Meta tags:
description= cuda content on DEV Community;
keywords= software development, engineering, cuda;
Headings (most frequently used words):
on, cuda, gemma, gpu, and, tesla, t4, in, nvidia, the, why, it, gb, part, gib, at, bf16, model, rust, than, laptop, to, qat, memory, posts, dev, community, int4, embeddings, serve, e2b, 86, 30x, how, fast, can, 421m, parameter, decision, run, benchmarked, laya, across, gpus, exploring, result, visibility, of, fixed, latency, instructions, sm120, architecture, just, let, into, here, that, bigger, deal, sounds, two, native, tracks, not, wrapper, over, vs, core, cpu, re, measured, abba, order, 1x, reverse, engineering, modifying, binary, minimum, gce, vm, script, drive, finding, random, island, with, geometry, weights, decode, 79x, faster, same, nvjpeg2000, different, numbers, timer, boundaries, frames, flight, qwen3, 27b, 256k, 50, tps, 24, understanding, vram, bandwidth, your, won, fit, an, 2021, takes, from, sm_120, shared, cliff, fp8, kv, cache, crashes, vllm, workstation, blackwell, trending, guides, resources,
Text of the page (most frequently used words):
cuda (34), and (24), gemma (23), gpu (20), the (17), comments (17), follow (16), min (15), read (15), xbill (15), nvidia (14), comment (13), dev (12), sep (11), add (10), with (9), for (9), rust (9), tesla (9), vllm (9), why (8), than (6), qat (6), bf16 (6), part (6), gib (6), #laptop (6), memory (6), that (5), model (5), reactions (5), google (5), developer (5), experts (5), reaction (5), community (4), g5g (4), same (4), parameter (4), qwen3 (4), geometry (4), cpu (4), workstation (4), blackwell (4), fp8 (4), sm_120 (4), posts (4), performance (4), aug (4), create (3), software (3), tracks (3), your (3), llm (3), how (3), fast (3), can (3), 421m (3), decision (3), run (3), benchmarked (3), laya (3), across (3), gpus (3), reverse (3), engineering (3), modifying (3), binary (3), exploring (3), result (3), visibility (3), fixed (3), latency (3), instructions (3), sm120 (3), architecture (3), 27b (3), 256k (3), tps (3), just (3), let (3), into (3), here (3), bigger (3), deal (3), sounds (3), weights (3), decode (3), 79x (3), faster (3), finding (3), random (3), island (3), int4 (3), embeddings (3), serve (3), e2b (3), 30x (3), core (3), measured (3), abba (3), order (3), minimum (3), gce (3), script (3), drive (3), nvjpeg2000 (3), different (3), numbers (3), timer (3), boundaries (3), frames (3), flight (3), 2021 (3), takes (3), from (3), shared (3), cliff (3), cache (3), crashes (3), jahn (3), vram (3), aarush (3), karak (3), michał (3), piszczek (3), fyodor (3), serzhenko (3), machinelearning (3), muhammad (3), adil (3), stjepan (3), juan (3), torchia (3), ashraf (3), comradepenguin (3), bhushan (3), kinge (3), menu (3), account (2), log (2), built (2), code (2), what (2), serving (2), accelerating (2), running (2), ec2 (2), graviton2 (2), amd (2), cpp (2), side (2), ability (2), sort (2), top (2), latest (2), relevant (2), sign (2), llamacpp (2), understanding (2), bandwidth (2), won (2), fit (2), two (2), native (2), not (2), wrapper (2), over (2), search (2), place, where, coders, share, stay, date, grow, their, careers, made, love, 2016, 2026, ruby, rails, powers, other, inclusive, communities, open, source, forem, terms, use, privacy, policy, conduct, mlh, shop, free, postgres, database, contact, about, showcase, organization, accounts, advertise, help, education, videos, challenges, home, space, discuss, keep, development, manage, career, pure, jax, changes, between, turing, ada, doesn, throughput, physical, workloads, tensorrt, mscred, im2col, gemm, custom, pytorch, extension, engine, streams, 744b, models, consumer, hardware, sglang, llama, plus, pass, gemma4, cheapest, aws, has, arm, you, probably, want, intel, one, rtx, 5090, survival, guide, xformers, trap, installing, graviton, walk, through, trending, guides, resources, mcp, aiinfrastructure, geospatial, gpucomputing, gcp, hex, benchmarking, gpuprogramming, english, programming, opensource, right, left, older, post, hide, close, powered, algolia, navigation, skip, content,
Text of the page (random words):
cuda dev community skip to content navigation menu search powered by algolia search log in create account dev community close cuda follow hide create post older cuda posts 1 2 3 4 5 6 7 8 9 posts left menu sign in for the ability to sort posts by relevant latest or top right menu gemma 4 on a tesla t4 part 3 int4 embeddings serve e2b in 2 86 gib at 2 30x bf16 xbill xbill xbill follow for google developer experts sep 30 gemma 4 on a tesla t4 part 3 int4 embeddings serve e2b in 2 86 gib at 2 30x bf16 gemma vllm cuda machinelearning 9 reactions comments add comment 11 min read how fast can a 421m parameter decision model run i benchmarked laya across nvidia gpus bhushan kinge bhushan kinge bhushan kinge follow sep 24 how fast can a 421m parameter decision model run i benchmarked laya across nvidia gpus machinelearning cuda performance opensource comments 1 comment 6 min read exploring result visibility of fixed latency instructions on the sm120 architecture comradepenguin comradepenguin comradepenguin follow sep 18 exploring result visibility of fixed latency instructions on the sm120 architecture cuda nvidia gpu 1 reaction comments add comment 10 min read nvidia just let rust into cuda here s why that s a bigger deal than it sounds ashraf ashraf ashraf follow sep 18 nvidia just let rust into cuda here s why that s a bigger deal than it sounds rust ai cuda programming 1 reaction comments add comment 5 min read cuda rust two native tracks not a wrapper over c juan torchia juan torchia juan torchia follow sep 17 cuda rust two native tracks not a wrapper over c english rust cuda gpuprogramming 1 reaction comments add comment 5 min read a 4 gb laptop gpu vs a 6 core cpu on gemma 4 re measured in abba order 4 1x xbill xbill xbill follow for google developer experts sep 23 a 4 gb laptop gpu vs a 6 core cpu on gemma 4 re measured in abba order 4 1x gemma llamacpp cuda benchmarking 10 reactions comments 2 comments 11 min read reverse engineering nvidia modifying a cuda binary stjepan stjepan stjepan follow sep 6 reverse engineering nvidia modifying a cuda binary nvidia gpu cuda hex comments add comment 4 min read gemma 4 on a tesla t4 part 2 the minimum gce vm and a script to drive it xbill xbill xbill follow for google developer experts sep 22 gemma 4 on a tesla t4 part 2 the minimum gce vm and a script to drive it gemma vllm gcp cuda 13 reactions comments add comment 13 min read finding a random island with geometry and cuda muhammad adil muhammad adil muhammad adil follow aug 19 finding a random island with geometry and cuda gpucomputing geospatial cuda geometry comments add comment 4 min read gemma 4 on a tesla t4 qat weights decode 1 79x faster than bf16 xbill xbill xbill follow for google developer experts sep 18 gemma 4 on a tesla t4 qat weights decode 1 79x faster than bf16 gemma vllm cuda machinelearning 8 reactions comments 1 comment 9 min read same nvjpeg2000 different numbers timer boundaries and frames in flight fyodor serzhenko fyodor serzhenko fyodor serzhenko follow sep 18 same nvjpeg2000 different numbers timer boundaries and frames in flight cuda gpu performance cpp comments add comment 13 min read qwen3 8 27b at 256k 50 tps on a 24 gb gpu michał piszczek michał piszczek michał piszczek follow aug 17 qwen3 8 27b at 256k 50 tps on a 24 gb gpu aiinfrastructure llm cuda performance 1 reaction comments 1 comment 11 min read understanding gpu memory vram bandwidth and why your model won t fit aarush karak aarush karak aarush karak follow aug 13 understanding gpu memory vram bandwidth and why your model won t fit gpu cuda vram memory 1 reaction comments add comment 2 min read gemma 4 on an 2021 4 gb laptop gpu qat takes it from 9 5 gib to 1 6 xbill xbill xbill follow for google developer experts sep 10 gemma 4 on an 2021 4 gb laptop gpu qat takes it from 9 5 gib to 1 6 gemma llamacpp mcp cuda 11 reactions comments 3 comments 13 min read the sm_120 shared memory cliff why fp8 kv cache crashes vllm on workstation blackwell jahn jahn jahn follow aug 25 the sm_120 shared memory cliff why fp8 kv cache crashes vllm on workstation blackwell cuda llm gpu performance comments add comment 3 min read sign in for the ability to sort posts by relevant latest or top trending guides resources installing rust for vllm on graviton a g5g walk through rtx 5090 survival guide sm_120 cuda 12 and 13 side by side and the xformers trap the cheapest cuda gpu on aws has an arm cpu and you probably want the intel one serving gemma4 with rust on vllm running gemma 4 on ec2 g5g graviton2 amd with nvidia gpu the sm_120 shared memory cliff why fp8 kv cache crashes vllm on workstation blackwell qwen3 8b on workstation blackwell vllm vs sglang vs llama cpp plus an fp8 pass gemma 4 on an 2021 4 gb laptop gpu qat takes it from 9 5 gib to 1 6 same nvjpeg2000 different numbers timer boundaries and frames in flight running gemma 4 on ec2 g5g graviton2 amd with nvidia gpu gemma 4 on a tesla t4 part 2 the minimum gce vm and a script to drive it a 4 gb laptop gpu vs a 6 core cpu on gemma 4 re measured in abba order 4 1x gemma 4 on a tesla t4 part 3 int4 embeddings serve e2b in 2 86 gib at 2 30x bf16 finding a random island with geometry and cuda i built a cuda engine that streams 744b parameter ai models on consumer hardware gemma 4 on a tesla t4 qat weights decode 1 79x faster than bf16 nvidia just let rust into cuda here s why that s a bigger deal than it sounds qwen3 8 27b at 256k 50 tps on a 24 gb gpu exploring result visibility of fixed latency instructions on the sm120 architecture reverse engineering nvidia modifying a cuda binary gpu accelerating mscred with cuda im2col gemm and a custom pytorch extension how fast can a 421m parameter decision model run i benchmarked laya across nvidia gpus accelerating physical ai workloads with nvidia cuda and tensorrt g5g vs g6 for llm serving the same code and 3 7x the throughput gemma 4 in pure jax what changes between turing and ada and what doesn t dev community a space to discuss and keep up software development and manage your software career home dev challenges dev videos dev education tracks dev help advertise on dev organization accounts dev showcase about contact free postgres database dev shop mlh code of conduct privacy policy terms of use built on forem the open source software that powers dev and other inclusive communities made with love and ruby on rails dev community 2016 2026 we re a place where coders share stay up to date and grow their careers log in create account
|
