[ Print Article! ]

Intel's Core i7/Core i5 'Lynnfield' Processors
September 07, 2009

Summary: Don't want to fork over $400+ for a Core i7-920 CPU and X58 motherboard? If so, we've got good news for you. Today's arrival of Lynnfield makes many of the original Bloomfield-based Core i7 CPUs practically obsolete. They can OC like mad, and offer more aggressive Turbo Modes, and the P55 platform accompanying Lynnfield is much cheaper too.

All isn't perfect with Lynnfield however. There are a few reasons why you may want to opt for Bloomfield. We'll discuss all these topics and more inside!


Intel's Core i7/Core i5 Lynnfield ProcessorsPage:: ( 1 / 19 )

[image]

<% print_image("01"); %><% print_image("02"); %>

Nehalem’s new architecture shatters performance records

Last November, Intel shook up the CPU landscape with the introduction of their first Core i7 processors based on their then new next-generation Nehalem microarchitecture.

Nehalem brought with it a number of firsts for Intel. It was the first Intel processor to ship with an integrated memory controller -- previously this function was found on the North Bridge of the system chipset. By moving the memory controller from the chipset directly onto the processor itself, memory latency is dramatically reduced, thus improving performance. Intel actually one-upped AMD by outfitting the first Nehalem CPUs with a triple (rather than dual) channel memory controller.

Nehalem was also Intel's first processor to feature an L3 cache. 8MB of L3 cache is shared amongst all four processing cores. This last level cache stores data that the processor may need to use later, saving time as the CPU doesn’t have to tap into slower system memory.

Intel also introduced Turbo Mode with Nehalem. One lesson both Intel and AMD have learned the hard way is despite their work over the years with game developers, only a small percentage of games on the market today take advantage of more than two processing cores. In these cases, the other cores are essentially wasted.

With both AMD and Intel moving from two, to four, and next year, six processing cores, there’s very real potential that many of these additional cores will sit idling completely untapped by the software.

In these cases, Nehalem uses Turbo Mode to provide an instant performance boost. With Turbo Mode, the CPU can automatically shut down the cores that aren’t being used and overclock the core(s) that are being taxed. A built-in power control unit (PCU) watches the processing cores for aspects such as CPU utilization and temperature. If it detects that only one core is being used, it shuts off the other three cores and can bump up the speed of the one active core by up to 266MHz; so the one active core on a 2.67GHz Core i7-920 for example is clocked up to 2.93GHz. If the PCU detects that the core's power usage, current, or temps are too high at that level, it will automatically drop the active core down to just one speed bump (+133MHz), knocking you down to 2.80GHz.

We’ll be discussing Turbo Mode in much more depth later in this review.

[image]

<% print_image("03"); %><% print_image("04"); %>

Nehalem was also Intel’s first native quad-core CPU. Previously Intel merely slapped two dual-core processors together onto the same package, linked by the front-side bus (FSB). This native design improves communication between the four cores.

The final significant first Intel introduced with Nehalem is its Quick Path Interconnect (QPI), which replaced the FSB. QPI is a high-speed, point-to-point interconnect that provides connections between processors in multi-processor server systems and also links the CPU to the North Bridge (IOH chip) of the system chipset. Each QPI link boasts up to 6.4 Gigatransfers/second (one link is the equivalent of 12.8GB/sec of bandwidth), since it’s bi-directional, QPI effectively delivers 25.6GB of total bandwidth. In comparison, the latest Penryn Core 2 Quad processors relied on a 1333MHz FSB which provided up to 10.7GB/sec of peak bandwidth.

Intel incorporated a number of other additions and new features into Nehalem, including better prediction of branch instructions, SSE 4.2, and the return of Hyper-Threading, but for the sake of space, we’re going to suggest you flip back to our Nehalem preview article for more info on the rest of the CPU’s features. The bottom line is that thanks to its new architecture, Nehalem was able to exceed all our expectations when it came to performance. The CPU was universally hailed as being the world’s best processor by media outlets across the globe.


Nehalem goes mainstream

Intel’s first Core i7 Nehalem CPUs relied on their high-end Bloomfield core. From the get-go, Bloomfield was intended for use in the fastest, most expensive high-end PCs. Intel refers to Bloomfield as the processor for “extreme gaming and enthusiast platforms”.

These flagship processors relied on pricey motherboards based on Intel’s X58 chipset and supported triple-channel memory (you could actually run the processor in single or dual-channel memory mode, but this reduced performance). At launch X58 motherboards sold for over $300, and triple-channel memory kits would set you back over $120: that’s well over $400 before you even factor in the cost of the processor!

Fortunately prices on X58 motherboards have come down significantly in the past 9 months, you can now find motherboards and triple-channel memory kits for about half of what you would’ve paid when Core i7 launched last November. These lower prices have helped boost Bloomfield’s adoption rate among gamers and enthusiasts significantly, but it’s still out of reach for the mainstream consumer who may not want to shell out over $1,000 for a decent Core i7 PC based on the Bloomfield/X58 platform.

For this user, Intel has concocted their Lynnfield core with P55 chipset.

Like Bloomfield, Lynnfield is based on Intel’s Nehalem architecture, only it’s been tweaked to make the CPU+platform more affordable for the mainstream market. Instead of Bloomfield’s triple-channel memory controller, Intel removes one of the memory controllers in Lynnfield, knocking it down to a dual-channel design. As a result, you only need 2 sticks of memory instead of 3 for the best performance.

At the platform level, the P55 chipset relies on a single chip, instead of the traditional North Bridge/South Bridge (2-chip) design of X58. For motherboard manufacturers, an expensive 8-layer PCB with six DIMM slots for memory is no longer needed for your high-end mobo (the cheapest X58 motherboards sometimes rely on 6-layer PCBs), a 6-layer PCB is all that’s required with just 4 DIMM slots (4-layer PCB if you’re willing to forego SLI/CrossFire).

Those are just a couple of the changes between Lynnfield and Bloomfield though. Read on for more details…



DMI, PCIe, and memory controllerPage:: ( 2 / 19 )

Intel’s designed Nehalem to be completely modular. As such, features can be added or removed from the core architecture to service intended markets. Server CPUs for example can feature more cache, more cores (up to six currently), more QPI links, etc than traditional desktop Nehalem variants like Bloomfield or Lynnfield.


DMI takes center stage

In the case of Lynnfield, to cut costs Intel takes the basic Bloomfield core and removes features like the QPI link and triple-channel memory controller, and replaces it with DMI (Direct Media Interface) and a dual-channel controller.

DMI has actually been in use for years now in previous Intel chipsets. It’s a chip-to-chip interconnect that Intel has used previously to link the North Bridge chip (the IOH) with the South Bridge chip in their chipsets. In the case of the X58 platform, DMI is used to link the X58 IOH with the ICH10R South Bridge chip. You can see DMI in action on X58 in this block diagram:





As you can see, the DMI bus is limited to 2GB/sec of bandwidth. That pales in comparison to QPI’s 25.6GB/sec. However, it turns out that 25.6GB/sec of bandwidth is overkill for most desktop users. Unless you’re running a server with multiple processors that need to communicate with each other, or you’re running multiple graphics cards (2-Way, 3-Way SLI/CrossFire or Quad SLI/Quad CrossFire) most users won’t tap into the potential bandwidth advantages QPI offers.

With one x16 PCIe graphics card potentially drawing up to 16GB/sec of bandwidth though, DMI wasn’t sufficient for handling PCIe graphics duties.

To solve this problem, Intel’s integrated PCI Express functionality directly onto the Lynnfield CPU. You can see it on the right-side of Lynnfield in this die shot:



Now compare the Lynnfield die layout to Bloomfield’s layout:



Lynnfield supports up to 16 PCI Express 2.0 lanes, providing single-card or dual-GPU functionality. The 16 lanes can be configured in x16 (in the case of 1 card) or dual x8 (in the case of two graphics cards) configurations, with full support for CrossFire and SLI. Additional PCIe functionality is handled by the P55 chipset, which supports up to 8 x1 PCIe devices (500MB/sec each). You can see how the Lynnfield/P55 platform is configured in this block diagram:





With just 16 lanes of PCIe functionality present in Lynnfield, support for 3-Way SLI/CrossFire configurations isn’t possible. Users who would like to run 3-Way SLI will have to purchase a P55 motherboard with NVIDIA’s nForce 200 chip built-in. From polling motherboard manufacturers, it doesn’t sound like there’s going to be a ton of boards to choose from for users who would like to go this route; right now we know EVGA will offer their P55 FTW 200 and P55 Classified 200 boards, and ASUS will offer workstation-oriented P55 boards with the nForce 200 chip integrated on the motherboard, but that’s about it.

From what we gather, the motherboard manufacturers generally feel that someone who will want 3-Way SLI is going to want Intel’s X58 platform. After all, X58 can natively drive anywhere from 2x16 PCIe cards to 4x8 cards.

Dual-channel memory controller

As we’ve mentioned previously, for Lynnfield Intel removes one of the three memory channels found in Bloomfield, knocking it down to 2-channels (128-bits wide). However, Lynnfield’s dual-channel memory controller officially supports faster DDR3-1333MHz memory.

Technically, Bloomfield’s triple-channel controller only supports DDR3-1066MHz RAM.

When configured with dual-channel DDR3-1333MHz memory, Lynnfield’s peak memory bandwidth is 21.2GB/sec. That’s still not as high as Bloomfield’s triple-channel 192-bit wide DDR3-1066MHz, which offers up to 25.6GB/sec of peak memory bandwidth, but it’s close.

Of course, both the P55 and X58 chipsets can run with faster DDR3-1600 and higher memory modules, but technically that’s outside the specifications of the chipset. That hasn’t stopped motherboard makers and memory manufacturers from peddling their fastest DDR3-2000MHz memory kits though. Even Intel’s gotten in on this action, with their P55 and X58 motherboards supporting memory speeds outside the official specifications of the chipset.

Like Bloomfield, Intel says DDR3 memory modules must run at 1.65V or lower. Anything higher, and you risk damaging the memory controller integrated inside your CPU.

With the third memory controller and QPI removed, Lynnfield needs fewer pins than Bloomfield – 1156 versus 1366. As such, a completely different socket is required, LGA-1156. After first launching with Prescott Pentium 4 CPUs over 5 years ago, LGA-775 is finally being replaced.


Enhanced Turbo ModePage:: ( 3 / 19 )


With Lynnfield relying on a totally different 1156-pin socket than Bloomfield, critics pointed out giving two different cores the same name is confusing, possibly leading to uninformed users accidentally purchasing the wrong motherboard for their CPU. Many just yearned for the days when you could tell what features a CPU supported based on its name: buy a Core 2 Quad for example, and you know you’re getting a 4-core processor; AMD uses the X2, X3, X4 designation to denote the number of processing cores.

So why are we harping on Lynnfield’s Core i7 designation on a page that’s supposed to be dedicated to Turbo Mode? Because as you’ll see in the benchmarks later in this article, Lynnfield’s new Turbo Mode allows it to keep up with Bloomfield CPUs in many benchmarks despite its disadvantage in peak bandwidth.

How is Lynnfield able to do this? Simple. Lynnfield’s Turbo Mode sports more aggressive scaling.

As we mentioned at the outset of this article, Nehalem’s Turbo Mode can dynamically OC the processing cores based on factors like usage, power, and temperature. In the Core i7-920 for example, if the CPU is running a single-threaded app that’s only pushing 1 of the CPU’s 4 cores, and the processor is coming nowhere near its temp and power thresholds, it can OC that 1 active core by up to two speed bins (+266MHz). Your 2.66GHz Core i7-920 automatically becomes a 2.93GHz CPU.

If only two cores are being used, and the temp and power conditions are met, it will OC the two active cores by one speed bin (+133MHz), 2.80GHz.

Turbo Mode does this by upping the clock multiplier, which is normally locked at 20 in the case of the Core i7-920 (20.0x multiplier x 133MHz base clock (bclk) = 2.67GHz). Under the max Turbo Mode scenario with only one core being used, the Core i7-920’s multiplier is increased to 22x, yielding 2.93GHz. If 2 or more cores are being taxed, Turbo Mode increases the multiplier to 21x, giving you a final clock speed of 2.80GHz.

The following chart outlines the maximum Turbo Mode speeds for Intel’s Bloomfield-based Core i7 CPUs:

Bloomfield Core i7 - Max Turbo Frequency
CPUStock Clock SpeedMax Speed w/4 Cores ActiveMax Speed w/3 Cores ActiveMax Speed w/2 Cores ActiveMax Speed w/1-Core Active
Core i7-975 Extreme3.33GHz3.46GHz3.46GHz3.46GHz3.6GHz
Core i7-965 Extreme3.2GHz3.33GHz3.33GHz3.33GHz3.46GHz
Core i7-9503.06GHz3.2GHz3.2GHz3.2GHz3.33GHz
Core i7-9402.93GHz3.06GHz3.06GHz3.06GHz3.2GHz
Core i7-9202.66GHz2.8GHz2.8GHz2.8GHz2.93GHz


As you can see in the chart above, Intel’s flagship Core i7-975 Extreme Edition can use Turbo Mode to clock itself up to 3.6GHz max. The Core i7-940 uses Turbo Mode to clock up to 3.2GHz. It’s important to note that this is the best case scenario when all the temp and power conditions are met, and only one core is being used. If you’re running a dual or multi-threaded app that can take advantage of 2 or more cores, you’re limited to just one speed bump of 133MHz, with the i7-940 clocking itself up to 3.06GHz.

For Lynnfield, Turbo Mode is much more aggressive. It can bump the clock multiplier up to five speed bins. In the case of the Core i7-870 for example, it features a clock multiplier of 22.0, with a 133MHz bclk (22x133=2.93GHz). With Turbo Mode enabled, it can up the clock multiplier to 27.0, yielding a final clock speed of 3.6GHz. That’s the same max turbo frequency as Intel’s Core i7-975 Extreme Edition!

Lynnfield Core Max Turbo Frequency
CPUStock Clock SpeedMax Speed w/4 Cores ActiveMax Speed w/3 Cores ActiveMax Speed w/2 Cores ActiveMax Speed w/1-Core Active
Core i7-8702.93GHz3.2GHz3.2GHz3.46GHz3.6GHz
Core i7-8602.80GHz2.93GHz2.93GHz3.33GHz3.46GHz
Core i5-7502.66GHz2.8GHz2.8GHz3.2GHz3.2GHz


If you’re only pushing two cores, and your CPU temps and power consumption are within tolerances, Turbo Mode will increase the clock multiplier on the 870 to 26, yielding a final clock speed of 3.46GHz. If you’re using three or all four cores, Turbo Mode will increase the clock multiplier by two speed bins, to 24.0x (+266MHz) resulting in a final clock speed of 3.2GHz.

Like Bloomfield, overclocking won’t automatically disable Turbo Mode either. So if you can keep your CPU cool enough and power consumption remains in check, you can mix the two together for max performance.



Lynnfield CPU SKUsPage:: ( 4 / 19 )

[image]

<% print_image("05"); %>

Like the Bloomfield launch, Intel plans to offer three Lynnfield CPU SKUs today. Two of them are designated as Core i7, with one Core i5 being offered for the gamer on a budget. At the high-end is the Core i7-870, this chip features a 2.93GHz clock speed, supports Turbo speeds up to 3.6GHz, ships with 8MB of L3 cache, and will sell for $555. The midrange CPU is the Core i7-860, which runs at 2.80GHz, while the Core i5-750 is clocked at 2.66GHz:

Intel's Lynnfield CPUs
CPUStock Clock SpeedMax Turbo Speed# of Cores/ThreadsL3 CacheMemory Speed SupportMax TDPPrice
Core i7-8702.93GHz3.6GHz4/88MBDDR3-1333MHz95W$555
Core i7-8602.8GHz3.46GHz4/88MBDDR3-1333MHz95W$285
Core i5-7502.66GHz3.2GHz4/48MBDDR3-1333MHz95W$199



Lynnfield is built on Intel’s existing 45-nm manufacturing process, with the CPU sporting 774 million transistors and a die size of 296mm2. In comparison, Bloomfield boasts a smaller die size of 233mm2 and a lower transistor count of 731 million transistors.

You can blame the integrated PCIe controller for Lynnfield’s increased transistor count and larger die size.

With its larger die size, in theory Lynnfield could be more expensive for Intel to produce than Bloomfield, but since we don’t know how its yields compare to Bloomfield, we can’t definitely say one core is more expensive to produce than another. If Lynnfield’s yield rate is higher than Bloomfield’s, it could actually be cheaper to produce, despite its larger die size and transistor count.

Only Intel knows the answer to this, and they’re not talking. Trust us, we asked.

What’s in a name?

Starting with Nehalem, Intel is moving to a “simplified” brand strategy with Intel using the i7 modifier to denote high performance PCs, i5 for mainstream, and i3 for entry-level PCs. Intel plans to offer a mixture of different CPU cores for each modifier. Core i7 for instance will encompass the Bloomfield and Lynnfield cores for desktop PCs, while their upcoming Clarksfield CPU core for notebooks will also be designated as Core i7 when it debuts later this year. Intel’s Director of Corporate Marketing, Deborah Conrad, has said that the first Core i3 CPUs will debut early next year. Presumably these CPUs will be based on their upcoming 32-nm Clarkdale core.

Core i7 consists of the Core i7-900 series, which utilize the Bloomfield core, and the Core i7-800 series, which are based on Intel’s Lynnfield core. Both cores are quad-core, with the addition of Hyper-Threading bringing support for up to 8 threads.

Going forward Intel will deemphasize the number of physical cores, instead the number of threads will be more important. For instance, Core i5 CPUs will eventually encompass both dual and quad-core processors, with the quad-core Core i5 CPUs lacking Hyper-Threading support, yielding a max of 4 threads (1 thread per core) while dual-core Core i5 parts will support Hyper-Threading, so they will also support 4 threads.

It’s believed that Core i3 CPUs will feature two processing cores with Hyper-Threading support (2-core/4 threads), with the big omission being Turbo Mode.

Bottom line, if you want an 8-thread processor with all the goodies, you get Core i7, 4-threads with Turbo Mode will be designated Core i5, while 4-thread processors without Turbo are Core i3s.

Next year Intel will also introduce their first 6-core desktop CPU with support for Hyper-Threading, Gulftown. Rumor has it that this chip will be designated Core i9, although by now we all know how wrong the rumors were on Lynnfield’s branding.

Lynnfield vs Bloomfield

While Turbo Mode makes Lynnfield just as exciting as Bloomfield, with the obvious addition of the more affordable P55 platform, there’s still a couple of reasons why you may want to opt for a Core i7 rig based on Intel’s Bloomfield core.

For starters, Bloomfield’s X58 platform is going to give you full 16-lane operation for both GPUs (2x16) when running SLI or CrossFire. Lynnfield’s PCIe controller is limited to 8-lanes under the same scenario (2x8).

As GPUs become more powerful, with next-generation DX11 GPUs potentially shipping with twice the shaders and considerably more memory bandwidth than today’s DX10 cards, this could potentially become a bigger issue (especially as more next-gen games appear).

Also remember that all Core i7-900 CPUs not only feature QPI, it’s also unlocked.

Thanks to its triple-channel memory controller, Bloomfield offers the highest memory bandwidth of any Intel platform. And while it’s officially limited to DDR3-1066 speeds, with the right memory modules you can easily run DDR3-1333 or higher, just like Lynnfield (all our benchmarks were taken with Bloomfield running at 1066 memory speeds, while Lynnfield testbeds utilized DDR3-1333).

Finally, only Intel’s Core i7-975 Extreme Edition gives you the flexibility of an unlocked clock multiplier when OC’ing.



OC’ing and the new LGA-1156 socketPage:: ( 5 / 19 )

LGA-1156 infrastructure

Lynnfield ships with a new 1156-pin socket. With the new socket, Intel has tweaked the locking motion of the retention mechanism slightly -- clamping the socket down over the CPU requires a bit more force -- but processor installation is still a breeze and the new system probably does a slightly better job of locking the processor in place in the socket. The LGA-1156 retention mechanism slides underneath one bolt at the bottom of the socket, holding the processor in place. You then lock it all down with the lever on the right of the socket.

Intel’s Lynnfield processors continue to ship without pins. We’re huge fans of this feature as you no longer have to worry about bending a pin on the bottom of your CPU. With 1156 pins on the bottom of the CPU, trying to unbend one pin could’ve been a pretty nerve-racking experience. Over the years we’ve never bent a pin on any of the 2, 3, 4, or 5-series motherboards we’ve received either.

In terms of size, spacing on the new LGA-1156 socket is closer to LGA-775 than LGA-1366. Holes on LGA-1156 motherboards are located 75mm x 75mm apart. In comparison, LGA-1366 holes are spaced 80mm x 80mm from each other and LGA-775 72mm x 72mm.

You can see a comparison of how the new LGA-1156 socket compares to LGA-755 and LGA-1366 in this image sourced from EVGA:



Technically, you will need a new LGA-1156 heatsink to accommodate Lynnfield, although we do know that EVGA is including mounting holes for LGA-775 coolers on their P55 motherboards. Those of you with high-end Core 2 coolers will no doubt be happy to hear this. We wouldn’t be surprised if ASUS, Gigabyte, MSI and others follow suit to catch up with EVGA on this shortly.

Besides all-new LGA-1156 designs, cooler manufacturers will also be providing LGA-1156 brackets with their existing coolers also.

You can see how the Lynnfield CPU itself compares to Core 2 Quad (Yorkfield) and Core i7 Bloomfield in terms of size in these images:

[image]

<% print_image("06"); %><% print_image("07"); %>

Memory

Memory compatibility is the same as Bloomfield, only instead of having to buy three DDR3 memory modules for optimal performance you’ll only need two. Once again you’ll want to keep your DDR3 memory voltage below 1.65V in order to avoid damaging the processor.

Memory manufacturers will be providing specialized dual-channel DDR3 memory kits in time for the Lynnfield launch. We tested a pair of 2GB modules from Kingston Technology, the HyperX KHX1600C8D3K2/4GX for our Lynnfield testing (4GB RAM total).

[image]

<% print_image("08"); %>

These modules are rated for 1600MHz speeds with timings of CL8-8-8-24 at 1.65V and carry an MSRP of $132. If you’re willing to spend a little more, Kingston will also offer 2000MHz and 2133MHz dual-channel kits offering both CL8 and CL9 timings. Their flagship parts are 4GB modules rated for speeds of 2133MHz with CL8 timings and retail for $394.

In late September, Kingston also says they’ll be launching 8GB (2x4GB) DDR3-1333 and DDR3-1600 kits.

You can rest assured that OCZ, Corsair, GeIL and others will also have their 1333MHz, 1600MHz, 2000MHz, and 2133MHz Lynnfield kits ready to go also.


Overclocking

Considering how scaleable Core i7 Bloomfield has proven to be, we were eager to see how far we could push our Lynnfield CPUs. Sure enough, they didn’t disappoint one bit. First let’s start with the runt, Core i5-750.

Before attempting our OC’ing endeavors, we made sure to manually dial in the stock voltages for the CPU and system components, just to make sure that the Gigabyte P55-UD6 motherboard we used for testing didn’t automatically bump up the voltages to achieve stability. Once that was done, we dialed up the base clock (bclk), then loaded up the OS and followed our normal procedure of testing OC stability with Prime95, Cinebench, 3DMark Vantage, 3 runs of Crysis CPU2 benchmark, and 5 runs of Far Cry 2 small ranch looped. If the CPU passes this gauntlet of tests, we then crank up the bclk again, rinse and repeat until the system becomes unstable.

So how high did we take our Core i5-750 CPU on stock voltage? 3.82GHz (20x191):

[image]

<% print_image("09"); %>

Now for the max OC with voltage adjustment:

[image]
<% print_image("10"); %>

As you can see, with 1.45V of juice, we hit 4.08GHz with our Core i5 sample. We could actually run everything but Cinebench at even higher clock speeds. Every time we attempted to run the Cinebench benchmark we’d get a BSOD.

The Core i7-870 went even further than the Core i5-750. At stock voltage, we topped out at 3.89GHz (22x177):

[image]
<% print_image("11"); %>

1.168V of juice and a 3.89GHz clock speed sure looks sweet doesn’t it? Let’s see what happens when we crank up the voltage though:

[image]
<% print_image("12"); %>

We managed to hit 4.42GHz (22x201) with 1.41V of juice. That’s the highest speed we’ve managed to achieve with a Nehalem CPU, besting even the Core i7-975 Extreme Edition we tested earlier this summer. Once again, Cinebench prevented us from going any further, but we did go ahead and grab a CPU-Z screenshot at 4.55GHz for you to look at:

[image]
<% print_image("13"); %>

Our CPU voltage for the Core i5 is probably a bit on the high side. To be safe, you’ll probably want to stick closer to 1.4V, but we’ll know more about safe/unsafe Lynnfield voltages over time as we get more attuned to the new CPU and platform.

What kind of temps did we see at these speeds? Using a Thermalright MUX-120 LGA1156 cooler, the cores in our Core i5-750 chip idled between 44-47 degrees OC’ed (stock temps ranged from 30-33 degrees Celsius) while load temps peaked at 70-75 degrees (stock load temps ranged from 48-52 degrees Celsius).

With its higher clocks, the Core i7-870 chip ran a little hotter. OC’ed idle temps ranged from 50-54 degrees Celsius across the four cores, while load temps topped out between 77-89 degrees at 4.42GHz. In comparison, temps at stock voltage were 30-34 degrees Celsius at idle, and 50-56 degrees under load.

The Thermalright MUX-120 cooler performed tremendously, although we do wish it didn’t rely on Intel’s stock retention brackets. This move will appease users who don’t want to remove their motherboard from their system case in order to install the cooler though.




P55 MotherboardsPage:: ( 6 / 19 )

Motherboard manufacturers are really going all-out with their P55 motherboards. Many of the latest P55 boards are loaded with features you won’t even find on high-end X58 motherboards! The best part is, these high-end boards will sell for less than X58, with the cheapest P55 motherboards starting at $110 (MSI’s P55-CD53) and going up above $200+ from there.

We’ve received a ton of P55 motherboards so far, with more on the way. We’re going to start with ASUS first.


ASUS P55 lineup

Of all the motherboard manufacturers, it appears ASUS will have the most diverse array of P55 boards on the market. Looking for a motherboard designed for ultimate stability and reliability first and foremost? ASUS has that with their TUF series boards, beginning with the Sabertooth P55. Meanwhile, gamers and enthusiasts will be glad to hear that Republic of Gamers returns with the Maximus III Formula. ASUS’ P7P55 motherboards will also be offered for users looking to spend a little less money, while the HTPC crowd will be getting their own micro-ATX Maximus III Gene board which emphasizes the highest-quality audio. ASUS will also be offering workstation-oriented P7P55 WS boards.

It’s the P7P55 series motherboards that will draw the bulk of the sales though. With prices ranging from $139 for the entry-level P7P55D LE, all the way up to $279 for the P7P55D Premium (ASUS’ most expensive P55 motherboard), they’ve got all the main price points covered here, with most P7P55 series boards ranging in roughly $20 increments.

One feature ASUS is really pushing with the P7P55 series is Xtreme Design. For the end user, this means you’re getting a more reliable motherboard. ASUS uses the best components including all-solid capacitors, StackCool 3, and their built-in protection for electrostatic discharge that exceeds EU standards for electronic devices, and overcurrent protection.

[image]

<% print_image("14"); %><% print_image("15"); %>

In all honesty, some of these Xtreme Design features have been found on ASUS’ older boards in the past, but this is the first time that they’re really emphasizing it. One new addition is 2-ounce copper PCB. Previously this feature was only offered on ASUS’ WS workstation boards. Meanwhile, to ease memory installation, ASUS has implemented Q-DIMM – you’ll now find plenty of room between the graphics slot and memory DIMMs, just like the X58 Rampage II Gene.

These are just a couple of the new additions found on the P7P55 series motherboards.

The most talked about though is probably going to be the TurboV EVO processor. With this chip, the motherboard can automatically do all the overclocking itself. No more fiddling around in BIOS is necessary. Simply press the button and TurboV EVO automatically determines the highest stable clock speeds and voltages for your particular CPU and memory, while still delivering optimal component temps. On the P7P55D Deluxe board, ASUS even includes a wired remote control which you can use to OC the CPU’s bclk by hand, or load one of 3 custom profiles you’ve predefined. You can also adjust the motherboard’s EPU (power) profiles with the remote.

Besides the remote, the Deluxe board also features better cooling thanks to StackCool 3+, which features a 2+2-ounce PCB, 10-channel audio, and more powerful 16+3 phase power (lesser P7P55 boards feature 12+2 phase power).

[image]
<% print_image("16"); %><% print_image("17"); %>

The P7P55D Deluxe and P7P55D EVO both ship with 3 PCI Express graphics slots, with support for SLI and CrossFire. Here we should note that the third slot doesn’t provide 3-Way SLI support, instead it’s used to drive an additional display, with four lanes max going to this slot. Dual GigE ports are also provided along with ASUS Express Gate. You’ll also find dedicated buttons for power/reset, MemOK!, and clearing CMOS. Additional expansion slots include 2 x1 PCIe slots and two PCI slots.

The P7P55D EVO will carry an MSRP of $199 while the Deluxe board will sell for $229. (We’re not including pictures of our EVO board in this article as it’s an engineering sample that includes features that aren’t found on the retail board.)

Gamers and enthusiasts looking for an even higher-end motherboard may want to opt for the Maximus III Formula. This motherboard ships with features we’ve come to expect from ASUS ROG motherboards like SupremeFX X-Fi audio and the LCD Poster, and builds on it with new networking features like packet prioritization, which can be used to give the highest priority to gaming network traffic. With their GameFirst software, merely select what priority you’d like to devote to different games and other applications, a simple slider ranges from low to high, making it easy to customize and add different programs and their networking priority.

[image]
<% print_image("18"); %><% print_image("19"); %><% print_image("20"); %>

Another new feature that’s been added to the Maximus III is ROG Connect. With this feature you can use your notebook or netbook to overclock the Maximus III motherboard. Simply install the software on your notebook, hook it up to your mobo via the supplied USB cable, and you’re good to go. Once everything is up and running, you can manipulate BIOS settings from your notebook, or even watch POST codes on your notebook as your mobo boots up.

The Maximus III Formula will retail for $259.

We’ll be all the ASUS boards in much greater depth in a dedicated article.


EVGA and Intel P55 motherboardsPage:: ( 7 / 19 )

EVGA P55 FTW

Fresh off the success of their X58 motherboards, EVGA has a full complement of P55 motherboards, with prices ranging from $169.99-$229.99 on their ATX motherboards without nForce 200 (three boards are offered in $30 increments). Unfortunately we don’t have pricing on the micro-ATX and nForce 200 boards just yet.

Their flagship board most enthusiasts will probably want to opt for though is the $229.99 P55 FTW (For The Win) board.

[image]

<% print_image("21"); %><% print_image("22"); %>

As we mentioned earlier, the P55 FTW offers mounting holes for LGA-775 and LGA-1156 coolers. This is one of the only P55 motherboards on the market to support this feature. The board also features triple BIOS support. In case one BIOS fails, you’ll always have a backup BIOS, you can also use the secondary BIOS’ to compare BIOS features or for loading custom BIOS profiles. A 3-position BIOS selector switch is located on the bottom of the board, just underneath the PCI slot for this purpose.

OC’ers looking to send as much power to the CPU as possible will welcome the second 8-pin power connector. That’s right, the P55 FTW features 2x 8-pin PSU power connectors. EVGA also places dedicated read points on the top of the board, allowing you to get accurate voltages for the VCORE, DIMM, PCH, CPU_PLL, VTT, and Ground with a voltmeter or multimeter. Another interesting feature EVGA has integrated on the P55 FTW is its PCI Express disable jumpers. Why would you want to disable a PCI Express slot? Troubleshooting. In a water-cooled SLI system experiencing graphical glitches for instance, you’d have to take everything apart before you can even begin diagnosing the problem. Thanks to EVGA’s PCIe disable jumpers, you can easily disable the PCI Express slot(s).

[image]
<% print_image("23"); %><% print_image("24"); %><% print_image("25"); %>

For those of us that like to test and use motherboards on an open testbed, EVGA has provided built-in power, reset and clear CMOS buttons for quite some time on their boards. They’re now taking this to another level with ECP V2. This is an external card with those buttons built-in, as well as the debug LED, PCIe slot disable jumpers, and buttons for increasing or decreasing the CPU core voltage and VTT. Now you can easily boot up your P55 FTW testbed from the comfort of your chair without having to get up! A ribbon cable is used to connect the ECP V2 module to the motherboard.

Speaking of the debug LED, not only does it display hex codes during POST, it can also be used to monitor CPU temps once the motherboard has booted into the OS.

The EVGA P55 FTW features a 12-phase power design with support for SLI and CrossFire. Once again, a third PCI Express graphics slot is provided, but this is an x4 slot. P55 motherboards with 3 PCI Express graphics slots can run in either 1x16 or 2x8 and 1x4. Also provided are two PCI slots and one x1 PCIe slot. A heatsink with pulsating LED is located right next to the x1 PCIe slot.

The EVGA P55 FTW looks like another winner from EVGA.


Intel P55 Kingsberg

While Intel motherboards have come a long way from the days when their boards absolutely refused to offer OC’ing options in BIOS, they’ve still got a long way to go to catch up to the other P55 motherboards on the market. Intel knows this too.

As far as they’re concerned though they don’t need or even want to offer the most feature-packed, OC-friendly motherboard on the market. Instead they’re more concerned with selling the world’s best processors.

At the same time though Intel realizes there’s a market need for a baseline reference motherboard that can drive the motherboard manufacturers forward. Rather than getting their motherboards from ASUS or Gigabyte, many OEMs also insist on using Intel motherboards in their PCs.

[image]

<% print_image("26"); %><% print_image("27"); %><% print_image("28"); %>

Considering this, Intel plans to offer a range of P55 motherboards, including two micro-ATX designs and two ATX boards. Enthusiasts will definitely want to opt for the Extreme Series boards, as they’re the only Intel boards to offer full overclocking features in BIOS. The two Extreme Series boards are the ATX DP55KG Kingsberg desktop board, and the DP55SB Sharpsburg motherboard.

Both motherboards offer support for SLI and CrossFire, with debug LED display, and built-in Bluetooth support. On the backplate of the Kingsberg board you’ll find two eSATA ports, 8 USBs, 1 GigE connector, and full audio connectivity, including S/PDIF in/out.

The most interesting feature we found on the board was actually its tiny heatsink for the P55 PCH chipset. You’ll also notice the lack of support for legacy devices like parallel ATA IDE drives, and keyboard/mouse PS2 ports. This board, like other Intel boards from the past several years we’ve tested, is all USB/SATA.



Gigabyte P55 motherboardsPage:: ( 8 / 19 )

Last but certainly not least is Gigabyte, with their dazzling array of P55 motherboards.

[image]

<% print_image("29"); %><% print_image("30"); %>

Gigabyte’s really shaken things up with their P55-UD6 board, which features a 24-phase power design. Yes, you read that right, 24 power phases. That’s considerably more phases than other P55 motherboards. Gigabyte says their 24-phase power design “provides a fast transient response, providing quick and seamless power delivery during extensive CPU loading variations”. Faster transient response delivers smoother, more stable power to the CPU. The other key benefit of integrating 24-phases on the board is lower temps for the VRM circuitry, as the workload is spread across all 24 phases. This helps to ensure the longevity of the motherboard, and helps to lower component temps. Don’t forget this becomes even more important as you crank up the CPU clocks during OC’ing.

The P55-UD6 also continues to support Ultra Durable 3 features like 2-ounce copper PCB. Adding additional copper in the power and ground layers in the PCB improves signal quality, reducing EMI and providing better ESD protection, reduces energy waste, and helps to lower PCB temperatures. This in turn can all lead to better OC’ing results in theory. Other Ultra Durable 3 features include the use of 50,000 hour Japanese solid capacitors, lower RDS MOSFETs, and chokes with a ferrite core.

[image]
<% print_image("31"); %>

The other new feature Gigabyte adds to the P55-UD6 is termed “Smart 6”. The Smart 6 features that will probably interest enthusiasts the most is Smart QuickBoot, which can be used to speed up the POST process during bootup, or you can skip POST entirely and use Smart OS QuickBoot to boot up from hibernate mode. With this feature, the motherboard will suspend and hibernate when the user shuts down the PC. The system will then resume from suspend mode the next time the PC is turned on.

For overclocking, Smart QuickBoost offers 3 predefined OC profiles, a Fast, Turbo and Twin Turbo which will up the bclk from 133MHz to 140MHz, 150MHz, and 160MHz respectively.

Other Smart 6 features include Smart TimeLock (i.e. parental controls), Smart Recorder (records when the PC is turned on/off as well as when files are copied from the local HDD to an external storage device), Smart DualBIOS (stores up to 12 passwords and can be used to store important dates such as birthdays inside BIOS for reminder later) and Smart Recovery, which can be used to rollback your system to a previous date.


Another new feature added to the P55-UD6 is teaming for the motherboard’s dual GigE LAN ports. If you don’t need two GigE LAN ports, simply use teaming to double the speed of the one LAN port you’re using.

If all that weren’t enough, you can also use AutoGreen to pair your Bluetooth cell phone to your computer. Once installed, AutoGreen will automatically put your PC in standby or suspend mode once your Bluetooth cell phone is out of the PC’s range. Just get up and walk away (but don’t forget that cell phone!).

The P55-UD6 supports SLI and CrossFire, and like the other high-end P55 motherboards in this article, ships with 3 PCI Express graphics slots. Two PCI slots and two x1 PCIe slots round out the board’s expansion options.

Interestingly enough, the P55-UD6 also sports 6 DIMMs, although max memory capacity is the same as other P55 motherboards.

Besides its 24-phase power, the other area the P55-UD6 sets itself apart from others is its SATA connectivity. 10 internal SATA headers are located on the motherboard itself, with two eSATA/USB combo connectors located on the motherboard’s backplate. You also can’t miss the board’s debug LED and onboard power button (reset and clear CMOS switches are also provided on the board, although Gigabyte makes them as small as possible so you can’t accidentally hit them).

The P55-UD6 will carry an MSRP of $239.

[image]
<% print_image("32"); %><% print_image("33"); %>

If that’s too much to pay, Gigabyte will also offer the $169 P55-UD4P. This motherboard forgoes the 24-phase power, features fewer SATA ports (8), has less powerful heatsink-based P55 chipset cooling, skips the third PCI Express graphics slot (adding a third x1 slot in its place), and drops down to 4 DIMMs, but it does support the Smart 6 features, LAN teaming, Auto Green, and still supports SLI/CrossFire.

[image]
<% print_image("34"); %><% print_image("35"); %><% print_image("36"); %>

If you want to spend even less, Gigabyte also makes the P55-UD3R. This board lacks SLI support and ships with one Gigabit Ethernet port. The expansion options are also different, with four PCI slots, one x1 PCIe slot, and two PCIe graphics slots.

[image]
<% print_image("37"); %><% print_image("38"); %>

For HTPC use, Gigabyte will also offer two micro-ATX boards, the $149 P55M-UD4, which supports SLI/CrossFire and the $105 P55M-UD2. Thanks to their aggressive pricing, these boards will likely find a home in many budget Lynnfield systems as well.

As the “UD” designation implies, all the aforementioned Gigabyte motherboards will support Ultra Durable 3.



System SetupPage:: ( 9 / 19 )

Intel Core 2 Quad Q9650
Intel Core 2 Quad Q8400
Intel Core 2 Duo E8600
ASUS P5E3 Premium

4GB (2x2GB) OCZ DDR3 PC3-16000 Platinum @ DDR3-1333 Speeds

Intel Core i7-920
Intel Core i7-950
Intel Core i7-975 Extreme Edition
Gigabyte GA-EX58 Extreme

6GB (3x2GB) OCZ Reaper HPC DDR3-1600 @ DDR3-1066 Speeds

AMD Phenom II X4 965 Black Edition
AMD Phenom II X3 720 Black Edition
AMD Phenom II X2 550 Black Edition
Gigabyte GA-MA790FXT-UD5P

4GB (2x2GB) Corsair CM3X2G1600C9DHX @ DDR3-1333 Speeds

Intel Core i7-870
Intel Core i7-860
Intel Core i5-750
Gigabyte P55-UD6

4GB (2x2GB) Kingston KHX1600C8D3K2/4GX @ DDR3-1333 Speeds

ATI Radeon HD 4890 1GB
Catalyst 9.8

500GB Western Digital Caviar SE16

Windows Vista Ultimate 64-bit w/Service Pack 2


Benchmarks

Lost Planet
ARMA II
Crysis
Far Cry 2
Left 4 Dead
3DMark Vantage
Cinebench 10
WinRAR 3.9
LAME MT
DivX 7
VirtualDub 1.9.5



3DMark/Compression/Power UsagePage:: ( 10 / 19 )













Media Encoding/Rendering BenchmarksPage:: ( 11 / 19 )






Valve Particle Simulation Benchmark







ARMA IIPage:: ( 12 / 19 )

ARMA II – Direct3D







Far Cry 2Page:: ( 13 / 19 )

Far Cry 2 – Direct3D







CrysisPage:: ( 14 / 19 )

Crysis – Direct3D







Left 4 DeadPage:: ( 15 / 19 )

Left 4 Dead







Lost PlanetPage:: ( 16 / 19 )

Lost Planet – Direct3D







CrossFire Gaming PerformancePage:: ( 17 / 19 )

Far Cry 2




Crysis





Lost Planet







Overclocked PerformancePage:: ( 18 / 19 )












ConclusionPage:: ( 19 / 19 )


If we had to choose between the Core i5-750 and the Core i7-920, we’d pick up the Core i5-750 thanks to its ever so slightly better single-GPU gaming performance, and more importantly, its lower cost. Now obviously in the case of highly threaded apps the Core i7-920 easily outruns the Core i5-750, but these cases are pretty limited when it comes to games. Lost Planet and its MT Framework engine is really the best case scenario for Core i7-920. Everywhere else the Core i5 pulls slightly ahead of the i7-920 in single-GPU gaming. Thanks to Hyper-Threading, the i7-920 is able to outperform the i5-750 in media encoding/rendering tests, so if you’re the type who likes to mix your gaming with your video encoding and 3D rendering apps, you’re in for a very tough decision between the two CPUs.

Moving further up the ladder, the Core i7-860 represents the best combination of price, performance, and features in our opinion. Priced at $285, it delivers performance that’s quite competitive with Intel’s Core i7-950, which sells for nearly twice as much money as the i7-860. It doesn’t manage to outrun the 950, but at its price point it’s close enough that we’d take it anyway and use the money saved on a new GPU. We have a strong suspicion this chip is going to be even more popular than the Core 2 Duo E6600 was when Conroe was introduced back in the summer of 2006.

The Core i7-870 is an amazing chip that delivers great performance and the very best Turbo Mode settings Lynnfield has to offer. As a result, you get dramatic out-of-the-box scaling from this CPU, and it OCs like a beast too. We don’t know if Intel’s ever produced a better $560 CPU than the Core i7-870. With that being said though, in this economy we’d daydream about owning an i7-870, but actually purchase an i7-860 or a Core i5-750.

With Intel’s entire Lynnfield lineup delivering such ridiculous levels of performance at their price points, it’s hard to find a reason to pick up any Bloomfield CPU other than the Core i7-975 Extreme Edition. It’s got the highest clock speeds and the unlocked clock multiplier. Perhaps this is what Intel wanted all along?

The Lynnfield/P55 combo could end up killing Bloomfield/X58. As enthusiasts who crave maximum performance, we don’t want to see this happen, so hopefully Intel will incorporate some of Lynnfield’s new tricks into an enhanced Bloomfield derivative. Upping the triple-channel memory controller to DDR3-1333 would improve peak memory bandwidth to 32GB/sec. Improving Bloomfield’s Turbo Mode settings would be another biggie on our wish list. Until Intel improves Bloomfield though, sales are going to plummet significantly.

In Bloomfield’s defense, there are a couple of reasons why the CPU and X58 platform should continue to stick around though. In our multi-GPU testing, the Bloomfield CPUs did outperform their Lynnfield counterparts. For example, in Crysis testing at 1600x1200 with 4xAA and high settings, the Core i7-920 even managed to pull ahead of Core i7-870 by 5%.

Far Cry 2 also ran better on our Bloomfield/X58 testbed, with the i7-975 EE running 3% faster than the Core i7-870.

With high-end DX11 GPUs potentially delivering even more performance than the 4890, these margins should only increase in favor of Bloomfield.

The other advantage Bloomfield provides is a path to 6-core processing next year. Today’s X58 motherboards should be compatible with Intel’s 6-core Gulftown CPU, although if rumors are correct, this could be one pricey processor. Leaked Intel roadmaps posted online suggest that the Core i7-870 could be Intel’s fastest Lynnfield for quite some time, so your upgrade path may not be loaded with options for awhile if you pick up an 870 today.

And where does the debut of Lynnfield leave AMD? Looking at the benchmarks and Lynnfield price points, it certainly isn’t pretty for AMD. Back in August we predicted that the Phenom II 965 Black Edition would see its price reduced to $200 with the debut of Lynnfield. But judging by the benchmarks, AMD’s going to have to go even lower with the 965’s pricing in order to remain competitive. $150 sounds about right to us.

In order to become profitable though, AMD’s got to find a way to improve the ASP (average selling price) of their CPUs. More price cuts isn’t going to help that.

Lynnfield’s the real deal though. Now we see why Intel marketing designated this CPU as Core i7. We still think this move is going to confuse consumers, but when you’ve got a CPU this powerful, it would’ve been a disservice to the processor to call it anything less than Core i7.


© Copyright 2003 FS Media, Inc.
[ Print Article! | Close Window ]