{"id":661,"date":"2026-05-20T20:10:17","date_gmt":"2026-05-20T20:10:17","guid":{"rendered":"https:\/\/convly.ai\/rtx-5090-vs-rtx-5080-for-ai\/"},"modified":"2026-06-19T16:40:02","modified_gmt":"2026-06-19T16:40:02","slug":"rtx-5090-vs-rtx-5080-for-ai","status":"publish","type":"post","link":"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/","title":{"rendered":"RTX 5090 contro RTX 5080 per l'IA nel 2026: quale scheda Blackwell acquistare?"},"content":{"rendered":"<p>Inside NVIDIA&#8217;s Blackwell generation, AI builders face one clean decision: the <strong>RTX 5090<\/strong> o il <strong>RTX 5080<\/strong>. The 5090 costs roughly twice as much. It also has twice the VRAM. For AI, that second fact is the one that matters.<\/p>\n<p>The short answer: <strong>the 5080 is plenty for mainstream local AI; the 5090 exists for the people who need to run big.<\/strong><\/p>\n<div class=\"convly-tldr\">\n<h3>Punti chiave<\/h3>\n<ul>\n<li>The RTX 5090 has <strong>32 GB GDDR7<\/strong>; the RTX 5080 has <strong>16 GB<\/strong> \u2014 a 2x capacity gap.<\/li>\n<li>The 5090 is also <strong>~1.7\u20131.9x faster<\/strong> thanks to far more CUDA cores and bandwidth.<\/li>\n<li>Only the 5090 runs <strong>Llama 3 70B<\/strong> (4-bit) in VRAM; the 5080 cannot.<\/li>\n<li>The 5090 draws <strong>575 W<\/strong> and demands a 1000 W PSU; the 5080&#8217;s 360 W is far easier to build around.<\/li>\n<li>Buy the 5080 for 8B\u201313B models and image generation; buy the 5090 if you need 70B-class models or maximum speed.<\/li>\n<\/ul>\n<\/div>\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_84 counter-flat ez-toc-counter ez-toc-container-direction\">\n<label for=\"ez-toc-cssicon-toggle-item-6a38af04a6d73\" class=\"ez-toc-cssicon-toggle-label\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Attiva\/Disattiva<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #000000;color:#000000\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #000000;color:#000000\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewbox=\"0 0 24 24\" version=\"1.2\" baseprofile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/label><input type=\"checkbox\"  id=\"ez-toc-cssicon-toggle-item-6a38af04a6d73\"  aria-label=\"Attiva\/Disattiva\" \/><nav><ul class='ez-toc-list ez-toc-list-level-1' ><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#At_a_glance\" >At a glance<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#VRAM_decides_the_whole_comparison\" >VRAM decides the whole comparison<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#Speed_the_5090_is_also_simply_faster\" >Speed: the 5090 is also simply faster<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#Power_and_build_cost\" >Power and build cost<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#Who_should_actually_buy_the_5090\" >Who should actually buy the 5090?<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#The_real_cost_street_prices_and_electricity_over_time\" >The real cost: street prices and electricity over time<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#FAQ\" >Domande frequenti<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-8\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#Verdict\" >Verdict<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-9\" href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-5080-for-ai\/#Related_articles\" >Articoli correlati<\/a><\/li><\/ul><\/nav><\/div>\n<h2><span class=\"ez-toc-section\" id=\"At_a_glance\"><\/span>At a glance<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<table class=\"convly-vs\">\n<thead>\n<tr>\n<th>Specifiche<\/th>\n<th>RTX 5090<\/th>\n<th>RTX 5080<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Architettura<\/td>\n<td>Blackwell GB202<\/td>\n<td>Blackwell GB203<\/td>\n<\/tr>\n<tr>\n<td>Core CUDA<\/td>\n<td class=\"convly-vs-winner\">21,760<\/td>\n<td>10,752<\/td>\n<\/tr>\n<tr>\n<td>VRAM<\/td>\n<td class=\"convly-vs-winner\">32 GB GDDR7<\/td>\n<td>16 GB GDDR7<\/td>\n<\/tr>\n<tr>\n<td>Larghezza di banda della memoria<\/td>\n<td class=\"convly-vs-winner\">1.792 GB\/s<\/td>\n<td>~960 GB\/s<\/td>\n<\/tr>\n<tr>\n<td>FP16 Tensor (dense)<\/td>\n<td class=\"convly-vs-winner\">~419 TFLOPS<\/td>\n<td>~450 TFLOPS*<\/td>\n<\/tr>\n<tr>\n<td>TDP<\/td>\n<td>575 W<\/td>\n<td class=\"convly-vs-winner\">360 W<\/td>\n<\/tr>\n<tr>\n<td>Prezzo al pubblico consigliato (MSRP)<\/td>\n<td>$1,999<\/td>\n<td class=\"convly-vs-winner\">$999<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"font-size:.85rem;color:#6b6b6b;\">*Peak tensor TFLOPS figures vary by clock and sparsity mode; the 5090&#8217;s far larger core count makes it decisively faster in real workloads.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"VRAM_decides_the_whole_comparison\"><\/span>VRAM decides the whole comparison<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>For local AI, the question is never &#8220;how fast&#8221; before &#8220;does it fit.&#8221; Here the two cards split cleanly:<\/p>\n<ul>\n<li><strong>RTX 5080 \u2014 16 GB:<\/strong> runs <strong>Llama 3 8B<\/strong> at 8-bit, <strong>13B-class<\/strong> models at 4-bit, <strong>Stable Diffusion XL<\/strong> e <strong>Flux.1<\/strong>, and LoRA fine-tuning of 7B\u20138B models. It cannot hold a 70B model.<\/li>\n<li><strong>RTX 5090 \u2014 32 GB:<\/strong> does everything the 5080 does, plus runs <strong>Llama 3 70B<\/strong> at 4-bit (~40 GB? \u2014 see below), much longer context windows, larger fine-tunes, and big image and video models with room to spare.<\/li>\n<\/ul>\n<p>A clarification on 70B: a 70B model at Q4_K_M needs roughly 40 GB, which exceeds even 32 GB. But the 5090 runs 70B at more aggressive quantization (Q3\/IQ-class) fully in VRAM, and runs heavier quantizations with only light offload. The 5080, at 16 GB, is not in that conversation at all. For anything approaching 70B, the 5090 is the only consumer option.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Speed_the_5090_is_also_simply_faster\"><\/span>Speed: the 5090 is also simply faster<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Capacity aside, the 5090 has roughly <strong>double the CUDA cores<\/strong> e <strong>nearly double the memory bandwidth<\/strong>. That makes it much faster even on models that fit comfortably on both:<\/p>\n<table class=\"convly-vs\">\n<thead>\n<tr>\n<th>Workload<\/th>\n<th>RTX 5090<\/th>\n<th>RTX 5080<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Llama 3 8B Q4_K_M<\/td>\n<td class=\"convly-vs-winner\">~180 tok\/s<\/td>\n<td>~125 tok\/s<\/td>\n<\/tr>\n<tr>\n<td>Llama 3 13B-class Q4<\/td>\n<td class=\"convly-vs-winner\">~120 tok\/s<\/td>\n<td>~78 tok\/s<\/td>\n<\/tr>\n<tr>\n<td>SDXL 1024\u00d71024 (30 steps)<\/td>\n<td class=\"convly-vs-winner\">~25 it\/s<\/td>\n<td>~14 it\/s<\/td>\n<\/tr>\n<tr>\n<td>Llama 3 70B (quantized)<\/td>\n<td class=\"convly-vs-winner\">Runs in VRAM<\/td>\n<td>Does not fit<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Across workloads the 5090 lands around <strong>1.7\u20131.9x<\/strong> the 5080&#8217;s throughput \u2014 and on large models the comparison stops being about speed and becomes about possibility.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Power_and_build_cost\"><\/span>Power and build cost<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The performance comes at a real-world price beyond the MSRP. The 5090 draws <strong>575 W<\/strong>, demands a <strong>1000 W PSU<\/strong>, generates serious heat, and needs a case with genuine airflow. The 5080&#8217;s <strong>360 W<\/strong> is far gentler \u2014 an 850 W PSU and a normal mid-tower handle it easily. When you budget the 5090, budget the platform around it too.<\/p>\n<div class=\"convly-procons\">\n<div class=\"pros\">\n<h4>Choose the RTX 5090 if<\/h4>\n<ul>\n<li>You need to run 70B-class models locally<\/li>\n<li>You want maximum speed for image and video generation<\/li>\n<li>You do larger fine-tunes or need long context windows<\/li>\n<\/ul>\n<\/div>\n<div class=\"cons\">\n<h4>Scegli la RTX 5080 se<\/h4>\n<ul>\n<li>Your models are 8B\u201313B \u2014 the large majority of local AI<\/li>\n<li>You want a cooler, quieter, cheaper-to-build machine<\/li>\n<li>You would rather spend the $1,000 saved elsewhere<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<h2><span class=\"ez-toc-section\" id=\"Who_should_actually_buy_the_5090\"><\/span>Who should actually buy the 5090?<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Be honest about your workloads. If you run <strong>8B and 13B models<\/strong> and do Stable Diffusion, the 5080 handles all of it well \u2014 paying double for the 5090 buys speed you will enjoy but do not need. The 5090 earns its price for a specific user: someone who genuinely needs <strong>70B-class models<\/strong>, long contexts, or the fastest possible iteration on heavy generative work. For that person, no other consumer card competes. For everyone else, the 5080 is the rational pick.<\/p>\n<p><!--ai-enriched--><\/p>\n<h2><span class=\"ez-toc-section\" id=\"The_real_cost_street_prices_and_electricity_over_time\"><\/span>The real cost: street prices and electricity over time<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The sticker gap between these cards is wider in practice than the MSRPs suggest, and the purchase price is only the start of what a 24\/7 AI box actually costs you. Treat this as the part of the decision the spec sheet hides.<\/p>\n<p><strong>Street price, not MSRP.<\/strong> On paper the 5080 lists at $999 and the 5090 at $1,999 &#8211; a clean 2x. The 2026 GDDR7 memory shortage has broken that math. The 5080 has stayed comparatively close to MSRP, typically landing a few hundred dollars above $999, while 5090 board-partner cards routinely sell far above $2,000 &#8211; often 75% or more over MSRP, with the heavily-cooled models climbing higher still. The effective multiplier you pay has stretched well past 2x, frequently toward 3x. Always price the exact card in stock today; never budget off the launch MSRP.<\/p>\n<p><strong>Electricity is a recurring spec.<\/strong> The 5090&#8217;s 575W board power versus the 5080&#8217;s 360W is not just a PSU question &#8211; it is a monthly bill. For an always-on inference server, expect the 5090 to add a meaningful premium on your power bill over a year at typical US rates, and more in regions with expensive electricity. Idle draw is modest on both (the 5090 FE sits near 46W desktop-idle), so the cost only bites under sustained load.<\/p>\n<p><strong>You can claw most of it back.<\/strong> Inference is memory-bandwidth-bound, not compute-bound, so a power limit costs you far less speed than it saves in watts. Capping the 5090 around 400W typically sheds only single-digit-percent throughput while cutting roughly a third of the draw &#8211; the single highest-value tweak for a home AI rig.<\/p>\n<table class=\"convly-vs\">\n<tr>\n<th>Fattore costo<\/th>\n<th>RTX 5080 (16GB)<\/th>\n<th>RTX 5090 (32GB)<\/th>\n<\/tr>\n<tr>\n<td>Prezzo al pubblico consigliato (MSRP)<\/td>\n<td>$999<\/td>\n<td>$1,999<\/td>\n<\/tr>\n<tr>\n<td>Realistic 2026 street price<\/td>\n<td>Modestly above MSRP<\/td>\n<td>Well above MSRP<\/td>\n<\/tr>\n<tr>\n<td>Assorbimento della scheda<\/td>\n<td>360W<\/td>\n<td>575W<\/td>\n<\/tr>\n<tr>\n<td>Alimentatore consigliato<\/td>\n<td>850W<\/td>\n<td>1000W+<\/td>\n<\/tr>\n<tr>\n<td>Power-limit headroom<\/td>\n<td>Limitato<\/td>\n<td>~400W with ~10% speed loss<\/td>\n<\/tr>\n<\/table>\n<p>The takeaway: the 5090 is the more expensive card to <strong>buy and to run<\/strong>, and that running cost is permanent. If a 16GB card covers your models, the 5080 wins on lifetime cost by a wide margin.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"FAQ\"><\/span>Domande frequenti<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<h3>Is the RTX 5090 worth double the price of the 5080 for AI?<\/h3>\n<p>Only if you need its 32 GB of VRAM \u2014 for 70B-class models, long contexts, or big fine-tunes. If your work is 8B\u201313B models and image generation, the 5080 does it well and saves you $1,000.<\/p>\n<h3>Can the RTX 5080 run Llama 3 70B?<\/h3>\n<p>No. With 16 GB of VRAM it cannot hold a 70B model even heavily quantized. Running 70B locally requires the 32 GB RTX 5090 or a multi-GPU setup.<\/p>\n<h3>How much faster is the 5090 than the 5080?<\/h3>\n<p>Roughly 1.7\u20131.9x in real AI workloads, driven by nearly double the CUDA cores and memory bandwidth. On models too large for the 5080, the 5090 is not just faster \u2014 it is the only one that runs them.<\/p>\n<h3>Does the RTX 5090 need a special power supply?<\/h3>\n<p>Yes. It draws 575 W and NVIDIA recommends a 1000 W PSU. The 5080&#8217;s 360 W is satisfied by a standard 850 W unit, making it much simpler and cheaper to build around.<\/p>\n<h3>Can the RTX 5090 fine-tune models, or only run them?<\/h3>\n<p>It can do both, within limits. The 32GB of VRAM makes a single 5090 a capable home fine-tuning card for parameter-efficient methods like QLoRA on models up to roughly 30-40B parameters. A 70B QLoRA run needs closer to 48GB and will not fit on one card &#8211; that requires two 5090s (with PCIe interconnect overhead, since consumer Blackwell has no NVLink) or a rented data-center GPU. The 5080&#8217;s 16GB restricts you to QLoRA on smaller models, making it an entry-level fine-tuning card at best.<\/p>\n<h3>Should I buy now at inflated prices or wait?<\/h3>\n<p>If you need the hardware to earn or learn today, buy the card that fits your models and stop watching the ticker &#8211; GPU pricing in 2026 has been driven by a memory shortage with no clean end date. If your workload genuinely fits in 16GB, the 5080 is the far safer purchase at current prices because you are not overpaying for VRAM you will not use. Only stretch to a marked-up 5090 if 32GB unlocks a model or context length you cannot otherwise reach.<\/p>\n<h3>Are two RTX 5080s better than one RTX 5090?<\/h3>\n<p>No, not for most people. Two 16GB cards do not merge into a single 32GB pool &#8211; the memory stays split across the PCIe bus, so a model that needs more than 16GB must be sharded with real coordination overhead, and you still pay for two cards, two slots, and more power. A single 5090 gives you one contiguous 32GB space plus far higher bandwidth, which is simpler and faster for the large-model, long-context work that justifies the card in the first place.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Verdict\"><\/span>Verdict<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Il <strong>RTX 5090<\/strong> is the most capable consumer AI GPU in existence \u2014 32 GB of VRAM and class-leading speed make it the only card that brings 70B-class models within reach of a desktop. But it is a specialist&#8217;s tool. For the workloads most people actually run, the <strong>RTX 5080<\/strong> delivers everything needed at half the price and a fraction of the power and build complexity. Buy the 5090 because you need its memory \u2014 not because it is the flagship.<\/p>\n<p><!--related-block--><\/p>\n<div class=\"convly-related\">\n<h2><span class=\"ez-toc-section\" id=\"Related_articles\"><\/span>Articoli correlati<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<ul>\n<li><a href=\"https:\/\/convly.ai\/it\/glm-5-2-vs-kimi-k2-7-for-coding-2026\/\">GLM 5.2 contro Kimi K2.7 Code: quale codificatore open source vince?<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/deepseek-v4-vs-qwen3-2026\/\">DeepSeek V4 contro Qwen3.7 Max: lo scontro del 2026<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/openai-vs-anthropic-cold-war\/\">OpenAI vs Anthropic in 2026: Models, Philosophy, and Which to Use<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/rtx-5090-vs-rtx-4090-for-ai-2026\/\">RTX 5090 vs RTX 4090 for AI: Stable Diffusion, LLM Inference, and Training Benchmarks (2026)<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/rx-7900-xtx-vs-rtx-4090-for-ai\/\">AMD RX 7900 XTX contro RTX 4090 per l'IA nel 2026: ROCm pu\u00f2 competere?<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/rtx-5080-vs-rtx-4080-super-for-ai\/\">RTX 5080 contro RTX 4080 Super per l'IA nel 2026: un vero salto generazionale o semplicemente un aggiornamento marginale?<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/rtx-5070-ti-vs-rtx-4070-ti-super-for-ai\/\">RTX 5070 Ti contro RTX 4070 Ti Super per l'IA nel 2026: lo scontro nella fascia media<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/it\/rtx-4090-vs-rtx-3090-for-ai\/\">RTX 4090 contro RTX 3090 per l'IA nel 2026: vale davvero la pena aggiornare?<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>The RTX 5090 has double the VRAM of the 5080 and double the price. For AI, that VRAM gap decides everything \u2014 here&#8217;s which Blackwell card fits your work.<\/p>","protected":false},"author":1,"featured_media":673,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[246],"tags":[281,284,256,326,251,357],"class_list":["post-661","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-comparisons","tag-ai-gpu","tag-blackwell","tag-local-llm","tag-rtx-5080","tag-rtx-5090","tag-vram"],"_links":{"self":[{"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/posts\/661","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/comments?post=661"}],"version-history":[{"count":5,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/posts\/661\/revisions"}],"predecessor-version":[{"id":1209,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/posts\/661\/revisions\/1209"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/media\/673"}],"wp:attachment":[{"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/media?parent=661"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/categories?post=661"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/convly.ai\/it\/wp-json\/wp\/v2\/tags?post=661"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}