Thankfully, Intel’s suffixes don’t require much explanation because of how rigid they are. If a processor lacks the K suffix, you can’t overclock it, pure and simple. Intel will combine suffixes when necessary, however. The Core Ultra 5 245KF is unlocked for overclocking but lacks integrated graphics, for example, while the Core i9-14900KS is a special edition of the unlocked Core i9-14900K.
CPU specs are only useful if you understand what they mean – and, more importantly, what they don’t. Unlike generations past, where big clock speed boosts and core count jumps were common, AMD and Intel have much more conservative spec bumps each generation, if there's a spec bump at all.
Specs are still important for neighboring comparisons, but it's important to understand what they mean in the broader context of a CPU's architecture.
Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test .
A CPU with more cores can execute more instructions in parallel, but how that manifests in real applications varies widely. Rendering applications like Blender and transcoding applications like Handbrake scale well with high core counts, for example. Games, on the other hand, don’t see large performance jumps past eight cores, and many games don’t even see a marked improvement beyond six cores.
AMD’s modern CPU cores are designed with simultaneous multi-threading, or SMT. This doubles the number of logical processes running on a physical CPU core. So, the 16-core Ryzen 9 9950X has 32 threads. Intel originally introduced SMT – Intel calls it Hyper-Threading – but its current-gen Arrow Lake CPUs don’t use Hyper-Threading. Each physical CPU core only has a single thread.
Traditionally, CPUs use a homogeneous architecture; each core has the same design. Intel has pushed toward heterogeneous architectures in the past several generations, however, mixing together high-performance P-cores with high-efficiency E-cores within a single CPU. These designs can boost overall core count by leveraging weaker cores instead of high-performance cores across the entire die.
From a shopping perspective, it’s important to keep the heterogeneous architecture in mind for Intel chips. The Core i9-14900K has 24 cores compared to the Ryzen 9 9950X’s 16. However, both chips have 32 threads – the E-cores on the Core i9-14900K lack Hyper-Threading.
Synchronizing the disparate components of a CPU architecture is a clock. Clock speed is a frequency – how many cycles are completed each second – and there are two numbers for a CPU. There’s a base and boost clock speed. In modern CPUs, the boost clock speed generally refers to the maximum clock speed on one or two cores, not the maximum clock speed for all cores operating at the same time. Often, one or two preferred cores will boost to higher speeds and handle heavy-duty calculations while the other cores operate at a lower frequency with less intensive work.
Clock speed is important, but it says less about how powerful a CPU is than it previously did. Critically, clock speed doesn’t say anything about how many instructions are executed per cycle. You can think about clock speed like the speed of a conveyor belt. You can increase the speed of the belt, but it might get too hot and break. Alternatively, you can keep the speed the same and widen the belt, or stack things on top of each other, to move more stuff at the same speed.
Rather than the belt speed, a better number to use would be how many items you're able to move. For CPUs, that number is instructions per clock, or IPC; you’ll also see it referred to as instructions per cycle. There are physical limitations of how high the clock speed can go, but if you're able to get more done with each clock cycle, the CPU runs faster.
IPC is a good indicator of architectural improvements from one generation to another, but you won’t find it listed on a spec sheet.
CPUs execute a lot of instructions very quickly, so it’s important that the data needed for those instructions are close to the cores executing them. That’s your CPU’s cache. It’s a small amount of SRAM located on the package of the CPU. Without cache, your CPU would need to use your system’s DRAM for everything, which is considerably slower and would cause a significant bottleneck in your system. Decades ago, CPUs didn't need cache because the system memory could keep up with the pace of instruction execution. With a modern CPU, constantly going out to system memory would make your PC feel unusable.
Cache is organized into levels, with the lowest-number level being the fastest and smallest. L1 is the smallest and fastest, L2 cache is slightly larger, and L3 cache is larger still. CPU cores generally have dedicated L1 and L2 cache, while the L3 cache is much larger and shared across the cores. More cache is generally better, but there are limitations to how much cache can go on-package. Not only is SRAM expensive, it also takes up precious die space and contributes to heat. That’s why packaging technology like AMD’s 3D V-Cache have been such a breakthrough.
Although cache is important, more cache doesn’t lead to universally higher performance. It contributes to higher performance in applications where new data is flowing through memory frequently, such as in games.
Power is a messy topic rife with misunderstandings in the world of CPUs. Generally, you’ll see power consumption shared as the Thermal Design Power, or TDP, of a CPU. However, TDP isn’t a power limit, and it doesn’t refer to power consumption. Rather, TDP refers to how much heat the CPU cooler needs to dissipate under maximum load. More power leads to more heat, so TDP is measured in power instead of temperature.
In use, your CPU will often consume less power than the TDP, and it can consume more power for brief periods of time as long as it stays within its thermal limit. To measure peak power consumption, AMD uses PPT, or Package Power Tracking, to note how much power the CPU socket can draw. Intel uses power levels, noted like PL1 and PL2. PL1 is synonymous with TDP, while higher level power limits show maximum power for transient spikes. For high-end, unlocked SKUs, Intel applies a power profile where PL1 = PL2. That means the CPU can sustain higher power consumption for longer than the specified window, assuming it has adequate access to cooling.
Finally, there’s a maximum safe operating temperature, often referred to as TJMax. Once your CPU hits its maximum operating temperature, it will reduce speeds in order to bring the temperature down. In the event the temperatures can’t drop, built-in safety mechanisms will shut down the PC.
Power is a complex topic, and specs don’t share the full picture of power consumption and operating temperatures. Here at Tom’s Hardware, we run a full suite of power and thermal tests for our CPU reviews , which provide a more accurate insight into what you can expect out of a chip. You should use power and temperature specifications to inform your cooler, case, and fan choices, not as hard and fast rules about power consumption.
Desktop CPUs are socketed, not soldered, so you’ll need to pick up a motherboard with the socket that matches your CPU. AMD is currently on Socket AM5 while Intel is on LGA 1851. Both are Land Grid Array (LGA) sockets, where the CPU features contact pads that press into spring-loaded pins in the motherboard socket. AMD previously used a PGA, or Pin Grid Array, socket where the pins were on the CPU itself. AMD abandoned this design with the release of Ryzen 7000 CPUs and sunsetting of Socket AM4.
The socket only defines physical compatibility between a CPU and motherboard; the chipset defines full compatibility. Here are the chipsets from AMD and Intel with the latest socket, and the CPUs they support:
A620, B650, B650E, X670, X670E, B840, B850, X870, X870E
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/pc-components/cpus/SPONSORED_LINK_URL
- https://www.tomshardware.com/pc-components/cpus/cpu-buying-guide#main
- https://www.tomshardware.com
- Newegg's triple component combo deal for $459 pairs an Asus X870 motherboard with 32GB of DDR5-6000 RAM, and a 240mm AIO cooler — $780 worth of parts if purchas
- Opt-In NVIDIA Software Enables Data Center Fleet Management
- Prusa Research introduces the Open Community License to protect open source 3D Printing hardware — new rules aimed at addressing industry abuses
- Critical motherboard flaw allows game cheats, Riot Games blocks 'Valorant' players that don't update BIOS — security patches pushed live by all major motherboar
- AMD's legacy Ryzen 7 5800X3D chips now sell for up to $800, more than a new 9800X3D — AM4 chip costs twice as much as MSRP, as enthusiasts flock to old DDR4 mem
Informational only. No financial advice. Do your own research.