Moving to Altera from Xilinx

I am interested to hear about experiences anyone may have had in
switching to Altera hardware and tools.

To be clear, the intent here isn't to start a "who's better" debate
but rather to understand what to expect from those who did switch,
regardless of the underlying issues.

In our particular case we'd be switching from V2 and V5 devices to
Arria and Stratix.  In general terms the applications involve image
processing.  I've seen Altera's approach to image processing IP and
tools and I have to say that I am impressed.  In contrast Xilinx's
image processing tools are more like a collection of app notes without
a real underlying structure.  The tools seem more user friendly...but
I don't have real experience with them yet.  I was surprised that
Altera discourages low-level design.  We've used low level
instantiation and explicit control of interconnect paths quite
effectively with Xilinx in order to squeeze performance out of modules
that needed it.  I suppose that there might be a point in trying to
stay high level in that designs have improved reusability and, if
chips are fast enough it really is better to avoid low-level work.

Thanks,

-Martin

On 15 Sep, 16:51, m <martin.use...@gmail.com> wrote:
> I am interested to hear about experiences anyone may have had in
> switching to Altera hardware and tools.
>
> To be clear, the intent here isn't to start a "who's better" debate
> but rather to understand what to expect from those who did switch,
> regardless of the underlying issues.
>
> In our particular case we'd be switching from V2 and V5 devices to
> Arria and Stratix. =A0In general terms the applications involve image
> processing. =A0I've seen Altera's approach to image processing IP and
> tools and I have to say that I am impressed. =A0In contrast Xilinx's
> image processing tools are more like a collection of app notes without
> a real underlying structure. =A0The tools seem more user friendly...but
> I don't have real experience with them yet. =A0I was surprised that
> Altera discourages low-level design. =A0We've used low level
> instantiation and explicit control of interconnect paths quite
> effectively with Xilinx in order to squeeze performance out of modules
> that needed it. =A0I suppose that there might be a point in trying to
> stay high level in that designs have improved reusability and, if
> chips are fast enough it really is better to avoid low-level work.

Certainly will help transitioning from X to A.

Jon

On Sep 15, 11:08=A0am, Jon Beniston <j...@beniston.com> wrote:
> On 15 Sep, 16:51, m <martin.use...@gmail.com> wrote:
>
>
>
> > I am interested to hear about experiences anyone may have had in
> > switching to Altera hardware and tools.
>
> > To be clear, the intent here isn't to start a "who's better" debate
> > but rather to understand what to expect from those who did switch,
> > regardless of the underlying issues.
>
> > In our particular case we'd be switching from V2 and V5 devices to
> > Arria and Stratix. =A0In general terms the applications involve image
> > processing. =A0I've seen Altera's approach to image processing IP and
> > tools and I have to say that I am impressed. =A0In contrast Xilinx's
> > image processing tools are more like a collection of app notes without
> > a real underlying structure. =A0The tools seem more user friendly...but
> > I don't have real experience with them yet. =A0I was surprised that
> > Altera discourages low-level design. =A0We've used low level
> > instantiation and explicit control of interconnect paths quite
> > effectively with Xilinx in order to squeeze performance out of modules
> > that needed it. =A0I suppose that there might be a point in trying to
> > stay high level in that designs have improved reusability and, if
> > chips are fast enough it really is better to avoid low-level work.
>
> Certainly will help transitioning from X to A.
>
> Jon

Our development group did a wholesale platform change from V2 Pro X to
Stratix II GX around 2-3 years ago, and it went very well.  The change
was mainly due to the fundamental issues that we were having with
basic Xilinx IP blocks not working, as well as issues w/ the Rocket IO
in the Virtex 2 Pro X devices.  What also helped is amount of local
support that we got from Altera (coupled w/ the lack of support from
Xilinx).  In fairness to our local FAE, he was very good, but the way
priorities shake out amongst companies being supported locally, we
weren't the highest on the totem pole.

Anyways, we have not had a single errata or IP-based issue since the
switch.  Everything has pretty much worked as advertised, and we have
been including plenty of the advanced DSP blockset modules in DSP
Builder / Simulink.  It has taken some time to "cut our teeth" on the
DSP tools, but we haven't come across anything that we couldn't figure
out ourselves.  We mainly use FIR filters and NCOs, so not much of the
image processing that you were asking about.  With respect to your
comment about low-level design, the DSP builder tool really encourages
coming up with a high-level DSP design, which is then instantiated in
your project w/ a single component (typically using a .tcl script).
What I have found is that, in DSP Builder versions < 8.0, I needed to
do some manual massaging of the models in order to hit timing (the
Matlab simulations would work great, but then fail timing miserably in
place/route).

From what I understand, the 8.0 version of DSP Builder is a lot better
than the previous releases (takes care of timing, optimizing using
multi-cycle, etc), but we haven't upgraded, since we're in the middle
of a project right now.  I definitely am looking forward to checking
out the new features, though.

One issue that I've been curious about is why these forums seem to be
dominated by issues people are having with Xilinx parts.  From what I
understand, the market share is something like 60/30 between (Xilinx
and Altera, respectively), with the other guys cleaning up the rest.
I'm wondering why I don't see that sort of general trend in the posts
on here.  Is it because people are going down other channels for
Altera support?  Are there just more issues w/ Xilinx?  Who knows.

Anyways, I just thought I would toss my $.02 worth.

Jeff

On Sep 15, 1:47 pm, jeffjcan...@gmail.com wrote:

> One issue that I've been curious about is why these forums seem to be
> dominated by issues people are having with Xilinx parts.  From what I
> understand, the market share is something like 60/30 between (Xilinx
> and Altera, respectively), with the other guys cleaning up the rest.
> I'm wondering why I don't see that sort of general trend in the posts
> on here.  Is it because people are going down other channels for
> Altera support?  Are there just more issues w/ Xilinx?  Who knows.

I think this might be as simple as Xilinx having more low-cost eval
boards on the market, so that hobbyist, student, and professional
development type projects are more likely to end up in their silicon,
and more likely to end up asking basic questions about it here.
Wheras in more professional usage problems that can't  be solved in
house might be more likely to end up with the FAE's than on the
newsgroups.

On Sep 15, 1:54=A0pm, cs_post...@hotmail.com wrote:
> I think this might be as simple as Xilinx having more low-cost eval
> boards on the market, so that hobbyist, student, and professional
> development type projects are more likely to end up in their silicon,
> and more likely to end up asking basic questions about it here.
> Wheras in more professional usage problems that can't =A0be solved in
> house might be more likely to end up with the FAE's than on the
> newsgroups.

That's possible, but I suspect also that to a large degree it's a
question of inertia. Comp.arch.fpga has always been X dominated,
whereas I suspect many A users frequent Altera's web Forums (for
better or worse).

Tommy

On Mon, 15 Sep 2008, cs_posting@hotmail.com wrote:

|-----------------------------------------------------------------------|
|"[..]                                                                  |
|                                                                       |
|[..]                                                                   |
|Wheras in more professional usage problems that can't  be solved in    |
|house might be more likely to end up with the FAE's than on the        |
|newsgroups."                                                           |
|-----------------------------------------------------------------------|

I had a simple problem created by Xilinx policy which could only be
resolved by a Xilinx employee (I do not control Xilinx's faulty
website therefore I can not fix it), which an idiotic Xilinx Field
Application Engineer did not manage to fix. Fortunately I started
another webcase on exactly the same complaint and was assigned a
different field application engineer, this time one who could
think. Unfortunately this incompetence of Xilinx's adversely delayed
me by weeks.

Colin Paul Gloster wrote:

> Fortunately I started
> another webcase on exactly the same complaint and was assigned a
> different field application engineer, this time one who could
> think. Unfortunately this incompetence of Xilinx's adversely delayed
> me by weeks.

Unfortunately, that sort of problem is not restricted to Xilinx.

I had a question to do with Quartus service packs and patches. I opened a
service request with Altera and got a 'read the doco' response. I
re-iterated the question, trying to explain the doco was ambiguous. The
exchange went back and forth several times - the "robot" on the other end
was clearly not interested in finding the answer for me - merely throwing
more links at me to try to find the answer myself.

Eventually I mentioned it in passing to a local FAE, and he found the
answer for me almost immediately.

I think Xilinx/Altera farm out the web support to the same people that
handle telephone contract cold-calling and credit card enquiries.

I can just imagine some student/backpacker with a phone on his/her desk
and three lights on it...

* FPGA support
* Credit Card Query
* Phone bill query

:(

Regards,

-- 
Mark McDougall, Engineer
Virtual Logic Pty Ltd, <http://www.vl.com.au>
21-25 King St, Rockdale, 2216
Ph: +612-9599-3255 Fax: +612-9599-3266

agreed 100%

http://www.callcentermovie.com/

On Sep 18, 11:00=A0am, Jochen <JFren...@harmanbecker.com> wrote:
> agreed 100%
>
> http://www.callcentermovie.com/

R O F L !
I especially like the part where the call center guys had to call tech
support.

-Dave Pollum

Jochen wrote:

> http://www.callcentermovie.com/

LOL! Yeah, that about nails it! :)

Regards,

-- 
Mark McDougall, Engineer
Virtual Logic Pty Ltd, <http://www.vl.com.au>
21-25 King St, Rockdale, 2216
Ph: +612-9599-3255 Fax: +612-9599-3266