Wednesday, September 30, 2026

AWS Adds Graphics G7 to WorkSpaces Core Managed Instances: What Is Confirmed?

AWS has announced Graphics G7 support for WorkSpaces Core Managed Instances.

The announcement identifies the hardware, intended workloads and four Regions, but its comparison with G6 does not establish a performance gain for a particular application.

Data Center News & Trends

What did AWS add?

AWS announced support for Graphics G7 instances in Amazon WorkSpaces Core Managed Instances in a notice published September 30, 2026. The instances use NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs and Intel Xeon 6 processors. This adds a managed-instance option for professional graphics work; it does not, by itself, demonstrate how a particular application will perform.

AWS names CAD/CAM, 3D rendering, scientific visualization, video editing and AI-assisted design as potential workloads. It specifies 32 GB of GDDR7 memory per GPU and six instance sizes spanning 1–8 GPUs, 8–192 vCPUs and 32–768 GB of system memory.

GPU memory and system memory are different quantities, and the announcement does not map those family-wide ranges to individual sizes. AWS also lists Linux and Windows support and mentions bring-your-own-license, without detailing eligibility for each operating system.

AWS — Amazon WorkSpaces Core Managed Instances adds support for NVIDIA Blackwell GPU

Where and how does AWS say G7 can be selected?

AWS lists four Regions for the offering: US East (N. Virginia), US East (Ohio), US West (Oregon) and Europe (Spain). It says more Regions will be added but gives no schedule. The list does not guarantee that G7 is currently selectable in every account or partner solution.

According to AWS, a user can specify a G7 instance type when creating a new WorkSpaces Core Managed Instance. Alternatively, a user of a WorkSpaces Core partner solution can look for G7 types when that solution makes them available. The partner route is therefore conditional, not a statement that every partner already supports G7.

What remains unverified about performance?

AWS claims “up to 2.1× better performance” for graphics-intensive workloads compared with previous-generation G6 instances. That is a company claim, not a guaranteed gain across applications: the announcement supplies neither a benchmark setup nor an independently verified result for a reader’s workload.

AWS also describes G7 memory bandwidth as “2.67× faster” without explicitly identifying the comparison baseline, so that figure should not be presented as a confirmed G6 comparison.

The useful distinction is between an available option and a proven workload result. Before choosing G7, check whether it is offered in the relevant Region and access route, whether a suitable size and operating-system or licensing arrangement is available, and whether performance evidence reflects the application you intend to run.

Until then, the announced specifications and AWS’s performance claims answer different questions.

Sources

Amazon RDS for SQL Server Developer Edition Adds Multi-AZ Support: What the Announcement Establishes

AWS announced Multi-AZ support using Always On Availability Groups for specified SQL Server Developer Editions.

The edition and version limits are clear, but regional availability, cost and failover results for a particular deployment need separate checks.

Data Center News & Trends

What did AWS announce?

Amazon RDS for SQL Server now supports Multi-AZ deployments using Always On Availability Groups with SQL Server Developer Edition, according to AWS. Its announcement appeared in the source feed on September 24, 2026; that publication date does not establish a separate feature-launch date.

AWS says the configuration maintains a synchronous standby replica in a different Availability Zone and provides automatic failover during infrastructure failure. That describes supported behavior, not a measured failover time or uptime result for a particular deployment.

AWS — Amazon RDS supports Multi-AZ for SQL Server Developer Edition

Which editions and versions are included?

AWS names these boundaries for Multi-AZ with Always On Availability Groups:

  • SQL Server Developer Edition 2019 and 2022: supported.
  • SQL Server 2025 Enterprise Developer Edition: supported.
  • SQL Server 2025 Standard Developer Edition: not supported.

AWS presents the capability as a way to build and test high-availability configurations, including failover behavior, in non-production environments. It says Developer Edition includes Enterprise Edition functionality without Enterprise Edition licensing costs. That licensing statement does not mean an RDS deployment has no other costs.

What must a team verify before relying on it?

The announcement does not report a deployment-specific failover test or achieved uptime. It also directs readers elsewhere for applicable AWS Regions and RDS for SQL Server pricing rather than providing a region list or a total-cost calculation.

For a proposed non-production test, the useful distinction is between eligibility and results:

  • Confirm the exact edition and major version against AWS’s supported list.
  • Check availability in the intended Region and the applicable RDS pricing.
  • Test the required failover behavior in the environment where the configuration will be used.

Support for a configuration makes that test possible; it does not settle whether a particular deployment meets the team’s failover requirements.

Sources

How to Evaluate a Data Center UPS: Define the Protected Load and Power Path First

A UPS rating does not tell you which equipment is protected.

Use a decision matrix to separate the IT loads that need protection, their actual electrical connections, and the operating conditions behind an efficiency claim. Leave site-specific capacity, runtime and transfer performance unverified until the relevant records are available.

Data center deep-dive series

What must be decided before comparing UPS ratings?

Start with the equipment that needs backup power, then trace each of its power inputs through the proposed electrical distribution system. Only after that boundary is clear can you compare efficiency under the loads the system is expected to carry. The U.S.

Department of Energy’s Best Practices Guide for Energy-Efficient Data Center Design recommends first assessing whether all equipment—or only part of a data center—requires a UPS to reduce UPS-related losses. A rating alone cannot answer that question.

U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

How do you record what is protected—and what is not?

Servers, storage and network equipment are useful categories for the protected-load list, not a rule that every device must be connected to a UPS. DOE notes that the share of IT power needing UPS protection can differ substantially between, for example, a scientific computing facility and a financial institution.

Use “excluded” only where the protection requirement has been established; use “unverified” where the requirement or connection is unknown.

Protected IT load

Record for each proposal
Identify which server, storage and network equipment requires UPS protection, and why.
Exclude or defer when
Exclude equipment only if it has been determined not to require that protection; otherwise leave its status unverified.

Power inputs and path

Record for each proposal
Trace each protected input through its actual distribution connections. Identify the relevant utility service, switchboard or switchgear, alternate source, UPS and PDU, as applicable.
Exclude or defer when
Do not infer that both inputs of a dual-power-supply server are protected. Defer that claim until each input’s connection is confirmed.

Backup arrangement

Record for each proposal
Identify the proposed UPS technology, the generator’s role if one is present, and the required backup duration.
Exclude or defer when
Do not infer a battery system or a particular runtime from the word “UPS.”

Operating load

Record for each proposal
Record initial and future loads, including expected part-load or low-load operation and the redundancy arrangement.
Exclude or defer when
Do not rank proposals by a single rated-load efficiency figure when their operating conditions differ.

Efficiency boundary

Record for each proposal
State the UPS operating mode and whether the figure covers the UPS alone or also includes downstream distribution losses.
Exclude or defer when
Do not treat figures with different equipment boundaries as directly comparable.

Decision evidence

Record for each proposal
Check site load and connection records against equipment specifications and test evidence for capacity, runtime and transfer performance.
Exclude or defer when
Leave missing site-specific values unverified rather than declaring the proposal suitable or unsuitable.

This is a list of components to trace, not a fixed wiring sequence. DOE describes these as elements of a typical distribution system and explains that a PDU can distribute power received from a UPS or generator to multiple devices. UPS backup technology may use batteries, rotary machines, fuel cells or other approaches.

IBM’s Data centers overview also describes generators as a possible response to more severe outages. Neither description establishes how a particular site is connected.

IBM — What Is a Data Center? | IBM

When are two efficiency figures comparable?

Compare them at a defined protected load, expected part-load condition, redundancy state and operating mode. DOE advises considering initial and future loads—particularly part-load and low-load conditions—when selecting electrical equipment. Redundancy can change the load carried by each UPS even when the protected load stays the same.

In a DOE teaching example of battery-based N+1 systems, two UPS units each operate at 30% load factor; serving the same load with three smaller units puts each at 40%. The example illustrates a possible efficiency difference under those conditions, not a guarantee that three units are better at every site.

Mode and equipment boundary matter too. DOE describes a more efficient line-conditioning mode available on some double-conversion UPS systems; it should not be assumed to exist on every product or to provide the same power-conditioning conditions as double-conversion operation.

Power is also lost within the UPS, while a PDU with a built-in transformer can lose energy as heat during voltage conversion. Transformer load-factor guidance applies to configurations that actually contain that transformer, not to every PDU.

Keep facility PUE separate from UPS efficiency. In DOE’s definition, site-power PUE is the facility’s total annual energy divided by the annual energy drawn by all IT equipment. It helps characterize supporting infrastructure; it does not report the efficiency of one UPS or the useful work performed by IT equipment.

A proposal comparison should therefore identify both the operating conditions and which losses its efficiency figure includes.

What evidence is needed for a final choice?

Plan how a claimed efficiency could be checked after installation. DOE describes more detailed monitoring of losses along the electrical path, including transformers, UPS equipment and PDUs with transformers. It advises specifying meter accuracy and calibration status, trending measurements, and retaining data long enough to obtain annual energy totals.

The measurement points must match the equipment boundary used in the comparison.

Efficiency cannot substitute for protection. DOE notes that reliability and the ability to accommodate high power density can take priority over energy efficiency for critical data center loads.

The useful decision order is to establish the protected load, verify each input’s actual path, compare losses under equivalent operating conditions, and leave unsupported values open. Without site load and wiring records and the proposed equipment’s specifications and test evidence, capacity, backup duration and transfer performance remain questions to resolve—not conclusions supplied by a general design guide.

Sources

Related reading

AWS DRS Adds Support for Graviton Source Servers: What the Announcement Establishes

AWS says Elastic Disaster Recovery now supports Graviton-based arm64 source servers.

The announcement describes the recovery path and where the capability is offered, but it does not establish a recovery result for any particular workload.

Data Center News & Trends

What changed, and what does the date mean?

AWS says Elastic Disaster Recovery (DRS) now supports disaster recovery for Graviton-based arm64 source servers. That gives teams planning recovery for Graviton workloads a support path to investigate. The distinction is between the capability AWS has announced and a recovery outcome a team has verified for its own application.

September 25, 2026 is the publication date in the AWS What's New feed. The announcement does not give a separate date when the capability became active, so the feed date should not be treated as an activation date.

AWS — AWS Elastic Disaster Recovery now supports AWS Graviton-based source servers

How does AWS describe the arm64 recovery path?

According to AWS, DRS automatically detects arm64 source servers and recovers them onto Graviton instances, preserving the workload architecture. AWS says recovery works as it does for other servers, with nothing extra to configure. That describes the announced workflow; it does not show that every application will recover successfully.

AWS says the capability is available in all AWS Regions where DRS is offered, at no additional cost. This is a statement about the added capability, not a claim that DRS itself is free or available in every AWS Region.

What remains to be verified for a recovery plan?

The announcement includes no workload-specific recovery test results or measured recovery times. “Supported by DRS” is therefore different from “verified for this application”: the announcement cannot tell a team whether its workload will recover successfully or how long recovery will take.

Before relying on the capability in a plan, check the conditions that apply to the configuration in the AWS Elastic Disaster Recovery User Guide and test recovery of the workload concerned. The announced support makes that assessment relevant; the workload test is what can establish a result for the plan.

Sources

EC2 R8i and R8i-flex in Germany: What AWS Announced—and What Remains to Be Checked

AWS announced that EC2 R8i and R8i-flex are available in the AWS European Sovereign Cloud (Germany) region.

The change is regional availability, not a demonstrated performance or cost result for every deployment. Here is how the two families differ in AWS’s description, and what to verify before choosing one.

Data Center News & Trends

What changed in the announcement?

In its notice published September 25, 2026, AWS announced that Amazon EC2 R8i and R8i-flex instances are available in the AWS European Sovereign Cloud (Germany) region. The notice’s publication date is known; a separate rollout date has not been verified.

For a reader considering memory-intensive workloads in that region, the announcement puts both families on the list of options to examine. It does not establish that a needed size can be launched in a particular account, or that a workload will achieve a particular performance or cost result.

AWS — Amazon EC2 R8i and R8i-flex instances are now available in additional regions

How does AWS distinguish R8i-flex from R8i?

AWS describes the families in terms of resource use and instance size:

  • R8i-flex is a memory-optimized Flex family that AWS positions for applications that do not fully use all compute resources. Its stated family size range runs from large through 16xlarge.
  • R8i is positioned for memory-intensive workloads, particularly those needing the largest sizes or sustained high CPU use. AWS describes 13 family sizes, including two bare-metal sizes and 96xlarge.

These descriptions help narrow the choice, but a family’s stated size range is not confirmation that every size is launchable in the intended account in Germany.

Do AWS’s comparison figures predict a deployment’s results?

No. AWS claims that R8i and R8i-flex offer up to 15% better price-performance and 2.5 times the memory bandwidth compared with previous-generation Intel-based instances. It also claims 20% higher performance than R7i and, compared with R7i for named workloads, up to 30% faster PostgreSQL databases, 60% faster NGINX web applications and 40% faster AI deep-learning recommendation models.

Those are AWS’s comparisons, not independently verified results for a deployment in the newly announced region. In particular, an “up to” figure for a named workload should not be treated as the expected gain for a different application—or as a measured cost saving.

What should a deployment decision depend on?

AWS lists Savings Plans, On-Demand and Spot as purchase options, but that list does not give the applicable price or effective cost of a particular deployment. The useful distinction is between a family being announced for a region and the conditions under which it meets a specific deployment’s needs. Check:

  • Whether the required R8i or R8i-flex sizes are available to the intended account in the Germany region.
  • Which purchase terms and prices apply to the proposed deployment.
  • Whether results from the actual workload meet its performance requirements.

If any of those answers is unfavorable or still unknown, the regional announcement alone is not a sufficient basis for choosing the instance family.

Sources

How to Map a Data Center Value Chain: A Hypothetical Expansion

Start a data center value-chain map with what the expansion must deliver, not a list of companies.

This hypothetical example separates IT equipment, facility systems and network connections from the work of designing, installing and operating them—and leaves unverified contracts blank.

Data center fundamentals series

What does the hypothetical expansion need to deliver?

A useful value-chain map starts with outcomes: run an application, store its data and connect it to users. It then separates the IT equipment and facility systems that make those outcomes possible from the work needed to design, install and operate them. That classification does not identify a supplier or contract partner.

Suppose an organization wants to add computing and storage for a new application in a server room it already operates, while securing a connection to an external network. This is a hypothetical request, not a real project.

The first questions concern the application and its storage needs, the endpoints and distances of its connections, and initial and future loads. Without those answers, server counts, electrical capacity and link speeds remain unknown.

The Fiber Optic Association — The FOA Reference For Fiber Optics - Fiber Optic Network Design

U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

IBM — What Is a Data Center? | IBM

What belongs in each row of the map?

The table is a worked classification of that request, not a standard purchasing sequence. “Unconfirmed” means that the example supplies neither an assigned party nor a document establishing its responsibility.

Run the application and retain data

Related elements
Servers, storage and internal network
Design, installation or operating work
Choose a configuration; establish deployment and management responsibilities
Party and evidence to confirm
IT design and operating parties; scope of work
Still unknown
Equipment counts, storage configuration and capacity

Supply power

Related elements
Incoming supply and distribution; backup source, UPS or PDU where needed
Design, installation or operating work
Check whether existing infrastructure can serve the added load and which equipment needs UPS protection
Party and evidence to confirm
Electrical design and installation parties; design documents
Still unknown
Load, capacity and protection scope

Remove added heat

Related elements
Airflow and cooling systems
Design, installation or operating work
Assess initial and future loads and the suitability of existing airflow
Party and evidence to confirm
Facility design and operating parties; design documents
Still unknown
Heat load, cooling capacity and method

Connect systems and users

Related elements
Internal network and building entrance to an external network
Design, installation or operating work
Define endpoints, distance, speed and transmission equipment; coordinate the connection
Party and evidence to confirm
Network designer and external connection provider; scope of work
Still unknown
Link count, speed, cable type and interfaces

Complete and accept installation

Related elements
Cable routes and terminations if fiber installation is in scope
Design, installation or operating work
Record the route and connections; plan tests and capture acceptance results
Party and evidence to confirm
Installer; test plan and acceptance documents
Still unknown
Installation scope and link-specific tests

Keep the service operating

Related elements
IT and facility systems; connection records where applicable
Design, installation or operating work
Assign maintenance, monitoring, change and incident responsibilities
Party and evidence to confirm
Operating parties; contract and operating records
Still unknown
Division of continuing responsibilities

The unresolved column matters as much as the equipment column: it prevents a plausible component list from being mistaken for an approved design or an awarded contract.

Which equipment choices does the request leave open?

For IT, distinguish servers that run the application, storage for its data and the network that connects systems and users. Storage need not mean drives attached to servers. Direct-attached storage (DAS), network-attached storage (NAS) and a storage area network (SAN) are different configurations that can coexist; the request does not select one.

For facility power, keep the supply and distribution path separate from the IT equipment it serves. Backup sources, uninterruptible power supplies (UPSs) and power distribution units (PDUs) may be relevant, but the design must establish which loads need UPS protection and whether the existing path has capacity.

Cooling likewise requires an assessment of added heat and airflow, not an automatic choice of technology. Direct liquid cooling transfers heat to fluid near its source; some liquid-cooled approaches still need air cooling for remaining heat. Nothing in this example establishes that liquid cooling will be installed.

Finally, distinguish connections among servers and storage inside the facility from the connection through the building entrance toward an external network. As the Fiber Optic Association’s design guidance explains, communications requirements, distance and transmission equipment inform cable choices. Without defined endpoints and requirements, neither a fiber type nor an external connection provider can be assigned here.

Why map design, acceptance and operations separately?

An equipment list does not say who will turn requirements into an installable design. If a fiber connection is part of the project, its design may involve choosing a route and transmission equipment, coordinating with facility and electrical personnel, and checking whether permissions or inspections are needed. Those are tasks to allocate, not evidence that one firm performs them all.

Nor is delivery of components the same as acceptance of an installed link. For fiber work within scope, route and termination records, a test plan agreed during design, and results incorporated into acceptance documentation form a distinct deliverable. The particular tests depend on the link and agreed scope.

After acceptance, maintenance, monitoring, changes and incident response still need owners; fiber records and restoration plans may also need upkeep. Do not assume that the installer becomes the operator.

Can the map identify the companies that benefit?

Not from this example alone. It begins with a self-operated server room, but a different service model would change what must be assigned. Colocation provides space, power, cooling and network connectivity for customer-owned IT; whether management services are included needs checking.

In a managed arrangement, a provider can administer dedicated leased servers, storage and networking equipment. Cloud services instead provide access to shared IT resources over the internet. These models do not imply identical ownership, deployment or operating responsibilities.

Knowing what a UPS, storage system or network link does is not evidence of who supplied it, who contracted for it or whose revenue it generated. When applying the map to an actual project, use design documents, the scope and contract, and the chosen ownership model to fill in the responsible parties and acceptance and operating duties.

Until then, blank company fields are useful: they show precisely which claims still need verification.

The Fiber Optic Association — The FOA Reference For Fiber Optics - Data Centers -

Sources

Related reading

AWS’s Aurora Serverless Scaling Announcement: Up to 16 ACUs Added in a Second

AWS says Aurora Serverless can add capacity in larger steps, but the announcement does not measure the effect on a particular application.

Here is what the capacity figures mean, which platform versions AWS says are covered, and what remains to be checked.

Data Center News & Trends

What does the one-second scaling claim mean?

AWS says Aurora Serverless can add up to 16 ACUs to its current capacity within a second. The 16 ACUs are the maximum addition in that interval, not the resulting total capacity. AWS separately says capacity can continue rising to as much as 256 ACUs as workload demand grows; it does not say the service reaches 256 ACUs in one second.

The AWS feed published the notice on September 30, 2026. The announcement text does not give a separate date on which the change occurred.

AWS — Amazon Aurora serverless now scales faster to support agentic AI and other bursty workloads

Which clusters does AWS say are covered?

AWS says the enhancement is enabled by default, without configuration changes, on Aurora Serverless clusters running platform version 3 or 4. For existing clusters on versions 1 or 2, AWS describes a direct upgrade path to version 4; it does not say the enhancement is already enabled on those older versions.

To check a cluster’s platform version, AWS points to the instance configuration section of the AWS Management Console or the RDS API’s ServerlessV2PlatformVersion parameter. That version check establishes whether a cluster falls within the stated scope; it does not measure an application benefit.

What results does the announcement leave unverified?

AWS says Aurora Serverless scales down to zero when a workload finishes and describes it as suited to bursts separated by long idle periods. That description does not establish cost savings for a particular cluster. The notice also gives no application-specific before-and-after latency, throughput or cost results.

For prices and Region availability, it directs readers to Amazon Aurora Pricing rather than listing those details.

For a team running a bursty service, the useful next distinction is between eligibility and effect: check the cluster’s platform version to understand whether AWS says the change applies, then use that service’s workload and operating records to assess any effect on latency, throughput or cost. A faster capacity increase alone cannot establish those outcomes.

Sources