NVIDIA MAINTENANCE · TPM AFTER NVIDIA ENTERPRISE SUPPORT
Nvidia Maintenance — vendor-independent service for DGX, Mellanox, Spectrum and Quantum InfiniBand
We service Nvidia hardware after Nvidia Enterprise Support expiration — vendor-independent, with OEM original parts and SLA up to 24×7×4. DGX AI servers (from DGX-1 to current H200), Mellanox networking (ConnectX adapters, Spectrum Ethernet, Quantum InfiniBand) — also in the EOSL period. Typical TPM topics at the compute layer: BMC/iLO/iDRAC firmware updates without OEM login (lockout risk after contract end), DIMM failures as the most frequent server defect, and BIOS CPU microcode patches for Spectre/Meltdown-class vulnerabilities.
3
Product categories
9+
Product lines
4
Detail pages
30–60 %
Savings vs. Nvidia Enterprise Support / DGX Premium Support
What we cover in Nvidia hardware
3 categories · 9 product lines · 3 with EOSL · 9 detail pages
Mellanox · Network Adapter
2 linesSpectrum / Quantum · Switching
3 lines EOSL · 1| Category | Product line | Models | EOSL |
|---|---|---|---|
| DGX · AI Compute | DGX-1 / DGX-2 (älter) | DGX-1 (Pascal/Volta) · DGX-2 (Volta - 16x V100) | yes |
| DGX · AI Compute | DGX A100 | DGX A100 320GB · DGX A100 640GB | yes |
| DGX · AI Compute | DGX H100 / H200 | DGX H100 · DGX H200 | no |
| DGX · AI Compute | DGX GH200 / GB200 | DGX GH200 · DGX GB200 (NVL72) | no |
| Mellanox · Network Adapter | ConnectX-4/-5/-6/-7 (Adapter) | ConnectX-4 · ConnectX-5 · ConnectX-6 Dx · ConnectX-7 | no |
| Mellanox · Network Adapter | BlueField DPU (Data Processing Unit) | BlueField-2 · BlueField-3 | no |
| Spectrum / Quantum · Switching | Spectrum Ethernet (SN-Serie) | Spectrum SN2000 · SN3000 · SN4000 · SN5000 (10/25/100/200/400G) | no |
| Spectrum / Quantum · Switching | Quantum InfiniBand (QM-Serie) | Quantum-1 (HDR 200G) · Quantum-2 (NDR 400G) · QM8700/QM9700 | no |
| Spectrum / Quantum · Switching | ältere Mellanox SX-/SB-Switches | SX1012/SX1024/SX1036 · SB7700/SB7800 | yes |
Detail pages for key product lines
These product lines have their own maintenance detail page with model overview, EOSL table and FAQ. All other lines are also covered — see portfolio below.
Nvidia
DGX-1 / DGX-2 (älter)
DGX-1 (Pascal/Volta) · DGX-2 (Volta - 16x V100)
Detail page & FAQNvidia
DGX A100 + 2more lines
DGX A100 320GB · DGX A100 640GB · DGX H100 / H200 · DGX GH200 / GB200
Detail page & FAQNvidia
ConnectX-4/-5/-6/-7 (Adapter) + 1more line
ConnectX-4 · ConnectX-5 · ConnectX-6 Dx · ConnectX-7 · BlueField DPU (Data Processing Unit)
Detail page & FAQNvidia
Spectrum Ethernet (SN-Serie) + 2more lines
Spectrum SN2000 · SN3000 · SN4000 · SN5000 (10/25/100/200/400G) · Quantum InfiniBand (QM-Serie) · ältere Mellanox SX-/SB-Switches
Detail page & FAQNvidia product portfolio at a glance
We service all product lines — including older generations long past end-of-service-life. Every line with OEM original parts, German-speaking service and SLA up to 24×7×4.
DGX · AI Compute
DGX-1 / DGX-2 (älter)
DetailsDGX-1 (Pascal/Volta) · DGX-2 (Volta - 16x V100)
First DGX generations. EOSL reached — TPM only option for active fleets.
DGX A100
DetailsDGX A100 320GB · DGX A100 640GB
Ampere generation with 8x A100 GPUs, widely deployed AI platform in DACH research environments.
DGX H100 / H200
DetailsDGX H100 · DGX H200
Hopper generation with 8x H100/H200 GPUs for current LLM and training workloads.
DGX GH200 / GB200
DetailsDGX GH200 · DGX GB200 (NVL72)
Grace Hopper / Blackwell Superchip platforms for hyperscale AI.
Mellanox · Network Adapter
ConnectX-4/-5/-6/-7 (Adapter)
DetailsConnectX-4 · ConnectX-5 · ConnectX-6 Dx · ConnectX-7
High-performance network adapters (25/50/100/200/400 Gigabit) for server connectivity.
BlueField DPU (Data Processing Unit)
DetailsBlueField-2 · BlueField-3
DPU cards with ARM cores for network offload and storage acceleration.
Spectrum / Quantum · Switching
Spectrum Ethernet (SN-Serie)
DetailsSpectrum SN2000 · SN3000 · SN4000 · SN5000 (10/25/100/200/400G)
Ethernet switches for data center and HPC networks.
Quantum InfiniBand (QM-Serie)
DetailsQuantum-1 (HDR 200G) · Quantum-2 (NDR 400G) · QM8700/QM9700
InfiniBand switches for AI/HPC clusters with lowest latency.
ältere Mellanox SX-/SB-Switches
DetailsSX1012/SX1024/SX1036 · SB7700/SB7800
Predecessor generations, EOSL reached.
Nvidia Enterprise Support / DGX Premium vs. TechCare TPM
How does TPM compare to the OEM contract?
You save 41–50% per yearNvidia Enterprise Support / DGX Premium vs. TechCare TPM · example calculation
Enterprise Support / DGX Premium
TechCare TPM
Example NVIDIA DGX A100 with 4-year Enterprise Support: EUR 18,000–24,400/year OEM, EUR 10,700–12,200/year with TechCare TPM.
- EOSL hardware: On EOSL hardware (DGX-1, DGX Station A100, V100/T4 generation) 50–65 % savings vs. Enterprise Support — NVIDIA EOSL extension on DGX limited.
- Multi-year contract: 3-year TPM contract: additional 5–10 % discount, 5-year contract: 10–15 %. Plus fixed-price lock against OEM list-price increases.
Stays with Nvidia
- NVIDIA AI Enterprise software subscription
- DGX OS / Base Command subscription
- NVIDIA TAC software tickets
Moves to TechCare
- Hardware defects (memory, disk, PSU, mainboard, modules)
- 24×7×4 on-site SLA with German-speaking field service
- Spare-parts logistics DACH + global hub network
- Fixed-price contract in 48 h, one inventory — one contract
Why Nvidia maintenance via TechCare instead of Nvidia Enterprise Support / DGX Premium Support
Nvidia hardware is mainly deployed in DACH research data centers, universities, pharma companies and (increasingly) AI startups. DGX platforms are very CapEx-intensive (DGX H100 from 350,000 EUR) — premature refresh is rarely economically justifiable when hardware still functions. Nvidia Enterprise Support is above-average expensive.
We service all Nvidia platforms with OEM original parts — DGX mainboards, PSUs, cooling components, GPU trays, NVMe modules, Mellanox adapters, Spectrum linecards, Quantum optics. Note: Nvidia software stack (CUDA, NGC containers, AI Enterprise subscription) stays separately with Nvidia.
Compute-specific TPM topics: BMC/iLO/iDRAC firmware updates without OEM subscription, DIMM failure diagnostics (the most common server defect after PSU), and BIOS CPU microcode patches for Spectre-/Meltdown-class vulnerabilities — also post-EoSL via mirror repositories.
1
Contract for all Nvidia lines
30–60 %
Savings on current models
bis 70 %
Savings on EOSL models
48 h
Fixed-price quote
FAQ on Nvidia maintenance
Which Nvidia hardware do you service?
We service the complete DGX family (DGX-1 to DGX GB200), Mellanox network adapters (ConnectX-4 to ConnectX-7, BlueField DPUs), Spectrum Ethernet switches (SN2000 to SN5000) and Quantum InfiniBand switches.
What does TPM cost compared to Nvidia Enterprise Support?
30 to 60 percent savings vs. Nvidia Enterprise Support, up to 70 percent on EOSL platforms. A DGX A100 with 24×7×4 costs 35,000 to 55,000 EUR/year with Nvidia, 14,000 to 25,000 EUR with TechCare.
Can you service DGX-1/DGX-2?
Yes. DGX-1 (Pascal/Volta, released 2016/2017, EOSL reached) and DGX-2 (Volta, 2018, EOSL reached or imminent) are still productive in many DACH research environments.
What about CUDA, NGC and Nvidia AI Enterprise?
Nvidia software stack runs separately via Nvidia. CUDA and NGC containers are partly freely available. Nvidia AI Enterprise and Base Command Manager require active Nvidia contract. We service only the hardware layer.
Which SLA levels for DGX and HPC networking?
For DGX and Quantum InfiniBand we generally recommend 24×7×4 because failures can halt AI training jobs for days. For individual Mellanox adapters or Spectrum edge switches 5×9 NBD may suffice.
Do you also service the GPU cards in the DGX?
In DGX platforms GPUs are integrated modules (HGX boards with 8x GPU + NVLink). We service the complete server including HGX boards with GPUs — these are not individually swappable like in standard servers. For individual GPU defects in HGX boards a complete board swap is typically needed.
Do we lose the warranty or software licenses?
No. Factory warranty and Nvidia Enterprise Support are contractual services. Nvidia software licenses run separately and are not affected by a hardware maintenance switch.
Service performance
Real actuals Q1 2026 — straight from our ITIL ticketing.
99,2 %
Tickets resolved within agreed response time
2,4 h
Avg. first response on 4h SLA tier
88 %
First-time fix on initial dispatch
97 %
Spare part on site within 4 h, DACH depots
Related topics
Core service
Hardware maintenance in detail
How we set up TPM: contract, SLA tiers up to 24×7×4, engineer pool, spare parts strategy — the service behind the vendor hubs.
View service →Compliance
Audit documentation in the contract pack
DORA, NIS2, KRITIS, BAIT, TISAX, ISO 27001 — what the auditor checks, what TechCare delivers automatically with the contract.
View audit topics →Practice
Real references
Three templates from European customers 2024–25 — how the migration from OEM to TPM ran in reality.
View cases →