Spole\u010dnost Nvidia hled\u00e1 cesty, jak sn\u00ed\u017eit po\u017eadavky na pam\u011b\u0165 VRAM. Jednou z nich by mohla b\u00fdt neuronov\u00e1 komprese NTC, kter\u00e1 u\u0161et\u0159\u00ed i 85 % m\u00edsta.<\/p>\n
Pam\u011bti VRAM nen\u00ed skoro nikdy dost a spole\u010dnost Nvidia se sna\u017e\u00ed naj\u00edt zp\u016fsoby, kter\u00fdmi by mohla zajistit, \u017ee se toho do pam\u011bti vejde v\u00edce. To by mohlo pomoci zv\u00fd\u0161it v\u00fdkon u karet, kter\u00e9 maj\u00ed vzhledem ke sv\u00e9mu v\u00fdkonu men\u0161\u00ed mno\u017estv\u00ed pam\u011bti, teoreticky se nab\u00edz\u00ed GeForce RTX 5060 Ti 8GB<\/a>, RTX 5070 12GB<\/a> nebo RTX 5080 16GB<\/a>. Proto uk\u00e1zala novou technologii NTC (Neural Texture Compression), kter\u00e1 by m\u011bla b\u00fdt mnohem lep\u0161\u00ed ne\u017e kompresn\u00ed metody jako BCn.<\/p>\n \n\t\ti \n\t\ti Porovn\u00e1n\u00ed BCn a Nvidia NTC komprese<\/p>\n Nvidia natr\u00e9novala mal\u00e9 neuronov\u00e9 s\u00edt\u011b, kter\u00e9 jsou schopny ulo\u017eit pot\u0159ebn\u00e9 textury ve zlomku velikosti p\u016fvodn\u00edch dat, a to i ve srovn\u00e1n\u00ed s komprimovan\u00fdmi tradi\u010dn\u00edmi metodami. V\u00fdsledky prezentovala nap\u0159. na sc\u00e9n\u011b tosk\u00e1nsk\u00e9 vily, kde tradi\u010dn\u00ed BCn komprese \u00fadajn\u011b vy\u017eadovala 6,5 GB dat, zat\u00edmco NTC komprese jen 970 MB, tedy 85 % p\u016fvodn\u00edho mno\u017estv\u00ed. V\u00fdsledek je hodn\u011b podobn\u00fd, v\u00fdrazn\u011bj\u0161\u00edch rozd\u00edl\u016f si patrn\u011b nev\u0161imnete. Osobn\u011b mi v\u00fdsledek z NTC p\u0159i\u0161el b\u00fdt akor\u00e1t v n\u011bkter\u00fdch m\u00edstech barevn\u011b plo\u0161\u0161\u00ed (nap\u0159. kv\u011btin\u00e1\u010d vpravo od dve\u0159\u00ed je m\u00e9n\u011b barevn\u00fd, p\u016fsob\u00ed to jako siln\u011bj\u0161\u00ed barevn\u00e9 od\u0161um\u011bn\u00ed u fotek).<\/p>\n \t\t\tNvidia NTC<\/p>\n V dal\u0161\u00ed sc\u00e9n\u011b uk\u00e1zala rozd\u00edly p\u0159i stejn\u00e9m objemu dat, konkr\u00e9tn\u011b 970 MB\u00a0VRAM. Zde jde vid\u011bt, \u017ee NTC m\u00e1 mnohem bohat\u011bj\u0161\u00ed textury, v\u00edce detail\u016f i lep\u0161\u00ed barvy. Tady naopak BCn p\u016fsob\u00ed mnohem v\u00edce \u201cod\u0161um\u011bn\u011b\u201d, a to jasov\u011b i barevn\u011b (v p\u0159edchoz\u00ed sc\u00e9n\u011b byla NTC dohn\u00e1na mo\u017en\u00e1 a\u017e p\u0159\u00edli\u0161 daleko). Tv\u016frci je\u0161t\u011b dod\u00e1vaj\u00ed, \u017ee syst\u00e9m je deterministick\u00fd, tedy poka\u017ed\u00e9 textury zkomprimuje stejn\u011b, nejde tedy o AI syst\u00e9m, kter\u00fd m\u016f\u017ee poka\u017ed\u00e9 d\u00e1t jinou odpov\u011b\u010f.\u00a0<\/p>\n \t\t\tNvidia NTC<\/p>\n V\u00fdhodou nov\u00e9ho syst\u00e9mu NTC jsou nejen ni\u017e\u0161\u00ed n\u00e1roky na VRAM, ale tak\u00e9 to umo\u017en\u00ed zmen\u0161it velikost her, tedy i v\u00fdsledn\u00fdch instalac\u00ed nebo patch\u016f. Na \u00falo\u017ei\u0161t\u011b (HDD\/SSD) se tak vejde v\u00edce her, rychleji se mohou st\u00e1hnout z internetu. P\u0159i stejn\u00e9 kvalit\u011b zaberou textury v pam\u011bti m\u00e9n\u011b dat, resp. p\u0159i stejn\u00e9 spot\u0159eb\u011b VRAM mohou b\u00fdt textury detailn\u011bj\u0161\u00ed. A p\u0159\u00edpadn\u011b je mo\u017en\u00e9 zkombinovat ob\u011b v\u00fdhody najedou a o trochu sn\u00ed\u017eit spot\u0159ebu VRAM a o trochu zv\u00fd\u0161it detailnost textur.\u00a0<\/p>\n V\u00edce detail\u016f o t\u011bchto metod\u00e1ch najdete ve videu, kter\u00e9 Nvidia zve\u0159ejnila na sv\u00e9m YouTube kan\u00e1le.<\/p>\n","protected":false},"excerpt":{"rendered":"Spole\u010dnost Nvidia hled\u00e1 cesty, jak sn\u00ed\u017eit po\u017eadavky na pam\u011b\u0165 VRAM. Jednou z nich by mohla b\u00fdt neuronov\u00e1 komprese…\n","protected":false},"author":2,"featured_media":49287,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[7702,1902,74,73,78,75,77,76],"class_list":{"0":"post-49286","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-veda-a-technika","8":"tag-geforce-rtx","9":"tag-novinka","10":"tag-science","11":"tag-science-technology","12":"tag-technika","13":"tag-technology","14":"tag-veda","15":"tag-veda-a-technika"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@cz\/116370234824886538","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/posts\/49286","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/comments?post=49286"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/posts\/49286\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/media\/49287"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/media?parent=49286"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/categories?post=49286"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/cz\/wp-json\/wp\/v2\/tags?post=49286"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\n\t\tZdroj: YouTube (Nvidia)\n\t<\/p>\n
\n\t\tZdroj: YouTube (Nvidia)\n\t<\/p>\n