Ivy Bridge on air: The Core i7-3770K overclocked on four motherboards

Overclocking is a sort of rite of passage for PC enthusiasts. You’re not truly hardcore until you’ve pushed your hardware beyond the stock speeds defined by the manufacturer. Back in the day, that meant fiddling with DIP switches, navigating arcane BIOS interfaces, and sometimes even making hardware modifications with a #2 pencil. Also, we had to walk barefoot through several feet of snow just to get our hands on the hardware; Newegg wasn’t around, and Amazon only sold books.

Today, clock boosting couldn’t be easier. AMD and Intel both offer CPUs with fully unlocked multipliers, the holy grail of overclocking. Pushing one’s CPU past its default speed requires little more than turning up the multiplier. That can be done through increasingly user-friendly firmware interfaces and Windows software. Some boards will even overclock themselves—with and without your permission.

The obvious question, of course, is how Ivy Bridge fits into the overclocking picture. Intel’s Sandy Bridge CPUs have offered a decent amount of clock speed headroom since their debut early last year, and all eyes are on Ivy to see if she can replicate that feat. To find out what the new platform has to offer overclockers, we’ve spend some quality time pushing the new Core i7-3770K on Z77 motherboards from Asus, Gigabyte, Intel, and MSI. Read on to see what we’ve learned.

Free as in headroom

There are all kinds of overclockers in the enthusiast community. Some spend ridiculous amounts of time—and money—on elaborate contraptions that chill CPUs using liquid nitrogen. Others plumb their PCs with intricate networks of water-cooling tubing. Those camps tend to seek out the absolute limits of their CPUs. However, most enthusiasts seem content to tap the “free” clock speed headroom that can be exploited with affordable air cooling. Rather than probing the extremes, we’re going to see how far the Core i7-3770K goes with the sort of air tower one might expect to find inside the average enthusiast’s PC.

After rummaging through the collection of heatsinks that has accumulated in the Benchmarking Sweatshop, we settled on a Thermaltake Frio. This is the same cooler we use on our storage test systems, and at $48 online, it’s eminently affordable. The radiator is lined with more fins than we’d care to count, plus five heatpipes and dual 120-mm fans.

Where’s the second spinner? It interfered with the taller heatspreaders on our Corsair Vengeance DIMMs, so the Frio had to make do with just a single fan for much of our testing. The clearance issue affected all four of the Z77 boards in our labs. We also had trouble getting a couple of larger Noctua coolers to fit, so the problem isn’t unique to the Frio. Most motherboards put their DIMM slots too close to the socket to allow massive air coolers to coexist peacefully with taller memory modules.

As it turns out, the second fan didn’t end up making much of a difference. When we switched to a dual-fan setup with a pair of low-profile Kingston DIMMs, we didn’t notice a substantial decrease in CPU temperatures. The second fan didn’t help the CPU hit higher speeds, either.

That’s not to say that our overclocking endeavors weren’t limited by our system’s cooling setup. Although we managed to get up to 5GHz on a couple of the boards, the CPU temperature climbed high enough to boil water. Thermal throttling was rampant, forcing us to back off to 4.9GHz, the highest speed our Core i7-3770K would sustain under load.

To hit 4.9GHz, we used a 49X multiplier with Ivy’s stock 100MHz base clock. It’s also possible to increase the CPU frequency by turning up the base clock speed, but since that clock affects other system components, it’s best to stick to multiplier tweaking. We were never able to squeeze more than a few MHz out of Sandy Bridge’s base clock, and Intel says Ivy has a similar range.

Our Core i7-3770K needed 1.35V (which CPU-Z detected as 1.368V) to remain stable at 4.9GHz. The system was stressed with AIDA64’s built-in torture test running alongside the rthdribl HDR lighting demo. That tandem made for a good quick test, but 7-Zip proved to be the final arbiter. When we ran it to gather final performance numbers for each motherboard, the app spit out errors with configurations that were previously thought to be stable.

I was a little surprised to hit obvious thermal limits with the chip running at such a relatively low core voltage. Then I looked at the power consumption. Here are some wattage numbers for the entire system, sans monitor, taken from the wall socket. A multi-core Cinebench render was used to generate the load.

Yeah, that’s a pretty big difference under load. 4.4GHz represents the highest speed we achieved at the default voltage, which CPU-Z reported as 1.128V. That 500MHz jump over the Core i7-3770K’s 3.9GHz peak Turbo frequency consumed an additional 17W at the wall socket. Going up another 400MHz with a voltage bump of 0.24V sent the reading on our watt meter up by nearly 100W.

Processor power consumption is proportional to the product of frequency and the square of the voltage, so increasing the latter can really cause the wattage to rise. We crunched the numbers, and our results fit the formula nicely. Serious cooling will be required to push Ivy to its limits. At the same time, taking Ivy up to around 4.4GHz should require only modest cooling, provided the CPU voltage is untouched.

Distinctly different experiences

Now that we’ve established the Core i7-3770K’s limits with our cooling setup, it’s time to get all touchy feely with a stack of Z77 motherboards: Asus’ P8Z77-V, Gigabyte’s Z77X-UD3H, Intel’s DZ77GA-70K, and MSI’s Z68A-GD65. We’ll be concentrating on CPU overclocking, since we’ve already published an in-depth comparison of the Asus, Gigabyte, and MSI boards in addition to a quick look at the Intel one.

Let’s start with the Asus board, because it’s the only one that was stable at 4.9GHz. We didn’t go straight there, though. First, we tapped into the board’s auto-overclocking mechanism, which is activated via the OC Tuner firmware option. The board quickly overclocked itself to 4.2GHz using a 41X multiplier and a 103MHz base clock. OC Tuner didn’t increase the core voltage, and the system was perfectly stable.

Next, we tried our hand at manual tuning. There are two options: firmware and software. The P8Z77-V’s firmware interface is one of the best around, and it’s quick to navigate for folks who are familiar with traditional BIOS layouts. The firmware automatically increases the CPU voltage as the frequency rises, a function that isn’t matched by Asus’ Windows tweaking software.

Since we’ve already spent quite a lot of time poking around Asus’ firmware, we opted for the software route. Asus’ Windows utilities have come a long way in the last few years, and the TurboV EVO app is excellent. The interface is easy to use and nicely matches Asus’ other apps. Options abound, including an auto-overclocking mechanism for Ivy’s integrated GPU. There’s also a separate application with fine-grained control over the power circuitry, and it proved invaluable in getting our CPU stable at 4.9GHz.

Ivy overclocking is best confined to the CPU multiplier, so that’s the approach we took in our manual sessions. The P8Z77-V ran our Core i7-3770K up to 4.4GHz before additional voltage was required to maintain stability under load. Hitting 4.8GHz called for 1.3V, and 1.35V was needed to keep the BSODs away at 4.9GHz. At that speed, we had to set the load-line calibration to “high” to prevent Z-Zip from generating errors.

The P8Z77-V actually booted into Windows with the CPU running at 5GHz, but it crashed instantly under load, and applying extra voltage didn’t help. Already, the CPU was riding the edge of the throttling threshold. We even tried tweaking some of the other system voltages, to no avail.

Next up: the Gigabyte Z77X-UD3H. Again, we started with the board’s auto-tuning mechanism. This one is accessible through Gigabyte’s EasyTune software, which includes a couple of pre-baked settings in addition to an auto option that cranks the frequency and runs its own stability test along the way.

After rebooting and running its Windows-based stability test several times, the UD3H settled on 4.7GHz using a 45X multiplier paired with a 104MHz base clock. CPU-Z reported a CPU voltage of just 1.248V. The system almost made it through our stress test, but when the rthdribl window was closed after a five-minute load, the system promptly crashed. Upon rebooting, the machine scaled itself back to 4.4GHz (43x103MHz) at the same voltage. This configuration survived our torture test without fussing.

Gigabyte evidently needs a better stress test for its auto-tuning mechanism. Falling back to a slower speed is a nice recovery, but the auto-tune scheme should produce a stable result the first time around. Auto-tuning mechanisms are best for newbies looking for a trouble-free overclock and seasoned enthusiasts seeking a stable starting point for their own manual tuning.

Speaking of manual tuning, we bounced between Gigabyte’s software and firmware before favoring the latter. The EasyTune software messed with the temperature readings we were getting from AIDA64, and the app feels like it hasn’t been updated in ages. Gigabyte does have a new power tuning utility for Windows, but it’s a separate app that’s much clumsier than EasyTune. Fortunately, the motherboard’s new firmware interface is a real treat to use.

Mouse-friendly sliders permeate the “3D” interface, and there’s an “advanced” layout if you prefer the comfort of an old-school BIOS. Like the Asus board, the Gigabyte automatically increases the CPU voltage so long as the multiplier is tweaked via the firmware. You’ll need to add voltage manually when overclocking with EasyTune.

4.4GHz once again proved to be the CPU’s limit at stock voltage. However, we couldn’t get 4.9GHz stable no matter how much juice was pumped through the processor. Windows wouldn’t load until the CPU was given 1.35V, but that wasn’t enough to stave off the BSODs under load. Adding voltage only resulted in throttling and program errors. Adjusting the load-line calibration didn’t make the system more stable, and neither did tweaking other power settings and system voltages.

In the end, the fastest stable configuration was 4.8GHz on 1.35V. The board was happy booting Windows and running our stress test on 1.325V, but 7-Zip errors persisted until we applied more voltage.

Intel’s DZ77GA-70K is the only board of the bunch that doesn’t have an automatic overclocking system. We’ve actually seen auto-tuners on Intel boards prior to the Sandy Bridge generation, but the company’s recent efforts seem focused on a sort of auto/manual hybrid accessible via the firmware. The main interface offers a CPU clock speed slider that users are free to drag up to 4.5GHz. The slider increases the multiplier without touching the base clock or the voltage. It does, however, ramp up the current limit so the CPU isn’t starved for amperage.

For our pseudo-auto overclock, we dragged the slider all the way to the right and rebooted. The system returned at 4.5GHz, and it remained BSOD-free for the extent of our stress test. Hitting higher speeds would require getting our hands a little dirtier.

I really like Intel’s new motherboard firmware, and its advanced overclocking options offer plenty of multiplier headroom to hit higher speeds. The drop-down menus for the voltage settings are a little cumbersome due to the sheer number of options, though. Intel would do well to allow users to key-in voltages directly. It would be nice to see an “auto” setting that increases the CPU voltage automatically as the frequency rises, too.

Ultimately, most of our time was spent overclocking the Intel board using its Windows software. The Extreme Tuning Utility is very slick, offering just enough options along with a nice monitoring panel. There’s also an integrated stress test, although we elected to use our own.

Pushing higher than 4.5GHz at stock voltage proved impossible, but an extra 100 millivolts got the CPU to 4.7GHz without issue. 4.8GHz required another 100mV, which produced a 1.336 CPU voltage according to CPU-Z. That was enough juice to get the CPU to load Windows at up to 5GHz. However, the system crashed under load. Applying more voltage resulted in throttling, which struck again when we ran our 7-Zip and x264 tests at 4.9GHz. We had to sacrifice another 100MHz to get a rock-solid system.

Finally, we come to the MSI Z68A-GD65, which has the easiest auto-overclocking mechanism of the bunch. All the user has to do is hit the OC Genie button in the corner of the board. When the button is depressed, the board boots using an MSI-optimized overclocking profile. Users can tweak this profile themselves, but we went with MSI’s defaults for this leg of the overclocking journey.

OC Genie proved very conservative, taking the Core i7-3770K up to 4.2GHz by adjusting only its multiplier. The system endured our stress test without breaking a sweat, so we started tweaking settings manually.

MSI’s ClickBIOS II firmware is a big improvement over the company’s initial efforts, and I like the fact that it has an accompanying Windows application with an identical interface. There’s something to be said for maintaining consistency between the firmware interface and Windows utilities. Unfortunately, the Windows app takes forever to load up and apply changes. Modifying the multiplier requires a reboot, which isn’t the case for other Windows tuning apps, including MSI’s own Control Center software.

Control Center still takes longer than it should to load, but changes are applied almost instantly. Multiplier tweaking was all it took to get the Core i7-3770K up to a stable 4.5GHz. Higher frequencies required more voltage, exposing an issue with Control Center. Setting 1.2V in the app produced 1.192V in CPU-Z, but bumping up to 1.3V resulted in a reported voltage of only 1.208V. Assigning a 1.3V CPU voltage in the firmware produced 1.304V in CPU-Z, so we dropped Control Center for the remainder of our testing.

The system was stable up to 4.7GHz on 1.2V, and it hit 4.8GHz on 1.3V. Adding another 100MHz required 1.35V and seemed to be throttle-free. However, 7-Zip wouldn’t run for more than a minute without producing an error. Upping the CPU voltage didn’t help, and neither did fiddling with the other system voltages and power settings. In the end, we had to drop to 4.8GHz to banish the 7-Zip errors.

A quick look at performance

If you’ve read our review of the Core i7-3770K, you know how its performance stacks up at stock speeds. You’ve also seen how much additional performance can be gained by turning the frequency up to 4.9GHz. At that speed, Ivy Bridge nearly matches the performance of the Core i7-3960X, which is based on Sandy Bridge-E silicon that offers two more memory channels and CPU cores.

So, what about the configurations we’ve discussed in this article? 7-Zip and the x264 encoding benchmark were run on all the boards at their highest stable settings. The Asus was tested with the CPU at 4.9GHz, while the others had it clocked at 4.8GHz.

100MHz amounts to around 2% when you’re running at speeds this high, so it’s no surprise to see all the scores so close. The Asus P8Z77-V generally comes out on top thanks to its clock speed advantage. However, it has the lowest frame rate in the second pass of the x264 test.

Overclocking on the sly

For quite some time, we’ve complained that Asus’ motherboard firmware engages in overclocking behind the user’s back. If a manual memory multiplier is set, the CPU’s single-core Turbo multiplier is applied to all-core loads. The Core i7-3770K runs at 3.9GHz when all its cores are occupied instead of the default 3.7GHz. This constitutes overclocking, according to Intel, and it shouldn’t be done without the user’s consent. Even worse, it violates good practices for enthusiast firmware: modifying one setting should never change another, and especially not one that’s completely unrelated.

Asus doesn’t ask permission when applying this “multicore enhancement.” The firmware’s all-core Turbo frequency display is changed to reflect the overclocked speed, but the user isn’t given an explicit message about what’s going on. At least this feature can be disabled in the firmware; we just wish it weren’t enabled by default.

When defending this behavior, Asus has insisted that other motherboard makers engage in similar dirty tricks. We didn’t see any evidence of that when testing Z77 boards with a Sandy Bridge CPU. However, we did catch one more offender when we switched to Ivy Bridge. Gigabyte’s Z77X-UD3H plays the same game with Turbo multipliers if the memory speed is set manually. Our Core i7-3770K runs at 3.9GHz with an all-core load when the memory is set to run at 1600MHz. When the memory divider is left at “auto,” the CPU speed tops out at 3.7GHz when all cores are active.

Does the Gigabyte firmware ask permission? Nope. Indeed, nothing in the firmware even informs the user that the CPU has been overclocked. Although the status window displays a 39X multiplier for all-core loads, it does so regardless of the memory configuration—including when the board is using the correct 37X multiplier.

We haven’t had time to grill Gigabyte about this behavior, which can only be corrected by setting the CPU’s per-core multipliers manually. Ugh. It’s hard to view this trend as anything other than an underhanded attempt to inflate benchmark scores. There’s more evidence that Asus and Gigabyte are pushing boundaries, too. According to CPU-Z, the Z77X-UD3H’s default base clock speed is 100.88MHz, while the P8Z77-V is clocked at 100.52MHz. I’m not going to get too worked up over sub-MHz increases to clock speed, but it’s worth noting that MSI nails the 100MHz default exactly. The Intel board runs a smidgen slower, at 99.78MHz.

Conclusions

Intel’s 22-nm process allows Ivy Bridge to consume much less power than its predecessor when running at the same speed. The 3D transistors purportedly offer better performance at low voltages, which is great for mobile applications but perhaps not ideal for overclocked desktop rigs. That said, I’m reasonably impressed with how well Ivy overclocks. The Core i7-3770K was stable at 4.4-4.5GHz without so much as an extra millivolt applied to the CPU. Those speeds only increased system power consumption by a modest amount, so there’s no need to invest in an aggressive cooling solution. With our single-fan Frio config, CPU temperatures were in the 50-60°C range under load.

Moving beyond 4.5GHz requires additional voltage, and that’s when Ivy really starts to heat up. With the CPU voltage increased to a seemingly conservative 1.35V, system power consumption soared, and thermal throttling reared its ugly head. We’d recommend liquid cooling to keep high temperatures (and the associated throttling) at bay when approaching 5GHz. Even at 4.8-4.9GHz, with both of the Frio’s fans installed, our CPU eclipsed 90°C with regularity. Although we’ve only overclocked one Ivy Bridge CPU, our findings jibe with what we’ve been hearing from the motherboard makers.

Our Core i7-3770K hit similar limits on all four boards. The Asus was the only one stable at 4.9GHz, but the others were only 100MHz behind. At stock voltage, the Intel and MSI boards both managed 4.5GHz, while the other two topped out at 4.4GHz. You should be able to get a good overclock out of any of the boards we’ve tested. The experience is the only thing that will differ substantially.

Of all the boards, the Asus P8Z77-V offers the most enjoyable overclocking experience The firmware is easy to use, and the accompanying Windows software is the best in the business by far. If only Asus didn’t insist on an unseemly “multicore enhancement” that overclocks the CPU without the user’s permission.

Gigabyte plays a similar trick with the Z77X-UD3H, and its implementation is even sneakier. The auto-overclocking software is too aggressive, as well, and the EasyTune Windows utility is overdue for an update. That said, Gigabyte’s new firmware makes manual overclocking easy. I’d love to see the interface ported to Windows, where it would be a step up from EasyTune and the accompanying 3D Power app.

Then again, MSI’s Windows-based ClickBIOS implementation is painfully slow on the Z77A-GD65. The Control Center app is snappier, but the voltage adjustments didn’t work properly for us. Fortunately, the firmware interface is solid. The ability to define a custom profile for the OC Genie button is a particularly nice touch.

OC Genie is the easiest auto-overclocking scheme around, but there’s merit to Intel’s hybrid approach, which let us take the CPU up to 4.5GHz by dragging a single slider on the main firmware screen. The DZ77GA-70K can handle manual overclocking, too, and Intel’s firmware and Windows utility both offer a good user experience without any niggling issues. They’re a little short on options compared to the equivalent Asus offerings, though.

Since all the boards reached similar clock speeds, I’d be hesitant to decide between them based on the overclocking experience alone. Ideally, you’re only going to be overclocking the CPU once. And I’d recommend it. Ivy Bridge appears to have a decent amount of “free” clock speed headroom—and even more if your cooler is up to the challenge.

Comments closed
    • daneyuleb
    • 8 years ago

    Did I miss something? The entire introduction mentions all the various overclocks you can achieve (4.9, 5, 4.4) but never once mentions what the original stock speed is?

    • BIF
    • 8 years ago

    That cooler is like the tires on a monster truck.

    I wonder how much bigger they can make the fans before the government requires manufacturers to cage them for the sake of fingers, insects, and small pets.

    • jamsbong
    • 8 years ago

    Very good article.
    Judging from the results, there is NO point going for ivy bridge. I have a Sandy Bridge i5-2500k, can easily o/c to 4.6GHz with a small Volt bump. The power consumption is fine and my passive liquid cooling meant it is cool and actually quiet.
    Gaming wise, there is no point for me to upgrade.
    Video encoding improves a bit with ivy but then if you want an ultimate video encoding muncher, you might as well get a Sandy Bridge E and overclock it.

      • gmskking
      • 7 years ago

      You are correct sir. I’m loving my 2600K @ 5.0GHz.

    • tootercomputer
    • 8 years ago

    Looking it all over, I tend to agree with Scott’s comment in a different article on IB that it is only an incremental boost in performance from SB.

    • Auril4
    • 8 years ago

    These heatsinks are beginning to look like giant chillers. We really need new cpu/gpu technologies that run cooler. Heatsinks like that should tell us that we are not as advanced on CPU technology as we would like to think we are.

      • UberGerbil
      • 8 years ago

      These heatsinks are only necessary when running the CPUs way beyond their design limits — and then only when upping voltage. If you don’t increase voltage you can overclock quite a lot without increasing power draw very much, and a modest heatsink — even the stock Intel one — works fine. The fact that we can pack 64bit quad cores (and fairly performant graphics) into laptop chassis that are no larger (and often smaller) than those that used to house a single-core 486 should tell us that we are more advanced on CPU technology than you think we are.

      • yogibbear
      • 8 years ago

      All-in-one antec kuhler / corsair H70 says hi!

    • ish718
    • 8 years ago

    I guess we have to wait for Haswell…

    • Jigar
    • 8 years ago

    I think i’ll just wait for the price drop of 2600K or 2700K and grab any one of them.

      • NeelyCam
      • 8 years ago

      Not a bad idea.

    • RtFusion
    • 8 years ago

    Was a little too lazy to look but what is the default voltage(s) for the 3770K?

      • Firestarter
      • 8 years ago

      Stock voltages, under load:
      i7-3770K: [url<]http://www.computerbase.de/bildstrecke/40123/3/[/url<] i5-3570K: [url<]http://www.computerbase.de/bildstrecke/40123/5/[/url<]

      • mboza
      • 8 years ago

      1.128V (CPU-Z) in the article, 1.368V when overclocked, 21% voltage increase, so expect a 47% power increase from the voltage.

    • Bensam123
    • 8 years ago

    Weird… I ended up reading this before the IB article as it showed up first…

    Seems as though a lot of this testing was limited by the cooling capabilities of the cooler you picked out. Perhaps rehashes with a better cooler are in order… or simply benchmarking different coolers based on this.

    • Duck
    • 8 years ago

    Do you think that these 3D transistors are responsible for this runaway like power consumption when voltage is increased?

    It seems to make sense to me, intuitively. Like they let tonnes more current flow without resistance when small voltage increases are applied.

      • vargis14
      • 8 years ago

      Good question Duck.But i would think it would think less resistance would lower temps not increase them.From what i know increased resistance increases heat since it has a sorta barrier to get though in a sense to get from a to b.Lower resistance i would think would lower temps since less of a barrier to get trough.
      Am i making any sense:)
      I could be totally wrong on top of making any sense.

        • Duck
        • 8 years ago

        Less resistance increases heat. Try shoving plain copper wire into a power outlet. Sparks will be had.

        • NeelyCam
        • 8 years ago

        P=V^2/R. Less resistance -> more power.

        But that’s not what is happening here. Power consumption of CMOS logic circuits is proportional to CfV^2. You increase the voltage by 40%, power consumption doubles. This is not specific to trigate – it just comes from charging/discharging load capacitances

          • Bensam123
          • 8 years ago

          Oh god I love when you go all nerdy.

      • OneArmedScissor
      • 8 years ago

      Even dinky Westmere dual-cores do this:

      [url<]https://techreport.com/r.x/cpu-roundup-2010q1/oc-cine-power-peak.gif[/url<] Running the voltage as high as it will go on any CPU is the equivalent of opening the flood gates, tri-gates or not. :p

        • NeelyCam
        • 8 years ago

        A voice of reason. I find myself agreeing with you 95% of the time these days… what happened?

          • OneArmedScissor
          • 8 years ago

          Everybody started spamming boring things that don’t compel you to buy them…or argue over them.

    • swaaye
    • 8 years ago

    Geoff did you try using S3 sleep at all while overclocked? SB chips have troubles sleeping when overclocked while PLL overvoltage is enabled (or auto enabled). This usually restricts people to 4.3-4.4 GHz if they want to use sleep mode. I’m curious if Intel fixed this quirk that they acknowledged with SB.

      • Airmantharp
      • 8 years ago

      I’d like to know this as well- given how fast my system starts up with an Intel 320 SSD, and that I do most of my general browsing on my laptop, I just shut the system down when it’s not in use gaming or processing photos.

      • IlluminatiASUS
      • 8 years ago

      ASUS – Reply…. we found no value internally in enabling PLL overvoltage to extend overclocking headroom beyond the approx 4.8GHz real world value. We do offer this option within our UEFI but would advise you leave it disabled or AUTO.

      With that noted we have tested our boards with PLL overvoltage enabled at 4.6GHz and found sleep S1,S3 and S4 to work correctly. Keep in mind though variance in IMCs can affect this so there may be still be configurations that have issues with PLL over voltage. As this is an enthusiast level option you must be aware of what can occur if this is used.

      Hope this helps,

    • Firestarter
    • 8 years ago

    [quote<]With the CPU voltage increased to a seemingly conservative 1.35V, system power consumption soared, and thermal throttling reared its ugly head.[/quote<] I don't know Geoff, I'd say that 20 to 25% extra voltage isn't quite as conservative as you think. An extra 0.2V may not have been a lot for a Pentium III era 2.0V CPUs, but for processors with a 140x more transistors, each of which are positively tiny compared to, well, anything, I'd say that 0.2V extra voltage amounts to a pretty aggressive overclock.

      • NeelyCam
      • 8 years ago

      This. That’s a pretty hefty voltage increase proportionally. This is not like the CPUs of the past that run at 1.8V in stock.

      It seems to me that the first IBs out have limited overclocking potential because of some sort of a speed path that can’t be solved without large voltage increases that then hit the thermal limits. Maybe subsequent steppings clear up some of those bottlenecks and allow higher clocks without voltage increases.

        • Firestarter
        • 8 years ago

        Pretty much any CPU/GPU has that speed path problem though. A large part of what AMD and Intel do is finding the perfect compromise between fast, power hungry circuits and slower but more efficient circuits, all while trying to find the slowest path and making it just fast enough. If they’d go balls out with the power hungry circuits they could probably create a CPU that clocks a bit higher than what we have now, but it’d leak like crazy and be very hard to cool. It would also be pretty useless for mobile computing.

    • Anonim1979
    • 8 years ago

    Those Ivy Bridge chips need to have heat spreader removed.
    Intel probably limits their overclocking potential on purpose by using crappy paste inside.

      • Peldor
      • 8 years ago

      It would be interesting if Intel sold the K series without head spreaders perhaps. Do the extreme overclockers generally remove the heatspreader?

        • Duck
        • 8 years ago

        No you would wreck it. It doesn’t have thermal paste, it’s more like thermal cement.

          • Airmantharp
          • 8 years ago

          People have been yanking heat-spreaders off of CPUs for years; it was especially prevalent with Pentium 4’s.

          Things like freezing the CPU have been known to work without too much trouble or damage to the die, but obviously some more research is needed here. Also, whatever solution used to cool the die needs to have better performance than the heat-spreader, and the new lowered top of the CPU will need to be taken into account when mounting the cooling solution.

          • Goty
          • 8 years ago

          AKA solder.

    • tay
    • 8 years ago

    When you overclock, does the Turbo also get pushed up the same number of speed steps?

      • Firestarter
      • 8 years ago

      I guess Ivy Bridge is a lot like Sandy Bridge in this regard: When you overclock, you essentially set the turbo speed for 4 cores, while ignoring the thermal limits for the CPU.

        • tay
        • 8 years ago

        I see thanks for the response. So does the speed stepping still work? When you overclock to 4.4 Ghz, is your base frequency now 4.1 Ghz?

        So you went from 3.5 –> 3.9 to 4.1–>4.4? Thanks.

          • Duck
          • 8 years ago

          No, I read it as 4.4GHz would be both the base clock and the turbo clock.

            • Firestarter
            • 8 years ago

            Well, you’ll never see the base clock, 4.4ghz would just be the clock your processor runs at whenever it decides that it is not idle.

            • tay
            • 8 years ago

            Ahh that sucks. Does that mean all processor speedstep states are gone? Will it idle at 4.4 Ghz as well? If so then I will avoid the K series.

            Thanks again!

            • Airmantharp
            • 8 years ago

            My SnB 2500k is set to run at 4.8GHz; when it’s ‘idle’, it down-clocks to 1.6GHz. I’m not sure if that answers your question, but it works incredibly well.

            • Duck
            • 8 years ago

            What about idle voltage?

            • Airmantharp
            • 8 years ago

            I’d have to get back to you on that- it’s been going strong for over a year now.

            • tay
            • 8 years ago

            Thanks for the original reply. That is great to hear about the down-clocking. As Duck mentioned idle voltage is even more important. Thanks!

      • IlluminatiASUS
      • 8 years ago

      Adjustments all depend on what is defined with the UEFI of the board as well as whether you chose a offset or manual voltage.

      If you use a manual voltage even if you maintain all cstates and speed step you will not have a lower idle vid it will be the same at idle as it is at load.

      If you are smart and use an offset voltage it will drop accordingly depending on the load. Additionally loadline calibration affect both of these although it can be trickier with the offset as the voltage already varied based on the multiplier.

      Long story short you can have both the voltage and multi drop if you tune the board correctly but if you go for brute force with a fixed frequency this can also be defined.

        • tay
        • 8 years ago

        Best answer thanks! Been a while since I overclocked (Phenom II unlocked).

    • chuckula
    • 8 years ago

    Not too shabby. This pretty much confirms the rumors that the early Ivy Bridge parts are not the greatest overclockers ever, but reaching 4.7 Ghz at 1.3 volts is nothing to sneeze at either.

    How are the temperatures on the chips during normal operation & overclocking?

      • ybf
      • 8 years ago

      other sites have already posted numbers using LN2, and they got well over 6 gig out of the 3770K. ymmv, proportional to your cooling budget. kegerator, anyone?

      • Wirko
      • 8 years ago

      The CPUs were more overclocking friendly before Intel invented that “K” suffix. My E6400 went from 2133 to 3200 MHz effortlessly – while it was new, anyway. That was 50% or a full GHz over stock speed, using stock cooler and raising the voltage to 1.26 volts. Yet it was far from exceptional at the time.
      In the other direction I could lower the voltage to 0.89 V at stock speed. That was before the “T” suffix came about. The guys at Anandtech have tried the same with the 3770K and had to stop at 0.9 V. With so much finer process tech I was expecting that it could get much closer to zero volts but no, it doesn’t.

        • OneArmedScissor
        • 8 years ago

        The minimum voltage is limited by design, not manufacturing process, but there’s a solution for that:

        [url<]http://www.xbitlabs.com/news/cpu/display/20110915185412_Intel_Unveils_Details_About_Research_Near_Threshold_Voltage_Processor.html[/url<] I believe the major culprit is low level cache. If you go too low, there goes your data. That's why everything has stuck to 0.7 to 0.8v at minimum for many years now, regardless of what kind of CPU, GPU, or manufacturing process. The important part is that, as you pointed out, you can now get several GHz out of such a low voltage, instead of just the 800 MHz which it was standard for CPUs to drop down to at idle for many years. Today, you just shut off the parts that aren't needed. Dropping the voltage much lower wouldn't necessarily get us too far now that even sub-watt phone chips can power gate a gajillion specialized parts. Put those two advances together, and that's why this year we'll see 2 GHz quad-core phones with practically everything integrated, including the modem. The low power side of things isn't stagnating, but blowing past the rate of progress we expected of PCs. Also, your Core 2 "overclocked" so well because that was still "back in the day" when they used the same silicon for every CPU, and differentiated them largely just with clock speed. That differentiation was just as arbitrary as what's done with features today, like disabling hyper-threading and turbo boost on perfectly good parts.

          • NeelyCam
          • 8 years ago

          [quote<]Put those two advances together, and that's why this year we'll see 2 GHz quad-core phones with practically everything integrated, including the modem.[/quote<] Dropping the voltage to low levels comes with slower clocks. That high-frequency operation will still require high voltage/power (which will come down later with further process shrinks). But these NTV-techniques could certainly help with load-dependent power scaling for some "connected stand-by modes"

      • Arclight
      • 8 years ago

      [quote<]How are the temperatures on the chips during normal operation & overclocking?[/quote<] Apparently this was a temp review for both stock and OCed frequencies. Go figure.... Edit: q.e.d.

    • madlemming
    • 8 years ago

    Do you use any custom settings for the 7zip torture test, or just it’s normal benchmark set for a lot of iterations?

    I’ve always used prime95 for stress testing, but if 7zip can find errors faster, nice.

      • Dissonance
      • 8 years ago

      Nothing special–just the built-in benchmark. The less-than-stable configs would error out in the first few iterations.

Pin It on Pinterest

Share This