On Nov 4, 11:40 am, austin <aus...@xilinx.com> wrote:
> All,
>
> I was puzzled that they used a push of the 65nm technology node (to
> 60nm).
>
> It is a huge investment for a FPGA device company to launch a new
> family, and to launch one with "the old" technology node means that
> the latest technology node is guaranteed to beat you on price, power,
> and/or performance (since 90nm, we do not get to choose all three, we
> are at best a two out of three for a new node).
That is a vastly overstated argument without *any* supporting
evidence. It has happened more than once that an FPGA vendor (I won't
mention any names that include an X) has pushed ahead with a new
process only to be burned by the problems it created. Maybe the new
process will see better performance (are parts still getting much
faster really?) or power (assuming the new process deals with the
problems it creates in this area), but price is highly dependent on
yield which is seldom good in a new process and only provides edible
fruit as it ripens. All this is on top of the issue of meeting an
announced schedule for availability of the parts. I remember at least
two generations of parts from Xilinx that were "shipping" except that
you couldn't get any for some six months. Altera tends to hold back a
bit longer until they are sure they will meet schedule and won't have
problems with the parts.
> True, S6 is optimized for power (first time we have ever used a low
> power process from a foundry), so bragging about performance is one
> way to shout very loudly "we sure burn a lot more power!"
>
> When we asked customers what their number one need was for S6, it was
> "lower the power!"
>
> I know that many like to use the latest Spartan node to replace the
> previous Virtex node (lower their bill of materials costs), but
> frankly, S6 was designed for a new markets, and not intended to
> cannibalize Virtex 5 sockets.
>
> Anyway, it is always fun to watch how these things play out,
>
> Austin
Yes, it makes for interesting reports even if not interesting parts.
Rick
Reply by ●November 7, 20092009-11-07
Michael <m.pont@rapiditty.co.uk> writes:
> different projects? Are people doing this switching, or are they
> finding their "ideal" FPGA provider and sticking with this?
I think the latter is quite common since many will fall into the trap
of using proprietary IP from a given vendor like NIOS/Microblaze,
Megawizard/Coregen components, specific IO macros, certain PLL
configurations, package options, programming solutions, etc.
Also there is some investment in learning the vendor provided tools,
even though if you know one it's easier to learn the another.
Being able to switch in the middle of the development cycle, or even
between projects requires a bit of planning, e.g. using a vendor
independent soft CPU and structuring your HDL so the vendor dependant
parts are in separate modules or use wrappers. This planning can of
course be an advantage if you at any point would migrate to lets say
an ASIC or a new FPGA vendor with some great features.
When you first start using FPGA's the differences mentioned above does
not seem obvious since there is so much new stuff to learn. You will
focus on getting your part up as quickly as possible and that
typically involves prototyping using a dev kit from a given vendor and
use as much ready IP (which is often vendor dependent) as possible.
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
Reply by Michael●November 7, 20092009-11-07
I've been eavesdropping on this interesting C4 / S6 discussion.
Anyone have any idea (and willing to share) how many developers are
chopping and choosing between Xilinx and Altera (and other FPGAs) for
different projects? Are people doing this switching, or are they
finding their "ideal" FPGA provider and sticking with this?
Michael.
Reply by Torfinn Ingolfsen●November 5, 20092009-11-05
Torfinn Ingolfsen wrote:
> Downloading as I write this.
> I run Ubuntu (which is not on the supported os list), but I'll try
> anyway and report how it goes.
Installation worked, with a couple of fixes:
- the installation script is using csh, which isn't installed on Xubuntu
(I'm usin Xubuntu 9.10) per default. I fixed it by installing tcsh (I
used synaptic).
- I used 'sudo quartus_free/install' after unpacking to run the install
script. I took all default values when installing.
> If it works in Ubuntu at all, I'll try it out to see how it works.
So far I have only tested that it (quartus) starts up, and that the
menus work etc.
It does. Cool!
--
Torfinn Ingolfsen,
Norway
Reply by Antti●November 5, 20092009-11-05
On Nov 5, 6:06=A0pm, austin <aus...@xilinx.com> wrote:
> Jon,
>
> Yes, TSMC is having some yield issues (at 45/40nm). =A0But, yield issues
> comes "with the territory." =A0It isn't the first time, and it won't be
> the last time. =A0In IC Design, the choice of technology was made three
> years ago, so yield was not a factor in the choice for C4.
>
> Ranting about a yield issue is not good for business, as you never
> know who might be next, and it isn't anyone at Altera's fault that
> TSMC has those issues (for the moment).
>
> Oh, I like that: =A0C4! Like the powerful explosive... =A0the FPGA
> guaranteed to make a bang!
>
> I digress. =A0Back to idle speculation.
>
> I believe that Altera was faced with our release of S6, both with and
> without transceivers, and realized they needed something in order to
> compete. =A0It did not take a rocket scientist to see where the low cost
> FPGA families would have transceivers in the future three years ago.
>
> Austin
I was wondering who brings out the "explosiveness" of the new
cyclones :)
well, for the moment Spartan-6 looks better than C4 as there is better
choice of packages
as Altera is no longer offering 8x8 mm packages, Xilinx had no 8x8 in
S3A what was a mistake
but Xilinx has re-introduced 8x8 packages in S6 again
sure, what is fun is the QFN148 packaged for Serdes enabled Cyclone, I
would love to see
S6 QFN68 with serdes :)
Antti
Reply by Torfinn Ingolfsen●November 5, 20092009-11-05
Ah! Thanks!
Downloading as I write this.
I run Ubuntu (which is not on the supported os list), but I'll try
anyway and report how it goes.
> I'm not working on FPGA development at the moment, but I like to keep
> track of what's new. If you try out the software, it would be nice to
> hear in this newsgroup how it goes.
If it works in Ubuntu at all, I'll try it out to see how it works.
--
Torfinn Ingolfsen,
Norway
Reply by David Brown●November 5, 20092009-11-05
Uwe Bonnes wrote:
> David Brown <david@westcontrol.removethisbit.com> wrote:
>> Torfinn Ingolfsen wrote:
>>> Antti wrote:
>>>> Hi
>>>>
>>>> Altera is promising 25% more fabric speed than S6 with their new
>>>> Cyclone IV
>>> And us hobbyist types are wondering: will Altera give us a _free_
>>> development environment for Linux this time?
>>> (just in case anyone from Altera drops in here occasionally).
>
>> There's a beta version available (I haven't tried it - I just noticed it
>> on their web site when wandering about).
>
> Any link (if you talk about a Linux version of Quartus II Web Edition Software)
>
> A search for linux "web edition" on the Altera site brings some hits, but I
> didn't see something downloadable...
>
> Thanks
Jon,
Yes, TSMC is having some yield issues (at 45/40nm). But, yield issues
comes "with the territory." It isn't the first time, and it won't be
the last time. In IC Design, the choice of technology was made three
years ago, so yield was not a factor in the choice for C4.
Ranting about a yield issue is not good for business, as you never
know who might be next, and it isn't anyone at Altera's fault that
TSMC has those issues (for the moment).
Oh, I like that: C4! Like the powerful explosive... the FPGA
guaranteed to make a bang!
I digress. Back to idle speculation.
I believe that Altera was faced with our release of S6, both with and
without transceivers, and realized they needed something in order to
compete. It did not take a rocket scientist to see where the low cost
FPGA families would have transceivers in the future three years ago.
Austin
Reply by Uwe Bonnes●November 5, 20092009-11-05
David Brown <david@westcontrol.removethisbit.com> wrote:
> Torfinn Ingolfsen wrote:
> > Antti wrote:
> >> Hi
> >>
> >> Altera is promising 25% more fabric speed than S6 with their new
> >> Cyclone IV
> >
> > And us hobbyist types are wondering: will Altera give us a _free_
> > development environment for Linux this time?
> > (just in case anyone from Altera drops in here occasionally).
> There's a beta version available (I haven't tried it - I just noticed it
> on their web site when wandering about).
Any link (if you talk about a Linux version of Quartus II Web Edition Software)
A search for linux "web edition" on the Altera site brings some hits, but I
didn't see something downloadable...
Thanks
--
Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de
Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt
--------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
Reply by David Brown●November 5, 20092009-11-05
Torfinn Ingolfsen wrote:
> Antti wrote:
>> Hi
>>
>> Altera is promising 25% more fabric speed than S6 with their new
>> Cyclone IV
>
> And us hobbyist types are wondering: will Altera give us a _free_
> development environment for Linux this time?
> (just in case anyone from Altera drops in here occasionally).
There's a beta version available (I haven't tried it - I just noticed it
on their web site when wandering about).