Maker Pro
Maker Pro

best processor for running Silvaco Smartspice

P

Paul Spitalny

Jan 1, 1970
0
Hi,
It is new computer time again. I design a lot of circuits (like sigma
delta converters) where I have to run very long transient spice (Silvaco
SMartspice) simulations(4-5 hours). I have been using a 2.5GHz pentium 4
with 1G of RAMBUS ram (500MHz bus). I need something faster. Any ideas
as to what the fastest possible machine might be?

THanks

Paul
 
J

Jim Thompson

Jan 1, 1970
0
Hi,
It is new computer time again. I design a lot of circuits (like sigma
delta converters) where I have to run very long transient spice (Silvaco
SMartspice) simulations(4-5 hours). I have been using a 2.5GHz pentium 4
with 1G of RAMBUS ram (500MHz bus). I need something faster. Any ideas
as to what the fastest possible machine might be?

THanks

Paul

P3's are faster than P4's... Intel screwed with the math routines.

Your best bet is an AMD chip with an equivalent speed rating... it'll
then do simulations 2X that the Intel chip will.

Surf back this past year... a number of benchmarks were posted.

...Jim Thompson
 
H

Helmut Sennewald

Jan 1, 1970
0
Jim Thompson said:
P3's are faster than P4's... Intel screwed with the math routines.

Hello Jim,
your story about the P3 was true with the first generation of P4, but
the current 3.x GHz P4s are much faster, because of the bigger cache
menory and the faster RAM interface.
Your best bet is an AMD chip with an equivalent speed rating... it'll
then do simulations 2X that the Intel chip will.

This Mathematica 5 benchmark could be valid for SPICE too.
A smaller number is better.
Athlon 64 3400 5.9s
Athlon XP 3200 6.8s
P4 3.2GHz 7.4s
P4 2.6GHz 8.9s
(Your P4 2.5GHz 9.xs ???)

http://www.tomshardware.com/cpu/20040106/athlon64_3400-32.html

It would be great if someone who has access to a P4 3.2GHz,
an Athlon XP 3200 and an Athlon 64 3200(3400) could do a speed
test with SPICE.

Best Regards,
Helmut
 
J

Jim Thompson

Jan 1, 1970
0
Hello Jim,
your story about the P3 was true with the first generation of P4, but
the current 3.x GHz P4s are much faster, because of the bigger cache
menory and the faster RAM interface.


This Mathematica 5 benchmark could be valid for SPICE too.
A smaller number is better.
Athlon 64 3400 5.9s
Athlon XP 3200 6.8s
P4 3.2GHz 7.4s
P4 2.6GHz 8.9s
(Your P4 2.5GHz 9.xs ???)

http://www.tomshardware.com/cpu/20040106/athlon64_3400-32.html

It would be great if someone who has access to a P4 3.2GHz,
an Athlon XP 3200 and an Athlon 64 3200(3400) could do a speed
test with SPICE.

Best Regards,
Helmut

Mark C?

I know my Athlon XP 1600 is almost twice as fast as a P4 running a
rather bitchy circuit in PSpice.

...Jim Thompson
 
P

Paul Spitalny

Jan 1, 1970
0
Helmut said:
Hello Jim,
your story about the P3 was true with the first generation of P4, but
the current 3.x GHz P4s are much faster, because of the bigger cache
menory and the faster RAM interface.




This Mathematica 5 benchmark could be valid for SPICE too.
A smaller number is better.
Athlon 64 3400 5.9s
Athlon XP 3200 6.8s
P4 3.2GHz 7.4s
P4 2.6GHz 8.9s
(Your P4 2.5GHz 9.xs ???)

http://www.tomshardware.com/cpu/20040106/athlon64_3400-32.html

It would be great if someone who has access to a P4 3.2GHz,
an Athlon XP 3200 and an Athlon 64 3200(3400) could do a speed
test with SPICE.

Best Regards,
Helmut
Hi Helmut and Jim,
Yes, as Helmut says, the early P4's are quite different from the new
ones. The new ones have faster ram (twice the bus speed) and also a much
large level 2 chache. I guess the only way to find out to to get access
to some different machines, unless anyone else there has done a similar
benchmarking to the kind we saw about a year or so ago in this newsgroup.

I talked with the guys at Silvaco (they make Smartspice) and they said
that the code is compiled using general double precision math, the kind
that P4's were not so good at previously. But, now..who knows!

Paul
 
J

Jim Thompson

Jan 1, 1970
0
On Mon, 02 Feb 2004 12:29:58 -0800, Paul Spitalny

[snip]
Hi Helmut and Jim,
Yes, as Helmut says, the early P4's are quite different from the new
ones. The new ones have faster ram (twice the bus speed) and also a much
large level 2 chache. I guess the only way to find out to to get access
to some different machines, unless anyone else there has done a similar
benchmarking to the kind we saw about a year or so ago in this newsgroup.

I talked with the guys at Silvaco (they make Smartspice) and they said
that the code is compiled using general double precision math, the kind
that P4's were not so good at previously. But, now..who knows!

Paul

I still have the benchmark circuit files. Would someone like to try
them on a new P4?

How do I tell a "new" P4 from an old one when I go shopping?

...Jim Thompson
 
T

Tim Auton

Jan 1, 1970
0
Paul Spitalny said:
[snip]
Yes, as Helmut says, the early P4's are quite different from the new
ones. The new ones have faster ram (twice the bus speed) and also a much
large level 2 chache. I guess the only way to find out to to get access
to some different machines, unless anyone else there has done a similar
benchmarking to the kind we saw about a year or so ago in this newsgroup.

I talked with the guys at Silvaco (they make Smartspice) and they said
that the code is compiled using general double precision math, the kind
that P4's were not so good at previously. But, now..who knows!

If they're using double-precision - ie 64, rather than 32-bit floating
point numbers - I'd expect a substantial benefit from a native 64-bit
processor. At least, I'd expect the benefit when the code has been
compiled for the 64-bit processor, which it probably isn't yet :)

I'd ask if Silvaco are intending to make a 64-bit version for the
AMD64 range (Opteron, Athlon FX) and whether it's multi-threaded. Even
as a 32-bit processor an Opteron is a formidable number-cruncher and a
dual-CPU Opteron box (which you'd need multi-threaded applications to
take advantage of) would be excellent for 32-bit work now and even
better when there are more 64-bit apps around.


Tim
 
T

Tim Auton

Jan 1, 1970
0
Jim Thompson said:
How do I tell a "new" P4 from an old one when I go shopping?

Simple answer: Look for HyperThreading.

Complicated answer: The answer changed today, with the release of the
first of the Prescott cored P4s, designated P4 x.xE. Northwood was the
most recent P4 core on the market yesterday. Prescott is built with a
90nm process (Northwood is 130nm), has double the on-chip cache and
some new instructions (SSE3). Confusing matters still further the
fastest Prescott-cored P4s haven't been released yet, so the
top-of-the-range 3.4GHz P4 EE (Extreme Edition, with more on-chip
cache [good for spice], basically a rebadged Xeon) is still based on
the Northwood core.


Tim
 
P

Paul Spitalny

Jan 1, 1970
0
Tim said:
How do I tell a "new" P4 from an old one when I go shopping?


Simple answer: Look for HyperThreading.

Complicated answer: The answer changed today, with the release of the
first of the Prescott cored P4s, designated P4 x.xE. Northwood was the
most recent P4 core on the market yesterday. Prescott is built with a
90nm process (Northwood is 130nm), has double the on-chip cache and
some new instructions (SSE3). Confusing matters still further the
fastest Prescott-cored P4s haven't been released yet, so the
top-of-the-range 3.4GHz P4 EE (Extreme Edition, with more on-chip
cache [good for spice], basically a rebadged Xeon) is still based on
the Northwood core.


Tim

Hi Tim,
You mentioned that the additional cache memory would improve things for
spice. Why is that? Would that help a pentium in the sense that its
floating point operations seem to be the thing that slow it down the
most (from what I understand) in a spice simulation.

Here's some verbiage I got from another newsgroup about floating point
math and the P4:

|The P4 is pretty much a wimp on everything fp, except vectorized,
|straightforward mul, add, sub, using SSE2.
|A lot of time consuming work like matrix/tensor multiplications,
|transformations etc. does fall into that category though. So it might
|be a good idea, to see to it, that the code is compiled with Intels
|auto vectorizing optimizing compiler.
|If everything is optimal, you can get 3-3.5 times the performance on
|single precision fp (this is the kind of performance you see in P4
|video encoding). On the other hand, branches, division, ruin it all.

|Just reading some single benchmark, is not going to be of any use to
|you. P4/Xeon benchmarks tend to be 100% SSE2, outrageously optimized
|and highly flattering for Intel. Real applications might be a
|different thing (scalar '387?). Unless you know what the code looks
|like, and how it is compiled, you cannot be sure to get the
|performance common benchmarks imply.
|If you write the software yourself, Intels compiler is a free
|download. Try it if you haven't already.


I may just go out and rent an Athlon for a day and do a spice test run.

P.
 
J

Jim Thompson

Jan 1, 1970
0
I still have the benchmark circuit files. Would someone like to try
them on a new P4?

...Jim Thompson
 
N

N. Thornton

Jan 1, 1970
0
Paul Spitalny said:
Hi,
It is new computer time again. I design a lot of circuits (like sigma
delta converters) where I have to run very long transient spice (Silvaco
SMartspice) simulations(4-5 hours). I have been using a 2.5GHz pentium 4
with 1G of RAMBUS ram (500MHz bus). I need something faster. Any ideas
as to what the fastest possible machine might be?

THanks

Paul

Hi

The fastest PC is currently a 5.25GHz CPU. There is a site about it,
but I dont have the ref. Its a liquid nitrogen cooled overclocked 3
point something CPU, with ordinary fridge compressor/heat exchange
cooling for the rest of the box. I dont know what else you do with the
machine, and thus whether ocing is a sensible route, but it certainly
would be the way to get max performance.


Regards, NT
 
P

Paul Spitalny

Jan 1, 1970
0
Jim said:
I still have the benchmark circuit files. Would someone like to try
them on a new P4?

...Jim Thompson
Hi Jim,
I don't have a new p4 (cause I still have not decided which way to go).
But, I could try the files on my p4 (2.5Ghz with rambus ram). I'd be
curious to compare to the run time on your athlon machine.

Then, as I go further into this investigation I can try to run the files
on some other machines too.

Are the files HSPICE compatible? I am not so sure of how my simulator
would handle a pspice file.

Thanks

Paul
 
J

Jim Thompson

Jan 1, 1970
0
Hi Jim,
I don't have a new p4 (cause I still have not decided which way to go).
But, I could try the files on my p4 (2.5Ghz with rambus ram). I'd be
curious to compare to the run time on your athlon machine.

Then, as I go further into this investigation I can try to run the files
on some other machines too.

Are the files HSPICE compatible? I am not so sure of how my simulator
would handle a pspice file.

Thanks

Paul

HSpice will put you at a distinct disadvantage ;-) (It'll run just
fine, maybe a wee bit slower than PSpice, but it'll run.)

E-mail your address to me and I'll send the benchmark files.

Also anyone else who's interested.

I'd particularly like to see the results on the "new" P4.

...Jim Thompson
 
Top