( ESNUG 424 Item 4 ) -------------------------------------------- [03/09/04]
Subject: ( DAC 03 #31 ) Another User's Impressions Of Apache RedHawk-SDL
> Well, the pros of Apache's RedHawk-SDL are that you can run dynamic IR
> drop analysis without VCD. It takes the output from a PrimeTime run to
> stimulate the circuit. This is why we decided to look at it in the
> first place. We have never done dynamic analysis and it is becoming
> obvious that this is needed moving beyond .15um technology.
>
> The results RedHawk-SDL showed compared nicely to Simplex (which is what
> we use now) as far as static IR drop analysis. They only analyzed one
> large block of a chip for us. Then when they went into their dynamic
> analysis, the results looked much worse, 3x worse. Since the chip
> operates very well, we either have enough margin on this chip or the
> tool is too pessimistic. But the results were extremely interesting and
> they did this analysis is a very short period of time.
>
> The only problem I have is that it is not integrated into our point
> tools. Neither is Simplex and it causes problems. I am currently using
> Magma and their Blast Rail IR Drop analysis tool is very nice to use
> because you can run it anytime the data has changed very easily. RedHawk
> has the capability to do what-if scenarios but the changes would still
> have to be rolled back in your point tool eventually.
>
> - Ed Mahr of Conexant in http://www.deepchip.com/items/dac03-31.html
From: Hugh Mair <hugh=user domain=ti spot calm>
Hi, John
Sorry for the long delay in response. My impressions of Redhawk-SDL.
First the Good:
Redhawk is truly a differentiated product. I believe Andrew was really out
ahead of the curve here.
It delivers a robust solution to a tough problem that most other EDA vendors
have simply avoided addressing. I can't compare it to AstroRail, because it
doesn't even come close. Haven't looked at Cooltime, had a bad experience
with Sequence, but that's another story.
I just saw a presentation from a competitor on future capability of their
product, which, even though way optimistic (IMHO), was not even up to the
current capability of RedHawk.
A couple of basic points about Redhawk:
Its accurate. We have benchmarked Redhawk-SDL against the largest circuits
we can simulate in SPICE and have seen very good correlation. Note, we have
not benchmarked on-chip inductance.
Redhawk-SDL is fast; really fast! This is a big advantage and a refreshing
change.
Why I am interested in it:
I am involved in developing static & dynamic power management architectures
for sub-100 nm wireless products at TI. Almost all the techniques we are
using and planning to use dramatically change the landscape of the on-chip
power delivery.
I looked at Redhawk, not really for its advertised capabilities, but for the
differentiated approach to the problem. For me, the biggest value is not so
much what it does, but how it does it. We have been able to adapt the tool
to solve two out of three problems we outlined to Apache when we engaged
them around 9 months ago. Namely:
1. The modeling of dynamic IR drop on unconventional supplies (not
grid based) in response to specific power management events.
2. The modeling of dynamic IR drop of a power supply grid incorporating
active devices. These devices are modeled in a non-linear fashion.
We are working the third problem which requires more significant Apache code
changes, but at the same time, Redhawk-SDL leverages the core capabilities
for the product. They are the only folks who have a chance of solving this
problem anytime soon.
The Not So Good:
For us, the flow integration is just not there. We don't use voltage
scaling in our .libs, so even if we annotate instance specific voltage to
PrimeTime-SI, it has no impact (anyone got an SPDM library to spare?). Our
STA flow is SPEF into PT-SI, so Redhawk updated SDF output doesn't fly (and
I still don't trust SDF from bad experiences many years ago).
So far I have not been able to light the necessary fire under our design
teams to invest in the tool and realy examine their design more critically.
The Bad:
It is expensive. Yes, it may be less than a couple of metal masks in a
65 nm process, but it is a lot more than 'competing' products. If the
engineering community hasn't worked out how to fully utilize Redhawk-SDL,
you can imagine the EDA purchasing manager's reluctance. Enter Catch-22.
- Hugh Mair
Texas Instruments Dallas, TX
|
|
