What's new
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

Intel Core i7-980X Gulftown Six-Core 32nm Processor Review

Status
Not open for further replies.

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal

Intel Core i7-980X Gulftown
Six-Core 32nm Processor Review





Yes, it is finally here...well, almost. Although we have been given a green light to tease you with a review, actual availability of hexa- or six-core Core i7-980X processors is still a few weeks away. Nevertheless, now that the review embargo has ended - and it was another painful embargo for us since Gulftown and the Westmere-EP processors have been freely available in the enthusiast community (or eBay) for months - we can finally reveal (or least confirm) everything you have wanted to know about the soon to be available Core i7-980X Extreme Edition processor, codenamed Gulftown.

Although many had earlier suspected that Gulftown would sport the Core i9 moniker to distinguish it from the quad-core parts, Intel have stuck to the tried & true Core i7 brand. Being a Core i7 model, this processor supports both Hyper-Threading and Turbo Boost, which are the two features that Intel is using to differentiate between their Core i3/i5/i7 product lines. As we all knew it would be, this is a Socket LGA1366 processor and it will work on any Intel X58-based motherboard. At worst, a new BIOS will be needed on some motherboards, but if you have updated your BIOS anytime in the last 1-2 months chances are that you are ready to go with regard to Gulftown support.

The condensed specifics of the Core i7-980X are as follows and they read like a litany of must-haves for enthusiasts: a 32nm manufacturing process, fully unlocked multipliers, 3.33Ghz core clock, Turbo Boost up to 3.6Ghz, six cores plus Hyper-Threading, 12MB of L3 cache, triple-channel DDR3-1066 memory interface, 130W TDP, and a $999 USD price tag. The retail Core i7-980X package will also come with an impressive new stock CPU cooler; the DBX-B Thermal Solution.

This chip isn't targeted at your average computer user, not even your average enthusiast, who can barely take full advantage of the current quad-core/eight-threaded Core i7's potential. With six cores and twelve threads this is part for those who take their multi-tasking or benchmarking seriously. To make full use of this processor's capabilities you need to have software that supports more than eight threads, and they are few and far between right now. As you will see in the coming pages, Gulftown will particularly appeal to those who are obsessed with multi-threading performance but are also very conscious about power consumption, [email protected] -bigadv crunchers come to mind.

Even if you can't fathom a thousand dollar piece of hardware in your future this is still a review worth reading since Gulftown models will trickle down to reasonable price points in due time.


 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Gulftown - Core i7-980X

Gulftown - Core i7-980X




The Core i7-980X that we are reviewing today is the new Top of the Line model in the Core i7 realm. There have been rumours of lower-end six-core chips, but as of yet we have no official information to confirm or deny those rumours. There will be some attractive six-core Xeon variants though, so do keep an eye out for those if your budget doesn't allow for a $999 USD processor. Those looking for 32nm quad-cores would also be well-advised to keep an eye on the Xeon front.

Despite being based on the new 32nm manufacturing process, Intel have been rather conservative with the core clock, simply matching the 3.33Ghz frequency of the Core i7-975 Extreme Edition. Being an Extreme Edition part, the i7-980X does have fully unlocked multipliers though, so any perceived deficit in core clock can fixed with a few key strokes in the BIOS. We suspect that the 3.33Ghz speed was chosen for marketing reasons, allowing Intel to release a slightly faster Gulftown Extreme Edition model down the road, as they did with Bloomfield. As our overclocking results will demonstrate, they surely didn't settle on 3.33Ghz due to a lack of frequency headroom.

Enough talking though, let's take a closer look at how this new Gulftown chip compares to the rest of the Core i7 lineup.


As you can see, the Core i7-980X really only distinguishes itself from the Core i7-975XE by its smaller manufacturing process, two more cores, additional cache, and AES instruction set. It has the same core clock, same Turbo Boost frequencies, same memory frequency support, same TDP, and the now familiar $999 Extreme Edition pricetag. This was all expected though, since Gulftown is based on Westmere, which is merely 32nm derivative of the Nehalem microarchitecture. We will have to wait until 2011 with Sandy Bridge to see some truly substantial architectural changes.

Gulftown was designed to be multi-threading monster with even greater overclocking headroom then we have seen with the current 45nm Bloomfield-based Core i7 lineup, you can't go wrong there in you're an enthusiast.



Click on image to enlarge

Needless to say that new Gulftown Core i7 processors look exactly like the previous Bloomfield Core i7 ones. On the last line of integrated heatspreader, the numbers "951" reveal that our engineering sample was manufactured in the 51st week of 2009. Retails chips will surely all have 2010 manufacture dates, which will be identified by "001", "002", etc.


Click on image to enlarge

Those are 1366 contact points, hence the LGA1366 socket name. Interestingly, as you can see in the second image, Gulftown (on the left) has a lot less 'large' micro SMD resistors on the underside than Bloomfield does.


Gulftown / Bloomfield / Lynnfield / Clarkdale / Phenom II - Click on image to enlarge

Although Lynnfield LGA1156 is actually the chip with the largest core out of the entire Nehalem-based lineup (because of the integrated PCI-E controller), the LGA1366 package is by far the biggest because it was designed with future headroom in mind, specifically being able to handle six-core and maybe even eight-core dies.


Don't expect to see the stock core clock often with Turbo Boost enabled - Click on images to enlarge

Here is the obligatory CPU-Z screenshots. Our particular sample idled at 0.944V and utilized 1.152V under full load, occasionally peaking to 1.184V. This actually about the same as on D0 stepping Bloomfield's, which is slightly disappointing since we would have liked to see the vCore drop a bit with the new 32nm process. Although CPU-Z is not able to read the revision on this engineering sample chip, it is a B1, which is the final retail stepping.

On a side note, although Gulftown has a large & lovely 12MB L3 cache, it does come with an approximately 14% latency penalty compared to Bloomfield's L3 cache, which will obviously have a slight effect on clock-per-clock performance in some workloads.

Next let's take a look at the brand new CPU cooler that Intel have designed for this particular processor.
 
Last edited:

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Intel DBX-B Thermal Solution

Intel DBX-B Thermal Solution



Much to our surprise, Intel have finally gotten serious about bundling a quality CPU cooler with their $999 Extreme Edition processor. Bundled with every retail Core i7-980X will be the new Intel DBX-B Thermal Solution.



Click on images to enlarge

As you can see, this is a proper aluminium tower heatsink with copper heatpipes, no longer the low-profile circular fin array design that Intel have used seemingly forever. The DBX-B is just under 5.5 inches tall, 2.75 inches thick, and 4 inches wide. Our sample weighed in at 677 grams.

The stock Nidec fan measures 100mm and features a blue LED, as we have grown accustomed to on all Extreme Edition coolers. Intel claim idle CPU acoustics under 20 dBA at 800 RPM and maximum acoustics under heavy load below 35 dBA at 1800 RPM. While the idle figure seems about right, at its highest rotation we found this fan to be quite high pitched. Thankfully, this CPU cooler has a built-in fan speed switch with a Quiet mode that keeps fan noise under control.

Interestingly, the fan blades were designed in such a way to push relatively little air thru the heatsink. Instead, the angle of the blades causes a lot of the airflow to be directed towards the other components on the motherboard, like the MOSFETs. What will this mean for CPU cooling performance? We will check that out in our Temperature Testing section.



Click on images to enlarge

The bottom mounting bracket is made from somewhat flimsy plastic, but it's the flexible variety that won't break unless you do it on purpose. Installation of the CPU cooler is very straightforward. All you need to do is place the aforementioned mounting bracket under the motherboard, line it up with the socket holes, place the heatsink on the processor, push down the thumbscrews, and then tighten. Although you can tighten the thumbscrews by hand, it can somewhat difficult to tighten the rear ones on some motherboards where the MOSFET cooler can get in the way. With this in mind, the DBX-B has two holes on the top that go all the way through the heatsink and allows for a long thin screwdiver to tighten the rear thumbscrews.


Click on images to enlarge

The copper base on our sample was quite polished and slightly convex. This isn't a manufacturing fault though, it's simply done to ensure that the base makes strong contact with the center (where the CPU die is) of the processor's integrated heatspreader (IHS), which is also always convex. Simply making both the heatsink base and the IHS perfectly flat might seem logical, but it is impossible to do economically on a mass production scale.

There were machining marks on the base, but you couldn't feel them. This CPU cooler doesn't feature a H.D.T (Heatpipe Direct Touch) design, instead the heatpipes are merely soldered to the base.


Click on images to enlarge

As you can see, there's no comparing the new DBX-B to the previous Bloomfield and Lynnfield/Clarkdale CPU coolers. It's a huge improvement in every sense of the word.


Click on images to enlarge

For further comparisons sake, here is the DBX-B next to the ubiquitous Thermalright Ultra-120 eXtreme. Does it stand a chance of competing with it? Well that's what we are going to find out in our Temperature Testing section.
 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Westmere Microarchitecture - Gulftown Edition

Westmere Microarchitecture - Gulftown Edition




In 2007, Intel unveiled the Tick-Tock Model as a demonstration of the company's dedication towards continued rapid technological innovation. The "tick" is a shrinking of the previous architecture manufacturing process (65nm --> 45nm --> 32nm) and the "tock" is a new architecture. First launched in November 2008 with the Bloomfield Core i7-900 series, the Nehalem microarchitecture saw its second variant launched in September 2009; the Lynnfield Core i5-700 series & i7-800 series. Lynnfield was nothing radically new, it merely used Nehalem's modular design to integrate the PCI-Express controller into the CPU die.

In January 2010, the "tick" to 2008's "tock" appeared in the form of the Clarkdale, the first Westmere-based processors. Westmere shrinks the Nehalem architecture down to 32nm, which is a 41% smaller manufacturing process than 45nm. This is important because manufacturing cores on a smaller process allows for cooler running chips, better power efficiency, higher frequency scaling, and perhaps most importantly more cores on a CPU package, which is what Gulftown is all about.


Bloomfield die on the left, Gulftown die on the right - Not to scale


Fundamentally speaking, Gulftown carries forward much of what made the Bloomfield processors successful, but adds an extra two cores, an extra 4MB of L3 cache, and is manufactured on Intel's ultra-modern 32nm high-k+ metal gate transistor technology. Obviously, it doesn't have Lynnfield's integrated PCI-Express controller and it doesn't have an integrated GPU like Clarkdale does. We will have to wait for Sandy Bridge to see both those features built into a hexa-core processor. Frankly though, those are two elements we can live without, since the X58 IOH is more than capable of handling PCI-E duty and there's no place for an IGP on a high-performance platform.

In quantifiable terms, Gulftown's CPU die size is 248mm². By comparison, Lynnfield comes in at 296mm² and Bloomfield measures 263mm². Thanks to the 32nm process Intel have managed to stuff 50% more cores and cache, yet reduce the overall die size by about 6%. Transistor-wise, this is the first consumer-oriented processor to break the 1 billion mark. The Gulftown CPU die is packed with 1.17 billion transistors. This is a healthy increase from Lynnfield's 774 million transistors and Bloomfield's 731 million transistors. On an interesting sidenote, AMD's 'Istanbul' six-core Opteron processors, which are manufactured on a 45nm process, have 904 million transistors and a whopping 346mm² die size.

Now many of you are probably looking at the die pictures and saying "It is pretty but what am I looking at exactly?". A valid question, so let's take a look at the Gulftown CPU core layout:


Anyone familiar with Bloomfield's die layout will see nothing but similarity with Gulftown, and this is because of how the Nehalem microarchitecture was designed. Which leads us to...


Part of the reason that Intel is able to add and remove parts so easily is because the Nehalem architecture is dynamically scalable, and it was designed with modularity in mind. What this means is that Intel can custom create processors based on the needs of the market without having to go design a brand new chip from scratch. They can add or remove cores, L3 cache, memory channels, memory controllers, power management features, and even integrated graphics. Therefore, Intel have the ability to add new blocks to the core without having to go to the drawing board and redesigning the whole layout. Basically, they are only limited by how much stuff they can actually fit on one CPU package. Think of it as a multi-million dollar Lego set.


If you are unfamiliar with the features and technology present in Nehalem-based processors, or simply want a more in-depth explanation as to what Intel have added/tweaked on Gulftown, the following section should interest you.
 
Last edited:

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Westmere Microarchitecture - Gulftown Edition pt.2

Westmere Microarchitecture - Gulftown Edition pt.2



The Nehalem microarchitecture, upon which Westmere-based chips like Gulftown are derived,
was highly touted as one of the most significant architectural overhauls ever. Although sharing significant roots with the original P6 microarchitecture that was debuted in the Pentium Pro in 1995, and the later Core/Penry microarchitecture, Intel's engineers have added significant performance-oriented features, like an integrated memory controller, a completely new system interconnect, and a multi-level shared cache. They also focused a great deal on the chip's power efficiency capabilities.

As discussed on the previous page, Gulftown has fundamentally the same features and technology as the Bloomfield chips. It has a triple-channel DDR3 memory interface, it has Hyper-Threading, and the same relatively mild implementation of Turbo Boost. The sole difference is one new instruction set, AES-NI.

Nevertheless, let's examine some of these features and technologies:


  • QPI

For the Nehalem architecture, Intel has foregone the legacy front side bus in favour of the QuickPath Interconnect (QPI). The QPI is a high-speed, low-latency point-to-point processor link. From a technical standpoint, the QPI is a bi-directional 20-bit wide bus that is integrated onto the processor itself. The result? An incredibly fast interconnect that will improve overall bandwidth while reducing latency. This high-speed interface is used to access the distributed shared memory, it helps cores communicate with each other, and it also links up with the X58 northbridge; now known as the IO Hub (IOH).

Being a high-end performance part like the i7-965 Extreme Edition and i7-975 Extreme Edition, the new i7-980X features the faster 6.4 Gigatransfers per second (GT/s) QPI link, which has a theoretical maximum bandwidth of 25.6GB/s; equivalent to Nehalem's triple-channel DDR3-1066 memory bandwidth. The lower-end i7-920/930/940/950 models all feature a 4.8GT/s QPI interface with 19.2GB/s of bandwidth. The benefits of the faster QPI link are mostly seen in graphically intensive applications, specifically when multiple graphics cards are installed.

  • Integrated Memory Controller (IMC)


As we have come to expect from Nehalem-based chips, Gulftown features an integrated memory controller. As a result, the memory has a direct link to the processor, which not only means significantly lower latency, but much higher bandwidth as well. Current Core i7 processors feature a triple-channel memory interface, and each channel can support one or two DDR3 modules. This means that memory modules should be installed in sets of three, not two as has been the norm since the dual-channel memory architecture was first introduced back in 2003.

While there was much speculation that Intel would increase the stock memory frequency to DDR3-1333, it remains at DDR3-1066, which means 25.6GB/s of memory bandwidth. This amount of memory bandwidth proved to be overkill on Bloomfield processors, so it will be interesting to see if the situation is any different with Gulftown.

  • SSE4.2 & AES-NI

Building upon Penryn's implementation of SSE4.1, which was focused on improving video encoding, image/video editing, faster 3D game physics, etc...the Nehalem architecture adds 7 new instrutions, namely Accelerated String and Text New Instructions (STTNI) and Application Targeted Acceleration (ATA), which focus on faster XML parsing, faster search and pattern matching, and other cryptic processor functions.

A brand new addition to the Westmere core are the Advanced Encrytion Standard New Instructions (AES-NI). There are 12 new instructions designed to accelerate tasks that use the AES algorithm, such as whole disk encryption/decryption, internet security, VoIP, etc. Baiscally, this essentially allows the processor to do real-time high-security encryption/decryption with little to no effect on system performance.


  • Hyper-Threading


Nehalem also brought Hyper-Threading (HT) back from the dead, and it's a huge factor on Gulftown, turning a 6-core processor into a 12 thread crunching workhorse. With HT enabled, a processor with four physical cores is viewed by the operating system as having eight logical cores. A core usually processes the pieces of the different threads one after another, however an HT-enabled core can process two threads in a simultaneous manner. While Hyper-Threading did not perform particularly well on the Pentium 4, Nehalem's architecture was designed to remove many of the processing bottlenecks. Depending on the workload, and how effectively multi-threaded an application is, the performance increases could be 20% or higher.


  • Power Control Unit (PCU)



Nehalem’s Power Control Unit (PCU) is an extremely innovative power management feature that uses an on-chip micro-controller to actively manage the power and performance of the entire processor with the help of numerous integrated power sensors. The PCU can dynamically alter the voltage and frequency of the CPU cores to lower power consumption or provide performance boost in conjunction with the new Turbo Mode feature. Also, thanks to a development know as Power Gates, idle cores can be completely shut down and placed in a C6 sleep mode while other cores continue working. This is noteworthy because C6 mode had previously only been featured on mobile processors.


  • Turbo Mode



Turbo Boost is arguably the most discussed feature brought forth by Nehalem. Basically, all Core i7 LGA1366 processors come with two additional speed bins, which is to say that they have two higher multipliers that they can use under certain scenarios. For example, if you are using a single-threaded application, the PCU will down-clock or shut down the unused cores, thereby freeing up power and lowering heat output while "overclocking" that one core that is in use. If an application is multi-threaded and the cores are not running too hot, the PCU will overclock all the cores up one speed bin.

If you are more visually-inclined, the following illustration should help explain the Turbo Boost implementation:



Taken as a whole, Gulftown doesn't really bring any major new updates, but there weren't really any glaring feature/technology omissions on Bloomfield to begin with. We would have definitely liked to see Lynnfield's more aggressive Turbo Boost implementation make its way to Gulftown, but since the i7-980X has unlocked CPU multipliers, I don't foresee many people being limited in their quest for extra performance.
 
Last edited:

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Test Setups & Methodology

Test Setups & Methodology


For this review, we have prepared four different test setups, representing all the popular platforms at the moment, as well as most of the best-selling processors. As much as possible, the four test setups feature identical components, memory timings, drivers, etc. Aside from manually selecting memory frequencies and timings, every option in the BIOS was at its default setting.

Intel Core i7 "Gulftown" Test Setup​


Intel Core i7 "Bloomfield" Test Setup​


*The 1802 BIOS was only used for the Everest, ScienceMark, and Power Consumption tests.​

Intel Core i5 & Core i7 "Lynnfield" Test Setup​



Intel Core i5 "Clarkdale" Test Setup​


Intel Core 2 Test Setup​



AMD Phenom II AM3 Test Setup​



For all of the benchmarks, appropriate lengths are taken to ensure an equal comparison through methodical setup, installation, and testing. The following outlines our testing methodology:

A) Windows is installed using a full format.

B) Chipset drivers and accessory hardware drivers (audio, network, GPU) are installed followed by a defragment and a reboot.

C)To ensure consistent results, a few tweaks were applied to Windows Vista and the NVIDIA control panel:
  • Sidebar – Disabled
  • UAC – Disabled
  • System Protection/Restore – Disabled
  • Problem & Error Reporting – Disabled
  • Remote Desktop/Assistance - Disabled
  • Windows Security Center Alerts – Disabled
  • Windows Defender – Disabled
  • Screensaver – Disabled
  • Power Plan - High Performance
  • NVIDIA PhysX – Disabled
  • V-Sync – Off

D) Programs and games are then installed & updated followed by another defragment.

E) Windows updates are then completed installing all available updates followed by a defragment.

F) Benchmarks are each run three times after a clean reboot for every iteration of the benchmark unless otherwise stated, the results are then averaged. If they were any clearly anomalous results, the 3-loop run was repeated. If they remained, we mentioned it in the individual benchmark write-up.

Here is a full list of the applications that we utilized in our benchmarking suite:
  • 3DMark06 Professional v1.1.0
  • 3DMark Vantage Professional Edition v1.0.1
  • Cinebench R10 64-bit
  • Crysis v1.21
  • Far Cry 1.02
  • HyperPi 0.99b
  • High Definition Experience and Performance Ratings Test 2009 (HDxPRT 2009) (Adobe Elements 7.0/QuickTime Player/iTunes 8.0.1/Sorenson Squeeze/PowerDVD 8/DiVX engine)
  • Lame Front-End 1.0
  • Lavalys Everest Ultimate v5.02.1834 Beta (v5.30.1977-2043 Beta for Bloomfield, Clarkdale, Lynnfield, Gulftown)
  • Left 4 Dead
  • PCMark Vantage Advanced 64-Bit Edition (1.0.0.0)
  • Photoshop CS4 Extended (64-bit)
  • ScienceMark 2.0 Build 21MAR05
  • Street Fighter 4 Demo
  • Supreme Commander v1.1.3280
  • Valve Particle Simulation Benchmark
  • WinRAR 3.8.0
  • World in Conflict v1.010
  • x264 HD Benchmark v1.0

That is about all you need to know methodology wise, so let's get to the good stuff!
 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Feature Test: Hyper-Threading (HT)

Feature Test: Hyper-Threading (HT)



Twelve-threaded goodness

The Nehalem microarchitecture marks the return of Hyper-Threading (HT), which is a feature that was first implemented on the Pentium 4 "Northwood" but with little success. Thankfully, as we demonstrated in our original Core i7 review, with it's shorter, faster, more efficient pipeline (able to simultaneously process up to four instructions), Nehalem-based processors can truly make use of Hyper-Threading in certain workloads. To demonstrate this, we have benchmarked a small selection of multi-threaded applications with HT enabled and disabled.


The results speak for themselves. In highly multi-threaded applications, HT can make a significant difference, speeding up specific workloads by 8% to 35%. However, most contemporary games simply aren't designed to take advantage of more than four threads, much less twelve. This obviously negates any possible HT-related performance gains. However, Intel is obviously pushing software developers to ensure that more and more programs are designed to take advantage of all available threads, and we are aware of a handful of interesting titles that will be able to take full advantage of Gulftown's multi-threading prowess.
 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Feature Test: Intel Turbo Boost

Feature Test: Intel Turbo Boost - Turbocharged



3.33Ghz Core i7-980X Turbo'ing up to 3.59Ghz

For those of you who skipped the microarchitecture section, let's recap what Turbo Boost is, what it does. Turbo Mode is an exciting feature that automatically unlocks additional speed bins (multipliers) and allows the processor to self-overclock based on thermal conditions and workload. For example, if the Power Control Unit (PCU) senses that only one core is active and the other three are in an idle state, it will use the unused power and thermal headroom to overclock that single active core to ensure superior single-threaded performance. Conversely, if you are running a multi-threaded application, the PCU will measure the thermal headroom and if the processor is running cool enough it will overclock all six cores. As on the Bloomfield processors, Gulftown's Turbo Mode can provide a 266Mhz speed boost in single-threaded workloads and 133Mhz in multi-threaded applications.

Although the results will be fairly self-evident, let's check out the performance gains that Turbo Mode provides on this top-end Core i7-980X model. As per the above, thermal conditions permitting, it will run one core at 3.59GHz for single-threaded workloads, and 3.46Ghz in multi-threaded applications.


As you can see, there are some worthwhile performance improvements in multi-threaded applications, and some more noticeable speeds boosts in single-threaded applications like SuperPI. Some people may consider the Turbo Mode feature a mere gimmick, and perhaps it is for enthusiast users. However, as we have said in the past, no ones is going to begrudge Intel for giving all users a free 133-266Mhz speed boost. The fact that it's automatic and seamless is ideal for novice computer users. We would have definitely preferred to see Intel outfit Gulftown with more aggressive Turbo Boost profiles though, à la Lynnfield, since as we will demonstrate in our overclocking section, these new 32nm cores have a lot of frequency headroom.
 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Clock-per-clock: Bloomfield vs. Gulftown

Clock-per-clock: Bloomfield vs. Gulftown



Now as we have already established, at first glance Gulftown is basically just a six-core Bloomfield with an extra 4MB of L3 cache and manufactured on a 32nm process. However, what we are interested in finding out is whether there are any hidden changes that improve overall performance, maybe Uncore tweaks or lower cache latencies, etc. In order to determine this, we came up with a duel: Core i7 Bloomfield vs. Core i7 Gulftown. To ensure that both chips were competing on an equal playing field, we set identical frequencies and timings for both processors, and we also disabled Gulftown's 2 extra cores.


Equal: Quad-Core Bloomfield vs. Quad-Core Gulftown

Now as you can see, despite losing two of its cores the Gulftown's extra 4MB of L3 cache remains (one of the advantages of having a shared cache pool), but that shouldn't skew the results too much. It should be mentioned that Gulftown's large L3 cache has a 14% higher latency than Bloomfield's, which will obviously have a slight effect on clock-per-clock performance in some workloads.


In our admittedly very small selection of applications, we didn't notice any worthwhile performance differences between Bloomfield and Gulftown. That wasn't unexpected though since Intel never claimed otherwise. We can however say that in certain games there is likely to be improvements in the minimum framerate due to Gulftown's extra cache. Also, thanks to Gulftown's Advanced Encrytion Standard New Instructions (AES-NI), you might see as much as a 100% improvement in AES encryption and decryption with certain applications. Overall though, short of the AES-NI instructions, there is no significant performance difference between Bloomfield and Gulftown from a clock-per-clock standpoint in most applications.
 

MAC

Associate Review Editor
Joined
Nov 8, 2006
Messages
1,106
Location
Montreal
Synthetic Benchmarks: CPU & Memory

Synthetic Benchmarks: CPU & Memory



Lavalys Everest Ultimate

Everest Ultimate is the most useful tool for any and all benchmarkers or overclockers. With the ability to pick up most voltage, temperature, and fan sensors on almost every motherboard available, Everest provides the ability to customize the outputs in a number of forms on your desktop. We selected two of Everest's seven CPU benchmarks: CPU Queen and FPU Mandel. According to Lavalys, CPU Queen simple integer benchmark focuses on the branch prediction capabilities and the misprediction penalties of the CPU. It finds the solutions for the classic "Queens problem" on a 10 by 10 sized chessboard. At the same clock speed theoretically the processor with the shorter pipeline and smaller misprediction penalties will attain higher benchmark scores. The FPU Mandel benchmark measures the double precision (also known as 64-bit) floating-point performance through the computation of several frames of the popular "Mandelbrot" fractal. Both tests consume less than 1 MB system memory, and are HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


Needless to say that the i7-980X's first result raised some eyebrows. Despite trying out all the latest Beta versions of Everest, we consistently got the same results. While the FPU Mandel results are right where we would expect them to be, the CPU Queen results were about 30% higher than expected. We suspect that there's probably an issue with this particular test being run across 12 logical threads.

Lavalys Everest Ultimate

Everest Ultimate is the most useful tool for any and all benchmarkers or overclockers. With the ability to pick up most voltage, temperature, and fan sensors on almost every motherboard available, Everest provides the ability to customize the outputs in a number of forms on your desktop. In addition to this, the memory benchmarking utility provides a useful tool of measuring the changes to your memory sub-system.


Our memory bandwidth results were a little lower than expected when compared to the i7-975, especially with regard to copy speeds. We suspect it might be an issue with the BIOS, or perhaps Gulftown's Uncore/integrated memory controller is slightly different than Bloomfield's, perhaps due to its being utilized by 2 additional cores.


On the latency front, the i7-980X also proved to be a little slower than the i7-975.


Will ScienceMark 2.0 paint a different picture? Let's find out.

ScienceMark v2.0

Although last updated almost 3 years ago, and despite its rudimentary interface, ScienceMark v2.0 remains a favorite for accurately calculating bandwidth on even the newest chipsets.


The memory bandwidth rankings are fairly similar to Everest's, which is to say the the i7-980X does appear to have slightly lower overall memory bandwidth than the i7-975.



Surprisingly, the i7-980X achieved a lower latency figure than the i7-975 in ScienceMark, but just barely so.

These are just synthetic numbers though, real-life applications and games are what count. Let's check those out next.
 
Status
Not open for further replies.
Top