Do you like Virtex-5 ?

Tim <tim@nooospam.roockyloogic.com> writes:

>If you think I'm exaggerating, try the Avnet site:
<...>
>Spartan-3E parts: prototype parts only and no higher volume pricing

And funny enough, Avnet gave us a first price estimation for quantity 1000 of an
1600E higher than the prototype price in their webshop. I don't know what they
are smoking...

-- 
         Georg Acher, acher@in.tum.de
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias

George Acher posted on Fri, 16 Feb 2007 13:12:14 +0000 (UTC):

"And funny enough, Avnet gave us a first price estimation for quantity 1000 of a
n
1600E higher than the prototype price in their webshop. I don't know what they
are smoking..."

I checked the price of a component (not an F.P.G.A. though) we needed
on a few websites in December 2006. One of them was Avnet's and
Avnet's website had the engineering (prototype, commercial quality)
model priced at $11122.80 whereas Avnet's website had the radiation
hardened space qualified (better quality) version priced much less at
$3738. I performed this search due to miscommunication: we had
actually already bought the radiation hardened version before December
for far less than $5000 (though I do not know for how much exactly,
nor from whom).

Jim Granville posted on Fri, 16 Feb 2007 13:56:22 +1300:
          
"Peter Alfke wrote:

> Do you like Virtex-5 ? Then please vote for it...
>
> The editors of Electronic Design News think Virtex-5 is an innovative
> product;
> they have nominated for their 17th Annual EDN Innovation Awards:
>
>  the Xilinx Virtex-5 Design Team for "Innovator of the Year," and
>
>  the Virtex-5 LXT platform for "Innovation of the Year" (in the
> Digital ICs category)
>
[..]
  They also miss a [none of the above] vote, so readers cannot indicate
if they agree with the shortlists."

A very good point.

"Digital ICs, programmable logic, and memory:
  PEX 8548 PCI Express switch (PLX Technology)
  MR2A16A MRAM (Freescale)
  Virtex-5 LXT FPGAs (Xilinx)

and since the award is for innovation :

Adj.
1. innovative - ahead of the times;
2. innovative - being or producing something like nothing done or
experienced or created before;

then the clear winner (by a large margin) is the MRAM.

That's far closer to innovative than "another iteration in FPGAs" ?"

Nah, the concept of randomness was mentioned hundreds of years ago but
called "chaos".

It would be great to have Jim Granville as a reviewer for what passes
as research in Europe: I attended doctoral workshops here in which
many of the abstracts were written with the word "innovative" or
"novel" to describe something which is a minor adaptation of existing
work, and many of the abstracts were written as if it is amazing that
civilization even managed to survive before the brilliant Ph.D. candidate
managed to come up with something which will supposedly change the
world.

Sincerely,
Colin Paul Gloster

Tim wrote:

> Peter Alfke wrote:
>
>> Do you like Virtex-5 ? Then please vote for it...
>
>
> What do I do if I think Virtex-5 is really cool but no normal user can 
> take delivery of one?
>
> Most awards (e.g. the Oscars) are for products which are in meaningful 
> production. Which doesn't seem to be the case for Virtex-5, and for 
> Virtex-4, if distributor stockholdings are anything to go by.

I don't know what the pricing on the Virtex-5 is going to look like, but 
there's
a Virtex II Pro XC2VP100 chip in the Digi-Key catalog for $8400.  I'm sure
there is somebody, somewhere, who needs to miniaturize some system for an
exotic, incredibly space-constrained application that can use something like
this, or maybe the NSA needs it to crack codes or something.  But, I can't
possibly imagine a commercial product or even a research project that could
rationally select a part that expensive.  I also shudder to think of the 
CPU time
and memory it will take to compile the config for that thing!

We just bought a used minivan for slightly more that that price, and 
expect to
get a LOT more use out of it than an $8400 chip.

Jon

Peter Alfke wrote:

>On Feb 15, 5:06 pm, Tim <t...@nooospam.roockyloogic.com> wrote:
>  
>
>>Peter Alfke wrote:
>>    
>>
>>>Do you like Virtex-5 ? Then please vote for it...
>>>      
>>>
>>What do I do if I think Virtex-5 is really cool but no normal user can
>>take delivery of one?
>>
>>Most awards (e.g. the Oscars) are for products which are in meaningful
>>production. Which doesn't seem to be the case for Virtex-5, and for
>>Virtex-4, if distributor stockholdings are anything to go by.
>>    
>>
>
>Put in an order for 1000 pieces of any Virtex-5 LX or LXT (except
>perhaps for the biggest -330 parts)  with the ES suffix (early silicon
>=with errata sheet), and watch us ship from inventory.
>
>  
>
Oh, so I need to order 1000 pieces to get a prototype built?

Jon

Tim wrote:

> Peter Alfke wrote:
>
>>
>> Put in an order for 1000 pieces of any Virtex-5 LX or LXT (except
>> perhaps for the biggest -330 parts)  with the ES suffix (early silicon
>> =with errata sheet), and watch us ship from inventory.
>
>
> Not practical. There is no published pricing for these parts and 
> distribution refuses to give pricing without an order.
>
> To be blunt, Peter, you know that I have the highest regard for 
> Xilinx, but the way you conduct your sales business these days is very 
> disheartening. An unblinking concentration on the major accounts is 
> just fine, but for the great majority of users:
>
> - you cannot buy the Virtex-5
> - you cannot buy most of the Virtex-4 range
> - you cannot buy the Spartan-3E
> - you cannot buy the Spartan-3A
> - the Xilinx online store gives a new meaning to "store"
>
> If you think I'm exaggerating, try the Avnet site:
>
> Virtex-5 parts: all out of stock, no pricing
> Virtex-4 parts: out of 243 parts (mostly priced!) just 59 are available
> Spartan-3A parts: "Part is not found as a stocked item"
> Spartan-3E parts: prototype parts only and no higher volume pricing
>
This is a really disturbing trend.  For a while I thought this was a 
temporary disturbance
due to the RoHS debacle, but I think it is far more serious than that. 
 I think the entire
electronic industry in the US is in a death spiral.  There are just a 
RAFT of parts that I
can no longer get, that were stocked by several distributors just a year 
ago.  (TI, and
others, may be partially responsible for this by producing literally 
DOZENS of variants
on ordinary parts.  Some of the old CMOS 74HC parts are now available in 
5, 6 even
7 different packages, 3 temp ranges, and dozens of varying electrical 
and speed ranges.
So, a 74HC04 now expands to fill half a page of type that requires a 
microscope to
read in the digi-key catalog.  I have a small pile of the wrong size 
chips that I bought
by mistake because I can't remember the package suffix, but that's my 
fault.  Who can
possibly stock all these variants?

I can't count the times in the last year or two that I've bought 5 parts 
to make a prototype,
designed and debugged the board, and then had a mad scramble to try to 
obtain a small
production quantity of several of the parts.  Once I had Analog Devices 
quote me
120 weeks to obtain a certain part!  A frantic search turned up a TI 
pin-compatible
replacement at 1/6th the price!  But, it sure got my blood pressure up 
until I had
determined the substitute would actually work properly.

The distributors have warehouses full of non-RoHS compliant parts they 
need to get
rid of before they can fill them back up with more stuff they won't be 
able to move.
And, I have all these non-franchised distributors of questionable 
reliability constantly
calling me trying to sell me something!  I NEVER buy from them, but they 
KEEP
calling.

Jon

Colin Paul Gloster wrote:

> I checked the price of a component (not an F.P.G.A. though) we needed
> on a few websites in December 2006. One of them was Avnet's and
> Avnet's website had the engineering (prototype, commercial quality)
> model priced at $11122.80 whereas Avnet's website had the radiation
> hardened space qualified (better quality) version priced much less at
> $3738. I performed this search due to miscommunication: we had
> actually already bought the radiation hardened version before December
> for far less than $5000 (though I do not know for how much exactly,
> nor from whom).

Impressive. And six significant figures in the $11122.80 pricing! What 
sort of component? As I recall, the most expensive Xilinx FPGA is around 
$9800.

Jon Elson wrote:

> Tim wrote:
<snip>
>> If you think I'm exaggerating, try the Avnet site:
>>
>> Virtex-5 parts: all out of stock, no pricing
>> Virtex-4 parts: out of 243 parts (mostly priced!) just 59 are available
>> Spartan-3A parts: "Part is not found as a stocked item"
>> Spartan-3E parts: prototype parts only and no higher volume pricing
>>
> This is a really disturbing trend.  For a while I thought this was a 
> temporary disturbance
> due to the RoHS debacle, but I think it is far more serious than that. I 
> think the entire
> electronic industry in the US is in a death spiral.

It has certainly cause a step effect, that has moved more stuff on this
side to MOQ/Mfg leadtimes.

> There are just a 
> RAFT of parts that I
> can no longer get, that were stocked by several distributors just a year 
> ago.  (TI, and
> others, may be partially responsible for this by producing literally 
> DOZENS of variants
> on ordinary parts.  Some of the old CMOS 74HC parts are now available in 
> 5, 6 even
> 7 different packages, 3 temp ranges, and dozens of varying electrical 
> and speed ranges.
> So, a 74HC04 now expands to fill half a page of type that requires a 
> microscope to
> read in the digi-key catalog.
<snip>

Logic has suffered from some false starts.
Poor product management meant for a while, some vendors thought 5V
was going to die -
It took a while, but now most devices spec 1.65-5.5V logic
capability, and hot-plug is a common new feature too.

  There is no real excuse for temperature variants on logic (or 
SPLD/CPLD) these days, and most Logic
devices (eg Philips/NXP) are SINGLE part numbers/devices, with TWO
sets of data columns. One for 85'C and one for 125'C.
They do not bother with two speed grades - it's not like they have
signifcant yield issues these days!

  That makes strong sense, as it simplifies the production and stocking, 
and speeds design wins.

  Not all vendors are as up-to-date in their thinking, and some still 
see  an industrial temp device, as license for more margin.

  Since this started as a Xilinx thread, if we look at their OnLine 
"store",[CoolRunnerII example] they show ONLY commercial grade devices 
?! - where does that leave industrial customers ?
  Are Xilinx yields really so variable, that they need 3 part numbers for
each package-die variant ? - much more sensible to rationalise that down 
to 1, and greatly imrove the chances a designer will find it in stock!.

-jg

> And funny enough, Avnet gave us a first price estimation for quantity 1000 of an
> 1600E higher than the prototype price in their webshop. I don't know what they
> are smoking...

Don't blame Avnet, all pricing comes from Xilinx or the local rep.

The reality is that _if_ Xilinx thinks your product/project is
worthwhile you get samples,
small quantities for pilot production runs, and a good price for
volume.  If they don't,
then you get to twist in the wind.

To be fair, this is a pretty common industry wide practice and Xilinx
has always done very
well by me.

Since Peter and Austin want to self-promote (or guerilla market) their
way to Innovative
Product of Year I suggest they put their money where their mouths are
and put any
Virtex-5 part up on the web store for a little less than the usual
tourist pricing.  After
all, we've all heard how they've got _so_ many extra in stock they are
looking to unload.

If they do that, they've bought my vote.

So what about it Peter and Austin, man or mouse?

Ricky.

ricky,

Peter and I take this issue of pricing and availability quite seriously.

We have begun discussions on the issue.

If you think you are upset by this, try to place yourselves in the shoes 
of the hundreds of hardware ICDES folks, and hundreds of software folks, 
and hundreds of production test and characterization engineers...you get 
the picture.

I went to some of the other nominee's websites, and on one, I could: 
order a sample, order parts, have a representative contact me (the 
buttons I could click on).  10K pricing was listed.  I think there is 
significant room for improvement here.

Austin