||Intel Core 2 Duo E8500 Review
March 30, 2008 Chris Crazipper Angelini
Summary: Thanks to its 45-nm manufacturing process, Intel's Wolfdale core ushers in higher frequencies, more cache, better overclocking, and power-savings. So how far were we able to push the E8500? How does 4GHz+ sound! Full benchmarks inside!
| Introduction||Page:: ( 1 / 14 )|
Intel’s silicon process technology is the drum line keeping the company on a steady path. Its cadence—the tick-tock model of silicon and microarchitecture—helps explain Intel’s relentless march forward and continued success.
In short, the cadence model is broken down into two year periods. In the first year, one design team focuses on shrinking the process technology of an existing microarchitecture, working to deliver a smooth transition. In the second year, that perfected manufacturing node is applied to an all-new microarchitecture. Hence the tick and the tock.
The cadence model is working so well for Intel that we as enthusiasts hardly notice it in action. It’s much easier to just enjoy all of the exciting products coming from the company.
Nevertheless, Intel recently started shipping new processors based on the latest 45nm tick, and enthusiasts are going to want to keep an eye out. Why? Because the family of 45nm processors is more than just a simple die shrink, to begin. Dropping to a smaller process lets Intel cut power consumption, at the same time allowing room for more transistors on each die. So, while 45nm is the big news with Penryn, Intel lists nine other enhancements that’ll make this particular rim shot more potent than past process shifts.
Most significant are larger caches across the board. The Wolfdale design—Intel’s 45nm dual-core configuration—features 6MB of L2. Yorkfield, comprised of two Wolfdales on a single package, gets 12MB.
Intel’s Advanced Digital Media Boost consists of two components. First, a new shuffle engine helps speed up SSE instructions employing certain operations. Second, the addition of an SSE4 instruction set promises to further accelerate multimedia applications properly optimized to take advantage of the feature. In other words, don’t expect to see the benefits of SSE4 today, but rather once the game guys have a chance to work the technology into their titles.
Some of the other improvements include a faster divider, boosts in the speed that Intel’s Virtualization Technology enters and exits virtual machines, higher front side bus speeds, and a couple of power-saving technologies.
| A Slow, Steady Trickle||Page:: ( 2 / 14 )|
Last year, Intel introduced the Core 2 Extreme QX9650 based on its quad-core Yorkfield configuration. The idea then was to show off Penryn on the desktop in its full glory. The QX9650 runs at 3 GHz and rides on a 1333 MHz front side bus. It’s currently available online, albeit at a price point in excess of $1,000.
The rest of the desktop Penryns were to follow in the first quarter of 2008. As we crest the end of Q1 and carry over into Q2, a quick search online reveals that, indeed, Intel is already selling the dual-core Wolfdales and quad-core Yorkfields, all manufactured at 45nm.
|<% print_image("02"); %>||<% print_image("03"); %>|
Though the Core 2 Extreme QX9650 sits high and mighty on a four-digit price point, Intel’s other 45nm offerings are actually quite affordable. You can find a quad-core Yorkfield, also on a 1333 MHz bus, for less than $300. Dual-core Wolfdales sell as low as $215.
AMD is currently hawking its higher-end Phenom chips at the same prices as Intel’s entry-level models, giving AMD a modest edge in value. However, we haven’t yet explored the performance of Intel’s smaller process technology and tweaked microarchitecture, so there’s no telling how the competing CPUs will size up to each other.
On the Block: Core 2 Duo E8500
Premium Core 2 Duo processors based on the 65nm Conroe core all belonged to the E6300 series. Wofldales are instead identifiable by their E8000 model designators. The current flagship in Intel’s dual-core family is the Core 2 Duo E8500, clocked at 3.166 GHz. An E8400 at 3 GHz and an E8200 at 2.667 GHz are also available.
If you’re searching online and find Wolfdale-based processors with X5000, E5000, or E3000 model numbers, take note that those are Xeons. Intel is using the Wolfdale design to power some of its 1P and 2P server/workstation platforms. Just keep in mind that 3000-series Xeons work in LGA775 motherboards. However, the 5000-series chips are designed for DP operation and drop into the LGA771 interface. They won’t work with your enthusiast P35 or X38 board.
The chip we’re looking at today is the top-end E8500 running at 3.16 GHz on a 1333 MHz front side bus. In moving to 45nm, Intel employs smaller high-K transistors (versus its old polysilicon transistors) and is able to shrink the E8500’s die down to 107 square millimeters, at the same time increasing transistor count to 410 million. Compare the 65nm Conroe with its 143 millimeter die and 291 million transistors.
The changes to Intel’s manufacturing process are enabling smaller transistors, yes. They’re also making it possible for the company to add complexity through a 6MB shared L2 cache and several microarchitecture enhancements.
The new core requires comparable voltage in order to achieve faster clock speeds, thereby boosting performance and minimizing power consumption. Old dual-core Conroe chips employed VID ranges between .962 and 1.350V at the 3 GHz level. The Core 2 Duo E8500 we’re testing today sits between .85V and 1.3625V to hit 3.16 GHz. Even with that fat cache, it still sports a 65W TDP.
Intel lists a bevy of supported features for the E8500 on its processorfinder.intel.com engine. Dual-core, Enhanced Speed Step, EM64T, Virutalization Technology, the Execute Disable Bit, C1E, and the Thermal Monitor 2 are all included. You’ve already seen most of those. Again, Wolfdale isn’t a revolutionary move. Rather, it’s a tick—a showcase of Intel’s 45nm manufacturing. That means we’re expecting some magical things when it comes to overclocking.
| Overclocking, Pricing, and System Setup||Page:: ( 3 / 14 )|
Overclocking the Core 2 Duo E8500
The first thing to check before you start trying to tweak any of the Penryn-based 45nm CPUs is that you have a brand new BIOS. With Wolfdale and Yorkfield, Intel is back to using half multipliers. It turns out that at least a handful of motherboards out there began their lives equipped for CPUs that use whole number multipliers.
There’s no upping the multiplier here. A retail E8500 is stuck at 9.5x, so your only way to coax extra speed from the platform is to crank up that front side bus. We started off slow, keying in 10 MHz bumps. Eventually, we jumped to 325 MHz, then 350, 375, and 400 MHz. The highest bus speed that’d boot was around 450 MHz, yielding a nearly 4.3 GHz clock speed. We throttled down to 437 MHz, however, to achieve stability in all of our tests. The result was an FSB speed in excess of 1700 MHz and a core clock of 4.16 GHz.
No doubt, the scalability of Wolfdale is largely attributable to the materials changes tied into Intel’s 45nm node. According to a technology piece published by EETimes, the principal features of the new process are the use of hafnium-based high-k dielectrics and titanium nitride for the NFET/PFET replacement gates. The new materials help combat leakage—a problem that would have been hard to manage with SiON gate dielectrics only three atomic layers thick. The piece claims Intel is eyeing leakage improvements of 10x or more.
Those same changes will carry into the introduction of Intel’s next tock—the Nehalem microarchitecture—and then scale into 32nm manufacturing. Given the relative smoothness with which the company moved from 65nm to 45nm, we’d expect the Penryn generation to continue scaling well.
Prices on the Core 2 Duo E8500 are currently hovering around $300, while the 3 GHz E8400 sits around $260.
For the sake of comparison, you can snag an OEM Core 2 Quad Q9300, based on Intel’s 45nm Yorkfield design, for about $290. The trade-offs include a lower multiplier (7.5x versus the E8500’s 9.5x), which will likely hamper overclocking to some extent, a 95W TDP, and a lower stock clock speed.
On the AMD side, your options consist of the brand new Phenom X4 9850 at 2.5 GHz. You won’t get as much out of the 65nm K10-based chip when it comes to overclocking. But at its default settings (and attractive $235 price point), you’re looking at a very solid alternative.
Intel Core 2 Quad Q6600 (2.4 GHz)
Intel Core 2 Duo E8500 (3.16 GHz)
AMD Phenom 9850 Black Edition (2.5 GHz)
AMD Phenom 9600 Black Edition (2.3 GHz)
AMD Phenom 9500 (2.2 GHz)
ASUS STRIKER II Formula nForce 780i Motherboard
Gigabyte MA790FX-DQ6 Motherboard
ASUS P5E-VM HDMI Motherboard
2GB OCZ Technology DDR2-1066 CAS5 Memory (2x1GB)
Gigabyte GV-RX387512H Radeon HD 3870 512MB
Seagate Barracuda 7200.11 1TB SATA 3 Gbps Hard Drive
Windows Vista x32, current as of March 28th, 2008 with Windows Update
Desktop resolution 1600x1200, 32-bit color, 85Hz refresh
We disable Vista’s UAC and generate an image using Norton Ghost 11 to create the same basic benchmark platform for each test bed. The image is frozen with the latest Windows Updates and deployed to each system. The appropriate drivers are then loaded to the machines.
Unreal Tournament III
Lost Planet: Extreme Condition
Half-Life 2: Episode 2
Company of Heroes
Call of Duty 4
Windows Media Encoder
| 3DMark06||Page:: ( 4 / 14 )|
3DMark06 is optimized for threading, so Intel’s Core 2 Duo E8500 hits the stage with an inherent disadvantage. Nevertheless, it puts up a reasonable battle against AMD’s faster Phenom X4 9850 Black Edition and Intel’s aging Core 2 Quad Q6600.
When it’s cranked up to 4.17 GHz, the E8500 slides into a first place finish in the Overall score.
At 3.16 GHz, the Core 2 Duo E8500 takes a last place finish in the raw CPU benchmark. Cranked up to an overclocked 4.17 GHz, however, the dual-core contender takes a commanding lead.
If you’re willing to overclock, the E8500 is quite the scrappy CPU in this stacked synthetic test. Let’s see how it performs in a more balanced real-world situation.
| Unreal Tournament III||Page:: ( 5 / 14 )|
Unreal Tournament III
Our Unreal Tournament III testing is done using the botmatch technique with a map populated by 24 players.
Although Unreal Tournament III takes advantage of quad-core processors, the E8500’s higher frequency and larger cache give it sufficient horsepower to blow away everything else at 800x600.
Shifting to 1280x1024 virtually vaporizes the Core 2 Duo’s advantage, although it’s still the fastest chip at 4.17 GHz.
Running at 1600x1200 favors Intel’s Core 2 Quad Q6600, followed by the E8500, and then the Phenoms.
| Crysis||Page:: ( 6 / 14 )|
We ditched the baked benchmarks for this test, recorded our own timedemo, and used it instead. Unfortunately, Crysis turns off the AI when you’re playing back demos, so you can expect actual game play to better tax these chips.
Late last year, Crytek made a big deal of multi-core support in Crysis, claiming 64-bit platforms and quad-core CPUs would deliver the most rewarding game experience possible. If anything, our benchmarks show that really isn’t the case. In fact, the Core 2 Duo again tears through our timedemo test faster than anything else at 800x600. Given the overclocked result at that resolution, we’re going to guess processing power isn’t the limiting factor there.
If there was any doubt at 800x600, it disappears at 1280x1024 where most of our CPUs hit a 38 frame per second brick wall. The Phenom 9600 and 9500 duck in a couple of frames lower.
The story changes little at 1600x1200, where everything hovers between 20 and 21 frames per second—unplayable for the most part without more graphics muscle.
| Lost Planet||Page:: ( 7 / 14 )|
Lost Planet: Extreme Condition
Same song, different key. Lost Planet is completely graphics bound from 800x600 all the way up to 1600x1200. Enjoy this one at 1280x1024 or add a second Radeon HD 3870 to give the platform a bit more muscle.
| Half-Life 2||Page:: ( 8 / 14 )|
Half-Life 2: Episode 2
Late in 2006, Valve’s Gabe Newell came out and evangelized the use of multi-core processors, specifically committing to optimizing his Source engine for CPUs with four cores. We’ve heard very little from Valve regarding multi-core updates to the engine since then. However, our benchmark results at 800x600 would suggest that you’ll still get more mileage from a faster dual-core chip with a larger cache, as the Core 2 Duo blows right past everything else we threw at it.
Things settle down as we shift to 1280x1024. The Core 2 Duo E8500 still leads the pack, but by a much smaller margin. We’re clearly graphics-bound here, though. There’s a solid chance Intel’s 45nm Wolfdale would extend its lead given a pair of graphics cards.
Naturally, 1600x1200 is a wash, especially with full anti-aliasing and anisotropic filtering.
| Company of Heroes||Page:: ( 9 / 14 )|
Company of Heroes
From our review of the Phenom X4 9850:
“Another title adapted for DirectX 10 support with a serious emphasis on graphics performance. From one chip to the next, you won’t notice much difference running Company of Heroes. If anything, you’ll want to add a second Radeon card for CrossFire support or pick up an SLI setup.”
The only exception to that quote comes at 800x600, where the Core 2 Duo hammers out a small advantage.
| Call of Duty 4||Page:: ( 10 / 14 )|
Call of Duty 4: Modern Warfare
“We concluded that Call of Duty 4 was limited by our Radeon HD 3870 back when we tested AMD’s Phenom 9600 and the same applies here as we run the Phenom X4 9850 through its paces. From 800x600 to 1280x1024 to 1600x1200 with the eye candy cranked, there’s not much wiggle room. Actually, that’s not entirely true. If you want to really get picky about it, the Core 2 Duo E8500 steps out in front of the pack and manages to keep one foot in front of the other processors at each successively higher resolution.”
| PCMark Vantage||Page:: ( 11 / 14 )|
PCMark Vantage is the latest synthetic system metric from Futuremark based on a number of real-world workloads. The benchmark is thread-aware and comprised of several test suites that measure performance under Windows Vista.
If you’re looking at the overall PCMarks score, Intel’s quad-core Q6600 puts down the highest number, followed by the E8500. AMD’s Phenoms aren’t far behind, though. You see the same general theme carry through the other benchmark suites with the exception of TV and Movies, where the new Phenom X4 9850 overclocked yields top marks.
Overclocked, Intel’s Core 2 Duo E8500 shakes up the chart a little bit by claiming the top PCMark score, in addition to several of the individual test suites. Tweaked for maximum performance, Wolfdale is a force to be reckoned with.
| Windows Media Encoder 9||Page:: ( 12 / 14 )|
Windows Media Encoder 9
While we’re all about gaming, it’s no secret that the latest titles stifle most attempts to compare processors, especially at the resolutions at which enthusiasts actually play. At 1600x1200, most folks are usually best off spending more money on the fastest video card they can buy.
Power users do more than game, though. Audio and video encoding are much better measures of processor performance. The encoder .dll used in Windows Media Encoder 9 is optimized for up to four threads, making it a great test of the dual- and quad-core chips on the bench.
Despite the fact that it’s up against a complete lineup of quad-core CPUs, the dual-core E8500 puts down the fastest encoding results. Overclocking the chip to 4.17 GHz shaves an additional 17 seconds off the Core 2 Duo’s time. It looks like there’s still a market for dual-core processing, regardless of whether you’re a gamer or multi-media aficionado.
| Ballistics Report: Core 2 Duo E8500||Page:: ( 13 / 14 )|
Raw Frequency: Today, everyone is preaching the virtues of parallelism and threading. Rightly so, in our opinion. But there’s still a software development push behind the movement, and we’re still not at a point where quadrupling on-die resources shows a significant gain over a higher-clocked two-core CPU except in synthetic tests specifically designed to show off what four cores can do. At least, that’s the case on the desktop. Higher-end SMB and enterprise apps are a different story…
Overclockability: The Core 2 Duo E8500’s standard 3.16 GHz shows really well—the clock speed would be put to even better use if we were testing with a more powerful graphics subsystem, too. Best of all, the microarchitecture seems incredibly scalable. It wouldn’t be difficult to take the chip to 4.3 GHz using more voltage, a beefier cooler, and maybe an X38- or X48-based motherboard, since the 780i doesn’t stretch out as well. Even still, we’ll take our 4.17 GHz using Intel’s stock cooler.
Technology: Intel has its cadence going, and this Core 2 Duo E8500 is showing up right at the beginning of a new generation of 45nm parts. With that said, there aren’t any growing pains to speak of. You drop the chip in a motherboard that supports it (not all LGA 775 boards do), update the BIOS to be sure you’re using half multipliers, and go. The power-saving features, larger cache, and multimedia enhancements are already baked in.
Power: At 65W, it’s hard to complain about a dual-core 3.16 GHz chip on a 1333 MHz front side bus. We’re used to the 125W and 130W Extreme Editions. Would someone buy this thing some carbon credits, please?
Dual-Core: In a perfect world, we’d take the 65W TDP, the 3.16 GHz, and add two more cores just so the E8500 would be ready for real-world software optimized for more than two threads. Unfortunately, that list of positives is made possible by scaling back to a dual-core configuration.
Price: At roughly $300, the Core 2 Duo E8500 isn’t a cheap chip. In fact, it’s probably the priciest dual-core processor Intel sells, short of mobile Meroms. If saving money is more your modus operandi, check out the E8400 or E8200, which definitely cost less and will likely yield a solid overclocking experience.
| Final Verdict||Page:: ( 14 / 14 )|
Intel Core 2 Duo E8500
Print Article! | Close Window ]