Sign in with
Sign up | Sign in

Intel’s Core i7 Platform, Configured, And Priced

Overclocking: Core i7 Vs. Phenom II
By

What We Recommend

Intel Platform

Motherboard

Gigabyte GA-EX58-DS4

Processor

Intel Core i7 920

Cooler

Thermalright Ultra 120 Extreme

Memory

Corsair XMS3 6 GB (3 x 2 GB) DDR3-1600

Graphics

AMD Radeon HD 4870 X2

Storage

Samsung Spinpoint F1 1TB 7,200 RPM

Power

PC Power and Cooling 750 W

Optical

Lite-On DH-4O1S


Some of the component decisions made in this comparison were influenced by hardware on-hand in the Tom’s Hardware SoCal lab. But we made this project a team effort between Thomas Soderstrom, Don Woligroski, and myself using retail hardware to verify the viability of our overclocks using less expensive components.

The back-side of Intel's Core i7 920The back-side of Intel's Core i7 920

As a result, we’re benchmarking with Asus’ Rampage II Extreme and an engineering-sample processor. That particular motherboard isn’t going to be the budget-buyer’s choice, nor is an i7 920 ES going to be representative of retail product. But if you flip over to Don’s $1,250 System Builder Marathon setup, you’ll see he was able to achieve similar overclocking results with the less-expensive Gigabyte X58 motherboard and a boxed i7 920. The real shocker there was that his retail CPU featured memory multipliers that my sample didn’t have—a fact that that Thomas demonstrated in his $2,500 System Builder Marathon story.

As configured, our value-oriented Intel system costs just north of $1,550 at the time of writing. Bear in mind that only four components differ between the AMD and Intel builds: motherboard, processor, heatsink, and memory. We’ve kept everything else the same in the interest of comparing performance in an overclocked environment.

Overclocking i7

There is no longer much mystique behind Core i7 overclocking. The non-Extreme models have locked multipliers, so extra performance comes from Bclk adjustments. Tweak your Bclk, your processor voltage, memory multiplier (the retail boxed product gives you a lot more flexibility than a tray or engineering sample chip does), and QPI voltage.

We were able to boot at 4 GHz and run fairly stable at 3.9 GHz. However, even at the 3.8 GHz setting on which we eventually settled, we saw load temperatures peaking at 85 degrees Celsius. For day-to-day usage, that’s plenty aggressive, thank you very much. Our final settings were a 20x CPU multiplier, 190 MHz Bclk, and stock voltages all of the way around.

Hopefully, these settings take into account the variance you'll see from one CPU to another, since we wouldn't expect every 920 to hit 4 GHz.

Display all 221 comments.
This thread is closed for comments
Top Comments
  • 17 Hide
    bernardv , January 12, 2009 7:23 AM
    This Phenom II overclock is very very far from optimal. Here are some points:
    1. 790FX board should be used, FX is the top end chipset.
    2. Ph2 doesn't support ACC, check the forums where skilled people got this CPU beyond 4GHz on air, they know what they were doing.
    3. If you're doing bench you should try to get an optimal RAM speed as well. Your RAM was running 30% below stock 1066!!! You must raise core do better.
  • 14 Hide
    cangelini , January 12, 2009 5:41 AM
    Uncle,
    With the Black Edition, that's really the value there. You want to use the multiplier for a clean, easy overclock. Once you've found your ceiling, use the reference clock to fine-tune the setup.
  • 13 Hide
    jtnstnt , January 12, 2009 6:17 AM
    I'm curious about the heatsink selection on the Phenom II. I looked up the Ajigo MF091 and it didn't look too impressive. While the Intel system gets the Thermalright 120mm True extreme. Whats up with that?
Other Comments
  • 0 Hide
    Roffey123 , January 12, 2009 5:00 AM
    At last a P2/i7 comparison to shut the people whining for one up. In all I think its a fair comparison - although I am disappointed that AMD's OC potential wasn't as good as they made it out to be. But as all overclockers know - not all chips are the same or have the same OC potential.

    The 3DS Max benchmark was puzzling as well, considering that i7 has 8 threads at its disposal. Do you guys have any idea as to why the AMD beat the i7 in that test; despite having only half as many threads?
  • 0 Hide
    unclefester , January 12, 2009 5:17 AM
    I thought it funny they use an ATI card on the Intel board (not that it makes a difference).
    Would have liked to see what the PII can gain or not gain by using the FSB not just the multiplier.
    Didn't see anything about memory timings either. As most PII OC's have been done at @880MHz.
    Overall a decent review.
    So Ford vs Chevy goes on.
  • -1 Hide
    apache_lives , January 12, 2009 5:37 AM
    I would love to see a high end overclocked Core 2 Quad 9 series in the tests here to see a price comparison - cheaper and already established platform etc.

    Also note that i7 platform should also feel more alive and responsive (un-benchmarkable) thanks to high memory bandwidths and more threads to balance things out like the P4C's and CL2 Dual Channel DDR1 - reguardless if it performed lower then the AMD A64's at the time it felt more responsive!

    Intel also has that more solid platform to back that CPU and options for Crossfire AND SLi so its more the premium option, but that cheaper AMD setup is too close to the i7 for my liking - a few shifts in prices and models should show an interesting result in the new few weeks.

    AM3 should be interesting, higher headroom + lower power etc (like 939 to AM2).
  • 14 Hide
    cangelini , January 12, 2009 5:41 AM
    Uncle,
    With the Black Edition, that's really the value there. You want to use the multiplier for a clean, easy overclock. Once you've found your ceiling, use the reference clock to fine-tune the setup.
  • -4 Hide
    vekere , January 12, 2009 5:42 AM
    I hear AMD is going to hire Sylar to look at the new Core i7 processor so this year we will have an answer.
  • 1 Hide
    kschoche , January 12, 2009 5:51 AM
    The WINRAR tests are so close to exactly half the time on the Intel chip compared to the AMD chip that I'd love to see how the results show up if you turn off hyperthreading, not that its a particularly meaningful result, but would be cool to look at still.
  • 13 Hide
    jtnstnt , January 12, 2009 6:17 AM
    I'm curious about the heatsink selection on the Phenom II. I looked up the Ajigo MF091 and it didn't look too impressive. While the Intel system gets the Thermalright 120mm True extreme. Whats up with that?
  • -5 Hide
    jtnstnt , January 12, 2009 6:18 AM
    Please explain.
  • 6 Hide
    apache_lives , January 12, 2009 6:38 AM
    jtnstntI'm curious about the heatsink selection on the Phenom II. I looked up the Ajigo MF091 and it didn't look too impressive. While the Intel system gets the Thermalright 120mm True extreme. Whats up with that?


    Heh very good point there, although it depends on wether the limits of the OC was architectural or thermal
  • 0 Hide
    jtnstnt , January 12, 2009 6:44 AM
    I was wondering because i didn't see an explanation in the article.
  • 3 Hide
    cangelini , January 12, 2009 7:09 AM
    jtnstntI'm curious about the heatsink selection on the Phenom II. I looked up the Ajigo MF091 and it didn't look too impressive. While the Intel system gets the Thermalright 120mm True extreme. Whats up with that?


    The Ajigo sports a copper base, four heat pipes, and an aluminum fin array. You're right--it's not a particularly spectacular heatsink, though it was the best Phenom II-compatible cooler in the lab. As a result, I made it a point to mention that stepping up to a better cooler might have yielded better overclocking, but would also erode some of AMD's price advantage as well.

    Heatsink aside, after talking to a number of colleagues from other sites at this year's CES, it sounds like the overclocks here are indeed representative of what others were able to do, even using different heatsinks. I'll look into getting more in the way of air cooling and hope for better results.
  • 17 Hide
    bernardv , January 12, 2009 7:23 AM
    This Phenom II overclock is very very far from optimal. Here are some points:
    1. 790FX board should be used, FX is the top end chipset.
    2. Ph2 doesn't support ACC, check the forums where skilled people got this CPU beyond 4GHz on air, they know what they were doing.
    3. If you're doing bench you should try to get an optimal RAM speed as well. Your RAM was running 30% below stock 1066!!! You must raise core do better.
  • -1 Hide
    arkadi , January 12, 2009 7:27 AM
    jtnstntI'm curious about the heatsink selection on the Phenom II. I looked up the Ajigo MF091 and it didn't look too impressive. While the Intel system gets the Thermalright 120mm True extreme. Whats up with that?

    P2 did go more than 60c, that what the article said any way. So i don't think it meter
  • 0 Hide
    falchard , January 12, 2009 8:01 AM
    I think you got the wrong end of the silicon with the chip. Anandtech was able to reach a stable 3.9 GHz with 1.52 volts.
  • 6 Hide
    cangelini , January 12, 2009 8:24 AM
    bernardv

    Bernard, 790GX is actually a newer platform than 790FX, and the only way stepping *back* to that one could have made a difference would have been if we were testing multiple graphics cards, in which case the chipset's PCI Express connectivity would have benefited it.

    Re: ACC--not sure if you were aware, but the enhancements rolled into ACC that made the Phenom a better overclocker are *already* rolled into Phenom II. Therefore, it doesn't need to support ACC since those capabilities are in-hardware.

    Re: RAM speed, we're using DDR2-1066. Per this page, that is the maximum official supported spec for the AM2+ version of the chip. Not sure if you mis-read that we were using DDR2-800, perhaps.

    All the best!
  • 6 Hide
    cangelini , January 12, 2009 8:26 AM
    falchardI think you got the wrong end of the silicon with the chip. Anandtech was able to reach a stable 3.9 GHz with 1.52 volts.


    And our own German lab was able to run 3.8 GHz at 1.55V and even boot at 4 GHz. And there's the variability of overclocking for ya. Everyone's experience is going to be different, and I can assure you that an extra 100-200 MHz would not change the tide of this comparison.
  • 2 Hide
    eodeo , January 12, 2009 8:32 AM
    Could you please disclose what are you testing in Max 9, maybe even share the scene so we could also test it? By your own scoring, i920 @ 3.8ghz in this test is slower than c2q 6600 @ 2.4ghz. I know something whent wrong, but I dont know what.
  • 2 Hide
    cangelini , January 12, 2009 8:38 AM
    eodeoCould you please disclose what are you testing in Max 9, maybe even share the scene so we could also test it? By your own scoring, i920 @ 3.8ghz in this test is slower than c2q 6600 @ 2.4ghz. I know something whent wrong, but I dont know what.


    Sure! From the Test Setup page (page 4) of this story:
    Version: 9.0, Rendering Dragon Image at 1920x1080 (HDTV)

    In essence, this is a single image being rendered out at 1920x1080. If you email me, I'd be happy to send along the source data (cangelini at bestofmedia dot com).
  • 0 Hide
    nashville , January 12, 2009 8:39 AM
    hey, in your ph2 review, power consumption on i7 was measured from 12v rail. i read uncore/cache dont get power from 12v rail, is this correct. if it is, will affect your measurements?
  • 0 Hide
    eodeo , January 12, 2009 8:53 AM
    Thanks for the quick response Chris! Will do.
Display more comments