( ESNUG 448 Item 2 ) -------------------------------------------- [11/11/05]
Subject: More Users on the Apache, Sequence, ChipVision, Golden Gate Demos
> As I mentioned in ESNUG 447 on Monday, here's the first of the videos
> shot at DAC with Pallab and I pestering the EDA vendors. In this case
> it's power tools from Apache, Sequence, Magma, Golden Gate, ChipVision,
> Prolific and Atrenta.
>
> http://www.deepchip.com/demos/dac_power_tools.fhtml
>
> The interviews range from a quickie 2 minute chat with Sequence to a
> half hour in-depth discussion with Magma about Blast Power; but most
> run ~5-7 mins.
From: Dinesh Venkatachalam <dinesh=user domain=legendsilicon spot gone>
Hi, John,
All the demo's were pretty technical/mktg mixed. I can't complain too much.
The ChipVision demo was the most interesting since it was C-based analysis.
Their tool had promise but as all C-based tools, the C-code needs to written
in a certain format that keeps the parser etc happy. Our C-code is written
by "algorithm folks" who care less about structure and more about the
algorithm, so it was not a viable solution unless we rewrote the code. Cost
of the ChipVision tool was also reasonable
Sequence Power Theatre claimed the most, including links to layout. The
problem I had with this tool is that we have already done all of the possible
power reductions that the tool could suggest or find, so the Sequence tool
essentially becomes something which reports which modules use how much power
(similar to a waveform display). For a one week post-partum analysis, we
felt the cost was too much. I think the Sequence tool is well suited for a
larger company where multiple designs were given to the power analysis group
to produce reports
We have a Magma Blast Create flow with Blast Fusion backend done in the US
or abroad. The backend company has used Blast Power and provided us with
power numbers and they have matched our lab results so far. We are looking
into getting Blast Power in-house to take care of IR drop issues and power
mesh as we go into lower geometries.
Golden Gate was focused on the backend and more suited for a layout house
especially if it is an add on tool. The only reason that Blast Power is
being considered herewas that we already have a Magma flow.
- Dinesh Venkatachalam
Legend Silicon Corp. Fremont, CA
---- ---- ---- ---- ---- ---- ----
From: Sergio Golombek <sergio=user domain=airgonetworks spot gone>
Hi John,
That Golden Gate demo is good and presents some new ideas in power-plan
synthesis, and also power minimization techniques . It appears to have
everything included in order to make a good power analysis flow, including
2.5-D RC extractor, integrated STA with crosstalk analysis, IR drop and
even DRC. It also appears to have a very large capacity. (10M gates flat!)
However, I didn't see much in the way of dynamic power analysis. I'd like
sometime to evaluate it and see what they can add to our Magma/Apache flow.
- Sergio Golombek
Airgo Networks, Inc. Palo Alto, CA
---- ---- ---- ---- ---- ---- ----
From: Peter Jeng <peter.p.jeng=user domain=intel spot gone>
Hi, John,
I did go through Apache Redhawk demo. This demo was well received. I think
this Redhawk tool has some good features such as dynamic IR drop analysis
which is interesting to me. I will follow up with this tool at Q1/06 time
frame. Thanks.
- Peter Jeng
Intel Corp.
---- ---- ---- ---- ---- ---- ----
From: [ Dr. Pepper ]
Hi, John,
Apache impressed me with its PSIwinder addition to Redhawk and that Redhawk
is proven in the TSMC Reference Flow 5.0. My question is the TSMC Reference
Flow 6.0. In the 6.0 flow it looks like PrimeRail may also be up to snuff.
The overall amount of information trumpeting PrimeRail's arrival for 6.0 on
the web is "scant" at best. Is PrimeRail ready to compete with Redhawk? Any
inputs from other users? Must be anon.
- [ Dr. Pepper ]
---- ---- ---- ---- ---- ---- ----
From: Dvor Efrat <dvor.efrat=user domain=infineon spot gone>
Hi John,
The Sequence demo was more technical (naturally) and it was interesting for
me to understand their flow and how they cope with power analysis &
verification (IR drop) in terms of gate level simulations input (what kind
of inputs do they support, how correlated "low accuracy" inputs like RTL
simulations are related to heavy gate level simulations inputs results).
Sequence mentioned an internal "event simulation" machine which handles
different simulation inputs / user defined switching activity -- but I'm
not sure how can they propagate the boundary switching activity beyond the
first FF's in the logic paths.
For me, if I'm to use this, I'll need to "play around" to get a feeling
about this. Sequence did try to present "good correlation" between their
"low accuracy" flow and their "high accuracy" flow. But the details shown
in the Sequence demo suggested that although they did get total power
correlation, their inner parts estimations were off (some higher, some lower;
eventually it summed up to ~1W both cases...)
Hope my 2 cents were worth it. Anyway - thanks for the useful demos!
- Dvor Efrat
Infineon
---- ---- ---- ---- ---- ---- ----
From: [ Red Dragon ]
Hi John,
In rough view, the demo helped me to be quickly familiar with the tool.
I think Prolific may be useful to meet tight timing and power requirements,
but it depends on it's pratical performance. I won't know this until I try
it. Keep me anonymous.
- [ Red Dragon ]
---- ---- ---- ---- ---- ---- ----
From: Nir Granite <nir.granite=user domain=intel spot gone>
Hi John,
The Sequence demo is quite good. There is some tech meat (could be more to
my opinion). The tools look attractive and professional from the Demo.
- Nir Granite
Intel Corp.
Index
Next->Item
|
|
