Google Cloud is thrilled to announce that the C4 machine series, the most powerful general-purpose virtual machine (VM) for users of Compute Engine and Google Kubernetes Engine (GKE), is now generally available. With up to 20% higher price-performance for general-purpose workloads and 45% better price-performance for CPU-based inference compared to equivalent generally available virtual machines (VMs) from other top hyperscalers, C4 VMs are designed from the bottom up and fine-tuned to achieve industry-leading performance. Additionally, C4 offers granular management and enterprise-grade reliability to serve customers’ most important business tasks.
Backed by Google’s Titanium, C4 virtual machines (VMs) offer the performance and adaptability required to manage most workloads in conjunction with the N4 machine series. C4 offers scalable storage with up to 500k IOPS and 10 GB/s throughput on Hyperdisk Extreme, as well as high performance connection with up to 200 Gbps of networking bandwidth thanks to Titanium offload technology. Featuring the newest generation of performance with Intel’s 5th generation Xeon processors (code-named Emerald Rapids) offering predetermined forms in high-cpu, standard, and high-mem configurations, C4 instances scale up to 192 vCPUs and 1.5TB of DDR5 memory.
C4 virtual machines (VMs) offer the performance, capabilities, and reliability required to satisfy your business demands, whether you’re managing large amounts of data processing and analytics, operating CPU-based inference, supporting video streaming or game servers with significant traffic, or running strong databases.
Performance that leads the industry
With up to 20% higher price-performance on the projected SPECrate 2017_int_base benchmark, C4 VMs outperform equivalent GA offerings from other cloud providers across a wide range of workloads, allowing you to optimise your demanding workloads and make the most of your cloud resources.
With Google Cloud’s high-performance Hyperdisk storage and Titanium networking, C4 virtual machines (VMs) offer the newest advancements in database workloads. They can perform up to two times as well on MySQL transactions per minute and twice as well on throughput-focused Redis workloads, allowing you to generate significantly more queries per second with each VM. When it comes to data analytics, C4’s speed enhancements result in a quicker time to insights, which makes it possible for you to handle and examine your data more effectively. The improvements in computation and network latency for gaming and streaming result in less buffering, smoother gameplay, and an overall better user experience.
Control, predictability, and responsiveness
Notable gains in consistency, predictability, and responsiveness are made with C4. Because of this, C4 is a perfect fit for applications where milliseconds count, such as high-frequency trading, gaming, and mission-critical corporate systems that must operate reliably and consistently even in the face of latency.
It provides enhanced infrastructure performance, reliability, security, and lifecycle management with Titanium. By shifting storage to the Titanium adapter, you free up host resources to handle your workloads. Additionally, the great majority of infrastructure maintenance updates are carried out with almost no impact on your workloads thanks to the use of hitless upgrades and live migration, reducing interruptions and ensuring predictable performance. In comparison to earlier generations, C4 offers up to 80% improved CPU responsiveness for real-time applications, enabling speedier transactions and more fluid gaming and streaming.
In order to speed up single-threaded workloads like financial and Monte Carlo simulations, as well as machine learning activities like recommendation systems, image classification, and SLM training, tuning, and inferencing, the C4 boasts the fastest single-core max turbo frequency of any previous generation virtual machine, reaching up to 4.0 GHz. Additionally, these VMs are the first to implement the new Turbo API frequency control capability, which is presently in preview for Compute Engine.
With the Turbo API, you can adjust the frequency boost to your liking, providing you with precise control over performance and the ability to strike a balance between consistency and peak performance. Stable performance can be prioritized by capping at all-core turbo frequency (3.1 GHz) for web servers and databases, while workloads like batch processing or scientific simulations can benefit from being able to boost to the greatest frequency.
Boosting CPU-based Interpretation
Comparing C4 VMs to earlier generations, there has been a significant improvement in machine learning inference performance and cost-efficiency. Advanced Matrix Extensions (AMX), the newest integrated accelerator from Intel, is supported by C4. This allows for increased throughput, which dramatically accelerates CPU-based machine learning inference. Regarding recommendation inference, C4 virtual machines exhibit a price-performance uplift of up to 40% when compared to C3, and a price-performance advantage of up to 45% when compared to the closest competitor offering. C4 VMs provide a familiar operating environment and give more deployment flexibility across task types than GPUs.
Customers will be able to use C4 VMs to optimise their small-to-medium machine learning installations without sacrificing efficiency thanks to these more potent and reasonably priced inference capabilities.
The newest Hyperdisk storage technology, now offered in Hyperdisk Balanced and Hyperdisk Extreme variants, is standard with next-generation storage with Hyperdisk C4 virtual machines. For a variety of workloads, Hyperdisk Balanced offers an optimal combination of cost-effectiveness and performance. By offering extremely low latency and high throughput for even the most demanding workloads, Hyperdisk Extreme pushes the envelope in storage performance. Titanium allows C4 VMs to fully utilise Hyperdisk, enabling scalable, high-performance I/O on Hyperdisk Extreme up to 500k IOPS and 10 GB/s throughput.
The advantages of Hyperdisk go beyond its performance. You may dynamically adjust IOPS and bandwidth in real time using Hyperdisk to ensure that your workloads have the resources they require. Optimising utilisation and cost-efficiency, hyperdisk storage pools (for hyperdisk balanced volumes) allow for flexible administration and distribution of storage resources.
In comparison to conventional storage options, Hyperdisk not only offers better performance and flexibility but also dramatically lowers your total cost of ownership (TCO). Your C4 virtual machines will benefit from an unmatched storage experience thanks to the marriage of Titanium’s storage acceleration and Hyperdisk’s cutting-edge capabilities.
Energizing the future collectively
Making the complex but necessary choice of virtual machine is essential to your success. When combined, the C4 and the previously disclosed N4 VMs handle the entirety of your general-purpose workloads, enabling you to reduce expenses without sacrificing workload or performance requirements. C4 VMs are the best option for applications requiring superior performance, exact control and predictability, or sophisticated functionality. For everything else, N4 offers outstanding value and a superb mix of versatility, cost-effectiveness, and performance.
Start using C4 virtual machines now
Us-central1 (Iowa), Us-east1 (South Carolina), Us-east4 (North Virginia), Europe-west4 (Netherlands), Europe-west1 (Belgium), and Asia-southeast1 (Singapore) are currently offering C4 virtual machines. Google Cloud cordially invite you to discover the unmatched power, dependability, and control of C4 and to witness the explosive future of cloud computing.
When generating a new virtual machine or GKE node pool in the Cloud console, choose the C4 machine family under General Purpose machine family to begin using C4 instances. Visit Google Cloud’s regions and zones page to stay informed about current regional availability, or get in touch with a Google Cloud sales person for more details.
0 Comments