[{"data":1,"prerenderedAt":1191},["ShallowReactive",2],{"blog-post-nl-\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens":3,"blog-posts-sidebar-nl":719},{"id":4,"title":5,"body":6,"categories":694,"date":705,"description":706,"extension":707,"heading":708,"image":709,"meta":710,"navigation":711,"originalUrl":712,"path":713,"seo":714,"slug":715,"stem":716,"updated":717,"__hash__":718},"blogNl\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens.md","Paiton MoE-benchmarks: MI300X versus H200 en B200",{"type":7,"value":8,"toc":683},"minimark",[9,17,24,27,34,37,40,46,49,154,160,165,168,360,368,374,377,383,387,390,394,397,402,405,417,421,525,530,536,546,552,555,581,587,590,610,616,620,627,630,640,645],[10,11,12,16],"p",{},[13,14,15],"strong",{},"Korte samenvatting:"," We hebben Paiton gebenchmarkt met onze nieuwe MoE-ondersteuning op Qwen\u002FQwen3-30B-A3B-Instruct-2507 om de inferentieprestaties in verschillende opstellingen te vergelijken. Elke configuratie werd vijf keer per batchgrootte uitgevoerd en we rapporteren het gemiddelde over de runs.",[18,19,21],"h3",{"id":20},"waarom-deze-benchmark",[13,22,23],{},"Waarom deze benchmark",[10,25,26],{},"De meeste gepubliceerde cijfers zijn gebaseerd op synthetische prompts of eenvoudige testdatasets. Wij richten ons, zoals altijd, op realistische conversationele workloads om het werkelijke gedrag van latency en throughput zichtbaar te maken. Het geteste model is Qwen\u002FQwen3-30B-A3B-Instruct-2507, uitgevoerd met Paiton's aangepaste kernelruntime.",[10,28,29,30,33],{},"Ondertussen vloeien miljarden naar glimmende NVIDIA-racks, terwijl wij ze ",[13,31,32],{},"verslaan"," met de goedkopere AMD MI300X van de vorige generatie.",[10,35,36],{},"Als uw doel betere resultaten per dollar is, en niet een bepaald logo per rackunit, betaalt u letterlijk een meerprijs om trager te gaan.",[10,38,39],{},"Terwijl anderen zich haasten om de nieuwste GPU's te kopen, richten wij ons op het ontsluiten van het volledige potentieel van zowel de huidige als de vorige generaties. GPU-leveranciers pushen vaak nieuwe hardware voordat de vorige generatie volledig is geoptimaliseerd en wij zijn hier om dat te veranderen.",[18,41,43],{"id":42},"methodologie",[13,44,45],{},"Methodologie",[10,47,48],{},"Om de geldigheid en toepasbaarheid van onze bevindingen te garanderen, hebben we ons aan een nauwgezette benchmarkingmethodologie gehouden:",[50,51,52,59,65,71,148],"ul",{},[53,54,55,58],"li",{},[13,56,57],{},"Getest model:"," We selecteerden Qwen\u002FQwen3-30B-A3B-Instruct-2507, een representatief groot MoE-model, zodat de resultaten relevant zijn voor hedendaagse LLM-deployments.",[53,60,61,64],{},[13,62,63],{},"Dataset:"," In plaats van synthetische gegevens hebben we realistische conversatiesporen uit de ShareGPT-dataset gebruikt. Deze keuze is van cruciaal belang voor het nauwkeurig beoordelen van de prestaties in scenario's die daadwerkelijke gebruikersinteracties nabootsen.",[53,66,67,70],{},[13,68,69],{},"Uitvoerlengte:"," Om de consistentie te behouden en de typische conversatielengtes weer te geven, werd de uitvoerlengte beperkt tot 256 tokens voor alle inferentieruns.",[53,72,73,76],{},[13,74,75],{},"Hardware- en softwareconfiguraties:",[50,77,78,103,126],{},[53,79,80,83],{},[13,81,82],{},"AMD MI300X (geoptimaliseerd voor Paiton):",[50,84,85,91,97],{},[53,86,87,90],{},[13,88,89],{},"ROCm-versies:"," Getest met zowel 6.4.1 als 7.0.0 om de impact van verschillende softwarestacks te evalueren.",[53,92,93,96],{},[13,94,95],{},"VRAM:"," 192 GB HBM, biedt aanzienlijke geheugencapaciteit voor grote modellen.",[53,98,99,102],{},[13,100,101],{},"Softwarestack:"," Paiton's aangepaste kernelruntime met MoE-ondersteuning, naast standaard vLLM v0.10.0 en stacks van concurrenten voor een volledige vergelijking.",[53,104,105,108],{},[13,106,107],{},"NVIDIA H200 (referentieplatform):",[50,109,110,116,121],{},[53,111,112,115],{},[13,113,114],{},"CUDA-versie:"," CUDA 13, vertegenwoordigt de nieuwste NVIDIA-softwareomgeving.",[53,117,118,120],{},[13,119,95],{}," 141 GB HBM.",[53,122,123,125],{},[13,124,101],{}," vLLM 0.10.2 en stacks van concurrenten.",[53,127,128,131],{},[13,129,130],{},"NVIDIA B200 (referentieplatform):",[50,132,133,138,143],{},[53,134,135,137],{},[13,136,114],{}," CUDA 13.",[53,139,140,142],{},[13,141,95],{}," 180 GB HBM.",[53,144,145,147],{},[13,146,101],{}," vLLM 0.10.2, gebouwd vanuit de broncode, en stacks van concurrenten.",[53,149,150,153],{},[13,151,152],{},"Benchmarkprocedure:"," Voor elke batchgrootte in de set {1, 2, 4, 8, 16, 24, 32, 64, 128, 256} werden inferentietaken vijf keer end-to-end uitgevoerd. Doorvoer (tokens\u002Fsec) en latentie werden nauwgezet geregistreerd, waarbij het gemiddelde van de runs werd gerapporteerd om de statistische variantie te verminderen. Cruciaal was dat de tokenisatie- en generatie-instellingen in alle configuraties identiek bleven om een eerlijke vergelijking te garanderen.",[18,155,157],{"id":156},"resultaten",[13,158,159],{},"Resultaten",[161,162,164],"h4",{"id":163},"analyse-van-doorvoer-en-kostenefficiëntie","Analyse van doorvoer en kostenefficiëntie",[10,166,167],{},"De volgende tabel toont de gemiddelde doorvoer (tokens\u002Fsec) die is waargenomen voor elke configuratie over verschillende batchgroottes:",[169,170,171,203],"table",{},[172,173,174],"thead",{},[175,176,177,183,188,193,198],"tr",{},[178,179,180],"th",{},[13,181,182],{},"Batchgrootte",[178,184,185],{},[13,186,187],{},"Paiton (MI300X) ROCm 6.4.1 \u002F vLLM 0.9.0",[178,189,190],{},[13,191,192],{},"MI300X ROCm 7.0 \u002F vLLM 0.10.0",[178,194,195],{},[13,196,197],{},"NVIDIA H200 CUDA 13 \u002F vLLM 0.10.2",[178,199,200],{},[13,201,202],{},"NVIDIA B200 CUDA 13 \u002F vLLM 0.10.2",[204,205,206,224,241,258,275,292,309,326,343],"tbody",{},[175,207,208,212,215,218,221],{},[209,210,211],"td",{},"1",[209,213,214],{},"189,37",[209,216,217],{},"162,24",[209,219,220],{},"189,51",[209,222,223],{},"180,11",[175,225,226,229,232,235,238],{},[209,227,228],{},"2",[209,230,231],{},"347,77",[209,233,234],{},"299,53",[209,236,237],{},"331,47",[209,239,240],{},"339,00",[175,242,243,246,249,252,255],{},[209,244,245],{},"4",[209,247,248],{},"580.07",[209,250,251],{},"496,88",[209,253,254],{},"551,72",[209,256,257],{},"610,16",[175,259,260,263,266,269,272],{},[209,261,262],{},"8",[209,264,265],{},"1.397,40",[209,267,268],{},"1.122,12",[209,270,271],{},"1.277,69",[209,273,274],{},"1.457,11",[175,276,277,280,283,286,289],{},[209,278,279],{},"16",[209,281,282],{},"2.855,09",[209,284,285],{},"1.989,90",[209,287,288],{},"2.222,61",[209,290,291],{},"2.840,19",[175,293,294,297,300,303,306],{},[209,295,296],{},"32",[209,298,299],{},"4.613,32",[209,301,302],{},"3.209,90",[209,304,305],{},"3.862,38",[209,307,308],{},"4.588,88",[175,310,311,314,317,320,323],{},[209,312,313],{},"64",[209,315,316],{},"7.234,05",[209,318,319],{},"5.573,63",[209,321,322],{},"6.687,79",[209,324,325],{},"8.368,82",[175,327,328,331,334,337,340],{},[209,329,330],{},"128",[209,332,333],{},"9.554,87",[209,335,336],{},"8.717,82",[209,338,339],{},"10.489,65",[209,341,342],{},"13.884,28",[175,344,345,348,351,354,357],{},[209,346,347],{},"256",[209,349,350],{},"14.672,35",[209,352,353],{},"12.587,27",[209,355,356],{},"16.129,35",[209,358,359],{},"21.980,31",[10,361,362],{},[363,364],"img",{"alt":365,"src":366,"title":367},"","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-3.jpg","Grafiek",[161,369,371],{"id":370},"waarnemingen",[13,372,373],{},"Waarnemingen",[10,375,376],{},"Hoewel de ruwe throughputcijfers over de hele linie competitief zijn, brengt een grondigere kostenanalyse belangrijke verschillen aan het licht. Met MoE-ondersteuning levert Paiton een robuuste throughput die soepel schaalt en ook bij hogere batchgroottes een voorspelbare latency behoudt. Sommige concurrerende stacks vertonen daarentegen meer variatie en mogelijk een lagere piekefficiëntie.",[18,378,380],{"id":379},"tokeneconomie",[13,381,382],{},"Tokeneconomie",[161,384,386],{"id":385},"kostenanalyse-op-aanvraag","Kostenanalyse op aanvraag",[10,388,389],{},"Voor een betekenisvolle vergelijking analyseerden we de kosten per 1 miljoen tokens, een cruciale metric voor productiedeployments. We combineerden de goedkoopste geloofwaardige on-demandprijs per GPU-uur voor elke GPU-familie met onze gemeten throughput bij batchgrootte 32. We kozen batchgrootte 32 omdat dit voor veel conversationele MoE-deployments een gangbaar werkpunt is.",[161,391,393],{"id":392},"transparante-formule","Transparante formule",[10,395,396],{},"De kosten per 1 miljoen tokens worden berekend met behulp van de volgende formule:",[10,398,399],{},[13,400,401],{},"$ per 1 miljoen tokens = (1.000.000 × R) \u002F (T × 3600)",[10,403,404],{},"Waar:",[50,406,407,412],{},[53,408,409],{},[13,410,411],{},"R = on-demand prijs van $\u002Fuur per GPU",[53,413,414],{},[13,415,416],{},"T = tokens\u002Fsec (per GPU)",[161,418,420],{"id":419},"on-demandresultaten-goedkoopste-geloofwaardige-aanbiedingen","On-demandresultaten (goedkoopste geloofwaardige aanbiedingen)",[169,422,423,452],{},[172,424,425],{},[175,426,427,432,437,442,447],{},[178,428,429],{},[13,430,431],{},"GPU",[178,433,434],{},[13,435,436],{},"Doorvoer (tokens\u002Fsec)",[178,438,439],{},[13,440,441],{},"$\u002FGPU\u002Fuur",[178,443,444],{},[13,445,446],{},"Tokens per $",[178,448,449],{},[13,450,451],{},"$ per 1 miljoen tokens",[204,453,454,472,489,507],{},[175,455,456,461,463,466,469],{},[209,457,458],{},[13,459,460],{},"Paiton MI300X",[209,462,299],{},[209,464,465],{},"$ 1,50",[209,467,468],{},"11.071.968",[209,470,471],{},"$ 0,090",[175,473,474,479,481,483,486],{},[209,475,476],{},[13,477,478],{},"Standaard AMD MI300X",[209,480,302],{},[209,482,465],{},[209,484,485],{},"7.703.760",[209,487,488],{},"$ 0,130",[175,490,491,496,498,501,504],{},[209,492,493],{},[13,494,495],{},"NVIDIA H200",[209,497,305],{},[209,499,500],{},"$ 2,59",[209,502,503],{},"5.368.559",[209,505,506],{},"$ 0,186",[175,508,509,514,516,519,522],{},[209,510,511],{},[13,512,513],{},"NVIDIA B200",[209,515,308],{},[209,517,518],{},"$ 3,75",[209,520,521],{},"4.405.325",[209,523,524],{},"$ 0,227",[10,526,527],{},[363,528],{"alt":365,"src":529,"title":367},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-4.jpg",[10,531,532],{},[533,534,535],"em",{},"Goedkoopste on-demand prijzen met batchgrootte 32 doorvoer → Paiton MI300X leidt op $\u002F1 miljoen tokens.",[10,537,538,541,542,545],{},[13,539,540],{},"Samenvatting voor beslissers:"," Paiton op MI300X komt uit op ongeveer ",[13,543,544],{},"$ 0,090 per 1 miljoen tokens",", aanzienlijk minder dan NVIDIA H200 (circa $ 0,186) en B200 (circa $ 0,227). Dit kostenverschil levert een tastbaar investeringsrendement op en is geen marginale verbetering.",[18,547,549],{"id":548},"waarom-paiton-wint",[13,550,551],{},"Waarom Paiton wint",[10,553,554],{},"De overtuigende prestaties en kostenefficiëntie van Paiton kunnen worden toegeschreven aan verschillende belangrijke architecturale en software-optimalisaties:",[50,556,557,563,569,575],{},[53,558,559,562],{},[13,560,561],{},"MoE-optimalisatie op kernelniveau:"," Paiton's aangepaste kernels maximaliseren de rekenintensiteit en geheugenlokaliteit. Door de overhead van generieke bibliotheken te omzeilen, bieden ze een efficiënter uitvoeringspad voor MoE-modellen, wat zich rechtstreeks vertaalt in betere prestaties.",[53,564,565,568],{},[13,566,567],{},"Strategische HBM-inzet op MI300X:"," De royale 192 GB High Bandwidth Memory (HBM) van de AMD MI300X wordt strategisch gebruikt door Paiton. Deze ruime geheugencapaciteit zorgt ervoor dat expertweights en key-value (KV) caches “hot” blijven in het geheugen, waardoor de behoefte aan dure geheugenswaps drastisch wordt verminderd. Dit is een cruciale factor bij het bereiken van duurzame hoge prestaties en bijgevolg lagere operationele kosten.",[53,570,571,574],{},[13,572,573],{},"Minder kernelaanroepen:"," Paiton combineert meerdere bewerkingen in grotere kernels die rekening houden met de experts. Daardoor zijn tijdens runtime veel minder kernellanceringen nodig. Dat verlaagt de latency, verbetert de GPU-bezetting en verhoogt de rekenintensiteit. Het resultaat is een hogere throughput en stabielere prestaties dan bij stacks die veel kleine kernels uitvoeren.",[53,576,577,580],{},[13,578,579],{},"Onafhankelijkheid van ROCm-versies:"," Paiton's aangepaste kernels zijn ontworpen om grotendeels onafhankelijk te blijven van wijzigingen tussen specifieke ROCm-versies. Dat biedt meer stabiliteit en flexibiliteit bij deployments, omdat de prestaties ook bij updates van de onderliggende ROCm-softwarestack consistent blijven.",[18,582,584],{"id":583},"reproduceerbaarheid-en-prijstransparantie",[13,585,586],{},"Reproduceerbaarheid en prijstransparantie",[10,588,589],{},"We streven naar transparantie en reproduceerbaarheid.",[50,591,592,598,604],{},[53,593,594,597],{},[13,595,596],{},"Doorvoer-\u002Flatentiegegevens:"," Alle onbewerkte gegevens zijn opgenomen in dit bericht.",[53,599,600,603],{},[13,601,602],{},"Formules:"," De formules die worden gebruikt voor kostenberekeningen worden expliciet vermeld.",[53,605,606,609],{},[13,607,608],{},"Prijzen:"," On-demand prijzen weerspiegelen de laagst geloofwaardige openbare vermeldingen die beschikbaar waren op het moment van deze benchmark.",[18,611,613],{"id":612},"conclusie",[13,614,615],{},"Conclusie",[161,617,619],{"id":618},"de-nieuwe-basislijn-voor-kosten-per-token","De nieuwe basislijn voor kosten per token",[10,621,622,623,626],{},"Voor organisaties die grote Mixture-of-Experts-modellen in productie inzetten, vormt Paiton op AMD MI300X de nieuwe ",[13,624,625],{},"referentie voor kosten per token",". De benchmarkresultaten bevestigen dat Paiton andere oplossingen achter zich laat wanneer modellen zoals Qwen3-30B-A3B-Instruct-2507 op schaal worden aangeboden. Dat voordeel geldt zowel voor on-demandclouddeployments als voor eigen on-premises infrastructuur. De combinatie van sterke prestaties en aanzienlijke kostenbesparingen maakt Paiton bijzonder waardevol voor een optimaal rendement op investeringen in LLM-inferentie.",[10,628,629],{},"Bovendien optimaliseren we onze oplossingen voortdurend en verwachten we in de nabije toekomst nog indrukwekkendere resultaten. Houd ons in de gaten voor aankomende updates, inclusief gedetailleerde FP8-resultaten, die we zeer binnenkort zullen posten.",[10,631,632,639],{},[633,634,638],"a",{"href":635,"rel":636},"https:\u002F\u002Fai.eliovp.com\u002Fpaiton",[637],"nofollow","Neem contact op om de mogelijkheden te bespreken"," :)",[10,641,642],{},[13,643,644],{},"Referenties",[646,647,648,655,662,669,676],"ol",{},[53,649,650],{},[633,651,654],{"href":652,"rel":653},"https:\u002F\u002Fwww.supermicro.com\u002Fen\u002Fproducts\u002Fsystem\u002Fgpu\u002F8u\u002Fas%20-8125gs-tnmr2",[637],"Supermicro GPU-systeem AS-8125GS-TNMR2",[53,656,657],{},[633,658,661],{"href":659,"rel":660},"https:\u002F\u002Fwww.amd.com\u002Fen\u002Fproducts\u002Faccelerators\u002Finstinct\u002Fmi300\u002Fmi300x.html",[637],"AMD Instinct MI300X",[53,663,664],{},[633,665,668],{"href":666,"rel":667},"https:\u002F\u002Fgithub.com\u002FROCm\u002Fvllm",[637],"ROCm\u002Fvllm GitHub",[53,670,671],{},[633,672,675],{"href":673,"rel":674},"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm",[637],"vllm-project\u002Fvllm GitHub",[53,677,678],{},[633,679,682],{"href":680,"rel":681},"https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-30B-A3B-Instruct-2507",[637],"Hugging Face Qwen3-30B-A3B-Instruct-2507",{"title":365,"searchDepth":684,"depth":684,"links":685},2,[686,688,689,690,691,692,693],{"id":20,"depth":687,"text":23},3,{"id":42,"depth":687,"text":45},{"id":156,"depth":687,"text":159},{"id":379,"depth":687,"text":382},{"id":548,"depth":687,"text":551},{"id":583,"depth":687,"text":586},{"id":612,"depth":687,"text":615},[695,696,697,698,699,700,701,702,703,513,495,697,704],"Alle","Kunstmatige intelligentie","Paiton","AI-benchmarks","AMD MI300X","Kosten per token","Inferentie-optimalisatie","Mixture of Experts","MoE","Qwen3","2025-09-26T13:36:18","Vergelijk Qwen3-30B-A3B MoE-benchmarks van MI300X met Paiton, H200 en B200: throughput en kosten per miljoen tokens.","md","Stop met te veel betalen: Paiton MI300X MoE verslaat H200\u002FB200 in kosten per 1 miljoen tokens","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fhulkvshulkpaitonwins.webp",{},true,"https:\u002F\u002Feliovp.com\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens\u002F","\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens",{"title":5,"description":706},"stop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","blog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens",null,"s16UGCZutwUBPNBr5EttIDbL2ow1wfFOQfdbWOAZxHU",[720,734,744,756,767,779,788,801,832,844,866,884,903,922,940,956,968,970,985,997,1006,1014,1029,1041,1052,1063,1073,1086,1096,1109,1120,1130,1141,1150,1162,1173,1182],{"path":721,"title":722,"description":723,"date":724,"slug":725,"image":726,"originalUrl":727,"categories":728},"\u002Fblog\u002Fpaiton-qwen38-mxfp4-dflash2-r9700","Qwen3.8: 400,7 tok\u002Fs op één R9700 | Paiton","Qwen3.8 op één R9700: 400,7 tok\u002Fs met ROCm 10 en vLLM 0.29, plus publieke 200K\u002F220K-chatprofielen. Benchmarks, beperkingen en startopdrachten.","2026-09-16T07:30:00Z","paiton-qwen38-mxfp4-dflash2-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen38-mxfp4\u002Fupdate-2026-09-19\u002Fhero.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-mxfp4-dflash2-r9700",[697,729,730,731,701,732,733],"AMD Radeon","Lokale AI","AI-inferentie","Grote taalmodellen","vLLM",{"path":735,"title":736,"description":737,"date":738,"slug":739,"image":740,"originalUrl":741,"categories":742},"\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700","Qwen3.8 GGUF in vLLM: sneller antwoord op één Radeon","Draai de originele NEO CODER MAX GGUF met Paiton in vLLM op een R9700. Bekijk de gemeten responstijden, beeldinvoer en lokale installatie.","2026-09-14T07:30:00Z","paiton-qwen38-neo-gguf-vllm-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-neo-gguf\u002F00-hero-neo-gguf-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700",[697,729,730,743,733],"GGUF",{"path":745,"title":746,"description":747,"date":748,"slug":749,"image":750,"originalUrl":751,"categories":752},"\u002Fblog\u002Fpaiton-minimax-h3-radeon-ai-pro-r9700","MiniMax H3 op Radeon: 15 seconden video met stereogeluid","Paiton genereert lokaal 15 seconden MiniMax H3-video met stereogeluid op één Radeon AI PRO R9700 in 5 min 33 s, met 16,7% minder wachttijd dan stock.","2026-09-09T07:30:00Z","paiton-minimax-h3-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-minimax-h3\u002F00-featured-minimax-h3-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-minimax-h3-radeon-ai-pro-r9700",[697,729,730,753,754,755],"Videogeneratie","MiniMax H3","ComfyUI",{"path":757,"title":758,"description":759,"date":760,"slug":761,"image":762,"originalUrl":763,"categories":764},"\u002Fblog\u002Fpaiton-flux2-klein-radeon-ai-pro-r9700","Lokale FLUX.2 klein op Radeon AI PRO R9700: sneller beelden genereren met minder VRAM","Paiton genereert FLUX.2 klein-beelden van 1024 × 1024 in 1,054 seconden op een R9700, met 16,2% minder generatietijd en 33,4% minder piekallocatie in Torch.","2026-09-07T09:00:00","paiton-flux2-klein-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-flux2-klein\u002Ffox-paiton.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-flux2-klein-radeon-ai-pro-r9700",[697,729,730,765,766,755],"Beeldgeneratie","FLUX",{"path":768,"title":769,"description":770,"date":771,"slug":772,"image":773,"originalUrl":774,"categories":775},"\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700","Ornith 1.5 haalt 44,6 tok\u002Fs op één Radeon AI PRO R9700","Paiton draait Ornith 1.5 35B A3B op één Radeon AI PRO R9700 met 44,63 outputtokens per seconde, 27% sneller en met 21,3% lagere gemodelleerde kosten.","2026-09-05T09:00:00","paiton-ornith15-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F00-featured-ornith15-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700",[697,696,729,731,776,777,701,732,733,778],"GPU-prestaties","Inferentielatentie","Kostenefficiëntie",{"path":780,"title":781,"description":782,"date":783,"slug":784,"image":785,"originalUrl":786,"categories":787},"\u002Fblog\u002Fpaiton-qwen38-radeon-ai-pro-r9700","Paiton: 21% meer Qwen3.8-doorvoer op Radeon AI PRO R9700","Paiton draait AMD's Qwen3.8 27B op één Radeon AI PRO R9700 met 39,77 outputtokens per seconde. Dat levert 21% meer throughput en 17,4% lagere gemodelleerde kosten op.","2026-09-04T09:00:00","paiton-qwen38-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-r9700\u002F00-featured-paiton-r9700.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-radeon-ai-pro-r9700",[697,696,729,731,776,777,701,732,733,778],{"path":789,"title":790,"description":791,"date":792,"slug":793,"image":794,"originalUrl":717,"categories":795},"\u002Fblog\u002Fai-data-center-power-requirements-gpu-per-megawatt","Stroomvereisten voor AI-datacenters: de GPU-per-MW-illusie","Waarom verschillen GPU-aantallen per megawatt? Lees hoe PUE, piekbelasting, opslag, netwerken en koeling de inzetbare AI-capaciteit bepalen.","2026-07-27T23:52:00","ai-data-center-power-requirements-gpu-per-megawatt","\u002Fasset\u002Fimages\u002Fblog\u002Fai-data-center-power-requirements-gpu-per-megawatt\u002Fgpu-per-megawatt-illusion.webp",[695,796,797,798,799,800],"AI-infrastructuur","Datacenters","ModFlex","HPC","AMD Helios",{"path":802,"title":803,"description":804,"date":805,"slug":806,"image":807,"originalUrl":808,"categories":809},"\u002Fblog\u002Fpaiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x","Wan2.2-videogeneratie: Paiton op AMD MI355X","Vergelijk Wan2.2-videogeneratie op AMD MI355X met Paiton en NVIDIA B200 via Diffusers. Lees hoe we diffusiemodellen optimaliseren.","2026-06-10T14:04:04","paiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonwan2.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x\u002F",[695,696,697,810,811,812,813,814,815,816,817,818,431,819,820,821,822,823,824,825,697,826,827,828,829,830,831],"14B","AMD","B200","Benchmarks","Blackwell","Compute","Diffusie","Eliovp","Generatieve AI","Hardware","Inferentie","Instinct","MI355x","NVIDIA","On-premises","Optimalisatie","Soevereine AI","T2V","Tekst-naar-video","Tuning","Video-generatie","Wan2.2",{"path":833,"title":834,"description":835,"date":836,"slug":837,"image":838,"originalUrl":839,"categories":840},"\u002Fblog\u002Ffrom-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure","ElioVP in De Tijd: chipoptimalisatie en datacenters","Lees hoe De Tijd ElioVP belicht, van de oorsprong in chipoptimalisatie tot het werk aan modulaire datacenters en koeling voor hoge vermogensdichtheid.","2026-02-10T20:48:12","from-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fphysicalnewspaper.webp","https:\u002F\u002Feliovp.com\u002Ffrom-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure\u002F",[695,696,841,842,811,843,841,823],"Modulaire DC","Niet gecategoriseerd","De Tijd",{"path":845,"title":846,"description":847,"date":848,"slug":849,"image":850,"originalUrl":851,"categories":852},"\u002Fblog\u002Fprivacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit","AI en privacy: een strategische prioriteit in de Benelux","Privacyrisico's van generatieve AI, vertrouwen, dataopslag en governance. Waarom bedrijven in de Benelux veilige AI strategisch moeten benaderen.","2026-01-29T13:51:11","privacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fheaderimage.webp","https:\u002F\u002Feliovp.com\u002Fprivacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit\u002F",[695,696,853,842,854,855,856,857,858,859,860,861,817,862,818,863,864,865],"Trending","AI Act","Antropomorfisme","AVG","Benelux","ChatGPT","Cyberbeveiliging","Databeheer","Gegevensbeveiliging","GDPR","Microsoft Copilot","Privacy","Shadow AI",{"path":867,"title":868,"description":869,"date":870,"slug":871,"image":872,"originalUrl":873,"categories":874},"\u002Fblog\u002Fitsme-bij-ons-is-het-its-not-me-en-dit-is-waarom","Waarom wij itsme niet gebruiken: privacy en soevereiniteit","Waarom ElioVP itsme niet gebruikt: onze afwegingen rond identiteitsmetadata, cloudafhankelijkheid, privacy en datasoevereiniteit.","2025-11-27T09:32:14","itsme-bij-ons-is-het-its-not-me-en-dit-is-waarom","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffrontimage.webp","https:\u002F\u002Feliovp.com\u002Fitsme-bij-ons-is-het-its-not-me-en-dit-is-waarom\u002F",[695,875,876,877,859,878,879,880,862,881,882,883,864],"AWS","Belgian Mobile ID","CLOUD Act","Datasoevereiniteit","Digitale identiteit","eIDAS","itsme","Liberty Global","MyGov.be",{"path":885,"title":886,"description":887,"date":888,"slug":889,"image":890,"originalUrl":891,"categories":892},"\u002Fblog\u002Ffield-report-the-reality-of-building-agentic-ai-in-2025","Praktijkrapport: de realiteit van Agentic AI bouwen in 2025","Praktijklessen over lokale AI-agents in 2025 gaan in op workflowontwerp, observability, modeltraining, hallucinaties en GPU-geheugenlimieten.","2025-11-25T14:03:39","field-report-the-reality-of-building-agentic-ai-in-2025","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffieldreport.webp","https:\u002F\u002Feliovp.com\u002Ffield-report-the-reality-of-building-agentic-ai-in-2025\u002F",[695,696,893,853,894,895,896,897,898,899,900,901,826,902],"Oplossingen","Agentic AI","AI-techniek","AI-strategie","Autonome agenten","Bedrijfs-AI","Lokale LLM","Modelverfijning","AI op locatie","VRAM-optimalisatie",{"path":904,"title":905,"description":906,"date":907,"slug":908,"image":909,"originalUrl":910,"categories":911},"\u002Fblog\u002Fthe-synthetic-unicorn-bubble","De synthetische unicornzeepbel","Een analyse van investeringsrisico’s bij AI-neoclouds: circulaire financiering, infrastructuurclaims, contractvoorwaarden en due diligence.","2025-11-24T19:22:28","the-synthetic-unicorn-bubble","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fsyntheticunicorn.webp","https:\u002F\u002Feliovp.com\u002Fthe-synthetic-unicorn-bubble\u002F",[695,696,853,912,913,914,915,916,917,918,919,920,921],"AI Infrastructure","AI Neocloud","Circulaire financiering","GPU Cloud","Beleggingsrisico's","Opstartwaardering","Synthetische bubbel","Technische analyse","Vaporware","Durfkapitaal",{"path":923,"title":924,"description":925,"date":926,"slug":927,"image":928,"originalUrl":929,"categories":930},"\u002Fblog\u002Fbuilding-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power","NVIDIA GB300 NVL72: modulair datacenter in vier maanden","Ontdek een modulair datacenterontwerp voor NVIDIA GB300 NVL72, met redundante voeding, hybride koeling en een bouwplanning van vier maanden.","2025-11-20T14:10:19","building-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fsuperpodmodflexfrontimage.webp","https:\u002F\u002Feliovp.com\u002Fbuilding-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power\u002F",[695,841,842,931,796,932,933,934,935,936,937,938,939],"150 kW-rack","DLC","Hoge dichtheid","Vloeistofkoeling","Modulair datacenter","NVIDIA Blackwell Ultra","NVIDIA GB300","NVL72","Snelle implementatie",{"path":941,"title":942,"description":943,"date":944,"slug":945,"image":946,"originalUrl":947,"categories":948},"\u002Fblog\u002Fwhy-cuda-translation-wont-unlock-amds-real-potential","CUDA-vertaling versus AMD-gerichte optimalisatie","Waarom CUDA-compatibiliteit niet hetzelfde is als AMD-prestaties: over ROCm, HIP, kerneloptimalisatie en hardwaregerichte afstemming.","2025-11-12T14:48:37","why-cuda-translation-wont-unlock-amds-real-potential","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fchatgpt-image-nov-11-2025-09_16_10-pm-1.webp","https:\u002F\u002Feliovp.com\u002Fwhy-cuda-translation-wont-unlock-amds-real-potential\u002F",[695,696,697,842,699,696,949,950,951,952,953,954,697,955],"CUDA-vertaling","FP8","GPU-optimalisatie","High-performance computing","HIP","Kerneltuning","ROCm",{"path":957,"title":958,"description":959,"date":960,"slug":961,"image":962,"originalUrl":963,"categories":964},"\u002Fblog\u002Fpaiton-the-simplest-way-to-supercharge-ai-inference","Paiton: snellere AI-inferentie in uw bestaande stack","Lees hoe Paiton aansluit op bestaande inferentiestacks, met AMD MI300X-benchmarks en vergelijkingen van prestaties per dollar.","2025-11-11T10:31:22","paiton-the-simplest-way-to-supercharge-ai-inference","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaiton-powaaah.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-the-simplest-way-to-supercharge-ai-inference\u002F",[695,696,697,731,965,699,778,966,701,954,697,967,733],"AMD Instinct","Hoge throughput","SGLang",{"path":713,"title":5,"description":706,"date":705,"slug":715,"image":709,"originalUrl":712,"categories":969},[695,696,697,698,699,700,701,702,703,513,495,697,704],{"path":971,"title":972,"description":973,"date":974,"slug":975,"image":976,"originalUrl":977,"categories":978},"\u002Fblog\u002Fagentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en","Lokale Agentic AI: van inbox naar actie","Lokale AI-agents zetten e-mails, documenten en beelden om in tickets, rapporten en acties, met modellen op maat van uw gegevens en systemen.","2025-09-16T13:09:00","agentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ffrontfotoblog.webp","https:\u002F\u002Feliovp.com\u002Fagentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en\u002F",[695,696,893,842,894,979,980,981,982,899,901,826,983,984],"Schadedetectie","Documentverwerking","E-mailautomatisering","Factuurextractie","Ticketautomatisering","Workflowautomatisering",{"path":986,"title":987,"description":988,"date":989,"slug":990,"image":991,"originalUrl":992,"categories":993},"\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","MI300X FP8-benchmarks: GPU-partitionering met Paiton","Bekijk hoe Paiton presteert met Llama 3.1 8B FP8 op gepartitioneerde MI300X-GPU's, vergeleken met NVIDIA H200 en B200.","2025-07-31T13:32:57","mi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-1.webp","https:\u002F\u002Feliovp.com\u002Fmi300x-fp8-data%e2%80%91parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach\u002F",[695,696,697,842,994,811,812,995,996,822,823,697,733],"AI","H200","MI300X",{"path":998,"title":999,"description":1000,"date":1001,"slug":1002,"image":1003,"originalUrl":1004,"categories":1005},"\u002Fblog\u002Fapplicable-ai-for-businesses","Toepasbare AI voor bedrijven","Ontdek hoe ElioVP lokale AI voor bedrijfsprocessen bouwt, met modeltraining op maat en automatische schadedetectie voor de logistiek.","2025-07-09T21:35:30","applicable-ai-for-businesses","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fscherm_afbeelding-2025-07-09-om-23.30.17.webp","https:\u002F\u002Feliovp.com\u002Fapplicable-ai-for-businesses\u002F",[695,696,893,894,979,980,981,982,899,901,826,983,984],{"path":1007,"title":1008,"description":1009,"date":1010,"slug":1011,"image":365,"originalUrl":1012,"categories":1013},"\u002Fblog\u002Fintroducing-paitons-free-evaluation-models","Maak kennis met de gratis evaluatiemodellen van Paiton","Test Paiton met gratis evaluatiemodellen voor AMD-GPU's. Vergelijk de prestaties voor tekst, beeldanalyse en beeldgeneratie met uw eigen workloads.","2025-07-07T11:26:13","introducing-paitons-free-evaluation-models","https:\u002F\u002Feliovp.com\u002Fintroducing-paitons-free-evaluation-models\u002F",[695,696,697],{"path":1015,"title":1016,"description":1017,"date":1018,"slug":1019,"image":1020,"originalUrl":1021,"categories":1022},"\u002Fblog\u002Fpaiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b","Llama 3.1 405B: sneller starten met Paiton op MI300X","Bekijk Paiton-benchmarks voor Llama 3.1 405B op acht AMD MI300X-GPU's, met opstarttijd, tensorparallelisme, throughput en latency.","2025-06-12T20:15:23","paiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fservingscreenshot.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b\u002F",[695,696,697,842,731,699,1023,950,1024,1025,1026,697,1027,1028],"Koude start","Grafiekcompilatie","Llama 3.1 405B","LLM-optimalisatie","Opstartlatentie","Tensor-parallellisme",{"path":1030,"title":1031,"description":1032,"date":1033,"slug":1034,"image":1035,"originalUrl":1036,"categories":1037},"\u002Fblog\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x","Paiton FP8 verslaat NVIDIA's H200 op AMD's MI300X","Vergelijk Paiton op AMD MI300X met NVIDIA H200 voor Llama 3.1 70B FP8: throughput, wachttijd tot het eerste token en latency bij diverse batchgroottes.","2025-06-08T19:12:40","paiton-fp8-beats-nvidias-h200-on-amds-mi300x","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fblognewfp8.webp","https:\u002F\u002Feliovp.com\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x\u002F",[695,696,697,842,699,1038,898,818,776,777,732,1025,1039,1040],"Koude startoptimalisatie","Model serving","vLLM-optimalisatie",{"path":1042,"title":1043,"description":1044,"date":1045,"slug":1046,"image":1047,"originalUrl":1048,"categories":1049},"\u002Fblog\u002Fmi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm","MI300X, H200, RX 7900 XTX en n300s: vLLM-benchmarks","Vergelijk MI300X, H200, RX 7900 XTX en Tenstorrent n300s met vLLM: throughput, gemodelleerde tokenkosten en hardwarebeperkingen voor Llama 3 8B.","2025-05-09T14:03:58","mi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fcomparisontenstor.webp","https:\u002F\u002Feliovp.com\u002Fmi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm\u002F",[695,696,697,893,842,811,996,823,1050,1051],"RX7900XTX","tenstorrent",{"path":1053,"title":1054,"description":1055,"date":1056,"slug":1057,"image":1058,"originalUrl":1059,"categories":1060},"\u002Fblog\u002Fclusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights","ClusterP&L: financiële modellen voor GPU-clusters","Ontdek hoe ClusterP&L kosten, rendement en investeringsscenario's voor GPU-clusters modelleert, met risicoanalyses en exporteerbare rapporten.","2025-05-03T10:52:22","clusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fcomparisonscenarios.webp","https:\u002F\u002Feliovp.com\u002Fclusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights\u002F",[695,696,841,893,812,995,1061,823,1062],"MI325X","P&L-calculator",{"path":1064,"title":1065,"description":1066,"date":1067,"slug":1068,"image":1069,"originalUrl":1070,"categories":1071},"\u002Fblog\u002Fcranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200","AMD MI300X versus NVIDIA H200: Qwen3-32B met Paiton","Vergelijk Qwen3-32B-benchmarks op AMD MI300X met Paiton en NVIDIA H200, met resultaten voor throughput, latency en hardwarekosten.","2025-05-02T21:10:30","cranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002F3ac59a73-2466-4422-b7e5-ef2e4a8ca58e.webp","https:\u002F\u002Feliovp.com\u002Fcranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200\u002F",[695,696,697,994,811,995,1072,823,697,733],"MI300",{"path":1074,"title":1075,"description":1076,"date":1077,"slug":1078,"image":1079,"originalUrl":1080,"categories":1081},"\u002Fblog\u002Fpower-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw","Modulaire datacenters voor NVIDIA NVL: 1 tot 2 MW","Ontdek modulaire datacenterontwerpen voor NVIDIA NVL-systemen, met aandacht voor vermogen, vloeistofkoeling, redundantie en uitrolplanning.","2025-05-02T14:09:59","power-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Feliovp_critical-1mw-pod_rev-2_transparent.webp","https:\u002F\u002Feliovp.com\u002Fpower-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw\u002F",[695,841,1082,912,933,799,934,935,1083,1084,938,1085],"1-2MW datacenter","NVIDIA Blackwell","NVIDIA NVL","Precisiekoeling",{"path":1087,"title":1088,"description":1089,"date":1090,"slug":1091,"image":1092,"originalUrl":1093,"categories":1094},"\u002Fblog\u002Fexamining-ai-agents-in-the-medical-field-ai-that-speaks-dicom","AI-agents in de medische wereld: AI die DICOM spreekt","Ontdek een lokale AI-agent die DICOM-gegevens opzoekt en bekijk tests van beeldmodellen met geanonimiseerde medische beelden.","2025-04-11T14:45:48","examining-ai-agents-in-the-medical-field-ai-that-speaks-dicom","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fhealthcareblog-1.webp","https:\u002F\u002Feliovp.com\u002Fexamining-ai-agents-in-the-medical-field-ai-that-speaks-dicom\u002F",[695,696,893,842,994,811,1095],"Zorg",{"path":1097,"title":1098,"description":1099,"date":1100,"slug":1101,"image":1102,"originalUrl":1103,"categories":1104},"\u002Fblog\u002Feliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs","Amerikaanse invoerheffingen en AI-leveringszekerheid in 2025","Lees ElioVP's visie uit april 2025 op Amerikaanse invoerheffingen en leveringszekerheid voor AI-servers, HPC-systemen en modulaire datacenters.","2025-04-04T10:01:27","eliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ftariffsshipping.webp","https:\u002F\u002Feliovp.com\u002Feliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs\u002F",[695,853,994,811,1105,1106,1107,1108],"Invoer","Taiwan","Invoerheffingen","Trump",{"path":1110,"title":1111,"description":1112,"date":1113,"slug":1114,"image":1115,"originalUrl":1116,"categories":1117},"\u002Fblog\u002Fwhy-ai-agents-are-the-future","Waarom AI-agenten de toekomst zijn","Ontdek AI-agents voor ERP, CRM, financiën en klantondersteuning, met praktijkvoorbeelden en een traject van procesanalyse tot pilot en uitrol.","2025-03-23T22:06:59","why-ai-agents-are-the-future","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ferp2.jpeg","https:\u002F\u002Feliovp.com\u002Fwhy-ai-agents-are-the-future\u002F",[695,696,893,994,1118,1119],"AI-agenten","ERP",{"path":1121,"title":1122,"description":1123,"date":1124,"slug":1125,"image":1126,"originalUrl":1127,"categories":1128},"\u002Fblog\u002Fthe-rise-of-open-source-ai-model-optimization","De opkomst van open-source AI-modeloptimalisatie","Verken trends in opensource-AI-optimalisatie: kwantisatie, mixture-of-experts-modellen, hardwaregerichte afstemming, RAG en edge-AI.","2025-03-22T20:59:23","the-rise-of-open-source-ai-model-optimization","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Friseofopensource.jpeg","https:\u002F\u002Feliovp.com\u002Fthe-rise-of-open-source-ai-model-optimization\u002F",[695,696,853,1129,811,431,823],"AI-nieuws",{"path":1131,"title":1132,"description":1133,"date":1134,"slug":1135,"image":1136,"originalUrl":1137,"categories":1138},"\u002Fblog\u002Fintroducing-our-benchmarking-tool-powered-by-dstack","Maak kennis met onze benchmarktool, gebouwd op dstack","Ontdek onze benchmarktool met dstack: herhaalbare vLLM-tests, automatische parameterreeksen en prestatierapporten voor lokale GPU's en de cloud.","2025-03-20T14:21:59","introducing-our-benchmarking-tool-powered-by-dstack","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fbenchmarktool.jpeg","https:\u002F\u002Feliovp.com\u002Fintroducing-our-benchmarking-tool-powered-by-dstack\u002F",[695,696,697,994,811,1139,1140,996,697],"benchmark","LLM",{"path":1142,"title":1143,"description":1144,"date":1145,"slug":1146,"image":1147,"originalUrl":1148,"categories":1149},"\u002Fblog\u002Foptimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200","QwQ-32B optimaliseren (door Qwen): AMD MI300X versus NVIDIA H200","Vergelijk throughput en latency van QwQ-32B op AMD MI300X met Paiton en NVIDIA H200, bij kleine batches en meer gelijktijdige aanvragen.","2025-03-19T21:41:44","optimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaiton4.jpeg","https:\u002F\u002Feliovp.com\u002Foptimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200\u002F",[695,696,697],{"path":1151,"title":1152,"description":1153,"date":1154,"slug":1155,"image":1156,"originalUrl":1157,"categories":1158},"\u002Fblog\u002Feliovp-featured-on-amd-tech-talk-podcast","Eliovp te gast in de AMD Tech Talk-podcast","Beluister Elio Van Puyvelde en Jim Greene in de AMD Tech Talk-podcast over het ontstaan van ElioVP en de hardware- en softwarediensten voor AI.","2025-03-19T07:53:39","eliovp-featured-on-amd-tech-talk-podcast","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Ftechtalkjimgreene.jpeg","https:\u002F\u002Feliovp.com\u002Feliovp-featured-on-amd-tech-talk-podcast\u002F",[695,811,1159,1160,1161],"Jim Greene","Podcast","Tech Talk",{"path":1163,"title":1164,"description":1165,"date":1166,"slug":1167,"image":1168,"originalUrl":1169,"categories":1170},"\u002Fblog\u002Ffurther-optimizing-amd-powered-inference-with-paiton","AMD-inferentie verder optimaliseren met Paiton","Bekijk Paiton-benchmarks voor DeepSeek R1 Distill Llama 8B op AMD MI300X, gericht op throughput en latency bij kleinere batchgroottes.","2025-03-13T06:18:30","further-optimizing-amd-powered-inference-with-paiton","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost3.webp","https:\u002F\u002Feliovp.com\u002Ffurther-optimizing-amd-powered-inference-with-paiton\u002F",[695,696,697,811,1171,1172,995,996,1061,697,733],"DeepSeek","H100",{"path":1174,"title":1175,"description":1176,"date":1177,"slug":1178,"image":1179,"originalUrl":1180,"categories":1181},"\u002Fblog\u002Fa-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b","Paiton-benchmarks: DeepSeek R1 Distill Llama 3.1 8B","Vergelijk standaard- en Paiton-versies van DeepSeek R1 Distill Llama 3.1 8B op AMD MI300X, met benchmarks voor throughput en latency.","2025-01-31T09:11:02","a-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost2.webp","https:\u002F\u002Feliovp.com\u002Fa-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b\u002F",[695,696,697,811,1171,1172,995,996,1061,697,733],{"path":1183,"title":1184,"description":1185,"date":1186,"slug":1187,"image":1188,"originalUrl":1189,"categories":1190},"\u002Fblog\u002Fai-model-optimization-with-paiton","AI-modeloptimalisatie met Paiton","Lees hoe Paiton modelcompilatie, aangepaste kernels en kernelfusie inzet om AI-inferentie op AMD GPU's te optimaliseren.","2025-01-30T19:53:25","ai-model-optimization-with-paiton","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fpaitonpost1.webp","https:\u002F\u002Feliovp.com\u002Fai-model-optimization-with-paiton\u002F",[695,696,697,811,1172,995,996,1061,697,733],1789853168185]