Where they actually differ
The most decisive specs, ranked by how much they matter — before the full spec table.
| Spec | Intel - Arc B580 | NVIDIA - GeForce RTX 5090 |
|---|---|---|
| CUDA / Stream cores | 2560 | 21760 |
| Ray tracing cores | 20 | 170 |
| TDP | 190 W | 575 W |
| 3DMark Time Spy | 9400 | 32500 |
| VRAM | 12 GB | 32 GB |
| Memory bus | 192 bit | 512 bit |
Between the Arc B580 and GeForce RTX 5090, the Arc B580 comes out ahead in tdp / boost clock, while the GeForce RTX 5090 wins on cuda / stream cores / ray tracing cores. Overall, the Arc B580 scores 59.2 and the GeForce RTX 5090 scores 84.2 on our Mars Score.
Which should you buy?
The GeForce RTX 5090 is the clear pick on raw spec strength, leading by 25.0 points on the Mars Score. The margin is wide enough that the underdog only makes sense if its individual strengths line up with what you specifically need.
Pick the Arc B580 if you care most about tdp / boost clock. The biggest gaps in its favor are tdp, boost clock. The single most decisive one: TDP — 385.00 W less. 190 vs 575.
Pick the GeForce RTX 5090 if you push the hardware — gaming, on-device AI, heavy multitasking — and the extra compute headroom is worth the trade-offs. Its strongest claims are cuda / stream cores, ray tracing cores, 3dmark time spy, among others. The single most decisive one: CUDA / Stream cores — 750.00% more. 21760 vs 2560.
There are no meaningful ties on the headline specs — every popular-comparison metric has a winner.
Bottom line: the GeForce RTX 5090 is the safer default. The other model is the right choice only when its specific advantages line up with your priorities.
Which should you buy: Intel - Arc B580 or NVIDIA - GeForce RTX 5090?
NVIDIA - GeForce RTX 5090 leads the current Mars Score by 25.0 points.The more useful decision is the trade-off behind that result: compare vram, cuda / stream cores, boost clock, tdp against the way you will actually use the product.
Choose Intel - Arc B580 if…
Intel - Arc B580 has the measurable edge on tdp w, boost clock mhz.
Choose NVIDIA - GeForce RTX 5090 if…
NVIDIA - GeForce RTX 5090 has the measurable edge on cuda cores, ray tracing cores, timespy score.
This is a structured comparison of the published catalog values. It is not a claim of identical real-world performance across every workload; use the linked spec drill-downs and methodology page to inspect the inputs.
Why Intel - Arc B580 wins
- ▲TDP — 385.00 W less. 190 vs 575.
- ▲Boost clock — 18.40% more. 2850 MHz vs 2407 MHz.
Why NVIDIA - GeForce RTX 5090 wins
- ▲CUDA / Stream cores — 750.00% more. 21760 vs 2560.
- ▲Ray tracing cores — 750.00% more. 170 vs 20.
- ▲3DMark Time Spy — 245.74% more. 32500 vs 9400.
- ▲VRAM — 166.67% more. 32 GB vs 12 GB.
- ▲Memory bus — 166.67% more. 512 bit vs 192 bit.
- ▲Has DLSS.
Spec comparison
Core
| Spec | Intel - Arc B580 | NVIDIA - GeForce RTX 5090 |
|---|---|---|
| Chipset | BMG-G21 | GB202 |
| Base clockⓘ | 2670 MHzlead | 2017 MHz |
| Boost clockⓘ | 2850 MHzlead | 2407 MHz |
| CUDA / Stream coresⓘ | 2560 | 21760lead |
| Ray tracing coresⓘ | 20 | 170lead |
Memory
| Spec | Intel - Arc B580 | NVIDIA - GeForce RTX 5090 |
|---|---|---|
| VRAMⓘ | 12 GB | 32 GBlead |
| VRAM typeⓘ | GDDR6 | GDDR7lead |
| Memory busⓘ | 192 bit | 512 bitlead |
Power & Form
| Spec | Intel - Arc B580 | NVIDIA - GeForce RTX 5090 |
|---|---|---|
| TDPⓘ | 190 Wlead | 575 W |
| Recommended PSUⓘ | 600 Wlead | 1000 W |
| Length | 272 mmlead | 304 mm |
| Slot size | 2-slot | 2-slot |
Benchmarks
| Spec | Intel - Arc B580 | NVIDIA - GeForce RTX 5090 |
|---|---|---|
| 3DMark Time Spyⓘ | 9400 | 32500lead |
| DLSSⓘ | false | truelead |
| FSRⓘ | truelead | false |
Frequently asked
Which has more VRAM, Arc B580 or GeForce RTX 5090?+
GeForce RTX 5090 ships with 32GB GDDR7 vs 12GB.
Which scores higher in 3DMark Time Spy, Arc B580 or GeForce RTX 5090?+
GeForce RTX 5090 scores 32500 vs 9400.
Which draws less power, Arc B580 or GeForce RTX 5090?+
Arc B580 draws 190W TDP vs 575W.