When you’re shopping for a new PC, don’t meekly settle for the default processor recommended by the configurator. Picking the right CPU is a personal decision that you shouldn’t enter into lightly. And with so many options to choose from, you need to know what you’re getting into when you settle on a chip for your system. We’ve rounded up eight of the leading processors on the market and put them through a battery of rigorous tests to help you shop with confidence.
Even after you’ve made the decision to opt for an Intel or an AMD processor, however, you still aren’t out of the woods: Within each company’s chip lineup sit a number of different families of CPUs. And you’ll find plenty of options within each of these product families that run the gamut of specifications and features. These details give each chip its speed, cost, and name-but you’ll run into trouble if you start trying to analyze a processor’s potential based on these three characteristics alone.
It’s unwise to assume that a processor’s stated frequency, or the speed at which the CPU executes its instructions, is a reliable guide to its speed versus competing chips. While that assumption might turn out to be true in some cases, a chip’s actual architecture can torpedo it. Depending on the make and model, a “slower,” or lower-frequency multicore CPU can beat out the general performance of a seemingly faster processor that’s been split into fewer cores–a four-core CPU versus a dual-core CPU, to put it simply. Also, processors that seem to indicate superiority by their naming conventions can be anything but: It all depends on how and when they were created, not on their labels.
And that’s just the beginning of the sources of confusion. How does automatic overclocking play a role in today’s conventional CPUs? How does the price-to-performance ratio play out on seemingly more expensive Intel chips versus AMD’s less pricey designs?
The questions can be as numerous as they are confusing for anyone unfamiliar with processor architectures. But that’s okay: We’ve taken eight classic examples–four each from the Intel and AMD lineups–and run them through our bevy of WorldBench 6 tests to deliver a guide you can use to understand your options, and to help you pick your next processor. We discuss the processors in order, from speediest to slowest, as decided by their WorldBench 6 benchmark scores. For a summary of our findings, click on the thumbnail at upper left here to see our full chart.
Processors aren’t limited to PCs, so we also take a look at CPUs for smartphones and tablets; that chart is on the third page of this article, which includes a brief glossary of processor terms, plus discussions of upcoming chip technologies and the use of server chips as an alternative to the standard chips covered here.
Intel Core i7-980X
Benchmark Score: 147
Whoosh! This category leader is so fast, it’ll sprout wings and fly.
At $999, the 3.33GHz Core i7-980X (with automatic overclocking to 3.6GHz) is Intel’s flagship offering in consumer processors. It’s the first desktop-grade CPU with six physical cores, but hyperthreading yields 12 virtualized ones. The six cores share 12MB of integrated L3 cache. The end result is a measurable performance boost for optimized applications–and a score of 147 on our WorldBench 6 tests.
A 32-nanometer design, the 980X is fully compatible with existing X58-chipset motherboards running an LGA 1366 socket. X58-based motherboards use Intel’s quickpath interconnect (QPI) architecture. QPI replaces the frontside bus design of the now antiquated Core 2 chips, for faster (and greater) bandwidth for routing communications between the motherboard, system components, and the CPU.
Intel splits its mainstream desktop and mobile CPU offerings into three sections: The Core i7, Core i5, and Core i3 lines, representing, respectively, high-end, midrange, and entry-level products. That doesn’t always mean that a Core i5 chip will be slower than a Core i7, however, as you’ll see later. Nor does a Core i7 always have four physical cores. Intel’s mobile i7 line defaults to two cores unless the chip has a QM or XM extension.
Intel Core i7-870
Benchmark Score: 127
A significant price drop with a minor performance loss against Intel’s top-line CPU makes this 45-nanometer, quad-core model an excellent buy.
Intel’s 2.93GHz (with automatic overclocking to 3.6GHz) Core i7-870 processor earns second place in our rankings. The differences between this $564 processor and the Core i7-980X, though important, result in only a 14 percent decrease in performance from that of the Core i7-980X. The i7-870 drops its internal L3 cache to 8MB. In addition, it carries a 45-nanometer designation (the size of the space between transistors). The smaller this number, the smaller and more numerous the transistors that can be packed onto a processor.
Also, this midrange CPU eschews Intel’s quickpath interconnect in favor of a direct media interface (DMI). You won’t notice that change as much as you might the switch to a dual-channel memory controller, which limits you to four memory slots on your motherboard instead of six.
AMD Phenom II X6 1090T
Benchmark Score: 118
AMD’s first-ever six-core desktop CPU beats all of the company’s other offerings, but are its advantages strong enough to justify an upgrade? You decide.
AMD enters the performance game in third place with its brand-new, six-core, 3.2GHz Phenom II X6 1090T processor. Code-named “Thuban,” this $299, 45-nanometer CPU can hit speeds of up to 3.6GHz as a result of Turbo Core, the AMD equivalent of Intel’s automatic overclocking technology (which is called Turbo Boost).
The X6 1090T’s 6MB L3 cache is half that of Intel’s six-core chip, and it takes a performance hit of 20 percent compared with Intel’s Core i7-980X. The X6 1090T also lacks an answer for hyperthreading, a critical technological component of Intel’s top-shelf CPU lineups.
Performance gaps aside, the X6 1090T wins the compatibility race against Intel by a mile. You can drop this CPU into any existing motherboard with AM3 or AM2+ sockets (after a BIOS upgrade). The combination of a simple upgrade path and the CPU’s super-low price of $299 makes for a great alternative to upgrading your existing motherboard and memory, if you just want to reap the benefits of a new, up-to-date processor.
Intel Core i3-540
Benchmark Score: 115
Confusion lurks within Intel’s Core processor lineup. Case in point: The Core i3-540 beats the general performance of the otherwise higher-class Core i5-750!
As mentioned earlier, naming conventions don’t always translate into measurable performance improvements. Intel’s Core i3-540 is a 3.06GHz dual-core CPU, and a member of the company’s Clarkdale line of chips. Released four months after Lynnfield (the family of the Core i7-870), Clarkdale is based on a 32-nanometer production process, versus Lynnfield’s 45-nanometer design.
The $145 Core i3-540 lacks Turbo Boost, so the chip can’t run faster than 3.06GHz. Despite this limitation, and the chip’s 4MB of L3 cache, it still tops the performance of Intel’s Core i5-750 CPU by 8.5 percent on WorldBench 6. (The quad-core i5-750 does do 25 percent better on Cinebench, the multicore-optimized processor benchmark, however.)
The Core i3-540 is really geared for users who aren’t enthusiasts–not only does it yield formidable dual-core performance, but it also comes with a GPU core that is packed right alongside the CPU. This sandwich of performance and integrated graphics could cost you a new motherboard, though, because the widely used P55 chipset doesn’t support the Core i3-540’s GPU core. Look to Q57, H55, or H57 chipsets instead.
AMD Phenom II X4 945
Benchmark Score: 110
If multithreaded applications aren’t your deal, AMD’s Phenom II X4 945 is a powerful alternative to its top-shelf Phenom II X6 1090T processor.
AMD released its first Phenom II processors, the quad-core “Deneb” CPUs, in January 2009. Triple-core X3 processors (whose performance falls between that of their dual-core and quad-core cousins) followed. Arriving a few months later was the 3.0GHz X4 945 processor, which brought full AM3-socket support. The benefit of this design is that AM3 will allow you to stick DDR3 memory in your system.
This $140 quad-core chip runs a mere 7 percent slower on our WorldBench 6 tests than AMD’s top-shelf, six-core Phenom II X6 1090T. The chips have similarities–the X6 is akin to an X4-class processor with two extra cores bolted on. Both share 6MB of L3 cache and run on a 2GHz HyperTransport architecture (the AMD analog to Intel’s QPI), but the X6 1090T runs at 3.2GHz, with automatic overclocking up to 3.6GHz. On Cinebench, the six-core X6 1090T outperforms the quad-core X4 945 by 60 percent.
AMD Athlon II X4 635
Benchmark Score: 110
The Athlon II X4 635 shows that omitting a processor’s L3 cache isn’t necessarily a crippling blow to performance.
Does your system really need a ton of L3 cache in order to achieve dazzling performance? For all intents and purposes, AMD’s 2.9GHz Athlon II X4 635 processor is the functional equivalent of its 3.0GHz Phenom II X4 945. Both CPUs are based on the AM3 socket, though the X4 945 sports a 15X multiplier instead of the Athlon II X4 635’s 14.5X, for a slight difference in clock speeds.
The $120 X4 635 runs on a smaller die size, compared with the Phenom II X4 945. The biggest difference between the two chips is the omission of the Phenom II family’s L3 cache. This doesn’t affect the overall WorldBench 6 score of the Athlon II X4 635, however, as it’s able to tie the Phenom II X4 945’s score of 110. That said, Cinebench results show that the latter CPU is a stronger multitasker, but only to the tune of a 9.5 percent increase in performance.
Intel Core i5-750
Benchmark Score: 106
You don’t get multithreaded processing power with the i5-750; as a result, this quad-core suffers on multitasking tests.
This 2.66GHz quad-core CPU, a member of Intel’s Lynnfield family of chips, is based on the same characteristics as the previously mentioned Core i7-870 CPU, with one key difference: This CPU has no hyperthreading–its four physical cores are all you’re going to get, rather than the eight “virtual cores” that would otherwise appear in your operating system when using a Core i7-870.
How much does that difference affect this $199 chip’s performance? Our WorldBench 6 tests showed a 17 percent drop by the i5-750 from the i7-870, which is almost double the 9 percent difference between the two CPUs’ stock clock speeds (the i5-750, with Turbo Boost, can raise its frequency to 3.2GHz when needed). The difference on our multithreaded Cinebench test is more pronounced, as the i5-750 delivers a score 40 percent lower than that of the i7-870.
In addition to the chip’s dual-channel memory controller (which means only four DIMM slots), the socket-1156 CPU’s internal PCI Express controller can dish out one full x16 PCI-E connection for graphics or split two x8 connections. That won’t affect your performance unless you plan to run a pair of graphics cards in parallel.
AMD Athlon II X2 255
Benchmark Score: 101
This low-end CPU isn’t that much slower than the Phenom II X6 1090T, except with multithreaded apps, where it’s game over.
AMD’s $75, 3.1GHz dual-core Athlon II X2 255 doesn’t come with any onboard L3 cache. And its L2 cache (memory that’s typically smaller, faster, and located closer to the core) is split into 1MB per core. As a 45-nanometer, Socket AM3 model, it’s backward-compatible with any motherboard based on Socket AM3 or AM2+.
As for performance, when compared with AMD’s top-shelf, six-core Phenom II X6 1090T, the X2 255 is only 15 percent slower on our general tests–but 70 percent slower on Cinebench, our multicore-focused measure. For another comparison, although the X2 255 runs at a higher clock speed than the quad-core, 2.9GHz X4635, the latter’s two extra cores help it deliver an 8.9 percent increase in performance on our WorldBench 6 tests, and a whopping 83.6 percent increase on our Cinebench test.
More: CPU Terms, New Chip Tech on the Horizon, Server Chips
Discussions of CPUs can quickly sound pretty technical. Here are some of the most-common terms used in describing processors.
Cache: Internal memory in the CPU, housing frequently accessed data and instructions for superfast turnaround on memory requests.
Chipset: Components that work together, alongside the CPU, to form a cohesive computing unit.
Clock speed: The speed at which a CPU executes its instructions, usually measured in gigahertz (billions of cycles per second); also known as the chip’s “frequency.”
Code name: A company’s nomenclature for a new series of chips; may refer to a processor lineup (“Gulftown” or “Deneb”), a microarchitecture (Intel’s “Nehalem”), or a platform (AMD’s “Dragon”).
Core: The part of the processor that reads and executes program instructions.
Die: The physical surface area on which a semiconductor circuit is fabricated. Smaller die sizes reduce chip manufacturing costs and chip power consumption.
GPU: Graphics processing unit-the chip that processes graphics and video. It may be located on the CPU, as part of the motherboard chipset, or on a separate (“discrete”) graphics card.
Nanometer: One billionth of a meter; the unit used to gauge the distance between the narrow connections tying transistors on a CPU.
Socket: The electrical interface on the motherboard, where the CPU sits; usually backward-compatible over chip iterations.
New Technology on the Horizon
What does the future of CPU technology look like? In a word, integration. Expect Intel to push toward stronger integration of its CPUs and GPUs, as in its Core i3 and i5 lines. The 32-nanometer Sandy Bridge CPU is the next “tock” in the company’s “tick/tock” development strategy, in which it refines the microarchitecture on one hand and launches a brand-new design on the other. Beyond Sandy Bridge is the 22-nanometer Ivy Bridge.
On Sandy Bridge processors, the GPU will actually rest on the same hunk of silicon as the CPU-right now, the Core i3, i5, and i7 platform features a 45-nanometer graphics core that’s split onto a separate die. A new Advanced Vector Extension instruction set-an evolution of the set for Nehalem processors-should bring performance boosts to media encoding, 3D modeling, and video and audio processing.
AMD also plans to get into the CPU/GPU mash-up party (it bought GPU and chipset developer ATI Technologies in 2006). Its forthcoming microprocessor design, code-named Fusion, will come in two varieties. Llano will be a 32-nanometer quad-core design featuring a DirectX 11- compatible GPU, similar to the ATI Radeon 5000. Ontario is a 40-nanometer design that blends a two-core CPU and a DirectX 11 GPU together for ultra-low-power mobile processing-a direct shot at Intel’s Atom platform.
Llano chips should continue AMD’s Turbo Core automatic overclocking. But to lower the power draw of its chips, the company is developing a new power-gating technology to let individual cores shut down when not needed.
Have you ever considered using a server-class processor as the heart of your desktop PC? Depending on the kind of system you’re after, you might unlock even more computer power if you opt for a server-class chip instead of a consumer CPU. But you may have to fork over some extra cash for it and consider how it affects your overall system configuration.
Intel’s Xeon-class processors can be powerful substitutes for a regular CPU–a Xeon W3580 can go onto an analogous LGA 1366-socket motherboard. The W3580 costs the same as its consumer equivalent, the Core i7-975 Extreme Edition, and it comes with support for both server-grade Error Correction Code memory and triple-channel memory up to 1333GHz speeds.
As for AMD, you could opt for one of its latest 8- or 12-core Opteron CPUs ($1000-plus for the 12-core variant). You’ll have to pick up a G34-socket motherboard, as no server chip is directly compatible with a Socket AM2+ or AM3 design. A six-core Opteron chip might offer faster clock frequencies, but it will be a pricer option, and one that requires a Socket F motherboard. AMD’s $300 “Thuban” line of consumer CPUs may be a better option.
Phone CPUs & Specs
What’s inside your smartphone or tablet? Consumers are increasingly spending time with the computer in their pocket, rather than their laptop. As smartphones and tablets become more and more capable for both work and play, we thought no discussion of processors would be complete without considering the latest mobile offerings. So we looked at the processors in many popular smartphones and tablets, ranking them, in our estimation, from most powerful to least.
This chart (click on the thumbnail at left) starts with a column of processor models. Most smartphone and tablet processors are SoCs (System-on-Chip designs), meaning that they combine CPU, graphics, RAM, and often other features into a single multilayered package. So we list the actual CPU within each SoC, its clock frequency, its GPU, and representative products with those internals. Individual products may slightly vary the specs (increasing or decreasing CPU clock speed, or the amount of RAM, for example). In most cases, products are readily available, but a couple of the mobile processors have only recently been announced, with no shipping products yet named for them.
–Joel Durham Jr.
When you purchase through links in our articles, we may earn a small commission. This doesn't affect our editorial independence.