Table of Contents
Introduction
Desktop PC hardware has undergone massive growth in recent years, becoming more complex in designs, capabilities, and use cases. The latest products deliver faster processing, advanced connectivity, integrated hardware optimized for AI, and other advancements that improve workflows and processes. These advancements can be difficult for users unfamiliar with the latest developments and trends. Current products may differ from previous purchases, which can create information asymmetry between modern hardware and their specific needs. A basic understanding of current products and their use cases can help mitigate the risk of selecting incompatible, inefficient, or unnecessarily expensive components for their system and processes.
This article provides an overview of the current state of desktop CPUs, RAM, storage, and GPUs for those looking to stay current with the latest PC and workstation hardware. There are other components that go into a workstation, including motherboards, power supplies, and peripheral devices like network interface cards, capture cards, and accelerator cards. However, individuals often choose these based on reliability and specific technical requirements tied to their workflows, making them difficult to cover comprehensively in an article like this.
We included a table of contents to help with navigation, as this article is longer than our usual publications. It’s also worth noting that this article is an overview of the latest hardware in each category and does not provide benchmark performance analysis or personalized hardware recommendations for specific workflows. If you’re looking for more in-depth information about specific components, feel free to explore our articles, blogs, podcasts, and workflow-based solutions.
Processor (CPU)
The central processing unit (CPU), or processor, processes data, manages tasks, and executes instructions. Its architecture and design determine how efficiently it handles workloads that influence overall system performance and responsiveness. Most professional workflows need fast processing, and different classes of CPUs, specifically consumer, workstation, and server-class CPUs, are designed with unique architectures, configurations, and features that impact performance, scalability, security, and price. These differences help users understand what each class of CPU offers so they can select a processor that aligns with their specific workload and processing requirements.
Intel and AMD are the primary CPU manufacturers for desktops, workstations, and servers. It’s also worth noting that Qualcomm has made advancements with its ARM-based processors, particularly in its Snapdragon lineup. While these processors are primarily used in mobile and low-power computing, Qualcomm appears to be exploring their potential in the PC desktop hardware market.
Intel’s latest lineup includes the Core™ Ultra Series for consumer systems, Xeon® W for workstations, and Xeon® Scalable for servers. AMD offers Ryzen™ for consumer systems, Ryzen™ Threadripper™ and Threadripper™ PRO for workstations, and EPYC™ for server environments.
Below is an overview of the latest consumer and workstation CPUs, providing context on products currently available.
Consumer CPUs
Consumer-class CPUs offer a range of capabilities, from entry-level processors for basic computing tasks to high-performance models for more demanding workloads. They typically balance performance, power efficiency, and cost, making them suitable for systems where affordability and versatility are priorities. Many include integrated graphics (iGPUs), handling basic graphical tasks and displays without requiring a dedicated GPU or providing specialized media decoding/encoding capabilities. Additionally, consumer CPUs often feature higher clock speeds for improved performance in lightly threaded applications. Despite the common term used to classify these CPUs, they are also extremely common in workstations and can be a terrific option depending on the exact workflow.
Currently, the Intel Core Ultra 200 Series and AMD Ryzen 9000 Series are the latest consumer CPU product lines on the market. Each has its own strengths and weaknesses, and neither is better across the board. The performance of this class of CPU has increased greatly in recent years, to the point that the distinction between top-tier consumer-class CPUs and entry-level workstation-class CPUs is not always clear.
Below, we compare the specifications of these CPUs to highlight their key features and differences. While the table provides a helpful overview, it does not cover every detail. Visit Intel Ark for a detailed comparison of Core Ultra 200 Series CPUs. Similarly, AMD’s site provides a complete product stack and specifications for the Ryzen 9000 Series CPUs.
Feature | Intel Core Ultra 200 Series | AMD Ryzen 9000 Series |
---|---|---|
Architecture | Arrow Lake Lion Cove / Skymont microarchitecture | Zen 5 |
Socket | LGA 1851 | AM5 |
CPU Core Counts | 6 – 8 P Cores & 8 – 16 E Cores | 6 – 16 Cores |
Max Memory Support (RAM) * | Up to 192 GB DDR5 6400 MT/s (2 memory channels) | Up to 192 GB DDR5 5600 MT/s (2 memory channels) |
PCIe Lanes | 24 (total) | 24 (usable) |
Integrated Graphics | Intel Arc Xe-LPG (4 Xe cores) | Radeon RDNA 2 (2 cores) |
AI Acceleration | Intel AI Boost NPU | N/A |
L3 Cache | Up to 36MB | 32 up to 128MB (X3D variant) |
MSRP | $260 – $600 | $279 – $749 |
* Maximum memory support, as listed by Intel and AMD, reflects the theoretical capability of the platform. Achievable configurations may vary based on motherboard design, DIMM compatibility, and memory density availability.
AMD Ryzen 9000 Series
The AMD Ryzen 9000 Series is built on the Zen 5 architecture, making it AMD’s latest consumer-based CPU family, with support for DDR5 memory and PCIe 5.0 for faster communication with GPUs, NVMe storage, and other connected devices. They include integrated graphics, offering hardware-accelerated decoding for modern codecs like AV1, VP9, H.264, and H.265 (HEVC).
The Ryzen 9000 Series offers a variety of SKU families, including “X”, “non-X”, and “X3D”.
- “X3D” SKUs: These variants feature 3D V-Cache™ technology, which significantly increases the L3 cache by stacking additional cache layers directly onto the processor die. This design enhances memory capacity and efficiency, particularly for workloads that benefit from larger cache sizes. AMD recently announced additional X3D products, bringing higher core-count models into this series.
- “X” SKUs: These models offer higher clock speeds and support overclocking, providing flexibility for users who want to maximize performance. This is the most common model found in workstations.
- Non-X SKUs: The design for non-X models is for lower power consumption and does not support overclocking, making them a good choice for energy and budget-conscious users.
Related Content
- Puget Systems Podcast – AMD Ryzen 9000 Series Benchmark Roundup
- AMD Ryzen 9000 Content Creation Review
- AMD Ryzen 9000: Performance vs Previous Generations
- Power Draw, Cooling, and Efficiency: AMD Ryzen 9000 Series Processors
- AMD Ryzen 7 9800X3D Content Creation Review
- AMD Ryzen 9600X and 9700X Content Creation Review
Intel Core Ultra 200S Series
The Intel Core Ultra 200 Desktop Series, based on Intel’s Arrow Lake architecture, features a hybrid design with high-performance P-Cores and power-efficient E-Cores. This design balances performance across single-threaded, lightly-threaded, and multi-threaded workloads. P-Cores handle demanding tasks, while E-Cores focus on lower-priority operations, helping with multi-tasking workloads using less power. They take up less physical space on the CPU, giving the processor additional cores and flexibility compared to a processor with only P-Cores. The Core Ultra processor also includes an integrated Neural Processing Unit (NPU) to handle simple AI-driven tasks.
The Intel Core Ultra 200 Series offers a range of product families and SKUs tailored to different user needs. Across all SKUs, the Intel Core Ultra 200 Series supports DDR5 memory with increased capacities and speeds as well as PCIe 5.0 for faster communication with GPUs, NVMe storage, and other peripherals. The series is available in the following SKU families:
- Normal (non-K) SKUs: Designed for slightly lower performance, these models do not support overclocking but provide reliable performance for standard workloads.
- “K” SKUs: These models support overclocking and include integrated graphics, making them ideal for users who want flexibility in tuning performance and access to an iGPU. The iGPU integrates Intel Quick Sync Video (QSV) technology, a hardware-based video encoding and decoding feature that accelerates video processing tasks—providing faster decoding and encoding support for modern codecs such as AV1 and HEVC.
- “F” SKUs: These processors do not support overclocking and also lack integrated graphics, focusing on stability and power efficiency.
- “KF” SKUs: Similar to “K” models except that the “KF” processors do not include integrated graphics.
Related Content
- Puget Systems Podcast – Intel Core Ultra Content Creation Review
- Intel Core Ultra 200S Content Creation Review
- Power Draw, Cooling, and Efficiency: Intel Core Ultra 200S
Workstation CPUs
Workstation-class CPUs are designed to handle specialized tasks, offering increased processing power for multi-threaded applications, bandwidth scalability, reliable performance during intensive workloads, and efficient resource management. These processors typically offer higher core and thread counts, designed to deliver the highest raw compute performance available in a standard desktop, supported by high memory bandwidth, additional PCIe bandwidth, and enhanced I/O capabilities. Unlike consumer CPUs, workstation processors do not include integrated graphics (iGPU) – instead dedicating all available silicon to CPU performance and relying on discrete GPUs (dGPU) for both video output and assistance with certain intensive computational tasks. These processors feature high core and thread counts, ECC memory support for data integrity, and extensive PCIe lane counts that enable the integration of multiple GPUs and other expansion cards.
Current workstation-class CPUs available include Intel’s Xeon™ W series, AMD’s Ryzen™ Threadripper 7000, and Threadripper PRO 7000 WX-series. As before, here is a comparison of the core specifications of these workstation-class CPUs. While the table provides an overview, it does not include all available features or configurations for each processor family.
Features | Intel Xeon W Series | AMD Ryzen Threadripper 7000 | AMD Ryzen Threadripper PRO 7000 WX |
---|---|---|---|
Architecture | Sapphire Rapids Golden Cove microarchitecture | Zen 4 | Zen 4 |
Socket | LGA 4677 | sTR5 | sTR5 |
CPU Core Counts | 8 – 60 Cores | 24 – 64 Cores | 12 – 96 Cores |
Max Memory Support (RAM) * | Up to 4 TB DDR5 4800 MT/s (8 memory channels) RDIMM | Up to 1 TB DDR5 5200 MT/s (4 memory channels) RDIMM | Up to 2 TB DDR5 5200 MT/s (8 memory channels) RDIMM |
PCIe Lanes | 64 – 112 | 88 (usable) | 144 (usable) |
L3 Cache | 22.5MB – 112.5MB | 128MB – 256MB | 64MB – 384MB |
MSRP | $609 – $5,889 | $1,499 – $4,999 | $1,399 – $10,000 |
* Maximum memory support, as listed by Intel and AMD, reflects the theoretical capability of the platform. Achievable configurations may vary based on motherboard design, DIMM compatibility, and memory density availability.
AMD Ryzen Threadripper 7000 Series
AMD built the Ryzen Threadripper 7000X series on the Zen 4 architecture, designing it for professionals and enthusiasts who need higher CPU core counts or additional platform scalability compared to consumer platforms like Intel Core or AMD Ryzen. Threadripper 7000X strikes a balance between core performance, memory bandwidth, and PCIe connectivity without being as expensive as the Threadripper PRO line. The 7000X series offers up to 64 cores and quad-channel DDR5 memory support, along with 48 PCIe 5.0 and 32 PCIe 4.0 lanes. You can review the full list of Threadripper 7000X processors on AMD’s website.
Related Content
- Puget Systems Podcast – AMD Ryzen Threadripper 7000 Content Creation Review
- AMD Ryzen Threadripper 7000 Content Creation Review
- AMD Ryzen Threadripper 7000: Performance vs Previous Generations
- Power Analysis: AMD Ryzen Threadripper 7000
AMD Ryzen Threadripper PRO 7000WX Series
AMD built the Threadripper PRO 7000 WX-series on the Zen 4 architecture and 5nm process, expanding on the Threadripper 7000X series with more cores, memory channels, and PCIe lanes to maximize performance in professional and enterprise environments. It offers up to 96 cores, eight-channel DDR5 memory support, and 128 PCIe 5.0 lanes. Threadripper PRO shortens processing times by distributing tasks across its high core and thread count, these features make it suited for engineering, simulations, VFX production, and data science workloads that need heavy computations. You can view the full list of Threadripper PRO 7000WX processors and their specs at AMD.
Related Content
- Puget Systems Podcast – AMD Ryzen Threadripper PRO 7000 WX-Series Content Creation Review
- AMD Ryzen Threadripper PRO 7000 WX-Series Content Creation Review
- AMD Ryzen Threadripper 7000 vs Threadripper PRO 7000 WX for Content Creation
- AMD Zen4 Threadripper PRO vs Intel Xeon-w9 for Science and Engineering
- AMD Threadripper PRO Memory Channel Performance Scaling
Intel Xeon W-2500 and W-3500 Series
Intel divides the Xeon W series, built on the Sapphire Rapids architecture, into W3, W5, W7, and W9 tiers, each designed to meet specific professional requirements ranging from entry-level tasks to high-performance workloads. These processors offer up to 60 cores/120 threads and up to 112 PCIe 5.0 lanes, enabling scalability and performance for multi-threaded workloads in fields such as engineering, scientific research, and creative applications. They support AI acceleration technologies, including Deep Learning Boost and Advanced Matrix Extensions (AMX), which can improve performance in machine learning and other data-driven tasks.
The differences between the Xeon W-2500 and Xeon W-3500 series processors are based on their price and performance. The W-2500 provides scalability for users, offering higher memory capacities and core/thread counts than a consumer CPU. The W-3500 is a step up from the W-2500 series CPUs, designed for high end workstations that require maximum memory capacities, larger core / thread counts, and top tier performance.
Intel divides the Xeon W lineup into X and non-X SKUs, which differ in their capabilities.
- X Variants: These models generally offer the highest clock speeds and allow users to unlock overclocking capabilities to maximize performance. It’s worth noting that the default clock speeds are still a good choice for maximum performance, however, Intel doesn’t officially support overclocking and users should be aware of the inherent risks such as instability and voided warranties if they decide to overclock their CPUs.
- Non-X Variants: These models do not support overclocking, and generally run a little slower, prioritizing stability and efficiency over raw performance.
Beyond the W series, Intel’s Xeon lineup includes several processor families designed for other various workloads.
- Xeon Scalable Processors: Targeted at enterprise and data center environments, offering scalability, security, and reliability for critical applications like cloud services.
- Xeon Max Processors: Optimized for high-performance computing and AI workloads, featuring high-bandwidth memory for deep learning and simulation tasks.
- Xeon D Processors: Designed for edge computing and embedded systems, providing compact and energy-efficient solutions for space or power-constrained environments.
- Xeon E Processors: Focused on small businesses, offering a balance of performance and cost for general-purpose tasks.
For a more detailed comparison of Intel Xeon W and other Xeon processors, you can explore the full list of processors on Intel.
Related Content
Server CPUs
Server-class CPUs support large-scale computing environments. While users do not typically use them in desktop systems, these CPUs remain relevant for workflows requiring centralized resources for multiple users (private cloud) or heavy compute needs. These processors offer extremely high core counts and configure into single-, dual-, or even quad-CPU setups. Since this class of CPU rarely appears in desktop form factors, we’ve excluded it from this article as it falls outside the scope of the article.
CPU Cooling
CPU cooling plays a critical role in maintaining system stability and optimal performance. Most modern processors feature some form of ‘boost’ technology, which allows them to exceed their base clock speed as long as they don’t exceed a maximum power budget and temperature range. If the CPU gets too hot, it reduces (throttles) the clock speed in order to prevent damage to the CPU. Proper cooling helps avoid thermal throttling and can help extend the lifespan of hardware components. Choosing the right cooling solution depends on factors like thermal design power (TDP), chassis compatibility, and desired noise (measured in dB) levels.
In general, the system integrator you purchase your computer from should ensure it is equipped with adequate cooling. However, it is useful to know that workstations typically use two primary types of coolers: air coolers and liquid coolers.
Air coolers utilize heatsinks that mount directly to the processor. Those heatsinks consist of fins with a large surface area that dissipates heat. Heatpipes often connect the fins to move heat more quickly, while one or sometimes two fans blow cool air across the fins. The only moving part is the fan, so with a large, high-quality model, they can be very quiet – but cheaper models can also be quite noisy. They have fewer failure points than an AIO cooler but also usually have a lower top-end limit to how much they can cool, given the space and mounting restrictions within an enclosure.
AIO (all-in-one) liquid coolers are a single, closed-loop unit that utilizes a pump to move coolant and heat from the processor to a radiator mounted to the chassis. The radiators generally have up to three fans in 120 or 140mm sizes. This means that they have more moving parts, which can result in higher noise levels and more potential failure points – but they can also provide significantly better cooling performance, making them an ideal choice for high-performance systems or processors with higher thermal requirements. This improved cooling capacity helps maintain stable performance during long, heavy workloads by reducing thermal throttling, ensuring the system runs efficiently without overheating.
Custom liquid coolers take a step beyond AIO coolers and typically appear in high-end DIY setups. Like AIO coolers, they use a pump and coolant to transfer heat from the processor to a radiator – but they utilize separate water blocks, tubing, and other components in a bespoke setup. This makes them much more complicated to configure but also allows users to cool GPUs, chipsets, and other hot-running components. These systems offer superior cooling performance and flexibility when properly set up but require regular maintenance and are more prone to potential issues such as coolant leaks. Their larger size and custom nature require more planning to ensure they fit properly within the case and maintain efficient airflow, while also making them unsuitable for systems that ship long distances.
Memory (RAM)
RAM (random access memory) is a type of volatile memory that temporarily stores data while the computer is powered on and clears when powered off. It provides much faster data access than storage devices like hard drives or SSDs. Determining the best RAM configuration for a user’s system can be complicated. The exact frequency and timings depend on the CPU, motherboard / chipset, and the total number of memory modules (DIMMs). Each of these factors affects the memory controller’s behavior and its ability to operate efficiently and reliably. For instance, most CPUs can operate memory at a higher frequency when just one DIMM is installed per memory channel. However, installing two DIMMs in each channel slows the memory as a whole.
For most professional workloads, it is better to prioritize having enough RAM first and worry about memory speed as a secondary factor. We recommend against utilizing ‘overclocked’ memory—modules designed to run faster than the CPU’s memory controller can handle—as that can negatively impact overall system stability.
The latest generation of RAM, DDR5, improves upon DDR4 with increased capacity, faster speeds, and better power efficiency. There are many types of RAM, but we will focus on three common classifications that you’ll see in desktop systems: UDIMM, CUDIMM, and RDIMM.
UDIMM (unregistered DIMM) is the standard memory type used in desktop systems. As unbuffered memory, it communicates directly with the memory controller without the use of intermediary hardware, offering lower latency but less stability compared to buffered or registered memory. DDR5 UDIMM modules currently support capacities of up to 48GB per module and operate at base speeds starting at 4,800 MT/s. It has a theoretical transfer rate of up to 8,800 MT/s, as defined by the JEDEC standard – though no modules with JEDEC settings that high are available yet.
CUDIMM (clocked unregistered DIMM), is the newest type of DDR5, designed for systems requiring high-performance memory with enhanced signal stability. It includes a clock driver (CKD) that regenerates and stabilizes clock signals, ensuring data accuracy and integrity at high speeds while reducing the risk of errors and crashes. This design allows CUDIMMs to operate reliably at higher frequencies than UDIMMs, particularly in systems with multiple DIMMs or overclocked configurations. By supporting faster clock speeds, CUDIMMs enable quicker data transfers between the RAM and the memory controller – but tend to cost more due to added manufacturing complexity. DDR5 CUDIMM modules are beginning to enter the market offering capacities up to 64GB per module.
RDIMM (registered DIMM) is a type of memory module that uses a register to buffer data signals and reduce the electrical load on the memory controller. This design enables systems to manage larger memory capacities with greater stability and reliability. DDR5 RDIMM modules can support up to 256GB of RAM per module, with base speeds starting at 4,800 MT/s and higher frequencies possible depending on system specifications. However, because RDIMM modules prioritize memory capacity and stability, they do so at the cost of lower average speeds and slightly higher latency. Additionally, only specific motherboards and CPUs support these modules, so you should verify compatibility before selecting this type of memory.
In addition, there is also a variation of memory called ECC (Error Correction Code), which is designed to detect and correct memory errors, ensuring data integrity and system reliability. Available in UDIMM and RDIMM configurations, it offers increased stability, especially in environments handling large datasets or complex computations. While ECC RAM operates at slightly lower speeds, the trade-off is enhanced reliability. However, it requires compatible motherboards and CPUs, and is often considered overkill for most desktop workstations.
When considering RAM for a system, it’s important to consider compatibility, capacity, and speed as the three main priorities. If RAM is not compatible with the motherboards and CPU, the system may fail to boot, operate unstably, or not recognize its capacity or speed.
An insufficient capacity of RAM can cause performance bottlenecks, forcing the system to rely on slower storage devices, which significantly impacts overall performance. Memory-intensive applications require sufficient amounts of RAM to run smoothly, and a shortage can result in noticeable slowdowns or system instability. It’s best to have more memory as this can provide headroom to accommodate future software advancements and system demands.
Conversely, choosing RAM with speeds or capacities beyond what the motherboard or CPU supports can result in instability, failure to boot, or reduced performance if the memory defaults to lower speeds. Mismatched DIMM configurations or unsupported memory profiles may also cause errors, crashes, or reduced reliability.
Related Content
Storage
Storage refers to the hardware that stores data on a computer, either temporarily or permanently. In desktop PCs, storage typically consists of internal drives installed in the chassis and connected to the motherboard through interfaces like SATA or PCIe, but similar drives are often available in external enclosures as well. These drives store the operating system, applications, and work files such as documents, photos, and videos. Currently, users can install three primary types of storage drives within a workstation and store files externally in multiple ways.
Platter Drives
Hard disk drives (HDDs) use magnetic platters to store data, providing high-capacity storage at a low cost. This makes them a popular choice for storing large, less frequently accessed data. Modern HDDs offer storage capacities ranging from 1TB to over 30TB. However, their mechanical design limits read and write speeds to a maximum of about 250 MB/s. Due to the falling cost and rising capacities of faster alternatives, this type of drive rarely appears inside modern workstations and has largely moved to large external storage arrays.
SATA SSD
Solid state drives (SSDs) provide faster data access than HDDs by using NAND flash memory instead of spinning disks. They offer storage capacities ranging from 120GB to 8TB, with read and write speeds around 400 to 550 MB/s, making them significantly faster than HDDs though more limited in capacity. Their non-mechanical nature makes them even faster for random read/writes which is what affects the “snappiness” felt in modern systems. People commonly use SATA SSDs for entry-level computers and general storage needs, offering a balance between performance and cost-effectiveness.
NVME SSD (M.2)
Non-Volatile Memory Express (NVMe) drives refer to a type of solid-state drive that use the NVMe protocol over the PCIe interface, offering much faster read and write speeds than SATA SSDs can achieve. These drives help improve load times, speed up data transfers, and make systems more responsive, which makes them a great choice for builds focused on performance. The most common form of these drives is called M.2, a compact drive about the size of a stick of gum that usually slots directly onto a motherboard, minimizing space inside the computer. Larger U.2 drives, which use the same protocol, are shaped more like traditional SSDs and install in drive bays, connecting to the system via a data cable.
The latest Gen 4 and Gen 5 NVMe M.2 drives push performance even further, with Gen 4 reaching speeds up to 7,000 MB/s and Gen 5 exceeding 12,000 MB/s. They typically come in capacities ranging from 250GB to 8TB and connect to motherboards via M.2 PCIe slots. While Gen 5 drives deliver exceptional performance, they are still rare – and their higher speeds generate more heat, so proper cooling is important to avoid throttling.
The falling cost of NVMe drives in recent years has made them by far the most common type of drive you will find in modern, high-performance workstations. Lower-performance SATA drives can save a small amount of budget, but the loss of speed often isn’t worth it for the minor cost savings. As capacities increase, the price difference between SATA and NVMe can also widen – so for high capacity drives, the older tech can still make sense.
Network-Attached Storage (NAS)
Network-attached storage (NAS) is a term referring to dedicated storage systems connected to a local network, allowing multiple users to access shared data. By keeping data on the local network, this approach maintains faster access and greater control than cloud storage. A key feature of NAS is its ability to use RAID (redundant array of independent disks) for redundancy and protection against hardware failures.
A NAS box typically consists of a specialized enclosure or chassis housing multiple storage drives – HDDs, SATA SSDs, or NVMe SSDs – depending on performance and capacity needs. Internally, a NAS is basically a mini computer complete with its own CPU, RAM, and motherboard, all running a customized operating system tailored for storage and networking.
These systems will include a high-speed network interface of some kind – most typically, the same Ethernet connection that computers commonly use. These come in several speeds, and you would want to match or exceed the speed your local network router is capable of. Many models also support advanced features such as drive expansion through additional, daisy-chained units or caching with NVMe SSDs for enhanced performance.
Cloud Storage
Cloud storage provides scalable, remote data solutions by hosting data on servers accessed via the Internet. Like on-premises storage, it relies on similar hardware, but at a much larger scale, and retrieves and manages data over internet connections rather than a local network. This allows cloud storage to scale easily without requiring end users to perform physical hardware upgrades, making it ideal for offsite backups, disaster recovery, and syncing data across devices.
For modern workflows, cloud storage allows for real-time collaboration and access to files from almost anywhere. Teams can store, share, and edit files directly on cloud platforms, ensuring seamless workflows regardless of location. However, even with the fastest internet available, it takes time to download and upload files to cloud storage, and the costs can be high depending on how much space you need and how often you need to access the data.
Hybrid cloud solutions further enhance storage options by combining the speed and control of local storage with the scalability and accessibility of the cloud. This approach allows organizations to store large, frequently accessed files locally for high-speed performance while leveraging the cloud for collaboration, remote backups, and disaster recovery.
Related Content
- Puget Systems Podcast – Storage & Server Solutions with Chris Stevenson
- How to Configure Storage and Cache File Locations in Premiere Pro
- Understanding Storage for Video Editing
Graphics Card (GPU)
The main chip on a graphics processing unit (GPU), also commonly called graphics or “video” cards, specializes in accelerating graphical and computational workloads. With thousands of cores optimized for parallel processing, GPUs efficiently handle tasks that involve processing large amounts of data simultaneously. This includes rendering images and video frames, performing matrix calculations for machine learning, simulating particle or object movement in physics simulations, encoding and decoding video, and processing pixels in image editing.
GPUs fall into two categories: integrated GPUs (iGPUs) and discrete GPUs (dGPUs). Integrated GPUs reside within the CPU and share system memory, while discrete GPUs function as standalone hardware components with dedicated VRAM (Video Random Access Memory) for high-speed graphical data processing. This section focuses on discrete GPUs because they offer dedicated resources and significantly higher performance, making them more suitable for scenarios that require increased graphical and computational power compared to the shared and limited resources of iGPUs.
Like CPUs, discrete GPUs can be broadly categorized into three main classes based on their intended use cases: Consumer, Professional, and Data Center. Each class is built with specific goals in mind, such as performance, reliability, scalability, and cost-effectiveness.
Consumer GPUs
Consumer GPUs cater to individual users, offering high performance at a lower cost compared to other types of GPUs. High-end models deliver significant processing power and large VRAM capacities, mid-tier models balance performance and affordability, and entry-level GPU models offer fewer capabilities but are more budget-friendly. Manufacturers build consumer GPUs with thermal solutions balanced between cooling and noise levels, and are not intended for 24/7 operation under sustained heavy workloads.
Consumer GPUs are available in product lines like AMD Radeon™, Intel Arc™, and NVIDIA GeForce™, with options ranging from budget-friendly models to higher-performance variants. Third-party brands such as ASUS, Gigabyte, MSI, and PNY manufacture graphics cards with these GPUs under license, often adding their own cooling designs or overclocking features. While high-end consumer GPUs can sometimes match or exceed workstation-class GPUs in specifications like clock speeds, core counts, and memory bandwidth, they differ in their functional design and intended use.
Here is a comparison of specifications between consumer graphics cards and their manufacturers. While the table provides an overview, it does not include all available features or configurations for each product. Additionally, there are certain features that differ between manufacturers, making direct comparisons challenging. We have grouped these features to highlight comparable aspects.
Features | AMD Radeon RX 7000 series | Intel Arc B Series | NVIDIA GeForce RTX 40-series | NVIDIA GeForce RTX 50 -series |
---|---|---|---|---|
Architecture | RDNA 3 | Xe2 | Ada Lovelace | Blackwell |
Shader Units | Up to 6144 | Up to 2560 | Up to 16384 | Up to 21760 |
Ray Tracing Accelerators | 32 – 96 | 18 – 20 | Gen 3 up to 191 TFLOPS | Gen 4 up to 318 TFLOPS |
TFLOPS (FP32) | 61.39 TFLOPS | 13.67 TFLOPS | 82.58 TFLOPS | 104.8 TFLOPS |
AI Accelerators (AMD) XMX Engines (Intel) Tensor Cores (NVIDIA) | 64 – 192 | Up to 160 | Gen 4 up to 1321 AI TOPS | Gen 5 up to 3352 AI TOPS |
VRAM (memory config) | 8 – 24 GB DDR6 | 10 – 12 GB DDR6 | 8GB DDR6 – 24 GB DDR6X | 12 – 32 GB DDR7 |
Memory Bandwidth | Up to 960 GB/s | Up to 456 GB/s | Up to 1008 GB/s | Up to 1792 GB/s |
Decoder Encoder | VCN (Decode) VCE (Encode) up to x2 Encoder | Intel Quick Sync Video | Gen 5 x1 Decoder Gen 8 up to 2x Encoder | Gen 6 up to x2 Decoder Gen 9 up to x3 Encoder |
MSRP | $269 – $999 | $219 – $249 | $299 – $1599 | ~$549 – $1999 |
AMD Radeon RX 7000 Series
AMD recently announced their Radeon RX 9000 series graphics cards at CES 2025 in Las Vegas. These cards are not yet available for purchase, and we will update this article when the full information is available along with the official press release.
The AMD Radeon RX™ 7000 series GPUs, based on the RDNA 3 architecture and released in November 2022, are AMD’s current consumer GPU in the market. The RX 7000 series emphasizes rasterized gaming performance and includes features such as Infinity Cache, which reduces data latency by storing frequently accessed data closer to the GPU cores, and a chiplet-based design, which separates GPU cores and memory controllers to improve scalability and energy efficiency. These GPUs also support encoding and decoding of modern video codecs, including AV1, H.264, and H.265 for video playback, streaming, recording, and editing.
The VRAM capacity of different Radeon RX 7000 models suits various tiers of performance and workloads. Entry-level and mainstream models feature up to 8GB of VRAM, performance-tier GPUs provide up to 16GB, and enthusiast-level options, like the RX 7900 XTX, include up to 24GB of VRAM. AMD uses naming conventions and product SKUs to classify models within the lineup based on their capabilities and performance tiers.
- XT: Indicates a higher-performance variant within a specific tier, with increased cores, clock speeds, and memory capacity.
- XTX: Represents the flagship or top-tier variant, offering the highest performance in the lineup.
- GRE (Golden Rabbit Edition): A variant that balances performance and efficiency, often with limited availability based on region.
- Non-Suffix Models: Base versions of the GPUs, offering balanced performance at a lower price point, typically with fewer cores and lower clock speeds compared to their XT/XTX counterparts.
Refer to AMD’s website for a full list of Radeon RX 7000 series products and specifications and Radeon for creators for a full list of codecs supported for hardware decoding and encoding.
Related Content
- AMD Radeon RX 7900 XTX 24GB Content Creation Review
- NVIDIA GeForce RTX 40-series vs AMD Radeon RX 7000 Series for Content Creation
Intel Arc B Series
The Intel Arc B-series (codenamed Battlemage) GPUs, announced in December 2024, are Intel’s second-generation consumer graphics cards built on the Xe2-HPG architecture. They offer improvements in power efficiency and performance compared to the previous Arc A-series (Alchemist). The B-series is built to deliver competitive performance at a budget-friendly price point. Updates to the Xe2 architecture and refined software optimizations make it an attractive option for consumers seeking value-oriented GPUs. The series includes the Arc B580, which launched in December 2024, and the Arc B570, scheduled for release in January 2025.
Refer to Intel’s product page for a full list of specifications for the B580 and B570 and refer to their Product Support page for a full list of codecs support by Arc series GPUs.
Related Content
NVIDIA GeForce RTX 40 & 50-Series
The NVIDIA GeForce RTX™ 40- and 50-series GPUs, built on the Ada Lovelace and Blackwell architectures respectively, represent NVIDIA’s current consumer GPU lineup. The RTX 50 series builds on the RTX 40 series with architectural improvements, introducing more efficient processing methods and refined rendering capabilities, utilizing both AI and hardware processing for peak performance. They serve as an alternative to professional GPUs in some workloads, especially among the content creation space. However, workstation-class GPUs remain the preferred choice for professional systems requiring maximum stability and uptime due to their certified drivers and specialized support. It’s also worth noting that certain applications require NVIDIA GPUs with CUDA cores for compatibility, making them the only choice for certain workflows.
NVIDIA offers the GeForce RTX 40 and 50 series GPUs in a few different variants, which typically feature higher performance and specs than their equivalent base model. It’s worth noting that these variants apply to the RTX 40 and 50 series GPUs only as other models of GeForce products may have different naming conventions.
- Base Models: Standard versions that balance performance and cost, suitable for a wide range of use cases.
- Ti Models: NVIDIA will refresh base models, noting “Ti” to indicate a revision of the product.
- Super Models: NVIDIA often enhances their GeForce model line between major launches, using “Super” to indicate newer version of the model with small boost in performance.
While these GPUs feature technologies that support gaming, they also offer other feature suited for professional workloads. With VRAM capacities ranging from 8GB to 24GB in the RTX 40 series and 12GB to 32GB in the RTX 50 series, they support workloads that demand fast parallel processing and significant memory capacity. Real-time ray tracing performance benefits from AI-driven optimizations, which reduce the computational load by predicting and approximating how light interacts with surfaces. Both 40 and 50 series graphics cards include multiple NVENC and NVDEC encoders and decoders, offering support for modern video codecs such as AV1, H.264, and H.265 in multiple bit depths and chroma subsampling formats.
Visit NVIDIA’s website for a full list of specs comparing the GeForce RTX 50 series GPUs to other models and check out NVIDIA GPU Matrix for a list of codecs supported by their NVENC and NVDEC decoders and encoders.
Related Content
- NVIDIA GeForce RTX 5090 Content Creation Review
- NVIDIA GeForce RTX 40 SUPER Series Content Creation Review
- NVIDIA GeForce RTX 4090 24GB Content Creation Review
- NVIDIA GeForce RTX 4080 16GB Content Creation Review
- 1-7x NVIDIA GeForce RTX 4090 GPU Scaling
Professional GPUs
Professional GPUs are built for workflows that require stability, higher precision calculation support, and multi-card compatibility. They offer higher VRAM capacities to handle large datasets and often include ECC memory for error correction, which helps maintain data integrity during computations. These GPUs handle tasks like CAD, scientific simulations, video editing, visual effects, and 3D rendering, where system instability during intensive tasks can lead to data loss or workflow interruptions. They often come with certified drivers to ensure consistent performance and compatibility with professional software, including NLE software, CAD tools, rendering engines, and simulation platforms. Additionally, they maintain reliable performance during extended workloads. While they are more expensive than consumer GPUs, their features are necessary for workflows where accuracy and stability are essential.
Current professional GPUs available in the market includes Intel Arc™ Pro, AMD Radeon™ PRO, and NVIDIA RTX™ (formerly Quadro). These GPUs feature VRAM capacities of up to 48GB, capping out higher than consumer-class cards, which enables them to better handle high-resolution textures, complex 3D models, simulations, and other advanced workloads that require sufficient memory capacities. Although they may not quite achieve the same raw processing speeds as some consumer GPUs, they prioritize stability and reliability, making them suitable for professional environments requiring consistent and accurate results.
Features | AMD Radeon PRO W7000 Series | Intel Arc Pro A-Series | NVIDIA RTX Ada Generation |
---|---|---|---|
Architecture | RDNA 3 | Xe HPG | Ada Lovelace |
Shader Units | Up to 6144 | Up to 2048 | Up to 18,176 |
Ray Tracing Accelerators | 32 – 96 | 8 – 16 | Gen 3 up to 142 |
TFLOPs (FP32) | 61.32 TFLOPs | 8.397 TFLOPs | 91.06 TFLOPs |
AI Accelerators (AMD) XMX Engines (Intel) Tensor Cores (NVIDIA) | 64 – 192 | 128 – 256 | Gen 4 up to 568 |
VRAM (memory config) | 8 – 48 GB DDR6 | 6 – 12 GB DDR6 | 16 – 48 GB DDR6 |
Memory Bandwidth | Up to 864 GB/s | Up to 256 GB/s | Up to 960 GB/s |
Decoder Encoder | VCN (Decode) VCE (Encode) up to 2x Encoder | Intel Quick Sync Video | Gen 5 up to 3x Decoder Gen 8 up to 3x Encoder |
MSRP | $429 – $3999 | N/A | $625 – $6800 |
AMD Radeon PRO W7000 Series
AMD Radeon PRO W7000 series GPUs, announced in April 2023, are AMD’s most current professional graphics cards available. Built on the RDNA 3 architecture, the W7000 series handles the processing of engineering, architectural design, media production, and scientific research workflows, which require compatibility with industry-specific software, precision, and stable performance. Additionally, AMD provides AI support through these GPUs, integrating accelerators optimized for generative AI and inference workloads, supporting computation formats like FP16 and INT8. These GPUs maintain compatibility with a wide range of ML frameworks, enabling various AI applications such as model training, inference, and real-time processing. While these cards don’t quite reach the performance metrics of certain NVIDIA RTX Ada GPUs, they offer unique support for various software packages and make a good option if an NVIDIA or Intel Pro GPU is not supported.
Visit AMD’s website for a full list of specifications for their PRO W7000 Series GPUs.
Related Content
- AMD Radeon PRO W7800 and W7900 Content Creation Review
- AMD Radeon PRO W7600 and W7500 Content Creation Review
- Stable Diffusion Performance – NVIDIA RTX vs Radeon PRO
Intel Arc Pro A-Series
Intel is a recent entrant in the professional GPU market. The Arc Pro A-series GPUs, first launched in August 2022 with the Pro A40 and Pro A50, followed by the Pro A60 in June 2023, target entry-level professional workloads. Driver support for the Arc Pro A-series is still evolving and offers less coverage compared to the ecosystems of NVIDIA and AMD. These GPUs provide an option for professionals with light to moderate workloads, supporting open-source frameworks and integrating with Intel’s software ecosystem, such as oneAPI, to facilitate development across Intel platforms.
Visit Intel’s website for a full list of products and specifications of Pro A-Series GPUs.
NVIDIA RTX Ada Generation
NVIDIA’s professional GPUs, including the RTX™ Ada Generation series announced in December 2022, come with hardware features such as CUDA, Tensor, and RT Cores, and provide driver compatibility tailored to professional workflows and industry-specific software. These GPUs, with VRAM capacities up to 48GB of DDR6, handle large datasets and software that require high VRAM capacities, such as simulations, rendering, real-time ray tracing, and AI-driven tasks. CUDA cores accelerate parallel processing of large datasets, supporting tasks such as simulations, rendering, and real-time data analysis, RT Cores perform ray tracing calculations by simulating light interactions in 3D environments, while Tensor Cores accelerate ML computations, supporting FP operations, deep learning training, inference tasks, and other AI-driven processes.
NVIDIA provides a set of tools and technologies in RTX Ada Series GPUs, including CUDA and OptiX drivers, which provide compatibility with certain professional software applications to enhance computational and rendering processes. OptiX, a ray tracing API, improves rendering performance and precision by calculating light paths and interactions, supporting complex visualizations in fields such as architectural design, product prototyping, and VFX production. Within Autodesk Maya, SolidWorks, AutoCAD, and CATIA, CUDA is the only driver that the applications support for hardware-accelerated simulation and rendering tasks, making RTX Ada cards the only available option for users.
Visit NVIDIA resources for a full list of specifications of RTX Ada GPUs.
Related Content
- NVIDIA RTX Ada Generation Content Creation Review
- NVIDIA RTX 6000 Ada vs RTX A6000 for Content Creation
- LLM Inference – NVIDIA RTX GPU Performance
- GPU Rendering: NVIDIA RTX 6000 Ada Performance
Datacenter GPUs
Datacenter GPUs support large-scale, resource-intensive environments where multiple users or systems share computational resources. They feature high memory bandwidth, scalability, and increased support for multi-GPU configurations, with capabilities such as GPU virtualization to enable shared access across virtual machines and error correction to ensure data integrity. Built for continuous 24/7 operation, server GPUs include robust thermal and power management systems to maintain reliability in demanding workloads.
Data center GPUs support hybrid systems and cloud environments. Their architecture is built for shared, multi-user environments and tasks requiring large-scale parallel processing, which differ from the dedicated, single-user workloads typically found in desktop systems and traditional workflows. For this reason, we have excluded datacenter GPUs from this article.
Conclusion
This article provides an overview of current PC hardware components, covering their design, capabilities, and use cases in desktop and workstation systems. Hopefully you’ve found it helpful, whether you’re a new user unfamiliar with how these components function, an enthusiast seeking a refresher, or someone catching up on the latest trends.
However, it is important to recognize that this article does not encompass every component. Motherboards, power supplies, specialty hardware, and other peripherals are also critical to system performance and reliability. Factors influencing their selection include the user’s industry, workload demands, long-term reliability considerations, and budget constraints. These components often align with specific technical requirements tied to individual workflows, making them difficult to address comprehensively in a single article, and have therefore been excluded.
Ideal hardware configurations can also vary significantly depending on user needs and professional applications. A VFX artist may rely on a multi-GPU setup for rendering complex 3D models and simulations, while a video editor might prioritize hardware configurations optimized for their footage resolution and editing software. Similarly, data scientists might require high memory capacities and AI accelerators to handle large datasets efficiently, whereas CAD designers typically need hardware certified for precision and compatibility with specialized software.
As PC hardware continues to evolve with advancements in processing power, connectivity, and AI-driven technologies, there are an increasing number of options to pick from. Staying informed about emerging technologies and industry trends allows you to make informed decisions when building or upgrading your system, but it can be a daunting task if you do not have the time or personal interest to keep up. If you are looking for a new workstation and have questions about building the right system for your workflows, we highly recommend speaking with one our consultants.
Finding the perfect workstation doesn’t have to be complicated. Explore our solutions page for a curated selection of recommended systems for a multitude of applications and workflows, or visit our custom configuration page if you already know the ideal hardware for your needs. If you need assistance with tailoring a system to a unique workflow or have any other questions, we encourage you to reach out to our dedicated technology consultants.
Related Content
Latest Content
- NVIDIA GeForce RTX 5080 Content Creation Review
- Understanding Modern Desktop PC Hardware for Workstations
- NVIDIA GeForce RTX 50-Series | Known Software Issues
- NVIDIA GeForce RTX 5090 Content Creation Review
- NVIDIA GeForce RTX 50-Series | Features for Content Creators
- Windows 11 24H2 File Explorer Woes