( ESNUG 459 Item 6 ) -------------------------------------------- [12/14/06]
Subject: ( ESNUG 458 #5 ) Precision is better & at 1/3 price of Synplicity
> Precision appeared to give good results at first. However, it's tendency
> to quietly and ruthlessly optimise parts of the design away cost us time
> and money. We ended up chasing a number of what we thought were bugs with
> our design -- that turned out to be bugs in the output from Precision.
>
> Another issue was the instability between revs. A minor rev upgrade could
> render a previously fully functional FPGA image crippled. ... I haven't
> worked with Precision for a few months now, but I think I would have to
> think long and hard before using Precision RTL on any future project.
From: [ I am the Furby King ]
Hi, John,
Several years ago (perhaps in 2000?) we pitted the old Leonardo Level 3
against the Precision and noticed at that time that the two were a wash in
terms of performance. Neither compared with Synplify, but we felt that we
needed a 2nd synthesis tool in-house so that we could have a fall-back
tool in case Synplify had a bug or couldn't handle something.
Over the years, we have noticed Precision getting better with each release.
Today, we feel it has far surpassed the capabilities that Leonardo had. In
fact, we feel so confident about Precision that we are actually migrating
from Synplify to it. This means that the roles are going to be reversed and
Synplify will be our "fall-back" tool. We use Xilinx, Altera, Actel, and
Lattice, and have great results with Precision. We are also Precision
Physical customers, but that's a different story. It's a great tool that
we can't live without when we encounter a timing crunch situation.
We have also other reasons for migrating to Precision from Synplify.
1.) First, but not necessarily foremost, is the cost. Synplify costs
us quite a bit more than Precision. Why pay more for equivalent
performance? I won't give away the actual numbers because I'm
afraid that the vendors would know who I am based on those (you
know, different customers get different discounts and such). But
I can tell you that Synplify Pro (WAN) is currently 2.53x the cost
of Precison (WAN) lease cost per year for adding it to our multi-
year lease agreement.
2.) Second, is the quality of Synplify as of late. We have found bugs in
nearly every Synplify release over the last few of years. Some were
minor, but lots were major. It appears to me that after Synplicity
went public, they quickly realized that "Oh crap!, we don't have
anything to sell." They had saturated the FPGA synthesis market and
found themselves surviving off of a maintenance revenue stream. So,
Synplicity started to create niche products to sell. These may be
great tools (I know that some were not) but regardless, they lost
focus on their flagship product - FPGA synthesis. The quality of
Synplify over the last few years proves the lack of focus, at least
for us. So, Synplify is no longer worlds above its competition in
terms of performance and quality any more.
Synplicity's more expensive as well. Top it off with their recent
announcement of a 10% price hike and we were pushed over the edge.
After Synplicity hikes the price 10%, the Precision to Synplify Pro
price ratio will go up to 2.78x. It sounds incredible, but I have
it on paper.
In response to the guy's story on how Leonardo did what Precision couldn't,
all designs are different. We have 2 tools because sometimes there are
things that 1 tool does that another can't. It's the nature of the field.
You're never going to find a tool that is perfect for all scenarios. I just
encourage the user community out there to remember that 1 success story (or
1 failure story) does not tell the entire story.
We have over 100 unique users of FPGA synthesis here & overall, the results
are in favor of Precision.
Feel free to use all or part of this message anonymously.
- [ I am the Furby King ]
---- ---- ---- ---- ---- ---- ----
From: Christian Bara <christian.bara=user domain=fr.thalesgroup spot calm>
Hi, John,
Since tool performance can change over time, we run regular benchmarks on
10 of our most significant FPGA designs regularly. Every quarter we look
at all the "serious" FPGA synthesis tools on the market: Mentor, Synplicity,
Altera, and Xilinx.
For all our typical FPGA designs, in 2006, Precision gave us better results
in at least one domain ("performance" or "area") while in the other domains
it remained at least as good as other tools. Precision Synthesis RTL
typically shows a 10-20% average improvement on frequency or area over
Synplicity, Xilinx and Altera. Altera Quartus remains the closest tool in
term of QoR, but not its VHDL parser.
Since we only use VHDL in Thales, the performance of an FPGA synthesis tool's
VHDL parser is important part of QoR to us. Precision RTL's VHDL parser is
by far the best tool in this area -- it support the real VHDL standard as it
should. For example, packages such as "textio" or "math" are supported and
this is very important for us for most of our designs. This saves us a lot
of time, for example in the case of memory declaration/initialization in the
RTL code. We cannot even compile some of our designs with the other FPGA
synthesis tools! In Thales we carefully follow the VHDL standard, using
automatic design checking tool and specification and standard tracking.
A full standard VHDL code support is crucial for our internal process suport
and for tasks such as reuse of our internal IPs, etc.
Precision RTL has the best VHDL support I've seen anywhere.
- Christian Bara
Thales Group Land & Joint Systems Colombes, France
Index
Next->Item
|
|
