Intel Core i7-2600k Review 2nd Generation Core Processor – Performance Refined

Posted Mar 6, 2011
12Shares
Share Tweet Submit

Intel’s tick-tock clock work like release schedule has given its main rivals quiet a lot of headaches. Intel’s clock is running on time (mostly) and despite a scare (the p67 sata bug) or two its keeping time.

Earlier this year Intel unleashed a salvo of new processor based on a new(ish) architecture called the 2nd generation core architecture (nee: Sandy Bridge). This builds on the success of the core design and adds several desirable features, including a highly revamped graphics core.

A little later we will look at the performance characteristics of the current flagship processor, the Core i7 2600K.

What Has Changed With 2nd Generation Core Architecture (Read: What Will Matter To You)

Intel made several changes to the processor as well as the graphics core that makes these processors so blindingly fast. The numbers will speak for this. But to the end users the most important changes are:

  1. The processors naming convention
  2. The over-clocking specifics
  3. The motherboard compatibility

1 & 2. The Naming Convention & Over Clocking Specifics

All new Intel 2nd Generation core architecture based processors will use the following naming convention.

BRAND

SERIES

GENERATION

MODEL

MODIFIER

INTEL CORE

i3/i5/i7

2

xxx

K/T/S

INTEL CORE

i7

2

600

K

The last row shows how the processor to be reviewed today is named.

The ‘K’ series processors have an unlocked multiplier, just like the Extreme Edition processors or AMD black editions. This is important as for all of these processors; this is the only way to over-clock. By making the clock generator a part of the chipset (the x67), Intel has effectively locked all the bus speeds together. Changing the base clock will change the PCI and PCI-e clocks as well making things rather impossible to over-clock via this route. The only way to effectively over-clock is by changing the multiplier. And these are only unlocked on the K series of processors. Thankfully Intel is not charging a exuberant price premium for the “K” series.

A welcome side effect of this is that now the processor and memory clocks are independent of one another. Gone are the days of paying tons for high speed memory to run it synchronously with the processor (1:1 ratio for nerds!). Your processor might be after-burning at 5ghz, while your memory will cruise at a leisurely 1066mhz without breaking a sweat!

The “T” modifier is for 45 Watt TDP processors, while the “S” modifier is for 65 Watt TDP processors. Neither the T or S series processors are available for sale right now.

Lastly processors that do not carry any modifiers are the standard multiplier locked variants of the “K” series.
Intel has also improved the turbo technology. The turbo works for processor as well as the graphics core and has the ability to work beyond the boundary set by the processor’s TDP.

3. Motherboard Compatibility

While AMD has given the AM2+ socket an extended life cycle, Intel seems to insist on a new socket for every new series of processors. Intel has changed no less than 3 sockets since the death of the venerable socket 775. Intel had the LGA 1366 socket for the Nehalem series, the LGA 1156 socket for the Clarksdale/ Lynnfield series. And now it has the LGA 1155 socket for the Sandy Bridge series of processors. And this is not the end of it! Intel will have a new LGA 2011 socket for the Sandy Bridge “E” editions processors (Extreme Editions). Intel might argue that their architecture change is so significant that it demands a socket redesign.

Under The Hood –Changes That Make Sand Bridge Scream Performance (Read: What Matters To Intel)

This is how these processors look from inside (ie: the die shot)

IMPROVEMENTS TO THE PROCESSOR CORE

1. The Micro-op cache: Instructions are stored in a special cache as they are decoded. When a new instruction is loaded on the processor, it first checks to see if it has decoded a similar instruction and uses it. This saves power and processing cycles.

2. Improved Branch prediction: Coupled with the micro-op cache is a redesigned branch prediction unit.

3. Reworked out of order execution unit: This allows Intel to save power and transistor for newer circuitry

4. AVX Instructions: Advanced Vector Extension instructions extend the SSE instruction set. It allows for faster image & audio manipulation

5. Ring Bus: This allows the L3 cache to be shared between the processor and graphics core.

6. Systems Agent: Those familiar with socket 1155 board might have seen a UEFI entry for something called ‘system agent’ voltage. System agent is the new name for everything that was called ‘uncore’ in the past. This included the memory controller, PCI-e lanes among other things. There are improvements to the memory controller that allows for better latencies as well as the power control circuitry of the processor. The systems agent also feature a full set of 16 lanes for PCI-e x16 slots, which can be split into x8,x8 for SLI or crossfire configurations.

7. Graphics Sub-system: For many of us the graphics sub-system will be a non-starter. It is not meant to play games with eye candy, heck even without eye candy. It is much better than the graphics core found in the Clarksdale processor, but nowhere near discrete video cards.

8. Media Engine: Tied to the graphics sub-system is the media engine. This helps in video transcoding, something akin to GPU transcoding that can be done using discrete video cards (from ati and Nvidia)
As you can see Intel has invented new terminologies to go along with its new architecture. They definitely add to the processor’s prowess, but most of us are interested in how fast it will go, rather than what makes it go that fast!

This image summarizes what has been mentioned above.

The New Platform X67

Together with the new socket is a new chipset the x67. It comes in two broad flavors; the P67 and the H67. A Z67 variant will also appear. There is at least one other variant (the Q67), but these three will form the bulk of the shipping chipsets. Not to be outdone by locking the clock generator on the chipset, Intel also limits the over-clocking capabilities of these chipsets. As before only the ‘H’ (and this time the ‘Z’) series enable the use of onboard graphics!

CHIPSET

PROCESSOR OC

GRAPHICS OC

P67

Yes

No

H67

No

Yes

Z68

Yes

Yes

Z68 is not out yet. Given the debacle with the x67 Sata bug (soon to be over), Z68 will probably be delayed a bit.

The Intel Core I7-2600k Processor

As promised, this is the new Intel processor.

The ‘confidential ‘moniker shows it’s an engineering sample (ES). The performance characteristics should be similar to shipping processors.

CPU-Z goodness.

Processor Specs

Speed (Turbo)

3.4 ghz (3.8 ghz)

Cores (Threads)

4 (8)

L3 Cache

8 MB

TDP

95 Watts

Graphics (Turbo)

850 mhz (1.35 ghz)

Price

Rs. ~30,000

Speed (Turbo)

3.4 ghz (3.8 ghz)

Processor Testing

Processors

Core i7-2600K

Core i5-2500K

Core i7-950

Boards

Asus Sabertooth P67 (Core ix-2xx series)

Asus Rampage II Gene (Core i7-950)

Memory

Corsair DDR3-1333mhz 2×4 GB. Total 8 GB

Corsair DDR3-1333mhz 3×2 GB. Total 6 GB (950)

Video

Ati 6950-2GB

Power Supply

Corsair HX620

Cooler

Stock! Cooler Master Hyper 212+

OS

Windows 7 x64

Given that these processors are power efficient, a hyper 212+ should yield decent over-clocking results!

*NOTE: As the tests are done on a P series chipset board, the performance of the onboard GPU can’t be assessed. Its capabilities will be explored in a later article.

Testing Philosophy

Rather than boring you, the reader, with meaningless number, the results will shown as an improvement over an equally priced Intel high end processor available today i.e. Core i7-950. Its score in all tests will be taken as baseline 100%. Scores for both stock and over-clocked Core i5-2500K and Core i7-2600K are based relative to that 100%

Test Suite

Synthetic

Sandra 2011, X264 Benchmark (HD V3), Cinebench, 3Dmark11 –Physics Test

Real World

7-Zip

Gaming

Far Cry 2,Crysis -Warhead

Legend

Core i7-950

Baseline (100%) in all charts

Core i5-2500K

Default Clock (3.3 ghz)

Core i5-2500K*

Over-clocked (4.5 ghz)

Core i7-2600K

Default Clock (3.4 ghz)

Core i7-2600K*

Over-clocked (4.5 ghz)

Synthetic Tests

Synthetic tests are designed to show off a processor’s (or for that matter any piece of computer hardware’s) prowess. New equipment usually does much better as it reflects architectural improvements. These numbers do not translate to as large real world performance gains simply because real world applications are bogged down by more than just the processor.

Sandra 2011

We at WCCF-Tech like Sandra. It has an extensive benchmarking suite which tests the processor as well as the memory sub-systems. Prepared to be dazzled by a torrent of graphs which show how powerful the new processors are as compared to their predecessor.

The 2600K handily beats the 950, while the 2500 performs on par. Considering we have a 4 cores / 4 threads vs 4 cores / 8 threads battle royale, the 2500K does very, very well. Over clocked, both processors knock the 950 out of the ring!

X264 Benchmark Hd V3

This benchmark measures the encoding performance of the processor. It offers a standardized benchmark as the clip as well as the encoder used is uniform.

The 950 is pretty much caught napping. 2nd generation stock processors do much better, over-clocked they wipe the floor with 950’s silicon!

Cinebench R11.5 Cpu Test

Cinebench is based on Maxon’s Cinema 4D. It is used to compare graphics as well as processor performance. The CPU test uses all available processors to render a 3D scene.

Nothing to see here folk, more of the same. The 950 is only able to muster its 8 thread to the level of the 4 cores of a 2500k. Everything else has already crossed the finish line

3dmark11 Physics Test

This is the latest incarnation of one of the oldest graphics benchmarking suites. The latest incarnation supports directx 11. It has a physics test that emulates physics on the processor.

Running out of superlatives here. I’ll let the graph do the talking this time!

Gaming Tests

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

1680×1050

Benchmark

Inbuilt ‘Ranch Small’ CPU

Rendering Path

DX 9 –Medium

Things do look more cramped here. The difference between stock and over-clocked processor is not as extensive as in synthetic tests. But still both 2nd generation core processors handily beat the 950. And that too in a game that is very multi core aware.

Crysis -Warhead

1680×1050

Benchmark

Customized

Rendering Path

DX 9; Physics Set to Enthusiast

Again, games are not the best way to show off performance improvements as they are bogged down by graphics sub-systems. But even here the 950 is panting, trying hard to catch up to its newer brethren, but failing in the end!

Real World

Some might argue against using 7-zip’s compression and decompression benchmark as a ‘real world’ test. But if you try and think about it for a minute, the benchmark does show how fast the program will either compress or decompress, while negating the impact of disk transfers.

7 -Zip Compression

7 -Zip Decompresison

The 4C/ 8T 950 does better here, at least when compared to a stock 2500K.

Overclocking – Hello Uefi, Bye Bye Bios!

Gone are the days of tinkering with the base clock, memory dividers and arcane voltages. Intel has made over-clocking as easy as 1,2,3 (or A,B,C). Gone are the days of the good old BIOS (at least if you are using x67 Asus board; Gigabyte and Intel still continue to use the old BIOS, or BIOS like interface in the case of Intel).

As can be seen from the charts above, the 2600K was able to reach 4.5ghz on stock cooler, the same frequency that a 2500K was able to achieve.
Oh, UEFI stands for Unified Extensible Firmware Interface and is the fashionable replacement for the old Basic Input Output System (BIOS). On most incarnations (except Intel) it features a nice mouse driven interface.

The 2600K clocks to 4500ghz with its stock cooler with a bump to the core voltage.

The 2500K does the same, albeit with lower core voltage. The hyper-threading nature of the 2600K needs more juice to attain stability.

Overclocking On After Market Coolers

We coupled the Core i7-2600K to a Cooler Master Hyper 212+ (a cooler that can be had for as little as Rs. 2400/ US$ 30). We were able to achieve speeds of 4800GHz.

Conclusions

The performance gains seen here are reminiscent of those seen when Intel launched its core2 architecture with the core2 duo series of processors. The 2nd generation of core processors does things faster, while guzzling less energy. They run cooler and over-clock easier (the K series)
There are no reasons not to recommend either of these two processors. If you do a lot of image/ video manipulation the 2600K is a blessing. For those who want to game the 2500K offers amazing performance for an even more amazing price!

Share on Facebook Share on Twitter Share on Reddit