Lattice MachXO3L - is it available anywhere ?

[Brane2]

I've downloaded documetnation on lower part of newcomers - MachXO3.

New Diamnond 3.1 has support for chips, but I can't find chips anyhwere.

I tried Farnell and Mouser and couple of other adresses, but so far can't find anything.


Had anyone managed to spot any MachXO3L model being actually offered ?

 

[Brane2]

Nevermind that.

Have found them at Digikey. They were "hidden" as CPLDs, so I haven't found them previously.

Prices seem a bit insane, but so are their prices for MachXO2, so I suppose these will get to Mouser these days for some normal price.

 

[rickman]

On 3/29/2014 2:40 PM, Brane2 wrote:

Nevermind that.

Have found them at Digikey. They were "hidden" as CPLDs, so I haven't found them previously.

Prices seem a bit insane, but so are their prices for MachXO2, so I suppose these will get to Mouser these days for some normal price.

Not sure what you think of as "insane" prices. What FPGAs do you find
much lower than $10 at quantity 1? This seems to be what I have been
paying for 3kLUT XP parts. I did a survey a few months ago and didn't
find much that was cheaper.

--

Rick

 

[Brane2]

Dne ponedeljek, 31. marec 2014 16:56:03 UTC je oseba rickman napisala:

Not sure what you think of as "insane" prices. What FPGAs do you find

much lower than $10 at quantity 1?

I don't remeber exact numbers, but I remmeber that prieces were unusually high.
So I checked their prices for MachXO2, which I could compare to Mouser's and from this I concluded that main reason is Digikey's higher provisions, not Lattice's pricing policy.


BTW, Mouser has ice40 with 3520 LUTS for $6.2 for 1, which drops under $5 at 50 pcs... Not exactly MachXO3, but could be enough for many applications...

 

[rickman]

On 4/1/2014 10:43 AM, Brane2 wrote:

Dne ponedeljek, 31. marec 2014 16:56:03 UTC je oseba rickman napisala:

Not sure what you think of as "insane" prices. What FPGAs do you find

much lower than $10 at quantity 1?


I don't remeber exact numbers, but I remmeber that prieces were unusually high.
So I checked their prices for MachXO2, which I could compare to Mouser's and from this I concluded that main reason is Digikey's higher provisions, not Lattice's pricing policy.


BTW, Mouser has ice40 with 3520 LUTS for $6.2 for 1, which drops under $5 at 50 pcs... Not exactly MachXO3, but could be enough for many applications...

Er, I still haven't gotten used to the updated T-bird controls... so you
will get a copy of this to your email as well.

Just remember that unlike the other Lattice non-volatile products, the
ice40 line is NOT flash but rather one time programmable. It still has
RAM so you can always load it from external memory I believe.

--

Rick

 

[Brane2]

Dne torek, 08. april 2014 23:30:49 UTC je oseba Tim napisala:

XO3 is clearly having a few neonatal problems, as always in this

industry, so they have launched those parts with a lot of commonality

with the XO2 as the XO3L range.

<SNIP>

The bigger XO3 parts (XO3H?) should

implement the extra stuff that they forgot to remove from the XO3L data

sheet ;-)

I was checking around for higher end of XO3 and was told that Lattice is about to introduce new Sapphire series with HW mult, SERDES etc in May.

It's not 100% clear whether this is simply "XO3H" or are they talking about "FPGA, formerly known as ECP4"

Both could make sense.

3Gb/s SERDES is getting old, so it would be nice to have it available as an option in cheap MachXO and faster 6Gb/s version in ECP4/Sapphire.

OTOH, they might want to ephasize that Sapphire is simply XO3 that went throgh extreme heat treatment, just as they make Synthetic sapphire (Al2O3). ;o)

 

[Tim]

On 01/04/2014 15:43, Brane2 wrote:

Dne ponedeljek, 31. marec 2014 16:56:03 UTC je oseba rickman napisala:

Not sure what you think of as "insane" prices. What FPGAs do you find

much lower than $10 at quantity 1?


I don't remeber exact numbers, but I remmeber that prieces were unusually high.
So I checked their prices for MachXO2, which I could compare to Mouser's and from this I concluded that main reason is Digikey's higher provisions, not Lattice's pricing policy.

The MachXO2 has some neat hard coded blocks, worth it to you if they are
worth it to you ;-) The PLL is particularly flexible, with register
access to many parameters. And there's hard coded support for I2C and SPI.

We use XO2s in a Raspberry Pi expansion card, not unusually difficult to
design.

XO3 is clearly having a few neonatal problems, as always in this
industry, so they have launched those parts with a lot of commonality
with the XO2 as the XO3L range. My take is that the XO3L is most
relevant to really big volumes - it doesn't have a compelling low-volume
advantage over the very similar XO2. The bigger XO3 parts (XO3H?) should
implement the extra stuff that they forgot to remove from the XO3L data
sheet ;-)

 

[Brane2]

Dne sreda, 09. april 2014 04:58:36 UTC je oseba rickman napisala:

<SNIP>

That will be easy to tell. The XPx and XOx lines have on board flash.

The ECPx line is only RAM based. I believe the XPx line is based on the

It will be easy to tell, but I was trying to _predict_ , which is at this point more like guess. ;o)

From what was said earlier in this thread Lattice wants to stop going

head to head with X and A and is shooting for a different market where

having the most advanced process technology is not such a big advantage.

That may be a niche market for now, but I expect it has legs.

Yes, but as said, main Lattice's strategy is to use one process.

They were at 65nm for MachXO2 and ECP3

Now that they bought SIliconBlue, they had to allocate 40nm production capacities somewhere for ice40.

Since 40nm has advanced to sweet spot for clients like Lattice, they started working on 40nm XO2 low cost successor - XO3.

So they made density jump like XO-XO2 to something like 3x and added MIPI, 3G SERDES etc.


And now seems good point to renew ECP3, if X and A have left enough space for it. To me it seems this might be the case.

If they decided to "change gear" with ECP3, then they will probably upgrade SERDES, few other bits and add density or two at high end. And probably lower price points for same densities.

If they can sell ECP3 now, they should be able to sell such updated 40nm ECP4.


<SNIP>

What exactly would be the market for these smaller chips with high end
SERDES?

If you mean ECP4, they are not that small. They are reaching densities and packages where SERDES doesn't lift the price that much any more.

If you mean XO3, you musundeestood me. XO3 is to get 3G SERDES.
What would be the point ? Some cheap/er low-volume peripherals for PCIe, maybe SATA etc.

 

[rickman]

On 4/8/2014 9:49 PM, Brane2 wrote:

Dne torek, 08. april 2014 23:30:49 UTC je oseba Tim napisala:

XO3 is clearly having a few neonatal problems, as always in this

industry, so they have launched those parts with a lot of commonality

with the XO2 as the XO3L range.

SNIP

The bigger XO3 parts (XO3H?) should

implement the extra stuff that they forgot to remove from the XO3L data

sheet ;-)


I was checking around for higher end of XO3 and was told that Lattice is about to introduce new Sapphire series with HW mult, SERDES etc in May.

It's not 100% clear whether this is simply "XO3H" or are they talking about "FPGA, formerly known as ECP4"

That will be easy to tell. The XPx and XOx lines have on board flash.
The ECPx line is only RAM based. I believe the XPx line is based on the
ECPx devices with the flash added as a RAM shadow. They load a lot
faster than the X and A parts. I get the impression they won't be
coming out with an XP3 line just like they don't have an ECP4 line.
From what was said earlier in this thread Lattice wants to stop going
head to head with X and A and is shooting for a different market where
having the most advanced process technology is not such a big advantage.
That may be a niche market for now, but I expect it has legs.

Both could make sense.

3Gb/s SERDES is getting old, so it would be nice to have it available as an option in cheap MachXO and faster 6Gb/s version in ECP4/Sapphire.

What exactly would be the market for these smaller chips with high end
SERDES?


> OTOH, they might want to ephasize that Sapphire is simply XO3 that went throgh extreme heat treatment, just as they make Synthetic sapphire (Al2O3). ;o)

--

Rick

 

[rickman]

On 4/9/2014 2:11 AM, Brane2 wrote:

Dne sreda, 09. april 2014 04:58:36 UTC je oseba rickman napisala:

SNIP
That will be easy to tell. The XPx and XOx lines have on board flash.
The ECPx line is only RAM based. I believe the XPx line is based on the

It will be easy to tell, but I was trying to _predict_ , which is at this point more like guess. ;o)

Can't say I follow what you mean. What are you trying to predict exactly?

From what was said earlier in this thread Lattice wants to stop going
head to head with X and A and is shooting for a different market where
having the most advanced process technology is not such a big advantage.
That may be a niche market for now, but I expect it has legs.
....snip...

If they can sell ECP3 now, they should be able to sell such updated 40nm ECP4.

It is not are they selling ECP3, the question is how many new design
wins? That is what FPGA vendors live and die by. They know that for
each generation they have a window of opportunity to lock in some large
fraction of their business for this line. Those days for ECP3 are over.
So the question is can they make money on ECP4 or has that ship sailed?

What exactly would be the market for these smaller chips with high end
SERDES?

If you mean ECP4, they are not that small. They are reaching densities and packages where SERDES doesn't lift the price that much any more.

If you mean XO3, you musundeestood me. XO3 is to get 3G SERDES.
What would be the point ? Some cheap/er low-volume peripherals for PCIe, maybe SATA etc.

I am taking Lattice at their word that they intend to focus on the
mobile market with smaller, more cost effective devices with lower power
consumption. I fail to see where high speed SERDES fit in with that
strategy.

Can you use generic SERDES for these interfaces? I remember reading
some time back that the SERDES in Xilinx chips could not be used for
SATA, not because of speed limitations, but because the SERDES is not
compatible. I know little about SATA but it seems like a signficant
oversight to me.

--

Rick

 

[Brane2]

It is official now- Lattice is out with ECP5.

Which seems awfull lot like XO3H under new name, which suggests that this is new high-end of their offer. Highest model has 85K luts.

Seems reasonable. At 90K or higher, ECP was never that cheap, so this makes sense. They also kept only smallest, cheapest packages.

They seem determined at staying in that niche. Which might not be so bad, we just have to see prices.

 

[Brane2]

BTW, XO3L so far seems as heavy dissapointment.

Not only it wasn't improved WRT to XO2 ( there is nothing from what was promised), they removed FLASH block as it was done in XO2 and plopped in NCVM block, which is "multi-time progammable". Which means TWICE.

XO2 offered at least 10k programming cycles.
So not only is this not good for prototyping, now it's understandable why there is no talk about UFM.

It still might be that XO3L is just "Celeron" part of the spectrum and that fully fledged XO3 is to come.

But prices at Digikey do not seem to follow that reduction in silicon and resources. XO3L is priced around XO2 of the same declared size.

So far, this is dissapointing.
Unless this is just uverture to XO3 and Sapphire. One can still hope. ;o)

 

[Brane2]

Dne četrtek, 10. april 2014 21:10:16 UTC je oseba Brane2 napisala:

> Unless this is just uverture to XO3 and Sapphire. One can still hope. ;o)

This is freshly added on Lattice's site:

"The MachXO3-H FPGA family is a higher density family that will have more advanced features such as PCIe, Gigabit Ethernet, etc."

If this is not just some marketing BS maneuvering, there is still a glimpse of hope for the May announcement. )