Intel made low-power processors really interesting a few years back by launching a couple of 65 W Core 2 Quads that offered the same performance as its 95 W models. Sure, you had to pay quite a bit more for the privilege of owning one. But the idea of slipping an extra-cool chip into a diminutive desktop or home theater machine made the premium a worthwhile consideration.
Once the company introduced its Nehalem architecture, it was able to more easily exploit available thermal headroom using Turbo Boost. By increasing clock rate according to workload, Intel didn’t leave as much performance on the table. A 3 GHz CPU, for example, could be made to run at 3.4 or 3.5 GHz when only one of its cores was active, operating faster (and hotter) with the ultimate goal of dropping back to idle sooner.
It then debuted power-optimized versions of its CPUs based on the Lynnfield design. But instead of dropping TDP and maintaining performance, the company was forced to cut its base clocks and scale back Turbo Boosted frequencies in an effort to keep power down. Even as Intel gave you less performance, it continued charging more for those S-family models in a double-blow to value. We quickly called the company out in Intel Core i5-750S: Since When Does The S Mean Slow?
The company countered our criticism by claiming the “power-optimized” models were conceptualized as complements to the dual-core Clarkdale-based CPUs. Core i5-750S was supposed to be the first quad-core Nehalem-based model able to fit within an 82 W TDP. That didn’t justify charging more for it, we answered back.

And apparently, Intel heard us. Its new Ivy Bridge-based desktop line-up consists of 14 models. Seven of those are low-power SKUs, and none of them cost more than the standard versions. Instead, Intel charges the same price, asking only that you choose between more speed and lower power.
That’s a decision we can live with because, for most of us, the vanilla 77 W models already represent a significant savings over last generation’s 95 W Sandy Bridge flagships. Choosing higher-performing third-gen Core chips becomes almost universal.

What if you’re a system builder and you need a guarantee that your Ivy Bridge-based processor won’t exceed 65 or even 45 W, though? That’s an entirely legitimate concern, particularly as the all-in-one form factor picks up steam. In that case, you simply have to give up a little speed and go with the –S- or –T-class parts.
We got our hands on a number of Core i5s to complement our coverage of Ivy Bridge in Intel Core i7-3770K Review: A Small Step Up For Ivy Bridge, including two 77 W samples, one 65 W Core i5-3550S and a 45 W Core i5-3570T. The plan is to run all four i5s through our benchmark suite to gauge the impact of scaling down power on performance, and then to determine if the slower “optimized” chips are any more efficient.
The first issue to address popped up while I was working on my launch piece last month. Before Ivy Bridge-based processors were even showing up on shelves, we were already getting reports that Intel’s retail boxes were printed with 95 W TDPs, and not the 77 W limits the company tried claiming.
Intel responded with the following:
“Third-generation Intel quad-core standard power processors have a TDP of 77 W. In some cases, you may continue to see references to a 95 W TDP. Intel has requested that original equipment manufacturers continue to design platforms based on Intel 7-series Express chipsets to a 95 W TDP target to ensure compatibility with second-generation Intel processors.”
So, platforms continue to be designed to support 95 W Sandy Bridge-based parts, but Ivy Bridge is 77 W, right?
Technically, yes. However, Intel did seem to goof up. It should have been using a 77 W spec on its boxed processors. We received the following shortly after publishing our Core i7-3770K story:

The company seemingly used 95 W to indicate platform support, when it should have been citing the specification for the Ivy Bridge-based parts themselves. So, expect to see those five models (the -3550K should probably be -3570K) listed as 77 W parts moving forward.
Core i5-3570T: 45 W
From the bottom, Core i5-3570T is our lone 45 W sample. Intel achieves its aggressively low thermal ceiling by dropping the chip’s base clock to 2.3 GHz and only allowing Turbo Boost to kick up to 3.3 GHz on a single core when headroom allows for it. With four cores active, the chip is limited to 2.9 GHz.

Core i5-3550S: 65 W
A 65 W TDP gives the Core i5-3550S the flexibility to run at a more aggressive 3 GHz base clock rate. Turbo Boost subsequently pushes the chip up to 3.7 GHz when a single thread is active. With four cores taxing the CPU, frequency is dialed in at 3.3 GHz.

Core i5-3550: 77 W
Stepping up to the highest 77 W thermal ceiling opens up enough flexibility to operate the Core i5-3550 at 3.3 GHz. Turbo Boost facilitates an aggressive 3.7 GHz ceiling, which reflects really well in single-threaded applications. When all four of the chip’s cores are active, the -3550 runs at up to 3.5 GHz.

Core i5-3570K: 77 W
Flagship of the third-gen Core i5 family, Intel’s -3570K features a base clock rate of 3.4 GHz and a maximum Turbo Boosted frequency of 3.8 GHz. It achieves 3.6 GHz with cores active (and the available thermal headroom to not violate its TDP, of course).

| Test Hardware | |
|---|---|
| Processors | Intel Core i7-3770K (Ivy Bridge) 3.5 GHz (35 * 100 MHz), LGA 1155, 8 MB Shared L3, Hyper-Threading enabled, Turbo Boost enabled, Power-savings enabled |
| Intel Core i5-3570K (Ivy Bridge) 3.4 GHz (34 * 100 MHz), LGA 1155, 6 MB Shared L3, Turbo Boost enabled, Power-savings enabled | |
| Intel Core i5-3550 (Ivy Bridge) 3.3 GHz (33 * 100 MHz), LGA 1155, 6 MB Shared L3, Turbo Boost enabled, Power-savings enabled | |
| Intel Core i5-3550S (Ivy Bridge) 3.0 GHz (30 * 100 MHz), LGA 1155, 6 MB Shared L3, Turbo Boost enabled, Power-savings enabled | |
| Intel Core i5-3570T (Ivy Bridge) 2.3 GHz (23 * 100 MHz), LGA 1155, 6 MB Shared L3, Turbo Boost enabled, Power-savings enabled | |
| Intel Core i7-2700K (Sandy Bridge) 3.5 GHz (35 * 100 MHz), LGA 1155, 8 MB Shared L3, Hyper-Threading enabled, Turbo Boost enabled, Power-savings enabled | |
| Thermal Paste | Zalman ZM-STG1 |
| Motherboard | Gigabyte Z77X-UD5H (LGA 1155) Intel Z77 Express Chipset, BIOS F7 |
| Memory | G.Skill 16 GB (4 x 4 GB) DDR3-1600, F3-12800CL9Q2-32GBZL @ 9-9-9-24 and 1.5 V |
| Hard Drive | Intel SSD 510 250 GB, SATA 6 Gb/s |
| Graphics | Intel HD Graphics 4000 |
| Intel HD Graphics 3000 | |
| Intel HD Graphics 2500 | |
| Power Supply | Cooler Master UCP-1000 W |
| System Software And Drivers | |
| Operating System | Windows 7 Ultimate 64-bit |
| DirectX | DirectX 11 |
| Graphics Driver | HD Graphics Driver For Windows 7 (15.26.8.64.2696) |
In addition to testing all four Core i5 processors, we also re-ran our results using the Core i7-3770K and Core i7-2700K using Intel HD Graphics 4000/3000.

| Game Benchmarks And Settings | |
|---|---|
| Batman: Arkham City | Game Settings: Lowest Quality Settings, Anti-Aliasing: Disabled, V-sync: Disabled, DirectX 11 Mode, 1280x720, Built-in Benchmark |
| The Elder Scrolls V: Skyrim | Game Settings: Low Quality Settings, FXAA disabled, V-sync: Disabled, 1280x720, 25-second playback, Fraps |
| World of Warcraft: Cataclysm | Game Settings: Good Quality Settings, Anti-Aliasing: 1x AA, Vertical Sync: Disabled, 1280x720, Demo: Crushblow to The Krazzworks, DirectX 11/9 |
| Audio Benchmarks and Settings | |
| iTunes | Version: 10.4.10, 64-bit Audio CD ("Terminator II" SE), 53 min., Convert to AAC audio 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) |
| Video Benchmarks and Settings | |
| HandBrake CLI | Version: 0.9.5 Video: Big Buck Bunny (720x480, 23.972 frames) 5 Minutes, Audio: Dolby Digital, 48 000 Hz, Six-Channel, English, to Video: AVC Audio: AC3 Audio2: AAC (High Profile) |
| MainConcept Reference v2.2 | 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, 2 Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV |
| Application Benchmarks and Settings | |
| WinRAR | Version: 4.11 RAR, Syntax "winrar a -r -m3", Benchmark: 2010-THG-Workload |
| WinZip 16 | Version: 16.0 Pro WinZip CLI, Benchmark: 2010-THG-Workload |
| 7-Zip | Version 9.22 beta LZMA2, Syntax "a -t7z -r -m0=LZMA2 -mx=5", Benchmark: 2010-THG-Workload |
| Adobe Premiere Pro CS 5.5 | Paladin Sequence to H.264 Blu-ray Output 1920x1080, Maximum Quality, Mercury Playback Engine: Software Mode |
| Adobe After Effects CS 5.5 | Version: CS5.5 Tom's Hardware Workload, SD project with three picture-in-picture frames, source video at 720p, Render Multiple Frames Simultaneously |
| Adobe Photoshop CS 5.1 (64-Bit) | Version: 11 Filtering a 16 MB TIF (15 000x7266), Filters:, Radial Blur (Amount: 10, Method: zoom, Quality: good) Shape Blur (Radius: 46 px; custom shape: Trademark sysmbol) Median (Radius: 1px) Polar Coordinates (Rectangular to Polar) |
| ABBYY FineReader | Version: 10 Professional Build (10.0.102.82) Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages |
| 3ds Max 2012 | Version: 10 x64 Rendering Space Flyby Mentalray (SPECapc_3dsmax9), Frame: 248, Resolution: 1440 x 1080 |
| Adobe Acrobat X Professional | PDF Document Creation (Print) from Microsoft PowerPoint 2010 |
| SolidWorks 2010 | PhotoView 360 Render 01-Lighter Explode.SLDASM (SolidMuse.com) Image Output Resolution: 1920x1080, Render: Preview Quality “Good”, Final Render Quality “Best” |
| Visual Studio 2010 | Compile Chrome project (1/31/2012) with devenv.com /build Release |
| Synthetic Benchmarks and Settings | |
| PCMark 7 | Version: 1.0.4 |
| 3DMark 11 | Version 1.0.3 |
| SiSoftware Sandra 2012 SP3 | CPU Test=CPU Arithmetic/Multimedia, Memory Test=Bandwidth Benchmark, Cryptography, Cache Latency |

It turns out that PCMark yields some of the most interesting results—and not necessarily in a good way.
If you flip back to the twelfth page of Intel Core i7-3770K Review: A Small Step Up For Ivy Bridge, you’ll notice that the Core i7-3770K scores here are higher than in that story. Moreover, the Core i5-3570K blows away the Core i7-2700K in this story and the launch coverage as well. What. The. Heck. Right? We were using a flippin’ GeForce GTX 680 for the launch, and basic HD Graphics 4000 here. You mean to tell me that, according to PCMark, the system with integrated graphics is better?
Let’s step through the sub-tests for more detail. We did some digging on this and have answers.

Alright. Productivity comes first, including Storage, Web browsing/decrypting, and Text editing components. Sandy Bridge winds up behind four different Ivy Bridge-based setups, three of which are Core i5s, and two of which employ HD Graphics 2500. It’s frankly difficult to imagine that this would reflect real-world performance, and we’ll keep these numbers in mind as we start firing up our own workloads.

Here’s where things start to get really wonky (though the oddest results are yet to come). Although the Creativity suite includes Storage, Image manipulation, and Video transcoding workloads, there’s a clear step down from HD Graphics 4000-, 3000-, and 2500-equipped CPUs.
Naturally, our first question was to Futuremark: is PCMark benefiting from Quick Sync? The company’s response: yes, using Microsoft’s Media Foundation transforms, hardware acceleration is utilized. Futuremark does not make public the weightings for each piece of the test, but it seems unlikely that one Quick Sync-enabled Ivy Bridge chip with HD Graphics 2500 should score less than half of a Quick Sync-enabled Ivy Bridge chip with HD Graphics 4000, particularly when transcoding is only one of three variables.

The Entertainment test is broader; it includes Video playback, Storage, Graphics, and Web browsing. All of these integrated graphics engines support DirectX 10, at least, so none of the tests are getting dropped. And yet, there’s still a huge gap between the top and bottom, clearly segmented by HD Graphics 4000, 3000, and 2500. We’ll simply have to accept that Graphics (gaming) and Video playback are the biggest determinants of performance in this one.

Futuremark uses three pieces for its Computation suite: Video transcoding (downscaling), Video transcoding (high-quality), and Image manipulation. Not only does the test consequently suggest that the HD Graphics 4000-equipped processors are more than five times faster than the ones with HD Graphics 2500, but, again flipping back to Intel Core i7-3770K Review: A Small Step Up For Ivy Bridge, the benchmark puts a Core i5 almost four times above a Core i7-3960X-based machine with a GeForce GTX 680.
Bottom line: because PCMark is a black box, there’s no way to see how heavily Futuremark weighs video transcoding in its sub-tests. But judging by the disparity between Sandy Bridge and Ivy Bridge, the importance placed on Quick Sync is way too high. Will this benchmark show up again in the future? Probably, but more selectively, now that we know any fixed-function transcode acceleration can throw off its results so blatantly.

In a diagnostic like Sandra 2012, Hyper-Threading gives the two Core i7s a clear advantage, even though they’re not running all that much faster than the next-lowest finisher, Core i5-3570K. From there, optimizations for power affect performance on a fairly gradual scale.

The effects are similar in the Multi-Media module, where the two Core i7s achieve superior integer and floating-point performance.

Intel has a habit off stripping off features from lower-end SKUs, and we’ve seen AES-NI serve as one of those differentiators that Intel drops. Fortunately, the Core i5s all retain their hardware-accelerated AES, so each contender achieves performance in-line with memory bandwidth.

And as we might have guessed from the Cryptography result, all of our numbers are fairly even (as we’d expect from a common memory controller operating at a consistent data rate).


All of the 95 W and 77W CPUs perform fairly similarly in Photoshop. The only drop-offs happen at 65 W and 45 W, where lower thermal ceilings keep Turbo Boost from doing as much for performance, since our test is fully-threaded.

Hyper-Threading and cache play much more of a role in Premiere Pro, where the two Core i7s get their work done in less than 10 minutes. Three of the Core i5s appear to serve up similar performance, while the 45 W model’s restrictive TDP again keeps that chip from accelerating as fast as it’d need to in order to keep up.
Remember that we’re using integrated graphics, so Adobe’s Mercury Playback Engine isn’t enjoying hardware acceleration. Finally, with CS 6, we’ll see OpenCL support start speeding up certain models from AMD, too.

Our After Effects test doesn’t really do a whole heck of a lot with Hyper-Threading, so the Core i7-2700K drops behind a couple of 77 W Core i5s, as Ivy Bridge’s architectural improvements play more of a role than the former flagship’s eight threads or larger cache.
Again, the only time performance really suffers is when the 45 W chip’s thermal envelope holds it back.

The two Core i7s rule in 3ds Max, which does utilize all of the cores we throw at it. The i5s lag back a little bit. And as you start putting the squeeze on TDP, the 65 W and 45 W Core i5s just can’t be pushed as hard.

Demonstrating similar behavior as 3ds Max, SolidWorks’ PhotoView 360 also exploits all eight of the Core i7’s threads, translating into a quantifiable performance advantage. The two 77 W i5s aren’t far behind, though. Only as you start limiting TDP does the workload take longer to finish.

Both Core i7s stand out yet again, followed by a predictable order of Core i5s organized by thermal design power.

Same story here, though now the advantage attributable to Hyper-Threading appears even more pronounced.

I was reviewing CPUs back when Intel introduced Hyper-Threading on the desktop, and let’s just say that it didn’t always turn into a performance advantage. In Visual Studio, though, it helps shave several minutes off of our Google Chrome compile job. The Core i5s (even the 45 W one) remain closer together.

Up until now, all of our tests have emphasized performance in the context of a threaded test—that is, an application able to take advantage of a multi-core processor. Those are the situations where you really want a quad-core chip able to handle eight threaded concurrently.
When it comes to converting a PowerPoint file to PDF, though, more cores don’t help. That’s why the Core i7-2700K drops to fourth place and the Ivy Bridge-based chips start jumping ahead of it. Although the improvements to Intel’s newest architecture are slight, they’re enough to affect our result by a couple of seconds.

We’re working on our WinZip 16.5-based benchmark right now, which will take advantage of OpenCL acceleration on AMD processors and make better use of multi-core CPUs. For now, though, we’re left with WinZip 16.0. Similar to the Acrobat-based test on the previous page, this metric does not do a good job of exploiting available processing resources. Not only does it take an exorbitantly-long time to finish, but it also doesn’t demonstrate much difference between our contenders. The only exception is the 45 W Core i5-3570T.

WinRAR is more efficient, but it still doesn’t show off a ton of difference between five of these six CPUs.

Freely-available 7-Zip shows off the largest gaps between processor features, by far. The two Core i7s, armed with Hyper-Threading and an 8 MB shared L3 cache, stand out as exceptional performers. The Core i5s file in after, and the 45 W CPU doesn’t even take a massive penalty for its limited thermal ceiling.

A single-threaded metric, Lame naturally reflects the best performance from the highest-clocked, most efficient architecture. But because Turbo Boost pushes these CPUs so fast when only a single thread is active, they mostly turn in very similar results. Only the lower-clocked Core i5-3570T lags behind.

The same story applies to iTunes, another single-threaded app. Slight tweaks to the Ivy Bridge architecture even allow a couple of Core i5s to outperform Core i7-2700K.

Once we switch over to a more threading-optimized test, the results fall back to what we saw in the content creation apps: Hyper-Threading gives the Core i7s a nice little boost, while the quad-core i5s file into place accordingly. Again, the only chip that lags back is the Core i5-3570T.

HandBrake tells us a similar story.

The Sandy Bridge architecture’s HD Graphics 3000 doesn’t support DirectX 11, so there’s no way to include Intel’s Core i7-2700K in these benchmarks.
However, we do get a good sense for how HD Graphics 2500 compares to 4000 (not well). The difference isn’t quite 2x; however, we wouldn’t expect very playable performance out of anything with the lower-end implementation.
Interesting also is that the Core i7 gets a slight performance boost in the graphics component of 3DMark—which typically wouldn’t be reflected by a 100 MHz disparity. More likely is that the i7’s 8 MB cache is showing a slight advantage over the i5’s 6 MB repository.







We see four very different things in these charts.
First, the fastest solution (according to the Performance suite score) is a Core i7-3770K with a Radeon HD 6570. Without question, it’s better to get a low-end discrete card than to rely on any integrated solution.
Then, the A8-3850 with built-in Radeon HD 6550D graphics comes in second place in the graphics benchmarks. Where AMD’s A8 really suffers is the Physics suite. Lackluster processor performance hammers the rate at which this threaded test renders frames, yielding a dead-last finish.
As we might expect, the Core i7-3770K with its native HD Graphics 4000 engine places third in the graphics test, achieving half the score of the discrete card. It’s also interesting to note that the same CPU sheds more than 2000 points in the Physics suite, depending on whether its GPU is turned on or not. Remember that the graphics core shares resources like cache and system memory bandwidth.
Lastly, the HD Graphics 2500 solution simply underperforms.



I ran these tests originally for our Ivy Bridge launch coverage, and then added HD Graphics 2500 results to fill in the blanks. After seeing just how bad the second-tier Ivy Bridge-based graphics engine performed, though, I stuck to just a single resolution: 1280x720.
The outcome isn’t pretty. Even in the face of relatively modest detail settings, HD Graphics 2500 isn’t something you’d want to use for gaming. At Batman’s lowest preset, our Core i5-3550 isn’t playable. In World of Warcraft, it’s both choppy and stuttery. Skyrim doesn’t offer much of an improvement.
Bottom line: whereas HD Graphics 4000 served up frames rates that made mainstream titles like WoW and Skyrim smooth enough to play at meager settings, HD Graphics 2500 just doesn’t make the proposition very compelling.

Our six processors fall within five watts of each other at idle—last-generation’s Core i7-2700K interestingly the second-place finisher.
Intel admits that it didn’t do much of anything to cut Ivy Bridge’s power consumption beyond its adoption of 22 nm lithography. It’s not surprising, then, that idle power use doesn’t really change compared to last generation.

Under load, however, the story is drastically different. A 10-minute Linkpack workload reveals 155 W maximum power consumption from our Core i7-2700K compared to 138 W on the Core i7-3770K. Incidentally, that’s just 1 W off from the 18 W separating Sandy Bridge’s 95 W TDP and Ivy Bridge’s 77 W ceiling.

Past explorations of overclocking show that tuning clock rate and upping voltage affect Ivy Bridge’s temperature much faster than Sandy Bridge’s. At stock settings, though, the difference isn’t as pronounced. And this is with the boxed cooler, too (not a big aftermarket model that’d dissipate heat more effectively).
If you’ve already read my launch coverage of Intel’s Ivy Bridge architecture, then you probably also saw that I charted power consumption across all of our benchmarks, facilitating an average, a precise time measurement, and a representation of watt-hours based on the product of those two.
The only hiccup we encountered was that Intel’s older Core i7-2700K wrapped up faster than the Core i7-3770K. This was because the -2700K doesn’t support DirectX 11, and consequently skipped through 3DMark 11. Thus, I’m leaving Sandy Bridge out of this comparison altogether.
It’s typically pretty hard to read these line charts, particularly with more than 30 or 40 minutes worth of data from five different CPUs crammed in. However, the peaks are perhaps most telling. We clearly see that the 77 W Core i7-3770K spikes the highest, followed by the Core i5-3550, the Core i5-3570K and Core i5-3550S fairly close together, and finally the Core i5-3570T.

In an effort to make it easier to digest that consumption information, I averaged together the lines for each CPU. The result is surprisingly subtle.
As we’d expect, the Core i7-3770K uses the most power. The other two 77 W models follow behind closely. Interestingly, the 65 W Core i5-3550S ends up less than 1 W behind the Core i5-3570K. The most significant reduction comes from the Core i5-3570T, which drops down to 82.9 W of system power, on average.

Of course, the compromise you make when you cut power is generally a corresponding loss of performance. We see a gradual scale down from the Core i7-3770K at just under 38 minutes to the Core i5-3570T, which takes almost 47 minutes to finish our benchmark suite.
All of those results trump what we saw in my original review of the Core i7-3770K. In fact, even a Core i7-3960X mated to a GeForce GTX 680 took more than an hour and nine minutes to wrap up. So, what the heck happened? Knowing that the major difference between that platform and these is integrated graphics, that PCMark 7 is able to exploit Quick Sync, and that the results we garnered for PCMark were so much higher than anything seen before, it’s safe to assume that the most significant time-savings comes from Futuremark's synthetic metric.

We can take those average power consumption numbers and multiply them by the fraction of an hour taken to complete our in-house suite to come up with energy used in watt-hours.
What we find is that the two fastest processors—both K-series SKUs—use their superior performance to finish up workloads faster. The fact that they use slightly more power, on average, than the purported low-power parts is completely counteracted by their ability to drop back to idle sooner.
Although Core i5-3550 is the least-efficient model in our comparison, the –T and –S parts fail to impress. There’s really only one reason to buy either SKU, and I’ll get into that as we wrap up.
We’ve seen Intel’s approach to low-power processors evolve over the past few years.
Back in ’09, the company was selling 65 W Core 2 Quads that delivered just as much speed as its 95 W models. We covered two of those chips in Core 2 Quad Gets Efficient: Enter The Q8200S And Q9550S.
Then, in 2010, Intel tried to get tricky. When we published Intel Core i5-750S: Since When Does The S Mean Slow?, the company unhappily tried to justify the existence of a low-power Lynnfield-based model to us. But it was charging an extra $60 for a Core i5 that ran slower than the non-S version, and only cut 12 W from the original i5-750’s TDP. That’s the very definition of paying more for less.
This time around, we’re again asked to accept low-power Core i5s with –S and –T suffixes that underperform the non-appended chips. But at least we’re not being assailed by higher prices on those same SKUs. The Core i5-3750T (45 W) and -3550S (65 W) are both listed as $205 products. The Core i5-3550 costs $205, while the Core i5-3570K is a $225 part.
In general, though, it’s smarter to spend an extra $20 on the unlocked K-series part that runs at higher clock rates, includes the more capable HD Graphics 4000 engine (with its correspondingly-faster Quick Sync feature), and even achieves superior energy efficiency at its stock frequency.
Why on earth would you want one of those low-power parts then? Only one reason: to cram the performance you do get from them into a smaller form factor. In some environments, a 77 W chip simply doesn’t work. Even though our average power numbers show all of these CPUs to be fairly close to each other, our load consumption results demonstrate that it’s possible to push higher-end Ivy Bridge CPUs all the way up to their thermal ceilings. By trimming voltage and frequency, Intel prevents the –S and –T models from dissipating as much heat, ideally increasing their utility in all-in-ones, HTPCs, and embedded applications.
If you’re not looking at a strict thermal limit, skip those parts altogether in favor of Intel’s Core i5-3570K. It only costs $20 bucks more and does a number of things better. The days when you could buy a lower-TDP Core 2 Quad that didn’t compromise performance are over. Today, Intel’s –S and –T SKUs are all about dipping in under power limits at the expense of speed. This time around, at least, those parts don’t add the insult of a higher price tag, too.
