FitMyLLM
AMD· GCN 1.0

AMD Radeon HD 8970 OEM

Running LLMs on the Radeon HD 8970 OEM — the long read: which models fit at which quantisation, and the settings worth changing. · Or what a budget buys

VRAM
3 GB
BUDGET
BANDWIDTH
264
GB/S
MODELS Q4
81/449
18%
7B Q4 SPEED
~34
FAST
▸ MODEL COVERAGE @ Q418% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

Average speeds at Q4 quantization. Actual performance varies by model architecture and context length.

3B
~78
TOK/S
7B
3.9GB NEEDED
14B
7.9GB NEEDED
32B
18.0GB NEEDED
70B
39.4GB NEEDED
▸ MEASURED RIG REPORTS

We rent the machine and time every model on it: decode, VRAM peak, concurrency, watts and cost per million tokens. Yours may already be one of them — and two are free to read in full.

SEE THE REPORTS
▸ SPECIFICATIONS
VRAM
3 GB
BANDWIDTH
264 GB/s
FP16 COMPUTE
3.8 TFLOPS
TDP
250W
MEMORY
GDDR5
ARCHITECTURE
GCN 1.0
STREAM PROCESSORS
2,048
COMPUTE UNITS
32
PCIE
Gen 3 x16
81
FAST MODELS · >30 TOK/S
Real-time chat speed
81
USABLE · >10 TOK/S
Comfortable for all tasks
81
TOTAL COMPATIBLE
Fit in VRAM at Q4
▸ DON’T WANT TO BUY?

Test Radeon HD 8970 OEM (or anything bigger) without committing. Pay by the second, cancel anytime.

Spin up in ~60s. Pay by the second. Cancel anytime.

Some links are affiliate links — we may earn a small commission at no extra cost to you. This helps keep FitMyLLM free and independent.

▸ COMPATIBLE MODELS· 81
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
2607
TOK/S · 18% VRAM
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
1738
TOK/S · 19% VRAM
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
1676
TOK/S · 19% VRAM
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
1676
TOK/S · 19% VRAM
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
917
TOK/S · 21% VRAM
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
869
TOK/S · 22% VRAM
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
769
TOK/S · 22% VRAM
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
700
TOK/S · 23% VRAM
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
700
TOK/S · 23% VRAM
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
700
TOK/S · 23% VRAM
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
670
TOK/S · 23% VRAM
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
652
TOK/S · 24% VRAM
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
618
TOK/S · 24% VRAM
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
469
TOK/S · 26% VRAM
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
469
TOK/S · 26% VRAM
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
451
TOK/S · 27% VRAM
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
413
TOK/S · 28% VRAM
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
413
TOK/S · 28% VRAM
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
391
TOK/S · 29% VRAM
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
391
TOK/S · 29% VRAM
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
391
TOK/S · 29% VRAM
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
378
TOK/S · 29% VRAM
S
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
317
TOK/S · 31% VRAM
S
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
313
TOK/S · 32% VRAM
S
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
290
TOK/S · 33% VRAM
S
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
270
TOK/S · 34% VRAM
S
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
250
TOK/S · 35% VRAM
S
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
217
TOK/S · 38% VRAM
S
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
213
TOK/S · 39% VRAM
S
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
196
TOK/S · 41% VRAM
S
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
196
TOK/S · 41% VRAM
S
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
196
TOK/S · 41% VRAM
S
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
196
TOK/S · 41% VRAM
S
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
181
TOK/S · 43% VRAM
S
OPT 1.3B1.3B
OPT·2K CTX· CHAT
181
TOK/S · 43% VRAM
S
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
181
TOK/S · 43% VRAM
S
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
174
TOK/S · 44% VRAM
S
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
165
TOK/S · 45% VRAM
S
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
165
TOK/S · 45% VRAM
S
Qwen2.5-Coder-1.5B1.5B
QWEN·32K CTX· CHAT· TOOL_USE· CODING
156
TOK/S · 47% VRAM
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Radeon HD 8970 OEM.

WHAT THIS CARD IS WORTH

Radeon HD 8970 OEM holds 81 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is Falcon3-3B at Q4_K_M.

TOKENS/SEC PER $100
8B at Q4_K_M, so cards compare like for like
VRAM PER $100
what memory costs on this card
THE BIGGEST IT TAKES
Falcon3-3B3.23B · Q4_K_M2.6 GB55 tok/sEST
granite-4.0-h-micro 3.2B3.2B · Q4_K_M2.6 GB55 tok/sEST
Llama-3.2-3B3.2B · Q4_K_M2.7 GB55 tok/sEST
Falcon-H1 3B3.15B · Q5_K_M2.7 GB50 tok/sEST
Qwen 2.5 3B3.1B · Q5_K_M2.6 GB50 tok/sEST
SmolLM3-3B3.1B · Q4_K_M2.5 GB56 tok/sEST
Ministral 3B3B · Q4_K_M2.6 GB58 tok/sEST
StarCoder2 3B3B · Q5_K_M2.5 GB52 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
M2 (8GB)5 GB$5992.8 tok/s per $100
M3 (8GB)5 GB$5992.8 tok/s per $100
M1 (8GB)5 GB$4992.2 tok/s per $100
RTX 2060 6GB6 GB$15026.7 tok/s per $100
Radeon HD 8970 OEM3 GB

Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.

▸ DEVICE UNDER TEST

AMD Radeon HD 8970 OEM — 3 GB VRAM.

RADEON HD 8970 OEM SPEC
BRAND
AMD
VRAM
3 GB GDDR5
BANDWIDTH
264 GB/s
FP16 COMPUTE
3.8 TFLOPS
FP32 COMPUTE
3.8 TFLOPS
STREAM PROCESSORS
2,048
TDP
250 W
ARCHITECTURE
GCN 1.0
▸ AI CAPABILITY
81/ 449 models @ Q4

With 3 GB VRAM and 264 GB/s bandwidth, this GPU handles models up to 3.1B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~30 tok/s.

§ 01TOP MODELS FOR RADEON HD 8970 OEM
81 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
Qwen 2.5 3B3.1B2.4 GB7637.2
SmolLM3-3B3.1B2.4 GB7630.5
Ministral 3B3B2.3 GB7829.6
StarCoder2 3B3B2.3 GB789.5
Granite 4.1 3B3B2.3 GB7816.6
xLAM-2 3B Function-Calling3B2.3 GB78
Jamba 2 3B3B2.3 GB78
Dolly v2 3B2.8B2.2 GB845.6
StableLM Zephyr 3B2.79B2.2 GB8414.9
Zephyr 3B2.79B2.2 GB8414.4
OPT 2.7B2.7B2.1 GB8728.0
Phi-2 2.7B2.7B2.1 GB8724.1
Zamba2 2.7B2.7B2.1 GB8748.0
Granite 3.0 2B2.63B2.1 GB8935.8
gemma-2-2b2.6B2.1 GB9022.9
LFM2 2.6B2.6B2.1 GB9016.3
Granite 3.1 2B2.53B2.0 GB9337.8
Granite 3.3 2B2.53B2.0 GB9320.5
Gemma 1 2B2.51B2.0 GB9320.2
CodeGemma 2B2.51B2.0 GB9322.9