Quick answer: The RTX 5090 earned its keep for VRAM-bound work — 34B LLMs at Q5-Q6, Flux.2’s 4-bit build with ControlNet headroom, beefier LoRA batches. Note what it does not unlock: a 70B at q4_K_M is 43GB and Flux.2 at BF16 is ~64GB, so neither fits 32GB either. But for image generation and most hobbyist workflows, the $2,700 premium over a 4090 wasn’t worth it. Honestly, six months in, we recommend the 5090 only if you’re VRAM-limited.
NVIDIA GeForce RTX 5090
32GB GDDR7If you're hitting OOM on a 4090 with 34B models at Q5, Flux.2 plus ControlNet, or larger training batches, the 5090's 32GB is the only consumer card that fixes it. Otherwise, skip.
Affiliate links — we may earn a commission at no extra cost to you. Amazon ships globally; Shopee SG covers Singapore & ASEAN.
Who this is for
This is for the person sitting on a perfectly good RTX 4090 wondering if the 5090 upgrade is worth $4,900. Or the buyer choosing between them for a fresh AI rig. We’ve run both daily since January 2026 — SDXL, Flux, Llama 70B inference, LoRA training, some PyTorch research code — and the answer is more nuanced than the launch-week reviews implied.
What actually improved (the wins)
Let’s start with what the 5090 genuinely changed.
32GB GDDR7 is the headline, and it earned it. The 4090’s 24GB is a real ceiling, and the extra 8GB moves several workloads across it. Be precise about which ones, though: a 70B at q4_K_M is a 43GB download, so it does not fit a 5090 either — that card tops out at q3_K_S (31GB), while a 4090 holds no 70B at all, since even q2_K is 26GB. Where the 32GB genuinely changes the answer is one tier down. Flux.2 is not that tier either — its transformer is ~64GB at BF16 and ~32GB at FP8, so a 5090 runs the same 4-bit build a 4090 does, just with room left for a ControlNet. The real wins are 34B LLMs at Q5-Q6, and SDXL LoRA training with batch size 4 and the text encoder unfrozen. The gain is real; it is just one tier, not a 70B unlock.
FP8 training is finally usable on consumer hardware. Blackwell’s native FP8 tensor cores aren’t just a checkbox — we’ve trained LoRAs in FP8 with measurable VRAM savings and ~1.7x throughput vs BF16 on the same card. The 4090 can do FP8 via software (Transformer Engine emulation) but it’s clunky and the speedup evaporates. If you’re serious about training, this matters. We’ve covered this trade-off in more depth in our best GPU for PyTorch guide, where FP8 native support genuinely shifts the recommendation.
Memory bandwidth at 1,792 GB/s vs 1,008 GB/s is a real LLM inference boost. Llama 70B Q4 went from “barely usable” (~12-15 tok/s with offloading on a 4090) to “actually fine” (~35-40 tok/s on a 5090, no offloading). For a 13B model, 5090 hits ~140 tok/s vs 95 on the 4090. That 35-46% spread on LLM tok/s is consistent across our RTX 4090 vs 5090 head-to-head testing and matches the Knightli benchmarks the community has been citing.
For AI research workloads, the 32GB unlocks experiments you couldn’t run before. Bigger context windows, full-fp16 attention on longer sequences, MoE models with more experts loaded — the kind of work we cover in our best GPU for AI research deep-dive simply wasn’t possible on a 4090 without dropping to quantization or splitting across cards. If you’re doing actual research workloads (not just running pretrained models), the 5090’s VRAM is a structural advantage.
What didn’t change as much as advertised
Now the contrarian half — and this is where the launch-week articles oversold the card.
Image generation only improved 20-25%. SDXL went from 6.5s/img to ~4.0s/img. Flux dev from 18s to ~14s. That’s nice, but not life-changing. If you’re cranking out images, you’ll notice. If you generate a few per session, you genuinely will not feel $2,700. Look — if you only do image gen, save your $2,700 and buy a 4090.
Gaming-grade BF16 isn’t 2x faster. Despite the spec sheet showing ~130 TFLOPS FP16 vs 82.6 on the 4090 (a ~57% theoretical jump), real-world BF16 training throughput sits closer to +35-45% in our runs. The Blackwell scheduler and driver stack are still maturing — we’ve seen kernels regress between driver versions. Honestly, the 5090 disappointed us here. We expected 1.7-2x, got 1.4x.
The premium hurts more once you account for PSU and case. 575W TGP vs 450W means a lot of 850W PSUs that ran a 4090 fine are now marginal. Add ~$150 for a quality 1000W unit, factor in transient spikes that have crashed some 1000W units (real reports, not theory), and the swap lands near $2,850 over keeping the 4090. For 25% faster image gen.
Check RTX 4090 price on Amazon→Buy on Shopee SG→For the majority of AI hobbyists, the 4090 is still the right answer. It’s not slower in any way that breaks workflows — it’s slower in ways that add seconds, not minutes. We still recommend it as the default in our best GPU for AI cluster guide, and after 6 months with both cards we’re not walking that back.
Workload-by-workload verdict
Here’s how we’d actually advise per workflow, based on six months of daily use:
| Workload | Recommendation | Why |
|---|---|---|
| Image gen (SDXL, Flux dev) | RTX 4090 | 20-25% speedup doesn’t justify $2,700. Both fit the models comfortably. |
| Image gen (Flux.2 + ControlNet, video models) | RTX 5090 | Neither card runs Flux.2 above 4-bit; 32GB is what leaves room after it loads. |
| LLM inference (≤13B) | RTX 4090 | Both run flat-out. 95 vs 140 tok/s is a “nice to have.” |
| LLM inference (34B-70B) | RTX 5090 | This is where 32GB earns the upgrade. Q4 70B is usable, not painful. |
| LoRA fine-tuning (small models) | RTX 4090 | Workflows fit. Speedup exists but isn’t transformative. |
| Full fine-tuning / FP8 training | RTX 5090 | Native FP8 + bandwidth + VRAM all compound here. Real win. |
| AI research (long context, MoE, novel architectures) | RTX 5090 | The 8GB extra unlocks experiments. Not optional. |
| Mixed hobbyist (some of everything) | RTX 4090 | Honest answer. Most workflows aren’t VRAM-limited. |
Common mistakes we’ve watched people make
After six months of watching the 4090 → 5090 upgrade discourse, four mistakes keep coming up.
1. Buying the 5090 just for image generation. If your SDXL/Flux workflow runs fine on a 4090, you are buying 25% speed for $2,700. Save the money — at this gap it is most of a second machine.
2. Underestimating the PSU and thermal upgrade. That 850W gold-rated PSU you bought for your 4090 build is marginal at 575W card TGP plus a modern CPU. Plan for $150-200 of platform upgrades, not just the GPU swap.
3. Assuming “newer architecture” means “always faster.” Blackwell drivers were rough through Q1 2026. We saw PyTorch training kernels actually slower than Ada Lovelace on certain ops until the April driver. If you need rock-stable today, the 4090’s mature software stack is genuinely an asset.
4. Buying the 5090 because the 5080 is bad. The RTX 5080 at 16GB is a real letdown for AI — too little VRAM for the price. Don’t let that push you up to the 5090 unconsidered. The 4090 is the actual sweet spot in that lineup, not the 5090 by default.
Final verdict
| RTX 4090 | RTX 5090 | |
|---|---|---|
| VRAM | 24GB GDDR6X | 32GB GDDR7 |
| Bandwidth | 1,008 GB/s | 1,792 GB/s |
| Image gen | 6.5s/img SDXL | 4.0s/img SDXL |
| 70B inference | Painful | Actually usable |
| FP8 training | Software emulation | Native, ~1.7x faster |
| TGP | 450W | 575W |
| Street price | ~$2,200 | ~$4,900 |
| Our verdict | Default pick for most | Worth it only if VRAM-bound |
NVIDIA GeForce RTX 5090
32GB GDDR7The 5090's case is narrow but real: 34B LLMs at higher quants, Flux.2 with ControlNet headroom, FP8 training, and research workloads that wouldn't fit on a 4090 at all. If that's not you, the 4090 is the smarter buy.
Affiliate links — we may earn a commission at no extra cost to you. Amazon ships globally; Shopee SG covers Singapore & ASEAN.
One-sentence verdict: The RTX 5090 is a real upgrade for VRAM-bound and FP8-training workflows — but for the average AI hobbyist running SDXL and 13B models, the 4090 still wins on value, and we’d buy it again.
Frequently asked questions
Is the RTX 5090 worth upgrading from a 4090 for AI?
Only if you’re VRAM-limited on 24GB — think 34B models at Q5-Q6, Flux.2’s 4-bit build with a ControlNet loaded, or FP8 training. It will not get you a 70B at q4_K_M (43GB) or Flux.2 at full precision (~64GB); neither fits 32GB. For SDXL and 13B models, the upgrade is roughly 25% faster and not worth $2,700 plus PSU costs.
How much faster is the RTX 5090 than the 4090 for image generation?
Roughly 20-25% in our 6-month testing — SDXL drops from about 6.5s to 4.0s per image, Flux dev from 18s to 14s. Real, but not transformative.
Does the RTX 5090 actually deliver on FP8 training?
Yes — Blackwell’s native FP8 tensor cores give roughly 1.7x throughput vs BF16 with real VRAM savings. The 4090 can emulate FP8 in software but the speedup is much smaller.
Do I need a new power supply for the RTX 5090?
Likely yes. The 5090’s 575W TGP plus transient spikes makes 850W PSUs marginal. Budget for a quality 1000W unit, adding roughly $150-200 to your upgrade cost.