Research

DGX Spark vs M5 Max vs RTX 5090: Local LLM Throughput

Three-way 2026 throughput comparison of NVIDIA DGX Spark, Mac Studio M5 Max 128GB, and GeForce RTX 5090 32GB for local LLM inference. Decode speed, prefill, memory ceiling, and which model sizes each one can actually hold.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: July 2026

The two-way DGX Spark versus Mac Studio comparison misses the machine most people actually buy. A GeForce RTX 5090 is faster than either at everything that fits in its 32GB of GDDR7, and slower than both at everything that does not. That single trade, raw speed against memory ceiling, is the whole decision.

The Three Machines

SpecNVIDIA DGX SparkMac Studio M5 Max 128GBGeForce RTX 5090
Memory available to the model128GB unified LPDDR5X128GB unified LPDDR5X32GB GDDR7
Memory bandwidth~273 GB/s~546 GB/s~1,790 GB/s
Largest model at Q4, comfortably~120B class~120B class~30B class
70B Q4 fully residentYesYesNo, requires offload
Software stackCUDA, TensorRT-LLM, vLLMMLX, Metal, llama.cppCUDA, TensorRT-LLM, vLLM, llama.cpp
Approximate street price~$3,000~$3,499~$2,500 plus host system

Memory bandwidth is the number that predicts decode speed, and the RTX 5090 has roughly 6.5x the DGX Spark and 3.3x the M5 Max. Memory capacity is the number that predicts whether a model runs at all, and there the ordering reverses.

Decode Throughput, Q4 Quantisation, Single Stream

ModelDGX SparkM5 Max 128GBRTX 5090 32GB
8B class110-125 tps95-110 tps142-178 tps
30B class50-62 tps40-50 tps55-75 tps
70B Q435-45 tps25-32 tpsOffload required, single-digit to ~15 tps
120B class20-28 tps14-19 tpsNot practical

The RTX 5090 wins decisively up to about 30B and falls off a cliff at 70B, where the model no longer fits in 32GB and layers spill to system RAM across PCIe. Once that happens the 5090 is slower than a Mac Studio that holds the whole model in unified memory, and the gap is not close.

Prefill, Where the Gap Is Widest

Prefill, also called prompt processing, is compute-bound rather than bandwidth-bound, and it is what determines how long you wait before the first token appears on a long prompt. The RTX 5090 processes an 8B-class model at roughly 7,200 tokens per second of prompt against the DGX Spark's ~1,200 and the M5 Max's ~400.

For agentic workloads that re-send a large context on every turn, this dominates the felt experience far more than decode speed does. A 40,000-token context costs about 6 seconds of prefill on the 5090, 33 seconds on the Spark, and roughly 100 seconds on the M5 Max. See time-to-first-token benchmarks for the full breakdown.

How To Choose

RTX 5090 if your working set is 30B and under, you want the fastest possible iteration, and you already have or will build a PC around it. This covers most coding-agent work in 2026, because the strong sparse coding models activate few enough parameters to run well here.

DGX Spark if you need 70B-plus resident, you fine-tune, or your research code is CUDA-only. The Spark is the only one of the three that does all three well.

Mac Studio M5 Max if the machine doubles as your daily driver, you value silence and idle power, or you develop for Apple platforms. It is the slowest of the three on raw throughput and the only one you would also happily use as a desktop.

Note that total parameters determine whether a model fits and active parameters determine how fast it runs. That distinction decides more of this comparison than any spec above. See active versus total parameters.

Brand Visibility Implications

Every configuration above runs frontier-adjacent open-weight models with no API call anyone can observe. As the entry price for credible local inference falls toward the cost of a gaming GPU, a growing share of the questions people ask about brands is answered from weights on a desk rather than from a monitored endpoint. Parametric recall, what the model absorbed at training time, becomes the only lever, and it cannot be changed after the fact. See the local-LLM visibility blind spot.

Methodology

Vendor specifications come from NVIDIA, Apple, and model-card publications. Throughput figures aggregate community benchmark reporting from the llama.cpp discussions, the MLX repository, and published independent test suites. Single-stream decode unless stated otherwise. Ranges rather than point values are used wherever independent runs disagree, which is most of the time: quantisation format, prompt length, thermal state, and runtime version each move these numbers by more than the differences being measured. Treat every figure as an order-of-magnitude guide, not a specification. Updated quarterly.

How Presenc AI Helps

Presenc AI instruments brand visibility on locally served open-weight models as well as cloud APIs, so teams whose buyers run local-first inference can see mention rates and recommendation drift that endpoint monitoring never reaches.

Frequently Asked Questions

Up to about 30B parameters, yes, and by a wide margin: roughly 142-178 tps against 95-110 tps on 8B-class models, plus far faster prompt processing. At 70B and above the answer reverses, because 32GB of GDDR7 cannot hold the model and the 5090 falls back to slow PCIe offload while the 128GB Mac Studio keeps everything resident.
The RTX 5090 has roughly 1,790 GB/s against approximately 546 GB/s on the M5 Max and approximately 273 GB/s on DGX Spark. Memory bandwidth is the main predictor of decode speed, which is why the 5090 leads on every model size it can actually hold.
DGX Spark, because it combines 128GB of unified memory with the full CUDA training stack and native low-precision support. The RTX 5090 fine-tunes smaller models faster but runs out of memory sooner. Apple Silicon is the weakest of the three for training work.
Agentic loops re-send a large context on every turn, so prompt processing runs repeatedly while decode produces only a short response each time. On a 40,000-token context, prefill costs roughly 6 seconds on an RTX 5090, 33 seconds on DGX Spark, and around 100 seconds on an M5 Max, which dominates the total wait.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Start monitoring today.