[{"data":1,"prerenderedAt":2082},["ShallowReactive",2],{"blog-post-nl-\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach":3,"blog-posts-sidebar-nl":1524},{"id":4,"title":5,"body":6,"categories":1500,"date":1510,"description":1511,"extension":1512,"heading":1513,"image":1514,"meta":1515,"navigation":1516,"originalUrl":1517,"path":1518,"seo":1519,"slug":1520,"socialImage":1521,"stem":1522,"updated":1521,"video":1521,"__hash__":1523},"blogNl\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach.md","MI300X FP8-benchmarks: GPU-partitionering met Paiton",{"type":7,"value":8,"toc":1488},"minimark",[9,13,22,25,28,34,80,83,88,99,106,109,249,252,258,261,269,277,280,291,299,302,308,311,319,322,328,334,337,346,349,356,362,369,375,378,381,387,390,396,402,405,411,417,420,426,433,436,439,444,447,452,455,461,466,776,782,786,978,984,995,1000,1003,1006,1092,1105,1110,1146,1152,1155,1267,1272,1277,1299,1305,1308,1311,1314,1321,1338,1344,1370,1376,1379,1382,1385,1388,1394,1397,1406,1409,1417,1423],[10,11,12],"p",{},"Bij ElioVP willen we AI-inferentie tot het uiterste drijven en elke prestatie in een plug-and-play-runtime verpakken.",[10,14,15,16,21],{},"In ",[17,18,20],"a",{"href":19},"\u002Fnl\u002Fblog\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x","ons vorige artikel"," toonde Paiton's FP8-pipeline op AMD's MI300X duidelijk betere resultaten dan NVIDIA's H200. Daarna gingen we opnieuw aan de slag.",[10,23,24],{},"Deze keer draaien we Llama-3.1-8B-Instruct-FP8-KV, een lichtere FP8-gekwantiseerde Llama-variant, niet op 8 GPU's maar op 64 virtuele GPU's die uit één MI300X-server zijn gepartitioneerd.",[10,26,27],{},"Met vLLM en Paiton's kerneloptimalisaties verwachtten we een bescheiden winst bij het opschalen van multi-tenancy. Het resultaat was veel sterker dan verwacht en kwam dicht bij een gelijkspel met NVIDIA's B200.",[10,29,30],{},[31,32,33],"strong",{},"Waarom hebben we dit gedaan?",[35,36,37,44,50,56,62,68,74],"ul",{},[38,39,40,43],"li",{},[31,41,42],{},"Maximale benutting",": Partitioneer de hardware zodat elke tenant alleen betaalt voor precies het VRAM en de rekenkracht die nodig zijn.",[38,45,46,49],{},[31,47,48],{},"Elastische multi-tenancy",": Start geïsoleerde vGPU's in enkele seconden en vermijd vertraging door noisy neighbours en concurrentie om resources.",[38,51,52,55],{},[31,53,54],{},"Granulaire SLA's",": Pas de QoS per slice aan, ultralage latentie voor chatbots, bulkdoorvoer voor batchtaken, zonder te jongleren met hardware.",[38,57,58,61],{},[31,59,60],{},"Kostenefficiënt schalen",": Stem uw computeomgeving en budget nauwkeurig af door kleinere GPU-partities te huren in plaats van de volledige GPU.",[38,63,64,67],{},[31,65,66],{},"Snelle CI\u002FCD-provisioning",": Integreer GPU-slices in uw pipeline voor directe A\u002FB-tests, blue-green deployments en regressiebenchmarks.",[38,69,70,73],{},[31,71,72],{},"Foutisolatie",": Beperk OOM-fouten en driverproblemen tot één slice, zodat één problematische taak niet de volledige server stillegt.",[38,75,76,79],{},[31,77,78],{},"Toekomstbestendige flexibiliteit",": Herpartitioneer wanneer nodig om nieuwe modelgroottes of kwantisatieformaten te ondersteunen, zonder ingrijpende hardware-upgrade.",[10,81,82],{},"Met die bouwstenen op hun plaats onderzochten we hoe ver Paiton inferentie op een gepartitioneerde MI300X kon opschalen. De resultaten spreken voor zich.",[10,84,85],{},[31,86,87],{},"Doelen",[35,89,90,93,96],{},[38,91,92],{},"Evalueer de schaalbaarheid van Paiton op MI300X bij gebruik van GPU-partitionering.",[38,94,95],{},"Meet de latentie en doorvoer van Llama 3.1 8B in FP8-formaat met behulp van vLLM.",[38,97,98],{},"Valideer de geheugenefficiëntie en kernelfusievoordelen van plug-and-play Paiton-modellen.",[100,101,103],"h3",{"id":102},"benchmarkopstelling-en-methodologie",[31,104,105],{},"Benchmarkopstelling en methodologie",[10,107,108],{},"Onze benchmarkingmethode volgt een duidelijke reeks regels en stappen. Dit zorgt ervoor dat onze tests open en reproduceerbaar zijn.",[35,110,111,128,168,179,195,211,217],{},[38,112,113,116,117],{},[31,114,115],{},"Hardwareconfiguratie",":\n",[35,118,119,122,125],{},[38,120,121],{},"8x AMD MI300X",[38,123,124],{},"8 x NVIDIA H200",[38,126,127],{},"8 x NVIDIA B200",[38,129,130,116,133],{},[31,131,132],{},"Inferentiebibliotheek",[35,134,135,144,151,160],{},[38,136,137,138],{},"AMD MI300X (Paiton): ",[17,139,143],{"href":140,"rel":141},"https:\u002F\u002Fgithub.com\u002FROCm\u002Fvllm",[142],"nofollow","vLLM v0.9.0",[38,145,146,147],{},"AMD MI300X (AITER): ",[17,148,150],{"href":140,"rel":149},[142],"v0.9.2",[38,152,153,154,159],{},"NVIDIA H200: ",[17,155,158],{"href":156,"rel":157},"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm",[142],"v0.10.0"," (V1-modus)",[38,161,162,163,167],{},"NVIDIA B200: ",[17,164,166],{"href":156,"rel":165},[142],"v0.10.1"," (V1-modus, vanuit de broncode gebouwd voor ondersteuning van de B200-architectuur)",[38,169,170,173,174],{},[31,171,172],{},"Taalmodel:"," ",[17,175,178],{"href":176,"rel":177},"https:\u002F\u002Fhuggingface.co\u002Famd\u002FLlama-3.1-8B-Instruct-FP8-KV",[142],"Llama-3.1-8B-Instruct-FP8-KV",[38,180,181,116,184],{},[31,182,183],{},"Driverstapel",[35,185,186,189,192],{},[38,187,188],{},"AMD MI300X: ROCm 6.4.2",[38,190,191],{},"NVIDIA H200: CUDA 12.8.1",[38,193,194],{},"NVIDIA B200: CUDA 12.8.1",[38,196,197,200],{},[31,198,199],{},"Framework:",[35,201,202,205,208],{},[38,203,204],{},"AMD MI300X: Torch 2.7.1+rocm6.3",[38,206,207],{},"NVIDIA H200: Torch 2.7.1+cu128",[38,209,210],{},"NVIDIA B200: Torch 2.9.0.dev+cu128",[38,212,213,216],{},[31,214,215],{},"Batchgrootte",": 1024",[38,218,219,222,223],{},[31,220,221],{},"Meetprotocol",": elke benchmark werd 10 keer uitgevoerd en de cijfers die we rapporteren zijn algemene gemiddelden. Dit helpt het effect van tijdelijke systeemwijzigingen te verminderen. Onze zorgvuldige meetstappen omvatten:\n",[35,224,225,231,237,243],{},[38,226,227,230],{},[31,228,229],{},"Opstarttijden",": belangrijk om te controleren hoe lang het duurt om het model te laden en het systeem gereed te maken.",[38,232,233,236],{},[31,234,235],{},"Cold-Start TTFT (Time to First Token)",": meet hoe lang het duurt vanaf een nieuw verzoek totdat het eerste gegenereerde token verschijnt. Dit is van cruciaal belang voor hoe snel interactieve toepassingen reageren.",[38,238,239,242],{},[31,240,241],{},"Steady-State TTFT",": Controleert de TTFT nadat het systeem stabiel heeft gedraaid, waarbij typische prestaties bij constant gebruik worden weergegeven.",[38,244,245,248],{},[31,246,247],{},"End-to-end-latencymetrics",": geven een volledig beeld van de tijd voor een compleet inferentieverzoek, vanaf het verzenden van de input tot de uiteindelijke output.",[10,250,251],{},"Deze gedetailleerde methode biedt een goede manier om de specifieke prestatiegegevens van Paiton te controleren in drukke, gepartitioneerde GPU-omgevingen.",[100,253,255],{"id":254},"gegevensparallellisme-zonder-partitionering",[31,256,257],{},"Gegevensparallellisme zonder partitionering",[10,259,260],{},"Eerst probeerden we vLLM's ingebouwde optie “–data-parallel-size” te gebruiken. Al snel bleek dat die niet meteen werkte en ingrijpende aanpassingen zou vereisen. Daarom kozen we een andere aanpak.",[10,262,263,264,268],{},"Om de benchmarks uit te voeren over 8 containers met behulp van vLLM, hebben we eerst de officiële NGINX-loadbalancing-gids gevolgd (",[17,265,266],{"href":266,"rel":267},"https:\u002F\u002Fdocs.vllm.ai\u002Fen\u002Fstable\u002Fdeployment\u002Fnginx.html",[142],")",[270,271,272],"ol",{},[38,273,274],{},[31,275,276],{},"NGINX-configuratie",[10,278,279],{},"Dit is de load-balancingconfiguratie die we gebruikten in \u002Fetc\u002Fnginx\u002Fnginx.conf:",[281,282,288],"pre",{"className":283,"code":285,"language":286,"meta":287},[284],"language-text","upstream backend {\n    least_conn;\n    server vllm0:8000 max_fails=3 fail_timeout=10000s;\n    server vllm1:8000 max_fails=3 fail_timeout=10000s;\n    server vllm2:8000 max_fails=3 fail_timeout=10000s;\n    server vllm3:8000 max_fails=3 fail_timeout=10000s;\n    server vllm4:8000 max_fails=3 fail_timeout=10000s;\n    server vllm5:8000 max_fails=3 fail_timeout=10000s;\n    server vllm6:8000 max_fails=3 fail_timeout=10000s;\n    server vllm7:8000 max_fails=3 fail_timeout=10000s;\n}\n\nserver {\n    listen 80;\n    location \u002F {\n        proxy_pass http:\u002F\u002Fbackend;\n        proxy_set_header Host $host;\n        proxy_set_header X-Real-IP $remote_addr;\n        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;\n        proxy_set_header X-Forwarded-Proto $scheme;\n    }\n}\n","text","",[289,290,285],"code",{"__ignoreMap":287},[270,292,294],{"start":293},2,[38,295,296],{},[31,297,298],{},"De Docker-containers lanceren",[10,300,301],{},"We hebben het volgende script gebruikt om 8 containers te starten met behulp van incrementele apparaat- en poortnummers:",[281,303,306],{"className":304,"code":305,"language":286,"meta":287},[284],"#!\u002Fbin\u002Fbash\n\necho \"Starting vLLM containers with incremental configuration…\"\n\nfor i in {0..7}; do\n    device_num=$((128 + (i * 8)))\n    device_path=\"\u002Fdev\u002Fdri\u002FrenderD${device_num}\"\n    port=$((8080 + i))\n    container_name=\"vllm${i}\"\n\n    echo \"Starting container ${container_name} on port ${port} with device ${device_path}…\"\n\n    docker run -itd \\\n        –ipc host \\\n        -v \u002Fdata:\u002Fdata \\\n        –network vllm_nginx \\\n        -e VLLM_ROCM_USE_AITER=True \\\n        -e HF_HOME=root\u002F.cache\u002Fhuggingface \\\n        -e HF_HUB_CACHE=\u002Froot\u002F.cache\u002Fhuggingface\u002Fhub \\\n        –device=\u002Fdev\u002Fkfd \\\n        –device=${device_path} \\\n        –group-add video \\\n        -p ${port}:8000 \\\n        –name ${container_name} \\\n        rocm\u002Fvllm:latest \\\n        vllm serve \\\n        amd\u002FLlama-3.1-8B-Instruct-FP8-KV \\\n        –num-scheduler-steps 10 \\\n        –kv-cache-dtype fp8 \\\n        –max-model-len 4096\n\n    if [ $? -eq 0 ]; then\n        echo \"✓ Container ${container_name} started successfully\"\n    else\n        echo \"✗ Failed to start container ${container_name}\"\n    fi\n\n    echo \"—\"\ndone\n\necho \"All containers started. Summary:\"\necho \"Containers: vllm0 through vllm7\"\necho \"Ports: 8081 through 8088\"\necho \"Devices: renderD128 through renderD184 (in steps of 8)\"\n",[289,307,305],{"__ignoreMap":287},[10,309,310],{},"De regel device_num=$((128 + (i * 8))) was nodig door achtergebleven renderdevicevermeldingen in \u002Fdev\u002Fdri\u002F na eerdere GPU-partitionering. Ook na het resetten van de partities keerden de device-ID's niet terug naar hun oorspronkelijke waarden. Daarom moesten we elk devicepad verschuiven om naar de beschikbare rendernodes te verwijzen.",[270,312,314],{"start":313},3,[38,315,316],{},[31,317,318],{},"Benchmarking",[10,320,321],{},"Ten slotte hebben we de volgende opdracht uitgevoerd om alle containers te benchmarken:",[281,323,326],{"className":324,"code":325,"language":286,"meta":287},[284],"for i in {1..10}; do\n    echo \"=== Running benchmark iteration $i\u002F10 ===\"\n    python3 ~\u002Fvllm\u002Fbenchmarks\u002Fbenchmark_serving.py \\\n      –backend vllm \\\n      –model amd\u002FLlama-3.1-8B-Instruct-FP8-KV \\\n      –dataset-name sharegpt \\\n      –dataset-path ~\u002Fvllm\u002FShareGPT_V3_unfiltered_cleaned_split.json \\\n      –num-prompts 1024 \\\n      –random-range-ratio 1.0 \\\n      –percentile-metrics ttft,tpot,itl,e2el \\\n      –sharegpt-output-len 256\n    echo \"=== Completed iteration $i\u002F10 ===\"\n    echo\ndone\n",[289,327,325],{"__ignoreMap":287},[100,329,331],{"id":330},"gegevensparallellisme-met-partitionering",[31,332,333],{},"Gegevensparallellisme met partitionering",[10,335,336],{},"De eerste cruciale stap was het partitioneren van onze GPU's.",[10,338,339,340,345],{},"Dit was heel eenvoudig en gemakkelijk te doen volgens ",[17,341,344],{"href":342,"rel":343},"https:\u002F\u002Finstinct.docs.amd.com\u002Fprojects\u002Famdgpu-docs\u002Fen\u002Flatest\u002Fgpu-partitioning\u002Findex.html",[142],"de officiële documentatie van AMD",".",[10,347,348],{},"Stappen:",[270,350,351],{},[38,352,353],{},[31,354,355],{},"Stel de rekenpartities in.",[281,357,360],{"className":358,"code":359,"language":286,"meta":287},[284],"sudo amd-smi set –gpu all –compute-partition CPX\n",[289,361,359],{"__ignoreMap":287},[270,363,364],{"start":293},[38,365,366],{},[31,367,368],{},"Stel de geheugenpartities in.",[281,370,373],{"className":371,"code":372,"language":286,"meta":287},[284],"sudo amd-smi set –memory-partition NPS4\n",[289,374,372],{"__ignoreMap":287},[10,376,377],{},"Wacht een paar seconden en klaar!",[10,379,380],{},"Resultaat:",[10,382,383],{},[384,385],"img",{"alt":287,"src":386},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Frocmoverview.webp",[10,388,389],{},"Klaar voor gebruik!",[10,391,392,393,268],{},"Zoals vermeld in de vorige sectie hebben we, om de benchmarks uit te voeren over meerdere containers met behulp van vLLM, eerst de officiële NGINX load-balancing-handleiding gevolgd (",[17,394,266],{"href":266,"rel":395},[142],[270,397,398],{},[38,399,400],{},[31,401,276],{},[10,403,404],{},"Hier is de taakverdelingsconfiguratie die we gebruikten in \u002Fetc\u002Fnginx\u002Fnginx.conf:",[281,406,409],{"className":407,"code":408,"language":286,"meta":287},[284],"upstream backend {\n    least_conn;\n    server vllm0:8000 max_fails=3 fail_timeout=10000s;\n    .\n    .\n    .\n    server vllm63:8000 max_fails=3 fail_timeout=10000s;\n}\n\nserver {\n    listen 80;\n    location \u002F {\n        proxy_pass http:\u002F\u002Fbackend;\n        proxy_set_header Host $host;\n        proxy_set_header X-Real-IP $remote_addr;\n        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;\n        proxy_set_header X-Forwarded-Proto $scheme;\n    }\n}\n",[289,410,408],{"__ignoreMap":287},[270,412,413],{"start":313},[38,414,415],{},[31,416,298],{},[10,418,419],{},"We hebben het volgende script gebruikt om 64 containers te starten met behulp van incrementele apparaat- en poortnummers:",[281,421,424],{"className":422,"code":423,"language":286,"meta":287},[284],"#!\u002Fbin\u002Fbash\n\n# Script to run vLLM containers with incremental device, port, and name changes\n# Runs 64 containers with device=\u002Fdev\u002Fdri\u002FrenderD128 increasing in steps of 64\n# Port starting at 8081 and increasing by 1 each time\n# Container name starting at vllm0 and increasing by 1 each time\n\necho \"Starting vLLM containers with incremental configuration…\"\n\nfor i in {0..63}; do\n    # Calculate device number (renderD128, renderD192, renderD256, etc.)\n    device_num=$((128 + i))\n    device_path=\"\u002Fdev\u002Fdri\u002FrenderD${device_num}\"\n\n    # Calculate port (8081, 8082, 8083, etc.)\n    port=$((8081 + i))\n\n    # Calculate container name (vllm0, vllm1, vllm2, etc.)\n    container_name=\"vllm${i}\"\n\n    echo \"Starting container ${container_name} on port ${port} with device ${device_path}…\"\n    docker run -itd \\\n        –ipc host \\\n        -v \u002Fdata:\u002Fdata \\\n        –network vllm_nginx \\\n        -e VLLM_ROCM_USE_AITER=True \\\n        -e HF_HOME=root\u002F.cache\u002Fhuggingface \\\n        -e HF_HUB_CACHE=\u002Froot\u002F.cache\u002Fhuggingface\u002Fhub \\\n        –device=\u002Fdev\u002Fkfd \\\n        –device=${device_path} \\\n        –group-add video \\\n        -p ${port}:8000 \\\n        –name ${container_name} \\\n        rocm\u002Fvllm:latest \\\n        vllm serve \\\n        \u002Fdata\u002F.cache\u002Fhuggingface\u002Fhub\u002Fmodels–amd–Llama-3.1-8B-Instruct-FP8-KV\u002Fsnapshots\u002Ffa42f9a9105c545755fea25cf69f49ac8c8b40e1\u002F \\\n        –num-scheduler-steps 10 \\\n        –kv-cache-dtype fp8 \\\n        –max-model-len 4096\n\n    # Check if container started successfully\n    if [ $? -eq 0 ]; then\n        echo \"✓ Container ${container_name} started successfully\"\n    else\n        echo \"✗ Failed to start container ${container_name}\"\n    fi\n\n    echo \"—\"\ndone\n\necho \"All containers started. Summary:\"\necho \"Containers: vllm0 through vllm63\"\necho \"Ports: 8081 through 8144\"\necho \"Devices: renderD128 through renderD4160 (in steps of 64)\"\necho \"\"\necho \"To check container status: docker ps\"\necho \"To view logs: docker logs <container_name>\"\n",[289,425,423],{"__ignoreMap":287},[270,427,429],{"start":428},4,[38,430,431],{},[31,432,318],{},[10,434,435],{},"Hetzelfde script als in de vorige sectie.",[10,437,438],{},"De opstelling in actie:",[10,440,441],{},[384,442],{"alt":287,"src":443},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Flh7-rt.googleusercontent.com-dc3c57fbfc36.gif",[10,445,446],{},"Paiton MI300X",[10,448,449],{},[384,450],{"alt":287,"src":451},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Flh7-rt.googleusercontent.com-d1bd449add79.gif",[10,453,454],{},"Standaard MI300X",[100,456,458],{"id":457},"benchmarkresultaten",[31,459,460],{},"Benchmarkresultaten",[462,463,465],"h4",{"id":464},"zonder-partities-8-gpus","Zonder partities, 8 GPU's",[467,468,469,516],"table",{},[470,471,472],"thead",{},[473,474,475,481,486,491,496,501,506,511],"tr",{},[476,477,478],"th",{},[31,479,480],{},"Metric",[476,482,483],{},[31,484,485],{},"Paiton ∆",[476,487,488],{},[31,489,490],{},"Standaard",[476,492,493],{},[31,494,495],{},"∆ versus standaard",[476,497,498],{},[31,499,500],{},"H200",[476,502,503],{},[31,504,505],{},"∆ versus H200",[476,507,508],{},[31,509,510],{},"B200",[476,512,513],{},[31,514,515],{},"∆ versus B200",[517,518,519,552,584,616,648,680,712,744],"tbody",{},[473,520,521,527,530,533,538,541,546,549],{},[522,523,524],"td",{},[31,525,526],{},"Benchmarkduur(en) ↓",[522,528,529],{},"4.812",[522,531,532],{},"11.029",[522,534,535],{},[31,536,537],{},"+129,20%",[522,539,540],{},"11.84",[522,542,543],{},[31,544,545],{},"+146,05%",[522,547,548],{},"4,59",[522,550,551],{},"-4,61%",[473,553,554,559,562,565,570,573,578,581],{},[522,555,556],{},[31,557,558],{},"Verzoekdoorvoer (req\u002Fs) ↑",[522,560,561],{},"213,55",[522,563,564],{},"94.308",[522,566,567],{},[31,568,569],{},"+126,44%",[522,571,572],{},"83,22",[522,574,575],{},[31,576,577],{},"+156,61%",[522,579,580],{},"225,99",[522,582,583],{},"–5,50%",[473,585,586,591,594,597,602,605,610,613],{},[522,587,588],{},[31,589,590],{},"Uitvoertokendoorvoer (tok\u002Fs) ↑",[522,592,593],{},"53851.639",[522,595,596],{},"23809.63",[522,598,599],{},[31,600,601],{},"+126,18%",[522,603,604],{},"20940.86",[522,606,607],{},[31,608,609],{},"+157,16%",[522,611,612],{},"56989.26",[522,614,615],{},"-5,52%",[473,617,618,623,626,629,634,637,642,645],{},[522,619,620],{},[31,621,622],{},"Totale tokendoorvoer (tok\u002Fs) ↑",[522,624,625],{},"101941.667",[522,627,628],{},"45047.076",[522,630,631],{},[31,632,633],{},"+126,30%",[522,635,636],{},"39674.51",[522,638,639],{},[31,640,641],{},"+156,94%",[522,643,644],{},"107827.34",[522,646,647],{},"-5,46%",[473,649,650,655,658,661,666,669,674,677],{},[522,651,652],{},[31,653,654],{},"Gemiddelde TTFT (ms) ↓",[522,656,657],{},"543.799",[522,659,660],{},"4252.513",[522,662,663],{},[31,664,665],{},"+682,47%",[522,667,668],{},"3027.49",[522,670,671],{},[31,672,673],{},"+456,96%",[522,675,676],{},"1245,55",[522,678,679],{},"+129,05%",[473,681,682,687,690,693,698,701,706,709],{},[522,683,684],{},[31,685,686],{},"Gemiddelde TPOT (ms) ↓",[522,688,689],{},"15.075",[522,691,692],{},"16.872",[522,694,695],{},[31,696,697],{},"+11,92%",[522,699,700],{},"26,70",[522,702,703],{},[31,704,705],{},"+77,02%",[522,707,708],{},"10.27",[522,710,711],{},"-31,87%",[473,713,714,719,722,725,730,733,738,741],{},[522,715,716],{},[31,717,718],{},"Gemiddelde ITL (ms) ↓",[522,720,721],{},"15.025",[522,723,724],{},"16.509",[522,726,727],{},[31,728,729],{},"+9,88%",[522,731,732],{},"71.11",[522,734,735],{},[31,736,737],{},"+373,37%",[522,739,740],{},"32,62",[522,742,743],{},"+117,10%",[473,745,746,751,754,757,762,765,770,773],{},[522,747,748],{},[31,749,750],{},"Gemiddelde E2EL (ms) ↓",[522,752,753],{},"4317.43",[522,755,756],{},"8403.948",[522,758,759],{},[31,760,761],{},"+94,65%",[522,763,764],{},"9705.94",[522,766,767],{},[31,768,769],{},"+124,79%",[522,771,772],{},"3818,69",[522,774,775],{},"-11,51%",[10,777,778],{},[384,779],{"alt":287,"src":780,"title":781},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Faverage-throughput-2-2.webp","Grafiek",[462,783,785],{"id":784},"met-partities-64-vgpus","Met partities, 64 vGPU's",[467,787,788,818],{},[470,789,790],{},[473,791,792,796,801,805,810,814],{},[476,793,794],{},[31,795,480],{},[476,797,798,800],{},[31,799,485],{},"*",[476,802,803],{},[31,804,490],{},[476,806,807],{},[31,808,809],{},"∆ versus standaard (ratio)",[476,811,812],{},[31,813,500],{},[476,815,816],{},[31,817,505],{},[517,819,820,840,860,880,899,919,939,958],{},[473,821,822,827,830,833,836,838],{},[522,823,824],{},[31,825,826],{},"Benchmarkduur(en)",[522,828,829],{},"7.875",[522,831,832],{},"17.294",[522,834,835],{},"2,20",[522,837],{},[522,839],{},[473,841,842,847,850,853,856,858],{},[522,843,844],{},[31,845,846],{},"Verzoekdoorvoer (req\u002Fs)",[522,848,849],{},"130.234",[522,851,852],{},"59.727",[522,854,855],{},"2.18",[522,857],{},[522,859],{},[473,861,862,867,870,873,876,878],{},[522,863,864],{},[31,865,866],{},"Uitvoertokendoorvoer (tok\u002Fs)",[522,868,869],{},"33339.931",[522,871,872],{},"15047.115",[522,874,875],{},"2,22",[522,877],{},[522,879],{},[473,881,882,887,890,893,895,897],{},[522,883,884],{},[31,885,886],{},"Totale tokendoorvoer (tok\u002Fs)",[522,888,889],{},"62667.914",[522,891,892],{},"28497,62",[522,894,835],{},[522,896],{},[522,898],{},[473,900,901,906,909,912,915,917],{},[522,902,903],{},[31,904,905],{},"Gemiddelde TTFT (ms)",[522,907,908],{},"1082.885",[522,910,911],{},"6255.879",[522,913,914],{},"5,78",[522,916],{},[522,918],{},[473,920,921,926,929,932,935,937],{},[522,922,923],{},[31,924,925],{},"Gemiddelde TPOT (ms)",[522,927,928],{},"20,99",[522,930,931],{},"31.289",[522,933,934],{},"1,49",[522,936],{},[522,938],{},[473,940,941,946,948,951,954,956],{},[522,942,943],{},[31,944,945],{},"Gemiddelde ITL (ms)",[522,947,928],{},[522,949,950],{},"31.13",[522,952,953],{},"1,48",[522,955],{},[522,957],{},[473,959,960,965,968,971,974,976],{},[522,961,962],{},[31,963,964],{},"Gemiddelde E2EL (ms)",[522,966,967],{},"6435.477",[522,969,970],{},"14067.724",[522,972,973],{},"2.19",[522,975],{},[522,977],{},[10,979,980],{},[981,982,983],"em",{},"*Opmerking: we werken eraan om deze cijfers verder te verbeteren.",[10,985,986],{},[981,987,988,989,994],{},"**Opmerking 2: dit is niet mogelijk met NVIDIA, of is op zijn minst erg moeilijk (",[17,990,993],{"href":991,"rel":992},"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm\u002Fissues\u002F6551#issuecomment-2237624342",[142],"meer informatie",").",[10,996,997],{},[384,998],{"alt":287,"src":999,"title":781},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Faverage-throughput-2-2-1.webp",[10,1001,1002],{},"Laten we dit nu eens bekijken vanuit een ROI-gedreven perspectief.",[10,1004,1005],{},"Als de resultaten tot nu toe nog geen indruk op u hebben gemaakt, zal deze vergelijking dat vermoedelijk wel doen. We gebruiken de MI300X-server als referentie om kosten en doorvoer te vergelijken met de H200 en B200.",[467,1007,1008,1032],{},[470,1009,1010],{},[473,1011,1012,1017,1022,1027],{},[476,1013,1014],{},[31,1015,1016],{},"Architectuur",[476,1018,1019],{},[31,1020,1021],{},"Kostenfactor versus Paiton",[476,1023,1024,800],{},[31,1025,1026],{},"Kostenefficiëntie throughput",[476,1028,1029],{},[31,1030,1031],{},"Kostenefficiëntie latency",[517,1033,1034,1048,1063,1078],{},[473,1035,1036,1041,1044,1046],{},[522,1037,1038],{},[31,1039,1040],{},"MI300X+Paiton",[522,1042,1043],{},"Referentie",[522,1045,1043],{},[522,1047,1043],{},[473,1049,1050,1054,1057,1060],{},[522,1051,1052],{},[31,1053,454],{},[522,1055,1056],{},"1x",[522,1058,1059],{},"+126,31%",[522,1061,1062],{},"+94,57%",[473,1064,1065,1069,1072,1075],{},[522,1066,1067],{},[31,1068,500],{},[522,1070,1071],{},"1.375x",[522,1073,1074],{},"+253,30%",[522,1076,1077],{},"+209,07%",[473,1079,1080,1084,1087,1090],{},[522,1081,1082],{},[31,1083,510],{},[522,1085,1086],{},"2x",[522,1088,1089],{},"+89,18%",[522,1091,705],{},[1093,1094,1095,1100],"blockquote",{},[10,1096,1097],{},[981,1098,1099],{},"*Kostenefficiëntie van throughput: het percentage extra totale tokenthroughput per dollar ten opzichte van elk platform.",[10,1101,1102],{},[981,1103,1104],{},"**Kostenefficiëntie van latency: het percentage betere end-to-end-latency per dollar ten opzichte van elk platform.",[10,1106,1107],{},[31,1108,1109],{},"Wat dit u vertelt",[35,1111,1112,1122,1136],{},[38,1113,1114,1117,1118,1121],{},[31,1115,1116],{},"Paiton levert 2,5× de throughput per dollar"," ten opzichte van een H200 en ",[31,1119,1120],{},"126% meer"," dan de standaardconfiguratie.",[38,1123,1124,1127,1128,1131,1132,1135],{},[31,1125,1126],{},"Latency per dollar"," is ",[31,1129,1130],{},"3,1× beter dan bij de H200"," en ",[31,1133,1134],{},"94% beter dan bij de standaardconfiguratie",". Elke bespaarde milliseconde draagt bij aan het rendement.",[38,1137,1138,1139,1141,1142,1145],{},"Het ",[31,1140,510],{},"-gat is reëel, maar onthoud dat het ",[31,1143,1144],{},"tweemaal"," zoveel kost. Paiton wint over de hele linie nog steeds op het gebied van kostenefficiëntie.",[100,1147,1149],{"id":1148},"kosten-per-miljoen-tokens",[31,1150,1151],{},"Kosten per miljoen tokens",[10,1153,1154],{},"Als we de beschikbare huurprijzen voor de verschillende systemen gebruiken, kunnen we de relatieve kosten per 1 miljoen tokens berekenen:",[467,1156,1157,1190],{},[470,1158,1159],{},[473,1160,1161,1165,1170,1175,1180,1185],{},[476,1162,1163],{},[31,1164,1016],{},[476,1166,1167],{},[31,1168,1169],{},"Doorvoer (tok\u002Fs)",[476,1171,1172],{},[31,1173,1174],{},"GPU-aantal",[476,1176,1177],{},[31,1178,1179],{},"Ong. uurkosten",[476,1181,1182],{},[31,1183,1184],{},"Inferentiekosten \u002F 1M-tokens",[476,1186,1187],{},[31,1188,1189],{},"Relatieve kosten",[517,1191,1192,1212,1230,1249],{},[473,1193,1194,1198,1200,1203,1206,1209],{},[522,1195,1196],{},[31,1197,1040],{},[522,1199,625],{},[522,1201,1202],{},"~8",[522,1204,1205],{},"$ 20,50",[522,1207,1208],{},"$ 0,06",[522,1210,1211],{},"REF",[473,1213,1214,1218,1220,1222,1224,1227],{},[522,1215,1216],{},[31,1217,454],{},[522,1219,628],{},[522,1221,1202],{},[522,1223,1205],{},[522,1225,1226],{},"$ 0,13",[522,1228,1229],{},"2,26× ↑",[473,1231,1232,1236,1238,1240,1243,1246],{},[522,1233,1234],{},[31,1235,500],{},[522,1237,636],{},[522,1239,1202],{},[522,1241,1242],{},"$ 28,20",[522,1244,1245],{},"$ 0,20",[522,1247,1248],{},"3,54× ↑",[473,1250,1251,1255,1257,1259,1262,1264],{},[522,1252,1253],{},[31,1254,510],{},[522,1256,644],{},[522,1258,1202],{},[522,1260,1261],{},"$ 48,60",[522,1263,1226],{},[522,1265,1266],{},"2,24× ↑",[10,1268,1269],{},[384,1270],{"alt":287,"src":1271},"\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-1.jpg",[10,1273,1274],{},[31,1275,1276],{},"Inzichten:",[35,1278,1279,1285,1290],{},[38,1280,1281,1284],{},[31,1282,1283],{},"Paiton"," levert 2,26× kostenbesparingen op vergeleken met de niet-geoptimaliseerde MI300X.",[38,1286,1287,1289],{},[31,1288,500],{}," kost 3,54× meer dan MI300X+Paiton per 1 miljoen tokens.",[38,1291,1292,1294,1295,1298],{},[31,1293,510],{}," is de duurste en kost meer dan ",[31,1296,1297],{},"2,24× meer"," dan de geoptimaliseerde MI300X-installatie.",[100,1300,1302],{"id":1301},"grote-overwinning-voor-amd",[31,1303,1304],{},"Grote overwinning voor AMD",[10,1306,1307],{},"Uitzoeken hoe MIG met vLLM op NVIDIA werkte, bleek bijzonder vermoeiend. Uiteindelijk liepen we vast op een NCCL-fout die onoplosbaar leek.",[10,1309,1310],{},"Hoewel MIG virtuele partities op ondersteunde NVIDIA GPU's toestaat, kwamen we, zoals eerder vermeld, aanzienlijke beperkingen tegen bij het gebruik ervan in combinatie met vLLM voor data-parallelle workloads. Meer specifiek kon vLLM de MIG-partities niet goed benutten voor gedistribueerde inferentie.",[10,1312,1313],{},"De AMD-architectuur maakte daarentegen een eenvoudige partitionering en gecontaineriseerde deployment van vLLM-instances mogelijk. Deze gestroomlijnde opstelling en ROCm-compatibiliteit maakten AMD direct veel geschikter voor echte multi-tenancy.",[10,1315,1316,1317,1320],{},"Dit vertegenwoordigt een ",[31,1318,1319],{},"grote overwinning voor AMD",", vooral voor bedrijven die geïsoleerde inferentieworkloads op gedeelde hardware willen implementeren zonder al te veel wrijving of compromissen.",[1093,1322,1323],{},[1093,1324,1325],{},[1093,1326,1327,1332,1335],{},[10,1328,1329],{},[31,1330,1331],{},"AMD heeft Intel methodisch voorbijgestreefd op het vlak van prestaties en is nu strategisch klaar om NVIDIA's leiderschap uit te dagen. We zijn er trots op die evolutie mee te ondersteunen.",[10,1333,1334],{},"***Kian Mohadjerin",[10,1336,1337],{},"Hoofd AI, Eliovp BV*",[100,1339,1341],{"id":1340},"belangrijkste-resultaten",[31,1342,1343],{},"Belangrijkste resultaten",[35,1345,1346,1352,1358,1364],{},[38,1347,1348,1351],{},[31,1349,1350],{},"De throughput"," schaalde vrijwel lineair tot 64 partities, dankzij Paiton's minimale geheugenoverhead en snelle kerneldispatch.",[38,1353,1354,1357],{},[31,1355,1356],{},"De latency"," bleef stabiel tijdens parallelle sessies. Dat toont de kracht van Paiton's planning per GPU en optimalisaties voor gedeeld geheugen.",[38,1359,1360,1363],{},[31,1361,1362],{},"Het geheugengebruik per partitie"," lag aanzienlijk lager dan bij standaard vLLM of andere runtimes, waardoor deployments met hoge dichtheid mogelijk werden.",[38,1365,1366,1369],{},[31,1367,1368],{},"De kosten per miljoen tokens"," zijn met meer dan 2× verlaagd vergeleken met geavanceerde systemen zoals de B200, wat het vermogen van Paiton aantoont om toonaangevende efficiëntie te leveren, zelfs op meer betaalbare AMD-hardware.",[100,1371,1373],{"id":1372},"conclusie",[31,1374,1375],{},"Conclusie",[10,1377,1378],{},"Dit experiment toont hoe Paiton met geavanceerde packaging- en optimalisatietechnieken het volledige potentieel van moderne hardware zoals de MI300X ontsluit. Llama 3.1 8B FP8 uitvoeren over 64 GPU-partities laat zien dat inferentieworkloads op grote schaal parallel kunnen draaien zonder veel prestaties of bruikbaarheid op te offeren.",[10,1380,1381],{},"Stel u het potentieel voor van Paiton in combinatie met AMD's aankomende MI355X. Met nog meer geheugenbandbreedte, rekenkracht en architecturale verbeteringen kan de combinatie van next-generation hardware en de Paiton-runtime high-performance AI-serving opnieuw definiëren.",[10,1383,1384],{},"Blijf op de hoogte van toekomstige updates terwijl we de mogelijkheden van Paiton uitbreiden.",[10,1386,1387],{},"Wilt u onze resultaten zelf verifiëren? Test Paiton en vraag een evaluatiemodel aan.",[100,1389,1391],{"id":1390},"prijzen",[31,1392,1393],{},"Prijzen",[10,1395,1396],{},"Bent u benieuwd naar de prijzen van Paiton, dan is onze formule vrij eenvoudig:",[467,1398,1399],{},[470,1400,1401],{},[473,1402,1403],{},[476,1404,1405],{},"50% van x% kosten bespaard per 1 miljoen tokens",[10,1407,1408],{},"We meten de kostenbesparing door de huidige throughput van de klant te vergelijken met de throughput na inschakeling van Paiton.",[10,1410,1411,1416],{},[17,1412,1415],{"href":1413,"rel":1414},"https:\u002F\u002Fai.eliovp.com\u002Fpaiton",[142],"Neem contact op en laten we praten"," :)",[100,1418,1420],{"id":1419},"referenties",[31,1421,1422],{},"Referenties",[270,1424,1425,1432,1439,1444,1450,1456,1462,1468,1474,1481],{},[38,1426,1427],{},[17,1428,1431],{"href":1429,"rel":1430},"https:\u002F\u002Fwww.supermicro.com\u002Fen\u002Fproducts\u002Fsystem\u002Fgpu\u002F8u\u002Fas%20-8125gs-tnmr2",[142],"Supermicro GPU-systeem AS-8125GS-TNMR2",[38,1433,1434],{},[17,1435,1438],{"href":1436,"rel":1437},"https:\u002F\u002Fwww.amd.com\u002Fen\u002Fproducts\u002Faccelerators\u002Finstinct\u002Fmi300\u002Fmi300x.html",[142],"AMD Instinct MI300X",[38,1440,1441],{},[17,1442,1443],{"href":19},"Paiton FP8 verslaat NVIDIA's H200 op AMD's MI300X",[38,1445,1446],{},[17,1447,1449],{"href":140,"rel":1448},[142],"ROCm\u002Fvllm GitHub",[38,1451,1452],{},[17,1453,1455],{"href":156,"rel":1454},[142],"vllm-project\u002Fvllm GitHub",[38,1457,1458],{},[17,1459,1461],{"href":176,"rel":1460},[142],"Hugging Face AMD Llama-3.1-8B-Instruct-FP8-KV",[38,1463,1464],{},[17,1465,1467],{"href":266,"rel":1466},[142],"vLLM Nginx-implementatie",[38,1469,1470],{},[17,1471,1473],{"href":342,"rel":1472},[142],"AMD GPU-partitioneringsdocumentatie",[38,1475,1476],{},[17,1477,1480],{"href":1478,"rel":1479},"https:\u002F\u002Frocm.blogs.amd.com\u002Fsoftware-tools-optimization\u002Fcompute-memory-modes\u002FREADME.html",[142],"ROCm Compute Memory-modi",[38,1482,1483],{},[17,1484,1487],{"href":1485,"rel":1486},"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm\u002Fissues\u002F6551",[142],"vLLM GitHub nummer 6551",{"title":287,"searchDepth":293,"depth":293,"links":1489},[1490,1491,1492,1493,1494,1495,1496,1497,1498,1499],{"id":102,"depth":313,"text":105},{"id":254,"depth":313,"text":257},{"id":330,"depth":313,"text":333},{"id":457,"depth":313,"text":460},{"id":1148,"depth":313,"text":1151},{"id":1301,"depth":313,"text":1304},{"id":1340,"depth":313,"text":1343},{"id":1372,"depth":313,"text":1375},{"id":1390,"depth":313,"text":1393},{"id":1419,"depth":313,"text":1422},[1501,1502,1283,1503,1504,1505,510,500,1506,1507,1508,1283,1509],"Alle","Kunstmatige intelligentie","Niet gecategoriseerd","AI","AMD","MI300X","MI355x","NVIDIA","vLLM","2025-07-31T13:32:57","Bekijk hoe Paiton presteert met Llama 3.1 8B FP8 op gepartitioneerde MI300X-GPU's, vergeleken met NVIDIA H200 en B200.","md","MI300X FP8 data-parallelle benchmarks (8 tot 64 GPU's): H200 achter zich, B200 binnen bereik","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fimage-2-1.webp",{},true,"https:\u002F\u002Feliovp.com\u002Fmi300x-fp8-data%e2%80%91parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach\u002F","\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach",{"title":5,"description":1511},"mi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach",null,"blog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","uQm1mJoHgsinnWfZU7M1U4QZ_l4R6SoEtDY0aUsAmfY",[1525,1537,1550,1560,1569,1583,1592,1601,1610,1620,1631,1641,1653,1663,1675,1684,1695,1723,1734,1756,1774,1793,1812,1829,1846,1857,1873,1888,1890,1899,1907,1922,1933,1944,1955,1965,1978,1988,2001,2012,2022,2033,2042,2054,2064,2073],{"path":1526,"title":1527,"description":1528,"date":1529,"slug":1530,"image":1531,"originalUrl":1532,"categories":1533},"\u002Fblog\u002Fpaiton-idol-radeon-ai-pro-r9700","De Radeon AI PRO R9700 doet ook 3D: Idol in 16,4 seconden","Idol maakt een 3D Gaussian splat in 16,44 seconden op één Radeon AI PRO R9700 met Paiton, tegenover 39,22 seconden voor standaard-Idol op dezelfde GPU.","2026-11-06T09:00:00Z","paiton-idol-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-idol\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-idol-radeon-ai-pro-r9700",[1283,1534,1535,1536],"AMD Radeon","Lokale AI","3D-generatie",{"path":1538,"title":1539,"description":1540,"date":1541,"slug":1542,"image":1543,"originalUrl":1544,"categories":1545},"\u002Fblog\u002Fpaiton-deepseek-v41-flash-amd-instinct-mi350p","Deepseek V4.1 Flash op één MI350P","DeepSeek V4.1 Flash op één MI350P met 144 GB: 77,9 tokens\u002Fs bij één verzoek en 1.107,6 in een batch van 512, met Paiton en expertberekeningen op CPU.","2026-11-03T12:00:00Z","paiton-deepseek-v41-flash-amd-instinct-mi350p","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-deepseek-v41\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-deepseek-v41-flash-amd-instinct-mi350p",[1283,1546,1547,1548,1549],"AMD Instinct","AI-inferentie","DeepSeek","Kwantisatie",{"path":1551,"title":1552,"description":1553,"date":1554,"slug":1555,"image":1556,"originalUrl":1557,"categories":1558},"\u002Fblog\u002Fpaiton-qwen38-flash-next-amd-instinct-mi350p","Qwen3.8 Flash Next op MI350P: dubbel zo snel | Paiton","Qwen3.8 Flash Next op één AMD MI350P: 2,24× de decodesnelheid van de geteste vLLM-configuratie bij één verzoek en circa 2,2 miljoen gedeelde KV-cachetokens.","2026-10-30T12:00:00Z","paiton-qwen38-flash-next-amd-instinct-mi350p","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-mi350p\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-flash-next-amd-instinct-mi350p",[1283,1546,1547,1559,1549],"Qwen",{"path":1561,"title":1562,"description":1563,"date":1564,"slug":1565,"image":1566,"originalUrl":1567,"categories":1568},"\u002Fblog\u002Fpaiton-qwen38-3bit-accuracy-prefix-caching-mtp","Qwen3.8 Flash Next: 3-bit dicht bij 16-bit nauwkeurigheid, met MTP-prefixcaching","Paitons nieuwe 3-bit Qwen3.8-gewichten benaderen 16-bit taakscores op twee R9700-kaarten. Native MTP-prefixcaching hervat een prompt van 200K in 0,68 seconde. Resultaten, grenzen en installatie.","2026-10-24T18:45:00Z","paiton-qwen38-3bit-accuracy-prefix-caching-mtp","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen38-3bit-mtp\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-3bit-accuracy-prefix-caching-mtp",[1283,1534,1535,1559,1549],{"path":1570,"title":1571,"description":1572,"date":1573,"slug":1574,"image":1575,"originalUrl":1576,"categories":1577},"\u002Fblog\u002Fnvidia-dsx-pod-vera-rubin-modular-data-centers","Nvidia confirms the modular DC direction","NVIDIA DSX POD brengt Vera Rubin AI-fabrieken naar modulaire datacenters. ElioVP bespreekt terreinvoorbereiding, koeling en de aanpak achter ModFlex.","2026-10-21T13:00:00Z","nvidia-dsx-pod-vera-rubin-modular-data-centers","\u002Fasset\u002Fimages\u002Fblog\u002Fnvidia-dsx-pod-modflex\u002Fhero.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fnvidia-dsx-pod-vera-rubin-modular-data-centers",[1578,1579,1580,1508,1581,1582],"Modular DC","AI-infrastructuur","ModFlex","Vera Rubin","Vloeistofkoeling",{"path":1584,"title":1585,"description":1586,"date":1587,"slug":1588,"image":1589,"originalUrl":1590,"categories":1591},"\u002Fblog\u002Fpaiton-qwen38-flash-next-4bit-radeon-ai-pro-r9700","Qwen3.8 Flash Next in 4-bit op 2 × Radeon R9700: getest tegenover 16-bit","Paiton draait Qwen3.8 Flash Next in 4-bit op twee R9700-kaarten: 96,8% GSM8K, 96,3% HumanEval, 189,8 tokens\u002Fs en een optionele RAM-prefixcache. Kwaliteit, snelheid en installatie.","2026-10-12T15:00:00Z","paiton-qwen38-flash-next-4bit-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen38-flash-next-4bit\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-flash-next-4bit-radeon-ai-pro-r9700",[1283,1534,1535,1559,1549],{"path":1593,"title":1594,"description":1595,"date":1596,"slug":1597,"image":1598,"originalUrl":1599,"categories":1600},"\u002Fblog\u002Fpaiton-qwen38-flash-next-radeon-ai-pro-r9700","Qwen3.8 Flash Next op 2 × Radeon AI PRO R9700: 216 tokens\u002Fs, 200K-context","Paiton draait Qwen3.8 Flash Next met 3-bit experts op twee R9700-kaarten: 216,3 tokens\u002Fs, 565,1 tokens\u002Fs samen en een aparte 200K-modus. Metingen, kwaliteit en installatie.","2026-10-10T10:30:00Z","paiton-qwen38-flash-next-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen38-flash-next\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-flash-next-radeon-ai-pro-r9700",[1283,1534,1535,1559,1549],{"path":1602,"title":1603,"description":1604,"date":1605,"slug":1606,"image":1607,"originalUrl":1608,"categories":1609},"\u002Fblog\u002Fpaiton-qwen38-w3a4-radeon-ai-pro-r9700","Qwen3.8 27B op 1 × Radeon AI PRO R9700: 3-bit gewichten, 20% snellere decode","Qwen3.8 27B op één R9700: 19,9% snellere gewogen decode, een nieuwe 4-bit cache, 200K-context en optionele beeldinvoer. Metingen en Paiton-installatie.","2026-09-27T09:00:00Z","paiton-qwen38-w3a4-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen38-w3a4\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen38-w3a4-radeon-ai-pro-r9700",[1283,1534,1535,1559,1549],{"path":1611,"title":1612,"description":1613,"date":1614,"slug":1615,"image":1616,"originalUrl":1617,"categories":1618},"\u002Fblog\u002Fpaiton-qwen-image-21-radeon-ai-pro-r9700","Qwen-Image 2.1 op 1 × Radeon AI PRO R9700: 2048×2048-beelden in 103 seconden","Genereer lokaal Qwen-Image 2.1-beelden van 2048×2048 pixels met 1 × Radeon AI PRO R9700. De Paiton v1.0.2-container mat 103,29 seconden per opgewarmd verzoek, inclusief de PNG-respons.","2026-09-23T09:00:00Z","paiton-qwen-image-21-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-qwen-image-21\u002Fhero-nl.webp","https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-qwen-image-21-radeon-ai-pro-r9700",[1283,1534,1535,1619,1559],"Beeldgeneratie",{"path":1621,"title":1622,"description":1623,"date":1624,"slug":1625,"image":1626,"originalUrl":1627,"categories":1628},"\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",[1283,1534,1535,1547,1629,1630,1509],"Inferentie-optimalisatie","Grote taalmodellen",{"path":1632,"title":1633,"description":1634,"date":1635,"slug":1636,"image":1637,"originalUrl":1638,"categories":1639},"\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",[1283,1534,1535,1640,1509],"GGUF",{"path":1642,"title":1643,"description":1644,"date":1645,"slug":1646,"image":1647,"originalUrl":1648,"categories":1649},"\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",[1283,1534,1535,1650,1651,1652],"Videogeneratie","MiniMax H3","ComfyUI",{"path":1654,"title":1655,"description":1656,"date":1657,"slug":1658,"image":1659,"originalUrl":1660,"categories":1661},"\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",[1283,1534,1535,1619,1662,1652],"FLUX",{"path":1664,"title":1665,"description":1666,"date":1667,"slug":1668,"image":1669,"originalUrl":1670,"categories":1671},"\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",[1283,1502,1534,1547,1672,1673,1629,1630,1509,1674],"GPU-prestaties","Inferentielatentie","Kostenefficiëntie",{"path":1676,"title":1677,"description":1678,"date":1679,"slug":1680,"image":1681,"originalUrl":1682,"categories":1683},"\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",[1283,1502,1534,1547,1672,1673,1629,1630,1509,1674],{"path":1685,"title":1686,"description":1687,"date":1688,"slug":1689,"image":1690,"originalUrl":1521,"categories":1691},"\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",[1501,1579,1692,1580,1693,1694],"Datacenters","HPC","AMD Helios",{"path":1696,"title":1697,"description":1698,"date":1699,"slug":1700,"image":1701,"originalUrl":1702,"categories":1703},"\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",[1501,1502,1283,1704,1505,510,1705,1706,1707,1708,1709,1710,1711,1712,1713,1714,1507,1508,1715,1716,1283,1717,1718,1719,1720,1721,1722],"14B","Benchmarks","Blackwell","Compute","Diffusie","Eliovp","Generatieve AI","GPU","Hardware","Inferentie","Instinct","On-premises","Optimalisatie","Soevereine AI","T2V","Tekst-naar-video","Tuning","Video-generatie","Wan2.2",{"path":1724,"title":1725,"description":1726,"date":1727,"slug":1728,"image":1729,"originalUrl":1730,"categories":1731},"\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",[1501,1502,1732,1503,1505,1733,1732,1508],"Modulaire DC","De Tijd",{"path":1735,"title":1736,"description":1737,"date":1738,"slug":1739,"image":1740,"originalUrl":1741,"categories":1742},"\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",[1501,1502,1743,1503,1744,1745,1746,1747,1748,1749,1750,1751,1709,1752,1710,1753,1754,1755],"Trending","AI Act","Antropomorfisme","AVG","Benelux","ChatGPT","Cyberbeveiliging","Databeheer","Gegevensbeveiliging","GDPR","Microsoft Copilot","Privacy","Shadow AI",{"path":1757,"title":1758,"description":1759,"date":1760,"slug":1761,"image":1762,"originalUrl":1763,"categories":1764},"\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",[1501,1765,1766,1767,1749,1768,1769,1770,1752,1771,1772,1773,1754],"AWS","Belgian Mobile ID","CLOUD Act","Datasoevereiniteit","Digitale identiteit","eIDAS","itsme","Liberty Global","MyGov.be",{"path":1775,"title":1776,"description":1777,"date":1778,"slug":1779,"image":1780,"originalUrl":1781,"categories":1782},"\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",[1501,1502,1783,1743,1784,1785,1786,1787,1788,1789,1790,1791,1717,1792],"Oplossingen","Agentic AI","AI-techniek","AI-strategie","Autonome agenten","Bedrijfs-AI","Lokale LLM","Modelverfijning","AI op locatie","VRAM-optimalisatie",{"path":1794,"title":1795,"description":1796,"date":1797,"slug":1798,"image":1799,"originalUrl":1800,"categories":1801},"\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",[1501,1502,1743,1802,1803,1804,1805,1806,1807,1808,1809,1810,1811],"AI Infrastructure","AI Neocloud","Circulaire financiering","GPU Cloud","Beleggingsrisico's","Opstartwaardering","Synthetische bubbel","Technische analyse","Vaporware","Durfkapitaal",{"path":1813,"title":1814,"description":1815,"date":1816,"slug":1817,"image":1818,"originalUrl":1819,"categories":1820},"\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",[1501,1732,1503,1821,1579,1822,1823,1582,1824,1825,1826,1827,1828],"150 kW-rack","DLC","Hoge dichtheid","Modulair datacenter","NVIDIA Blackwell Ultra","NVIDIA GB300","NVL72","Snelle implementatie",{"path":1830,"title":1831,"description":1832,"date":1833,"slug":1834,"image":1835,"originalUrl":1836,"categories":1837},"\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",[1501,1502,1283,1503,1838,1502,1839,1840,1841,1842,1843,1844,1283,1845],"AMD MI300X","CUDA-vertaling","FP8","GPU-optimalisatie","High-performance computing","HIP","Kerneltuning","ROCm",{"path":1847,"title":1848,"description":1849,"date":1850,"slug":1851,"image":1852,"originalUrl":1853,"categories":1854},"\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",[1501,1502,1283,1547,1546,1838,1674,1855,1629,1844,1283,1856,1509],"Hoge throughput","SGLang",{"path":1858,"title":1859,"description":1860,"date":1861,"slug":1862,"image":1863,"originalUrl":1864,"categories":1865},"\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","Paiton MoE-benchmarks: MI300X versus H200 en B200","Vergelijk Qwen3-30B-A3B MoE-benchmarks van MI300X met Paiton, H200 en B200: throughput en kosten per miljoen tokens.","2025-09-26T13:36:18","stop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","\u002Fasset\u002Fimages\u002Fblog\u002Fimported\u002Fhulkvshulkpaitonwins.webp","https:\u002F\u002Feliovp.com\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens\u002F",[1501,1502,1283,1866,1838,1867,1629,1868,1869,1870,1871,1283,1872],"AI-benchmarks","Kosten per token","Mixture of Experts","MoE","NVIDIA B200","NVIDIA H200","Qwen3",{"path":1874,"title":1875,"description":1876,"date":1877,"slug":1878,"image":1879,"originalUrl":1880,"categories":1881},"\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",[1501,1502,1783,1503,1784,1882,1883,1884,1885,1789,1791,1717,1886,1887],"Schadedetectie","Documentverwerking","E-mailautomatisering","Factuurextractie","Ticketautomatisering","Workflowautomatisering",{"path":1518,"title":5,"description":1511,"date":1510,"slug":1520,"image":1514,"originalUrl":1517,"categories":1889},[1501,1502,1283,1503,1504,1505,510,500,1506,1507,1508,1283,1509],{"path":1891,"title":1892,"description":1893,"date":1894,"slug":1895,"image":1896,"originalUrl":1897,"categories":1898},"\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",[1501,1502,1783,1784,1882,1883,1884,1885,1789,1791,1717,1886,1887],{"path":1900,"title":1901,"description":1902,"date":1903,"slug":1904,"image":287,"originalUrl":1905,"categories":1906},"\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",[1501,1502,1283],{"path":1908,"title":1909,"description":1910,"date":1911,"slug":1912,"image":1913,"originalUrl":1914,"categories":1915},"\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",[1501,1502,1283,1503,1547,1838,1916,1840,1917,1918,1919,1283,1920,1921],"Koude start","Grafiekcompilatie","Llama 3.1 405B","LLM-optimalisatie","Opstartlatentie","Tensor-parallellisme",{"path":1923,"title":1443,"description":1924,"date":1925,"slug":1926,"image":1927,"originalUrl":1928,"categories":1929},"\u002Fblog\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-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",[1501,1502,1283,1503,1838,1930,1788,1710,1672,1673,1630,1918,1931,1932],"Koude startoptimalisatie","Model serving","vLLM-optimalisatie",{"path":1934,"title":1935,"description":1936,"date":1937,"slug":1938,"image":1939,"originalUrl":1940,"categories":1941},"\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",[1501,1502,1283,1783,1503,1505,1506,1508,1942,1943],"RX7900XTX","tenstorrent",{"path":1945,"title":1946,"description":1947,"date":1948,"slug":1949,"image":1950,"originalUrl":1951,"categories":1952},"\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",[1501,1502,1732,1783,510,500,1953,1508,1954],"MI325X","P&L-calculator",{"path":1956,"title":1957,"description":1958,"date":1959,"slug":1960,"image":1961,"originalUrl":1962,"categories":1963},"\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",[1501,1502,1283,1504,1505,500,1964,1508,1283,1509],"MI300",{"path":1966,"title":1967,"description":1968,"date":1969,"slug":1970,"image":1971,"originalUrl":1972,"categories":1973},"\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",[1501,1732,1974,1802,1823,1693,1582,1824,1975,1976,1827,1977],"1-2MW datacenter","NVIDIA Blackwell","NVIDIA NVL","Precisiekoeling",{"path":1979,"title":1980,"description":1981,"date":1982,"slug":1983,"image":1984,"originalUrl":1985,"categories":1986},"\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",[1501,1502,1783,1503,1504,1505,1987],"Zorg",{"path":1989,"title":1990,"description":1991,"date":1992,"slug":1993,"image":1994,"originalUrl":1995,"categories":1996},"\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",[1501,1743,1504,1505,1997,1998,1999,2000],"Invoer","Taiwan","Invoerheffingen","Trump",{"path":2002,"title":2003,"description":2004,"date":2005,"slug":2006,"image":2007,"originalUrl":2008,"categories":2009},"\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",[1501,1502,1783,1504,2010,2011],"AI-agenten","ERP",{"path":2013,"title":2014,"description":2015,"date":2016,"slug":2017,"image":2018,"originalUrl":2019,"categories":2020},"\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",[1501,1502,1743,2021,1505,1711,1508],"AI-nieuws",{"path":2023,"title":2024,"description":2025,"date":2026,"slug":2027,"image":2028,"originalUrl":2029,"categories":2030},"\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",[1501,1502,1283,1504,1505,2031,2032,1506,1283],"benchmark","LLM",{"path":2034,"title":2035,"description":2036,"date":2037,"slug":2038,"image":2039,"originalUrl":2040,"categories":2041},"\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",[1501,1502,1283],{"path":2043,"title":2044,"description":2045,"date":2046,"slug":2047,"image":2048,"originalUrl":2049,"categories":2050},"\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",[1501,1505,2051,2052,2053],"Jim Greene","Podcast","Tech Talk",{"path":2055,"title":2056,"description":2057,"date":2058,"slug":2059,"image":2060,"originalUrl":2061,"categories":2062},"\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",[1501,1502,1283,1505,1548,2063,500,1506,1953,1283,1509],"H100",{"path":2065,"title":2066,"description":2067,"date":2068,"slug":2069,"image":2070,"originalUrl":2071,"categories":2072},"\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",[1501,1502,1283,1505,1548,2063,500,1506,1953,1283,1509],{"path":2074,"title":2075,"description":2076,"date":2077,"slug":2078,"image":2079,"originalUrl":2080,"categories":2081},"\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",[1501,1502,1283,1505,2063,500,1506,1953,1283,1509],1793962345598]