Now I am really confused!

This debate is getting long in the tooth. Just take the number of slices and divide by two. In the meantime, Altera has jumped on this bandwagon and they multiply their ALE numbers by 1.25. As long a there is Marketing, there will be "creativity" with numbers. Just grin and bear it! Peter Alfke

Now I am more confused. If I divide the number of slices by two, I get a number that is more than 4 times less than what is given in the data sheet.

I don't understand how Altera data sheets relate to your data sheets. If marketing is going to write your data sheets, can you at least get them to give an honest definition of the logic cell???

"Logic Cell = ((1) 4-input LUT + (1)FF + Carry Logic) * 1.125"

I don't care how you market, but right now your data sheets are not consistent and will confuse anyone who does not read this newsgroup.

Peter Alfke wrote:

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Our stuff still has a way to go before it catches up with the real smoke and mirror artists. Did the Citroen 2CV ever have a two horsepower engine, even when first introduced - whenever that was? Actually I don't understand that one... it's as if X advertised "Virtex-4: more than one LUT"

ok, I should have said multiply by two. I just think this argumenting about an extra 12.5%, silly as it may be, is not worth our combined efforts. Everybody knows where the factor comes from, most of us disagree with the reasoning. Grin and bear it. Do you want to argue about the meaning of "100%" in wine merchandizing? Do you want to fight the funny factor two in the "%proof" calculation? Or have you measured a "2 by 4 inch" piece of lumber? There is lots of fiddling with numbers going on...

Peter Alfke

Peter should have said multiply by two instead of divide by two. There are two LUTs in a slice, 4 slices in a CLB. So take the number of slices and multiply by 2 to get LUTs. It looks like Xilinx is also counting the IOBs in their logic cell count.

I agree Peter, this is not really worth much effort. I don't understand why you bring in unrelated issues such as the proof of spirits or the size of lumber. A 2x4 is sized according to a spec, no one says it is

2" x 4". The proof of spirits is not percentage, so I don't see why you are comparing the two, again, this is clearly stated. But the V2Pro data sheets defines a logic cell one way and are counted another way that is not discussed unless someone reads your posts here on the newsgroup. I understand who is doing it and why. At one point I thought it was just a matter of marketing getting into the data sheets. But if they contradict their own definition of what a logic cell is, then the data sheet contains falsehoods.

If you don't want to discuss it, then feel free not to. I am not trying to sound rude or insulting. I belive I have only stated facts. So please don't feel that I have a grudge, I am just trying to get the facts straight.

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

have

If people who market to engineers took to heart what engineers need - accurate information - then this discussion wouldn't involve *any* combined effort. I'd like to think the silicon vendors are more precise than the lumber industry or more informative than a centuries-old form of labeling alcohol content; after all, are the words "percent" and "proof" commonly used *together* in the wine industry? A troy ounce isn't an ounce. We know that. If definitions are provided and spec sheets adhere to those definitions, we're happy. If the spec sheet for a "2 by 4" specifies 2.00" x 4.00", we'll be unhappy with the undersized product - any spec sheet

*should* have the actual dimensions and tolerance.

Give an engineer precision and there won't be arguments. Complacency with marketing slop - inaccuracies for the sole purpose of justifying a position - may have a place in consumer advertising but is seen as marketing "lies" whenever we LEARN about the BLATANT inaccuracies of the marketing literature. Remember 19" displays (16.2" visible)? Suits forced the industry to properly specify the numbers. I know to go into the rows and columns to get my LUT or LE counts the way that I know they should be specified. It's because I know better now. What impression are we giving the new engineers looking to develop their first FPGA designs? Distorted information is expected from the silicon vendors? That's what they take away once they learn their original estimates were wrong because they didn't KNOW to do conflicting arithmetic on their datasheets.

We - as engineers - will never be truly complacent with marketing distortions propagated by our organizations or those we do business with.

I think there would be more respect and appreciation garnered from the engineering community if DATASHEETS were fully accurate and didn't include marketing distortions. I can see calling a device a "1Mgate" device where the definition of a gate varies with perspective, but don't tell me there are 12.5 lbs of manure in a 10 lb bag because it smells better to marketing.

This is a distortion we continue wrestling with because we cannot change it. We can only work around it as individuals. How about taking up with other marketing slants and use verbiage like "compare with units having 12.5% more resources" or similar sidestepping?

Those within the organizations selling FPGAs have to communicate that the numbers are "justified" because it's not right to call your own company's literature misleading. It's easy to "believe" that the numbers are okay because "that's the way it's always been" or that "everyone else is doing it" but it's wrong and there should be internal and external pressures to make the information *accurate* when accuracy is called for.

And from where I sit, I feel I'm significantly less frustrated than rickman.

Oh... and remember years ago when Brand A illustrated 3 registers in their I/Os? My initial work with the datasheets said my timing was good. It was only after getting further into the design that I realized those I/O cells only contained SINGLE registers and couldn't possibly achieve my required performance. Some high-management folks eventually visited us to hear our rants and apologize for the confusion. They not only lost a socket to Brand X for the technical aspects of the design requirements but lost first consideration on most future sockets from a previously "biased" engineer.

If I had the "correct" information in the front end of the design cycle, the marketing issues wouldn't have become engineering and financial issues. I would have had stronger respect for the currently offered solutions than I have now. Time heals many things.

Here's a radical thought : Given that 'marketing numbers' have caused, over time, at a minimum annoyance or bemusement, and at a maximum (above) outright customer flight, what about the idea that instead of excusing this silliness, it actually gets fixed, and data sheets get provided with accurate formula for available resource ? If that formula includes a [Real World Calculation] * MFF = HahSoThere we could live with that - even 'grin and bear it'. The MFF ( marketing fiddle factor ) would only be used by the marketdroids, and they alone would boast and compare the right hand side, whilst the designers would use the real world numbers in the multi term equation on the left hand side.

Marketing would not have to go 'cold turkey' as their number would still be in the data sheet - just now it is clearly identified :)

Or, will we see 'marketing MHz' and 'narketing ns' numbers next..... :(

-jg

Anybody old enough to remember Radio Shack's spec on the stereo amps? 150 WATTS to us means out put power to speakers. From Radio Shacks spec it was input power on the AC side.

Would it do any good if several people wrote (paper) letters to the marketing dept? Mayve we could convince some of them to use real numbers and brag about having honest info. They could even make up a fancy name for it. :) or :(

I think it terms of LUTs/FFs so an accurate table is what I'm looking for. Extra columns for RAM, multipliers, and such would be handy.

--
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

their

Jumping to our defense on this one, we clearly state the number of Adaptive Logic Modules (ALMs) in our Stratix II literature, and a give a separate "Equivalent Logic Element" count. We use 2.5 LEs per ALM to provide our customers with a way to easily compare the logic capacity of Stratix and Stratix II devices. This is not smoke and mirrors -- the Stratix II ALM is a radically different logic structure. The number of ALMs/LE will differ from design to design, but over 100+ user designs we found that 2.5 was a reasonable estimate.

Regards,

Paul Leventis Altera Corp.

John_H a écrit:

I do! I'm still angry with them about that one (though they actually put 3 registers in their latest chips)

--
  ____  _  __  ___
|  _  \_)/ _|/ _ \   Adresse de retour invalide: retirez le -
| | | | | (_| |_| |  Invalid return address: remove the -
|_| |_|_|\__|\___/

I'm not trying to take sides in this, but my original post was about how the data sheet defined "Logic Cells" one way and then counted them another. If the data sheet clearly says that the numbers are "adjusted" with some fudge factor, that is not so big a deal. But if you define an LC one way and count it another, that is just not honest.

Every FPGA has extras in them which extend the capacity of the LUTs. But when I am trying to count LUTs in my block diagram to see how large a part I will need, I already have all the "fudge factors" I can handle, I don't need more from the data sheets!

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

The deux chevaux were really a measure of taxation, had little to do with engineering horsepower (which has little to do with the performance of a horse). The Brits used a similar strange taxation measurement. Peter Alfke