AmpereOne is the most affordable server ARM. Better than AMD and Intel?

Ampere's success story began with the Altra series. These processors, equipped with Neoverse N1 and N2 cores operating at up to 3 GHz, showed impressive results. However, the company did not stop there. The new series AmpereOne became a logical and powerful continuation of the already proven Ampere Altra, offering significant improvements and innovations.

AmpereOne: A New Player in the Server Processor Market

AmpereOne is positioned as a direct and worthy competitor to industry giants like Intel's 6th-gen Xeon and AMD's EPYC 9004 (and possibly the upcoming Turin-based 9005). And guess what? They can really give those ossified x86 dinosaurs a run for their money.
The processor is manufactured using the advanced 5-nanometer process technology, which allows achieving an optimal balance between performance and energy efficiency.
One important thing, starting with AmpereOne, is that the core architecture has changed. If Ampere previously used licensed Arm Neoverse cores, now, under the Arm license as an architecture, they have developed their own core design called Siryn.

Since Ampere Computing places a special emphasis on the fact that their processors are extremely efficient in AI-related tasks, it is worth noting that their architecture includes vector units, in the case of AmpereOne – 2 blocks of 128 bits, they are used for tasks that are well parallelized across multiple threads, as a rule, these are operations that involve linear algebra (FP16, Bfloat16, Int8, Int16). In more understandable terms, this helps well in tasks related to AI and neural networks.

AmpereOne supports DDR5 memory with the ability to scale up to 12 channels. This provides high memory bandwidth and low latency, which is critical for many server applications. In addition, the processor is equipped with a PCIe Gen5 interface, which offers significantly increased bandwidth compared to previous generations.

But the main thing that makes AmpereOne good is the number of cores. AmpereOne is available in a version with 192 computing cores. Yes, you heard right — one hundred and ninety-two cores in one processor. And that's not the limit: the company plans to expand the CPU line to 256 cores in the future. For comparison, the top-end AMD EPYC 9754 offers 128 cores, and the Intel Xeon 6780E — 144 cores.

New LGA5964 Socket – A Step Towards Unification and Flexibility

New socket LGA5964developed for the updated AmpereOne series, is a logical continuation of the previous LGA4926. Although the basic principles remain the same, the new socket offers a number of improvements. First of all, it ensures compatibility with more powerful AmpereOne processors, while maintaining continuity in terms of ease of maintenance and upgrades. As before, users can easily upgrade their server systems by changing processors without replacing the entire motherboard. This not only maintains the flexibility of AmpereOne systems, but also protects customer investments, ensuring a long-term perspective for developing their infrastructure in line with growing business needs.

The LGA5964 socket expands scalability. While the compatibility principle remains the same as the previous generation, the new socket supports a wider range of processors, including future models with increased performance. This allows companies to start with basic configurations and gradually increase capacity, adapting to growing business needs without completely replacing the equipment.

The unified socket also simplifies the process of managing a fleet of server equipment. IT administrators can more effectively plan updates and allocate resources, dealing with a single platform. And this, believe me, can save a lot of gray hair for system administrators and DevOps engineers.

Energy efficiency is a key advantage of AmpereOne

Now let’s talk about what really sets AmpereOne apart from the competition: its exceptional power efficiency. Ampere Computing claims that AmpereOne processors are 50% more efficient than EPYC Genoa in terms of performance per watt. Yes, you read that right — by as much as 50%!

When compared to AMD's flagship EPYC Bergamo or Genoa, Ampere's new product offers up to 34% better utilization of compute power per rack. And what's especially important is that these impressive results are achieved not by increasing power consumption, but by a more efficient architecture and optimization for specific cloud computing tasks.

What does this mean in practice? This is a significant reduction in electricity costs. For large data centers, this can result in millions of dollars in savings annually. Less heat generation allows you to reduce the cost of cooling systems for data centers, which cannot operate without air conditioning systems. And given that cooling can account for up to 40% of the total energy consumption of a data center, this is a very significant savings. Separately, it is interesting that AmpereOne processors do not have a heat spreader cover, which is more typical for GPUs, and also simplifies cooling the processor crystal.

More efficient use of energy allows for higher server density. The fewer servers required to perform a task, the higher their density. This is especially important in the conditions of limited data center space, especially in large cities, where every square meter is worth its weight in gold.

Collaboration with Oracle and Qualcomm

An important aspect of AmpereOne's success is Ampere's strategic partnerships with major players in the technology market. Of particular note here is Ampere's close ties with Oracleone of the leaders in the field of cloud technologies and databases.

Oracle is not just showing interest in Ampere solutions — the company is actively implementing AmpereOne processors into its infrastructure. And this is a serious confirmation of the efficiency and reliability of these processors in real-world operating conditions. The use of AmpereOne in Oracle cloud services demonstrates that these processors are capable of handling the high loads and complex tasks typical of modern cloud applications.

The partnership with Oracle also gives Ampere access to a huge enterprise customer base. This could significantly accelerate the adoption of ARM processors in the enterprise segment. And this, in turn, could become a catalyst for further development of the ARM software ecosystem in the server segment.

No less interesting is the strategic alliance between Ampere and QualcommThis collaboration opens up new prospects for the development of energy-efficient server solutions for AI-related tasks.
Tests showedthat the 128-core Ampere Altra is as powerful as a Xeon-based server with an NVIDIA Tesla A10 accelerator when running the large LLaMA 3 language model. At the same time, the Ampere solution consumes only a third of the energy required by its competitor. Moreover, a ready-made server with Ampere Altra Max costs about €4,300, while the NVIDIA A10 alone will cost about €3,500. Thus, the Ampere solution is not only more energy efficient, but also allows you to save up to 30% of your budget with comparable performance.

Accessibility for a wider audience: democratizing server technology

Now let's talk about what really sets Ampere apart from its competitors – the accessibility of their solutions to a wide audience. Unlike traditionally expensive server solutions from Intel and AMD, Ampere strives to make its technologies accessible not only to large corporations with bottomless pockets, but also to small and medium businesses, as well as to PC users who have decided to move away from the usual x86.

But the most interesting thing is how far progress has gone in this direction. Now enthusiasts and small companies can buy a motherboard of the format MATX and assemble a server system unit based on AmpereOne. Yes, you heard right – a real ARM server in a desktop case that will easily fit under your desk. This opens up completely new possibilities for experimentation and innovation.

Ampere opens up access to high-performance computing for companies of all sizes. The more processor cores you have, the more services, microservices, virtual machines, and containers you can run. And the lower the price per core and overall performance, the more servers you can afford for your data center. Compare for yourself: 128 Ampere cores will cost a startup or a small company only ~4,300€ for a ready-made server. But 128 Intel Xeon 6756E cores cost ~7,600€ for the processor alone. And that's not counting the traditionally high prices for Intel motherboards. Plus savings on electricity: 190 W for Ampere One versus 225 W for Intel. The choice is obvious, isn't it?

By the way, pay attention to the maximum frequency of Ampere One, unfortunately, the base frequency and specifically for the 128-core model is not indicated even on the Ampere Computing website. Not to mention the hell that is happening with their TDP table. But so far it looks like the CPUs from Ampere One are good not only for multi-threading, but also for performance per core, which means 1C should hurry to port their products natively to ARM, with such frequencies everything will definitely fly.

The availability of AmpereOne helps to spread the ARM architecture in the server segment. In the long term, this may lead to more diversity in the market and stimulate competition. And competition, as we know, is the engine of progress. Perhaps this will force Intel and AMD to finally shake themselves up and start offering something truly innovative, and not another “revolutionary” processor with a 5% performance gain and a 20% increase in power consumption.

Imagine a small DevOps team that can deploy a full-fledged cluster of ARM servers right in the office for testing and development. Or a startup that can afford a powerful computing infrastructure without having to rent a rack in a data center.

Can AmpereOne really compete with x86?

Now let's talk about the most interesting part – performance. Can an ARM processor really compete with x86 in the server segment? Spoiler: yes, maybe, and how.

According to SPECint_base 2017 benchmarks, AmpereOne processors demonstrate impressive results. In some scenarios, they demonstrate performance on par with or even better than the top AMD EPYC and Intel Xeon models. And this is with significantly lower power consumption.

But where things get interesting is when we look at the specific workloads that are typical for cloud computing. Here, AmpereOne often stands head and shoulders above the competition. This is especially true for tasks involving big data, machine learning, and containerization.

For example, in tests using software used in storage systems and databases, AmpereOne shows performance on par with or even better than x86 processors, with significantly lower power consumption.

Of course, there are certain scenarios where x86 processors still have an advantage. For example, in some highly loaded single-threaded applications, Intel and AMD can show better results. But let's be honest – how many such applications are there in modern cloud environments? Most modern workloads scale well to many cores, and here AmpereOne feels like a fish in water.

Is everything ready for a mass transition to ARM?

One of the main concerns when considering the move to ARM architecture has always been software compatibility. And here, Ampere has a few tricks up its sleeve.

In recent years, we have seen an impressive growth in support for the ARM architecture in server software. Linux, which dominates the server segment, has long been a great performer on ARM processors. Windows Server 2025 has joined the party, finally adding support for ARM, albeit with some caveats, as is usually the case with Microsoft. In terms of tools for developing, deploying and managing applications, giants such as Docker, Kubernetes and Jenkins are fully compatible with ARM. In the virtualization field, VMware, with its entire product line, has supported ARM for quite some time, as has Proxmox. In fact, most of the open source software available for Linux traditionally works well on the ARM architecture.
Looking at this picture, one can't help but wonder: doesn't this all look like a large-scale industry preparation for the transition to ARM? After all, almost all key server software developers, from giants like Microsoft to small open-source projects, are actively investing in ARM support. It seems that the industry is seriously preparing for significant changes in the architecture of server systems.

Cloud providers like AWS, Azure, and of course Oracle are actively promoting ARM solutions. This means that more and more developers are optimizing their software to run on ARM processors.
Virtualization and containerization technologies greatly simplify the process of migrating applications between different architectures. Many applications can be simply rebuilt for ARM without the need to change the source code.

Conclusion

So, what do we have in the end? AmpereOne is not just another server processor, but a potential turning point in the development of server technologies. The combination of high performance, exceptional energy efficiency and affordability makes these processors an attractive choice for a wide range of applications – from large cloud data centers to compact server solutions for small businesses.

Of course, Ampere still faces some challenges. It needs to continue to optimize performance, expand the ecosystem of compatible software, and convince conservative corporate customers of the benefits of ARM architecture. But judging by what we see now, the company has every chance of success and a significant share of the server market.

If your company is already using Ampere servers, we would be very interested to hear about your impressions and any challenges you may have encountered. Your experience can be very valuable for those who are just considering the transition to ARM architecture in the server segment. What benefits have you noticed? What challenges have you encountered during the migration? How have ARM servers performed in real-world workloads? We look forward to hearing your thoughts in the comments!