XFX Speedster SWFT105 RX 6400
The XFX Speedster SWFT105 RX 6400 is XFX’s version of the Radeon RX 6400. At 155 mm it’s 17 mm shorter than the 172 mm reference design. Below you’ll find every current retailer price and the full spec sheet.
Prices update regularly and may differ at the retailer. This page contains affiliate links.
Prices include 19% VAT. Shipping is shown under each price, and offers are listed by price plus shipping.

Gaming performance
Gaming score: geometric mean of in-game FPS, normalized so the RTX 5090 = 100. Render and synthetic scores are in the Benchmarks section below.
Gaming scores are normalised performance indexes derived from Tom’s Hardware’s published GPU benchmarks. Performance per dollar and per watt are our own calculations on top of that data.
1080p Gaming
1440p Gaming
4K Gaming
The Radeon RX 6400 scores 11.9 at 1080p, 0.0 at 1440p, and 0.0 at 4K on a scale where 100 = the fastest GPU we test (RTX 5090). These scores apply to all board versions of this chip.
Benchmarks
Standardized benchmark scores for this chip, each compared to the fastest card we track and linked to its source. They apply to every board version.
3D & gaming
Compute & AI
Where to buy
Every in-stock retailer we track, cheapest first.
Specifications
GPU chip
- Chip
- Radeon RX 6400
- Architecture
- RDNA 2 Navi 24
- Process node
- TSMC N6
- Shader cores
- 768
- SM count
- 12
- Texture units (TMUs)
- 48
- Render outputs (ROPs)
- 32
- RT cores
- 12 · 1st gen
Memory
- VRAM
- 4GB GDDR6
- Memory bus
- 64-bit
- Bandwidth
- 128 GB/s
- Memory speed
- 16 Gbps
- Infinity Cache
- 16 MB
GPU die
- Foundry
- TSMC
- Die size
- 107 mm²
- Transistors
- 5.4 billion
- Generation
- Radeon RX 6000
Display & connectivity
- Max resolution
- 7680 × 4320
- Max displays
- 4
- HDCP
- 2.3
APIs & software
- DirectX
- 12 Ultimate (12.2)
- Vulkan
- 1.3
- OpenGL
- 4.6
- OpenCL
- 2.2
- Shader model
- 6.7
Platform features
- Upscaling
- FSR 3, FSR 3 Frame Generation
- Also runs
- XeSS 2, XeSS 2 Frame Generation
- Resizable BAR
- Yes
Video encoding & decoding
| Codec | Encode | Decode |
|---|---|---|
| H.264 / AVC | Not supported | Not supported |
| H.265 / HEVC | Not supported | Supported |
| AV1 | Not supported | Not supported |
| VP9 | Not supported | Not supported |
Decoder: VCN (HEVC decode only)
Board
- Boost clock
- 2321 MHz(reference)
- Base clock
- 1923 MHz
- TDP
- 53W
- Recommended PSU
- 350W
Physical
- Cooling
- Air
- Warranty
- 3 years
- Length
- Reference design:
172 mmThis card: 155 mm - Slot width
- 1 slots
- Outputs
- 1 HDMI 2.1, 1 Displayport 1.4a
Identifiers
- MPN
- RX-64XL4SFG2
- ASIN
- B09Y7358KJ
- EAN
- 0840191500169
- UPC
- 840191500169
Chip-level specs — cores, memory, die, APIs, media engine and platform features — are shared by every version of the Radeon RX 6400. The Board, Physical and Identifier boxes are specific to this XFX variant; a struck-through value shows where it differs from the reference design.
Local AI performance
What the XFX Speedster SWFT105 RX 6400 can run on your own machine, and how fast. Capacity and speed come from the Radeon RX 6400 chip, so they are the same on every board version; price per GB is this board’s own.
Largest model it can run (8K context)
- Q4 (the usual choice, near-original quality)
- ~3.1B
- Q8 (closer to original, half the model size)
- ~1.8B
- FP16 (unquantised, for training and research)
- ~0.9B
Parameter counts the card holds with room for an 8K conversation, counting the model weights plus its attention cache. A longer conversation needs more memory, so it lowers these figures.
Speed & value
- Speed, 8B model at Q4
- 19-23 tok/s
- Memory bandwidth
- 128 GB/s
- Price per GB of VRAM
- €60.51
Generation speed tracks memory bandwidth, because every word produced reads the whole model from memory. Shown as a range.
Models that fit, at Q4
| Model | Size | Memory needed | Est. speed |
|---|---|---|---|
| Falcon3 3B | 3.2B | 2.7GB | 47-55 tok/s |
| Llama3.2 3B | 3.2B | 3.0GB | 47-55 tok/s |
| Qwen2.5 3B | 3.1B | 2.2GB | 49-57 tok/s |
Popular models small enough for this card at Q4, one per family. Speed is left blank for mixture-of-experts models, where a bandwidth estimate is not reliable enough to publish.
Choosing a card for AI? Compare every GPU for local AI, including setups that pool the memory of two or more cards.
Affiliate disclosure: Some links on this page are affiliate links — MaxMyBuild may earn a commission at no extra cost to you if you buy through them. As an Amazon Associate, MaxMyBuild earns from qualifying purchases. For Amazon, current price and availability are shown on the Amazon product page; product prices and availability are accurate as of the date and time of purchase and are subject to change. Prices shown for other retailers are approximate and may differ at checkout. “Search” links are non-affiliate and provided for convenience only.
