Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
11 years ago

Cyclone V vs Kintex 7 (Resources and Performance)

Hi guys,

I'm having a hard time deciding which of these FPGA's to select. I'm thinking between the Cyclone V E A7 vs the Kintex 7 160T, both of which have a pretty close amount of resources.

I'd like to pick up a chip that maximizes resource count and Fmax performance while staying within the web-editions (no expensive licenses) and within a good price range.

To test, I made a few large designs and had them compiled in Quartus II and ISE targeting both devices. I found that resource utilization is usually matched between both devices. My main concern is the Fmax reports from both; I find that Kintex 7 generally has a better Fmax than the Cyclone V (about Cyclone V's Fmax + 100) even though I'm using the same design for both.

For the Kintex 7 I'm reading the post-PAR clock analysis report.

For the Cyclone V, I'm picking the smaller of the two Fmax's in (0*C/85*C) from the TimeQuest Timing Analyzer after compilation.

For both devices, I varied the speed grades and compilation settings.

To the point: it looks like the Kintex 7 160T and the Cyclone V E A7 have very similar resources and are at a similar price. The Kintex 7 appears to show a higher Fmax in its timing report, but I'd like to make sure I'm doing an even comparison. For applications, assume it'll be mostly logic and BRAM (I'll consider the additional features in each myself, i.e. hard memory controllers, transceivers, etc)

My questions:

1) It the TimeQuest Timing Analyzer a good estimate to the Cyclone V's best possible clock rate?

2) I'm not sure, but I think I've read that the Cyclone V can have a high internal clock rate. I tried putting in PLL's but it looks like the TimeQuest Timing Analyzer reports an estimate similar to the input clock itself.

3) Does anyone think the Cyclone V E A9 will become supported in the web edition in a later version? The higher LE/$ on this chip is probably worthy of a compromise for performance.

Perhaps if anyone has experience with both devices, it'd be helpful to hear some opinions.

Thanks!

- Henry

P.S. I hope I do not cause a Xilinx v Altera war!

4 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    11 years ago

    away from the smoke of an Altera/Xilinx battle I have achieved 491MHz speed in part of a design in kintex 7, this should enter records.

    generally Xilinx devices seem faster when comparing like by like but certainly need a lot of floor planning.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    11 years ago

    --- Quote Start ---

    away from the smoke of an Altera/Xilinx battle I have achieved 491MHz speed in part of a design in kintex 7, this should enter records.

    generally Xilinx devices seem faster when comparing like by like but certainly need a lot of floor planning.

    --- Quote End ---

    Can you comment on the speed grade for the Kintex 7 and if you had actually tested it on a board? I tried a very simple and pipelined design and the best reports timing reports from ISE were around 425 MHz (speed grade -3). For the same design on the Cyclone V (speed grade -6), I got 320 MHz. This design is only about 10% utilization. In a larger design utilizing about 70%, Kintex 7 gets 295 MHz whereas the Cyclone V gets 199 MHz.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    11 years ago

    --- Quote Start ---

    Can you comment on the speed grade for the Kintex 7 and if you had actually tested it on a board? I tried a very simple and pipelined design and the best reports timing reports from ISE were around 425 MHz (speed grade -3). For the same design on the Cyclone V (speed grade -6), I got 320 MHz. This design is only about 10% utilization. In a larger design utilizing about 70%, Kintex 7 gets 295 MHz whereas the Cyclone V gets 199 MHz.

    --- Quote End ---

    Yes we tested the running hardware and it is bit true. However as project integration is updated from time to time we do get violations which are tackled by either further floor planning or extra pipeline.

    The speed grade is in my notes at work and I will let you know on Tuesday as it bank holiday here in UK

    We use vivado, not ISE
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    11 years ago

    I did some further checking and found the Kintex 7 to outperform the Arria V at -3 speed grade. I'm really considering whether the TimeQuest timing analyzer is accurate or if I'm misusing it. Can anyone comment if the figures shown in the 0*C/85*C timing models right after a compilation is the fmax to be looking at? If the Kintex 7 can outperform an Arria V, I'm beginning to think picking up the Kintex 7 is a no brainer, but I'm worried I might be doing this comparison the wrong way. I'll see if I can grab a Cyclone V to do some actual testing, but I don't actually have a Kintex 7 to physically compare.