Sign in with
Sign up | Sign in

How Not To Overclock A Clarkdale

Phenom II X2 555 Vs. Pentium G6950: New Budget Dual-Core Titans
By

There's good news and bad news tied to our overclocking endeavors, as can often be the case when you push hardware harder than it's supposed to go. First, the good news: we achieved a substantial overclock of 4.5 GHz on our retail Pentium G6950 CPU in preparation for this story. We throttled it back to 4.41 GHz to keep the temperatures in check for our benchmarking runs, yielding an impressive 1.61 GHz over the stock frequency.

Now for the bad news: it doesn't count. We killed the processor. Bolstered by reasonable temperature reading and a bit of overconfidence, I pushed the Pentium too hard. My choice of voltage was likely the murder weapon: I set it to 1.475V. In my defense, these are still very new chips, and this is the first G6950 in our lab. There were no warning signs and the CPU lasted through a couple hours of Prime95 testing. Halfway through the overclocked benchmarking runs, the CPU simply died. Not while stress testing Prime95, mind you, but during the AVG virus scan benchmark. Therefore, there will be a follow-up to this article in the next couple of weeks when our replacement arrives. I will be going a second round with the G6950, and this time I'll be limiting myself to a maximum 1.4 V. I will also be testing the CPU on a budget Intel board in the same price range as the Asus M4A785TD-V EVO we used to test the Phenom II X2 555.

Obviously, the results we achieved at 4.41 GHz are not indicative of a sustainable overclock for the components we had on-hand. I did struggle with whether or not to share them in this article. In the end, I decided to include them for curiosity's sake. I know if I were a reader (and not the writer), I would have liked to at least see them. And despite that fact that they aren't really valid from a real-world standpoint, they do give us a glimpse of what a Clarkdale can do at high clocks. It is quite possible that some G9650 CPUs out there might make hit these frequencies more easily with a lower voltage setting.

How To Overclock A Phenom II Black Edition

Here's a CPU we have many more hours of overclocking experience with. At 1.5V, a setting proven quite tolerable by AMD's 45nm Phenom IIs as long as the temperatures are kept in check, our Phenom II X2 555 sample made it to a stable 4 GHz. It couldn't handle 4.1 GHz, though. Knowing that was close to the CPU's upper-bound, we pulled the multiplier back to 19 so that we could increase the HyperTransport reference clock and get a boost from the faster interconnect and memory speeds.

In the end, we took the Phenom II X2 555 to 4.065 GHz with a 213 MHz reference clock and a 713 MHz (1,426 MHz DDR) memory speed:

This provided a nice 850+ MHz boost over the stock speed, facilitating a sustainable 24/7 overclock. We should also mention that we left options like Cool'n'Quiet and C1E enabled, as we appreciate the flexibility of low power usage when the machine is not being taxed. This is something that most users will decide as well.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 130 comments.
This thread is closed for comments
Top Comments
  • 24 Hide
    jasont78 , January 25, 2010 3:29 AM
    good article we like to know ur human and can blow shit up
  • 21 Hide
    alchemy69 , January 25, 2010 3:34 AM
    Bring on the battle of the fanboys. I'll get the popcorn.
  • 20 Hide
    burnley14 , January 25, 2010 3:43 AM
    obarthelemyI see a bunch of overclocking articles... do you have any clue about how many of your readers overclock ? and how many of the public at large ?My guess from personal anecdote would be 10% and 0.01 % resp ?

    It's pretty much a free way to get better performance, so I'm glad they have so many articles about it.
Other Comments
  • 20 Hide
    footsoldier , January 25, 2010 3:25 AM
    Kudos to AMD! Gogogo!
  • 24 Hide
    jasont78 , January 25, 2010 3:29 AM
    good article we like to know ur human and can blow shit up
  • 21 Hide
    alchemy69 , January 25, 2010 3:34 AM
    Bring on the battle of the fanboys. I'll get the popcorn.
  • 20 Hide
    burnley14 , January 25, 2010 3:43 AM
    obarthelemyI see a bunch of overclocking articles... do you have any clue about how many of your readers overclock ? and how many of the public at large ?My guess from personal anecdote would be 10% and 0.01 % resp ?

    It's pretty much a free way to get better performance, so I'm glad they have so many articles about it.
  • 8 Hide
    Anonymous , January 25, 2010 3:43 AM
    That's great news for my next budget PC :D 
  • 14 Hide
    volks1470 , January 25, 2010 3:45 AM
    I'd say a decent majority do overclock, and this site isn't exactly for the general public. Not very people get what's going on here on Tom's.

    POLL!!
  • 8 Hide
    blackjellognomes , January 25, 2010 3:46 AM
    obarthelemyI see a bunch of overclocking articles... do you have any clue about how many of your readers overclock ? and how many of the public at large ?My guess from personal anecdote would be 10% and 0.01 % resp ?


    More like 50% and 5%, I think.
  • 11 Hide
    melangex3 , January 25, 2010 3:47 AM
    If you read this article, you are probably in a higher percentage group than the general public!
  • 14 Hide
    eddieroolz , January 25, 2010 4:00 AM
    Good showing by AMD!
  • 4 Hide
    terr281 , January 25, 2010 4:02 AM
    Most people, at least those who build systems for themselves, friends, and clients (who would read this article), probably overclock their cpu... even if it is only a very modest overclock on the included retail cooler. (Or, higher on an aftermarket one.)

    As such, I must agree that it is a good thing that AMD seems to still have a market. (As such, we won't find Intel being the only player in the CPU market... at least for the next year anyway.)

    With luck, AMD's shift to completely new chips will allow the company to keep a competitive presence in the low-end and mainstream market.
  • -3 Hide
    verrul , January 25, 2010 4:08 AM
    intel has to work harder on their low end mainstream efforts or they will continue to lose ground to amd
  • 8 Hide
    fatkid35 , January 25, 2010 4:29 AM
    i enjoy seeing a win here for amd. makes me happy. two wins actually.$100 dollar dual cores @ stock for stock it wins, due to it has higher stock clocks. secondly it survived the abuse put to it. even if the intel chip will clock higher, it failed. thats hard to forget. "its not the dog in the fight, its the fight in the dog."
  • 1 Hide
    Schip , January 25, 2010 4:32 AM
    Is it just me, or is there something weird on the "Benchmark Results: Synthetics" page. The table titled, "PCMark Vantage Hard Drive Test Score" shows the stock 555 performing better than the 555 when overclocked, which contradicts intuition and the paragraph that follows the table. Not a big deal, just thought I should point it out. Peace!
  • 6 Hide
    envolva , January 25, 2010 4:53 AM
    I think Tom's Hardware should focus on overclock for daily use. I would never go over 1.4V with a brand new processor, and I guess those who do wouldn't do it for daily use.

    So I would really like to see some limitations applied when comparing the value of each processor. Some limitations would apply like max voltages, max temps, power saving on. Disabling custom features like Intel's turbo boost or hyper threading would be fair game if it made the overclocking easier/safer.

    I appreciate the fact that you push the chip to the limit so the reader don't have to, but in the end the overclock results aren't really useful without guesswork of how much the performance would decrease when you apply daily use limitations. Can a Pentium G6950 keep 4.2GHz at 1.4V? Can the Phenom II 555 reach 3.8GHz at 1.4V?

    Personally I wouldn't go over 1.35V with my i7 920, but I understand each fabricant, and each processor have its own limits. I'm not aware of the AMD processor stock or max voltage, but in this case I'm guessing 1.4V is a fair number to impose as limit with these two competitors.
  • 3 Hide
    DarkMantle , January 25, 2010 5:27 AM
    On "Test Systems And Benchmarks" it says "Mushkin PC3-10700
    3 x 2,048MB, DDR3-1333, CL". Was this a mistake when writting the article or did you really tested 3 dual channel processors with 3 memory sticks?.
  • 6 Hide
    ta152h , January 25, 2010 5:39 AM
    Clarksdale is a big compromise, and I don't know why anyone would buy the Pentium without using the GPU. That's kind of the point.

    If not, you go to the faster Core 2 based Pentium. It's cheaper, runs faster, and isn't lobotomized like LGA1156 processor. At least with the Lynnfield you get the faster memory controller, but with the Clarksdale, you get abysmal memory performance and all the bad compromises of the Lynnfield, without the main benefit. Who'd want this except budget buyers who want to use the GPU?

    The Pentium G6950 is a real bomb. It's a horrible, brain-damaged processor that will be sold to the masses, because it can make for a cheap platform suitable for surfing. But when you quantify the performance, it's going to suck, bad. Better off with the older Pentiums, or an AMD product.

    Also, I'd be really curious about the Athlon X2s. The Athlon X4 is just an inferior Phenom at a lower price, but the Athlon X2 has the much larger L2 cache, which could make it a very interesting product - especially considering the price. It should also use slightly less power, saving even more money.
  • -3 Hide
    ta152h , January 25, 2010 5:48 AM
    What's with the weird L1 cache sizes anyway? The Athlon still uses 128K, 3 cycle L1 cache. And for the G6950, why do you have it 4 x 32K, and the Phenom II 2 x 128K? If you want to call the L1 cache seperate data and instruction, at least do it consistently, instead of making it confusing by applying it to the Pentium, but not to the Phenom.

    It's also worth noting in the thermal limits that the Pentium G6950 includes a lot more than the AMD product, including the PCI-E controller and GPU. It's not an apples to apples comparison.
  • 3 Hide
    carlhenry , January 25, 2010 6:34 AM
    ta152hWhat's with the weird L1 cache sizes anyway? The Athlon still uses 128K, 3 cycle L1 cache. And for the G6950, why do you have it 4 x 32K, and the Phenom II 2 x 128K? If you want to call the L1 cache seperate data and instruction, at least do it consistently, instead of making it confusing by applying it to the Pentium, but not to the Phenom. It's also worth noting in the thermal limits that the Pentium G6950 includes a lot more than the AMD product, including the PCI-E controller and GPU. It's not an apples to apples comparison.


    its apples to apples because they compared on the given price point, not on the feature set. it'd be apples to pineapples if you compared a 100$ cpu vs a 200$ cpu eh?
  • 8 Hide
    noob2222 , January 25, 2010 6:58 AM
    burnley14It's pretty much a free way to get better performance, so I'm glad they have so many articles about it.

    Well, obviously from this article itself, I wouldn't exactly use the term "FREE" since you would have just purchased a processor and fried it.

    Would be better to use the term gambling since nothing is guranteed.
Display more comments