In this article
Short answer
The AMD Ryzen 5 9600X is the best budget CPU for an AI PC build. It scored 63.36 tokens/s in Phoronix's llama.cpp 8B prompt test against 43.68 for the Intel Core Ultra 5 245K, and AMD rates it at 65W. Step up to the Ryzen 7 9700X for more cores, or choose the 245K when a faster kernel build or the NPU matters.
- In a single-GPU AI build the GPU sets the pace, so a mid-range CPU is usually enough.
- The 9600X and 9700X lead the 245K in llama.cpp and OpenVINO vision tests.
- The 245K compiles faster than the 9600X with 14 cores to six.
- None of the three includes a cooler.
- Best overallAMD Ryzen 5 9600X
- Best for faster builds and an NPUIntel Core Ultra 5 245K
- Best step-up chipAMD Ryzen 7 9700X
Comparison at a glance#
| Model | Cores / threads | Max boost | Socket | Default power | NPU |
|---|---|---|---|---|---|
| AMD Ryzen 5 9600X | 6c / 12t | 5.4 GHz | AM5 | 65W | None |
| Intel Core Ultra 5 245K | 14c / 14t (6P + 8E) | 5.2 GHz | LGA1851 | 125 W base, 159 W turbo | 13 TOPS |
| AMD Ryzen 7 9700X | 8c / 16t | 5.5 GHz | AM5 | 65W | None |
How we picked#
The aim was a CPU that will not hold back a single GPU in an AI PC at a modest budget. The picks come from AMD's and Intel's specification pages and from Phoronix's Linux results for llama.cpp, OpenVINO, TensorFlow and a kernel build. Core count, boost clock, power and platform cost counted. All three are sold boxed in India.
The picks#
AMD Ryzen 5 9600X
AMD- Cores6 cores, 12 threads
- BoostUp to 5.4 GHz
- SocketAM5
The Ryzen 5 9600X is a six-core, twelve-thread chip with a 5.4 GHz top boost clock and a 65W default TDP. It uses AM5, with PCIe 5.0 and 24 of its 28 native lanes usable, DDR5-5600 for two modules and a 256 GB memory ceiling. It also lists AVX512 among its instruction extensions, which matters for CPU inference.
In Phoronix's Q1 article it scored 63.36 tokens/s when llama.cpp read prompts for an 8B model, ahead of the 245K at 43.68. It reached 73.70 FPS in OpenVINO Person Detection, twice the 245K's 36.37. Where it falls behind is parallel work: the kernel build took 1402.57 seconds, against 916.49 for the 245K. For a GPU-led AI machine that is acceptable.
AMD recommends a premium air cooler and ships no cooler in the box. That keeps the platform simple, and a mid-priced tower cooler plus an AM5 board that lists the chip in its support list is enough; check the board maker's support list. Because the socket is shared with the Ryzen 9 chips, a later upgrade needs no new board.
| Spec | Detail |
|---|---|
| Cores and threads | 6 cores, 12 threads |
| Max boost clock | Up to 5.4 GHz |
| Base clock | 3.9 GHz |
| Platform (socket) | AM5 |
| L2 cache | 6 MB |
| L3 cache | 32 MB |
| Default TDP | 65W |
| Memory support | DDR5-5600 (two single-rank or dual-rank modules); DDR5-3600 with four modules |
| PCIe | PCIe 5.0; 28 native lanes total, 24 usable |
| Cooler | Not included; AMD recommends a premium air cooler |
Pros
- Second only to the 9700X in the llama.cpp 8B prompt test, at 63.36 tokens/s
- 65 W default power, with a premium air cooler recommended by AMD
- AVX-512 listed by AMD, as on the bigger Ryzen 9000 chips
- AM5 socket leaves room to move up to an 8 or 16-core chip
Cons
- Only six cores, so builds and parallel jobs run slowly
- Kernel build took 1402.57 seconds, the slowest of the three
- No NPU listed by AMD
Best for a first AI build where most spend goes to the GPU
Not for buyers who compile large projects often
Sources: Maker: amd.com (opens in a new tab), Benchmark: phoronix.com (opens in a new tab)
Intel Core Ultra 5 245K
Intel- Cores14 cores, 14 threads
- Boost5.2 GHz
- SocketFCLGA1851
The Core Ultra 5 245K has fourteen cores, six performance and eight efficient, running fourteen threads. Intel lists a 5.2 GHz top boost clock, 125 W of base power and 159 W of turbo power, on the LGA1851 socket. Its NPU is rated at 13 peak INT8 TOPS, with 30 peak INT8 TOPS for the platform.
The extra cores show in builds. Phoronix measured 916.49 seconds for the kernel build, well ahead of the 9600X and 9700X. In an OpenVINO GenAI test with Phi-3-mini INT4 it produced 23.02 tokens/s, above the 9600X at 20.40 and the 9700X at 21.45. That is the kind of workload where Intel's own toolkit helps.
On general CPU AI work it is the weakest here. llama.cpp prompt speed was 43.68 tokens/s and OpenVINO Person Detection ran at 36.37 FPS. Higher turbo power also means a stronger cooler and board than the AMD chips need. Intel's page does not name a cooler for this chip either.
| Spec | Detail |
|---|---|
| Cores and threads | 14 cores (6 Performance-cores + 8 Efficient-cores), 14 threads |
| Max boost clock | 5.2 GHz (P-core max turbo 5.2 GHz, E-core max turbo 4.6 GHz) |
| Base clock | P-core 4.2 GHz; E-core 3.6 GHz |
| Platform (socket) | FCLGA1851 |
| L2 cache | 26 MB (Total L2) |
| L3 cache | 24 MB Intel® Smart Cache |
| Base power | 125 W |
| Maximum turbo power | 159 W |
| Memory support | Up to DDR5 6400 MT/s |
| PCIe | 5.0 and 4.0; max 24 lanes |
| NPU | Intel AI Boost, 13 TOPS (INT8) |
Pros
- 14 cores in a 6P plus 8E layout, with 14 threads
- Fastest kernel build of the three, at 916.49 seconds
- 13 TOPS NPU and an integrated GPU as listed by Intel
- Ahead of the 9600X in an OpenVINO GenAI Phi-3 INT4 test
Cons
- Lowest llama.cpp prompt speed of the three, at 43.68 tokens/s
- 159 W maximum turbo power, against a 65W default rating on the AMD chips
- Slowest in OpenVINO Person Detection, at 36.37 FPS
Best for an Intel build that needs a mid-range chip with an NPU
Not for buyers who rely on CPU-side llama.cpp work
Sources: Maker: intel.com (opens in a new tab), Benchmark: phoronix.com (opens in a new tab)
AMD Ryzen 7 9700X
AMD- Cores8 cores, 16 threads
- BoostUp to 5.5 GHz
- SocketAM5
The Ryzen 7 9700X offers eight cores and sixteen threads at up to 5.5 GHz, with AMD's 65W default TDP. It uses the same AM5 platform as the 9600X, with 32 MB of L3 cache and 8 MB of L2. AMD recommends a premium air cooler.
It tops the group in nearly every CPU AI test Phoronix ran. The llama.cpp 8B prompt test hit 82.20 tokens/s, TensorFlow ResNet-50 reached 42.05 images/s and OpenVINO Person Detection ran at 91.56 FPS. In Whisper.cpp small-model transcription it took 233.29 seconds, against 276.16 for the 245K and 281.74 for the 9600X, where lower is better.
The kernel build is the exception: 1085.92 seconds, behind the 14-core 245K. If most of your time goes to inference and data work rather than compiling, the 9700X is the stronger buy. It is also the sensible stopping point before the 16-core chips, whose power and cooling needs are much higher.
| Spec | Detail |
|---|---|
| Cores and threads | 8 cores, 16 threads |
| Max boost clock | Up to 5.5 GHz |
| Base clock | 3.8 GHz |
| Platform (socket) | AM5 |
| L2 cache | 8 MB |
| L3 cache | 32 MB |
| Default TDP | 65W |
| Memory support | DDR5-5600 (two single-rank or dual-rank modules); DDR5-3600 with four modules |
| PCIe | PCIe 5.0; 28 native lanes total, 24 usable |
| Cooler | Not included; AMD recommends a premium air cooler |
Pros
- Highest scores of the three in llama.cpp, ONNX, TensorFlow and OpenVINO vision
- 8 cores and 16 threads give headroom for data loaders
- 65 W default power like the 9600X
- Quickest Whisper.cpp small-model time of the three
Cons
- Costs more than the other two, so less left for the GPU
- Kernel build trails the 14-core 245K
- No NPU listed by AMD
Best for a build that wants the best CPU-side AI speed in a modest tier
Not for buyers who want to spend the least on the CPU
Sources: Maker: amd.com (opens in a new tab), Benchmark: phoronix.com (opens in a new tab)
Performance and benchmarks#
These charts are Phoronix results from Linux. They show how much a budget chip gives up against the 9700X, and where Intel's 245K stands. Treat the figures as CPU-only; with a GPU running the model, the gaps shrink.
llama.cpp CPU LLM inference: Prompt processing speed (tokens/s, higher is better)
5 products. Highest: AMD Ryzen 9 9950X, 102.49 tokens/s. Lowest: Intel Core Ultra 5 245K, 43.68 tokens/s.
| Product | tokens/s |
|---|---|
| AMD Ryzen 9 9950X | 102.49 |
| AMD Ryzen 7 9700X | 82.2 |
| AMD Ryzen 5 9600X | 63.36 |
| Intel Core Ultra 9 285K | 47.21 |
| Intel Core Ultra 5 245K | 43.68 |
About this data: figures are copied from the named source, not measured by us, and the test system and settings are those the source states above. See all benchmark sources and credits.
In llama.cpp, prompt processing runs at 82.20 tokens/s on the 9700X, 63.36 on the 9600X and 43.68 on the 245K. The AMD chips benefit from AVX-512, which Phoronix credits for the Ryzen edge in CPU AI work.
OpenVINO CPU inference: Throughput (FPS, higher is better)
5 products. Highest: AMD Ryzen 9 9950X, 110.37 FPS. Lowest: Intel Core Ultra 5 245K, 36.37 FPS.
| Product | FPS |
|---|---|
| AMD Ryzen 9 9950X | 110.37 |
| AMD Ryzen 7 9700X | 91.56 |
| AMD Ryzen 5 9600X | 73.7 |
| Intel Core Ultra 9 285K | 49.97 |
| Intel Core Ultra 5 245K | 36.37 |
About this data: figures are copied from the named source, not measured by us, and the test system and settings are those the source states above. See all benchmark sources and credits.
In OpenVINO Person Detection the 9700X scored 91.56 FPS, the 9600X 73.70 and the 245K 36.37. This is an Intel framework, yet the AMD chips lead on this test.
OpenVINO GenAI CPU LLM inference: Generation speed (tokens/s, higher is better)
5 products. Highest: Intel Core Ultra 9 285K, 26.53 tokens/s. Lowest: AMD Ryzen 5 9600X, 20.4 tokens/s.
| Product | tokens/s |
|---|---|
| Intel Core Ultra 9 285K | 26.53 |
| AMD Ryzen 9 9950X | 25.16 |
| Intel Core Ultra 5 245K | 23.02 |
| AMD Ryzen 7 9700X | 21.45 |
| AMD Ryzen 5 9600X | 20.4 |
About this data: figures are copied from the named source, not measured by us, and the test system and settings are those the source states above. See all benchmark sources and credits.
OpenVINO GenAI with an INT4 language model reverses the order: the 245K produced 23.02 tokens/s, then the 9700X at 21.45 and the 9600X at 20.40. The spread is small, so any of the three runs this model at a similar pace.
TensorFlow CPU inference: Throughput (images/sec, higher is better)
5 products. Highest: AMD Ryzen 9 9950X, 52.55 images/sec. Lowest: Intel Core Ultra 5 245K, 30.48 images/sec.
| Product | images/sec |
|---|---|
| AMD Ryzen 9 9950X | 52.55 |
| AMD Ryzen 7 9700X | 42.05 |
| Intel Core Ultra 9 285K | 38.12 |
| AMD Ryzen 5 9600X | 35.2 |
| Intel Core Ultra 5 245K | 30.48 |
About this data: figures are copied from the named source, not measured by us, and the test system and settings are those the source states above. See all benchmark sources and credits.
TensorFlow ResNet-50 puts the 9700X first at 42.05 images/s, then the 9600X at 35.20 and the 245K at 30.48.
General compute: Linux kernel compile: Build time (seconds, lower is better)
5 products. Lowest: AMD Ryzen 9 9950X, 549.97 seconds. Highest: AMD Ryzen 5 9600X, 1402.57 seconds.
| Product | seconds |
|---|---|
| AMD Ryzen 9 9950X | 549.97 |
| Intel Core Ultra 9 285K | 577.27 |
| Intel Core Ultra 5 245K | 916.49 |
| AMD Ryzen 7 9700X | 1085.92 |
| AMD Ryzen 5 9600X | 1402.57 |
About this data: figures are copied from the named source, not measured by us, and the test system and settings are those the source states above. See all benchmark sources and credits.
For the kernel build, lower is better. The 245K's 14 cores give it 916.49 seconds, the 9700X 1085.92 and the 9600X 1402.57. If you compile a lot, core count matters more than the AI scores suggest.
Whisper.cpp speech-to-text gives a similar order. The 9700X took 233.29 seconds on the small English model, the 245K 276.16 and the 9600X 281.74, with lower times better. Transcription is a common side job in an AI PC, and a 9700X finishes it noticeably sooner. Across all the charts, the 245K is strongest where the work spreads over many cores and weakest where wide vector units carry the load.
Best Budget CPU for an AI PC Build (last checked November 2026).
Value for money#
The 9600X and the 245K sit in the same entry-to-mid tier, with the 9700X one step above in the mid-range tier. Prices change, so check current prices before buying. Between the first two, the 9600X gives more CPU-side AI speed, and the 245K gives more cores for compiling and an NPU.
Platform cost shapes value too. Both AMD chips are rated 65W, and AMD recommends a premium air cooler rather than liquid cooling. The 245K lists 159 W of maximum turbo power, so budget a better cooler and board. The cooler is not in any box.
For long-term value, the 9600X and 9700X share the AM5 socket with the 16-core Ryzen 9 chips, so an upgrade needs only a new CPU. The 245K sits on LGA1851, which is also current. In a single-GPU build, spend the larger share of the budget on the GPU, as covered in our budget GPU picks for machine learning.
A budget CPU carries a few jobs in an AI PC. It prepares data, runs the Python process that drives the GPU, serves small models on its own and handles everyday tasks. Phoronix's numbers show the cost of choosing a smaller chip is real but limited, and the largest losses come in compiling and CPU-only LLM prompts. Those losses disappear when the GPU runs the model.
How to choose#
- Set the GPU first. A CPU of this class is usually enough to feed a mid-range GPU.
- Decide whether you compile. If yes, the 245K's 14 cores or the 9700X's 16 threads help.
- Pick the software path. llama.cpp and OpenVINO vision favour the AMD chips; OpenVINO GenAI with small INT4 models favours the 245K.
- Check the cooler. AMD recommends a premium air cooler for the 9600X and 9700X. Intel lists a 159 W maximum turbo power for the 245K.
- Match the board. AM5 and LGA1851 boards are not interchangeable, so choose the CPU before the board.
- Plan memory. Two DDR5 modules suit these chips, and our best RAM for AI and machine learning guide covers capacity.
For a higher-end option see AI and ML CPU guide, and for a mid-range matchup read Core Ultra 7 270K Plus vs Ryzen 7 9700X for AI. All CPU guides sit on the AI processors page.
Common mistakes#
- Overspending on the CPU. For one GPU, a six or eight-core chip is enough.
- Ignoring compile time. Six cores finish builds slowly.
- Choosing by NPU alone. The 13 TOPS NPU on the 245K does not speed up most desktop AI tools.
- Forgetting the cooler. None of these chips ships with one.
- Buying an old-socket board. AM5 and LGA1851 are the live platforms.
- Pairing a small CPU with a very large GPU and expecting no limit. A six-core chip can hold back data loading when the GPU is fast, so match the two.
Verdict#
Best overall AMD Ryzen 5 9600XSee the details
The Ryzen 5 9600X is the best budget CPU for an AI PC build, with solid CPU-side AI speed at a low power rating and an upgrade path. Choose the Ryzen 7 9700X if your budget allows, since it leads almost every CPU AI test here. Choose the Core Ultra 5 245K if you compile often, use OpenVINO GenAI or want the NPU.
In India, the cooler and motherboard can cost as much as the processor itself, so price the whole platform at Indian retailers before deciding which budget chip to buy.
Frequently asked questions#
Is a six-core CPU enough for an AI PC?
Yes for a build with one GPU, because the GPU does the heavy model work. It struggles with long compiles and many parallel jobs, so pick more cores if you often build software or run several tasks at once.
Which budget CPU is fastest for llama.cpp?
Among these three the Ryzen 7 9700X, then the Ryzen 5 9600X. Phoronix measured both well ahead of the Core Ultra 5 245K. This applies to CPU-only runs, and matters less when a GPU runs the model.
Do I need an NPU?
No. The 245K's NPU is rated at 13 TOPS by Intel, and AMD shows no NPU on these desktop chips. Most local AI tools use the CPU or GPU.
Do these CPUs include a cooler?
AMD lists both Ryzen chips as boxed without one, and Intel names no cooler for the 245K. Budget for a separate cooler in every case, and check that it supports the socket on your chosen motherboard before you order.
Which socket is better for upgrades?
Both AM5 and LGA1851 are current, so neither is a dead end today. Check the board maker's support list for the chips you may buy later, because the maker confirms which processors a given board accepts.
Should I buy the 9700X instead of the 9600X?
If you run more than one job at once or want faster CPU inference, yes. For a GPU-led build the 9600X is enough, and the money saved can go towards a better graphics card or more memory.
Sources & credits#
We have not tested these products ourselves. Specs come from the makers; benchmark figures come from the sources we name.
- AMD, AMD Ryzen 5 9600X product page (opens in a new tab)
- Intel, Intel Core Ultra 5 245K specifications (opens in a new tab)
- AMD, AMD Ryzen 7 9700X product page (opens in a new tab)
- Phoronix, AMD Ryzen 9000 vs. Intel Core Ultra Arrow Lake On Linux (opens in a new tab)
Facts and figures in this article come from these sources. Articles are compiled from the makers' published specifications and the benchmarks named above, and checked by the NextAIGrid Editorial Team.
Disclaimer
This guide is for information only. Specifications and availability change, so check the maker's page. Found an error? Tell us.
How was this buying guide put together?
It draws on 4 named sources: maker specification pages and published third-party benchmarks, listed in the article. We have not tested these products ourselves. Specs come from the makers; benchmark figures come from the sources we name. Specifications and sources were last checked on 17 November 2026. Warranty terms and availability differ by brand and seller in India, so check the maker's India page for the warranty period and buy from an authorised seller. Read how we research and our editorial policy.
Spotted an error? Tell us. Confirmed fixes are listed on the corrections page.