Author Topic: Xilinx Spartan 7 seen  (Read 18482 times)

0 Members and 1 Guest are viewing this topic.

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: Xilinx Spartan 7 seen
« Reply #25 on: April 21, 2017, 07:18:01 pm »
Well these cost comparisons are always going to be like that. Historically FPGAs have been expensive so they were mostly used for prototyping and low volume/high margin products. The cost just didn't scale well into mass production and pretty soon a microcontroller or custom silicon becomes cheaper. When they say adding the processor only costs 19 cents they are probably suggesting that it is economical vs using a separate processor in a production device. It's pretty common for companies to quote component prices in the thousands or tens of thousands quantity.
 

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #26 on: April 21, 2017, 09:27:35 pm »
They're certainly want to convey an idea that it is economical, but why 19 cents, not 50, and not 7. What's the math behind this number?

Is Virtex-7 2000 economical because you can run multiple Microblazes on it. I don't think so.

BTW: Look at the suggested uses - single axis motor control, adaptive LED lighting, automotive data conversion. Who the hell is going to use FPGA for that when it can be done with a simple PIC for a fraction of the cost.

Looks like they're a little bit disconnected from reality. Hence the out-of-the blue 19 cents figure.

Still, Spartan-7 is faster and better than Spartan-6. If you want to use it for high speed, $18 might be considered economical, because there's nothing else at that price. I wish they had something for $10 or $5.

 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: Xilinx Spartan 7 seen
« Reply #27 on: April 21, 2017, 10:28:05 pm »
Beats me, I'm not a marketing guy, you'd have to talk to a one of them.
 

Offline kony

  • Regular Contributor
  • *
  • Posts: 242
  • Country: cz
Re: Xilinx Spartan 7 seen
« Reply #28 on: April 21, 2017, 11:39:46 pm »
One thing that anyone had not mentioned yet - it is basically Xilinx effort to cover all market segments served under the new (acrively supported and developed) toolchain, anything over this is just nice bonus thrown in.
 

Offline hamster_nzTopic starter

  • Super Contributor
  • ***
  • Posts: 2803
  • Country: nz
Re: Xilinx Spartan 7 seen
« Reply #29 on: August 19, 2017, 05:54:30 am »
After a long wait, Spartan 7 devboards are now shipping - XC7S25-CSGA324 or XC7S50-CSGA324, $89 or $109.

The fall either side of the XC7A35T-1CPG236C used on most Artix-7 boards, (e.g 80 or 120 DSP slices vs 90 on the Artix 7).

https://blog.digilentinc.com/new-product-the-arty-s7/ & http://store.digilentinc.com/arty-s7-spartan-7-fpga-for-makers-and-hobbyists/
Gaze not into the abyss, lest you become recognized as an abyss domain expert, and they expect you keep gazing into the damn thing.
 

Offline Sal Ammoniac

  • Super Contributor
  • ***
  • Posts: 1662
  • Country: us
Re: Xilinx Spartan 7 seen
« Reply #30 on: August 19, 2017, 07:07:19 am »
The Digilent Spartan 7 board is $10 more expensive than the Artix version, but the Spartan 7 they use has about 50% more logic cells/slices.
Complexity is the number-one enemy of high-quality code.
 

Offline MattSR

  • Regular Contributor
  • *
  • Posts: 95
  • Country: au
Re: Xilinx Spartan 7 seen
« Reply #31 on: August 19, 2017, 07:17:34 am »
How does the largest Artix-7 compare to the largest Spartan-7?
 

Offline hamster_nzTopic starter

  • Super Contributor
  • ***
  • Posts: 2803
  • Country: nz
Re: Xilinx Spartan 7 seen
« Reply #32 on: August 19, 2017, 08:37:48 am »
How does the largest Artix-7 compare to the largest Spartan-7?

Not saying anything, but for the currently available parts (7S25 and 7S50)....

If you compare the numbers in table 5 (Typical Quiescent Supply Current) of https://www.xilinx.com/support/documentation/data_sheets/ds189-spartan-7-data-sheet.pdf with table 5 in https://www.xilinx.com/support/documentation/data_sheets/ds181_Artix_7_Data_Sheet.pdf, they look identical.

Humm.... don't the same size dice have the same static power>

And if you compare the size of artix-7 and spartan-7 bitstream files https://www.xilinx.com/support/documentation/user_guides/ug470_7Series_Config.pdf, they look identical

Humm.... they is an equal number of configuration bits as their Artix-7 equivalents?

And then if pick a random timing spec, like table 31 (DSP48E timing), they look identical for the matching speed grades.

Humm... looks like they are using the same process?

Looks like they are doppelgangers - but at least they have different JTAG IDCODEs :)

I also recently saw this on Hack-a-day: http://hackaday.com/2016/05/11/dirt-cheap-dirty-decapping/
« Last Edit: August 19, 2017, 08:40:31 am by hamster_nz »
Gaze not into the abyss, lest you become recognized as an abyss domain expert, and they expect you keep gazing into the damn thing.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 4509
  • Country: au
    • send complaints here
Re: Xilinx Spartan 7 seen
« Reply #33 on: August 19, 2017, 10:07:09 am »
How does the largest Artix-7 compare to the largest Spartan-7?

Not saying anything, but for the currently available parts (7S25 and 7S50)....

If you compare the numbers in table 5 (Typical Quiescent Supply Current) of https://www.xilinx.com/support/documentation/data_sheets/ds189-spartan-7-data-sheet.pdf with table 5 in https://www.xilinx.com/support/documentation/data_sheets/ds181_Artix_7_Data_Sheet.pdf, they look identical.

Humm.... don't the same size dice have the same static power>

And if you compare the size of artix-7 and spartan-7 bitstream files https://www.xilinx.com/support/documentation/user_guides/ug470_7Series_Config.pdf, they look identical

Humm.... they is an equal number of configuration bits as their Artix-7 equivalents?

And then if pick a random timing spec, like table 31 (DSP48E timing), they look identical for the matching speed grades.

Humm... looks like they are using the same process?
Spartan 6 came in parts with and without transceivers, so it looks like the thats now the only differentiator between 7 series Spartan and Artix. The common dies for the XC7A12T/XC7A25T and XC7A15T/XC7A35T/XC7A50T could make for some interesting hacking (and has been mentioned before on here) but the config sizes suggests that the XC7S6/XC7S15 parts are a different cut so I don't see any possibility to take a cheap spartan and unlock lots of resources.
 

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #34 on: August 19, 2017, 02:48:01 pm »
looks like the thats now the only differentiator between 7 series Spartan and Artix

Spartans are supposed to be more power efficient and also they're slower. The Spartan prices on DigiKey are not that great :(
 

Offline JoeN

  • Frequent Contributor
  • **
  • Posts: 991
  • Country: us
  • We Buy Trannies By The Truckload
Re: Xilinx Spartan 7 seen
« Reply #35 on: August 19, 2017, 08:06:13 pm »
I also recently saw this on Hack-a-day: http://hackaday.com/2016/05/11/dirt-cheap-dirty-decapping/

Interesting.

"Unless pre-arranged, all decapping images are added to an open source die database under a Creative Commons 0 license. Decapped chips cannot be returned. "

Anyone know where this database is?  I can't locate it.
Have You Been Triggered Today?
 

Offline hamster_nzTopic starter

  • Super Contributor
  • ***
  • Posts: 2803
  • Country: nz
Re: Xilinx Spartan 7 seen
« Reply #36 on: August 19, 2017, 08:32:41 pm »
I also recently saw this on Hack-a-day: http://hackaday.com/2016/05/11/dirt-cheap-dirty-decapping/

Interesting.

"Unless pre-arranged, all decapping images are added to an open source die database under a Creative Commons 0 license. Decapped chips cannot be returned. "

Anyone know where this database is?  I can't locate it.

Have Tweeted them... will let you know what they say!
Gaze not into the abyss, lest you become recognized as an abyss domain expert, and they expect you keep gazing into the damn thing.
 

Offline JoeN

  • Frequent Contributor
  • **
  • Posts: 991
  • Country: us
  • We Buy Trannies By The Truckload
Re: Xilinx Spartan 7 seen
« Reply #37 on: August 19, 2017, 08:37:33 pm »
I also recently saw this on Hack-a-day: http://hackaday.com/2016/05/11/dirt-cheap-dirty-decapping/

Interesting.

"Unless pre-arranged, all decapping images are added to an open source die database under a Creative Commons 0 license. Decapped chips cannot be returned. "

Anyone know where this database is?  I can't locate it.

Have Tweeted them... will let you know what they say!

Thanks.  I could have done that I guess but I have never had business with them and I thought someone might know.  Maybe this service never took off and they don't have enough to make a serious showing yet.  It's cheaper than a professional job, but more than most hobbyists are willing to pay I think.  I'd like to acquire some acid and try it myself the more professional way.
Have You Been Triggered Today?
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 4509
  • Country: au
    • send complaints here
Re: Xilinx Spartan 7 seen
« Reply #38 on: August 19, 2017, 10:07:50 pm »
How does the largest Artix-7 compare to the largest Spartan-7?

Not saying anything, but for the currently available parts (7S25 and 7S50)....

If you compare the numbers in table 5 (Typical Quiescent Supply Current) of https://www.xilinx.com/support/documentation/data_sheets/ds189-spartan-7-data-sheet.pdf with table 5 in https://www.xilinx.com/support/documentation/data_sheets/ds181_Artix_7_Data_Sheet.pdf they look identical.

Humm.... don't the same size dice have the same static power>

And if you compare the size of artix-7 and spartan-7 bitstream files https://www.xilinx.com/support/documentation/user_guides/ug470_7Series_Config.pdf they look identical

Humm.... they is an equal number of configuration bits as their Artix-7 equivalents?

And then if pick a random timing spec, like table 31 (DSP48E timing), they look identical for the matching speed grades.

Humm... looks like they are using the same process?
Spartan 6 came in parts with and without transceivers, so it looks like the thats now the only differentiator between 7 series Spartan and Artix...
Spartans are supposed to be more power efficient and also they're slower. The Spartan prices on DigiKey are not that great :(
Except as shown above, within the 7 series they have the same quiescent power requirements as the same sized Artix parts and the same switching speeds, so they're neither more efficient or slower than the Artix product line. It looks like the only difference is the availability of the high speed transceivers, which used to be available in the Spartan branding of the previous product 6 series line.

Spartan and Artix together are a grouping of the low speed parts, and are indeed slower and lower power than the Kintex range for comparable sized parts.
 

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #39 on: August 20, 2017, 02:27:54 am »
Except as shown above, within the 7 series they have the same quiescent power requirements as the same sized Artix parts and the same switching speeds, so they're neither more efficient or slower than the Artix product line. It looks like the only difference is the availability of the high speed transceivers, which used to be available in the Spartan branding of the previous product 6 series line.

You're right. There's no grade -3 for Spartans, but grades -2 and -1 are identical to Artix.

Given the same speed, they're priced a little bit less, which is not bad if you don't use PCIe and GTH transceivers anyway:

$73.29 for XC7S50-2FGGA484I
$101. 99 for XC7A50T-2FGG484I

However, if XC7A50T is the same as XC7A15T, then

$52.36 for XC7A15T-2FGG484I

looks even better.

 

Offline asmi

  • Super Contributor
  • ***
  • Posts: 2728
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #40 on: August 22, 2017, 07:19:40 pm »
I also remember hearing that smaller devices will be available in BGA-196 1.0 mm pitch package which was specifically designed to be fully routed out on a 4 layer board.
Initially Xilinx announced even QFP-144 package, but it was quietly cancelled somewhere down the road.
Still can't wait to get my hands on these chips to play around with them! Manufacturing 6-layer PCBs is significantly more expensive at the moment, so I limit myself to 4-layer boards, and that package sounds like a great deal for me.

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2758
  • Country: us
Re: Xilinx Spartan 7 seen
« Reply #41 on: August 22, 2017, 09:08:24 pm »
I also remember hearing that smaller devices will be available in BGA-196 1.0 mm pitch package which was specifically designed to be fully routed out on a 4 layer board.
Initially Xilinx announced even QFP-144 package, but it was quietly cancelled somewhere down the road.
Still can't wait to get my hands on these chips to play around with them! Manufacturing 6-layer PCBs is significantly more expensive at the moment, so I limit myself to 4-layer boards, and that package sounds like a great deal for me.
If you don't need the performance of the Spartan 7, Xilinx's Spartan 3A is still quite affordable at about $11 each for the smallest device, in QFP-144 package.  You can also get the 3AN with internal configuration PROM for about $13.

I'll keep using these until they stop making them.  I've got a project I did in the XC3S50AN with 32-bit counters running at 150 MHz, and I don't think that was anywhere near the limit.

Jon
 

Offline asmi

  • Super Contributor
  • ***
  • Posts: 2728
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #42 on: August 22, 2017, 09:15:47 pm »
If you don't need the performance of the Spartan 7, Xilinx's Spartan 3A is still quite affordable at about $11 each for the smallest device, in QFP-144 package.  You can also get the 3AN with internal configuration PROM for about $13.

I'll keep using these until they stop making them.  I've got a project I did in the XC3S50AN with 32-bit counters running at 150 MHz, and I don't think that was anywhere near the limit.

Jon
No, I don't want to use such old ICs. I like new and shiny stuff :) Besides, I really like Vivado, and the fact that Microblaze softcore is free for all 7 series devices.

Offline Mr. Scram

  • Super Contributor
  • ***
  • Posts: 9810
  • Country: 00
  • Display aficionado
Re: Xilinx Spartan 7 seen
« Reply #43 on: August 22, 2017, 10:02:25 pm »
...on what looks to be a new Digilent Dev board



Video says parts are "available for order"
Did anyone notice the job title? "Cost-Optimized Portfolio Product Line Manager" With a title that long, you'd think the job must really stink :D
 

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #44 on: August 22, 2017, 10:24:10 pm »
I also remember hearing that smaller devices will be available in BGA-196 1.0 mm pitch package which was specifically designed to be fully routed out on a 4 layer board.
Initially Xilinx announced even QFP-144 package, but it was quietly cancelled somewhere down the road.
Still can't wait to get my hands on these chips to play around with them! Manufacturing 6-layer PCBs is significantly more expensive at the moment, so I limit myself to 4-layer boards, and that package sounds like a great deal for me.

If you don't mind leaving some close-to-center pins unconnected, you can get more usable pins from bigger packages (such as 484-pin BGA) on the 4-layer board than you would get from the 196-pin BGA.
 
The following users thanked this post: Someone

Offline asmi

  • Super Contributor
  • ***
  • Posts: 2728
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #45 on: August 22, 2017, 10:54:43 pm »
If you don't mind leaving some close-to-center pins unconnected, you can get more usable pins from bigger packages (such as 484-pin BGA) on the 4-layer board than you would get from the 196-pin BGA.
What's the point in spending money on a more expensive 484 package if you can only break out about half of those pins using cheap 0.15/0.15/0.3 process? Even with 256 package you can't break it out completely (I tried when I was designing this board).

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #46 on: August 22, 2017, 11:25:38 pm »
What's the point in spending money on a more expensive 484 package if you can only break out about half of those pins using cheap 0.15/0.15/0.3 process? Even with 256 package you can't break it out completely (I tried when I was designing this board).

The point is to save money by using 4-layer design instead of 6-layers. Of course, this only makes sense if you save on board production more than you have to pay for a bigger package.

For example, partial fanout of 484-BGA on 4-layer board may give you more pins and cost less than full fanout of 256-BGA on 6-layer board. Or 676-BGA on 4-layer board may give you more pins and cost less than 484-BGA on 6-layer board.
 

Offline asmi

  • Super Contributor
  • ***
  • Posts: 2728
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #47 on: August 22, 2017, 11:59:53 pm »
The point is to save money by using 4-layer design instead of 6-layers. Of course, this only makes sense if you save on board production more than you have to pay for a bigger package.

For example, partial fanout of 484-BGA on 4-layer board may give you more pins and cost less than full fanout of 256-BGA on 6-layer board. Or 676-BGA on 4-layer board may give you more pins and cost less than 484-BGA on 6-layer board.
I see what you mean but I'm not sure I approve such waste of resources. Not only 484 packages are significantly more expensive, but there are also cases where you need to route out specific pin groups (for example if you want to have DDR2/3 memory device, you are quite limited in selection of pins).

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #48 on: August 23, 2017, 12:47:53 am »
I see what you mean but I'm not sure I approve such waste of resources.

It depends on what are resources for you. If you save money by ordering a cheaper PCB, or if you save time by simplifying your routing job, this seems like resources are being saved, not wasted.

Pins? I don't think you really save them. When you buy a Xilinx part in a smaller package, it has all the same pins as the same part in a bigger package (e.g. XC7A15T-2FGG484C is exactly the same as XC7A15T-2FTG256C), only some of the pins are not bonded out.

Not only 484 packages are significantly more expensive

$20-30 isn't a big deal compared to $200 saved on PCB fabrication. Of course, if you go to production, the economy changes. But this is no brainer for prototyping.

, but there are also cases where you need to route out specific pin groups (for example if you want to have DDR2/3 memory device, you are quite limited in selection of pins).

DDR pins are abundant in 7-series. There are some pins that are in the middle, such as JTAG or config pins. But these are always the same regarding of the package, and always in the middle.

May be if Xilinx thought about people who's going to fan out their FPGAs, it would be easier with smaller parts. But it certainly looks like they locate their pins specifically to make fanouts as difficult as they can - all the big customers will have gazillion-layer boards anyway.
 

Offline asmi

  • Super Contributor
  • ***
  • Posts: 2728
  • Country: ca
Re: Xilinx Spartan 7 seen
« Reply #49 on: August 23, 2017, 01:06:42 am »
$20-30 isn't a big deal compared to $200 saved on PCB fabrication. Of course, if you go to production, the economy changes. But this is no brainer for prototyping.
It still offends my feelings as engineer, so I will stick to 256 unless I will actually need more pins :)

DDR pins are abundant in 7-series. There are some pins that are in the middle, such as JTAG or config pins. But these are always the same regarding of the package, and always in the middle.
It's not that simple. For memory interfaces there are quite strict rules as to which pins can be used. I haven't actually tried making a MIG design with 484 pin part to see if I can get away with only using pins from 3 outer "layers" of pins, but my gut feeling suggests it won't be possible because most DQ groups contain pins that are "deeper" than outer 3 rows, and you can only user pins from the single bank column for controller.

May be if Xilinx thought about people who's going to fan out their FPGAs, it would be easier with smaller parts. But it certainly looks like they locate their pins specifically to make fanouts as difficult as they can - all the big customers will have gazillion-layer boards anyway.
They did actually think of it, and even issued a document which outlines what's the best way to break out their ICs, and how many layers do they recommend depending on what PCB process you're going to use. When I was routing the board mentioned above, I had very little issues with routing. I especially like how they placed power/gnd pins next to each other in such a way that it's easier to place 0402 caps next to them under the package.


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf