Calculate the TPS of the CMP 50HX on local AI models
Every model in our catalogue assessed against this card at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from this card's memory bandwidth and the size of each model once compressed.
Calculated for this card
Largest model it holds
Ling-lite-1.5 ("Bailing")
16.8B · Q3_K_M · 38.1 tok/s
Fastest model
Gemma 3 QAT 1B
237 tok/s · 1B
What AI models can a CMP 50HX run?
Set the inputs, read the answer
More context means more memory for the conversation cache. Speed is for a fresh conversation and does not change with this setting.
Hides models that would only fit by being compressed below this point.
381 models match
Calculating| Quantisation | Fit | ||||||
|---|---|---|---|---|---|---|---|
|
237
tok/s
202–285 |
Gemma 3 1B | 1B | Mar 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
237
tok/s
202–285 |
Gemma 3 QAT 1B | 1B | Apr 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
237
tok/s
142–379 · low confidence |
HGRN 1B (WT 103) ≈ | 1B | Nov 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
237
tok/s
142–379 · low confidence |
LLama 3..2 Typhoon 2 1B ≈ | 1B | Dec 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
237
tok/s
142–379 · low confidence |
OLMo-1B ≈ | 1B | Feb 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
237
tok/s
142–379 · low confidence |
Pythia-1b ≈ | 1B | Apr 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
222
tok/s
133–356 · low confidence |
DeepSeekMoE-16B | 16B | Jan 2024 | 8.1 GB | 4k tokens | Q3_K_M | Tight |
|
220
tok/s
132–351 · low confidence |
OpenELM-1.1B ≈ | 1.1B | May 2024 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
216
tok/s
129–345 · low confidence |
DeciCoder-1B ≈ | 1.1B | Aug 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
216
tok/s
129–345 · low confidence |
SantaCoder ≈ | 1.1B | Jan 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
216
tok/s
129–345 · low confidence |
TinyLlama-1.1B (1T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
216
tok/s
129–345 · low confidence |
TinyLlama-1.1B (3T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
198
tok/s
119–316 · low confidence |
EXAONE 4.0 (1.2B) ≈ | 1.2B | Jul 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
198
tok/s
119–316 · low confidence |
MinerU2.5 ≈ | 1.2B | Sep 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
198
tok/s
119–316 · low confidence |
Pleias 1.0 1.2B ≈ | 1.2B | Dec 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
198
tok/s
119–316 · low confidence |
Pleias-RAG-1B ≈ | 1.2B | Apr 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
193
tok/s
164–231 |
Llama 3.2 1B | 1.2B | Sep 2024 | 2.2 GB | 131k tokens | Q8_0 | Comfortable |
|
190
tok/s
114–304 · low confidence |
MiniCPM-1.2B ≈ | 1.2B | Jun 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
DeepSeek Coder 1.3B ≈ | 1.3B | Jan 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
DeepSeek-VL-1.3B ≈ | 1.3B | Mar 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
DigiRL ≈ | 1.3B | Jun 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
GLA Transformer 1.3B ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
Janus 1.3B ≈ | 1.3B | Oct 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
Kosmos-2.5 ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
182
tok/s
109–292 · low confidence |
Otter ≈ | 1.3B | May 2023 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
Speeds are estimates for a single request — one conversation at a time — calculated from memory bandwidth, model size and quantisation. Real throughput varies with the inference runtime and its version. Figures published by hardware vendors measure many simultaneous requests and are much higher.
On record
CMP 50HX full specification
Everything on record for this board, ordered by how much it bears on running a language model rather than by how a spec sheet would list it. Memory comes first because it decides the outcome; the rest is context.
Memory
The two specifications that decide what this card can run and how quickly. Capacity sets which models fit; bandwidth sets how many tokens per second they produce once they do.
- Memory size
- 10 GB
- Memory bandwidth
- 560 GB/s
- Memory type
- GDDR6
- Memory bus width
- 320 bit
- Memory clock
- 1.75 GHz
The chip
Which processor is on the board and how it was manufactured. A smaller process size generally means more performance for the same power.
- Graphics processor
- TU102
- Architecture
- Turing
- Generation
- Mining GPUs
- Foundry
- TSMC
- Process size
- 12 nm
- Transistors
- 18.6 billion
- Transistor density
- 24,700 K/mm²
- Die size
- 754 mm²
- Released
- 24 June 2021
Clock speeds
How fast the processor runs. Worth far less here than on a gaming benchmark: generating text is limited by memory bandwidth, so a higher clock barely moves the result.
- Base clock
- 1.35 GHz
- Boost clock
- 1.55 GHz
Processing units
What the chip contains. These drive graphics performance and matter mainly for processing a long prompt rather than for producing the answer.
- Shading units
- 3,584
- Texture mapping units
- 192
- Render output units
- 80
- Streaming multiprocessors
- 56
- Tensor cores
- 448
- Ray tracing cores
- 56
- L1 cache
- 64 KB
- L2 cache
- 5 MB
Theoretical performance
Peak arithmetic rates published for the board. These are ceilings that no real workload reaches, and generating text reaches a small fraction of them because it is limited by memory rather than arithmetic.
- Half precision (FP16)
- 22.2 TFLOPS
- Single precision (FP32)
- 11.1 TFLOPS
- Double precision (FP64)
- 346.1 GFLOPS
- Pixel rate
- 124 GPixel/s
- Texture rate
- 297 GTexel/s
The board
What it takes to physically install and power the card — the practical constraints that decide whether it fits the machine you already own.
- Power draw (TDP)
- 250 W
- Suggested power supply
- 600 W
- Power connectors
- 2x 8-pin
- Bus interface
- PCIe 1.0 x4
- Slot width
- Dual-slot
- Dimensions
- 267 mm × 35 mm
Software support
Which graphics and compute interfaces the card supports. CUDA compute capability is the one that bears on inference: below 7.0 there are no tensor cores, and modern inference software falls back to slower code paths.
- CUDA compute capability
- 7.5
- DirectX
- 12.2
- OpenGL
- 4.6
- Vulkan
- 1.4
- OpenCL
- 3.0
- Shader model
- 6.8
Listings
Where to buy a CMP 50HX
No vendor is currently listing this card. Listings come from vendors who publish them here directly — browse the vendor directory to see who is selling what.
What the numbers mean
What the memory subsystem means for AI
Memory
10 GB
Bandwidth
560 GB/s
Largest model
Ling-lite-1.5 ("Bailing")
At 10 GB of GDDR6 the CMP 50HX is limited to the smaller end of the catalogue. About 9 GB is actually available to a runtime, and a model has to fit entirely inside it before generating anything at all.
The memory bus moves 560 GB/s across a 320-bit bus. That is the number that governs generation speed — arithmetic per byte read is small enough that the bus, not the cores, is what everything waits on.
The figure is the memory clock — 1.75 GHz here — multiplied by the bus width. It is why core counts predict generation speed so poorly.
The biggest thing it holds is Ling-lite-1.5 ("Bailing") (16.8B) at Q3_K_M compression, for about 38.1 tokens per second.
The chip and how it was built
The CMP 50HX is built on the TU102 graphics processor, using NVIDIA's Turing architecture, as part of the Mining GPUs generation.
The chip is manufactured by TSMC, on a 12 nm process, with a die measuring 754 mm², holding 18.6 billion transistors. A smaller process generally means more performance for the same power, though for language models it matters far less than the memory subsystem.
It was released in June 2021, roughly 5 years ago. Inference software support tends to follow hardware by a year or two, so a card of this age generally has mature, well-optimised code paths available to it.
Compute throughput, and why it matters less than it looks
FP16
22.2 TFLOPS
FP64
346.1 GFLOPS
Tensor cores
448
On paper the CMP 50HX reaches 22.2 TFLOPS at half precision and 11.1 TFLOPS at single precision. These are peak figures no real workload sustains, and generating text reaches only a small fraction of them — decoding is limited by memory rather than arithmetic, which is why a card can look enormously powerful here and still produce tokens at an ordinary rate.
Double-precision throughput is 346.1 GFLOPS. It has no bearing on running a language model — no inference runtime uses it — but it separates datacentre parts from consumer ones, since the latter deliberately restrict it.
The card carries 448 tensor cores across 56 streaming multiprocessors. These accelerate the matrix arithmetic at the heart of a transformer, and they are what make prompt processing — reading a long document before answering — dramatically faster than it would otherwise be.
Clocks run from 1.35 GHz at base to 1.55 GHz boosted. Worth far less here than on a gaming benchmark: raising the clock speeds up the arithmetic, and the arithmetic is not what generation is waiting on.
Cache and processing units
The CMP 50HX has 64 KB of L1 cache, backed by 5 MB of L2. Cache absorbs a share of the memory traffic that would otherwise hit the main bus, which is the one place on this page where a number other than bandwidth quietly affects generation speed — a large L2 lets more of the working set stay close to the cores.
There are 3,584 shading units, 192 texture mapping units, and 80 render output units. These drive graphics workloads and contribute to prompt processing, but they sit idle for much of the time a model spends generating a reply.
Power, size and installation
Power draw
250 W
The CMP 50HX is rated at 250 W, with a 600 W power supply suggested for the whole system. Running a language model keeps a card busy in bursts rather than continuously — it draws hard while generating and idles between requests — so sustained draw over a working day is usually well below the rated figure.
The board occupies a dual-slot, measuring 267 mm long, and needs 2x 8-pin. Worth checking against the case and power supply already in the machine, since the largest cards need considerably more of both than a typical desktop provides.
It connects over PCIe 1.0 x4. The interface governs how quickly a model is loaded from disk into the card, not how fast it runs once there, so a narrower link costs a few seconds at startup and nothing thereafter.
The extremes
The largest AI models a CMP 50HX can run
The biggest open-weight models that fit on this card, newest first. Each is shown at the best compression the card can hold.
The fastest AI models on a CMP 50HX
Where this card produces tokens quickest. Smaller models dominate here, because generating each token means reading the whole model out of memory once.
Step by step
How to work out the tokens per second of a CMP 50HX
You do not have to calculate anything by hand — the gputps.com calculator on this page has already worked it out for every model this card can hold. Reading off the answer takes six steps.
-
01
Search for the model you want
Every one of the 381 models this CMP 50HX runs is in the table above. Search narrows it by name or by size.
-
02
Decide how long your conversations run
Set the context to your real working length. Short questions cost almost nothing; a long document can consume a large share of the card's 10 GB.
-
03
Pin the comparison to one quality level
Compression is what lets bigger models fit. The quality control drops any model that needs more of it than you are willing to give.
-
04
Look at the range, not just the number
The figures are calculated, not measured. 237 tok/s on Gemma 3 QAT 1B is the fastest result on this card, and like every row it carries a range that reflects how much the runtime matters.
-
05
Check the memory column before committing
Compare what each model needs with the 10 GB this card provides. Tight means it works today; comfortable means it still works when the conversation grows.
-
06
Open the model to compare cards
Following a model through to its own page lists all the hardware that can run it, so you can see where the CMP 50HX sits against the alternatives.
Answers
CMP 50HX — common questions
What is the largest AI model a CMP 50HX can run?
The largest model in our catalogue that fits on a CMP 50HX is Ling-lite-1.5 ("Bailing") at 16.8B parameters, compressed to Q3_K_M. It generates roughly 38.1 tokens per second and needs about 8.9 GB of the card's memory.
How many tokens per second does a CMP 50HX produce?
It depends on the model. On a CMP 50HX the fastest model we track is Gemma 3 QAT 1B at about 237 tokens per second, while larger models run proportionally slower because each token requires reading the whole model out of memory once. Speeds are estimates for a single conversation at a time.
Can a CMP 50HX run a 7B model?
Yes. For example a CMP 50HX runs DeepSeek Coder 6.7B at Q4_K_M, using about 8.3 GB of memory and generating around 81.7 tokens per second.
Can a CMP 50HX run a 13B model?
Yes. For example a CMP 50HX runs DeepSeekMoE-16B at Q3_K_M, using about 8.1 GB of memory and generating around 222 tokens per second.
How much memory does a CMP 50HX have?
A CMP 50HX has 10 GB of GDDR6 memory. Around a tenth of that is reserved by the inference runtime and the driver, leaving roughly 9 GB available for a model and its conversation.
What is the memory bandwidth of a CMP 50HX?
The CMP 50HX has 560 GB/s of memory bandwidth, across a 320-bit memory bus. This is the single best predictor of how fast it generates text, because producing each token means reading the entire model out of memory once.
What type of memory does a CMP 50HX use?
It uses GDDR6 clocked at 1.75 GHz. HBM types are found on datacentre accelerators and carry far more bandwidth than the GDDR used on desktop cards, which is why they generate tokens considerably faster at the same capacity.
Who makes the CMP 50HX?
The CMP 50HX is a NVIDIA product, with the chip manufactured by TSMC, on a 12 nm process.
When was the CMP 50HX released?
The CMP 50HX was released in June 2021.
How much power does a CMP 50HX use?
The CMP 50HX has a rated board power of 250 W, and a 600 W system power supply is suggested. Generating text draws hard in bursts and idles between requests, so average consumption over a working session is normally well below the rated figure.
How much cache does a CMP 50HX have?
The CMP 50HX has 64 KB of L1 cache, and 5 MB of L2 cache. Cache absorbs part of the memory traffic that would otherwise reach the main bus, so a larger L2 gives a modest lift to generation speed beyond what bandwidth alone predicts.
What are the TFLOPS of a CMP 50HX?
The CMP 50HX is rated at 22.2 TFLOPS at half precision and 11.1 TFLOPS at single precision. These are peak arithmetic ceilings rather than achievable rates, and text generation reaches only a small fraction of them because it is limited by memory bandwidth instead.
How many tensor cores does a CMP 50HX have?
The CMP 50HX has 448 tensor cores across 56 streaming multiprocessors. They accelerate the matrix arithmetic a transformer is built from, which mainly speeds up processing a long prompt rather than producing the reply.
Does the CMP 50HX support CUDA?
Yes. The CMP 50HX reports CUDA compute capability 7.5. Capability 7.0 and above has tensor cores, which modern inference software uses; below that it falls back to slower code paths for quantised models.
What bus interface does the CMP 50HX use?
It uses PCIe 1.0 x4. This governs how fast a model is loaded onto the card rather than how fast it runs once loaded, so it costs a few seconds at startup and nothing during generation.
Is the CMP 50HX good for running local AI models?
Its memory limits it to smaller models and its bandwidth gives usable, if unspectacular, generation speeds. In total it runs 381 of the models we track. Whether that is enough depends entirely on which model you want — the table above answers that directly.
Can a CMP 50HX run a model that does not fit in its memory?
It can be split, with the overflow held in system memory — but that part drags the whole thing down, and none of the 10 GB figures on this page assume it.
Would two CMP 50HX cards be twice as fast?
No. A second CMP 50HX doubles the memory to 20 GB, which lets you hold models neither could hold alone, but generation does not split that way. These figures describe one card.
What AI models can a CMP 50HX run?
381 of the 679 open-weight language models we track fit on a CMP 50HX and can be run locally on it. The table on this page lists every one, with the memory it needs, the quantisation it runs at and an estimated generation speed.
The other direction
Looking at it from the other side?
This page starts from the hardware. If you already know which model you want and need to know what it takes to run it, start from the model instead.