Intel's Ivy Bridge architecture does give us improved graphics performance. However, aside from a slight boost to instruction-per-cycle (IPC) throughput, the new design offers little more to enthusiasts shopping for a stronger processor. As a result, the new 3.3 GHz Core i3-3220 is fairly similar to the older Core i3-2120, which also ran at 3.3 GHz.
The situation is a little different when you look at AMD's mainstream line-up. The company recently introduced its FX-4170, which features a base close 600 MHz faster and a Turbo Core frequency up to 500 MHz faster than FX-4100. Granted you step up to a 125 W thermal ceiling when you embrace AMD's dual-module solution, and you pay $10 more. But we expect enthusiasts chasing performance to gladly make those compromises when a better experience is available. So, is it?

New action in the budget-oriented processor space makes for an interest re-match. Ever since introducing its Sandy Bridge design, Intel has enjoyed a quantifiable advantage in the gaming space, its Pentium G860 beating AMD's Phenom II X4 955 and FX-8120 in our suite using high-end graphics (Picking A Sub-$200 Gaming CPU: FX, An APU, Or A Pentium?) Conversely, we've shown that an FX-4100 can keep pace with Intel's Core i3-2100 when you use a less expensive GPU, which becomes the bottleneck in games (AMD FX Vs. Intel Core i3: Exploring Game Performance With Cheap GPUs).
So, with the introduction of the Core i3-3220, we thought it'd be prudent to revisit both companies' $125 offerings to see how they do in terms of gaming and productivity.
| AMD FX-4170 | Intel Core i3-3220 | |
|---|---|---|
| Architecture: | Bulldozer | Ivy Bridge |
| Manufacturing Process: | 32 nm | 22 nm |
| Cores (Threads): | 4 (4) | 2 (4) |
| Base Clock Rate (Maximum Turbo): | 4.2 (4.3) GHz | 3.3 GHz |
| Processor Interface: | Socket AM3+ | LGA 1155 |
| L3 Cache: | 8 MB | 3 MB |
| Thermal Envelope: | 125 W | 55 W |
| Online Price: | $120 (Newegg) | $130 (Newegg) |
The match-up we have is both interesting and asymmetrical. The two processors are capable of executing four threads concurrently. Intel achieves this with two physical cores equipped with its Hyper-Threading technology to exploit underutilized resources, while AMD's FX-4170 employs two Bulldozer modules sporting a pair of integer cores, a shared floating-point unit, and a bunch of other shared resources. The FX-4170 includes 8 MB of shared L3 cache, while the Core i3 has 3 MB. AMD's FX-4170 operates at a base clock rate 900 MHz higher than Intel's offering, and it can accelerate a full 1 GHz higher under the influence of Turbo Core. The Core i3-3220 doesn't benefit from Intel's Turbo Boost technology at all, but instead relies on an architecture able to execute more instructions per cycle than AMD's. The entire FX family comes equipped with an unlocked ratio multiplier, useful for overclocking, while all of Intel's Core i3s don't accommodate overclocking at all, really.
If you consider the specifications on their own, the Core i3-3220 looks completely outclassed. But because the Ivy Bridge design enjoys far superior IPC than AMD's best effort, each core is made all the more effective, despite a substantial frequency deficit. To that point, there's also a colossal disparity in the power these two chips dissipate. The FX-4170 has a 125 W TDP, while the Core i3-3220, manufactured at 22 nm, has a 55 W ceiling. That's less than half of the FX.
We're using the same memory and storage subsystem on both platforms in order to normalize as many variables as possible. In addition to the Core i3-3220 and FX-4170, we're also including Intel's Core i5-3550 in our benchmark results. This gives us a reference point in order to gauge how these $125 budget-oriented CPUs measure up to a $200 model.
| Socket AM3+ | LGA 1155 | ||||
|---|---|---|---|---|---|
| CPU | AMD FX-4170 (Zambezi), 4.2 GHz Base Clock Rate, 4.3 GHz Turbo Core | Intel Core i3-3220 (Ivy Bridge), 3.3 GHz, Hyper-Threading enabled | |||
| Motherboard | Gigabyte 990FXA-UD5 Socket AM3+, Chipset: AMD 990FX | ASRock Fatal1ty P67 Performance LGA 1155, Intel P67 Express PCH | |||
| Networking | On-Board Gigabit LAN controller | ||||
| Memory | Corsair Vengeance, 2 x 4 GB, 1866 MT/s, CL 9-9-9-34-2T | ||||
| Graphics | AMD Radeon HD 7970 925 MHz GPU, 3 GB GDDR5 at 1375 MHz | ||||
| Hard Drive | Western Digital Caviar Black 750 GB 7200 RPM, 32 MB Cache, SATA 3Gb/s | ||||
| Power | ePower EP-1200E10-T2 1200 W ATX12V, EPS12V | ||||
| Software and Drivers | |||||
| Operating System | Microsoft Windows 7 x64, Service Pack 1, KB2645594 and KB2646060 installed | ||||
| DirectX | DirectX 11 | ||||
| Graphics Drivers | AMD Catalyst 12.8 | ||||
| Benchmark Configuration | |
|---|---|
| 3D Games | |
| Battlefield 3 | Campaign Mode, "Operation Swordbreaker" 60-Second Fraps Run Ultra Quality Defaults (4x AA, 16x AF), 1920x1080 |
| DiRT Showdown | Run with -benchmark example_benchmark.xml, Modified to two laps and no competing cars Ultra Quality Preset, 8x AA, 1920x1080 |
| Elder Scrolls V: Skyrim | Update 1.4.27, Celedon Aethirborn Level 6, 25 Seconds Fraps DX11, Ultra Details, 8x AA, 16x AF, FXAA enabled, 1920x1080 |
| StarCraft II | Custom map "Tom's Hardware Guide", 60 seconds Fraps Ultra Details, Extreme Quality, 1920x1080 |
| Metro 2033 | Full Game, Built-In Benchmark, "Frontline" Scene DX 11, High Preset, 4x MSAA, 4x AF, PhysX/DOF Disabled, 1920x1080 |
| Audio/Video Encoding | |
| iTunes | Version 10.4.1.10 x64: Audio CD (Terminator II SE), 53 minutes, default AAC format |
| Lame MP3 | Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s) |
| HandBrake CLI | Version 0.95: "Big Buck Bunny" (720x480, 23.972 FPS) 5 Minutes, Audio: Dolby Digital, 48 000 Hz, Six-Channel, English, to Video: AVC Audio: AC3 Audio2: AAC (High Profile) |
| MainConcept Reference | Version: 2.2.0.5440: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, Two-Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV |
| Productivity | |
| Adobe Photoshop CS5 | Version 12.1 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates |
| Autodesk 3ds Max 2012 | Version 12.0 x64: Space Flyby Mentalray, 248 Frames, 1440x1080 |
| WinZip | Version 15.5 Pro: THG-Workload (650 MB) to ZIP, command line switches "-a -ez -p -r" |
| WinRAR | Version 4.1: THG-Workload (650 MB) to RAR, command line switches "winrar a -r -m3" |
| 7-Zip | Version 9.22: THG-Workload (650 MB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5" |
| ABBYY FineReader | Version 10.0.102.82: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages |
| Synthetic Benchmarks and Settings | |
| 3DMark 11 | Version: 1.0.1.0, Benchmark Only |
| SiSoftware Sandra 2012 | Version 2011.10.17.80, CPU Test = CPU Arithmetic / MultiMedia, Memory Test = Bandwidth Benchmark |

Particularly when we use 3DMark 11's Extreme preset, the overall suite score is almost entirely dependent on graphics performance, and so the benchmark turns back near-identical results.
The Physics subtest, however, is very much processor-bound. Because the Core i5 shows up to this fight with four x86 cores, the threaded metric hands it a decisive win. The two quad-threaded models are more evenly matched.


SiSoftware's Sandra diagnostic puts the FX-4170 and Core i3-3220 fairly close together in several measurements as well. The one exception is Sandra's Multi-Media Int x16 iSSE4.1 test, where four integer clusters help facilitate theoretical performance close to what a Core i5 is able to achieve.


We know that the iTunes AAC conversion process is single-threaded. And although it's possible to make Lame a more parallelized workload by running multiple conversions concurrently, our benchmark isolates the performance of one core.
Not surprisingly, then, the FX is beaten handily. Its higher clock rate and modest Turbo Core state cannot make up for the fact that Intel's Ivy Bridge architecture gets substantially more done per clock cycle. As we've known for a year now, AMD's design relies on parallelism in order to compete.


When we give Bulldozer a workload that keeps its modular design fully utilized, it fares much better. In both HandBrake and MainConcept, the FX-4170 uses its four integer cores and higher clock rate to slip right past Intel's Hyper-Threaded Core i3. AMD cannot match a Core i5, but then again, remember that it's in a different price segment, too.


Our Photoshop and 3ds Max benchmarks readily take advantage of multi-core CPUs. Naturally, then, AMD's FX-4170 again slides right past the Core i3-3220 in both metrics. Again, Intel's Core i5 serves as a reference point for what a true quad-core chip based on Ivy Bridge can do, and really, the FX does impressively well.

FineReader is well-threaded also. However, more on-die resources don't seem to help AMD at all (nor does the higher clock rate). Perhaps there is an issue with branch prediction hampering the Bulldozer module concept. In any case, the FX-4170 loses to a dual-core Core i3-3220, which in turn is decimated by Intel's Core i5-3550.



We know that the latest version of WinZip has optimizations for threading, though we've seen those kick in most aggressively on platforms with OpenCL support (specifically those with AMD graphics). An older version like 15.5 is still largely single-threaded, which hurts the FX. WinRAR and 7-Zip are better able to take advantage each processor's architecture, and we see the FX and Core i3 perform similarly, for the most part.

Battlefield 3's single-player campaign is largely graphics limited, so the fact that all three platforms turn back similar results from our Radeon HD 7970 was to be expected. At 1920x1080 and Ultra quality settings, the graphics card plays more of a role in determining performance than any of these three CPUs.

We took some time to develop a DiRT Showdown test methodology that produces consistent numbers, and the results here suggest the game is, like Battlefield, fairly GPU-dependent. The Core i5 does establish a slight advantage, though it's hardly significant.

In Metro 2033, the four-core Core i5 enjoys a sizable advantage, indicating an engine able to utilize the chip's four x86 cores. Meanwhile, though the Core i3-3220 does lead AMD's FX-4170, the performance gap isn't very large.

The Elder Scrolls V: Skyrim is significantly more processor-bound than any other title in this story. The Core i5 stomps the dual-core/module competition, and Intel's Core i3-3220 runs away from the FX-4170 with very little effort. AMD's frame rates are smooth, but priced within $10 of each other, the Ivy Bridge-based Core i3 still scores a notable win.

Again, we see evidence that StarCraft II stands to benefit from a platform with a true quad-core processor. The Core i5 serves up higher minimum frame rates and a significantly better average, and the two dual-core/module CPUs tie, for all intents and purposes.
We're already intimately familiar with the Bulldozer and Ivy Bridge architectures, so nothing that we saw today is particularly surprising. Single-threaded applications are going to hum on Intel's chip, while applications able to tax the FX's four integer clusters are going to treat two Bulldozer modules more like a quad-core processor (not quite, though, as the Core i5's stellar performance demonstrates).
The real questions, then, are: what does a dramatically higher clock rate do for AMD's offering, specifically, what does that bump up to a 125 W TDP do for power consumption, and how does Intel compare, given its dual-core implementation?

Despite proof that some of the games we tested do take advantage of quad-core CPUs, the dual-core Core i3-3220 takes a lead in this discipline, mostly a result of Skyrim. The FX-4170, on the other hand, serves up better application performance, and by a larger margin. When you consider the way people use their PCs, we're inclined to put more value on the larger productivity win favoring AMD, particularly since the apps where an FX excels are threaded. Those are the workloads that require more processing power.
If we were to make our judgement on performance alone, AMD's FX-4170 would have the edge.
But there's another side to this story. It starts with power consumption, and ends with efficiency.

At idle, the FX-4170-based machine uses almost 20 W more than the Core i3-3220-based box. Under load, that gap grows to a staggering 103 W. The dual-module FX almost doubles the consumption of a quad-core Core i5, in fact.
Yes, the load comes from a largely-synthetic Prime95 run, and yes, it's unlikely you'll ever see such nasty power numbers on a day to day basis. Nevertheless, two times the power consumption really puts AMD's small performance advantage into context. Efficient, this CPU is not. Whether or not that matters to you is a personal decision.
What we're really looking forward to, though, is Vishera. Intel put its cards on the table earlier this year with its Ivy Bridge architecture, and Haswell-based chips won't show up until the second quarter of next year. More immediately, we're expecting FX CPUs based on AMD's Piledriver architecture this month. We've seen evidence that Piledriver may add up to 15% more performance in the same thermal envelope as Bulldozer. If that holds true, then a processor with two Piledriver modules in the same $125 range should help the company claw back some of the performance/watt deficit it currently suffers.
You can bet we'll revisit this topic when those chips start showing up.