{"id":27749,"date":"2020-02-06T11:06:10","date_gmt":"2020-02-06T19:06:10","guid":{"rendered":"https:\/\/blog.digilentinc.com\/?p=27749"},"modified":"2020-12-21T11:17:02","modified_gmt":"2020-12-21T19:17:02","slug":"six-reasons-you-should-consider-fpgas-over-asics-or-cpu-gpus","status":"publish","type":"post","link":"https:\/\/digilent.com\/blog\/six-reasons-you-should-consider-fpgas-over-asics-or-cpu-gpus\/","title":{"rendered":"Six Reasons You Should Consider FPGAs Over ASICs or CPU\/GPUs\u00a0"},"content":{"rendered":"

In this blog post, we\u2019ll be focusing on why\u00a0<\/span>FPGAs hit the sweet spot\u00a0<\/span>for processing power<\/span>.<\/span>\u00a0<\/span>Before we dive too deeply into this conversation, let’s address the \u201celephant in the room\u201d –\u00a0<\/span>everything exists for a reason, even ASICs<\/span>\u00a0(<\/span>A<\/span>pplication\u00a0<\/span>S<\/span>pecific\u00a0<\/span>I<\/span>ntegrated\u00a0<\/span>C<\/span>ircuit)<\/span>, CPUs (<\/span>C<\/span>entral Processing\u00a0<\/span>U<\/span>nit), and GPUs (<\/span>G<\/span>raphics\u00a0<\/span>P<\/span>rocessing\u00a0<\/span>U<\/span>nit). ASIC<\/span>s provide incredible\u00a0<\/span>p<\/span>erformance<\/span>\u00a0per watt<\/span>, and are thus very efficient, but they are also much more challenging to\u00a0<\/span>design<\/span>\u00a0<\/span>and\u00a0<\/span>are more pricey<\/span>\u00a0<\/span>than\u00a0<\/span>using\u00a0<\/span>an FPGA<\/span>\u00a0(<\/span>F<\/span>ield\u00a0<\/span>P<\/span>rogrammable\u00a0<\/span>G<\/span>ate\u00a0<\/span>A<\/span>rray<\/span>). On the other end of the spectrum, CPUs\/GPUs have the edge\u00a0<\/span>in<\/span>\u00a0ease of programming and design but\u00a0<\/span>lack the efficiency of the others.<\/span>\u00a0<\/span><\/p>\n

Without any further ado, let\u2019s get to the list:<\/span>\u00a0<\/span><\/p>\n

1)\u00a0<\/span>FPGAs are\u00a0<\/span><\/b>Re<\/span><\/b>configur<\/span><\/b>able<\/span><\/b>\u00a0–\u00a0<\/span>The configurability of your average FPGA leaves ASICs in the dust.\u00a0<\/span>In addition to the hard\/soft IP cores\u00a0<\/span>(an example: Arm Core on our\u00a0<\/span>Arty Z7)\u00a0<\/span>that are configured for a specific application, the real value lies in being able to reconfigure\u00a0<\/span>(and reconfigure again)\u00a0<\/span>after<\/span><\/i>\u00a0installation \u2013 something that\u00a0<\/span>ASICs just can\u2019t do<\/span>.<\/span>\u00a0<\/span><\/p>\n

\"fpga<\/p>\n

 <\/p>\n

2)\u00a0<\/span>FPGAs Work\u00a0<\/span><\/b>i<\/span><\/b>n Parallel \u2013\u00a0<\/span><\/b>One of the benefits of FPGAs that make them such a good tool for working\u00a0<\/span>with measurement systems and other edge computing applications that require\u00a0<\/span>the processing of a large amount of data<\/span>\u00a0like embedded vision is that they\u00a0<\/span>are able to<\/span>\u00a0process in parallel.\u00a0<\/span>CPUs\/GPUs work sequentially, processing one piece at a time, but with a well-configured FPGA you\u2019ll be able to<\/span>\u00a0simultaneously\u00a0<\/span>intake and process the next ba<\/span>tch of information before the first\u00a0<\/span>batch<\/span>\u00a0is done<\/span>, giving a low latency.\u00a0<\/span>\u00a0<\/span><\/p>\n

3)\u00a0<\/span>FPGAS Perform Time-Critical Processing<\/span><\/b>\u00a0– With the\u00a0<\/span>aforementioned low<\/span>\u00a0<\/span>latency,\u00a0<\/span>engineers and developers are able to use FPGAs for applications that require\u00a0<\/span>very time-critical calculations; like software-defined radio, medical devices, and mil-aero systems.\u00a0<\/span>When\u00a0<\/span>you don\u2019t have to wait as long on\u00a0<\/span>the processor\u00a0<\/span>t<\/span>o<\/span>\u00a0complete a calculation<\/span>,\u00a0<\/span>the output can be much more accurate.<\/span>\u00a0ASICs have even less latency, but again, they are only for a single specific application<\/span>. For prototyping and design, FPGA is the more forgiving choice.\u00a0<\/span>\u00a0<\/span><\/p>\n

4)\u00a0<\/span>FPGAs Have Optimal Performance per Watt<\/span><\/b>\u00a0\u2013\u00a0<\/span><\/b>When compared with a\u00a0<\/span>CPU or GPU, you will be getting high<\/span>er<\/span>\u00a0performance per watt (though it is closer when using floating point arithmetic)\u00a0<\/span>with an FPGA.\u00a0<\/span>This low power consumption can be nearly 3 to 4 times less than that of a GPU. The operating cost of an ASIC is far and away the best, but the high initial cost (sometimes in the millions) does a lot to offset that.<\/span>\u00a0<\/span><\/p>\n

5) <\/span>No OS Overhead \u2013\u00a0<\/span><\/b>If the latency and computational power of a CPU\/<\/span>GPU would be comparable to an FPGA, the inside track i<\/span>s<\/span>\u00a0lost by the necessity of running an operating system.\u00a0<\/span>The OS brings down the\u00a0<\/span>processing cost efficiency, as resources need to be dedicated to it, increasing the power used and decreasing the compute power.<\/span>\u00a0<\/span><\/p>\n

6<\/span>)\u00a0<\/span>FPGAs are Essentially Blank Canvases<\/span><\/b>\u00a0\u2013\u00a0<\/span><\/b>While ASICs need to have their functionality\u00a0<\/span>before<\/span><\/i>\u00a0manufacturing and CPUs\/GPUs\u00a0<\/span>are optimized for a narrow set of applications<\/span>, an FPGA\u2019s blueprint is\u00a0<\/span>almost completely\u00a0<\/span>user<\/span>\u00a0<\/span>defined.\u00a0<\/span>With the right\u00a0<\/span>knowledge of HDLs (<\/span>H<\/span>ardware\u00a0<\/span>D<\/span>esign\u00a0<\/span>L<\/span>anguage), an engineer can configure the FPGA f<\/span>abric to tackle any function, and in a lot of cases, multiple function<\/span>s<\/span>. In addition,\u00a0<\/span>FPGA<\/span>s<\/span>\u00a0ha<\/span>ve<\/span>\u00a0a huge interface\u00a0<\/span>flexibility that\u00a0<\/span>has recently been advanced even more with the rise\u00a0<\/span>in popularity of SoCs<\/span>\u00a0<\/span>(S<\/span>ystem\u00a0<\/span>O<\/span>n<\/span>\u00a0<\/span>C<\/span>hip<\/span>; Xilinx\u2019s Zynq 7000 SoC can be found on our\u00a0<\/span>ZedBoard<\/span>)<\/span>,<\/span>\u00a0which<\/span>\u00a0<\/span>actually<\/span>\u00a0include<\/span>\u00a0a<\/span>\u00a0CPU<\/span>\u00a0alongside the FPGA.<\/span>\u00a0<\/span>\u00a0<\/span>\u00a0<\/span><\/p>\n

So, yes, there are applications that might be better suited for an ASIC, a CPU, or a GPU, but\u00a0<\/span>for engineers that are fluent in HDLs, the FPGA hits the mark for price, processing power, and configurability. For those that\u00a0<\/span>are more versed in languages like C, Java, and Python? Keep an eye on our new\u00a0<\/span>Eclypse<\/span><\/a>\u00a0Z7<\/a>, which\u00a0<\/span>has high level API that\u00a0<\/span>will allow\u00a0<\/span>for\u00a0<\/span>an easier interaction between FPGAs and\u00a0<\/span>software languages (currently C and C++ are supported, with more coming later).\u00a0<\/span>\u00a0<\/span><\/p>\n

0<\/span><\/a><\/div><\/div>
<\/div><\/div>
<\/div>","protected":false},"excerpt":{"rendered":"

Is an FPGA the right tool for you? Dive into six ways that will help you discern if an FPGA is the way to go.<\/p>\n","protected":false},"author":50,"featured_media":27900,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[4267,35,1563],"tags":[],"ppma_author":[4502],"class_list":["post-27749","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-featured","category-fpga","category-guide"],"jetpack_featured_media_url":"https:\/\/digilent.com\/blog\/wp-content\/uploads\/2020\/02\/2020-FPGAsale.png","jetpack_sharing_enabled":true,"authors":[{"term_id":4502,"user_id":50,"is_guest":0,"slug":"davidh","display_name":"David Horn","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/917c337136844f075c76fcf4a0c3b94aa8c225366009ebf63c08fcb9ce6d0e52?s=96&d=mm&r=g","1":"","2":"","3":"","4":"","5":"","6":"","7":"","8":"","9":"","10":""}],"post_mailing_queue_ids":[],"_links":{"self":[{"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/posts\/27749","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/users\/50"}],"replies":[{"embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/comments?post=27749"}],"version-history":[{"count":0,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/posts\/27749\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/media\/27900"}],"wp:attachment":[{"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/media?parent=27749"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/categories?post=27749"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/tags?post=27749"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/digilent.com\/blog\/wp-json\/wp\/v2\/ppma_author?post=27749"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}