Adjusting PC Hyperthreading for Spice Simulation

"D from BC" schrieb im Newsbeitrag news: snipped-for-privacy@4ax.com...

Hello,

You can get what you asked for if you use the latest LTspiceIV. It will stress all your CPU-cores to its limits. Let's see when the first people will claim LTspice has burned their Quadcore-CPUs because of running long lasting simulation runs. :-)

None of the other SPICE programs you can afford will use more than one CPU-core.

Best regards, Helmut

Great!

I checked my LTspice version ..it's version 2.25d :( Ugh.. I'll get version 4.

Thanks

D from BC myrealaddress(at)comic(dot)com British Columbia Canada

I wonder if the other SPICE program writers are massively pissed about LTspice being free.

Perhaps the LT president and LTspice writer get all 4 tires spiked a little too often :P But I don't think it's that gangsterlike in technology.. :)

D from BC myrealaddress(at)comic(dot)com British Columbia Canada

I'm running LTspice4 now. Confirmed using help/about. However. in Windows task manager scad3.exe is running.. huh...

Anyways.. My spice is not running faster and Windows Task Manager is only showing CPU usage at 50%. iow...1 virtual core saturated.

I did spot the max treads option under the LTspice control panel. It's set to 2.

I haven't noticed a change in speed. :(

D from BC myrealaddress(at)comic(dot)com British Columbia Canada

With hyperthreading, a single thread will use the entire CPU in the absence of any competing threads (Windows reports that it's only 50%, but it's lying; you can confirm this by running two processes on an otherwise-idle system and noting that one process runs more slowly as the other process consumes more CPU).

If you have a CPU with two full cores (Pentium-M), the only way that you can utilise both cores is with multi-threaded code. You can't make the CPU behave as a single, faster core.

Not any more than than original free spice. I just downloaded the source, compiled and spiced.

Welcome to the world of virtual cores. What you REALLY have is the ability of the core to run different threads at once as long as they do not need the same resources. So, if you have a bunch of threads waiting for the floating point unit and nothing waiting for the integer unit, you will have no performance gains. Intel has loads of information on their website on this. The advantage comes when running multiple apps, odds are you will be able to utilize more of the core's overhead that way. You want to get REAL gains you need REAL cores, although the i7 is suppose to do much better with the new Hyperthreading. Vista thinks I have 8 cores to play with.

No. Engineers don't choose simulators, jerks in management do... that's why Cadence sales are so high, even though, IMNSHO, it's a royal piece-a-crap, grossly user unfriendly and butt slow.

...Jim Thompson

"D from BC" schrieb im Newsbeitrag news: snipped-for-privacy@4ax.com...

Hello,

I just tried a SMPS simulation. LTspiceIV has a caused a CPU usage of 100% on my dual core CPU on Win-XP. The gain regarding simulation speed depends on the circuit size. Small circuits will not gain much in speed. Large circuits will run up to 3 times faster on a quad core CPU compared to a single core CPU with the same clock frequency.

Best regards, Helmut

That's because with hyperthreaded CPUs, "50% utilization" of "one CPU" is still using (nearly) 100% of the total clock cycles available if nothing else is running on the "second" CPU.

In other words... you get 1 billion operations per second (or whatever). Hyperthreaded CPUs just give the appearance of two CPUs so that if a particular thread is waiting on, e.g., a memory read from DRAM (this can take hundreds of cycles), another thread can run in those hundreds of cycles. However, if the thread is never waiting on anything, it will consume all 1 billion cycles every second.

A 1GHz hyperthreaded CPU can compute pi no faster than a non-hyperthreaded CPU. (Nor will it be any slow.) But hyperthreaded is a win when you have multiple *different* threads running around when one starts having to wait and would otherwise just be wasting CPU cycles...

Oh, I think they're a little more ticked than before -- LTspice has many, many improvements over Berkeley SPICE, plus it has a decent (if simplistic) integrated graphing/probing environment, that had to be provided by various (not as tightly-integrated) add-ons with the Berkeley SPICE.

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

And amazingly, sometimes those jerks haven't even used SPICE in years (since, e.g., college)... and possibly not ever!

So maybe I have LTspice running at nearly full blast with my single core yet 2 virtual core ye old hyperthreading Pentium 4 Prescott.

D from BC myrealaddress(at)comic(dot)com British Columbia Canada

Try to disable Hyperthreading in the BIOS. In rare cases it will actually slow things down, like in SQL server

Cheers

After reading replies, I did this experiment..

Ran the simulation and looked at the simulation run speed at the bottom of the LTspice window.

LTspice reports simulation speed ~27us/s Windows shows a total CPU usage at 50%

Spawned another instance of LTspice and run the same simulation at the same time..

LTspice reports simulation speed ~15us/s Windows shows 100% CPU usage..

1 thread one core, the hyperthreading lets the thread use up all the mips as long as there's no completing threads.

Thanks for replies...That clears that up.

D from BC myrealaddress(at)comic(dot)com British Columbia Canada

As the kids say: THIS. So much THIS.

None of my other *programs* use more than one CPU core (that I have noticed).

And that's using Wine on linux! It was the first time I ever saw "400%" CPU usage. Good job Mike.

I don't think so. Ltspice is not very user friendly IMHO.

-- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... "If it doesn't fit, use a bigger hammer!"

--------------------------------------------------------------

Compared to entering netlists by hand it's quite friendly. :-)