up to 8

On a sunny day (Tue, 10 Feb 2009 12:20:15 -0800) it happened John Larkin wrote in :

Very few people do HD video editing or nuclear arms simulation yet. And my PC runs a lot more threads then 16,

Even this newsgroup has more.

In my view, more and more cores (then say 2 to 6) is a dead end road.

We need faster chips. Or something else entirely.

Well, some graphics cards do OK with a few hundred cores, but that's SIMD most of the time. It can certainly take the PC past the 1 TFLOPS mark on some tasks.

--
Dirk

http://www.transcendence.me.uk/ - Transcendence UK
http://www.theconsensus.org/ - A UK political party
http://www.onetribe.me.uk/wordpress/?cat=5 - Our podcasts on weird stuff

Last year someone here in sed claimed there will never be more than

4 cores on a die ...

Mark

Yes... I need spice at max speed..I'm getting tired of model trimming (make primitive) for fast run. I could use a 10Ghz CPU for my apps. Ever sim a bunch ltspice op amp models...It sims too slow for me.

I had a visitor say "Why is your computer running so slow?' I said, 'It's 3Ghz and it's doing lots of math.'

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

On a sunny day (Tue, 10 Feb 2009 21:01:16 +0000) it happened Dirk Bruere at NeoPax wrote in :

Yes, absolutely, as far as vector like operations go, and indeed some high end cards have H264 *encoding* support in hardware too. We discussed the possibility of using a FPGA to speed up database searches here in the past (by a large factor).

When one stays with video editing for example, usually that is one person, doing the editing of one thing at the time. Difficult to split over many processors or cores, although for example Cinelerra claims it supports 'render farms' (and I finally got Cinelerra working by downloading a special Linux version that has it pre-installed, never had much luck compiling it, but anyways I cannot run it as that would require yet an other PC, and I do not want virtualisers), But that is as far as normal users will ever get I think.

I would not mind a 100 GHz CPU (with suitable fast memory), some render stuff runs all night. OTOH you get used to it being ready in the morning :-)

On a sunny day (Tue, 10 Feb 2009 22:21:55 +0100) it happened "TheM" wrote in :

Even PS3 uses 7 of the cores in the Cell processor, been out for years. The point is: How _useful_ is all that? Looking at Sony's result for the year (and comparing against Wii for example), it brought them nothing.

Sure Intel can try to outwit AMD with bringing out more cores... But I would not want one, unless free and with a fraction of the power consumption. Else not even for free.

[snip]

What are you doing with an OpAmp that causes a slow simulation? That sort of circuit should sim about as fast as you can blink.

Provide a sample of "slow" and I'll test it here.

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
"It isn't that democrats are ignorant. Far from it... it's just that 
they know so much that just isn't so"     -Ronald Reagan

The GHz race is tapering off. The future of general-purpose processors, and even embedded processors, is more cores... LOTS of cores. I guess I'll post next when Intel or AMD hits 64.

Some people are applying GPUs as number crunchers, for specialized things like FFTs.

John

On a sunny day (Tue, 10 Feb 2009 14:16:37 -0800) it happened John Larkin wrote in :

AMD looks a bit dark ATM, they have made big losses, look at the numbers on the bottom of this page, it is in German, but the numbers speak for themselves, could be dead AMD:

Yes. And FPGAs

I recently ripped out a LT1720 comparator and replaced it with a behavioral- schmitt-triggered buffer- with-differential- input to get a faster sim..

Also, a LT1218 and an LT1354 and are on my hitlist to replace with more primitive models that run faster.. I don't need everything simulated.. However, I shouldn't be even trying those parts in the first place. I'm not buying them and I should use the universal op amp models and sim only the important parameters.

I'm trying out smps line harmonic analysis. If I'm doing it right, the run times are long since I need several line cycles (60hz) while the rest of the circuit blazes away at about

1Mhz. I click run and go for lunch.. The faster the CPU, the more sloppy I can be with the models.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

D,

Are you using LTspice? Or something else?

LTspice has various new devices included (many of them undocumented) that Mike specifically created to speed up switch-mode circuit simulations...

---Joel

Yup...LTspice 4 But I'm not expert on it, but I try.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

I was mildly disappointed when I upgraded from a single core AthlonXP2400+ to an Intel Quad Core Q6600. Theoretically it should have been about 8 times faster. In practice it barely seems a factor of 2, and that's taking into account the fact that most s/w doesn't support all 4 cores.

The real subjective speedup will come when SSDs replace HDDs and access time falls from 10mS to

And realtime multistream video compression for streaming over the Net (something I'm working on now).

--
Dirk

http://www.transcendence.me.uk/ - Transcendence UK
http://www.theconsensus.org/ - A UK political party
http://www.onetribe.me.uk/wordpress/?cat=5 - Our podcasts on weird stuff

What is your definition of "slow"? Runtime in seconds?

Max timestep setting?

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
 I love to cook with wine     Sometimes I even put it in the food

On a sunny day (Tue, 10 Feb 2009 23:03:32 +0000) it happened Dirk Bruere at NeoPax wrote in :

What bitrate, H264? What size format, how many fps? How many channels?

It's a 'the faster, the better' sort of thing. Put it this way... I post on here while waiting for simulations to complete.. I haven't timed it...Could be 5 to 10 minutes.. I click and do other things. Might be all just due to newbie settings..dunno..

I leave the maximum timestep box blank..

Well..that was good timing..One of my sims just finished..

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

H264 It's all dictated by domestic upload limits ie 400kb/s

--
Dirk

http://www.transcendence.me.uk/ - Transcendence UK
http://www.theconsensus.org/ - A UK political party
http://www.onetribe.me.uk/wordpress/?cat=5 - Our podcasts on weird stuff

On a sunny day (Tue, 10 Feb 2009 23:52:47 +0000) it happened Dirk Bruere at NeoPax wrote in :

You may perhaps be interested, if you use Linux, in what I do here:

This script should be started first on the receiving side, it listens on port

1234:

# listen-ts-c while [ 1 ] do echo "Hold ctrl C down to abort." netcat -l -p 1234 | mplayer -fs -vop pp=0x20000 -monitoraspect

1440:900 -cache 1000 - done exit 1

This will take an AVI movie, and send it in H264 format at about 340 kbps to IP_ADDRESS on port 1234:

You can make ffmpeg read from file too, almost any format:

#send H264 to an IP address port 1234 ffmpeg -f avi -i Buenos-Aires1-2001.avi \\

-f avi \\

-vcodec h264 -r 2 -b 220 -g 300 -bf 2 \\

-acodec mp2 -ab 64 -s 352x288 \\

-y /dev/stdout | \\ netcat IP_ADDRESS 1234

Wow, I just tried this, and it works!

You can use UDP protocol by simply adding the '-u' command line option to netcat on both sides.

The format is transport stream, .ts

Now you can send multiple programs to multiple ports, for example the second program to port 1235.

I dunno how to exactly put more then one program in the transport stream, but that should be possible. Maybe videolan-client has the routines for that.

If your source is not an .AVI, but in some other format, then change the option in ffmpeg that says: ffmpeg -f avi -i Buenos-Aires1-2001.avi to for example for mpeg2: ffmpeg -f mpegvideo -i some_movie.mpg

Typing: ffmpeg -formats shows the zillions of options.

As they say sometimes: 'I hope this helps'

And, you can have ffmpeg read from stdin too, I use that to pipe yuv format from for example a webcam, or basically from any mjpeg tools compatible utility. ffmpeg -f yuv4mpegpipe -i - \\