This page compares the OpenMetal Medium v4 bare metal server against the AWS m7i instance family, the closest match by general-purpose Intel Xeon spec profile. The comparison is structural, not just price-vs-price: it covers tenancy model, egress billing, storage persistence, management access, and total cost of ownership for sustained workloads. The m7i family runs on the same Intel Xeon Sapphire Rapids generation as the Medium v4, making this a like-for-like architecture comparison where the differences are in how the infrastructure is delivered and billed.
Key Takeaways
- Dedicated hardware vs shared tenancy: Every Medium v4 is a single-tenant server with full IPMI access. The m7i family runs on shared physical hosts by default; dedicated hosts cost significantly more .
- Fixed monthly pricing vs hourly billing: The Medium v4 charges one fixed monthly rate regardless of utilization. AWS m7i instances bill per hour on-demand, or require 1-3 year reserved instance commitments for discounts.
- Persistent NVMe vs EBS-only storage: The Medium v4 includes 6.4 TB of local Micron 7500 MAX NVMe that persists across reboots. The m7i family has no instance storage — all storage is EBS, which adds latency and per-GB/month cost .
- 95th-percentile egress vs per-GB transfer: OpenMetal bills egress on 95th-percentile burst. AWS charges $0.09/GB for the first 10 TB , creating cost escalation for egress-heavy workloads.
- Full IPMI access vs SSH/SSM only: The Medium v4 provides BIOS-level remote management — power control, console access, OS install, hardware diagnostics. AWS instances are accessible only via SSH or Systems Manager.
- HIPAA-eligible dedicated infrastructure: OpenMetal is HIPAA compliant at the organizational level on dedicated hardware. AWS HIPAA eligibility varies by service and requires separate BAA configuration .
Spec Comparison
| Component | OpenMetal Medium v4 | AWS m7i.8xlarge | AWS m7i.16xlarge |
|---|---|---|---|
| CPU | 24 physical cores / 48 threads (dedicated) | 32 vCPUs (shared host) | 64 vCPUs (shared host) |
| RAM | 256 GB DDR5-4400 | 128 GB | 256 GB |
| Storage | 6.4 TB Micron 7500 MAX NVMe (persistent) | EBS only (no instance storage) | EBS only (no instance storage) |
| Network | 20 Gbps private (LACP), 2 Gbps public | Up to 12.5 Gbps | Up to 25 Gbps |
| Tenancy | Single-tenant dedicated hardware | Shared (dedicated host extra) | Shared (dedicated host extra) |
| Remote Management | Full IPMI (power, console, BIOS, OS install) | SSH/SSM only | SSH/SSM only |
| Boot/Data Isolation | Dedicated RAID 1 boot drives | EBS boot volume (shared backend) | EBS boot volume (shared backend) |
| Storage Persistence | Local NVMe (persistent across reboots) | EBS only (adds cost + latency) | EBS only (adds cost + latency) |
Note on instance selection: The m7i.8xlarge is the closest match by vCPU count (32 vCPU vs 24 physical cores / 48 threads), while the m7i.16xlarge matches on RAM (256 GB). Neither provides local NVMe storage, persistent boot/data isolation, or IPMI access. All m7i instances require EBS for any storage, adding both cost and latency compared to local NVMe.
Processor: Dedicated Cores vs Shared vCPUs
Both the Medium v4 and the m7i family use Intel Xeon Sapphire Rapids (5th Gen) processors [VERIFY — confirm Silver 4510 generation; existing Storage Large v4 draft labels this CPU as Emerald Rapids]. The difference is in how the hardware is allocated:
The Medium v4 provides 24 dedicated physical cores (48 threads) across two sockets. The customer has exclusive access to the CPU, including BIOS-level configuration, NUMA topology control, and direct access to hardware accelerators (AVX-512, and AMX if available on the Silver 4510 ). No hypervisor overhead, no noisy-neighbor contention, no CPU steal.
AWS m7i instances provide vCPUs that map to threads on shared physical hosts. The m7i.8xlarge’s 32 vCPUs share the underlying CPU with other tenants. Performance can vary based on co-tenant activity, and there is no BIOS access, no NUMA control, and no ability to tune hardware-level settings. Dedicated Hosts remove the sharing but add significant cost .
Memory: 256 GB vs Cloud Instance RAM
The Medium v4 ships with 256 GB DDR5-4400 ECC across 16 physical DIMM slots. Memory is upgradeable by adding DIMMs to open slots — the server supports up to 4 TB without hardware replacement. ECC is standard.
The m7i.8xlarge provides 128 GB , half the Medium v4’s capacity. Reaching 256 GB requires the m7i.16xlarge (64 vCPUs) , which over-provisions CPU relative to the Medium v4’s 24 cores. On AWS, changing memory requires migrating to a different instance size and restarting the instance. There is no physical DIMM upgrade path.
Storage: Persistent NVMe vs Ephemeral/EBS
This is the most significant structural difference between the Medium v4 and the m7i family.
OpenMetal Medium v4: Local persistent NVMe
The Medium v4 includes 6.4 TB of Micron 7500 MAX NVMe (1.1M random read IOPS, 70 us typical read latency, sub-1ms at 99.9999%) plus dual 960 GB boot SSDs in RAID 1. The NVMe storage is local, persistent across reboots and power cycles, and isolated from boot I/O. No external storage service, no network-attached latency, no per-GB/month charges.
| Metric | Micron 7500 MAX (6400 GB) |
|---|---|
| Sequential Read | 7,000 MB/s |
| Sequential Write | 5,900 MB/s |
| Random Read IOPS | 1,100,000 |
| Random Write IOPS | 400,000 |
| Read Latency (typical) | 70 us |
| QoS | Sub-1ms at 99.9999% (6-nines) |
AWS m7i: EBS only
The m7i family has no local instance storage . All storage is Amazon EBS (Elastic Block Store), which is network-attached:
- EBS gp3: 3,000 baseline IOPS, $0.08/GB/month . 6.4 TB of gp3 costs approximately $512/month in storage charges alone .
- EBS io2 Block Express: Up to 256,000 IOPS , but at $0.065/IOPS/month provisioned . Matching the Micron 7500 MAX’s 1.1M IOPS on EBS is not practically possible or economically viable.
- Latency: EBS adds network hop latency compared to direct-attached NVMe. Typical EBS gp3 latency is single-digit milliseconds vs 70 microseconds for the Micron 7500 MAX.
For workloads where storage performance matters (databases, analytics, blockchain state), the Medium v4’s local NVMe provides an order-of-magnitude improvement in both latency and IOPS over standard EBS volumes.
Networking Comparison
| Dimension | OpenMetal Medium v4 | AWS m7i |
|---|---|---|
| Private bandwidth | 20 Gbps (2x 10 Gbps LACP bonded) | Up to 12.5-25 Gbps (varies by size) |
| Public bandwidth | 2 Gbps (burst up to 40 Gbps) | Shared (varies by instance) |
| Private traffic cost | Included (all VLAN traffic free) | Free within same AZ; $0.01/GB cross-AZ |
| Network SLA | 99.96% base (actual >99.99% since 2022) | 99.99% for multi-AZ |
| DDoS protection | Included (up to 10 Gbps per IP) | Shield Standard included; Advanced $3,000/mo |
Egress pricing: 95th-percentile vs per-GB
OpenMetal’s 95th-percentile billing is a structural cost advantage for sustained egress. Example at two traffic levels:
- 5 TB/month egress: On AWS, 5 TB at $0.09/GB = ~$450/month in transfer charges alone. On OpenMetal, included in the base allocation or billed at 95th-percentile (far lower than per-GB).
- 50 TB/month egress: On AWS, ~$4,500/month in transfer charges. On OpenMetal, 95th-percentile billing for the sustained rate, which is a fraction of the per-GB equivalent.
The cost gap widens linearly with egress volume. For CDN origins, API backends, data distribution services, and streaming platforms, OpenMetal’s egress model delivers significant savings.
Security and Confidential Computing Comparison
| Feature | OpenMetal Medium v4 | AWS m7i |
|---|---|---|
| Physical isolation | Dedicated single-tenant hardware | Shared host (Nitro hypervisor) |
| Intel SGX | Available [VERIFY on Silver 4510] | Nitro Enclaves (different model) |
| Intel TDX | Not confirmed for Medium tier | Not available on m7i |
| Measured boot | Hardware Boot Guard | Nitro Measured Boot |
| IPMI/BIOS access | Full IPMI | Not available |
| Memory encryption | TME-MK [VERIFY on 4510] | Nitro memory encryption |
OpenMetal’s security posture is based on physical isolation: the customer is the only tenant on the hardware, with BIOS-level access and no shared hypervisor. AWS Nitro provides strong virtualized isolation but operates on shared physical infrastructure by default.
HIPAA and Regulatory Compliance Comparison
| Dimension | OpenMetal | AWS |
|---|---|---|
| HIPAA scope | Organizational-level compliance; BAAs available | Per-service eligibility; BAA required per account |
| Physical isolation | Dedicated hardware by default | Shared; dedicated hosts available at premium |
| Facility certifications | SOC 1/2, ISO 27001, PCI-DSS, HIPAA (varies by location, held by facility operator) | SOC 1/2/3, ISO 27001, PCI-DSS, HIPAA (AWS-level) |
| HIPAA-eligible locations | Ashburn, Los Angeles (facility-level HIPAA) | Most US regions |
OpenMetal’s HIPAA compliance is at the organizational level — the entire infrastructure platform is covered. On AWS, HIPAA eligibility is per-service, and customers must configure each service for compliance separately .
When OpenMetal Wins
- Sustained 24/7 workloads: Fixed monthly pricing beats hourly billing for servers that run continuously. No reserved instance planning or commitment gymnastics.
- Egress-heavy workloads: 95th-percentile billing is structurally cheaper than per-GB for CDN origins, API backends, data distribution, and streaming services.
- Storage-performance-sensitive workloads: Local Micron 7500 MAX NVMe (1.1M IOPS, 70 us latency) vastly outperforms EBS gp3 at a fraction of the per-GB cost. Databases, analytics, and blockchain benefit directly.
- Compliance and isolation: Single-tenant dedicated hardware with IPMI and HIPAA eligibility provides a stronger isolation posture than shared-tenancy cloud instances.
- Predictable budgeting: Fixed monthly rate with optional 5-year price lock eliminates cost variability. No surprise egress bills, no reserved instance expiration planning.
When AWS Wins
- Scale-to-zero: Workloads that run intermittently (batch jobs, dev/test, event-driven) pay nothing when idle. OpenMetal’s fixed monthly rate applies regardless of utilization.
- Deep ecosystem integration: Lambda, DynamoDB, SageMaker, Kinesis, and other AWS-native services have no equivalent on bare metal. If your architecture depends on managed AWS services, the integration cost of replacing them exceeds the infrastructure savings.
- Global edge presence: AWS has 30+ regions and 400+ edge locations. OpenMetal operates in 4 regions. For workloads requiring sub-10ms latency to end users worldwide, CloudFront or similar CDN services are difficult to replicate.
- Fully managed operations: AWS handles OS patching, hardware replacement, and scaling automatically. OpenMetal’s base support covers hardware and provisioning, but the customer manages the OS and application layer (or purchases Assisted Management).
- Sub-$10k/month cloud spend: At lower spend levels, the operational overhead of managing dedicated servers may outweigh the cost savings from fixed pricing.
Cost Model and TCO
Cost model comparison
| Dimension | OpenMetal Medium v4 | AWS m7i Family |
|---|---|---|
| Pricing model | Fixed monthly, lock up to 5 years | On-demand hourly, or 1/3-year reserved |
| Egress | 95th-percentile billing; included base; $375/mo per additional 1 Gbps | Per-GB ($0.09/GB first 10 TB) |
| Private traffic | Included (no cost) | Free within same AZ; cross-AZ $0.01/GB |
| IPMI / remote console | Included | Not available |
| Hypervisor licensing | None (bare metal) | Included in instance cost |
| Local storage | 6.4 TB NVMe included | None (EBS required, per-GB/month) |
| Commitment | Monthly with optional 1-5 year price lock | On-demand or 1/3-year reserved |
| Ramp pricing | Available for migrations | Not available |
| DDoS protection | Included (up to 10 Gbps per IP) | Shield Standard included; Advanced $3,000/mo |
| Support | Base included; Assisted Management available | Business/Enterprise plans (extra) |
TCO estimate: 3-node cluster, 12 months
Estimate for a sustained workload running 24/7 for 12 months on a 3-node cluster:
| Cost Component | OpenMetal (3x Medium v4) | AWS (3x m7i.8xlarge Reserved 1yr) |
|---|---|---|
| Compute | Fixed monthly x 12 | ~$0.80/hr x 3 x 8,760 hrs |
| Storage (6.4 TB per node) | Included (NVMe) | EBS gp3: ~$512/mo x 3 ($0.08/GB) |
| Egress (5 TB/mo) | Included in base | ~$450/mo ($0.09/GB) |
| Egress (50 TB/mo) | 95th-pct overage if above base | ~$4,500/mo |
| Support | Base included | Business plan ~$100+/mo |
| DDoS | Included | Shield Standard included |
| IPMI | Included | N/A |
Key insight: The m7i family’s lack of local storage forces all workloads onto EBS, which adds $500+/month per node before accounting for IOPS provisioning. Combined with per-GB egress charges, AWS costs escalate quickly for storage-intensive or egress-heavy workloads.
Organizations spending above $20,000/month on public cloud infrastructure typically see close to 50% cost reduction when moving sustained workloads to OpenMetal. Organizations under $10,000/month may find that the operational overhead of managing dedicated hardware outweighs the cost savings.
Ready to Deploy a Medium v4?
Tell us about your workload and we’ll help you configure the right deployment.
Deployment Options
Bare Metal Dedicated Server
Deploy a Medium v4 as a standalone bare metal server with full root access and IPMI. Fixed monthly pricing with rate locks up to 5 years. Ramp pricing for migrations from AWS or other providers.
Hosted Private Cloud
Deploy a three-node Medium v4 Hosted Private Cloud cluster running OpenStack and Ceph, production-ready in under 45 seconds. Day 2 operations included. No VMware licensing.
Where to deploy
Deploy a Medium v4 in Ashburn, Los Angeles, Amsterdam, or Singapore. All locations offer the same fixed monthly pricing.
| Location | Region | Facility Certifications | Location Page |
|---|---|---|---|
| Ashburn, Virginia | US-East | SOC1/2 Type II, ISO 27001, PCI DSS, NIST 800-53, HIPAA | Ashburn |
| Los Angeles, California | US-West | SOC1/2, ISO 27001, PCI-DSS, HIPAA | Los Angeles |
| Amsterdam, Netherlands | EU-West | SOC Type 1/2, PCI-DSS, ISO 27001, ISO 50001, ISO 22301 | Amsterdam |
| Singapore | Asia | BCA Green Mark Platinum | Singapore |
Facility certifications are held by the facility operator. Proof of Concept clusters are available for testing before committing.
Get a Medium v4 Quote
Tell us about your infrastructure needs and we’ll provide a custom quote.
- Bare metal: Single-server or multi-server deployments with full root access and IPMI
- Hosted Private Cloud: Three-node OpenStack + Ceph clusters with Day 2 operations included
- Custom configurations: RAM upgrades, additional NVMe drives
Ramp pricing available for migrations. All deployments include fixed monthly pricing, 99.96%+ network SLA, and DDoS protection.
Product specifications, pricing, and availability may change due to market conditions and other factors. For the most current information, please contact the OpenMetal team directly.