The Intel Core i7 3770K Review

Posted Apr 23, 2012
22Shares
Share Tweet Submit

Intel, Chipzilla, the company that basically paid its corporate clients to stay away from AMD; no matter how you describe Intel, you must admit that of late the company has not put a foot wrong. In fact I would have to go back to the days of the Pentium 4 to find a product that was embarrassing for the chip giant. The ‘disaster’ that the Netburst architecture was, with its mile long pipeline design, must have shaken Intel from the core. If anything that was good for the company as well as consumer. Now before you decide to stop reading and jump to the comment section to scream heretic! Take a moment to read the next few lines.

Why good for consumers? When AMD launched its Athlon64 processors, which were much better than Intel’s offering of Pentium 4s it charged a premium price for its products because it knew they were good enough. The company that was once the champion of low-cost but decent performance processors was now charging premium prices for its offerings. Had Intel not been shook up, AMD would continue to have charged what it wanted without a care for us the poor consumers. Not only that, Intel’s own offering were and are to a large extent are priced (depending on market segment) about the same they were when Intel launched the Core2 series of desktop processors.

With the Netburst debacle, Intel realized that it was not worth fighting the ‘gigahertz war’. If you are old enough to remember once processors were sold by their designated speed. There was the 486-33, 486DX4-100, Pentium 66 and so on. It was only later in the days of Pentium 4 that Intel shifted to numbering its processor and with the Core series of processors the scheme of labeling processors by their speed was abandoned. It started its tick-tock cycle with a tick representing a die shrink (Sandy Bridge -> Ivy-bridge 32 nm to 22 nm) and a tock representing a new architecture (next year’s Haswell). Every year Intel has been treating us to a tick or tock and more or less has kept its schedule, even when the world’s economy was on life support, Intel was cruising along at a leisurely pace and turning in numbers that put a large smile on shareholders’ faces.

Today we have with us a ‘tick’ product representing a die shrink of Intel’s Sandy Bridge architecture. Intel calls this a tick+ as not only is this a die-shrink, Intel has tweaked the General as well as Graphics processing cores of the Ivy-bridge die-shrink as well.

The 3D Tri-Gate Factor

The general processor architecture remains largely unchanged from Sandy Bridge series. Those who are interested can look at our Intel Core i7-2600K review to over the architecture.

The biggest change from the previous generation is the deployment of Intel’s 22nm 3D tri-gate manufacturing node one which the consumers will never notice. The reduction in thermal dissipation profile is a direct benefit of this switch. The changes to the general processing capabilities include support for higher speed memory (up to 2800MHz), Low voltage DDR3 memory (DDR3L), improvements to the floating point and integer engines and more aggressive turbo frequencies.

The changes to the graphics core are much more pronounced. The GPU core now supports DirectX 11, more execution units (dubbed ‘HD4000’) and better quick sync performance. Finally improvements to system agent include support for PCI-e 3.0.

Nomenclature

Intel’s new processor will get a numerical naming scheme; 37xx, 35xx, 34xx and 33xx. The 37 and 35 series processors are quad-core offerings and will be available from launch. HD4000 however will not be uniformly implemented across the processors. For example while the 3770 processors will have HD4000 graphics core, only the unlocked ‘K’ variant of 3570 processor will have HD4000. The unlocked (non-K) variant will use HD2500. The alphabet suffix will continue to be the same (i.e.“K” for fully unlocked processors, ‘T’ & ‘S’ for low TDP variants).

Motherboard Support

The processor uses the same LGA1155 socket introduced with the Sandy Bridge line of processors. Motherboards based on the 6 series platform controller hubs (PCH) will need a EFI/BIOS update to support new processors. Intel, like always, has a brand new PCH to go with the new processor –the 7x series.

Just as before There will be Z series (over-clocks both processor and graphics cores) and H series (over-clocks only the graphics core) variants. What is missing is the P series; this was made redundant with the introduction of the “Z” series anyways. All variants will support PCI-e 3.0, integrated USB 3.0 controller. Some variants (read: Z77, H77) will support Intel’s rapid start technology (aka SSD caching), while others (read: Z75) will not.

Unless you must have PCI-e 3.0 functionality (and that your 6 series board does not, as the case will be with the majority of the older boards), the new 7 series boards will not bring any new tech to your computer setup.

The lineup… or the lack of it

Intel will probably not be able to/ or not willing to offer a top-to-bottom lineup of the new processors. If we are to believe the former it is probably due to the radically new manufacturing process. Those who toe the latter line are more likely to say that Intel wants to clear up inventory of last generation processors. In any event only quad-core offerings will be available initially followed by dual cores at a later date (H2/ Q3).

The Intel 3770K: The King of Ivy Bridge

The processor under our keen observation is the top of the line Intel offering for this generation the 3770K. With a base CPU clock of 3.5GHz and maximum single core turbo of 3.9GHz this processor is a direct performance as well as price segment replacement of 2700k. The core/thread count remains as before (4/8) as does L3 cache (8MB). The TDP has gone down to 77 Watts as compared to 95 Watts for 2700K. As has been mentioned before the GPU core has been upgraded to the new HD4000 engine.

Performance Review

Today we will be looking at the processor side of Ivy-bridge and focus on the graphics core review in a future article. Not many, reading this review, who will upgrade to the new processor, will be looking to use the integrated GPU to begin with.

Test Setup

We’ll be using a mix of synthetic and real world tests to see how the new processor stacks up to the 2600k.

  • Synthetic Sandra 2011
  • X264HD (V3)
  • Cinebench
  • 3D-Mark 11 –Physics test
  • Real World 7-zip
  • Gaming FarCry 2
  • Crysis -Warhead

These might not look like a ton of benchmarks that you will see on many review sites. But consider this: Everyone already knows this processor is faster, everyone probably also knows that the computer performance is not going to be a revolutionary step above its predecessor. Also the applications used test the various compute sub-system of the processor and thus will present as close to a complete picture of performance improvements over the previous generation. Churning out charts upon charts of data that shows exactly the same thing is redundant.

Test System

Processors –

  • Intel Core i7-2600k @stock & 4.5GHz
  • Intel Core i7-3770k @stock & 4.5GHz (over-clocked to 4.8GHz)
Motherboard Intel DZZ7GA-70K
Memory G.Skill ARES DDR3 1600/2133 MHz 2X4GB
GPU Nvidia GTX 480
PSU Thermaltake Tough Power XT 775
OS Microsoft Windows 7 x64
Ambient Temps 27C For All Tests
CPU Cooling Noctua NH-D14 With Stock Fans @ Full Speed
Case & Fans Xigmatek Pantheon/Noctua NF-P14 Exhaust

CPU-Z

Despite the recent hullaballoo over the TDP of Ivy Bridge processor, our ES reports the TDP as 77 Watts for the top of the line 3770K.

Idle & Load Temps

The system was turned on for about half hour with only background tasks running. Temperatures were measured using Hardware Monitor (1.19). For load temperatures, Wprime was run 4 times in loop prior to recording temperatures. The ambient temperature as listed above was 27C. The processor and case cooling is as listed above. The highest reported temperatures for idle and load are displayed below.

Over-clocking

Over-clocking second generation core processors (Sandy- and Ivy- Bridge) is a much simpler process (at least if you have an unlocked ‘K’ processor which we do). Simply adjusting the multiplier rather than tinkering the base-clock allows for extremely decent over-clocks. We are aiming for a speed of 4.5 GHz, a speed which we achieved with are engineering sample 2600k. One reason for this is to compare performance at over-clocked speeds.

This is the first Intel motherboard that ships with a GUI interface EFI which Intel calls ‘Visual BIOS’. Though it is a better offering than the text-based offering, it still needs some work. The EFI at boot seemed to think that one of the fans had stopped (which was not the case). A fellow reviewer had trouble getting a USB mouse to work (worked fine for us) in the EFI and had to use a PS/2 mouse which did not work once the system booted into Windows.

Apart from these kinks, the EFI is not as easy to use as that of Asus which are the best in business. That being said we were able to over-clock our 3770k to 4.5GHz with little effort.

The visual BIOS (EFI) has a plethora of settings including some decent ones for temperature monitoring and fan control. We’ll have in-depth coverage of the motherboard in a later review.

The temperatures at 4.5GHz are shown below. Test conditions are same for those as stock clock.

Maximum over-clock achieved

After a littler visual bios (EFI) tinkering we were able to achieve an over-clock of 4.8GHz. Though Windows stability was a little spotty; we were able to run Sandra tests but BSOD at times (once) while running 3Dmark11 CPU test.

Over-clocked Performance

We compared the performance of both the 2600k and 3770k at 4.5GHz using 7-zip, and Cinebench

7-zip Compression

The difference between the two processor when over-clocked shrinks to 12% from about 14.5% at stock. The values in red reflect stock clock performances.

7-zip Decompression

The difference between the two processor when over-clocked shrinks to 10% from about 12% at stock. The values in red reflect stock clock performances. In this benchmark the 2600K seems to gain more from OC as compared to 3770k

Cinebench

The performance delta again shrinks to 10% (from 13%). It seems that factors other than clock speeds interfere with higher clocked Ivy Bridge processors.

Over-clock impressions

The processor is as easy to clock as Sandy Bridge offerings. However at 4.5GHz, the performance delta between 2600k and 3770k shrinks by an average of 2%. It seems as if the new processor is being held back by. We are using 1600MHz memory, that could be one factor limiting the performance.

As all processors have different abilities to over-clock we would not emphasize purchasing a CPU based on its ability to over-clock. Stock clocks are what should be concern to you and those are coming up next:

Testing

 

Cinebench R11.5

This nifty application offers GPU as well as CPU tests. For CPU it tests various processing subsystems.

We have added the performance difference delta on the charts. Here, 3770K is about 13% faster as compared to 2600k.

X264HD V4.0

Pass 1 of this test measures processing abilities of the CPU, while Pass 2 reveals cache and memory controller performance

The performance improvements are around the 12% mark in this test.

3DMark 11: Physics

This test tests the processing capabilities of the CPU.

This time around the 3770K is able to garner an 18% lead over its predecessor. Architectural changes together with the bump in the clock speed are most likely responsible for this largish performance difference.

7-Zip

I’d like to consider this a ‘real’ world test. It measures in real-time how well the application can compress and decompress.

Compression benchmark

We are at the 15% mark, about as much as speculated prior to the release of the new processor

The difference is smaller this time around, hovering at 10%

Sis Software Sandra 2012SP3

Sandra though not as popular as other benchmarking and testing suites offer an excellent set of processor benchmarks. They are reproduced below.

Arithmetic

Inter-core Efficiency

Multimedia

Cryptography

Power Management Efficiency

Generally the 3770K out performs the 2600K.

Gaming

Games are what fast processors are all about. Well for most of us anyways. As games usually rely on the performance of the graphics card, high resolution testing is not an option. Gaming tests are run with low eye candy and using low(ish) resolutions.

Far Cry 2

  • Far Cry 2 1680×1050
  • Benchmark Inbuilt ‘Ranch Small’ CPU
  • Rendering Path DX 9 –Medium

The Dunya engine powering the game is multicore aware. The difference between the two processors is 5%. (Intel 2600k performance listed above as 100%). The difference is more due to improvements in clock speeds than architectural advances.

Crysis: Warhead

  • Crysis: Warhead 1680×1050
  • Benchmark Customized
  • Rendering Path DX 9; Physics Set to Enthusiast

Things are pretty much similar when running Crysis: Warhead.

Memory Speed & Processor Clock Comparison

To see how the processor reacts to memory speed we plugged in a set of G.Skill ARES 2133MHz memory. The amount of memory remained the same i.e. 8 GB (2x 4GB). The table below shows the performance difference at stock as well as 4.5GHz when using 1600 and 2133 MHz memory.

Test

1600MHz

2133MHz

Δ%

1600MHz-4.5GHz

2133MHz -4.5GHz

Δ%

X264HD (P2)

43.38

43.98

1.36%

50.02

50.43

0.81%

Cinebench

7.93

7.99

0.70%

9.01

9.12

1.20%

7-Zip C

23050

23793

3.12%

26817

27382

2.06%

7-Zip D

23055

23279

0.96%

26239

26683

1.66%

3DMark-11**

10032

11046

9.17%

10902

12288

11.27%

*P2= Pass 2; C = Compression; D= Decompression; **Physics test only

We only used pass 2 of X264HD which is truer representation of memory speeds performance. The biggest jump in performance is seen in the 3DMark 11 physics test. A jump of 9% and 11% at stock and over-clocks respectively. The gains are greater in 3 of the 5 tests (Cinebench, 7-Zip Decompression and 3DMark).

Conclusion

It was a foregone conclusion that Ivy Bridge would be faster as compared to Sandy Bridge processors. The difference is as was expected, hovering between 10-15% depending on the application being run. The processor over-clocks just as well as previous processors. What is of concern is that the performance delta at over-clocked speeds actually shrinks. This can be either due to memory speed (which is set to 1600MHz for all tests) or a kink in the system that has not been figured out (though unlikely). Over the next few weeks, when we have had more time with the processor we’ll explore this further.

We are yet to test the on-die GPU. The performance gains from HD2500 are said to be greater as compared to processor core gains. We think that anyone who buys a top of the line processor and is reading this is unlikely to use the iGPU. But just to be thorough, we will test the integrated GPU as well in the near future.

Share on Facebook Share on Twitter Share on Reddit