( ESNUG 593 Item 01 ) --------------------------------------------- [12/21/22]
Subject: Empyrean ALPS benches 2.7X to 38.4X over FineSim/PrimeSim/Spectre-X
EMPYREAN ALPS CRUSHES -X, -XT, & APS: Empyrean is different, because
it's competing on having both better speed and accuracy against the
other zippy SPICE tools; and I have a boatload of hand-on SPICE user
benchmarks to prove it!
For the GPU-assisted Empyrean ALPS-GT, the users found:
"ALPS-GT was 1.5X faster than Siemens AFS-XT (red)."
"ALPS-GT was 3.0X faster than Siemens AFS (vanilla)."
"ALPS-GT was 3.0X faster than Cadence Spectre-APS."
Even plain ALPS without the post-layout GPU boost is still beats
out CDNS/SNPS.
"ALPS was 1.5X faster than Spectre-X."
"ALPS was 4.0X faster than Spectre-APS."
"ALPS is 4.0X faster than Spectre."
"ALPS is ~3.0X faster than Spectre."
"ALPS is ~2.0X faster than HSPICE."
- Empyrean ALPS-GT crushes Spectre-X and AFS-XT is Best of 2020 #1c
From: [ Your Secret Santa ]
Hi, John,
I see it's been two years since you had any user SPICE benchmarks, so I
thought I'd add our recent August 2022 experiences to keep your DeepChip
readers up-to-date.
I use the ALPS true SPICE simulator to run the transient simulation for
our high-speed mixed signal circuits. ALPS inputs and outputs use the
standard SPICE format.
I ran two comparisons with 3 other SPICE simulators on our high-speed
divider, and share my results below. Empyrean ALPS outperformed against
the Cadence and Synopsys SPICE simulators I benchmarked by a wide margin.
CASE 1
Design size: Approx 20K transistors, 1,200K resistors, 800K capacitors
TSMC 5nm (RCC post-layout). To make the benchmarks apples-to-apples, I ran
all the SPICE simulations on 16 CPU cores.
As you can see above,
- ALPS is 12.0X faster than Synopsys FineSim
- ALPS is 1.9X faster than Synopsys PrimeSim
- ALPS is 2.9X faster than Cadence Spectre-X
CASE 2
Design size: Approx 10K transistors, 900K resistors, 600K capacitors
TSMC 5nm (RCC post-layout). I ran all simulations on 16 CPU cores.
As you can see above,
- ALPS was 12.0X faster than Synopsys FineSim
- ALPS was 7.3X faster than Synopsys PrimeSim
- ALPS was 2.7X faster than Cadence Spectre-X
- Synopsys PrimeSim (runlvl=4) failed on accuracy.
(Cadence Spectre-APS is our golden SPICE. Each SPICE (with the
exception PrimeSim runlvl=4) stayed within our +/- 1.5% accuracy target.
On top of these 2.7X to 12.0X speed-ups we saw with purely SW ALPS, when
we used the ALPS GPU accelerator (post-layout) it gave us an *additional*
3.2X speed-up over standard ALPS for Case 1. (So we altogether then got a
cumulative 8.6X to 38.4X speed-up when used the ALPS GPU accelerator on
our post-layout circuits.)
I usually run ALPS (and ALPS GPU) every day during our design stages.
Its Smart Matrix Solver helps a lot on post-layout simulation especially on
design stages that need many iterations for circuit tuning. With ALPS SPICE
runs, I can tweak my circuits 5 to 6 times a day. In contrast, with the
Synopsys and Cadence SPICE simulators, I can only do one 1 circuit tweak/day.
(Which is why I recommend ALPS for our post-layout transition simulations.)
And it's easy to use -- I just use ALPS inside the Cadence ADE command line
interface, and it helps me to complete my design verification more quickly.
- [ Your Secret Santa ]
P.S. Merry Christmas, John, assuming you celebrate Christmas.
P.P.S. Otherwise I wish you a Happy Holidays!
Related Articles:
Empyrean ALPS-GT crushes Spectre-X and AFS-XT is Best of 2020 #1c
Empyrean ALPS GPU crushing Cadence Spectre-APS is Best of 2019 #2c
Jason Xing unhappy his ICScape ALPS SPICE was forgotten last year
Users on Empyrean ALPS as SPICE benchmarks get #6 "Best of 2018"
BDA, Solido, MunEDA, and Silvaco get #5 for Best EDA of 2016
246 engineers on today's SPICE use vs. their future SPICE use
Join
Index
Next->Item
|
