Lattice XP2 finally announced

Jim,

The note in the "changed" history at the end of the datasheet (3E), says only the max has changed.

That is consistent with tighter process control. Also consistent with no change to the process center.

formatting link

page 161

Thanks for this opportunity to explain it.

Austin

Reply to
austin
Loading thread data ...

Jon,

Yes, V4 FX was very unpleasant for Xilinx (specifically, the MGT's). I have posted on this subject before. And, the issues with the MGT's, made many customers unhappy.

I look to the future, and try not to dwell on the past, but you are absolutely correct: and I offer my apology as a Xilinx employee, and promise that many things have already changed so that we never make that mistake again (witness the roll-out of V5 LXT).

Any abuse that is sent my way concerning the failure to realize the promises on V4 FX MGT is just something I and my colleagues will have to accept.

Austin

Reply to
austin

Antti,

Yes, Steve worked at Triscend.

Triscend had reached a point where it was either close the doors, and go out of business and put everyone on the street; or call up folks at Xilinx and ask if their excellent staff of employees could possibly find a home.

The doors still closed.

Since some were X-Xilinx employees, and we knew what excellent designers, etc. they were, it was a simple matter to ask them to come to

2100 Logic Drive the next day, rather than start looking for jobs.

We did the same with DynaChip.

The Triscend group has distinguished themselves with valuable additions to V5.

The Dynachip group is responsible for the design excellence of Virtex II and continuing.

It is not our fault that being a FPGA vendor is incredibly difficult. The history is littered with those who failed. Big names, like Intel, ATT Microelectronics, Motorola have all tried.

And, the valley is littered with the stock of little shops that also failed.

Austin

Reply to
austin

eh ok, maybe.

still a bit pitty that ARM is not taking off in FPGAs as Triscend closed doors and Altera discontinued the products.

Antti

Reply to
Antti

I think you should blame ARM's greed for that.

Reply to
mk

Antti,

ARM is nice, but how applicable to the day to day FPGA designer's SOC requirements?

With the Virtex line, the IBM PPC is the clear winner, with some 30% of designers using the PPC before VII Pro was introduced, and as much as

50% using the PPC now. IBM's comment to us was "Xilinx has done more for the PPC architecture in one year, than we did in ten."

The PPC lends itself to serious DSP programming, as it is quite similar to the processors sold just for DSP.

In the Spartan family, the PPC is not the clear winner, with ARM being used primarily in low cost/low power apps. However, licensing ARM cores doesn't fit the Spartan business model, as it would increase the cost to the customer beyond what the benefit might be.

And, with the chips getting more CLBs with each generation, we can wait and use soft processors until we see a clear winner emerging.

There has even been thoughts of hardening the MicroBlaze(tm) core, but most feel that the last thing the world needs is 'yam' (yet another microprocessor).

Austin

Reply to
austin

PPC, ARM, Microblaze.... I hope MIPS could be an option too! There are a lot of device using it and maybe using FPGA as glue for a hard MIPS inside the FPGA could lower such devices price...

Sandro P.S. I don't know MIPS Licesing cost...

Reply to
Sandro

some dont care about licensing:

there are

  • godson
  • godson
-2
  • some_similar_name

all MIPS derivates

formatting link

Antti there are also many MIPS cores for the use in FPGAs

Reply to
Antti

Sandro,

MIPS was one of the three six years ago, PPC, MIPS, and I forget the other one.

MIPS lost out to PPC, because our wireless, wired, and networking customers all used PPC.

Nothing wrong with MIPS, it was just that we already had a huge market that was already using PPC.

Austin

Reply to
austin

I know wikipedia sometimes is not the truth... but from

formatting link
it seems STM bought a licence from MIPS (?to build godson?) Sandro

Reply to
Sandro

My memory is that Steve was one of the main movers at Triscend. Possibly one of the founders. If Phil Freidin is reading, he may be able to fill in the details.

Reply to
Tim (one of many)

Interesting. From

formatting link
: "The average royalty rate fell to 6.7 cents in 2007, down from 7.9 cents in 2005"

Must be too high because X has some products on the horizon priced at $1 with embedded PPC ;-)

Reply to
Tim (one of many)

must be large difference in royalty, as Actel is charging over 100 USD per ARM enabled FPGA (in small qty)

Antti

Reply to
Antti

Hi Antti, Yes, but for a while customers did have a choice of both, and customers drove the decision by choosing NIOS over ARM. ARM was just too costly to do as a SoftCPU, and too restrictive in a HardCPU.

So the FPGA optimised SoftCPUs won.

Triscend had the same problems as the now trailing FpSLIC. Everyone looks and say 'yeah, nice idea', but the silicon struggles to make actual design wins, because you have locked three variables in silicon, and so have a VERY SMALL market sweet spot, as well as being VERY single-source locked. I'd call them a Salesman's device, not an Engineer's device :)

The recent growth in FLASH 32 bit Microcontrollers is even putting pressure on the SoftCPU design decision.

Some recent examples that have impressed me are

ARM9+FLASH with 'the works': [Atmel have one too, somewhere, in 're-work']

formatting link

Floating Point DSP + FLASH :

formatting link

Atmel AT91CAP - picks the higher volume, mask FPGA Niche,

formatting link

AVR32 - FLASH+EtherNET+OTG.USB

formatting link

-jg

Reply to
Jim Granville

besides these Microcontrollers, I see the MPU offerings are also bringing pressure on the 'FPGA with HardCPU' decision.

This new device from AMCC, has some impressive peripherals and MIPS/$$ ratios.

405EXr: 1,000 DMIPS PCI Express Gigabit Ethernet USB 2.0 On-the-Go 480 Mbps pricing will start at $14 in 10Ku volumes

formatting link

-jg

Jim Granville wrote:

formatting link

Reply to
Jim Granville

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.