Custom Search
Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
In article <gogoEpB6EG.IzL@netcom.com>, R. Mark Gogolewski <gogo@netcom.com> wrote: > >BTW, I _love_ the idea of Linux. However, if I ran a pure >Solaris group, the only $$ I would really save would be on hardware. >The EDA software won't get cheaper on Linux. The admin support >won't get cheaper, etc. , etc. In fact, I'll have to support >both for awhile when switching, so my admin costs go up. > I too want linux versions of the apps I need at work - viewlogic viewdraw, VCS, altera and xilinx tools, timing designer, etc, etc... I suspect that part of the problem is that EDA companies want to charge UNIX CAD prices for Linux software, rather than NT CAD software prices which seem to be much less (for less functionality in most cases). -- | Andrew Dyer <adyer@midway.com> or <adyer@mcs.net> | | Sr. Design Engineer (773) 961-1751 | | Midway Games, Inc. (773) 961-1890 (fax) | | 2727 W. Roscoe Ave., Chicago, IL 60618 |Article: 9251
I'm not too sure whether we represent a "typical" design house here, but
I can tell you one thing: our hardware budget every year is significantly
smaller than our CAD software budget. And don't get me wrong here: we're
not stuck for hardware. We've get new 8-, 16- and more processor
UltraSparc Servers to burn every quarter, with more memory and disk than
you can shake a stick at.
The point I'm trying to make here is that:
1) Linux is not necessarily cheaper to maintain than Solaris.
They're both Unix's at the end of the day. Unless you have a lot
of them, in which case it's nice to be able to use Solstice network
administration software, but something comparable probably also
exists for Linux.
2) CAD software will be expensive anyway.
3) Hardware is still relatively cheap (I know of one CAD company that
was giving away a free SS5 with every CAD license sold.)
To add to this you have the reasons not to run CAD software on a
Linux box:
1) How is Linux's MP support these days? I haven't checked the source
tree in a while, so I really don't know. We actually use our MP
UltraServers for heavy-duty DRC checking, etc. I'd prefer to give
the entire UltraSparc-III layout database to an 8-processor machine
than to a 1-processor machine, since I'd like to see it checked in
my lifetime :-) Likewise for other heavy-duty CAD software that we
(and I'm sure a LOT of others) use. It's not just Linux at fault
here - PC MP hardware is in it's infancy.
2) How much memory can you stick in a PC? Most machines here at SME
have at least 512MB of memory in them. Most of our processor ranch
machines are configured with a lot more than this. The machines
we use for large hspice, drc, rce, etc. runs have 4GB+ in there.
Not many PC's can handle this king of configuration unless you pay
a premium for it.
3) I know there are houses out there than run many different OSes on
different hardware platforms. We're not really very typical in
that regard - we run Sun hardware with Solaris everywhere - there
are no other machines used in SME. This makes our life simpler,
to some extent. I'd hate to throw another architecture/os into
the mix, and I'm sure most houses would feel the same way.
Running Linux on some other hardware might aleviate some of the above
issues, but this seems to me to be a "I'm doing it because I can" kind
of thing - I can see no reason to run Linux on a Sparc, say, since you
pretty much can't buy a Sparc without Solaris pre-installed anyway.
Likewise for Alpha, etc.
I actually like Linux a lot - I may work for Sun, but that doesn't bias
me against the Linux in anyway. I use it at home. I also use Solaris
and Win95 at home - different boxes for different tasks. I see a place
for both Unix's, but if I was to form a design house startup right now,
I'd be very wary about running whatever CAD software that exists for
Linux - I'd rather stick to a Sparc box.
Of course, give the option of Linux v. NT, I don't think I need tell
you which one I'd rather have...
Regards,
Robert.
--
Robert Walsh, Email: rjwalsh@eng.sun.com
Sun Microelectronics, Tel: +1-408-774-8116
901 San Antonio Road, MS USUN02-301, Fax: +1-408-774-8680
Palo Alto, CA 94303, USA. (PGP Key on Request)
Article: 9252In article <c7m6acpej.fsf@ite127.inf.tu-dresden.de>, Achim Gratz <gratz@ite.inf.tu-dresden.de> writes > >Is there such a thing as a programmable analog switch matrix/crossbar? >I find digital ones (Lattice, Aptix FPIC), analog muxes and complete >switch boxes for measurement equipment, some for analog mixers, but >both extremely expensive. I'd like to use these with a Zetex TRAC >device for lab exercises, so fmax=4MHz, crosstalk -60dB, Ron >uncritical if tightly distributed. Before I forget, programmable >resitors/capacitors would be a boon, too although I could use analog >muxes for these. > > >Achim Gratz. > AD75019 from Analog Devices is a remarkable 16x16 analog crosspoint switch comprising a matrix of 256 switches. It's serially configured with a simple protocol and runs from +/-12V supplies, switching signals of nearly rail to rail amplitude. I don't know the frequency or crosstalk specs offhand. Interestingly, the serial configuration bitstream has a minimum bit rate specification. We use a PIC16C84 to control ours, providing a parallel interface to the host system while keeping a copy of the configuration for readback. The device we use is (IIRC) in a 44 pin plcc and costs around $30 for one. Hope this helps. -- Keith WoottenArticle: 9253
paul: : I'd like to find the file format for the symbols created in Viewdraw. : These are ASCII files, and I've already done some reverse-engineering : and found some of the info. However, as long as this format has been : around, I'd be surprised if the information isn't already available. rk: and i'd like to have a good definition of the sch and wir files too. i'm writing a program to do some automagic and semi-automagic editing of a design. paul: : I've put in a request to Viewlogic, but haven't gotten a response yet. rk: i have received the silence, too. -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9254
andrew: : I suspect that part of the problem is that EDA companies want to charge : UNIX CAD prices for Linux software, rather than NT CAD software prices : which seem to be much less (for less functionality in most cases). rk: thought i just read in ee times that synopsys will charge same $ for unix and NT (marvelous). also said they would not do linux, no customer demand. -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9255
Furthermore, why is there an expectation that NT software should be cheaper than UNIX? And why, if Linux is essentially UNIX, would Linux be cheaper? It's the same functionality/code, right? Mark In article <01bd47c5$c95a2720$6e84accf@homepc>, rk <stellare@erols.com.NOSPAM> wrote: >andrew: >: I suspect that part of the problem is that EDA companies want to charge >: UNIX CAD prices for Linux software, rather than NT CAD software prices >: which seem to be much less (for less functionality in most cases). > >rk: >thought i just read in ee times that synopsys will charge same $ for unix >and NT (marvelous). also said they would not do linux, no customer demand. > >-------------------------------------------------------------- >rk > >"there's nothing like real data to screw up a great theory" >- me (modified from original, slightly more colorful version) >--------------------------------------------------------------Article: 9256
Reno, Thanks for the posts. I know there are quite a few attributes one can attach to nets, and to blocks, etc. My basic concern is whether XACT6 actually supports *any* of the net attributes. I have only ever used the L and SC= on nets. I have also used pin numbers, and they certainly work; IIRC a constraint file gets generated which contains the pin numbers. I have a good reason to think that both L and SC don't work. Next time I do a design, I will run XACT.EXE and have a look at the layout. That's always a pig of a job. >I think I should say it a bit more precisely (referring to my last mail): > >using vertical longlines: >Constrain all critical components (CLBs) in one column and lock the critical >net to a fixed pin of these CLBs with either the 'P' attribute or the >'MAP=PLO' resp. 'MAP=PLC' attribute combined with the use of CLBMAP mapping >symbol. The disadvantage is the effort to map and constrain your design (or >parts of it). But for best results you should always constrain 3K designs. >If you fix critical nets to B or C inputs of CLBMAP you have the biggest >chance to persuade PPR/APR to use a lonline (more resources are available in >3KA ICs). If PPR/APR doesn't it's now easy to correct the result in XDE. > >using horizontal longline: >Use a 'dummy'-TBUF (WAND) or feed your critical net through a TBUF (need the >'X' attribute) and PPR has to use a longline. If the extra delay is a >problem you can remove the TBUF in XDE. > >you see XDE is a nice tool. Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.Article: 9257
And *I* thought I was the only one on earth still running Viewdraw 4.1... :) I never tried win95 but can confirm Viewlogic 4 doesn't run under NT, as fully expected given the dongle, the DOS extender, and the video modes. >I am using rather old version of VL as a front end for XACT6. >Does anybody know how to make VL 4.1.3 running under DOS7 (from W95)? >Works fine with DOS5.0 and QEMM6.0 as memory manager. >I spent a lot of time testing different configurations but I failed. >There is no (or I can't find) info about it on ViewLogic home page. >Any help will be appreciated. Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.Article: 9258
On Sat, 28 Feb 1998 13:24:07 +0000, Jay Darmon <jdarmon@worldlink.nsp.net> wrote: >* The decision by Netscape to release the source code for Communicator 5.0 >has brought a lot of attention to Linux as well. See: >http://www.mozilla.org/ > >So far, though, the EDA market has been a tough one for Linux to move >into. It seems that the engineers (the users of EDA apps ) are quite >willing to use Linux, but the EDA companies want to push NT. The case for >EDA applications on Linux is compelling: > >4) Stability. Linux typically has uptimes measured in months while NT >crashes about twice a week (in my experience, and in the experience of >others I've talked to). > >5) Performance. When NT is doing any kind of disk access it seems to be >very unresponsive. [snip] Although what you state is a step in the right direction for free software, I have to point out that even Infoworld and the CAD/CAM community have severe tunnel vision. People seem to use free software and Linux interchangeably. Unfortunately, this is a little shortsighted since there are other efforts in the free software community going on, namely FreeBSD. As a user of both operating systems, I can testify that in a networked environment, you will get much better performance and stability out of FreeBSD than your average Linux distribution. Linux is the rage since software development is going on at a frenetic pace. Bugs are squashed at lightning speed. However, bugs are created in a similar fashion as a result of the Linux development model. As a CAD tool user, you want to be able to install an operating system and not mess with it. I'm not sure if managers would be excited about playing the kernel patch game to get the functionality they need. I realize that I'm being slightly unfair to Linux, but this is not that far off the mark. As a testimonial to the stability of FreeBSD, it is currently the operating system of choice for one of the largest FTP repositories on the Internet ftp.cdrom.com (wcarchive.cdrom.com) housing 1G of memory and 224GB of disk running on a P6/233. I'm pointing this out since Linux will be the OS that probably makes or breaks the free software reputation in the CAD industry. If I had to put my money down on an OS based on your bullets above, I'd bet on FreeBSD. To find out more about FreeBSD, I suggest you point your browser at http://www.freebsd.org. -ClintArticle: 9259
andrew: : >: I suspect that part of the problem is that EDA companies want to charge : >: UNIX CAD prices for Linux software, rather than NT CAD software prices : >: which seem to be much less (for less functionality in most cases). rk: : >thought i just read in ee times that synopsys will charge same $ for unix : >and NT (marvelous). also said they would not do linux, no customer demand. R. Mark: : Furthermore, why is there an expectation that NT software should be : cheaper than UNIX? And why, if Linux is essentially UNIX, would : Linux be cheaper? : : It's the same functionality/code, right? rk: unix versions of software have traditionally cost far more than the dos/win versions. yes, same code and all (in principle). but it is a different market. and, i believe, it will continue to be a different market. the unix boxes will probably stay ahead in performance of the win/intel boxes, although perhaps that may change in the future, who knows? in a recent benchmark done by integrated system design, march '98, showed that for verilog-xl, "in most cases, the sun workstation outperformed the pcs. from a cost/performance perspective, though, the PCs generally won. still, if you need the performance, the choice is clear." again, my opinion, chime in your $0.02, i think the larger and more demanding asics will require the fastest machines they can get. and for engineers designing those sorts of chips, the cost of the hardware platform and the expensive software for the design/analysis/simulation of state of the art chips relative to the cost of runs of the chips themselves, will not be the driving factor in tool selection. for a LARGE chip, simulation run times will be critical adding costs. and at the price per gate i've seen quoted, it's an expensive business. they will pay to stay on the leading edge, have the fastest and best software, and software companies will charge to keep them there. now, on the pc side, you will see the small to medium sized developments. these include people doing fpgas where the nre per gate is essentially zero. and people who will transfer their fpgas to asics for production. and you will see people designing this stuff in small businesses, where their model is investing only a moderate amount in tool sets. like my business, real small, just me! this is not the market for the best, most expensive tools. most people can get into business for perhaps $10,000. a fast pc, schematic entry package, say viewlogic, some backend tools, and an inexpensive or vendor supplied vhdl compiler. throw on a small laser jet and you're ready to go. ok, this stuff isn't as good as the latest synopsys tool, or ambit tool, and doesn't have super-duper networking and clustering (although i have a 100 mbps network to the wife's laptop ;) but here better is the enemy of good enough. and this market will never buy $90,000 tool sets in volume. however, there are a lot of people who can do this sort of work and vendors can sell moderate volumes of s/w at reduced prices and make more $. that's why it's expected. this will apply to regular businesses too. for example, at day job, i work for a "large company" of over 2,000,000 employees (not kidding U.S. Government). and i have bought the expensive tools but those purchases are few and far between. but inexpensive s/w for windows, not a problem, we can put a copy on everyone's desk (or get enough licenses or whatever). this has been the traditional model for stuff for at least the last 15 years or so. in eda, for example, during the '80s, people were buying apollo/mentor or their competitors. for the time, quite sophisticated, for the s/w and the os (apollo had some nice stuff). we were all well networked, still better than windows is today. and it cost about $35,000 a seat or so. very expensive. and then along came orcad at $500. simple. straightforward. and we could put a copy on everyone's desk for board design. i seem to remember that mentor was getting into the pc software for a bit, didn't really follow it. perhaps someone can remember what they did, how they did it, and how it worked out. -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9260
In article <gogoEpB6EG.IzL@netcom.com>, R. Mark Gogolewski <gogo@netcom.com> wrote: >Rick, > >Great points. > >You are very correct on the following two things: > > [] A port from any other Unix OS to Linux is essentially cake. Yep. > > [] Unix houses can very easily switch to Linux while NT houses > would have more difficulty. > Also true. >BTW, I _love_ the idea of Linux. However, if I ran a pure >Solaris group, the only $$ I would really save would be on hardware. >The EDA software won't get cheaper on Linux. The admin support >won't get cheaper, etc. , etc. In fact, I'll have to support >both for awhile when switching, so my admin costs go up. > Not really since there is a lot of overlap in the admin of Solaris and Linux - they're very similar. The Solaris admin's could easily pick up Linux admin, in fact they probably run Linux at home. NT and Solaris are very different though. philArticle: 9261
In article <01bd47e4$f45b9480$3280accf@homepc>, rk <stellare@erols.com.NOSPAM> wrote: >rk: >unix versions of software have traditionally cost far more than the dos/win >versions. yes, same code and all (in principle). but it is a different >market. and, i believe, it will continue to be a different market. <snip> Ahhh.... I completely agree here that there are certainly two major design markets: high-end (needing the best tools), and smaller companies/startups/groups/consultants who are less interested in having the very, very best (or really in spending the very, very most...). >From a software standpoint, I think the line you are drawing is not really the line between PCs and UNIX machines, but between differing goals in purchasing for the overall design environment. There is certainly a correlation given that the high-end market will pay the $$ for the best machines - and the best tools - given that the time spent running the tools must be kept to a min. The other portion of the market, in looking for a low-cost alternative, will actively pursue PCs and low-cost software as well. It seems that large companies, having a need for both types of projects, are following both approaches and use both NT and UNIX. Now, back to what I said before... >R. Mark: >: Furthermore, why is there an expectation that NT software should be >: cheaper than UNIX? And why, if Linux is essentially UNIX, would >: Linux be cheaper? >: >: It's the same functionality/code, right? I still believe that there should not be a price difference in the exact software running on two different OSs. However, it does seem that it is wise to offer different software packages for high-end (primary, but not limited to, UNIX) and low-end (primarily NT and desirous of Linux). Given all that, what we're we talking about? ;) MarkArticle: 9262
Is there any relatively bug-free design software out there? I've tried Xilinx Foundations and Altera Max Plus II, and have found both of them to have incredibly annoying bugs. Most of the worst bugs are not those that occur when my own designs are error-free. Rather, they are bugs that occur when my own designs have bugs, or when they don't obey some arcane rules. For example, in Foundations, if I name a state diagram file using more than 8 characters, I'd get a weird result instead a warning about the file name.Article: 9263
gogo@netcom.com (R. Mark Gogolewski) writes: > You are very correct on the following two things: > > [] A port from any other Unix OS to Linux is essentially cake. > > [] Unix houses can very easily switch to Linux while NT houses > would have more difficulty. > > BTW, I _love_ the idea of Linux. However, if I ran a pure > Solaris group, the only $$ I would really save would be on hardware. > The EDA software won't get cheaper on Linux. The admin support > won't get cheaper, etc. , etc. In fact, I'll have to support > both for awhile when switching, so my admin costs go up. We, essentially, are running a Solaris EDA shop (We also run NT, but that's mostly the Software guys). Simulations, Synthesis, etc are running on Ultra's. However, we are getting more and more Linux boxen, which are generally used as 'personal workstations' and fast X terminals. As for the admin support cost, we generally find it easier to admin a Linux box than a Solaris box, but that's related to previous Unix experience. I'd love to have DAI's Signalscan (and similar tools) running locally on the Linux box. Simulation is not as interesting, given that you need to beef up all the Linux boxen with lots 'o RAM. Kai -- Kai Harrekilde-Petersen <khp@dolphinics.no> #include <std/disclaimer.h> http://www.dolphinics.no/~khp/Article: 9264
I have no extra info beyond what you can work out by looking at the
files. However, do please post your final results - ESPECIALLY if you
can work out how the magic number is calculated.
This magic number is used to stop things like e.g. using a Xilinx/LCA
version of Viewdraw to open a sch or lib file which was created in the
unrestricted version.
I believe that unless you can generate the magic number, you will not
be able to generate valid Viewlogic files.
The sch file starts with something like
V 50
K 257863394400 FRED
Y 0
D 0 0 1653 1169
Where the 2nd line, after the "K", has the magic number. This, I
believe, is computed with an algorithm which takes in the filename
("FRED" above), a text string identifying the company selling the
restricted version (e.g. "Xilinx") and some other stuff I don't know.
I doubt Viewlogic will tell you, since they went to a lot of trouble
to prevent the restricted versions (sold for much less than their own
full version - $30k at one time) being used to design normal stuff,
with general libraries like LSTTL.
>I know this is a bit off-topic for this group, but I know there are
>many of you who use Viewlogic Viewdraw for schematic-based FPGA designs.
>
>I'd like to find the file format for the symbols created in Viewdraw.
>These are ASCII files, and I've already done some reverse-engineering
>and found some of the info. However, as long as this format has been
>around, I'd be surprised if the information isn't already available.
>
>I've put in a request to Viewlogic, but haven't gotten a response yet.
>Also, does anyone know of any newsgroups such as this which are focused
>on schematic capture, or EE-directed EDA tools in general.
>
>I've done some searching using AltaVista, but no cigar so far.
Peter.
Return address is invalid to help stop junk mail.
E-mail replies to zX80@digiYserve.com but
remove the X and the Y.
Article: 9265>>4) Stability. Linux typically has uptimes measured in months while NT >>crashes about twice a week (in my experience, and in the experience of >>others I've talked to). This is simply not true. I am the last one to defend MS but NT is very reliable. I use it all day. If you find regular crashes, as some people indeed do, you very probably have hardware problems, or you need a decent UPS. >>5) Performance. When NT is doing any kind of disk access it seems to be >>very unresponsive. Not true. There are lots of stupid things in NT but none of them make it in any way unsuitable for EDA. Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.Article: 9266
Hi, Are you looking for information about the Texas Instruments TMS320C6x digital signal processors? Please check out my website: http://www.scs.ch/~andrew/c6x.html Here you'll find the latest documentation and silicon availability information. There is plenty of stuff about doing hardware and software design with these processors, some application notes and a comprehensive bug list. Also, stacks of info about commercially available 'C6x processor boards and lots of other stuff ..... Have a look and please send me any comments. Don't forget to join my mailing list if you want to be notified when the site is updated ... Cheers, Andrew Phillips Supercomputing Systems AG Zurich, SwitzerlandArticle: 9267
> The Questions are:- > > 1) What file name should I use to contain the listing given above? Should > it be my_pkg.vhd, comp1.vhd, comp2.vhd or something else? There is no relation between the filename and the VHDL object(s) in a filename. So, a very bad idea could be to put everything in one file. Best is to have a "convention" and to split in multiple files. E.g.: - entities are stored in <entity_name>_ENT.vhd, - architectures are stored in <entity_name>_<architecture_name>.vhd - package definitions in <package_name>_PCK.vhd - package body in <package_name>_body.vhd - configurations in CFG_<entity_name>_<architecture_name>.vhd Or some other convention, as long as you have a convention, it's ok. > > 2) In my main program (in a separate file), how does it know where to > find the component definitions defined in the above listing? Look at the library concept in the LRM. You can analyze your entity/architectures in a library, e.g. MY_LIB. Then in your code you can access objects from that library with the - library MY_LIB; - use MY_LIB.MY_PKG.all; (Note that in your code you have "use IEEE; use IEEE.std..." ! The first of these two should be "library IEEE; use IEEE.std...") The library has to be bound to a certain directory, for most simulators. Look in the user's manual of your simulator. > > Thanks for any help and comments. > > C. Fung -- =================================================================== Jan Zegers === Easics === General Manager === VHDL-based ASIC design services === NEW Tel: +32-16-395 601 =================================== NEW Fax: +32-16-395 619 Interleuvenlaan 86, B-3001 Leuven, BELGIUM mailto:janz@easics.be http://www.easics.comArticle: 9268
rk: : >unix versions of software have traditionally cost far more than the dos/win : >versions. yes, same code and all (in principle). but it is a different : >market. and, i believe, it will continue to be a different market. : : <snip> mark: : Ahhh.... I completely agree here that there are certainly two : major design markets: high-end (needing the best tools), and smaller : companies/startups/groups/consultants who are less interested in : having the very, very best (or really in spending the very, very : most...). : : From a software standpoint, I think the line you are drawing is not : really the line between PCs and UNIX machines, but between differing : goals in purchasing for the overall design environment. There is : certainly a correlation given that the high-end market will pay the : $$ for the best machines - and the best tools - given that the time : spent running the tools must be kept to a min. The other portion of : the market, in looking for a low-cost alternative, will actively : pursue PCs and low-cost software as well. rk: agreed. but today, the unix s/w is generally more expensive. and the pc s/w is generally cheaper. and, the high end tools tend to run on unix. and the low end tools tend to run on pc. but the goals are, as you say, the key, not the OS. but with the volume of win machines in general as compared to unix boxes, it's simple economics that the $5k package can make more $ than the $90k if they both can do the job. and here do the job means process in a reasonable amount of time with a reasonable amount of hassle. and, actually getting back to my last msg, the small to medium sized designs are not as sensitive to performance as the million gate asic. mark: : It seems that large companies, having a need for both types of : projects, are following both approaches and use both NT and UNIX. rk: yup, at day job, that's what i got. admittedly, most of the pc's in our design group are running '95, with a few NT machines, mostly for easing the transition. the NT machines require about twice as much ram and wouldn't just upgrade on older dos/win 3.11 machines. also, the programming of custom h/w on the backplane for our home-brew add-on cards was more hassle in NT than '95. however, for the next upgrade, we'll probably all switch to NT since with memory cheap we all bought a lot and we have the software ready to program NT, including our custom h/w. mark: : Now, back to what I said before... : : >R. Mark: : >: Furthermore, why is there an expectation that NT software should be : >: cheaper than UNIX? And why, if Linux is essentially UNIX, would : >: Linux be cheaper? : >: : >: It's the same functionality/code, right? : : I still believe that there should not be a price difference in : the exact software running on two different OSs. However, it does : seem that it is wise to offer different software packages for : high-end (primary, but not limited to, UNIX) and low-end (primarily NT : and desirous of Linux). rk: in principle, yes, i agree. same s/w guys. they don't get paid more for unix coding than win programming [perhaps they should ;) ]. but, as i said earlier, marketing generally forces the prices down for windows. and that generally has been the case. perhaps someone with some extra time can look at eda and other s/w applications written for both win and unix and see what the differences are. and yes, i was surprised that synopsys will have identical pricing. especially with the competition on the win machines, with packages from synplicity, for example, that seem to be pretty good and run pretty fast. and, with vendor supplied tools being either free or inexpensive. for example, actel will give you a free vhdl synthesizer. it works. it's not synopsys but it works. and you don't have to go into the boss and ask for a big pile of $. cypress/quicklogic was selling theirs for $99. what does synopsys cost? $90,000 per license? synplicity is ~ $20,000 or so? it'll be interesting to see how it all works out. of course, i would like to be platform/OS independent and run the same tools for different size jobs. s/w learning curves are expensive to pay for. viewlogic and actel, for example, run on both unix and win '95/NT, and it makes it practical for designers working on either platform/OS to communicate. over to you, chet! -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) -------------------------------------------------------------- : : Given all that, what we're we talking about? ;) : : Mark : :Article: 9269
someone said: : >>4) Stability. Linux typically has uptimes measured in months while NT : >>crashes about twice a week (in my experience, and in the experience of : >>others I've talked to). peter: : This is simply not true. I am the last one to defend MS but NT is very : reliable. I use it all day. If you find regular crashes, as some : people indeed do, you very probably have hardware problems, or you : need a decent UPS. rk: at day job, i'm running two NT machines. basically never shut them off and rarely do they get in trouble. however, bad software can crash them, it's NOT crashproof, but that's another story. someone said: : >>5) Performance. When NT is doing any kind of disk access it seems to be : >>very unresponsive. peter: : Not true. rk: we haven't witnessed this problem. in fact, our #1 test engineer (day job) remarked positively how fast the NT machine seemed to run. was using it for real-time data acquisition and storing to disk (and we considered just adding more memory and storing to ram but having the disk grind away wasn't an observed problem). -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9270
It gives me great pleasure to announce that I now have no ties to any programmable logic vendor. As some of you will recall, I have been working for a Lucent distributor, Eurodis Bytech, but following my HDL 'convictions', I have (this week) changed jobs to work for the HDL solutions company, Saros Technology. Hopefully my views on programmable logic can now be treated with a slightly less sceptical view than they may have been in the past. ;-) Naturally, I look forward to working with all vendors, and their UK distributors in the provision of HDL solutions for ASIC and PLD to their customers, both present, and future. www.saros.co.uk Modelsim, Exemplar Leonardo/Galileo, Turbowriter, IP cores, TransEDA, etc. Stuart For those who don't give a stuff, sorry for the time-waste.Article: 9271
On Tue, 03 Mar 1998 12:28:12 -0800, Peter Alfke <peter.alfke@xilinx.com> wrote: >Don't believe what Marketing says about its competitor. You can be sure >that the facts have been massaged. There are hundreds of ways of >manipulating the truth without necessary creating outright lies, >although even that happens. ( Would you be so naive to believe what >Ford says about Chevrolet, or Toyota about Honda, or McDonald about >Burger King or vice versa ? ) Amen to that, but a sticky wicket approaches... >Be careful when you evaluate "equivalent" devices. One observer's >equivalence is another one's big difference. You could always just cut to the chase and count the number of 4ip LUTs. Or maybe the number of registers? Imperfect, but better than the marketing "specmanship" that has been going on recently. Let's try it shall we?.... Xilinx XC4062 - 4608 4ip LUTs, and a total of 5376 registers, of which 768 are internal. I/O is 384 maximum. Altera FLEX10K100A - 4992 4ip LUTs, and a total of 5398 registers, of which 406 are in the I/O blocks. I/O is 406 maximum. Similar devices yes? Courtesy of http://www.arrowsemi.com/ EPF10K100ARC2403 is $165 for 100-499 quantity. EPF10K100ARC2401 is $335 for 100-499 quantity. Courtesy of http://www.marshall.com/ XC4062XL-3HQ240 is $472 for 100+ quantity. XC4062XL-09HQ304 is $1218 for 100+ quantity. Trying smaller (5V this time) devices again such as the 10K20, 4013E, and OR2C12A: EPF10K20RC2084 is $42.50 for 100-499 quantity. EPF10K20RC2083 is $63.50 for 100-499 quantity. OR2C12A2S208DB is $51.50 for 26-100 quantity. 100+ not listed. OR2C12A4S208DB is $74.00 for 26-100 quantity. 100+ not listed. XC4013E-4PQ208C is $110 for 100+ quantity. XC4013E-1PQ208C is $270 for 100+ quantity. if I let an XL creep in : XC4013XL-3PQ208C is $64.60 for 100+ quantity. XC4013XL-1PQ208C is $113 for 100+ quantity. I tried for Spartan : XCS30-3PQ208C is $57.45 for 100+ quantity. Oh, and a EPF6016 is $22.95 in 100-599 quantity. Now, I know there is some speed grade difference, and your mileage may vary, but there is, shall we say, a "difference". >Don't equate chip size with cost. There are many other factors affecting >cost, some of them technology-oriented, some not. Cost is what matters, >not square microns. Agree. However, when a company runs on NET margins of 60%+, the product must be costing them buttons to manufacture. Perhaps some people are just prepared to make a lower margin? I wonder if the reverse applies where Xilinx is attacking the CPLD marketspace? Now if only people published revenue by density and product family... >Don't compare devices on the basis of today's single-quantity price, but >don't blindly accept high-volume futures either. Agree, although 100 up generally gives a reasonable finger in the air in distribution land. StuartArticle: 9272
On Thu, 05 Mar 1998 12:15:14 GMT, s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb) wrote: >Xilinx XC4062 - 4608 4ip LUTs, and a total of 5376 registers, of which >768 are internal. I/O is 384 maximum. Duh, I meant 768 are in the I/O blocks. StuartArticle: 9273
>Modelsim, Exemplar Leonardo/Galileo, Turbowriter, IP cores, TransEDA, >etc. Have fun. All nice expensive stuff :) Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.Article: 9274
>at day job, i'm running two NT machines. basically never shut them off and >rarely do they get in trouble. however, bad software can crash them, it's >NOT crashproof, but that's another story. It is not crashproof, of course. It is easily crashed when there are hardware-related problems. It is very hard to crash it with most normal windoze apps though. Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.
Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z