Back to Article

  • cooklaw - Saturday, February 21, 2009 - link Reply
  • 529th - Sunday, January 25, 2009 - link

    Thanks for the article/research. I've found this interesting. I also have a 3838A687 920 that I got on sale (229$) that I'm almost willing to lend for some overclocking results...???... but from what I make of this is that the motherboard tolerances are a bit more key to its overclockability; am i correct? Looking forward to your follow up article between the motherboards.

    P.S. i7 build is not up-n-running yet, ETA 6 months, lol
  • Rajinder Gill - Sunday, January 25, 2009 - link


    Just been testing a couple of ES 965's, 3830A and yet another 3838A. The 3838 is on a par with my weaker 920 using air/water cooling. The 3830A 965, is still not as good as the 3835A 920 retail. Out of 4 CPU's, I only have on that can run tight CAS at higher BCLK. WIth the 965's it's a little moot as you can use the higher multipliers, although I still find that the 3830A (the better 965), needs waaay more Uncore voltage and even then it's nowhere near Prime stable like the 3835A 920.

    Sub-zero cooling, the 965 3838 cold bugs very early, making it a poor clocker both sides of the spectrum. The 3830A 965 is a little better but I have not managed to get it stable for 3D past 4.6GHz.

  • eva2000 - Monday, February 09, 2009 - link

    Definitely can confirm that cpus will differ in terms of CPU VTT, tested 2x i7 920 3836A756 and CPU VTT handling differences are huge"> !

    i7 920 3836A756 #1

    * max memtest86+ bootable bclk = 224 requiring 1.58-1.60v CPU VTT
    * max super pi 32m bclk = 222bclk (without IOH/ICH voltage tweaks) & 228bclk (with IOH/ICH voltage tweaks) requiring 1.60-1.61v CPU VTT
    * max CPUZ Validation = 230bclk requiring 1.60-1.63v CPU VTT

    i7 920 3836A756 #2

    * max memtest86+ bootable bclk = 229bclk requiring 1.40-1.48v CPU VTT
    * max super pi 32m bclk = 225bclk (without IOH/ICH voltage tweaks) & ???bclk (with IOH/ICH voltage tweaks TBD) requiring 1.48-1.50v CPU VTT
    * max CPUZ Validation = 231bclk requiring 1.50v CPU VTT

    2nd cpu also likes vcore closer to cpu vtt for stability whereas 1st cpu didn't have such a strict requirement.
  • DrMrLordX - Sunday, January 25, 2009 - link

    That's very interesting. Did anyone get the coldbug with Penryn-class quads? This certainly puts Phenom II in a different light since it has been demoed at 6 ghz under LN2 and 6.3 ghz under LHe2 (suicide runs only, of course). Reply
  • CSGLEON - Saturday, January 24, 2009 - link

    i know this might be a little off topic but....
    would i be able to tell much of a differnce between a Intel Core 2 Quad Q9550 Yorkfield 2.83GHz and a Intel Core i7 920 Nehalem 2.66GHz? i play games and need to upgrad my cpu. but would have to upgrade everything for a i7 and spend a lot more money.

  • Iketh - Saturday, January 24, 2009 - link

    for the most part no Reply
  • aigomorla - Friday, January 23, 2009 - link

    i absolutley love i7 bashing.

    Its funny how only a handful of people actually got it for something useful.

    I used to hold the largest 4ghz yorkfield collection on this forum.

    I also hold one of the most expensive h2o systems on this forum/site.

    If you got an i7 because it was a FAD thing, and u had to jump the wagon, then shame on you.

    If you got the i7 knowing what your getting, because u actually use the 8 threads it offers, then its a win.

    Theres 2 different sides in people when it comes to computers.
    1. think there just an upgraded form of a calculator.
    2. think of it as art, and to tune and unlock that art, is a whole another world.
  • Cuhulainn - Sunday, January 25, 2009 - link

    I absolutley(sic) love bragging.

    So impressed.
  • marc1000 - Friday, January 23, 2009 - link

    somebody already asked of them? AMD has greater expertise with the IMC, even if the cores are not as powerfull as the cores of the i7. how weel do theyr new cpu overclocks? Reply
  • Rosaline - Friday, January 23, 2009 - link

    As I recall, the i7 940 CPUs come by default with faster uncore settings. Could this imply uncore/imc binning, and so give better overclocking performance? Reply
  • 7Enigma - Friday, January 23, 2009 - link

    Could you define this (you used it on the first page to describe the mobo's). I assume it means fine/good, but other than The Simpsons, I've never heard it used before! :)

    Duff man.....can't.....breath....
  • Rajinder Gill - Friday, January 23, 2009 - link

    Over here in the UK 'duff' is a slang term for 'faulty' or something that is not quite right.

    It's amazing the impact a cartoon can have on regional

  • strikeback03 - Friday, January 23, 2009 - link

    does "ace" mean something good? I forget if it was here or elsewhere I saw that one recently. Reply
  • Rajinder Gill - Friday, January 23, 2009 - link

    yeah 'ace' is good - to me Reply
  • 7Enigma - Friday, January 23, 2009 - link

    Whoops should have figured out that myself. I hastily read the sentence and misunderstood. Should have been able to easily see you meant the mobo's were the cuplrit and not the CPU's. Thanks for the clarification though. :) Reply
  • Nimbo - Friday, January 23, 2009 - link

    It is fresh in my mind the memorable article by Kris Boughton "Overclocking Intel's New 45nm QX9650: The Rules Have Changed"
    Published a year ago here at Anandtech. Impressive piece of information about overclocking. I was hopping something like that when I read the title, not anecdotal information about some chips overclocking better than others.
  • Rajinder Gill - Friday, January 23, 2009 - link

    Kris is working on something as we speak, although current work commitments mean he is away from base for a while. It'll be worth the wait ;)

  • kevinmarchibald - Friday, January 23, 2009 - link

    Also, the article mentions buying 940 CPUs as well. Was there less variance in overclockability, voltage needs, etc? Reply
  • kevinmarchibald - Friday, January 23, 2009 - link

    The article mentions "DDR3-1600 at CAS6 with the upcoming kits". I just got the Foxconn Bloodrage board and I had planned to buy the Mushkin DDR-1600 CAS7 3x1GB kit. How soon before these CAS6 kits come out and who will offer them? Is the drop from 4.375 latency to 3.75 (roughly a 15 percent reduction) worth it for overclocking and real world performance? Reply
  • Rajinder Gill - Friday, January 23, 2009 - link

    I think Gary meant running the Cas 7 ddr3 2000 kits at Cas 6 1600 as a compromise to overall VTT and VDimm voltages. Personally I have not seen any real gains from running cas below 8 over 1600Mhz memory speeds when in triple channel mode on this platform. Other than Everest reporting a boost, the real world apps seem more than happy, even at cas 9 -1600 actually. Does not mean to say that it'll stop people buying the performance stuff though, just that the variability in the IMC kinda points to compromised overclocking on the lower multiplier CPU's. Reply
  • Doormat - Friday, January 23, 2009 - link

    Do I really need 2000MHz DDR3 RAM or will 1600 do if I only overclock to 3.6Ghz?

    Those are the types of questions I'd like to see answered. If I've got a $750 where should I put my money? A $200 or $250 board? 1333 or 1600 RAM?
  • Rajinder Gill - Friday, January 23, 2009 - link

    1600 rated kits are probably the best bet or middle ground for decent overall performance vs cost. Triple channel 1600 speeds are more than enough to keep the i7 saturated with data. Reply
  • Joe Schmoe - Friday, January 23, 2009 - link

    I'm not really sure I got the point of the article. I just glanced at it before bed. If you are saying some chips from the same batch overclock better than others.. that has always been true. The IMC is the culprit this time instead of the cache.

    I think its interesting that the X58 boards all seemed to be about equal.

    Apparently, at around 350 dollars all the vendors start using the 20 cent capacitors instead of the 15 cent ones.

  • Rajinder Gill - Friday, January 23, 2009 - link

    If you look around the forums, you'll soon realize there are many people that don't or did not know, that's who the article is for really. Its actually on of the reasons I think the overclocking section should now be segregated on site from the mainstream stuff; different needs and different overall goals. Just that you have to appreciate that some are learning while others are already experienced. Reply
  • corporategoon - Thursday, January 22, 2009 - link

    Every article I've read on AnandTech lately that hasn't been written by Anand himself has been absolutely riddled with sentence fragments.

    Did you guys stop proof-reading your articles, or something? It's starting to make my brain hurt.
  • vacant78 - Thursday, January 22, 2009 - link

    On top of that, abbreviations jump from almost every sentence with almost none being explained (VTT, IMC,???). What happened to understendable communication? That article almost asks for a warning: "For computer journalists and tech gurus only". Shame. Reply
  • JarredWalton - Thursday, January 22, 2009 - link

    Not all articles go through the full editing process (re: me), so there are grammatical and spelling errors on occasion. Some times it's a question of timing: wait for additional editing, or just post something we feel our readers would like to hear. Reply
  • DigitalFreak - Thursday, January 22, 2009 - link

    I think they let the guys at DailyTech do article proofreading. :0/ Reply
  • The0ne - Thursday, January 22, 2009 - link

    After reading the article I would had to assume that most enthusiast at these technical levels would already have realize that manufacturing of the different parts would vary from batch to batch and sometimes within the batch itself. Having said this I'm not entirely sure if an article needs to be done about this subject if this is a known issue. I certainly expect the variances and would never expect to achieve the same result as another user.

    This applies to both the technical side of each of the parts and the manufacturing of them.
  • Rajinder Gill - Friday, January 23, 2009 - link

    Its range of swing we're seeing rather than the fact that- 'we've always known that batches vary'. Previous architectures have always been subject to variance, that's obvious to most. What surprised us with Core i7 thus far is that the level of voltage variance and voltage capability even on air and water cooling is far greater than anything we've seen before. Up to 0.2V applied VTT variance on the CPU's alone under 4GHz speeds is something that many have not taken into account. It's not just the variance in voltage though - it's also the maximum applied VTT the processors will boot at, some throw C1 boot code errors if you feed them too much.
    Again, as with all things pertaining to information, you either already know something or don't. This article is for those that may not have known, based upon some of the letters we get there are plenty of users that did not know hence why we stuck up this piece.
    Also, VTT trace loss - it's not really expected for $300+ overclocking dedicated boards to be skimping on copper to a vital signal power rail really. granted it's phase one and many companies possibly did not realize that the IMC was this power hungry when being pushed, it's worth pointing out so that perhaps we see an improvement in the future.

  • subflava - Thursday, January 22, 2009 - link

    Yes, overclocking is definately a case by case basis, but in general there are definately trends in batches/steppings/etc. I think the differences in the case of the i7 are worth pointing out because the magnitude of the variability seems to be much greater than with the Athlon/ least to this point in the early life of the i7. Reply
  • SlyNine - Friday, January 23, 2009 - link

    Got a Core I7 above 4ghz easly, Cannot get this Q6600 above 3.0ghz for the life of me, 3.1 is absolute tops another above crashes on boot. Reply
  • Aberforth - Thursday, January 22, 2009 - link

    The i7, DDR3 are all fictitious architectures designed to stay in the competitive market, there is no significant improvement with these in the real world scenarios, I must say all these multi-cores have very poor thread handling which is never efficient where as power saving is also a big bandwagon people like to jump on, we know Intel’s been doing that power saving stuff for a long time, they are full of it. Anyway...have fun tweaking it. Reply
  • SlyNine - Friday, January 23, 2009 - link

    On a Core I7 @3.5ghz I ripped and transcoded Transformers HD-DVD to a MP4 High Quality file 10Kbps with 5.1 AAC audio, in about 3.5hours.

    Lets see your dual core do that!
  • strikeback03 - Thursday, January 22, 2009 - link

    I'm not really sure what you are trying to say, but if you are looking for low-power stuff I'm sure there will be more articles on nvidia's Ion platform once they get them out the door to review. Reply
  • Exar3342 - Thursday, January 22, 2009 - link

    That is about the most uninformed comment I have read in a long time. Have you not seen the i7 results for 3D rendering and encoding? In addition to these significant improvements, gaming and general performance has also been upped. How is a new i7 CPU at 2.66 at or above in pretty much all benchmarks over a 3.2 C2Q. Go back to your Athlon and stop complaining about things you nothing about... Reply

Log in

Don't have an account? Sign up now