7 Series FPGAs

[forrestoff]

Hi,

Can someone explain the speed grades for the 7series? In the ds180 overview, it indicates a -2L grade; what does this mean?

Thanks.

[ywu]

The L at the end of speed grade number indicates "Lower" power.

[forrestoff]

Thanks for the response.

I understand that a speed grade is supposed to be relative and representitave metric of how fast an FPGA can perform. I don't understand why a speed grade is now badged another identifier that is not a measurment of speed. Assuming that -2L is not the same as -2, do you know which one is faster? Also, how is a -2L part differentiated between a -2 part by looking at the part number (pg. 12, ds180)?

It seems that it would have been much less confusing if they used a -1 and -2 speed grade, rather than this current scheme.

Sorry for my rant-ish response.

[danhoud]

The -nL speed grades are characterized for operation at lower Vccint.  This is reflected in the DC and AC characteristics in the datasheet.  -nL devices are guaranteed to operate as specified at .9V Vccint.  There is an attempt to make operation of a -1L and a -1 device the same, but this typically does not hold true across all parameters because of different VT operating curves for different fabric components.  A -nL device is preferred for applications where lower power is gained by sacrificing some speed performance.

[ywu]

I'm an engineer, so I don't want to get too much into the marketing territory. The simplest way to look at it is that -2L devices are screened -2 devices with lower leakage and operating at lower VCCINT (0.9v), which both result in lower power. There are no timing numbers for -2L devices in the datasheet yet, but if Virtex6 is any indication, -2L devices would be slightly slower than -2 but offer significant power reduction. You can take a look at WP389 for the complete 7 series power story. 

http://www.xilinx.com/support/documentation/white_papers/wp389_Lowering_Power_at_28nm.pdf

By the way, download Xilinx Document Navigator today to view all Xilinx documents.

---

forrestoff wrote:

Thanks for the response.

I understand that a speed grade is supposed to be relative and representitave metric of how fast an FPGA can perform. I don't understand why a speed grade is now badged another identifier that is not a measurment of speed. Assuming that -2L is not the same as -2, do you know which one is faster? Also, how is a -2L part differentiated between a -2 part by looking at the part number (pg. 12, ds180)?

It seems that it would have been much less confusing if they used a -1 and -2 speed grade, rather than this current scheme.

Sorry for my rant-ish response.

---

[forrestoff]

Thanks for the explanation.

I'm looking at a paragraph on page 9 and 10 of the WP389 and it suggests exactly what you mentioned--that the -2L parts are binned -2 parts with better leakage performance. Physics says that if you lower Vcc, your device is going to become slower (and just the opposite as well, within limits of course). The document agrees with this in Figure 2, pg. 5.

However, the document states the following, pg. 9, which is confusing if the above is correct:

"Devices that run at 0.9V are designated "-1L," based on their
speed grade. Their performance is similar to that of a -1 speed grade, but the “L”
signifies lower power."

This would suggest that the -1 and -1L are no different in terms of speed and that, perhaps, the -1L parts would perform faster than a -1 part if they were run at a nomal-power Vcc.

Do you think that's accurate to say?

Jumping back into the marketing territory--if you have a part that says it's lower power, why do you need to contain this information in the speed grade as well? This is just down-right confusing, but, agreed--probably only useful for a xilinx trivia match.

[forrestoff]

So for example, given a -1L device, would it operate like a -1 standard part at standard Vcc but also have the capability to be run at a lower Vcc for a hit in performance (below -1 performance)? This would make sense if it weren't for that paragraph on pg 9 on WP389.

[mcgett]

The -L devices are specified to run at 0.9V and all device specifications will only be guaranteed to work at this voltage. 

There will be separate speed files and data sheet specifications for these devices that will be the same for most parameters as the same speed grade in a regular 1.0V device, but there will be differences in some parameters.

White papers provided general information, but specifications are found in the data sheet.

[danhoud · Accepted solution]

Two points:

McGett's point regarding different performance designations, specifications and speeds files is absolutely correct,

i.e. a -1L is guaranteed to perform like a -1L, and performance when operated as a -1 is purely circumstantial.

You are approaching the binning strategy from the wrong direction.  A -1L is indeed a -1 which is fast enough to operate at similar performance when powered at 10% lower voltage.  This means that not all -1's can be binned to -1L requirements, and all -1Ls are not guaranteed by testing to operate at -1 requirements.  For 100% compliance to operating specs., they must be treated as independent product grades.

[forrestoff]

I understand, finally!  Thank you, all, for the clarification.