMarsh; 2nd January 2013 at 06:22 AM. Reason: Computationally effecient is the metric for supercomps
  Reply With Quote
Old 2nd January 2013, 07:24 AM   #2484
qusp is offline qusp  Australia
diyAudio Member
 
qusp's Avatar
 
Join Date: Oct 2009
Location: Brisbane, Australia
for a while graphics computing was heading the way of VLWL (very long word length) or CISC computing, but parallel RISC based cores running higher speeds were more appropriate for the tasks a home or even most enterprise solutions needed I guess.

I was working with a Silicon Graphics PowerChallenge and O2 in Shake/Flame on IRIX for a while when I did a Special effects/compositing Scholarship back in the late 90's

I believe they Owned Cray for a while too before their demise
  Reply With Quote
Old 2nd January 2013, 07:56 AM   #2485
diyAudio Member
 
Join Date: Jan 2011
Location: Germany
I am sure you can run different TRAN:s on each of those Linux boxes, but can you actually run one and only TRAN whose computational load s split by all these Linux boxes? I do not think so. But if so, no one would be more happy than me who does runs hour long TRAN analysis on daily basis!

IS there a SPICE around that can do TRAN and take advantage of multi core/CPU:s?

Quote:
Originally Posted by scott wurcer View Post
Our entire network of Linux boxes is configured as a compute farm and at night you could have dozens of machines running a transient analysis in parallel. Things that take days can take hours. The partitioning is special software but it does exist.
  Reply With Quote
Old 2nd January 2013, 08:20 AM   #2486
jcx is offline jcx  United States
diyAudio Member
 
Join Date: Feb 2003
Location: ..
Default matrix inversion can be parallelized

go to the LTspice Yahoo Group (you have to sign up) - the database shows some execution speed vs processor - multi core do sim factors faster on modern multicore PC (maybe not exactly by # of cores but worthwhile)

Last edited by jcx; 2nd January 2013 at 08:27 AM.
  Reply With Quote
Old 2nd January 2013, 08:59 AM   #2487
diyAudio Member
 
Kindhornman's Avatar
 
Join Date: Aug 2012
Location: Los Angeles, California
I can say now that the CAD software that I am using has started to take advantage of both multiprocessors and multithreading rendering and 3D surface development is much faster than it has ever been. Unless you had the money for a seat in Catia or a few other programs you just couldn't take advantage of the hardware we have today. I am looking forward to having more processors and thread and being able to offload some of the number crunching to these supper new GPU cards. The day that I don't have to wait for a screen to refresh will be a time saver for sure.
  Reply With Quote
Old 2nd January 2013, 03:19 PM   #2488
diyAudio Member
 
scott wurcer's Avatar
 
Join Date: Jan 2004
Location: cambridge ma
Quote:
Originally Posted by staccatiss View Post
I am sure you can run different TRAN:s on each of those Linux boxes, but can you actually run one and only TRAN whose computational load s split by all these Linux boxes?

Yes, last month we generated the eye diagram for a 10G TIA running a 2^31 bit pattern looking for inter-symbol interference. One circuit one LONG transient analysis. 3+days on a fairly loaded work station, a few hours in the farm.
__________________
"The question of who is right and who is wrong has seemed to me always too small to be worth a moment's thought, while the question of what is right and what is wrong has seemed all-important."

Last edited by scott wurcer; 2nd January 2013 at 03:28 PM.
  Reply With Quote
Old 2nd January 2013, 03:26 PM   #2489
diyAudio Member
 
scott wurcer's Avatar
 
Join Date: Jan 2004
Location: cambridge ma
Quote:
Originally Posted by RNMarsh View Post
Besides, who cares what a 30+ year computer can do compared to a PC today?
Soooorry, you seem to.
__________________
"The question of who is right and who is wrong has seemed to me always too small to be worth a moment's thought, while the question of what is right and what is wrong has seemed all-important."
  Reply With Quote
Old 2nd January 2013, 03:36 PM   #2490
RNMarsh is offline RNMarsh  United States
diyAudio Member
 
RNMarsh's Avatar
 
Join Date: May 2012
Location: 2457 Cascade Trail; Cool, CA. 95614
Quote:
Originally Posted by scott wurcer View Post
Soooorry, you seem to.
really? and here i thought it was you who brought that up. whatever.

SIM users can rejoice in the fact that standards are being used which are designed to allow for scaled down PC use in the future. The supercomputer use of RAID and Linux and the like will make closing the performance gap a little easier and even faster than before.
Cray is alive and well. Their memory systems are being sold to HP, Dell, IBM thru a standard Lustre client. Cray's Sonexion (memory) is available for a wide range of x86 Linux InfiniBand-attached compute clusters. 1TB I/O bandwidth. The trickle-down theory may not work in economics/politics but it is working well in the computer world.
Thx-RNMarsh

Last edited by RNMarsh; 2nd January 2013 at 03:56 PM.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Discrete OPAMP audio-gd Vendor's Bazaar 27 20th September 2012 04:02 PM
discrete opamp help blackpowderaudio Parts 0 16th December 2009 03:46 PM
THAT transistor headphone amp (250ma discrete opamp) design sanity check. Russ White Headphone Systems 19 13th December 2007 12:52 PM


New To Site? Need Help?

All times are GMT. The time now is 11:51 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2