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
TROLLING FOR USER DAC DIRT! --------------------------- It's once again that time of year, folks! For those of you who went to last week's DAC in Las Vegas, I'd like to ask you for your impressions for the Forth Annual ESNUG/DAC Awards. Please write me what you thought (and, yes, you'll be anonymous!) What did you think was the best/worst/ whatever DAC party, and why? Technically what was the most interesting EDA product/tool you saw on the DAC floor and why? The worst? The <make your own category> product? What was the biggest lie you heard from an EDA salesman and why? What was the best freebie? The worst freebie? The <make your own category> freebie? (And why?) What was the best personal quote? The best/worst/whatever DAC panel/paper/talk and why? Which vendor had the best DAC booth and/or floor show? The worst? The stupidest? The most content-free? Ever watch a product crash right in the middle of its demo? Whose? .... AND CALLING ON THE DARK SIDE, TOO! ---------------------------------------- Since turnabout is fair play, this year I'd like to add some voices from the "Dark Side" (the EDA vendors themselves) about what *they* saw in the customers at this year's DAC. Of course, you'll also be anonymous, too! What is your most troublesome customer story? What was the most outrageous customer request you got at DAC? Did the customer get it? Whose products do you wish you were selling? Whose are glad you're not selling? What was the most ridiculous claim you saw an EDA vendor try to make at DAC? - John Cooley Part Time EDA Consumer Advocate Full Time ASIC, FPGA & EDA Design Consultant =========================================================================== Trapped trying to figure out a Synopsys bug? Want to hear how 4258 other users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, _] [_ Verilog, VHDL and numerous Design Methodologies. Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion."Article: 3501
In article <4pjmjs$5tm@fstgal00.tu-graz.ac.at> maigner@sbox.tu-graz.ac.at (Manfred Aigner) writes: >I can't find double port RAMS in thr XACT 4000 libary. >How can I use this CLB facility if there is no symbol in the Libary ? >Thanks in advance >Manfred Aigner alias maigner@sbox.tu-graz.ac.at The dual port RAM and Synchronous RAM symbols are not in the XC4000 library, they are in the XC4000E library. If you don't have this library, then you don't have the Xilinx CD for the XC4000E, which is labeled XC4000E Pre-Release V1.0.0 and came out around the same time as the XACT 5.2/6.0 release (escape). The XC4000E software including libraries was supplied on a request only basis, initially with an NDA, but I don't think so any more. You should just need to ask your local salesman or FAE for a copy, assuming that you have a paid up license and maintainance contract. I've been using this stuff for about a year, and both the sync RAM and the dual-port stuff is great. All the best, Philip Freidin.Article: 3502
In article <1996Jun10.103315.11836@super.org> hubert.pujol@matra-transport.fr (Hubert Pujol) writes: >Hello, >I have an hold schematic (8 sheets) on DASH for a Xilinx 3064. I have files such as ".xnf", ".lca", ".rpt" ... >I want to get a new one for update purpose. >We reentry the schematic on compass and reroute the 3064. But net are rename and file differ from the hold ones. > >My problem is the following : >How can I be sure that the new design is equivalent to the first one ? > >The matter is not to do formal proof but to be able to get a same checksum for both designs. If you are running APR again on a netlist with different signal names, there is almost no way that you will get identical checksums for the resultant bitstream, as the device will be routed differently. To get the software to generate IDENTICAL output, you need to give it IDENTICAL input, and you need to force the simulated anealing seed value on the second run to whatever was either chosen or forced on the first run. If your two sets of schematics represent the same design, gate for gate (your goal, I believe), then you can at least compare the chip usage report and check that they indicate the same number of CLBs. You might want to try and compare the trimming reports. If you can get all the signal names to be the same, you might try and run the APR on the second schematic/netlist with the first design's .LCA file as a guide. this might get you very close to identical results. Good luck, Philip Freidin > >Does anybody get experiences for this problem ? >Thank you for your help. >Hubert Pujol >D. R. M. >Matra - Transport - International >48 - 56 rue Barbes >BP 531 >92542 Montrouge Cedex > >Tel 33 1 49 65 71 65 >Fax 33 1 49 65 72 39 >e-mail hubert.pujol@matra-transport.fr >--------------------------- >Article: 3503
In article <Pine.SUN.3.91.960611101320.18662A-100000@sushi> Rainer Scharnow <amigo@bintec.de> writes: >Is there anyone who had experience in constructing a crystal oscillator >(1...10 MHz) at a XC31xxA similar to the description in the Xilinx >Programmable Logic Data Book? My very special problem is I cannot use the >dedicated IOs XTAL1/2 for that purpose. > >E-regards > >--------------------------- >Rainer Scharnow >(amigo@bintec.de) >BinTec Commmunications GmbH >--------------------------- The XTAL1/2 pins have a special buffer designed specifically for driving a crystal as described in the data book. I have successfully built oscillators with the XC4000 (which doesn't have these pins) by configuring an inverter between two general purpose pins, and connecting the crystal to these two pins. There is a 2.2 M Ohm resistor in parallel with the crystal, and there are two capacitors to ground, one from each pin of the crystal, The values are 120 pF and 18 pF for a 2.4576 MHz crystal. Sorry, cant tell you which way round the two caps are. you will have to experiment with the cap and resistor values, depending on the crystal frequency. I seem to remember that you can leave the crystal out, and the circuit will oscillate anyway. you adjust the components till the RC oscillator is operating near the desired frequency, then add the crystal to lock it. Good luck, Philip FreidinArticle: 3504
> I have for a long time wondered if anyone will make FLASH-based FPGAs. Altera's (formerly Intel's) FLEX family is flash based.Article: 3505
In article <DsuGEo.1pA@world.std.com>, John Cooley <jcooley@world.std.com> wrote: > > TROLLING FOR USER DAC DIRT! > --------------------------- > > What was the best freebie? The worst freebie? > <make your own category> freebie? (And why?) Nomination for the worst freebie: I got a free water bottle (the bicycling kind) in exchange for listening to a Veritools salesperson. After the conference, I drove to the Grand Canyon with a friend for 3 days of camping. The first day we went on a 6 hour hike. With temperatures reaching 110 degrees in the valley, it is critical to bring lots of water. Many people die each year from dehydration, including one unlucky hiker just the day before. I brought three water bottles on that hike, including the Veritools one I got at DAC. Two hours into the hike, I discovered most of the water in the Veritools water bottle had leaked out. The damn thing was defective. Fortunately, I didn't die of dehydration, but it did soak everything in my backpack. I guess the moral is, don't entrust your life to a CAD vendor. -- Eric McCaughrin "There's no such thing as meric@mti.sgi.com bad weather -- just bad http://reality.sgi.com/employees/meric_mti clothing."Article: 3506
In article <4pi8ro$48hc@news-s01.ny.us.ibm.net>, 100410.763@compuserve.com wrote: > I need to build a UART into an Actel 1010. > > Ideally this should be 16450 or 16550A compatible. > Let me give you a general comment about implementing a standard peripheral in an FPGA: When Intel, National or Motorola designed dedicated peripherals, they tried to make these deices as versatile as possible, so that one standard device can serve a multitude of different applications. So they put in some mode control registers to define: Number of bits in the word, number of stop bits, parity,etc. When you implement such a UART function in an FPGA you have two choices: Build an exact replica of the general purpose device, which will then take up a surprising amount of space ( = cost ) because all that cheap dedicated flexibility designed into the original standard part must be implemented with relatively expensive programmable logic. Or you make your choices up front and implement a UART with exactly your desired features, e.g. 7 data bits, 1.5 stop bits, odd parity, etc and forget all the unnecessary flexibility, because you don't relly need it. That solution can be quite efficient. Think about this carefully, and don't be surprised or disgusted when the "complete" UART implementation in an FPGA or CPLD is unacceptably expensive. There is always another way to do it more cheaply. Peter Alfke, Xilinx ApplicationsArticle: 3507
> erik@blender (Erik de Castro Lopo) wrote in article <4pjblc$tqn@reader1.reader.news.ozemail.net.ozemail.com.au>... > Hi, > [stuff deleted] > Thanks in advance, > > Erik. Erik, what is your email address? erik@blender bounces. Richard Gregg -- Richard Gregg Computer Science Department University of Tasmania Email: rr_gregg@cs.utas.edu.auArticle: 3508
Hi, I'd like to hear from anyone who is using the Xilinn LogiCore PCI target Interface for the Xilinx XC4000E series FPGAs. I'd be particularly interested in finding out how easy the package is to use and how easy it is to add application specific logic with out destroying the routablility and the PCI timing constaints. I also got a quote from the Australian distributor of Xilinx on the XC4013E-3, of US$318 in 100 off quanities. This is considerably more than I would have expected after using the XC3000 and XC3100 series. Does this price seem reasonable? Thanks in advance, Erik. PS : Please reply to the newsgroup as I don't have a proper return email address.Article: 3509
Manfred Aigner wrote: > > -- > I can't find double port RAMS in thr XACT 4000 libary. > How can I use this CLB facility if there is no symbol in the Libary ? > Thanks in advance > > Manfred Aigner alias maigner@sbox.tu-graz.ac.at You need to use the 4000E libraries since dual port ram's are only implemented in the 4000E and 4000EX family Phil Roberts phil@cri-dsp.com http://cri-dsp.comArticle: 3510
On Tue, 11 Jun 1996, Philip Freidin wrote: > In article <Pine.SUN.3.91.960611101320.18662A-100000@sushi> Rainer Scharnow <amigo@bintec.de> writes: > >Is there anyone who had experience in constructing a crystal oscillator > >(1...10 MHz) at a XC31xxA similar to the description in the Xilinx > >Programmable Logic Data Book? My very special problem is I cannot use the > >dedicated IOs XTAL1/2 for that purpose. > > > >E-regards > > > >--------------------------- > >Rainer Scharnow > >(amigo@bintec.de) > >BinTec Commmunications GmbH > >--------------------------- > > The XTAL1/2 pins have a special buffer designed specifically for driving > a crystal as described in the data book. I have successfully built > oscillators with the XC4000 (which doesn't have these pins) by configuring > an inverter between two general purpose pins, and connecting the crystal > to these two pins. There is a 2.2 M Ohm resistor in parallel with the > crystal, and there are two capacitors to ground, one from each pin of > the crystal, The values are 120 pF and 18 pF for a 2.4576 MHz crystal. > Sorry, cant tell you which way round the two caps are. you will have to > experiment with the cap and resistor values, depending on the crystal > frequency. > > I seem to remember that you can leave the crystal out, and the circuit > will oscillate anyway. you adjust the components till the RC oscillator > is operating near the desired frequency, then add the crystal to lock it. > > Good luck, > Philip Freidin > > > > OK, I haven't been lazy for the last time... :-) Thanks very much and here is my working result (variations will be welcome): <--------o-------------------------------------+ | | | | | ____O____ / \ \ / / \ \ / / \ IBUF \ / OBUF (inv.) / \ \ / ----+---- Y | | / \ / \ = = = = | | = = = = = = = = = = = = = = = = = | | = = = = \ / PIN A \ / PIN B | | | 860 k | | +--------+ | o--------------+ +-------------o | +--------+ | | 4...8 MHz | | +-+ 560 | | I | | I +--------+ | o-------+ | | +------+ +-------o | I | | I +--------+ | | +-+ | | | ---+--- ---+--- 56 p 56 p ---+--- ---+--- | | | | | | --+-- --+-- E-regards --------------------------- Rainer Scharnow (amigo@bintec.de) BinTec Commmunications GmbH ---------------------------Article: 3511
Hi, I'm looking for information on types of faults that occur in SRAM based FPGAs such as XC3K/4K and AT&T ORCA. The type of information I'm looking for is 1. which resources fail most often ( local lines/CLBs/IOBs/pips/switch matrices) 2. What faults are most common (metal migration/ bridging etc). It would be helpful if there was some info on how closely do FPGA failures resemble those of PLAs or static RAMs. Pointers to books /papers/data sheets etc would be appreciated. Thanks for your time. regards, Gurpreet.Article: 3512
> Hi, > > I'd like to hear from anyone who is using the Xilinn LogiCore PCI target > Interface for the Xilinx XC4000E series FPGAs. I'd be particularly interested > in finding out how easy the package is to use and how easy it is to add > application specific logic with out destroying the routablility and the > PCI timing constaints. The design is such that the PCI logic is in the right half of the array leaving the right half for the application. If the application does not crowd the PCI half timing should be OK. > > I also got a quote from the Australian distributor of Xilinx on the > XC4013E-3, of US$318 in 100 off quanities. This is considerably more than > I would have expected after using the XC3000 and XC3100 series. Does this > price seem reasonable? The price is similar to prices I have seen in California. If price is a concern and the application is small the PCI can be moved to a 4010E. > > Thanks in advance, > > Erik.Article: 3513
In article <Pine.SUN.3.91.960611101320.18662A-100000@sushi>, Rainer Scharnow <amigo@bintec.de> wrote: > Is there anyone who had experience in constructing a crystal oscillator > (1...10 MHz) at a XC31xxA similar to the description in the Xilinx > Programmable Logic Data Book? My very special problem is I cannot use the > dedicated IOs XTAL1/2 for that purpose. > There is a simple, fast one-stage inverting buffer between XTAL2 and XTAL 1 without any input hysteresis, input pull-up. If you use other pins by configuring an inverter between them, you get more gain, an odd number of inverters, input hysteresis and other "stuff" that gets in the way of a simple oscillator function. But it can work. Consider using a canned oscillator. You can buy them for the same price as a crystal ( <$1 here in the US in a retail store, single quantity ) and they use less power and take insignificantly more space. That's the reason we deleted the crystal oscillator function from the XC4000 and later families. Peter Alfke, Xilinx ApplicationsArticle: 3514
ft63@dial.pipex.com (Peter) wrote: >>I believe Zycad makes flash based FPGAs, though I've never heard of anyone >>actually using them. >I have for a long time wondered if anyone will make FLASH-based FPGAs. >This would be the ideal thing: the easy in-place multiple >reprogrammability of Xilinx, with the facility to *not* have to do it >at every power-up. >Peter. Xilinx now advertises the XC9500 family of FPGAs that are FLASH-based ("Fast-Flash"). jmArticle: 3515
Just thought I'd let you know Xilinx will be shipping XACTStep v5.2.1/v6.0.1 software this month (June). This release will be a production release that fully supports the XC4000E (for the dual port rams required) and the XC9500 FLASH CPLD family. Also be sure to ask for the various application notes on the family and the rams. You can also browse www.xilinx.com. Hope this helps, good luck (or is that "good engineering"!)Article: 3516
Roberta Fulton wrote: > > Just thought I'd let you know Xilinx will be shipping > XACTStep v5.2.1/v6.0.1 software this month (June). [cut] Is this the release that runs under Windows95 (originally planned for April)?? -- Regards AndyGArticle: 3517
Ever wondered how fast you can read external SRAM from a xilinx? In the process of testing my PCI-based FPGA coprocessor board (http://www.research.digital.com/SRC/pamette/), I wanted to know how fast I could run the external SRAM. My board contains XC4010E-3 parts and Motorola MCM6206D 32kx8 SRAM organized in two banks 64k x 16 bits. Currently I'm using the 15ns grade. the SRAM address/data bits are interleaved to allow efficient pointer chasing. To test read speed I devised the following setup. The SRAM is filled with a known value, say 0xDEAD. Starting from two designated locations, two randomly selected sequences of addresses (each address different) are constructed which loop back on themselves with some short period, say 8. Perhaps this is clearer as follows, if r is the SRAM, then for any value x in the loops: x = r[r[r[r[r[r[r[r[x]]]]]]]]; and for any value y out of the loops: 0xDEAD = r[y]; (of course 0xDEAD can not be a value in the loop). I then constructed the following circuit: +----> check | +-----+ | +-----+ +-----------+ | IOB | | | IOB | | SRAM | +--+ in +-+-+ out +---------+addr data+----+ | | REG | | REG | | | | | +-----+ +-----+ +-----------+ | | | +---------------------------------------------+ The IOB registers were arranged to initialize to the two known values in the loops (I used 0x8000 and 0xffff). I then ran the circuit at differing clock speeds. Whenever I saw 0xDEAD on the "check" output I knew the circuit had failed. If I connected both IOB REGs to the global clock the circuit worked up to 55MHz. Configuring the "in" REG in NODELAY mode got me to about 60MHz. In principle the SRAM should be able to be cycled at 66MHz, its all just a matter of adjusting the clocks, which is after all what NODELAY does. My clock comes it on a BUFGP (dedicated primary global buffer input pin). I routed the input from of non-dedicated IOB associated with that input pin to the neighbouring BUFGS (secondary global buffer) and used this as the clock for the IOB "in" REG. Because both clocks come from global buffers I don't need to worry about skew within the clocks and the added delay to the BUFGS varies little from ppr to ppr. It just redues the time the value has to propagate inside the Xilinx -- fine on deeply pipelined designs -- so the trick seems a good one. With it the SRAM read test starts to fail at around 71MHz. Nevertheless I think I'll put 12ns SRAM on my next batch of boards. that way 66MHz will work reliably and maybe speed freaks who like to live dangerously can push out to 80MHz! Mark Shand shand@pa.dec.comArticle: 3518
Erik de Castro Lopo wrote: Dear Erik, I too am looking into the Logicore PCI design and have found some very interesting material on www.xilinx.com describing distributed material with there product. Also, I have, in house a loner design (hierarchical) in viewlogic schematics. This design is targeted at the xc4013e-3 device and incorperates a burst fifo with the target design. The 4013e has ~ 50% of its resources left over for user defined applications. The xilinx site has lots of information concerning the Logicore design and they will send you some free into by contaction pci@xilinx.com subject:pci packet HOpe this helps: Steve Gerson steve@cri-dsp.comArticle: 3519
SMACD'96 CALL for PAPERS --------------- 4th International Workshop on Symbolic Methods and Applications to Circuit Design --------------------------------------------------- Leuven, October 10-11, 1996 Dept. Elektrotechniek, ESAT-MICAS Katholieke Universiteit Leuven Leuven, Belgium General scope of the Workshop - - - - - - - - - - - - - - - Symbolic methods for the analysis and design of electronic circuits have been the subject of steadily growing interest over the past years. Progress in methodologies and new algorithms, and increasing efforts in computer implementation have led to more usable software. On the other hand, the fast evolution of the CAD domain, and the renewed interest in analog circuit design and design automation, has put new light on the symbolic approaches. In Europe, a growing number of teams are involved in research on symbolic simulation and applications to circuit design. Since 1991 the bi-annual International Workshop on Symbolic Methods and Applications to Circuit Design (SMACD) has been organized as a forum for exchange of new ideas and advances in the field of symbolic circuit analysis and its applications. Following successful workshops in Bagneux (France - 1991), Firenze (Italy - 1992), and Sevilla (Spain - 1994), the 4th edition will be organized in Leuven (Belgium) on October 10-11, 1996. The workshop is endorsed by IEEE and ECS. In addition, a selection of papers will be considered for a special issue of the IEEE Transactions on Circuits and Systems, part II. Topics of interest - - - - - - - - - - Papers on all aspects of symbolic analysis techniques and their applications in circuit design are welcomed. Topics of interest include, but are not limited to: - recent advances in symbolic analysis techniques, design tools, and methods - algorithms from computer algebra and graph theory applicable to symbolic circuit analysis - realizations of symbolic simulators for analog and digital circuits, including demonstrations - new methods of circuit design and design automation that make use of symbolic techniques - innovative applications of symbolic analysis The program also includes an invited speaker and several other technical and social events. Author schedule - - - - - - - - Prospective authors are invited to submit a 2-page summary of their contributions to: Danielle Vermetten, SMACD'96 Secretariat Department Elektrotechniek, ESAT-MICAS Katholieke Universiteit Leuven Kardinaal Mercierlaan 94 B-3001 Heverlee Belgium Tel : ++32-16-321077 Fax : ++32-16-321975 E-mail: smacd96@esat.kuleuven.ac.be Author schedule Submission of summaries .............................. June 24, 1996 Notification of acceptance........................... August 5, 1996 Submission of final papers....................... September 27, 1996 ORGANIZING COMMITTEE - - - - - - - - - - - Prof. G. Gielen Prof. W. Sansen Dr. P. Wambacq Katholieke Universiteit Leuven, Belgium SCIENTIFIC COMMITTEE - - - - - - - - - - - Dr. M. Berger, Bosch, Germany Prof. F. Fernández, CNM - Sevilla, Spain Prof. G. Gielen, Katholieke Universiteit Leuven, Belgium Prof. M Hassoun, Iowa State University, U.S.A. Prof. E.-H. Horneber, University of Braunschweig, Germany Dr. A. Konczykowska, CNET, France Prof. A. Liberatore, University of Firenze, Italy Prof. S. Manetti, University of Firenze, Italy Dr. R. Sommer, University of Kaiserslautern, Germany Dr. P. Wambacq, Katholieke Universiteit Leuven, Belgium Dr. Q. Yu, Crystal Semiconductors, U.S.A. ============================================================================= PRELIMINARY REGISTRATION FORM I wish to attend the SMACD'96 Workshop. Please, send program and application form when available. Name: Mailing address: Phone: Fax: E-mail: Please complete and return to SMACD'96 Secretariat (attn Danielle Vermetten) Department Elektrotechniek, ESAT-MICAS Katholieke Universiteit Leuven Kardinaal Mercierlaan 94 B-3001 Heverlee Belgium or by e-mail to : smacd96@esat.kuleuven.ac.beArticle: 3520
Just thought I'd let you know Xilinx will be shipping XACTStep v5.2.1/v6.0.1 software this month (June). This release will be a production release that fully supports the XC4000E (for the dual port rams required) and the XC9500 FLASH CPLD family. Also be sure to ask for the various application notes on the family and the rams. You can also browse www.xilinx.com. Hope this helps, good luck (or is that "good engineering"!)Article: 3521
In article <peter-1106961320340001@appsmac-1.xilinx.com>, Peter Alfke <peter@xilinx.com> wrote: >When you implement such a UART function in an FPGA you have two choices: > >Build an exact replica of the general purpose device, which will then take >up a surprising amount of space ( = cost ) because all that cheap >dedicated flexibility designed into the original standard part must be >implemented with relatively expensive programmable logic. > >Or you make your choices up front and implement a UART with exactly your >desired features, e.g. 7 data bits, 1.5 stop bits, odd parity, etc and >forget all the unnecessary flexibility, because you don't relly need it. >That solution can be quite efficient. > >Think about this carefully, and don't be surprised or disgusted when the >"complete" UART implementation in an FPGA or CPLD is unacceptably >expensive. There is always another way to do it more cheaply. What Peter didn't say here, and kudos for not blowing your own horn, is that one of the other ways is to use a reprogrammable FPGA (which does not suit the Actel application at all!), and store a number of useful configurations off-board, loading up only the necessary one at the time of invocation. This is possibly not useful for an application with a 16450, but for a serial port that functions as async 5 bits, async 8 bits, sync HDLC etc. this is a viable and very useful attribute of FPGA-based UARTs. Another thing Peter didn't say was that Xilinx has an app note on building a 16450 in their parts. I cannot comment on how much work is involved in making it go on an Actel part, but the version I implemented in XC5200 was not quite flawless, and required a little helping hand... There are other sources around on the Net as well, if you poke around on some of the HDL repositories, for instance. FWIW. Arnim -- Arnim Littek arnim@actrix.gen.nz Actrix Networks Ltd. fax +64-4-499-1130 uucp/PPP/SLIP/BBS accounts tel +64-4-499-1122Article: 3522
ANNOUNCEMENT Menlo Park, CA (June 13, 1996) In order to thank its customers for their continuous support, Fintronic USA Inc. announces today that for a limited time it will offer two copies of its most suc- cessful products for the price of one. Starting today until Sep- tember 15th 1996, this promotion will be available for FinSim- ECS, the full Verilog simulator with the integrated Enhanced Cy- cle Simulation (ECS) Technology which has recently finished on top in an independent benchmark performed by DA Solutions (please see the June 3rd edition of the EE Times, article titled ``HDL simulators rated'', page 16). The list price for FinSim-ECS is $20,000, or $24,250 if bundled with Signalscan, the debugging en- vironment from Design Acceleration. The same two for one promo- tion is available for FinSim-Developer, a full Verilog simulator with complete support for PLI and SDF and which has a list price of only $5,000 (or $6,000 if bundled with Signalscan). Both pro- ducts run on most popular platforms including SUN, HP, SGI, DEC, Sony NEWS, PowerPC with Windows NT, Intel x86 with Windows NT, Windows 95, Windows 3.1 or Linux. Also today, Fintronic USA is introducing a new product, FinSim- Station, which we feel will provide its users with the fastest simulation environment at the lowest price available in the in- dustry. It includes a Pentium Pro 200 Mhz, 64MB RAM (expandable to 512MB for only $9,000 extra), 2GB HD, 8x CD-ROM drive, 3D ac- celerated graphic adapter, ViewSonic 17PS monitor and our top performance Verilog simulator FinSim-ECS. ``The performance of our software on the Pentium Pro 200 Mhz platform is significantly better than on an UltraSparc 167 Mhz and this performance comes for only a fraction of the cost. Furthermore, we have seen many designers worrying about the memory requirements for tomorrow's million gates designs but few realize that half a Gigabyte of RAM for the Pentium Pro costs under $10K. Many people have yet to discover the tremendous potential of the Pentium Pro platform for EDA, and FinSim-Station is our attempt to prove this potential'' says Dr. Alec Stanculescu, president of Fintronic USA. Until September 15th 1996, FinSim-Station will have a promotional list price of $17,000 for Linux and $20,000 for Windows-NT. The Windows-NT version comes bundled with Signalscan. For more detailed and up to date information as well as online ordering forms, please consult our Web page at http://www.fintronic.com, call Dr. Alec Stanculescu at (415) 325-4474 x105 or fax us at (415) 325-4908.Article: 3523
There hasn't been much reconfigurable computing discussion lately so... Many of you might have heard Bob Parker talk about the "rule of thumb" performance numbers for FPGAs. The number was how many 8-bit adders you can fit in an FPGA times the frequency at which they can be clocked. This came from me and the thought behind it was most operators in a high level languages are simular to adds: adds, subtracts, compares, increments .. So if an FPGA can hold X number of 8 bit adders then that times the freqency is kind of like the number of 8-bit ops it can do. Although You can't get 100% utillization if you think about 50% utillization and then consider that all the data has to move (and this is an operation) then you start to come back to "I should just count the adders" type of thought. I think one of the reasons reconfigurable computing works is that the data flow is in the hardware. As opposed to a CPU that emulates the data flow of an algorithm in a set of registers. So if you don't think my "rule of thumb" is very good then what would be a good one? "Its different for every problem/algorithm" is not a good answer! Steve Casselman (I guess I havn't had a good flame for a while:)Article: 3524
xilinx supports dual port rams in their new 4000E architecture. You need to obtain an update from Xilinx. Some time I was given the following part number for the 4ke update PR-VL-STD-PC1-4E-C This was for the ViewLogic PC package Rich Lomas
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