NVIDIA GeForce RTX 4070 Super

The refresh that mostly helps first tokens: 7168 cores on the same 504 GB/s bus — prompt processing up ~20%, token speed and VRAM unchanged.

12 GB GDDR6X 504 GB/s 192-bit bus 220 W TDP

Technical specifications

GPU die
AD103
CUDA cores
7168
Boost clock
~2.48 GHz
VRAM
12 GB GDDR6X
Memory bus
192-bit
Memory bandwidth
504 GB/s
PCIe
4.0 x16
NVLink
none
TDP
220 W
Release
Jan 2024

What these specs mean for llama.cpp

What the refresh actually changed

The Super is the AD103 with 22% more CUDA cores (7168 vs 5888) — and the same 192-bit GDDR6X bus. Since the two inference phases have different bottlenecks, the refresh splits cleanly:

For chat — which is mostly token streaming — the Super feels like the 4070. For prompt-heavy work (RAG, agentic loops, long-document QA) it earns the delta.

The VRAM line, again

12 GB is 12 GB: 13B Q4 at 32K+ is the comfort zone, 27B is 7 GB over. Nothing about the refresh moves that line, which is the part LLM users care about most.

Buy rule Same price as the 4070: take the Super, free prompt processing. Any premium over the 4070: pass — you'd be paying for first-token latency, the least-used phase in a chat box.

The MTP question: if your model ships a built-in multi-token-prediction head (Qwen 3.5/3.6/3.8, Gemma 4), the one-flag speculative pass (--spec-type draft-mtp --spec-draft-n-max 2 --parallel 1) is worth +40–65% (estimated) on token generation here; no per-card A/B in the record; the estimate matches the 4070's bus class at n-max 2. The price is ~0.6–2 GB of VRAM, the output is bit-identical to the unaccelerated run, and models without MTP heads (Llama class) see nothing. Full tuning rules in the MTP guide.

Expected performance (Q4_K_M, single GPU)

ModelSize on GPUToken genToken gen (MTP est.)Prompt proc.Fits?
8B Q8_0~9.7 GB~55–72 t/s~77–119 t/s~700–900 t/s✅ 32K ctx
13B Q4_K_M~8.5 GB~35–46 t/s~49–76 t/s~750–950 t/s✅ 32K+ ctx
13B Q5_K_M~11.9 GB~25–33 t/s~35–54 t/s~600–800 t/s⚠️ 8K ctx
27B Q4_K_M~19 GB———❌ ~7 GB over

MTP est. = rough prior with --spec-type draft-mtp for models that ship MTP heads (Qwen 3.5/3.6/3.8, Gemma 4); Llama-class models get no MTP speedup. Actual gains depend on model, quant, context length, llama.cpp build and your card — sweep --spec-draft-n-max, don't trust the column blindly. MTP guide →

Figures assume full offload (-ngl 99), ~8K context and batch 2048. Token speed tracks the memory bus almost linearly, so treat ranges as class estimates — your llama.cpp build and the model's attention layout shift them. Token gen matches the RTX 4070 exactly; prompt proc. is ~20% higher.

27B models (Qwen class) — Q4 / Q5 / Q6 / Q8

QuantGGUF size1× 4070 S (12 GB)2× 4070 S (24 GB)
Q4_K_M~19 GB❌ ~7 GB over⚠️ ~16–20 t/s, 8K ctx
Q5_K_M~22 GB❌❌ ~2 GB over
Q6_K~25 GB❌❌
Q8_0~33 GB❌❌

Mirror image of the 4070: 27B Q4 at 2× (24 GB, 8K, unhurried), Q5 at 3× (36 GB, 8–16K). Multi-card token speeds match the 4070 — the Super's extra cores only help the prefill side of those builds.

Practical tips

Multi-card: 2×, 3×, 4×

PCIe 4.0, ~25–35% tax, 12 GB tiles — the 4070's arithmetic, prompt side faster.

BuildGPU powerVRAM13B Q827B Q4_K_M27B Q5_K_M
2× 4070 S440 W24 GB~40–52 t/s⚠️ ~16–20 t/s, 8K ctx❌ ~2 GB over
3× 4070 S660 W36 GB(overkill)~19–24 t/s, 16K+ ctx~14–18 t/s, 8–16K ctx

3× (36 GB) is the interesting configuration — 27B Q5 with a real context. Board requirement as the 4070: three working x8/x16 lanes.

The price check 2× 4070 Ti Super (570 W, 32 GB) does the same work faster with fewer cards. The 3-way Super build only wins where the extra 4 GB (Q5 at 16K+) is worth the third slot.

Offloading (CPU RAM)

When a model won't fit, -ngl N parks the tail layers on the CPU. The mental model that saves pain: offloaded layers don't run at a fraction of GPU speed — they run at CPU speed. Handy moves: peel the last 2–4 layers off to reclaim ~0.5–1 GB of VRAM for almost nothing; keep MoE experts on CPU with --n-cpu-moe; park the output embedding with -ot output=CPU on big models. Budget 32 GB system RAM for 27B-class offload, 64 GB for 70B. And never offload the KV cache — the session becomes unusable.

3-card example llama-server -m 27b-q5_k_m.gguf -ngl 99 -sm layer -c 16384 --tensor-split 1,1,1

Recommended llama.cpp command

llama-server -m llama-3.1-13b-q4_k_m.gguf -ngl 99 -c 32768 -b 2048 -ub 2048 --host 127.0.0.1 --port 8080

Verdict

The 4070 with a ~20% faster prefill and nothing else moved. As a purchase: same price, take it; premium price, skip it. As a platform: the 4070's verdict applies wholesale — a 13B card with a 12 GB ceiling and a multi-card road to 27B.

← RTX 4070 next: RTX 4070 Ti →