Spartan-3AN

I guess someone at Xilinx will reply as to what they consider to be their __technical__ edge.

What is the real shipping status of the Lattice parts?

For one thing you can't access the flash in XP-parts from the logic, as far as I know. The flash is just for configuration. At least that's what the FAE told me, when i asked if there were at least a few blocks I could put user data into. So that would be one advantage of the Spartan-3AN.

XPs are shipping, and in low quantities as well. Last time I asked my distributor, they had every part I asked for in stock. But it's been a while...

Tim,

It's true, flash is only for configuration. But still this flash can be used as mirror for the EBR (no write back though).

I'm using the XP6 in small volumes (50pcs). LT was something like

Luc

No doubt about it, the Lattice XP parts are a nice solution.

So what is the technical advantage of Spartan-3AN? Here are but a few points to consider.

• Spartan-3AN FPGAs offer significantly more in-system Flash memory, between 627K to 12,251K even after storing an uncompressed FPGA configuration bitstream. + Store two different FPGA configuration bitstreams o Dynamically select between the two from with the FPGA application using the MultiBoot feature o Fail-safe "live" in-system Flash updates + Code-shadow MicroBlaze code within the FPGA + Fully accessible from within the FPGA application
• Spartan-3AN In-System Memory is extremely robust + Intended for use by the FPGA application, not just configuration + 100,000 erase/program cycles + 20 year data retention
• Spartan-3AN FPGAs offer the power-saving Suspend mode + Retains FPGA configuration and application state + Fast response time
• Spartan-3AN FPGAs have embedded 18x18 hardware multipliers ? I don't believe that XP parts have multipliers
• Spartan-3AN FPGAs span a broader range of densities, up to 22.5K LUTs
• Spartan-3AN FPGAs have more I/O, more I/O per package, up to 502 I/ O pins
• Spartan-3AN FPGAs offer more on-chip RAM + More block RAM + More LUT RAM
• Pin-compatible with select members of the Spartan-3A FPGA family + Architecturally identical to Spartan-3A FPGA family

I wouldn't claim that either part is "instant on", although the Lattice XP part

-- Steve Knapp

This last point would seem important. So if one places external flash alongside a 3A, (That flash being larger, or smaller, (or faster?) than the in-package Flash of 3AN) you get all the same Flash-access features ?

-jg

Correct. You have many of the same Flash size advantages by connecting an external Flash to a Spartan-3A device. Obviously, however, it is no longer a single package. The Spartan-3AN In-System Flash memory requires no user-I/O pins. Instead, all access is through an internal design primitive called SPI_ACCESS.

I should also mention that Spartan-3AN FPGAs still support _all_ the other Spartan-3A configuration modes (Master Serial, Master Parallel, Slave Serial, etc.). Spartan-3AN FPGAs exclusively offer the Internal Master SPI mode.

-- Steve Knapp

Hmm, no PQ208 package.

Can you summarize the availability picture for the various parts for prototype quantities and for production quantities?

Tim,

From the press release:

"Pricing and Availability Customers can immediately begin designing systems with Spartan-3AN FPGAs using the ISE? 9.1i design tool suite and Spartan-3 Generation library of application-specific IP. Engineering samples are shipping now for the XC3S200AN, XC3S700AN and XC3S1400AN devices with all five devices in production by Q3 2007. At customer production timeframes at the end of

*250k resale volume, 3S200AN-4FT256C

We decided to try something different this time: have stock of the items noted on the shelf at the time of release.

Austin

Did you try in the "Xilinx" web-shop?

Every item is "No stock. Call for info"

So I should work on the basis that Xilinx press release statements of price and availability are rock-solid!

Would it be reasonable to track the Spartan-3A in the Web Shop or in Distribution and work on the basis that equivalent parts in the 3AN range will be coming along six months later?

Unlike buying a toaster, the price for an FPGA is never rock-solid, even for one moment in time for the same quantity. Prices are often negotiable if the difference between a socket win or loss can have an impact on business, present or future. Unfortunately you can't enter your parameters into a spreadsheet and come out with a single price valid for anyone with those same circumstances.

If you have a relationship with your sales rep, explore the options available to you. If Xilinx wants your business more than giving it to one of its competitors, your sales rep might negotiate a price your distributor will honor.

Much of the electronics component pricing is affected to this degree. Reps can help get a negotiated price that will keep the business without harming later business opportunities. I'd love to have complete, fair, unbiased pricing that's applicable across companies and countries. The reality is that support needs for different size (and experience) customers vary and the bottom line impact can be significant or hard to notice. The price will fit your need; if it doesn't, another vendor will fit your need. If you have a need that must be filled, it will be filled. Just not necessarily at the price you want or the vendor you prefer.

At least there's a stake in the sand.

Tim, Uwe,

I am told that parts were shipped to distributors, so they would be on hand on the day of the introduction. Why else would we be so fussy about the strict adherence to the release of information? Parts, software, documentation, cores -- it all has to come together. One item missing is unacceptable.

The X-Store is another matter, and as you well know, Peter and I both continue to spend time to try to improve that situation.

The real good news is that the 3AN hits the streets today with ES parts on the shelves of distributors. The lead times are stated at 2 to 4 weeks in all the paperwork I have seen (probably just to make sure we do not disappoint anyone, anywhere).

Remember, we are changing how we introduce new products, and hope to regain the trust of our customers.

Austin

I understand. Business schools study whether this sort of detailed pricing is a necessity or a make-work program for the sales and marketing department. It must surely absorb spectacular levels of scarce management time. For what we need as engineers, a set of pricing curves (with a few spot prices) would be just as good as real prices. It would be up to the folk in buying to force the pricing curve as low as possible.

For instance, many many years ago Peter A pointed out that, for the products out there at the time, speed grades cost about 10% each, and bigger packages (there weren't too many to choose between) cost around

Right now, I get the sense that for a given part speed grades cost a little more than 10% and the price includes a package element by being more or less linear with the usable pin count. (This is medium volume - high volume rules are different) This makes an engineer's life easy (!) because costs are roughly linear with benefits.

But comparing between parts and between ranges involves poring over the distributor's price charts. It would be much easier if the relative curves were graphed. For instance, Spartan-3, 3E, 3A approx the same price per LUT/per pin, 3AN is x% higher, faster speeds 15% higher. Or something similar.

And as complex as the pricing is, it isn't as complicated as airline ticket pricing :-0

So if I understand the datasheet well, the flash device is just brought in the same package, isn't it? It even seems to be an Atmel DataFlash (R) device. At the end .. the big winner is ... Atmel. They are now delivering configuration devices to both Xilinx and Altera. Probably this is one of the reasons that they are stepping out of the PLD business. (at least this is what I heard) One thing to be mentioned - no 'instant on' feature with the S3-AN as the boot time is still in the 100ms range - as with normal SPI flash.

I agree - being pin compatible with S3-A is a feature, no migration path between densities however is not so nice.

In other words, I need more to convince me to switch over ...

Luc

I don't think Atmel are exiting the PLD business - their recent releases in the CPLD ATF1502BE / 1504BE devices can foot it with the best. Most of their SPLD/CPLD have RoHS versions.

I can see they are more of a fading force in the FPGA area, with the road map there thinner, and not all devices are RoHS, so no, they are not going to threaten a Xilinx Spartan-3AN :)

That said, I know customers that still buy Atmel EPROMS, and most of those are also now RoHS.

The improved coupling between Flash and FPGA is a good thing, in both the 3A and 3AN series. It also reduces the pcb footprint, and has to reduce RFI issues, in those designs that continually access the flash. It's also a soft-border : you can design to use both 3A and 3AN.

-jg

Samples of the XC3S200AN, XC3S700AN, and XC3S1400AN FPGAs are available now through Xilinx sales partners, typically with 4 week leadtime.

The XC3S50AN and XC3S400AN are still to come, likely within the next

If you click on the Avnet and NuHorizons web sites, you'll likely see a "No Stock, Please Call" message. I encourage you to "please call".

Please allow me to shed my Xilinx mantle and speak OPINION.

Xilinx distributors do not carry stock on Engineering Samples because Xilinx maintains a "non-cancellable, non-returnable" policy with the distributors. Distributors have no incentive to stock inventory and consequently carry no inventor on engineering samples. All orders for engineering samples are placed with Xilinx as they come in.

I'd highly recommend downloading ISE 9.1i, Service Pack 2 (available now) before starting a design. Service Pack 3 (mid March/early April) will have improved in-system programming support.

-- Steve Knapp

If the xilinx store does not want to carry regular parts and the distributors do not want to carry ES parts, why not have the xilinx store carry the ES parts? That way something would be available to those who are willing to use ES parts. Since virtex 5 (non ES) is still not avaliable and virtex 4 is scarce, I am not expecting to see the 3AN in less than a year from now for us small folks.

Of course, since versions 8 and 9 of ise are broken fatally for my designs, that may not be such an issue. I am still hoping that xilinx makes the software work with schematics while I am converting the projects to vhdl.