Intel E8600 Review: E0 Stepping, the Last Great Dual Core?

Overall Score

Introduction: By JumpingJack

The computer enthusiast has had quite the treat over the past two years, Intel launched the successful Core 2 Duo, and AMD launched the Phenom processor in response. The head to head competition has driven prices to new lows for some of the highest performing processors yet built. However, building a processor takes enormous time, effort, resources, and validation. Not all processors are created perfect either, some contain bugs or errata or there is room for improvement to squeeze out one higher speed bin and improve performance or power consumption. For the beginner enthusiast, you may hear the term ‘stepping’, which simply refers to a revision of the product line. More technically, a stepping, in this context, is a revision to one or more of the photolithographic mask(s) which is used to develop patterns on the die that ultimately create transistors and wire them up to make the chip.

The rate and number of steppings is governed by many factors and each company handles the details differently. AMD, for example, start their stepping counts at the major release of a new family of processors, and continues the progression even through shrinks across process technologies. For example, the last 90 nm K8 core (Windsor) had a stepping designation of F3, which produced the AM2 line of CPUs, where as the 65 nm Brisbane line was stepping G0 at introduction. Brisbane was later revised up to G1 and was a nice improvement. Intel, on the other hand, typically starts the revision label at the beginning for an individual product and retires the product at the end of the life cycle. Conroe, the famous Core 2 Duo, launched with a B2 stepping designation which was then revised to B3 stepping at the onset of the quad core. B3 stepping was around for a short while when the stepping changed again to G0, which was a great improvement over it’s predecessor. Inklings of the B3 to G0 switch, and its improvements, will be seen in the review to come.

[IMG]E8600stockcropped.jpg[/IMG]

Today we are going to be looking at the new E0 stepping of Intel’s Penryn line, which is a revised version of the previous C0 stepping that has been on the market for quite sometime now. We will be comparing a new E0 stepping E8600 versus an older C0 stepping Xeon E3110 (identical to E8400). Intel does a remarkable job of documenting their revision history, more so than one would think. Any change to any product, no matter the detail, is published to their web site and is accessible for anyone to look through and read. These come in the form of product change notifications or PCNs, and the database may be searched at this website. Looking over the PCN database, Intel published the revision change for their Quad Core 9550 on July 25th, the Quad Core Xeon on August 1, and some desktop and mobile SKUs a little later. Because Intel unified their architecture such that the same chip makes a mobile, desktop or server part it is important to read all the PCNs of each product type in order to capture every detail. Simply slog through the PDFs and look for products that changed from C0 to E0 stepping.

Some of these changes include:

  • Release of different and new packaging materials that are now halide free (addressing environmental awareness)
  • New instruction extensions (XSAVE/SRSTOR)
  • Support for power status indicator (unknown at this time)
  • Speed path improvements enabling higher speed bins

The last one of this list is of the greatest interest, as it should hopefully mean some higher over-clocking potential. The brief list shown here can be found in PCN108488-00, PCN108663-00, and 108660-00.

Intel also documents these changed items in spec updates, though those updates typically appear close to or just after release. These can be found under the documentation for the processor series in question, for example www.intel.com/design/core2duo/documentation.htm will show specification updates, including errata fixed with each released stepping. The current specifications do not list any specifics for the E0 stepping, but according to the update schedule we can expect one very soon.

Ok, so there it is, what a stepping means in terms of the usage of the term as well as some of the details that are going into this revision. So without further ado… let’s get to overclocking.

Benchmarking/testing/overclocking, By tool_462

IHS
Intel E8600 ES:
[IMG]E8600ES.jpg[/IMG]

Intel E3110: (remember, this is identical to E8400)
[IMG]E3110.jpg[/IMG]

Bottom
E8600:
[IMG]E8600bottom.jpg[/IMG]

E3110:
[IMG]E3110bottom.jpg[/IMG]

Benchmarking Setup:

  • DFI P35-T2RL “Blood Iron” BIOS 03/11/2008
  • 4GB (2 x 2GB) G. Skill DDR2-1000 @ 2.1v and 1:1 5-5-5-15 for all tests
  • eVGA 9800GTX (800/1965/1175) (Core/shader/mem)
  • Western Digital 74GB 8MB Raptor
  • Water Cooling (Apogee GT + MCW30 + MCP655 + 2x 120mm Swiftech rad + 1/2” ID Tygon)
  • Vista Ultimate x64

Desktop:

A bit messy you see, but it could be worse. I thank Truman Capote’s true masterpiece “In Cold Blood” (seen to the left of the G15 keyboard) for getting me through three runs each of multiple benches, at multiple CPU speeds.
[IMG]rig1.jpg[/IMG]

Case:

All testing done fully enclosed as pictured.
[IMG]rig2.jpg[/IMG]

Initially I had planned on running stock vs. stock benches but decided the best route is to provide straight forward comparison between E0 and C0 at equal clocks, and then at their highest stable overclock while keeping voltage at or below 1.4v, a core voltage that should be safe with 45nm Wolfdale CPUs for 24/7 usage. My E3110 has run safely at 1.4vcore Folding @ Home (full load) for a couple months and has shown no signs of maximum overclock degradation at that voltage.

For this testing, all voltages were left at stock other than core voltage. I’m perfectly aware that increased stability can be gained through tweaking of other settings, but I wanted the chips to have a consistent testbed for comparison.

Benchmarks ran include:

  • Super Pi 1m/8m
  • Cinebench R10
  • SiSoft Sandra
  • WinRar
  • 3DMark 2006
  • 3DMark 2001 SE
  • Half Life 2: Lost Coast
  • Crysis

Settings and version information for each test can be seen on the graph titles.

Before I delve into an array of graphs and you ignore everything I write, let me discuss what most of you are reading this for: Overclocking.

First, I began by finding the maximum stable (3 hours of OCCT due to time constraints) overclock at each CPU’s stock core voltage.

With both CPU’s stock core voltage being 1.24v, it was an easy comparison. The E3110 (C0) achieved a stable overclock of 3.51GHz (390 x 9.) While the E8600 (E0) was able to obtain nearly identical FSB speeds at stock core voltage resulting in a core speed of 3.93Ghz (393 x 10) rock solid stable.

I was pleased to say the least, being what could most certainly be the last of the great dual cores (with Nehalem on the horizon and no word on making anything less than a quad core) still had some factory headroom left in it.

Particularly for gaming, a 3.93GHz Wolfdale is nothing to shake a stick at and at the present, will oust nearly every quad core (save the QX chips with unlocked multipliers that generally clock to this speed for 24/7 usage) in nearly every game for the time being.

I would have loved to do a temperature comparison between the two steppings, but, as many of you know, early releases of the 45nm CPUs from Intel were often plagued with stuck, inaccurate, or completely off-base temperature sensors. Sadly my E3110 is one of those CPUs and consistently reads below ambient which is entirely impossible with water cooling.

I will say that the E8600 was in no way difficult to keep cool, at stock voltage it never climbed over 40C at full load (assuming the temp sensor is correct!) This is with a modest water cooling setup, this CPU is certainly an air cooling gamer’s dream.

As I mentioned previously, I am comfortable with 1.4vcore with this generation of CPUs and chose to use that as my limiting factor in testing the 24/7 overclockability of the E8600. No other voltages were tweaked, which would net an even higher result.

The overclock attained at this voltage is referenced in the benchmark graphs as “24/7 Overclock” which equates to a 100% stable system with 1.4v vCore or less. Without being able to bump the CPU VTT, the E3110 was able to achieve an overclock of 4.0GHz (445 x 9) for 24/7 usage.

With even less voltage (1.392v) the E8600 was able to achieve a whopping 4.50GHz (450 x 10) that was 100% stable and remains stable through (currently 5 hours 40 minutes) of OCCT testing.

[IMG]E8600-45-1392vcropped.jpg[/IMG]

Being extremely pleased with these results I have grown excited to see what this CPU will do and will update this review with further results.

 

A teaser update: (before I even finish writing the review!)

4.75Ghz stable for 1 hour OCCT and every bench I’ve thrown at it with 1.456vcore! Surely more to come.

[img]475-1456cropped.jpg[/img]

I will let the graphs speak for themselves, remember the “24/7 Overclock” is running at 4.0Ghz for the E3110 and 4.5Ghz for the E8600. System was flawlessly stable throughout the entire test suite.

WinRar: (Higher is better)
[IMG]WinRar.jpg[/IMG]

Cinebench R10: (Higher is better)
[IMG]Cinebench.jpg[/IMG]

SuperPi 1m: (Lower is better)
[IMG]SuperPi1m.jpg[/IMG]

SuperPi 8m: (Lower is better)
[IMG]SuperPi8m.jpg[/IMG]

SiSoft Sandra Whetstone: (Higher is better)
[IMG]Whetstone.jpg[/IMG]

SiSoft Sandra Dhrystone: (Higher is better)
[IMG]Dhrystone.jpg[/IMG]

SiSoft Sandra MultiMedia Float: (Higher is better)
[IMG]MM-float.jpg[/IMG]

SiSoft Sandra MultiMedia Integer: (Higher is better)
[IMG]MM-int.jpg[/IMG]

3DMark 2006 CPU score: (Higher is better)
[IMG]3d06-cpu.jpg[/IMG]

As expected the CPUs perform nearly identically (well within the margin of error) in tests when running at identical settings. Nothing, performance wise, (other than overclocking obviously) was improved upon in the switch from C0 to E0 stepping. Once you factor in overclocking at equivalent voltages, you begin to see the advantage that E0 will have over previous steppings in terms of real world performance for the enthusiast. Let us move on to gaming related benchmarks.

Here I made my best effort to create CPU limited benches to test the processor specifically rather than putting too much of the load on the 9800GTX.

3DMark 2001 SE: (Higher is better)
[IMG]3d01.jpg[/IMG]

3DMark 2006: (Higher is better)
[IMG]3d06-Overall.jpg[/IMG]

Half-Life 2: Lost Coast: (Higher is better)
[IMG]HL2.jpg[/IMG]

Crysis:
[IMG]Crysis1024.jpg[/IMG]

Very similar results to the ones we see in the CPU intensive benchmarks. I plan on updating the Crysis and Half Life 2 benches with a “real world” gamer comparison at the native resolution which I play my games, of 1680 x 1050 and settings controlled where I keep them for gameplay. This will provide a direct visual relation of the FPS gains that can be had with a high clocked Wolfdale.

Bringing everything together, we can see that Intel has done another great job in improving an already powerful and efficient line of processors. You will be pleased to know that previous 45nm CPUs will now be released with the new E0 stepping and are rumored to be hitting the market mid-late August. As B3 Quads were cast to the wayside in search of the early G0’s, I suggest verifying that you are in fact purchasing an E0 stepping CPU before spending your hard earned cash.

With regards to the E8600, it is Intel’s highest clocked Core 2 Duo CPU and may reign as the baddest dual core on the block for a long time coming. Priced at $266 it is a reasonable buy, especially for gamers, if you factor that it will be the highest performing chip in single threaded games (currently the overwhelming majority) other than a few “golden” QX9650’s and QX9770’s which would require much better cooling to achieve the same speeds reachable on air cooling with a dual core E0 CPU.

Summary of improvements:

  • Improved power saving
  • Better overclocking at lower voltages
  • Improved temperature sensors

Please, please! Ask me questions and request more information where needed, I will gladly respond to PM’s or posts at XCPUs.com, my member name is tool_462. A big thanks to JumpingJack for writing the technical side of the introduction that I surely could not have managed to do. Also a thank you to XCPUs editors who are creating and improving our new front page as we speak, and more than likely editing this article for errors are bound to occur when you have worked 14 hours, and benched/wrote another 10 hours on top of that!

Discuss in the Forums

 

OVERCLOCKING UPDATE (8/12/2008 9:04 AM)

I have successfully gotten the chip up to 4.9GHz at 1.6V. It will run 3DMark01 and SuperPi at those speeds.

CPU-Z Validation.

Just as an update now that I’ve had more time to push the CPU and tweak settings.

 

I am currently running 24/7 (28 hours+ OCCT stable) at 4.60Ghz with 1.408 vcore showing in CPU-Z and 1.4125 vcore set in BIOS. This may actually be more vcore than I need as I had initially suspected instability of the CPU. As it turned out my RAM had reverted to stock voltage during a switch of RAM for benchmarking purposes. I am comfortable at 1.4v with 45nm Intel CPUs for 24/7 usage under water cooling and CPU PLL voltage is 1.18v which is very safe.

I have yet to hit 5Ghz at a respectable voltage, but Forum Wars are approaching (have arrived!) and the cooler temperatures of Minnesota’s fall/winter season bode well to OCing. There is a reason my additive infused water won’t freeze till -30F…

SHARE THIS POST

  • Facebook
  • Twitter
  • Myspace
  • Google Buzz
  • Reddit
  • Stumnleupon
  • Delicious
  • Digg
  • Technorati

Leave A Response