Um Ihnen die Funktionen unseres Online-Shops uneingeschränkt anbieten zu können setzen wir Cookies ein. Weitere Informationen
NVIDIA GPU-Vergleich
NVIDIA bietet mit Blackwell (B300, B200), Hopper (H200) und der RTX Pro mehrere Hochleistungs‑GPUs für moderne KI‑ und HPC‑Workloads an. Auf dieser Seite vergleichen wir die wichtigsten technischen Daten und Einsatzszenarien dieser GPUs und helfen Ihnen dabei, die passende Lösung für Ihr Rechenzentrum oder Ihren KI‑Cluster zu finden. Wenn Sie Unterstützung bei der Auswahl oder bei einer konkreten Server‑Konfiguration benötigen, beraten wir Sie gerne persönlich.
| NVIDIA Data Center GPUs | |||||
| B300 | B200 | H200 | RTX PRO 6000 Server Edition | ||
| GPU Architecture | NVIDIA Blackwell Ultra architecture | NVIDIA Blackwell architecture | NVIDIA Hopper architecture | NVIDIA Blackwell architecture | |
| Segment | Data center AI & HPC accelerator | Data center AI & HPC accelerator | Data center AI & HPC accelerator | Data center AI & HPC accelerator | Data center AI & visual computing accelerator |
| Form factor | SXM6 | SXM6 | SXM5 | PCIe | PCIe |
| GPU Memory | 288 GB HBM3e per GPU | 180 GB HBM3e per GPU | 141 GB HBM3e | 96 GB GDDR7 | |
| Memory bandwidth | Up to 8 TB/s | 7.7 TB/s | 4.8 TB/s | 1.6 TB/s | |
| Max TDP | 1200 W | 1400 W | 1200 W | 1000 W | 700 W | 600 W | 400 - 600 W |
| NVLink | Yes | No | |||
| Multi-Instance GPU (MIG) | Yes | Yes | Yes | Yes | Yes |
| Media engines | 7 JPEG Decoder 7 Video Decoder | 7 JPEG Decoder 7 Video Decoder | 7 JPEG Decoder 7 Video Decoder | 7 JPEG Decoder 7 Video Decoder | 4x NVENC 4x NVDEC |
| vGPU | Yes | ||||
| Confidential Computing | Yes | ||||
| NVIDIA AI Enterprise | Yes | ||||
| Last Generation GPUs | Preis | Preis | Preis | Preis | Preis |
| Preis Edu/Startup | Preis Edu/Startup | Preis Edu/Startup | Preis Edu/Startup | Preis Edu/Startup | |
| NVIDIA RTX PRO GPUs | |||||||
RTX PRO 4000 | RTX PRO 4500 | RTX PRO 5000 | RTX PRO 5000 72GB | RTX PRO 6000 Max-Q | RTX PRO 6000 Server Edition | RTX PRO 6000 Workstation Edition | |
| GPU Architecture | Blackwell | ||||||
| CUDA Cores | 8960 | 10496 | 14080 | 14080 | 24064 | ||
| Tensor Cores | 280 (5. Gen) | 328 (5. Gen) | 400 (5. Gen) | 400 (5. Gen) | 752 (5. Gen) | ||
| RT Cores | 70 (4. Gen) | 82 (4. Gen) | 100 (4. Gen) | 100 (4. Gen) | 188 (4. Gen) | ||
| Peak Single-Precision Performance | - | - | 67 TFLOPS | 67 TFLOPS | 110 TFLOPS | 120 TFLOPS | 125 TFLOPS |
| Peak Double-Precision Performance | - | - | - | - | - | - | |
| AI TOPS | - | - | 2142 AI TOPS | 2142 AI TOPS | 3511 AI TOPS | - | 4000 AI TOPS |
| RT Core Performance | - | - | 203 TFLOPS | 203 TFLOPS | 333 TFLOPS | 355 TFLOPS | 380 TFLOPS |
| Memory Size | 24 GB GDDR7 | 32 GB GDDR7 | 48 GB GDDR7 | 72 GB GDDR7 | 96 GB GDDR7 | ||
| Memory Bandwidth | TBD | 896 GB/s | 1344 GB/s | 1344 GB/s | 1792 GB/s | 1597 GB/s | 1792 GB/s |
| NVLink | Not supported | ||||||
| Thermal Solution | Blower Active Fan | Blower Active Fan | Blower Active Fan | Blower Active Fan | Blower Active Fan | Passive | Double Flow Through |
| NVENC | NVDEC | JPEG | 2x | 2x | 1x | 2x | 2x | 1x | 3x | 3x | 3x | 3x | 3x | 3x | 4x | 4x | 4x | ||
| MIG Support | - | - | 2x 24GB | 2x 36GB | 4x 24GB | 4x 24GB | 4x 24GB |
| Display Connectors | 4x DP 2.1b | ||||||
| Form Factor | 4.4"H x 9.5"L, Single Slot 2.7"H x 6.6"L, Dual Slot | 4.4"H x 10.5"L, Dual Slot | 5.4"H x 12"L, XHFL Dual Slot | ||||
| Max Power Consumption | 140 W | 200 W | 300 W | 300 W | 400 - 600 W | 600 W | |
| Graphics Bus | PCIe Gen 5 x16 | ||||||
| Last Generation GPUs | Preis | Preis | Preis | TBA | Preis | Preis | Preis |
| Last NVIDIA Data Center GPUs | |||||||
| A100/A800 | A30 | A40 | A10 | A16 | A2 | ||
| Design | Highest Perf Compute | Mainstream Compute | High Perf Graphics | Mainstream & Video with AI | High Density Virtual Desktop | Entry-Level Small Footprint | |
| Form Factor | SXM4 | PCIe Gen4 x16 2 Slot FHFL 3 NVLink Bridge | PCIe Gen4 x16 2 Slot FHFL 1 NVLink Bridge | PCIe Gen4 x16 2 Slot FHFL 1 NVLink Bridge | PCIe Gen4 x16 1 Slot FHFL | PCIe Gen4 x16 2 Slot FHFL | PCIe Gen4 x8 1 Slot LP |
| Max Power | 500W | 300W | 165W | 300W | 150W | 250W | 60W |
| FP64 TC | FP32 TFLOPS | 19.5 | 19.5 | 19.5 | 19.5 | 10 | 10 | NA | 37 | NA | 31 | NA | 4x 4.5 | NA | 4.5 |
| TF32 TC | FP16 TC TFLOPS | 312 | 624 | 312 | 624 | 165 | 330 | 150 | 300 | 125 | 250 | 4x 18 | 4x 36 | 18 | 36 |
| FP8/FP6 TC | INT8 TC TFLOPS/TOPS | NA | 1248 | NA | 1248 | NA | 661 | NA | 600 | NA | 500 | NA | 4x 72 | NA | 72 |
| FP4 TC | - | - | - | - | - | - | - |
| GPU Memory | 80 GB HBM2e | 80 GB HBM2e | 24 GB HBM2 | 48 GB GDDR6 | 24 GB GDDR6 | 4x 16 GB GDDR6 | 16 GB GDDR6 |
| Multi-Instance GPU (MIG) | Up to 7 | Up to 7 | Up to 4 | - | - | - | - |
| NVLink Connectivity | Up to 8 | 2 cards | 2 cards | 2 cards | - | - | - |
| Media Acceleration | 1 JPEG Decoder 5 Video Decoder | 1 JPEG Decoder 5 Video Decoder | 1 JPEG Decoder 4 Video Decoder | 1 Video Encoder 2 Video Decoder (+AV1 decode) | 1 Video Encoder 2 Video Decoder (+AV1 decode) | 4 Video Encoder 8 Video Decoder (+AV1 decode) | 1 Video Encoder 2 Video Decoder (+AV1 decode) |
| Ray Tracing | - | - | - | Yes | Yes | Yes | Yes |
| Transformer Engine | - | - | - | - | - | - | - |
| DPX Instructions | - | - | - | - | - | - | - |
| Graphics | For-in-situ viz (no vPC or RTX vWS) | For-in-situ viz (no vPC or RTX vWS) | For-in-situ viz (no vPC or RTX vWS) | Best | Better | Good | Good |
| vGPU | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Hardware Root of Trust | Internal with Option for External | Internal with Option for External | Internal with Option for External | Internal with Option for External | Internal with Option for External | Internal with Option for External | Internal with Option for External |
| Confidential Computing | (1) | - | - | - | - | - | - |
| NVIDIA AI Enterprise | Add-on | Add-on | Add-on | Add-on | Add-on | Add-on | Add-on |
| Preis | Preis | Preis | Preis | ||||
| Last NVIDIA RTX Desktop GPUs | |||||||
| RTX A400 | RTX A1000 | RTX 2000 Ada | RTX 4000 Ada | RTX 4500 Ada | RTX 5000 Ada | RTX 6000 Ada | |
| GPU Architecture | Ampere | Ampere | Ada-Lovelace | Ada-Lovelace | Ada-Lovelace | Ada-Lovelace | Ada-Lovelace |
| CUDA Cores | 768 | 2304 | 2816 | 6144 | 7680 | 12800 | 18176 |
| Tensor Cores | 24 (3. Gen) | 72 (3. Gen) | 88 (4. Gen) | 192 (4. Gen) | 240 (4. Gen) | 400 (4. Gen) | 568 (4. Gen) |
| RT Cores | 6 (2. Gen) | 18 (2. Gen) | 22 (3. Gen) | 48 (3. Gen) | 60 (3. Gen) | 100 (3. Gen) | 142 (3. Gen) |
| Peak Single-Precision Performance | 2.7 TFLOPS | 6.7 TFLOPS | 12 TFLOPS | 26.7 TFLOPS | 39.6 TFLOPS | 65.3 TFLOPS | 91.1 TFLOPS |
| Peak Double-Precision Performance | - | - | - | - | - | - | - |
| Tensor Performance | 21.7 TFLOPS | 53.8 TFLOPS | 191.9 TFLOPS | 327.6 TFLOPS | 634.0 TFLOPS | 1044.4 TFLOPS | 1457.0 TFLOPS |
| RT Core Performance | 5.4 TFLOPS | 13.2 TFLOPS | 27.7 TFLOPS | 61.8 TFLOPS | 91.6 TFLOPS | 151.0 TFLOPS | 210.6 TFLOPS |
| Memory Size | 4 GB | 8 GB | 16 GB GDDR6 | 20 GB GDDR6 | 24 GB GDDR6 | 32 GB GDDR6 | 48 GB GDDR6 |
| Memory Bandwidth | 96 GB/s | 192 GB/s | 224 GB/s | 360 GB/s | 432 GB/s | 576 GB/s | 960 GB/s |
| NVLink | Not supported | Not supported | Not supported | Not supported | Not supported | Not supported | Not supported |
| VR Ready | No | No | Yes | Yes | Yes | Yes | Yes |
| Virtual Workstation | No | No | No | No | No | Yes | Yes |
| Display Connectors | 4x mDP 1.4a | 4x mDP 1.4a | 4x mDP 1.4a | 4x DP 1.4a | 4x DP 1.4a | 4x DP 1.4a | 4x DP 1.4a |
| Form Factor | 2.7"H x 6.4"L, Single Slot | 2.7"H x 6.4"L, Single Slot | 2.7"H x 6.4"L, Single Slot | 4.4"H x 9.5"L, Single Slot | 4.4"H x 10.5"L, Dual Slot | 4.4"H x 10.5"L, Dual Slot | 4.4"H x 10.5"L, Dual Slot |
| Max Power Consumption | 50 W | 50 W | 70 W | 130 W | 210 W | 250 W | 300 W |
| Graphics Bus | PCIe Gen 4 x8 * | PCIe Gen 4 x8 * | PCIe Gen 4 x8 * | PCIe Gen 4 x16 | PCIe Gen 4 x16 | PCIe Gen 4 x16 | PCIe Gen 4 x16 |
| Preis | Preis | Preis | Preis | Preis | |||
| Preis Edu/Startup | Preis Edu/Startup | ||||||
Planen Sie ein neues KI‑Projekt oder möchten Sie eine bestehende Umgebung aufrüsten? Wir unterstützen Sie bei der Auswahl der passenden NVIDIA‑GPU‑Plattform – von RTX Pro 6000 bis hin zu B300‑basierten Systemen. Teilen Sie uns kurz Ihren Anwendungsfall mit (z. B. LLM‑Training, Inferenz, VDI oder Visualisierung), und wir erstellen ein individuelles Angebot.
Kontakt: Tel: +49 40 300672 - 0 | Fax: +49 40 300672 - 11 | E-Mail: info[at]delta.de
