[{"data":1,"prerenderedAt":1264},["ShallowReactive",2],{"blog-post-en-\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700":3,"blog-posts-sidebar-en":792},{"id":4,"title":5,"body":6,"categories":767,"date":778,"description":779,"extension":780,"heading":781,"image":782,"meta":783,"navigation":784,"originalUrl":785,"path":786,"seo":787,"slug":788,"stem":789,"updated":790,"__hash__":791},"blog\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700.md","Ornith 1.5 at 44.6 tok\u002Fs on One Radeon AI PRO R9700",{"type":7,"value":8,"toc":757},"minimark",[9,16,31,46,54,63,70,75,86,97,173,193,204,207,211,214,287,297,305,311,314,318,321,328,354,360,363,366,370,380,383,386,455,462,468,471,475,478,490,493,496,499,508,512,515,576,579,587,600,603,607,732,735,739,742,745,753],[10,11,12],"p",{},[13,14,15],"strong",{},"One workstation GPU. A 256-token response completed in a median 5.12 seconds. No cloud inference API.",[10,17,18,19,26,27,30],{},"Paiton now runs the public ",[20,21,25],"a",{"href":22,"rel":23},"https:\u002F\u002Fhuggingface.co\u002FCapicua25x\u002FOrnith-1.5-35B-A3B-MXFP4-Quark-RDNA4",[24],"nofollow","Ornith 1.5 35B A3B MXFP4 checkpoint"," at an average of ",[13,28,29],{},"44.63 output tokens per second"," on one AMD Radeon AI PRO R9700.",[10,32,33,34,37,38,41,42,45],{},"The fastest qualified stock vLLM configuration we obtained on the same machine reached ",[13,35,36],{},"35.13 output tokens per second",". That gives Paiton ",[13,39,40],{},"27.03% more output throughput"," and reduces modeled time-based cost per generated token by ",[13,43,44],{},"21.28%",".",[10,47,48,49,53],{},"Our earlier ",[20,50,52],{"href":51},"\u002Fblog\u002Fpaiton-qwen38-radeon-ai-pro-r9700","Qwen3.8 benchmark"," tested three different prompt shapes. This release answers a different question: can a roughly 35B mixture-of-experts model, its speculative draft and a useful serving runtime fit on one 32 GB card, and still feel fast enough for local use?",[55,56,57],"blockquote",{},[10,58,59,62],{},[13,60,61],{},"Benchmark boundary:"," 256 requested input tokens, 256 output tokens, one active sequence, temperature zero and thinking disabled. Chat templating produced 268 to 270 actual prompt tokens. The headline measures output generation, not prompt-processing speed, long-context coding, tool use or concurrent serving.",[10,64,65],{},[66,67],"img",{"alt":68,"src":69},"Paiton serves Ornith 1.5 at 44.63 output tokens per second on one Radeon AI PRO R9700, 27.0% ahead of the fastest qualified stock vLLM result.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F01-throughput-ornith15-eliovp.webp",[71,72,74],"h2",{"id":73},"a-35b-model-that-actually-fits","A 35B model that actually fits",[10,76,77,78,81,82,85],{},"Ornith 1.5 35B A3B is a mixture-of-experts model. The ",[13,79,80],{},"A3B"," label means that roughly 3 billion parameters are activated for each token. It does ",[13,83,84],{},"not"," mean the whole model occupies the memory of a 3B model; the full expert pool still has to be stored.",[10,87,88,89,92,93,96],{},"The tested MXFP4 target is approximately ",[13,90,91],{},"22.9 GB",", while the public DFlash draft adds about ",[13,94,95],{},"772 MB",". Together, they fit inside the R9700's 32 GB of VRAM with the qualified runtime and a reserved KV cache.",[98,99,100,113],"table",{},[101,102,103],"thead",{},[104,105,106,110],"tr",{},[107,108,109],"th",{},"What you need",[107,111,112],{},"Published package",[114,115,116,125,133,141,149,157,165],"tbody",{},[104,117,118,122],{},[119,120,121],"td",{},"GPU",[119,123,124],{},"One Radeon AI PRO R9700, 32 GB",[104,126,127,130],{},[119,128,129],{},"Runtime path",[119,131,132],{},"Linux, Docker and ROCm 7.14",[104,134,135,138],{},[119,136,137],{},"First-start download",[119,139,140],{},"22.9 GB target + 772 MB draft",[104,142,143,146],{},[119,144,145],{},"Temporary free disk",[119,147,148],{},"Approximately 48 GB during first-time staging",[104,150,151,154],{},[119,152,153],{},"Maximum context",[119,155,156],{},"8,192 tokens",[104,158,159,162],{},[119,160,161],{},"Access",[119,163,164],{},"Terminal chat + OpenAI-compatible API",[104,166,167,170],{},[119,168,169],{},"Qualified serving scope",[119,171,172],{},"Text only, one active sequence",[10,174,175,176,181,182,186,187,192],{},"The ",[20,177,180],{"href":178,"rel":179},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin",[24],"public runtime plugin"," is Apache-2.0-licensed. The ",[20,183,185],{"href":22,"rel":184},[24],"target checkpoint"," is MIT-licensed, while the ",[20,188,191],{"href":189,"rel":190},"https:\u002F\u002Fhuggingface.co\u002Fz-lab\u002FQwen3.6-35B-A3B-DFlash",[24],"DFlash draft"," is Apache-2.0-licensed. The Paiton compiler itself is not distributed.",[10,194,175,195,200,201,203],{},[20,196,199],{"href":197,"rel":198},"https:\u002F\u002Fhuggingface.co\u002Fornith-ai\u002FOrnith-1.5-35B-A3B",[24],"upstream Ornith team"," targets coding and agentic workflows and publishes 262,144-token serving examples with tool use. That makes it a relevant local target rather than a synthetic capacity demo, but those broader capabilities are ",[13,202,84],{}," qualified by this package. The published Paiton path is capped at 8,192 tokens, text only, and automatic tool choice is outside this release's tested scope.",[10,205,206],{},"That boundary matters. Buyers evaluating local AI usually need to know more than whether a checkpoint can be loaded: what fits, what starts reliably, which API is available and where the qualification ends.",[71,208,210],{"id":209},"where-the-27-comes-from","Where the 27% comes from",[10,212,213],{},"Most of the uplift is already present before speculative decoding is enabled.",[98,215,216,233],{},[101,217,218],{},[104,219,220,223,227,230],{},[107,221,222],{},"Runtime",[107,224,226],{"align":225},"right","Output tokens\u002Fs",[107,228,229],{"align":225},"Median TPOT",[107,231,232],{"align":225},"Difference vs stock",[114,234,235,249,265],{},[104,236,237,240,243,246],{},[119,238,239],{},"Fastest qualified stock vLLM",[119,241,242],{"align":225},"35.132",[119,244,245],{"align":225},"27.457 ms",[119,247,248],{"align":225},"Reference",[104,250,251,254,257,260],{},[119,252,253],{},"Paiton without DFlash",[119,255,256],{"align":225},"43.405",[119,258,259],{"align":225},"21.926 ms",[119,261,262],{"align":225},[13,263,264],{},"+23.55%",[104,266,267,272,277,282],{},[119,268,269],{},[13,270,271],{},"Paiton with DFlash, two-run mean",[119,273,274],{"align":225},[13,275,276],{},"44.628",[119,278,279],{"align":225},[13,280,281],{},"18.864 ms",[119,283,284],{"align":225},[13,285,286],{},"+27.03%",[10,288,289,290,293,294,45],{},"The Paiton target path delivers ",[13,291,292],{},"23.55% more throughput than stock"," on its own. DFlash then moves the result from 43.405 to 44.628 tokens per second, adding another ",[13,295,296],{},"2.82% over Paiton without speculation",[10,298,299,304],{},[20,300,303],{"href":301,"rel":302},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.06036",[24],"DFlash"," uses a lightweight block-diffusion draft model to propose several upcoming tokens in parallel. Ornith verifies those proposals, accepts the valid prefix and remains the authority for the generated output.",[10,306,307],{},[66,308],{"alt":309,"src":310},"DFlash proposes token blocks in parallel and Ornith verifies them. Paiton provides most of the measured gain; DFlash adds a further 2.8% over Paiton alone.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F04-dflash-ornith15-eliovp.webp",[10,312,313],{},"DFlash provides the final lift, but it is not the source of the full 27%. That is where the public implementation detail stops: Paiton's generated kernels, fusion choices, schedules and runtime mechanics remain proprietary.",[71,315,317],{"id":316},"it-starts-20-ms-later-but-finishes-266-seconds-sooner","It starts 20 ms later but finishes 2.66 seconds sooner",[10,319,320],{},"Generation speed is only half of the user experience. A local user also notices how long the model takes to begin and how long the complete answer takes to arrive.",[10,322,323,324,327],{},"For this workload, median time to first token increased from approximately ",[13,325,326],{},"290 ms to 310 ms",". Once generation started, the faster stream more than recovered that 20 ms delay:",[329,330,331,342,348],"ul",{},[332,333,334,335,338,339,45],"li",{},"Median time per output token fell from ",[13,336,337],{},"27.46 ms to 18.86 ms",", a ",[13,340,341],{},"31.3% reduction",[332,343,344,345,45],{},"Median end-to-end time fell from approximately ",[13,346,347],{},"7.78 seconds to 5.12 seconds",[332,349,350,351,45],{},"The complete 256-token response finished about ",[13,352,353],{},"2.66 seconds sooner",[10,355,356],{},[66,357],{"alt":358,"src":359},"Paiton starts the Ornith 1.5 response about 20 milliseconds later but completes it 2.66 seconds sooner.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F02-latency-ornith15-eliovp.webp",[10,361,362],{},"This is the honest latency story: a marginally slower first token and a materially faster complete answer. With speculative decoding, TPOT is a streaming-cadence metric rather than a direct inverse of complete-request throughput because accepted tokens can arrive in bursts. Complete-request throughput and end-to-end time remain the primary measures here.",[10,364,365],{},"Workloads dominated by large prompt ingestion need separate prompt-processing measurements; this article does not infer those from decode throughput.",[71,367,369],{"id":368},"more-output-from-the-same-active-hour","More output from the same active hour",[10,371,372,373,376,377,45],{},"At the measured rates, stock vLLM needs about ",[13,374,375],{},"7.91 active hours"," to generate one million output tokens. Paiton with DFlash needs about ",[13,378,379],{},"6.22 hours",[10,381,382],{},"That is why a 27.03% throughput increase becomes a 21.28% cost reduction: cost per token is the inverse of throughput.",[10,384,385],{},"Using the same illustrative ownership model as the Qwen3.8 article, with 5,000 productive inference hours, equal 450 W whole-system draw, purchase prices of $1,299 or €1,749, and electricity at $0.17 or €0.2558 per kWh, the result is:",[98,387,388,401],{},[101,389,390],{},[104,391,392,395,398],{},[107,393,394],{},"Interactive token economics",[107,396,397],{"align":225},"Qualified stock vLLM",[107,399,400],{"align":225},"Paiton + DFlash",[114,402,403,416,429,442],{},[104,404,405,408,411],{},[119,406,407],{},"Hours per million output tokens",[119,409,410],{"align":225},"7.91",[119,412,413],{"align":225},[13,414,415],{},"6.22",[104,417,418,421,424],{},[119,419,420],{},"Modeled US cost per million",[119,422,423],{"align":225},"$2.66",[119,425,426],{"align":225},[13,427,428],{},"$2.09",[104,430,431,434,437],{},[119,432,433],{},"Modeled European cost per million",[119,435,436],{"align":225},"€3.68",[119,438,439],{"align":225},[13,440,441],{},"€2.89",[104,443,444,447,450],{},[119,445,446],{},"Output over 5,000 active hours",[119,448,449],{"align":225},"632.4 million",[119,451,452],{"align":225},[13,453,454],{},"803.3 million",[10,456,457,458,461],{},"The same card therefore produces approximately ",[13,459,460],{},"171 million additional output tokens"," over 5,000 productive hours in this model.",[10,463,464],{},[66,465],{"alt":466,"src":467},"At 5,000 productive inference hours, the modeled European cost falls from €3.68 to €2.89 per million output tokens with Paiton and DFlash.","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F03-economics-ornith15-eliovp.webp",[10,469,470],{},"This is a time-and-ownership model, not a measured wall-power efficiency claim. It assumes equal system draw and excludes the host computer, idle time, cooling, maintenance, financing, tax and residual value.",[71,472,474],{"id":473},"why-the-stock-comparison-is-credible","Why the stock comparison is credible",[10,476,477],{},"We did not compare Paiton with a default or deliberately weak installation. The selected stock cell was the fastest stable result obtained after testing the available execution, graph, attention, recurrent and mixture-of-experts settings.",[10,479,480,481,485,486,489],{},"It used pinned vanilla ROCm vLLM, O2 compilation, graph capture at batch size one, Triton attention, Triton GDN decode, the unfused Triton MoE backend, DFlash disabled, and the same supported ",[482,483,484],"code",{},"AUTO"," performance level with the ",[482,487,488],{},"COMPUTE"," power profile.",[10,491,492],{},"Both sides used the same GPU, checkpoint, tokenizer, pinned vLLM revision, request data, random seed, output length, concurrency and GPU profile. The headline gives stock its fastest qualified cell and compares it with the mean of two fresh-server Paiton runs.",[10,494,495],{},"The stock cell completed all 12 measured requests. The two Paiton runs completed all 24 and returned every requested output token. A deterministic natural-language request before and after the Paiton runs remained readable, stopped naturally, produced finite log probabilities and was byte-identical.",[10,497,498],{},"That last check is a serving sanity gate, not an academic accuracy evaluation, a coding benchmark or proof of quality parity with the upstream model. Test the quantized checkpoint on your own prompts and acceptance criteria before deployment.",[10,500,501,502,507],{},"The public ",[20,503,506],{"href":504,"rel":505},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Fblob\u002Fmain\u002Fmodels\u002FOrnith-1.5\u002FBENCHMARKS.md",[24],"Ornith benchmark record"," contains the publication workload, stock settings and reproduction command.",[71,509,511],{"id":510},"run-ornith-15-locally","Run Ornith 1.5 locally",[10,513,514],{},"This is a checked Linux and Docker path for one Radeon AI PRO R9700 with ROCm 7.14. Start the server and open the included terminal chat with one command:",[516,517,522],"pre",{"className":518,"code":519,"language":520,"meta":521,"style":521},"language-bash shiki shiki-themes github-light github-dark","git clone --depth 1 https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin.git \\\n  && cd paiton-vllm-plugin \\\n  && .\u002Fmodels\u002FOrnith-1.5\u002Fserve-docker.sh --chat\n","bash","",[482,523,524,550,565],{"__ignoreMap":521},[525,526,529,533,537,541,544,547],"span",{"class":527,"line":528},"line",1,[525,530,532],{"class":531},"sScJk","git",[525,534,536],{"class":535},"sZZnC"," clone",[525,538,540],{"class":539},"sj4cs"," --depth",[525,542,543],{"class":539}," 1",[525,545,546],{"class":535}," https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin.git",[525,548,549],{"class":539}," \\\n",[525,551,553,557,560,563],{"class":527,"line":552},2,[525,554,556],{"class":555},"sVt8B","  && ",[525,558,559],{"class":539},"cd",[525,561,562],{"class":535}," paiton-vllm-plugin",[525,564,549],{"class":539},[525,566,568,570,573],{"class":527,"line":567},3,[525,569,556],{"class":555},[525,571,572],{"class":531},".\u002Fmodels\u002FOrnith-1.5\u002Fserve-docker.sh",[525,574,575],{"class":539}," --chat\n",[10,577,578],{},"The first start downloads the pinned public target and DFlash draft, then creates a ready-to-run copy in a persistent Docker volume. The same server exposes an OpenAI-compatible endpoint at:",[516,580,585],{"className":581,"code":583,"language":584,"meta":521},[582],"language-text","http:\u002F\u002F127.0.0.1:8000\u002Fv1\u002Fchat\u002Fcompletions\n","text",[482,586,583],{"__ignoreMap":521},[10,588,589,590,593,594,599],{},"Use model name ",[482,591,592],{},"ornith",". The public ",[20,595,598],{"href":596,"rel":597},"https:\u002F\u002Fgithub.com\u002FEliovp-BV\u002Fpaiton-vllm-plugin\u002Ftree\u002Fmain\u002Fmodels\u002FOrnith-1.5",[24],"package guide"," includes the direct Docker command, the DFlash-off option and the exact operating scope.",[10,601,602],{},"The image contains the runtime plugin and compiled Paiton artifact. It contains no model weights and no Paiton compiler source.",[71,604,606],{"id":605},"tested-configuration","Tested configuration",[98,608,609,619],{},[101,610,611],{},[104,612,613,616],{},[107,614,615],{},"Component",[107,617,618],{},"Tested value",[114,620,621,631,639,649,657,665,675,685,695,703,710,716,724],{},[104,622,623,625],{},[119,624,121],{},[119,626,627,628],{},"AMD Radeon AI PRO R9700, 32 GB, ",[482,629,630],{},"gfx1201",[104,632,633,636],{},[119,634,635],{},"Model",[119,637,638],{},"Ornith 1.5 35B A3B MXFP4 Quark RDNA4",[104,640,641,644],{},[119,642,643],{},"Model revision",[119,645,646],{},[482,647,648],{},"9e488f46c0f7969f84c9923ee0256311cd50316e",[104,650,651,654],{},[119,652,653],{},"Target \u002F draft size",[119,655,656],{},"22.9 GB \u002F 772 MB",[104,658,659,662],{},[119,660,661],{},"ROCm",[119,663,664],{},"7.14",[104,666,667,670],{},[119,668,669],{},"vLLM revision",[119,671,672],{},[482,673,674],{},"39bd959b582c85e78e7e0326d49042ce7c3c07ed",[104,676,677,680],{},[119,678,679],{},"Paiton image",[119,681,682],{},[482,683,684],{},"ghcr.io\u002Feliovp\u002Fpaiton-vllm-plugin:ornith15-mxfp4-rdna4-v1.0.0",[104,686,687,690],{},[119,688,689],{},"Tested digest",[119,691,692],{},[482,693,694],{},"sha256:f8feb0ea85e36f681eaf4f6c1d534551e4e0d1d98bf479f1cb6b6236a5893de2",[104,696,697,700],{},[119,698,699],{},"Tensor parallelism",[119,701,702],{},"1",[104,704,705,708],{},[119,706,707],{},"Maximum active sequences",[119,709,702],{},[104,711,712,714],{},[119,713,153],{},[119,715,156],{},[104,717,718,721],{},[119,719,720],{},"KV cache reservation",[119,722,723],{},"3 GiB",[104,725,726,729],{},[119,727,728],{},"Qualified scope",[119,730,731],{},"Text-only, single-user inference",[10,733,734],{},"The package fails closed outside its supported GPU, checkpoint and serving contract. Additional requests queue instead of entering an unqualified concurrent path.",[71,736,738],{"id":737},"from-a-workstation-result-to-production-economics","From a workstation result to production economics",[10,740,741],{},"The Qwen3.8 release showed Paiton improving three different prompt shapes on the R9700. Ornith adds a distinct proof point: a 35B MoE target and speculative draft fit on one 32 GB card, serve through a familiar API and generate at 44.63 output tok\u002Fs.",[10,743,744],{},"For local users, that means a faster private model endpoint without a hosted per-token dependency. For operators running AMD Instinct CDNA infrastructure, the leverage is larger: every percentage point compounds across accelerators, models and billions of generated tokens.",[10,746,747,748,752],{},"Running an AMD token factory? ",[20,749,751],{"href":750},"\u002Fproducts\u002Fpaiton","Bring us the model, traffic shape and current baseline",". We will show where throughput, latency and cost per generated token can move.",[754,755,756],"style",{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"title":521,"searchDepth":552,"depth":552,"links":758},[759,760,761,762,763,764,765,766],{"id":73,"depth":552,"text":74},{"id":209,"depth":552,"text":210},{"id":316,"depth":552,"text":317},{"id":368,"depth":552,"text":369},{"id":473,"depth":552,"text":474},{"id":510,"depth":552,"text":511},{"id":605,"depth":552,"text":606},{"id":737,"depth":552,"text":738},[768,769,770,771,772,773,774,775,776,777],"Paiton","Artificial Intelligence","AMD Radeon","AI Inference","GPU Performance","Inference Latency","Inference Optimization","Large Language Models","vLLM","Cost Efficiency","2026-09-05T09:00:00","Paiton serves Ornith 1.5 35B A3B at 44.63 output tok\u002Fs on one Radeon AI PRO R9700, 27% faster than tuned stock vLLM, with 21.3% lower modeled cost.","md","A 35B MoE at 44.6 tok\u002Fs on one 32 GB GPU","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-ornith15\u002F00-featured-ornith15-r9700.webp",{},true,"https:\u002F\u002Feliovp.com\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700","\u002Fblog\u002Fpaiton-ornith15-radeon-ai-pro-r9700",{"title":5,"description":779},"paiton-ornith15-radeon-ai-pro-r9700","blog\u002Fpaiton-ornith15-radeon-ai-pro-r9700",null,"wBIk4SuVhU12gj05P5HO3R9GXAGV9tzFOvN-O3VYWs4",[793,804,815,827,837,839,847,861,892,904,926,944,963,981,999,1015,1027,1043,1058,1070,1079,1087,1102,1114,1125,1136,1146,1159,1169,1182,1193,1203,1214,1223,1235,1246,1255],{"path":794,"title":795,"description":796,"date":797,"slug":798,"image":799,"originalUrl":800,"categories":801},"\u002Fblog\u002Fpaiton-qwen38-mxfp4-dflash2-r9700","Qwen3.8: 400.7 tok\u002Fs on R9700 | Paiton","Qwen3.8 on one R9700: 400.7 aggregate tok\u002Fs with ROCm 10 and vLLM 0.29, plus public 200K\u002F220K chat profiles. Benchmarks, limits and launch commands.","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",[768,770,776,802,774,803],"Qwen3.8","DFlash2",{"path":805,"title":806,"description":807,"date":808,"slug":809,"image":810,"originalUrl":811,"categories":812},"\u002Fblog\u002Fpaiton-qwen38-neo-gguf-vllm-r9700","Qwen3.8 GGUF in vLLM: Faster Responses on One Radeon","Run the original NEO CODER MAX GGUF in vLLM with Paiton on an R9700. Explore measured latency gains, image input and local deployment.","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",[768,770,813,814,776],"Local AI","GGUF",{"path":816,"title":817,"description":818,"date":819,"slug":820,"image":821,"originalUrl":822,"categories":823},"\u002Fblog\u002Fpaiton-minimax-h3-radeon-ai-pro-r9700","MiniMax H3 on Radeon: 15-Second Video With Native Sound","Paiton generates a 15-second MiniMax H3 video with stereo audio on one Radeon AI PRO R9700 in 5m 33s, with 16.7% lower latency than matched 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",[768,770,813,824,825,826],"Video Generation","MiniMax H3","ComfyUI",{"path":828,"title":829,"description":830,"date":831,"slug":832,"image":833,"originalUrl":828,"categories":834},"\u002Fblog\u002Fpaiton-flux2-klein-radeon-ai-pro-r9700","Local FLUX.2 klein on Radeon AI PRO R9700: Faster Image Generation with Less VRAM","Paiton generates 1024×1024 FLUX.2 klein images in 1.054 seconds on an R9700: 16.2% lower warm latency and 33.4% lower peak Torch allocation.","2026-09-07T09:00:00","paiton-flux2-klein-radeon-ai-pro-r9700","\u002Fasset\u002Fimages\u002Fblog\u002Fpaiton-flux2-klein\u002Ffox-paiton.webp",[768,770,813,835,836,826],"Image Generation","FLUX",{"path":786,"title":5,"description":779,"date":778,"slug":788,"image":782,"originalUrl":785,"categories":838},[768,769,770,771,772,773,774,775,776,777],{"path":51,"title":840,"description":841,"date":842,"slug":843,"image":844,"originalUrl":845,"categories":846},"Paiton: 21% More Qwen3.8 Throughput on Radeon AI PRO R9700","Paiton serves AMD’s Qwen3.8 27B at 39.77 output tokens\u002Fs on one Radeon AI PRO R9700, delivering 21% more throughput and 17.4% lower modeled cost.","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",[768,769,770,771,772,773,774,775,776,777],{"path":848,"title":849,"description":850,"date":851,"slug":852,"image":853,"originalUrl":790,"categories":854},"\u002Fblog\u002Fai-data-center-power-requirements-gpu-per-megawatt","AI Data Center Power Requirements: The GPU\u002FMW Illusion","Why do AI data-center proposals quote different GPU capacities? See how PUE, peak loads, storage, networking and cooling determine deployable compute.","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",[855,856,857,858,859,860],"All","AI Infrastructure","Data Centers","ModFlex","HPC","AMD Helios",{"path":862,"title":863,"description":864,"date":865,"slug":866,"image":867,"originalUrl":868,"categories":869},"\u002Fblog\u002Fpaiton-returns-to-its-diffusion-roots-optimizing-wan2-2-t2v-a14b-on-amd-mi355x","Wan2.2 Video Generation: Paiton on AMD MI355X","Compare Wan2.2-T2V-A14B video generation on AMD MI355X with Paiton and NVIDIA B200 using Diffusers, and explore our diffusion optimization approach.","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",[855,769,768,870,871,872,873,874,875,876,877,878,121,879,880,881,882,883,884,885,768,886,887,888,889,890,891],"14B","AMD","B200","Benchmarks","Blackwell","Compute","Diffusion","Eliovp","Generative AI","Hardware","Inference","Instinct","MI355x","NVidia","On-Premise","Optimization","Sovereign AI","T2V","Text-to-Video","Tuning","Video-Generation","Wan2.2",{"path":893,"title":894,"description":895,"date":896,"slug":897,"image":898,"originalUrl":899,"categories":900},"\u002Fblog\u002Ffrom-the-attic-to-the-front-page-eliovp-recognized-as-a-pioneer-in-chip-optimization-data-center-infrastructure","ElioVP in De Tijd: Chip Optimization and Data Centers","Read about De Tijd's coverage of ElioVP, from its origins in chip optimization to its work on modular data centers and high-density cooling.","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",[855,769,901,902,871,903,901,883],"Modular DC","Uncategorized","De Tijd",{"path":905,"title":906,"description":907,"date":908,"slug":909,"image":910,"originalUrl":911,"categories":912},"\u002Fblog\u002Fprivacy-is-geen-it-probleem-meer-het-is-een-strategische-prioriteit","AI Privacy: A Strategic Priority for Benelux Businesses","Explore generative AI privacy risks, trust, data retention and governance, and why Benelux businesses need a strategic approach to secure AI.","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",[855,769,913,902,914,915,916,917,918,919,920,921,877,922,878,923,924,925],"Trending","AI Act","Anthropomorphism","AVG","Benelux","ChatGPT","Cybersecurity","Data Governance","Data Security","GDPR","Microsoft Copilot","Privacy","Shadow AI",{"path":927,"title":928,"description":929,"date":930,"slug":931,"image":932,"originalUrl":933,"categories":934},"\u002Fblog\u002Fitsme-bij-ons-is-het-its-not-me-en-dit-is-waarom","Why We Do Not Use itsme: Privacy and Data Sovereignty","Why ElioVP does not use itsme: our assessment of identity metadata, cloud dependence, data sovereignty and authentication risks.","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",[855,935,936,937,919,938,939,940,922,941,942,943,924],"AWS","Belgian Mobile ID","Cloud Act","Data Sovereignty","Digital Identity","eIDAS","itsme","Liberty Global","MyGov.be",{"path":945,"title":946,"description":947,"date":948,"slug":949,"image":950,"originalUrl":951,"categories":952},"\u002Fblog\u002Ffield-report-the-reality-of-building-agentic-ai-in-2025","Field Report. The Reality of Building Agentic AI in 2025","Lessons from building on-premise AI agents in 2025 cover workflow design, observability, model training, hallucinations and GPU memory limits.","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",[855,769,953,913,954,955,956,957,958,959,960,961,886,962],"Solutions","Agentic AI","AI Engineering","AI Strategy","Autonomous Agents","Enterprise AI","Local LLM","Model Fine-Tuning","On-Premise AI","VRAM Optimization",{"path":964,"title":965,"description":966,"date":967,"slug":968,"image":969,"originalUrl":970,"categories":971},"\u002Fblog\u002Fthe-synthetic-unicorn-bubble","The Synthetic Unicorn Bubble","An analysis of AI neocloud investment risks, examining circular financing, infrastructure claims, contract terms and 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",[855,769,913,856,972,973,974,975,976,977,978,979,980],"AI Neocloud","Circular Financing","GPU Cloud","Investment Risks","Startup Valuation","Synthetic Bubble","Tech Analysis","Vaporware","Venture Capital",{"path":982,"title":983,"description":984,"date":985,"slug":986,"image":987,"originalUrl":988,"categories":989},"\u002Fblog\u002Fbuilding-the-engine-for-the-ai-race-the-4-month-path-to-nvidia-gb300-nvl72-power","NVIDIA GB300 NVL72: A Four-Month Modular Data Center Plan","Explore a modular data center design for NVIDIA GB300 NVL72, covering redundant power, hybrid cooling and a four-month deployment plan.","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",[855,901,902,990,856,991,992,993,994,995,996,997,998],"150kW Rack","DLC","High Density","Liquid Cooling","Modular Data Center","NVIDIA Blackwell Ultra","NVIDIA GB300","NVL72","Rapid Deployment",{"path":1000,"title":1001,"description":1002,"date":1003,"slug":1004,"image":1005,"originalUrl":1006,"categories":1007},"\u002Fblog\u002Fwhy-cuda-translation-wont-unlock-amds-real-potential","Why “CUDA” Translation Won’t Unlock AMD’s Real Potential","Why CUDA compatibility is not the same as AMD performance: explore ROCm, HIP, kernel tuning and the case for hardware-specific optimization.","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",[855,769,768,902,1008,769,1009,1010,1011,1012,1013,1014,768,661],"AMD MI300X","CUDA Translation","FP8","GPU Optimization","High Performance Computing","HIP","Kernel Tuning",{"path":1016,"title":1017,"description":1018,"date":1019,"slug":1020,"image":1021,"originalUrl":1022,"categories":1023},"\u002Fblog\u002Fpaiton-the-simplest-way-to-supercharge-ai-inference","Paiton: The Simplest Way to Supercharge AI Inference","Learn how Paiton integrates with existing inference stacks, with AMD MI300X benchmark results and performance-per-dollar comparisons.","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",[855,769,768,771,1024,1008,777,1025,774,1014,768,1026,776],"AMD Instinct","High Throughput","SGLang",{"path":1028,"title":1029,"description":1030,"date":1031,"slug":1032,"image":1033,"originalUrl":1034,"categories":1035},"\u002Fblog\u002Fstop-overpaying-paiton-mi300x-moe-beats-h200-b200-on-1m-tokens","Paiton MoE Benchmarks: MI300X vs H200 and B200","Compare Qwen3-30B-A3B MoE inference with Paiton on MI300X against H200 and B200, including throughput and cost per million 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",[855,769,768,1036,1008,1037,774,1038,1039,1040,1041,768,1042],"AI Benchmarks","Cost per Token","Mixture of Experts","MoE","NVIDIA B200","NVIDIA H200","Qwen3",{"path":1044,"title":1045,"description":1046,"date":1047,"slug":1048,"image":1049,"originalUrl":1050,"categories":1051},"\u002Fblog\u002Fagentic-ai-but-make-it-local-from-inbox-to-insight-to-action-en","Local Agentic AI: From Inbox to Action","Local-first AI agents turn email, documents and images into tickets, reports and actions, using models tailored to your data and systems.","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",[855,769,953,902,954,1052,1053,1054,1055,959,961,886,1056,1057],"Damage Detection","Document Processing","Email Automation","Invoice Extraction","Ticket Automation","Workflow Automation",{"path":1059,"title":1060,"description":1061,"date":1062,"slug":1063,"image":1064,"originalUrl":1065,"categories":1066},"\u002Fblog\u002Fmi300x-fp8-data-parallel-benchmarks-8-64-gpus-h200-left-behind-b200-within-reach","MI300X FP8 Benchmarks: GPU Partitioning with Paiton","Explore Llama 3.1 8B FP8 benchmarks on partitioned MI300X GPUs with Paiton, comparing throughput and latency against NVIDIA H200 and 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",[855,769,768,902,1067,871,872,1068,1069,882,883,768,776],"AI","H200","MI300X",{"path":1071,"title":1072,"description":1073,"date":1074,"slug":1075,"image":1076,"originalUrl":1077,"categories":1078},"\u002Fblog\u002Fapplicable-ai-for-businesses","Applicable AI for Businesses","Explore ElioVP's approach to local AI for business workflows, including custom model training and automated damage detection for logistics.","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",[855,769,953,954,1052,1053,1054,1055,959,961,886,1056,1057],{"path":1080,"title":1081,"description":1082,"date":1083,"slug":1084,"image":521,"originalUrl":1085,"categories":1086},"\u002Fblog\u002Fintroducing-paitons-free-evaluation-models","Introducing Paiton’s Free Evaluation Models","Test Paiton with free evaluation models for AMD GPUs. Compare text, vision and image generation performance using your own workloads.","2025-07-07T11:26:13","introducing-paitons-free-evaluation-models","https:\u002F\u002Feliovp.com\u002Fintroducing-paitons-free-evaluation-models\u002F",[855,769,768],{"path":1088,"title":1089,"description":1090,"date":1091,"slug":1092,"image":1093,"originalUrl":1094,"categories":1095},"\u002Fblog\u002Fpaiton-dramatically-faster-startup-and-performance-for-llama-3-1-405b","Llama 3.1 405B: Faster Startup with Paiton on MI300X","See Paiton benchmarks for Llama 3.1 405B on eight AMD MI300X GPUs, covering model startup, tensor parallelism, throughput and 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",[855,769,768,902,771,1008,1096,1010,1097,1098,1099,768,1100,1101],"Cold Start","Graph Compilation","Llama 3.1 405B","LLM Optimization","Startup Latency","Tensor Parallelism",{"path":1103,"title":1104,"description":1105,"date":1106,"slug":1107,"image":1108,"originalUrl":1109,"categories":1110},"\u002Fblog\u002Fpaiton-fp8-beats-nvidias-h200-on-amds-mi300x","Paiton FP8 Beats NVIDIA’s H200 on AMD’s MI300X","Compare Paiton on AMD MI300X with NVIDIA H200 for Llama 3.1 70B FP8, including throughput, first-token delay and latency across batch sizes.","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",[855,769,768,902,1008,1111,958,878,772,773,775,1098,1112,1113],"Cold Start Optimization","Model Serving","vLLM Optimization",{"path":1115,"title":1116,"description":1117,"date":1118,"slug":1119,"image":1120,"originalUrl":1121,"categories":1122},"\u002Fblog\u002Fmi300x-vs-h200-vs-rx-7900-xtx-vs-tenstorrent-n300s-with-vllm","MI300X vs H200 vs RX 7900 XTX vs Tenstorrent n300s with vLLM","Compare MI300X, H200, RX 7900 XTX and Tenstorrent n300s on Llama 3 8B with vLLM, including throughput, modeled token costs and hardware limits.","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",[855,769,768,953,902,871,1069,883,1123,1124],"RX7900XTX","tenstorrent",{"path":1126,"title":1127,"description":1128,"date":1129,"slug":1130,"image":1131,"originalUrl":1132,"categories":1133},"\u002Fblog\u002Fclusterpl-empowering-gpu-cluster-investors-with-real-world-financial-insights","ClusterP&L: Financial Modeling for GPU Clusters","Explore how ClusterP&L models GPU cluster costs, profitability and investment scenarios, with ROI metrics, risk simulations and exportable reports.","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",[855,769,901,953,872,1068,1134,883,1135],"MI325x","pnl calculator",{"path":1137,"title":1138,"description":1139,"date":1140,"slug":1141,"image":1142,"originalUrl":1143,"categories":1144},"\u002Fblog\u002Fcranking-out-faster-tokens-for-fewer-dollars-amd-mi300x-vs-nvidia-h200","AMD MI300X vs. NVIDIA H200: Qwen3-32B with Paiton","Compare Qwen3-32B benchmarks on Paiton-optimized AMD MI300X and NVIDIA H200, covering throughput, latency and hardware costs.","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",[855,769,768,1067,871,1068,1145,883,768,776],"MI300",{"path":1147,"title":1148,"description":1149,"date":1150,"slug":1151,"image":1152,"originalUrl":1153,"categories":1154},"\u002Fblog\u002Fpower-meets-precision-high-density-modular-data-center-for-nvidia-nvl-deployments-1-2-mw","Modular Data Centers for NVIDIA NVL: 1 to 2 MW","Explore modular data center designs for NVIDIA NVL systems, covering power capacity, liquid cooling, redundancy and deployment planning.","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",[855,901,1155,856,992,859,993,994,1156,1157,997,1158],"1-2MW Data Center","NVIDIA Blackwell","NVIDIA NVL","Precision Cooling",{"path":1160,"title":1161,"description":1162,"date":1163,"slug":1164,"image":1165,"originalUrl":1166,"categories":1167},"\u002Fblog\u002Fexamining-ai-agents-in-the-medical-field-ai-that-speaks-dicom","Examining AI agents in the medical field: AI that speaks DICOM","Explore a local AI agent demo for DICOM workflows, from patient and study retrieval to a comparison of vision models using anonymized medical images.","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",[855,769,953,902,1067,871,1168],"Healthcare",{"path":1170,"title":1171,"description":1172,"date":1173,"slug":1174,"image":1175,"originalUrl":1176,"categories":1177},"\u002Fblog\u002Feliovp-bv-your-trusted-partner-for-supply-chain-resilience-amidst-new-u-s-tariffs","U.S. Tariffs and AI Supply Chain Resilience: April 2025","Read ElioVP's April 2025 perspective on U.S. tariffs and supply chain resilience for AI servers, HPC systems and modular data centers.","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",[855,913,1067,871,1178,1179,1180,1181],"import","Taiwan","Tariffs","Trump",{"path":1183,"title":1184,"description":1185,"date":1186,"slug":1187,"image":1188,"originalUrl":1189,"categories":1190},"\u002Fblog\u002Fwhy-ai-agents-are-the-future","Why AI Agents Are the Future","Explore AI agents for ERP, CRM, finance and customer support, with practical use cases and a path from workflow assessment to pilot and deployment.","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",[855,769,953,1067,1191,1192],"AI Agents","ERP",{"path":1194,"title":1195,"description":1196,"date":1197,"slug":1198,"image":1199,"originalUrl":1200,"categories":1201},"\u002Fblog\u002Fthe-rise-of-open-source-ai-model-optimization","The Rise of Open-Source AI Model Optimization","Explore open-source AI optimization trends, from quantization and mixture-of-experts models to hardware-aware tuning, RAG and edge deployment.","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",[855,769,913,1202,871,121,883],"AI news",{"path":1204,"title":1205,"description":1206,"date":1207,"slug":1208,"image":1209,"originalUrl":1210,"categories":1211},"\u002Fblog\u002Fintroducing-our-benchmarking-tool-powered-by-dstack","Introducing Our Benchmarking Tool: Powered by dstack","Explore our dstack-powered tool for reproducible vLLM benchmarks, automated parameter sweeps and performance reports across local and cloud GPUs.","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",[855,769,768,1067,871,1212,1213,1069,768],"benchmark","LLM",{"path":1215,"title":1216,"description":1217,"date":1218,"slug":1219,"image":1220,"originalUrl":1221,"categories":1222},"\u002Fblog\u002Foptimizing-qwq-32b-by-qwen-amd-mi300x-vs-nvidia-h200","Optimizing QwQ-32B (by Qwen): AMD MI300X vs. NVIDIA H200","Compare QwQ-32B throughput and latency on AMD MI300X with Paiton and NVIDIA H200, from small batches to higher concurrency.","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",[855,769,768],{"path":1224,"title":1225,"description":1226,"date":1227,"slug":1228,"image":1229,"originalUrl":1230,"categories":1231},"\u002Fblog\u002Feliovp-featured-on-amd-tech-talk-podcast","Eliovp Featured on AMD “Tech Talk” Podcast","Listen to Elio Van Puyvelde and Jim Greene on AMD's Tech Talk podcast, discussing ElioVP's origins and its AI hardware and software services.","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",[855,871,1232,1233,1234],"Jim Greene","Podcast","Tech Talk",{"path":1236,"title":1237,"description":1238,"date":1239,"slug":1240,"image":1241,"originalUrl":1242,"categories":1243},"\u002Fblog\u002Ffurther-optimizing-amd-powered-inference-with-paiton","Further Optimizing AMD-Powered Inference with Paiton","Explore Paiton's DeepSeek R1 Distill Llama 8B benchmarks on AMD MI300X, focusing on throughput and first-token latency at smaller batch sizes.","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",[855,769,768,871,1244,1245,1068,1069,1134,768,776],"Deepseek","H100",{"path":1247,"title":1248,"description":1249,"date":1250,"slug":1251,"image":1252,"originalUrl":1253,"categories":1254},"\u002Fblog\u002Fa-first-look-at-paiton-in-action-deepseek-r1-distill-llama-3-1-8b","Paiton Benchmarks: DeepSeek R1 Distill Llama 3.1 8B","Compare stock and Paiton-optimized DeepSeek R1 Distill Llama 3.1 8B on AMD MI300X, with throughput and latency benchmarks across batch sizes.","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",[855,769,768,871,1244,1245,1068,1069,1134,768,776],{"path":1256,"title":1257,"description":1258,"date":1259,"slug":1260,"image":1261,"originalUrl":1262,"categories":1263},"\u002Fblog\u002Fai-model-optimization-with-paiton","AI Model Optimization with Paiton","Learn how Paiton uses model compilation, custom kernels and kernel fusion to optimize AI inference on AMD GPUs.","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",[855,769,768,871,1245,1068,1069,1134,768,776],1789853166443]