Why is Ivy Bridge so hot and bothered?

If you read our Ivy Bridge coverage carefully, you’ll know that we observed some rather high temperatures when overclocking the Core i7-3770K. With a single-fan air tower, our chip ran at a reasonable 50-60°C when clocked to 4.4-4.5GHz at its default voltage. However, when we pushed to 4.9GHz on 1.35V, the temperature soared past 100°C. Other reviews have observed similarly scorching temperatures, so it’s not just our sample.

As we noted earlier this week, the Core i7-3770K’s power consumption increases dramatically when it’s pushed to 4.9GHz—as it should, given the proportional relationship between CPU power consumption, frequency, and the square of the voltage. The additional power is dissipated as heat, which contributes to those high temperatures.

The question is: why does Ivy appear to reach higher temps when overclocked than Sandy Bridge? Is it just the additional power draw being turned into more heat? Is it a thermal density issue, since Ivy’s heat output is concentrated into a much smaller die area than Sandy’s?  Or is it something else?

Overclockers.com points out there’s another factor to consider: the interface between the CPU die and heat spreader. The site claims Ivy Bridge employs thermal paste between the die and heat spreader, which differs from the fluxless solder that lurks under the lid of Sandy Bridge CPUs. It’s possible that a less efficient heat transfer interface could be leading to higher temperatures for Ivy.

Curious, we asked Intel about the interface between the Ivy Bridge die and the heat spreader. Intel has confirmed to TR that Ivy uses a "different package thermal technology" than Sandy Bridge. The firm stopped short of answering our questions about why the change was made and how the thermal transfer properties of the two materials compare. However, Intel claims the combination of the new interface material and Ivy’s higher thermal density is responsible for the higher temperatures users are observing with overclocked CPUs.

Intel also points out Ivy Bridge has a higher TjMAX specification, which governs when the CPU starts throttling in order to protect itself from heat damage. The cut-off for the Core i7-3770K is 105°C, while the 2600K starts throttling at 100°C.

Intrigued, we decided to perform some basic testing, comparing overclocked Ivy Bridge Core i7-3770K and Sandy Bridge Core i7-2600K chips on the same motherboard with the same cooler.

To keep things even, we aimed for similar levels of power consumption for the two CPUs, which should translate directly into similar levels of heat to be dissipated. Fortunately, we were able to reach rough power parity at equal clock speeds. At 4.9GHz, the Ivy-based system draws 236W at the wall socket, and the Sandy rig pulls 231W. The CPUs require similar voltages, as well. Ivy needs 1.368V to hit 4.9GHz, while Sandy takes 1.381V, according to CPU-Z. (For what it’s worth, CPU-Z reports the Core i7-3770K’s default voltage as 1.024V, while the 2600K registers 1.240V. We need to increase Ivy’s voltage by about a third to hit 4.9GHz, but Sandy demands only a 12% boost.)

Turns out the difference in temperatures between the two was quite a bit wider than the 5°C difference between the thermal throttling thresholds. Our Ivy CPU flirts with 100°C at those settings, while the Sandy Bridge chip is currently running at about 80°C in the same test system.

It appears that, all other things being equal, Ivy Bridge runs hotter. That almost certainly means the cooling is less efficient.  We can’t say with certainty whether the problem is primarily the difference in thermal interface materials or in power density—or perhaps equal parts each—but it does look like Ivy Bridge chips present a new sort of challenge for would-be overclockers. Those wishing to push toward the 5GHz mark might be better served by sticking with their older 32-nm processors, at least for the time being.

Comments closed
    • flip-mode
    • 7 years ago

    The first thing Intel should do is go back to the soldier. At least remove that from the equation. I have a feeling it has a larger impact than some might think. Go back to the soldier and see where that get things and then we can start talking about the CPU’s thermal density.

    I doubt Intel’s going to get right on that just because I say so, though.

    • sschaem
    • 7 years ago

    Just to quote the recent shortbread

    “Overclockers.com on Ivy Bridge temperatures – it’s gettin’ hot in here”

    That was an eye opener.

    I bet SB-E wont use the same conductor.

    • DMF
    • 7 years ago

    Hotter ??

    Do what I do. Clean the cat hair out of your intake vents!

    [No! I’m out of cheezburgers! Now go away!]

    • ronch
    • 7 years ago

    How about we ask some of the other guys about this too? AMD, Nvidia, GF, IBM and TSMC?

    • ronch
    • 7 years ago

    If Intel’s fancy new 22nm process can’t reach high clocks at acceptable power levels, it worries me how much better the other guys with smaller budgets can do. And how would they improve performance further without including a water block with each chip sold.

    • ColeLT1
    • 7 years ago

    Just found and ordered a 2570k:
    [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16819116504[/url<]

    • FranzVonPapen
    • 7 years ago

    It’s probably those damn Tri-gate transistors! Or sunspots! Or aliens!

    [b<]No, it's the TIM[/b<] (solder vs. heatsink grease): [url<]http://www.overclockers.com/ivy-bridge-temperatures[/url<] Look for horses before zebras. The linked article points out this exact scenario has played out before, in the E6XXX and E4XXX lines.

      • Krogoth
      • 7 years ago

      It is a little bit of everything.

      I suspect that the largest contributor is still the tri-gate transistor and 22nm process. It is just power hungry when you crank-up the volts.

      Notice how Ivy Bridge’s power consumption skyrockets when you throw more volts? Combine that with a ~25% smaller surface area. It is really no surprise that the silicon roasts.

      Inferior TIM material just compounds the problem.

        • NeelyCam
        • 7 years ago

        It’s CfV^2. Every process is powerhungry when you crank up the voltage. How many times do I have to repeat this?!?

          • alphacheez
          • 7 years ago

          IVB get crazier (higher power consumption) at lower voltages than we’re used to probably because the unique structure of the transistors leads to greater leakage when pushed the nominal voltage. On the other side the 3D transistors also seem to do a god job of running at lower voltage and that could bode quite well for ULV chips. I’m hoping and expecting to see some awesome IVB notebook chips in the 35 W and lower range with performance much better than we’ve seen at such power consumption levels previously.

            • NeelyCam
            • 7 years ago

            [quote<]IVB get crazier (higher power consumption) at lower voltages than we're used to probably because the unique structure of the transistors leads to greater leakage when pushed the nominal voltage. [/quote<] New process tends to bring lower nominal voltages. This voltage scaling trend has been going on for a long time - there is nothing unusual about that. You can't look at 22nm at 1.3V and 90nm at 1.3V like it's an apples/apples comparison. New process - lower nominal voltage. 22nm trigate transistors aren't naturally "leakier" (in fact, they are less leaky because of the nature of tri-gate) - if it appears so it's because people only look at power consumption at 1.2V or 1.35V, forgetting that the nominal voltage had dropped. When voltage is increased beyond the nominal, power consumption scales with fV^2, and IB chips follow that trend.

            • Bensam123
            • 7 years ago

            Do you think tri-gate transistors are in their natural habitat at 5ghz?

            • NeelyCam
            • 7 years ago

            [i<]There is no natural habitat![/i<] There is no magical 4GHz or 5GHz barriers. Look - please take some semiconductor physics classes and logic circuit design classes before you continue arguing with me. I'm tired of repeating this sh*t over and over to you because you just don't know enough about the topic to understand, and you don't accept the fact that I do. I've tried to be nice, I've tried to simplify it, I've tried to explain the details... I'm done. If this was you trolling, then good job - I rarely lose my patience, but you made it happen.

            • ronch
            • 7 years ago

            Someone just overvolted Neely past his nominal voltage and now he’s blown a fuse.

            By the way, you’ve been replying to Krogoth, alphacheez and Bensam. Who ticked you off, exactly?

            • NeelyCam
            • 7 years ago

            Bensam

            • Bensam123
            • 7 years ago

            Someone gets frustrated because they don’t understand something, anything pisses them off.

            • NeelyCam
            • 7 years ago

            I +1’d you because this statement tends to be true in general. But that’s not the reason for my frustration here

            • Bensam123
            • 7 years ago

            You can hate on me all you want, but I don’t believe that things either work or they don’t. It’s entirely possible for performance to degrade when something doesn’t operate within its specified parameters, such as a battery for instance.

            Don’t get all high and mighty on me because you read a paper on 3D transistors either. I did read through the other news posts when you first started posting on 3D transistor tech.

            Unless you’re employed by Intel and make these transistors or helped design them, in which case you’d know exactly what’s happening here, you’re just as much on the outside as I am. Chances are if you actually knew what was going on you wouldn’t be here stumped and befuddled, just like the rest of us either.

            BTW I never said anything about 4ghz or 5ghz being a mystical barrier.

      • flip-mode
      • 7 years ago

      Is it easier to remove the cap with TIM rather than soldier? Seems like the optimal solution would be to remove the cap and get direct contact with the silicon.

        • Krogoth
        • 7 years ago

        I believe they both use epoxies.

        IMO, I wouldn’t risk removing the headspreader unless you are extra careful with your aftermarket HSF and you don’t plan on taking the system to LAN parties.

        I have seen too many horror stories from exposed cores in the Socket A and Socket 370 days.

    • WillBach
    • 7 years ago

    Hi Geoff,

    Since we’ve agreed that [url=https://techreport.com/articles.x/22836<]Ivy Bridge is female[/url<], why can't you use [url=http://gatherer.wizards.com/Handlers/Image.ashx?multiverseid=1290&type=card<]the Fire Elemental from MtG[/url<] that Melissa A Benson drew for the "hot and bothered" descriptor? Edit: used Wizards link.

    • Bensam123
    • 7 years ago

    Sort of interesting thought Scott/Geoff, some people have been mentioning how well the BD overclocks vs SB/IB, I wonder if a high performance analysis is in order? Although you guys originally overclocked the BD, that didn’t seem to get very far. Clock vs clock performance at high OC levels… It could make for an interesting comparison and outcome.

    People love when the underdog can give the competition a run for it’s money. 😀

    • MrDigi
    • 7 years ago

    bit-tech seems to have gotten better results from their overclocking.
    [url<]http://www.bit-tech.net/hardware/2012/04/23/intel-core-i7-3770k-review/8[/url<]

      • NeelyCam
      • 7 years ago

      Strange numbers for 2500k/2600k. Why would those systems run at 20W higher idle than IB?

        • Goty
        • 7 years ago

        A particularly power hungry mobo, perhaps? Doesn’t seem very likely, though, considering they’re both from the same manufacturer and series, meaning I couldn’t imagine there’d be that large a difference between them. Could just be Bit-Tech’s usual attention to detail (read: their numbers don’t often match up between reviews).

    • yogibbear
    • 7 years ago

    An engineer somewhere is shaking his fist at management/cost cutting somewhere else that removed the fluxless solder and then forgot to test overclockability in their proof of concept and then probably earnt themselves a healthy bonus for all the money they saved intel…. now the bad press is probably money they are losing as enthusiasts are less likely to mass adopt as they were expecting.

    • indeego
    • 7 years ago

    [quote<]Intel also points out Ivy Bridge has a higher T[sub<]jMAX[/sub<] specification,[/quote<] whatthe?

      • Firestarter
      • 7 years ago

      maximum temperature at the juncture? AFAIK it’s something like that

      • derFunkenstein
      • 7 years ago

      It’s the point at which the CPU throttles itself to save its life.

        • Corrado
        • 7 years ago

        I thought it was the point at which the CPU went to an overstock/liquidation store to buy women’s clothing and accessories.

        [url<]http://www.tjmaxx.com/[/url<]

    • clone
    • 7 years ago

    couldn’t Intel get a larger flatter footprint by using fewer layers, wouldn’t this alleviate the spot heating issues that seem to plague this CPU when overclocked?

    given the results I wouldn’t touch Ivy Bridge and am more motivated to buy an i7 series before they get discontinued…..it’s unfortunate that AMD has no way to take advantage of this opportunity.

      • flip-mode
      • 7 years ago

      Larger footprint? LOL. You could stick with 32 nm and get your larger footprint for sure.

        • clone
        • 7 years ago

        [b<]Larger footprint? LOL. You could stick with 32 nm and get your larger footprint for sure.[/b<] I'm so glad you made this comment, provides insight.

          • NeelyCam
          • 7 years ago

          What he meant is that yes – you could increase the footprint of the silicon and alleviate the heating issues.

          But the cost of silicon is largely footprint based – not volume based. So, doubling the footprint would improve the cooling quite a bit, but would also double the cost of silicon, and it would not be a good option for Intel (or consumers).

            • clone
            • 7 years ago

            I doubt that’s what he meant but regardless.

            if a processor was etched in 1 step using a disc of silicon going in one side and coming out the other with CPU’s fully finished on it I’d agree but from what I’ve read processors have insulating layers between each layer which implies a multi stage form of production that involves layering material.

            if their is layering then the material cost would remain the same if not be reduced.

            • NeelyCam
            • 7 years ago

            It’s still a bit unclear to me – are you proposing increasing the silicon area? Or keeping it the same?

            Most of the heat generation happens in the transistors that, in Intel’s case, are sitting in a silicon substrate. With flip-chip packaging, this substrate is touching the thermally conductive material (whichever it may be) that goes between the silicon chip and the heat spreader.

            Those layers you’re talking about are between the transistors and the socket. Removing some of them would save a bit of money (if one doesn’t change the area of the chip) but it wouldn’t affect cooling much at all in bulk silicon processes.

            • clone
            • 7 years ago

            I don’t know how many layers Ivy has, took a look didn’t find but for the sake of discussion say Ivy has 10 layers of transistors stacked on top of one another so that it’s dimensions are for the sake of discussion 10 X 10 X 5 mm

            I’m proposing integrating some of those layers… 2 if possible into the other 8 to increase the size of the thermal transfer area in this case going to 12.5 X 10 X 4mm increasing the surface by 25%… or going even further and integrating 4 layers into the other 6 making it 12.91 X 12.91 X 3mm increasing the surface to 166.66mm or 66.7% while also bringing the heat sources much closer to the surface for removal from the cpu.

            the cpu package wouldn’t have to be larger but the effective heat dissipation area would be more efficient.

            • NeelyCam
            • 7 years ago

            Oh, so you [i<]are[/i<] proposing the silicon area of the chip is increased. Well, the chip would get more expensive because you increased the area (as I already said, cost is area-based, not volume based).

            • chuckula
            • 7 years ago

            Ivy has exactly 1 layer of transistors. Ivy Bridge does include multiple material layers including metal layers that link transistors, but the transistors are in a planar array. The vast majority of ICs that are made today are also planar, including every CPU and GPU that you can buy on the market, no matter who makes it. Ivy Bridge’s heat problems are not due to the depth of the chip, but due to the fact that a huge number of transistors are packed very closely together and it is difficult to extract the heat from such a small area efficiently. Intel’s long-term solution is likely to just reduce the amount of heat through process improvements, but another solution would be to improve the cooling to remove the heat from the die more efficiently.

            There is no “stacking” of transistors in a planar chip. Some of the foundries are talking about using through-silicon vias to take multiple silicon dies and link them together in “3D” chips, but even those are simply multiple planar arrays of transistors that are cut from a wafer and put back together into a stack.

            • clone
            • 7 years ago

            interesting.

            • chuckula
            • 7 years ago

            P.S. –> When you hear “3D” transistors bandied about related to Ivy Bridge, that just means that the gates of each the transistors stick up vertically as “fins” and the gate wraps around the fins. The transistors themselves are still in a 2D array though. Think of it as a bunch of sharks swimming side-by-side with their fins (and lasers) sticking out of the water. The sharks are still in a 2D arrangement even though each shark has a “3D” fin.

            • NeelyCam
            • 7 years ago

            Clone was being sarcastic

            • TurtlePerson2
            • 7 years ago

            The way chips are manufactured you can only have one layer of transistors. The other layers are mostly metal and insulators (Si02).

            • flip-mode
            • 7 years ago

            That was pretty much exactly what I meant. In other words, yours was the first post I’ve seen – ever – that sort of intentionally sought a larger die size.

            • clone
            • 7 years ago

            doubtful but whatever.

            that said kind of still do if the end result is a much needed improvement in cooling efficiency.

            • NeelyCam
            • 7 years ago

            You doubt that flip-mode knows what he meant..? That’s just silly – almost as silly as suggesting increasing the chip size for nothing but improved cooling..

      • TurtlePerson2
      • 7 years ago

      By “fewer layers,” I assume you mean fewer metal layers. This isn’t really practical as it would increase routing complexity and delay. The added capacitance of closer wires would increase the power dissipation, which would cause an opposite effect of what you want. It’s possible that Intel used way too many metal layers on Ivy Bridge, but I’m going to assume that they’re smart enough to not do that.

      Also, given that it uses a flip-chip package, I wouldn’t be surprised to find out the majority of heat conduction is through the substrate. If this is true, then the number of layers on top of the chip doesn’t really have a large effect on thermal conduction.

    • Deanjo
    • 7 years ago

    Ivy is just going through menopause.

    • Arclight
    • 7 years ago

    [quote<]Those wishing to push toward the 5GHz mark might be better served by sticking with their older 32-nm processors, at least for the time being.[/quote<] Indeed.

    • ish718
    • 7 years ago

    [quote<]" We need to increase Ivy's voltage by about a third to hit 4.9GHz, but Sandy demands only a 12% boost." " Ivy needs 1.368V to hit 4.9GHz, while Sandy takes 1.381V, according to CPU-Z"[/quote<] I was assuming the new smaller chip would be able to hit similar clock frequencies at significantly lower voltages...

      • Flying Fox
      • 7 years ago

      Not a given, that’s why we have tests.

        • NeelyCam
        • 7 years ago

        I agree with ish. I think it should be a given, considering the 30% performance improvement Intel was advertising.

        Something is going on with this chip.

          • flip-mode
          • 7 years ago

          What 30% improvement did Intel advertise?

            • NeelyCam
            • 7 years ago

            +30% in transistor performance for the same power – this was when they announced trigate

            EDIT: Sorry – it was 37% at low voltages, but 18% at 1V:

            [url<]http://www.anandtech.com/show/4313/intel-announces-first-22nm-3d-trigate-transistors-shipping-in-2h-2011[/url<] So, it's, what, 2% at 1.35V then...?

            • flip-mode
            • 7 years ago

            I don’t think you’re following Intel’s intention in making that claim. Intel probably never intended for 22 nm @ 1.35 V. The question is, I think, what is the ideal voltage for the 22 nm process and at that voltage does it run 30% faster. Well, since we know IB will run at 4.5 GHz at it’s stock voltage of 1.024 you could set SB to 1.024 and see what clocks you could get out of it at that voltage. I think that’s an objective way to look at it rather than picking your own voltage that’s not only higher than IB’s stock, but also even higher than SB’s stock.

            • Rurouni
            • 7 years ago

            So basically Intel 22nm process can’t scale the clock speed as good as their 32nm process?

            • flip-mode
            • 7 years ago

            Hmm… not quite. They seem to scale the clocks just fine, and they do so at lower voltage than 32 nm Sandy, but the problem is that they get difficult to keep cool. TR’s article here is trying to find out why, and the answer seems to be two-fold: first of all the power density of IB is higher than SB – even though it uses less power. This is because it is also much smaller, and since power (heat) dissipation is partially dependent upon surface area and Ivy Bridge has less surface area to work with, the chip is more difficult to cool. The second part of the two-fold is the thermal interface material used in Ivy Bridge between the actual piced of silicon and the metal cap that covers it. Intel switched to some different kind of material. Sandy Bridge apparently used some kind of soldier and that would transfer heat extremely effectively. Ivy Bridge uses something different – what, exactly, Intel has not said – but it seems to be less effective at transferring heat than the soldier used in Sandy Bridge.

            • bwcbiz
            • 7 years ago

            1. They probably cut the quality of the thermal interface material because the chip generates less heat at it’s specified frequencies/voltages (TDP 77 vs. 95 w), so they could save a few cents per CPU without breaking the spec.
            2. Tick/tock. The real improvements in 22 nm process node come in the next generation when the chip is redesigned to take advantage of the process and work around the issues like this.
            3. Maybe Intel realized they made SB too overclockable for the good of the Sandy Bridge-E market?

            • flip-mode
            • 7 years ago

            Sure, those all seem plausible, even though I’m not so sure about #3 since SB-E came well after SB and Intel had plenty of time to target the thing exactly how they wanted. The problem with the “extreme” platform is that the mainstream platform has progressed to the point that it even satisfies most extreme users. I’m not convinced that a faster SB-E would pull too many people up market or if what Intel really needs to do is make the platform more affordable. There are scarce few desktop users that could legitimately make use of 64 GB of RAM who aren’t actually better served by the Xeon platform, and to some extent the need for 8 cores follows a similar logic. If you’re doing [i<]THAT[/i<] much rendering / transcoding / compiling / computation going on then you probably [i<]aught[/i<] to have it running on a server-grade platform rather than on your desktop. In other words, SB-E seems to be competing more with server platforms than desktop platforms. If I want 64 GB + and massive processing power, do I want a Supermicro based system or an Asus Republic of Gamers system? Personally, though, I'd love to have an SB-E rig with the RAM maxed out. If the motherboards were priced at the same level as Z77 mobos, I would definitely go that route without looking back. I just think Intel made the SB-E platform too expensive.

            • Rurouni
            • 7 years ago

            I’m not basing my thought because of the temperature only, but also on the fact that to scale to a similar clock, it needs to up the voltage a lot more percentage wise. So at certain point, it needs to use lots of power even though the transistor itself is smaller.

            • NeelyCam
            • 7 years ago

            ^This.

      • Dissonance
      • 7 years ago

      [i<]I was assuming the new smaller chip would be able to hit similar clock frequencies at significantly lower voltages[/i<] And you're not incorrect in that assumption. All you need to do is roll back one sentence: [i<]For what it's worth, CPU-Z reports the Core i7-3770K's default voltage as 1.024V, while the 2600K registers 1.240V[/i<] At its stock frequency, Ivy has a comparable clock speed to Sandy at a much lower voltage. It's only when overclocked past ~4.5GHz that Ivy needs [i<]any[/i<] additional voltage.

        • NeelyCam
        • 7 years ago

        [quote<]At its stock frequency, Ivy has a comparable clock speed to Sandy at a much lower voltage. It's only when overclocked past ~4.5GHz that Ivy needs any additional voltage.[/quote<] And this is the crux of the matter. Why does IB need so much additional voltage to overclock just a little bit? What is the bottleneck? I feel like there is something "wrong" with the chip design - something that could be fixed by followup steppings, unleashing an overclocking monster. The easy low-power overclocking up to some 4.5GHz points to a great process (while at the same time breaking through OAS' magic 4GHz barrier), but there are some speed/race paths somewhere that don't allow it to go faster without drastic measures.

          • Meadows
          • 7 years ago

          They could solder it, for starters.

            • NeelyCam
            • 7 years ago

            I’m not talking about the temperature increase – that just comes from increased voltage. I’m talking about not getting more frequency out of more modest voltage increase.

            SB can overclock more with a smaller voltage increase. Why?

            • Bensam123
            • 7 years ago

            New leaky transistor tech? I thought you explained and talked about this in other posts…

            • NeelyCam
            • 7 years ago

            I was talking about why the temperature increases with voltage, and later discussed with TurtlePerson2 about the reason why IB heats up easier than SB.

            But the reason why IB and SB have the same max clock frequency at the same voltage is still somewhat of mystery to me. With the new process and faster transistors I would expect the same voltage to yield higher clocks in essentially the same architecture…

            • Bensam123
            • 7 years ago

            …but you just asked why power usage increases, which would be leaky new transistors… Even if they operate better under spec, doesn’t mean they operate better when you push them beyond their engineered threshold.

            Temperature would be the thermal compound under the cap. It’s comorbid; there can be more then one answer here.

            They could be intertwined too… The new transistors aren’t as efficient at higher voltage so they leak more. The further you overclock it, the more heat it produces of a logarithmic nature (not linear).

            • NeelyCam
            • 7 years ago

            The word “leaky” is thrown around a bit much, I think. “Leaky” used to be the buzzword when the thin gate oxides were ‘leaking’ because of quantum tunneling effects (at the 90nm node or so).

            The “flip-side” of leakage was the transistor “channel” leakage (or ‘drain-source leakage’) that was caused by the lack of channel control by the gate, made worse by the so-called “short-channel effects” that were causing increased ‘leakage’ current through the channel of a transistor that was supposed to be off. Incidentally, this is related to the gate leakage in a way – thinning the gate dielectric makes the gate leak more but the channel leak less.

            Then HKMG came along. Dielectrics with a high dielectric constant allowed the gate to retain the control of the channel to keep the channel leakage low, even when the dielectric thickness was increased. This enables reduction of channel length while keeping it “off” sufficiently.

            Shrinking the channel length further means an increase in channel leakage again. Anything that exerts a stronger control over it could be used to keep the leakage in check – like SOI, ultra-thin-body structures, higher dielectric constants, or switching to a finfet structure (that exerts channel control from two/three directions instead of just one).

            Your comment about overclocking increasing heat combines way too many effects. Overclocking has a (largely) linear effect on power consumption [i<]unless[/i<] you have to increase voltage. Assuming temperature doesn't change when voltage is increased, IB transistors should be faster than SB transistors. Why aren't they? When you say "The new transistors aren't as efficient at higher voltage", what exactly do you mean? What's the underlying effect that makes them less efficient? I understand you try to explain to yourself what you see in logical terms based on what you know - that's what we all do. But at the moment your explanations are sort of all over the place...

            • Bensam123
            • 7 years ago

            I don’t understand why some of you are so stumped by this… Everything isn’t linear. Even if something acted in a certain way in a past, that doesn’t mean it will continue to do so well into the future. This is a new technology that changes how transistors work quite a bit. It’s not absurd to think there is a downside to new technology when operated outside of it’s specified envelope. Heck there are things that explode when you do that.

            My explanation isn’t all over the place, I see two different areas of contention, the thermal interface between the heat spreader and the CPU and the transistors. One deals specifically with heat and the other with power usage, both can coincide with eachother if the power usage goes up in a non-linear fashion, more wasted energy produces more heat.

            • NeelyCam
            • 7 years ago

            My god… when will you give up?

            V^2 is non-linear, and power goes up with fV^2. If you don’t believe me, read the TR overclocking article – Geoff also said that the power consumption follows the theoretical equation very well. Just because some Ian dude reviewing IB for Anandtech didn’t know the equation doesn’t mean it doesn’t hold.

            Why does the equation hold? Because that’s how CMOS circuits work. There are some second-order effects that aren’t considered in the equation, and it’s plausible that they could play a bigger role in some future process, but as I’ve stated before and as Geoff has stated before, [b<]IB power consumption follows the equation.[/b<] The root cause of high temperature [i<]could[/i<] be considered an area of contention (although the most likely reasons have already stated here a dozen times). But the power usage is [b<]NOT[/b<] an area of contention. [i<]So please don't bring it up anymore.[/i<]

            • just brew it!
            • 7 years ago

            As has already been pointed out, die surface area probably has a larger effect on temperature than the type of thermal compound used. The thermal interface is very thin; therefore the type of compound used has a very small effect (unless it sucks pretty massively).

            OTOH, thermal resistance is inversely proportional to surface area. So even with [i<]the same thermal compound[/i<] under the lid, if you reduce the die size by 25% while keeping the power roughly the same, your die temperature rise (relative to the temperature of the top surface of the lid) will increase by ~25%.

            • Bensam123
            • 7 years ago

            Yet, it’s something we can point out that’s different instead of saying this is voodoo. You personally don’t even know what kind of thermal compound it is, it just looks like cheap thermal paste, and if Intel wanted to they could even make it worse then cheap thermal paste for the sake of fudging overclocking.

        • yogibbear
        • 7 years ago

        Wait wait wait. Stop the press! The curve isn’t linear? ZOMG. Mind. BLOWN!

        • wierdo
        • 7 years ago

        Dave at RWT had an interesting take on this, tri-gate seems to change the dynamics of voltage tweaking for OC purposes:

        [url<]http://realworldtech.com/forums/index.cfm?action=detail&id=128738&threadid=128721&roomid=2[/url<] "...The first two facts mean that the benefits of cranking up the voltage aren't nearly as significant as they used to be. The latter means that there is less headroom than before..."

    • colinstu
    • 7 years ago

    Hopefully “IB-E” and Haswell / “Haswell-E” doesn’t try pulling this kind of stuff!

    • Krogoth
    • 7 years ago

    ITT: Butthurting overclockers with unrealistic expectations.

    Have overclockers considered the fact that 22nm process with Tri-gates is simply leaky when you feed it with more volts? The power consumption results from overvolting makes this painfully apparent. Considering that Ivy Bridge die has a smaller surface area to dissipate wattages comparable to Sandy-Bridge-E chips at similar clockspeeds. It is really no surprise that Ivy Bridge gets to be toasty when you try to push it hard.

    This isn’t the first time that a die-shrinkage didn’t give significant gains in clockspeed ceiling and drops in power consumption due to leakage issues. Prescott is perhaps the most famous example. It was barely faster clockspeed-wise than Northwood, but was stuck with an architecture that require high clockspeed to overcome its IPC inefficiencies.

    The bottom line is that overclockers were expecting the same clockspeed celling jump and power consumption drop that we saw with Nehelam => Sandy Bridge. However, this didn’t happened. It is likely because the 22nm process with Tri-Gates just is leaky. It probably would explain why Intel had to delay the Ivy Bridge launch by a fiscal quarter over “rumored” issues with the 22nm process.

      • NeelyCam
      • 7 years ago

      No. AMD fanbois who bought BD got butthurt. Intel fanbois don’t, because they use lube

        • FranzVonPapen
        • 7 years ago

        Guess those Intel fanbois are [i<]smart[/i<].

      • TurtlePerson2
      • 7 years ago

      FinFETs (tri-gates) have lower leakage than traditional MOSFETs. That lower leakage scales with voltage too.

    • CuttinHobo
    • 7 years ago

    “There are some who call me… TIM.”

      • cheerful hamster
      • 7 years ago

      Look at the bones!

    • cheerful hamster
    • 7 years ago

    Please forgive the slightly OT question, but I haven’t kept up with the Intel timelines. Are we slated for a socket 2011 refresh this year, presumably an Ivy Bridge-E chip? It seems like with the die shrink, they should be able to pack more cores on it.

      • kalelovil
      • 7 years ago

      No we aren’t.
      [url<]http://www.xbitlabs.com/picture/?src=/images/news/2012-03/intel_cpu_roadmap_ww08_bg.jpg[/url<]

        • cheerful hamster
        • 7 years ago

        Wow, if that chart is accurate, it doesn’t seem to bode well for Haswell if the SB-E chips are still going to be the Extreme and Premium Performance chips into mid-2013. Hopefully they have something up their sleeve along the lines of Gulftown, and release an 8-core SB-E for desktops. (I build DAWs, and digital audio benefits from as many cores as possible.)

          • NeelyCam
          • 7 years ago

          Haswell is fine – this is just Intel milking the fully amortized 32nm fabs to the max since there is no competition.

          • flip-mode
          • 7 years ago

          I don’t think it does you any favors to judge Haswell by whatever is happening on socket 2011. Haswell is going to be a mainstream product and has certain goals of it’s own (internal architecture, further SOC integration, improved IPC, lower power, etc) while socket 2011 is about a whole different set of criteria – massive memory bandwidth, core count, PCIe lane count, maximum memory capacity, and so on.

            • Airmantharp
            • 7 years ago

            We’d need to be looking out for an IvB-E or HsW-E at some point in the future, with possible pin and electrical compatibility with X79 LGA2011.

            Still, with IvB being so close in performance to SnB when overclocked, I don’t see much of a need to go beyond a decent 6-core SB-E CPU if you need something that X79 provides over Z68/Z77.

    • Bensam123
    • 7 years ago

    Awesome inside look you two. Stuff like this makes TR great.

    As others mentioned all that needs to be done now is to have the heatspreader chopped off in a dramatic fashion as you risk life and limb to achieve the absolute best possible in technological benchmarking. If they don’t use the same cement, wouldn’t it be easier to remove the heat spreader?

    The review aside, I would put money on Intel trying to limit the results of overclocking on these early samples with a different thermal compound. Chances are they actually overclock quite well, too well, so in order to give themselves headroom for future products they decided to add a soft cap that isn’t readily visible to end users. A tiny fraction of actual users will risk destroying their CPU by removing the heat spreader for a small boost in cooling.

    Maybe Intel wants that too? So there is risk involved with actually pushing the CPU a lot further as you can destroy your CPU merely by trying to overclock it all the way. That would end up netting them some money too as people pull off part of the die with the spreader.

    • maxxcool
    • 7 years ago

    Scott! Geoff! Cut that spreader off! get us some direct to die cooling tests! (if its possible)

    • Airmantharp
    • 7 years ago

    [s<]Alright, enough of this. People keep talking about how hot Ivy Bridge is in measured degrees Celsius compared to Sandy Bridge- and they're right. But here's the thing. Ivy Bridge, at 160mm2, is 26% smaller than Sandy Bridge, at 216mm2. Ivy Bridge, at 100c, runs 25% hotter than Sandy Bridge, at 80c, when run at the same clock-speeds, same voltages, with the same power draw, on the same system. See what I did there? 26% smaller die and 25% higher measured temperature, while using the same power? [u<][b<]Consider the law of thermodynamics upheld.[/u<][/b<][/s<] Edit: I'm wrong, temperatures need to be compared in Kelvin, thanks Voldenuit!

      • Voldenuit
      • 7 years ago

      [quote<]Ivy Bridge, at 100c, runs 25% hotter than Sandy Bridge, at 80c, when run at the same clock-speeds, same voltages, with the same power draw, on the same system.[/quote<] 100 C is not 25% higher than 80C, because Celsius is not an absolute scale. If you were to convert it to kelvin, 100 C = 373K is 5% higher than 80C = 353K. Even then, you can't do a straight comparison because the heat transfer is also a function of ambient temperature, and you are not dumping the heat into a 0K heatsink. Bottom line, you can't easily estimate heat output/transfer from die temperatures alone.

        • Airmantharp
        • 7 years ago

        Ah! you got me there, forgot about Kelvin (I don’t do this often, business major).

        Still, with the same power usage at the same CPU speeds and same voltage, it seems that Ivy doesn’t output any more heat than Sandy.

          • JustAnEngineer
          • 7 years ago

          You don’t really need to be using absolute temperature (degrees Kelvin or Rankine). You should be comparing the temperature of the processor to the temperature of the air being used to cool it.
          dQ/dt = U·A·∆T

          P.S.: The rate of heat transfer is equal to the overall heat transfer coefficient (set by the design, construction and cleanliness of the heat transfer surface) times the area that is transferring heat times the difference in temperature between the hot side and the cool side.

            • Voldenuit
            • 7 years ago

            Yeah, I realized that some time after I made my post. Good call.

            However, we already know that the heat transfer for the two systems is 231W and 236W (neglecting power supply efficiency and motherboard VRM heat output). Also, ΔT should be based on the exhaust air temp, not just the ambient air temp. All things being equal, it does seem that IB and SB put out the same (or similar) amount of heat energy, it’s just that IB gets hotter because of the smaller surface area and the (presumably) less efficient TIM between the die and the heatspreader.

            EDIT: On further thought, the ‘low’ temperature you pick will depend on which boundary condition you set on the analysis. If you want to derive the heat transfer coefficient of the cooler, you would use the ambient air temp and the area of the fins. If you want to derive the heat transfer coefficient of the heatspreader/TIM you would use the area of the heatspreader and either the exhaust air temp (making the assumption that the steady state temperature of the exhaust is equal to the sink temperature of the heatspreader) or ideally, the temperature of the heatsink pad itself (which is harder to measure). Doing this, we might be able to derive a close approximation of the heat transfer coefficient of IB’s TIM+heatspreader compared to SB+heatspreader. Who knows, we might even find that it performs ‘better’, contrary to our current expectations.

            • Airmantharp
            • 7 years ago

            Even if my understanding of physics and the underlying math isn’t so great, that is my point :).

            • JustAnEngineer
            • 7 years ago

            The normal convention is to use the log-mean temperature difference.
            [url<]http://en.wikipedia.org/wiki/Log_mean_temperature_difference[/url<]

            • Voldenuit
            • 7 years ago

            A couple points:

            * with LMTD, you have to use kelvin or rankine.
            * As the article says, LMTD is not valid for reboilers and condensers, so you can’t use it with heatpipe coolers (since the refrigerant in heat pipes goes through a phase change)

            • JustAnEngineer
            • 7 years ago

            You do not need to use absolute temperatures (degrees Kelvin or Rankine) for heat transfer calculations.

            ∆ is a [i<]difference[/i<]. You're subtracting one temperature from the other. It therefore doesn't matter where the zero of your scale is. The [i<]difference[/i<] is still the same number of units (∆ °F = ∆ °R = 1.8 ∆ °K = 1.8 ∆ °C). I can apply heat transfer equations to finite elements of the cooler or to the whole thing at once. As long as my frames of reference are good, I can still get valid results.

            • Voldenuit
            • 7 years ago

            While you don’t necessarily need to use absolute units for thermal analysis in general, you [i<]do[/i<] need to use absolute units for log mean temperature analysis, because it uses a ratio of temperatures in the denominator. You are correct that you can set your frame of reference anywhere, however, you will be analysing different parts of the system depending on what boundary conditions you choose.

            • JustAnEngineer
            • 7 years ago

            Incorrect.

            That’s a ratio of temperature [i<]differences[/i<][. 100°C = 373.15°K. 25°C = 298.15°K. 100°C - 25°C = 75°C. 373.15°K - 298.15°K = 75°K. [i<]/dayjob[/i<]

            • yogibbear
            • 7 years ago

            It’s good practice to though because almost every other equation involving temperature IS affected by not converting to absolute temperature. Though I agree that it’s not necessary.

            /dayjob as well. (Though I just click the simulate button and don’t care about the maths behind it these days as too busy dealing with compliance paperwork and other BS systematic processes to actually do real engineering anymore).

            • Airmantharp
            • 7 years ago

            Voldenuit, I’m not knocking you in any way, but when it comes to this stuff, I’m quite inclined to believe JAE. He’s the man.

            • Voldenuit
            • 7 years ago

            Oops. You’re right.

            /brainfart 😛

            • Sam125
            • 7 years ago

            [quote<]You do not need to use absolute temperatures (degrees Kelvin or Rankine) for heat transfer calculations.[/quote<] That's a really ambiguous statement. You're right that you don't need to convert units for unit deltas but otherwise they do need to be converted into Kelvin/Rankine. Lets not forget the Zeroth Law here. ; )

            • alphacheez
            • 7 years ago

            Using JAE’s equation, Airmantharp’s die area values and an ambient temp of 22 C I calculated an Overall Heat Transfer Coefficient (U in the equation) of ~18400 for Sandy Bridge and ~18900 for Ivy Bridge.

            This is model reduces the entire thermal gradient from the temperature sensor in the CPU to the ambient air as one system. That’s flawed but otherwise you have to get into the nitty gritty details of how each thermal junction and thermal pathway behaves and performs.

            I think the main thing holding IVB back is the power density. I remember hearing about the possible use of thin layer of diamond as a heat spreader between the CPU silicon and aluminum/copper heat spreader on the package and I think that might help substantially in this case.

            It will be interesting to see what people are able to find about this issue as more and more people look into the “hot” IVB CPUs.

        • Arag0n
        • 7 years ago

        You are partially right. The important thing to measure the temperature of Ivy and Sandy is the Delta from the room. You can’t get cooler than rooms temperature and your energy exchange is proportional to the temperature difference of temperature between the room’s air and your CPU cooler.

        Given a room at 15º or 25º Celsius, if Ivy is up to 80ºC and sandy is up to 60º, the difference should be calculated as (80-15)/(60-15)% = 44%.

        That means that Ivy is acting as a processor that has a 44% higher TDP even if they have the same.

          • Airmantharp
          • 7 years ago

          I see what you’re doing here- but again, temperature is not equal to TDP (your last statement).

          You’re saying that IvB has a 44% higher delta to room temperature than SnB; to put that in reference with the difference in die size of 26%, that would mean that IvB is 7% hotter than ambient when compared to SnB per mm2.

          That’s IvB at 144% over ambient in degrees Celsius compared to IvB, with 74% of the die area, so, 144 * .74 = 107; thus 7% of SnB over ambient.

            • Arag0n
            • 7 years ago

            I don’t think the area applies to the problem…. I’m not saying that temperature equals to TDP, what I said is that the temperature that a Ivy Bridge runs is equivalent to a Sandy Bridge with 44% more TDP. That means means that I’m comparing two sandy bridge with same area just different TDP. Sure you can apply the area to Ivy Bridge and say that they have a problem of size, but I’m not talking about it at all. The only thing I’m saying is that given a sandy bridge and ivy bridge and being totally unknown what is or is not inside, if you want to expect the behaviour between them, you must expect around a 44% more delta temperature to room with Ivy than Sandy.

      • Sam125
      • 7 years ago

      If the HSF were the same for both processors and the same TIM, heat spreader etcetc then what you wrote makes sense. There’s only one way to know for sure, which is to have someone get an IR camera and test the two CPUs. 😛

      • flip-mode
      • 7 years ago

      So did you ever come up with the final, correct number for the heat delta?

    • Anonymous Hamster
    • 7 years ago

    So what does it take to non-destructively remove the heat spreader?

      • UberGerbil
      • 7 years ago

      I’m certainly looking forward to the pics, non-destructive or otherwise.

        • phez
        • 7 years ago

        the first link has the pictures of the IHS removed with the TIM paste
        [url<]http://www.overclockers.com/ivy-bridge-temperatures[/url<]

          • UberGerbil
          • 7 years ago

          Thanks

          • maxxcool
          • 7 years ago

          mmm naked! Now, does it still work ?

            • Airmantharp
            • 7 years ago

            I was wondering that too!

            Also, if you wind up with a TIM’d IvB, how hard/possible would it be to re-attach the IHS with fluxless solder?

            • cheerful hamster
            • 7 years ago

            I can see it now…DIY modding with fluxless solder. That will make all the pencil-trick veterans feel like amateurs.

            • Duck
            • 7 years ago

            It might be easier than you think if you have access to a reflow oven.

            • cheerful hamster
            • 7 years ago

            Don’t we all?

            • Duck
            • 7 years ago

            I don’t. Do you?

            • cheerful hamster
            • 7 years ago

            Of course. It’s right there next to my particle accelerator.

            • ca_steve
            • 7 years ago

            [url<]http://spectrum.ieee.org/geek-life/hands-on/the-poor-mans-solder-reflow-oven[/url<]

            • Flying Fox
            • 7 years ago

            May be some AS5 or those diamond paste will be enough to do the trick?

            • MrDigi
            • 7 years ago

            You are much better with one thermal interface to the cooler. This is how most graphic cards are put together. The heat spreader just keeps the chip from getting damaged since it is a removable component. I would think a cooler with continuous contact would be ideal.

            • yogibbear
            • 7 years ago

            That’s what she said!

    • rrr
    • 7 years ago

    I hope this will be fixed in retail versions. Or if these are retail – in new stepping.

      • Firestarter
      • 7 years ago
    • rhysl
    • 7 years ago

    Anyone thinking , Intel has done this on purpose ?.. to keep SB and Ivy running together ?..

    Why Why Why would Intel put a crap interface for the Heat spreader on Ivy Bridge ?

      • NarwhaleAu
      • 7 years ago

      Its cheaper – and in most cases Ivy uses less power so doesn’t need to be soldered.

        • willmore
        • 7 years ago

        But this is in a ‘K’ chip.

      • clone
      • 7 years ago

      the thermal interface isn’t the problem… Ivy’s small die area is, I’d almost put money down that the tim is more efficient but not enough to offset the inherent difficulty that comes with transferring a lot of heat away from a tiny surface.

      Intel will be phasing out SB because Ivy will be cheaper (less silicon per CPU).

      running multiple product lines only serves to increase cost, once the manufacturing becomes mature SB will disappear.

        • flip-mode
        • 7 years ago

        The article states that Intel directly states the TIM plays a role: [quote<]However, Intel claims the combination of the new interface material and Ivy's higher thermal density is responsible for the higher temperatures users are observing with overclocked CPUs.[/quote<]

          • clone
          • 7 years ago

          gotta keep it in context… their is a web article comparing all of the different heatsink goo’s on the market.

          during testing the author also used toothpaste with the difference from best to worst being 2 degree’s….. I suspect that Intel’s new TIM is better than toothpaste while incrementally worse than the previous material.

      • kamikaziechameleon
      • 7 years ago

      from reading around there are a number of different variables that add up to contribute to this fact. There is no one reason.

      -die area
      -new thermal paste
      -new 3D switcher are apparently hotter(see some other posts below)

      I doubt any single variable could account for the 20 degree difference.

    • siberx
    • 7 years ago

    Are intel’s mobile chips still packaged as bare dies? If so, it reinforces the idea that these chips are very much geared for and intended for notebooks, as they wouldn’t have to pay the conductivity price of two TIM layers that their IHS-encased desktop counterparts would.

    • DeadOfKnight
    • 7 years ago

    Maybe if it were more efficient it would overclock better than intel wants it to. Maybe it would outperform their extreme edition processors on every level and that’s what they’re trying to avoid; who knows. I guess we’ll see when we get some subzero overclocking scores out. Honestly though I just think that the 22nm process doesn’t play as well with Ivy as we would have liked and we’ll have to wait for the Haswell architecture, especially a bigger chip (Haswell-E?), for better heat transfer when overclocking.

      • AustinW
      • 7 years ago

      I’ve seen a bunch of LN2 runs already over at xtremesystems. [url=http://www.xtremesystems.org/forums/showthread.php?280492-Ivy-3770K-amp-MSI-Z77A-GD65-LN2-Run<]Here[/url<] is one running at 6.4 GHz, for example.

      • DeadOfKnight
      • 7 years ago

      Ivy Bridge is still a great step ahead when it comes to anything but overclocking. Actually I think it’s even better at undervolting than Sandy. For those running at stock settings (and no one said you absolutely have to buy the “K” version of this chip) this will be great for energy efficiency and even better for battery life in notebooks. It’s still exciting for a mainstream (albeit higher-end mainstream) product that doesn’t have to be on the bleeding edge of performance.

      The same can be said about the GTX 680 which doesn’t overclock as well as AMD GPUs and doesn’t really raise the bar that much for performance over the GTX 580, but it is vastly more power efficient and slightly faster for about the same price. It’s fun to drool over the fastest chips when they come out, but I think it’s great that the industry as a whole has finally turned an eye towards power efficiency. Maybe everyone doesn’t care about their electric bill, but the implications this has for notebooks and mobile devices whose power is mostly limited by TDP is pretty damn impressive in my opinion.

      For those of you who actually need more processing power in your desktops and servers (although this isn’t a server chip), your time will come. The industry hasn’t abandoned you yet as far as I can tell.

      EDIT And yes I know the GTX 680 isn’t a true successor to the GTX 580. I’m still hoping for prices to drop as well. It’s still a more efficient design than fermi though.

    • ish718
    • 7 years ago

    I heard Ivy Bridge does the RROD.
    Tell Microsoft to stay away from 3D transistors with their next xbox…

      • NeelyCam
      • 7 years ago

      [url<]http://www.youtube.com/watch?v=Sk3I5IecWeQ[/url<]

        • NeelyCam
        • 7 years ago

        You guys just hate good music

    • flip-mode
    • 7 years ago

    Well, I guess the laws of physics cease to exist on top of your stove. Were these magic grits? Did you buy them from the same guy who sold Jack his beanstalk beans?

      • J-Honey
      • 7 years ago

      Are you sure about that five minutes?

      ARE YOU SURE ABOUT THAT FIVE MINUTES?!

    • kamikaziechameleon
    • 7 years ago

    waiting for that price drop on the old 2600k to pick one up.

      • wingless
      • 7 years ago

      I got mine at Microcenter the Friday before last for $199! Best purchase I’ve made in years.

        • JustAnEngineer
        • 7 years ago

        I got mine on January 8, 2011 for $330!
        Best CPU purchase I’ve made in years.

    • Wolfram23
    • 7 years ago

    Ivy Bridge is nowhere near as exciting as everyone expected. Personally I didn’t expect big performance increases, and I think the roughly 10% or so better is fine. However, I definitely expected lower power and temperatures as well as better overclocking, like we saw when we got SB compared to the Bloomfield and Lynnfield.

      • Visigoth
      • 7 years ago

      Seems like you don’t know the first thing about what a “tick” comprises. Sigh… :-/

        • Goty
        • 7 years ago

        Hmm, I’ve heard this argument before. Nevermind that the Bloomfield -> Lynnfield transition (aka, the “tick” after the Nehalem and before the SB “tock”) brought lower power consumption (significantly lower at idle), better turbo clocks, and a cheaper platform while not butchering the voltage scaling. No, IVB isn’t underwhelming because it does almost none of those things (it’s got power consumption under load down), but because we don’t know what comprises a “tick”.

          • modulusshift
          • 7 years ago

          2 data points do not a well defined and predictable set make.

            • Goty
            • 7 years ago

            I agree, but that argument goes both ways. 😉

    • kamikaziechameleon
    • 7 years ago

    No matter, I’m holding out for intel to finally release a 8 core processor. Don’t know why but in my minds eye that is a magical number.

      • UberGerbil
      • 7 years ago

      Finally? You can [url=http://ark.intel.com/search/advanced/?s=t&FamilyText=Intel%C2%AE%20Xeon%C2%AE%20Processor%207000%20Sequence&CoreCountMin=8&CoreCountMax=8<]get one now[/url<], and at [url=http://ark.intel.com/products/46495/Intel-Xeon-Processor-X6550-(18M-Cache-2_00-GHz-6_40-GTs-Intel-QPI)<]least one[/url<] has been available for two years.

        • alphacheez
        • 7 years ago

        I’m wondering if kamikazechameleon is upgrading from Dunnington…

    • Sam125
    • 7 years ago

    It’s not because of Intel’s 3-D transistor? I remember reading somewhere that the higher transistor density offered from 3D transistors would make it harder to cool.

    • Corrado
    • 7 years ago

    Did you measure the temp on the heatsink fins? If the IB is cooler, it clearly is not transferring heat as well as the SB is.

    • chuckula
    • 7 years ago

    I dismissed the first story I saw about the TIM just being cheap paste simply because the chip in question was an engineering sample and not a production chip. Now, however, I’m starting to wonder. Why would Intel do this?? The fluxless solder application for TIM does cost more, but I don’t see how it can cost that much more, especially on a smaller die where you’ll need less of the stuff than with Sandy Bridge.

    Maybe there’s some reason out there why using the old thermal packaging just doesn’t work with Ivy Bridge. Maybe it had decreased long-term reliability even though it would have helped to lower temperatures? I’m certainly no expert on how they do packaging so I don’t know what that reason is. If Intel just did this to save a buck, that’s just plain dumb (especially because they cut prices on IB.. why not just leave the prices the same and use the more expensive TIM???)

      • Kent_dieGo
      • 7 years ago

      My guess is the solder method has a non-zero failure rate. If they ruined just 1% of the chips in the solder process you can imagine how expensive that would be.

      • cocoviper
      • 7 years ago

      I’m betting it’s due to yield and testing. Sandy Bridge had a good amount of time to mature, but Ivy is so new that they may be sticking with a process that’s easily retestable if they have a problem (e.g. remove the heat-spreader and TIM, fix packaging errors like incorrectly attached die or decoupling capacitors, then attach a new TIM and heat-spreader and retest, possibly saving a $300-$1000 CPU). Reworking a soldered heat-spreader could be much more tricky. It’s possible they might transition to a soldered heat spreader after their yields improve in 3-5 months…

      • MrDigi
      • 7 years ago

      May be that the thermal paste is a dielectric which has a benefit over conductive solder.

    • Arag0n
    • 7 years ago

    Bulldozer had a higher TDP than Phenom x6, and seems that intel is having issues with their new process. Maybe someone was right for once, and below the 20nm the advantages of new node process aren’t so likely to show off. I kno bulldozer is 32nm, but that means that AMD saw the problems on node before intel, likely because intel uses HKMG and AMD still uses SOI.

    • kamikaziechameleon
    • 7 years ago

    If only we could get a comparative OC benchmark in 1 or 2 programs.

    • odizzido
    • 7 years ago

    Leave em alone guys. Intel is going through some hard times and simply cannot afford good materials.

    Or…wait what?…..

      • vargis14
      • 7 years ago

      Just remember not including R&D the switch from 32nm to 22nm will yield approx1/3 more chips per wafer.So its good for intel profits.
      I don’t like that they skimped on interfacing the IHS with ivy with it’s smaller surface area.
      But intel is in the business of making money,IVY will do that for them.
      Now if IVY performed like bulldozer people would be freaking out BIGTIME.
      Until we see a naked IVY benched we just will not know what difference the TIM material does.

    • TurtlePerson2
    • 7 years ago

    Scott and Geoff:

    You’re missing the biggest change between Ivy Bridge and Sandy Bridge. The use of FinFETs (or tri-gate transistors as Intel calls them) affects the temperature of the chips significantly. FinFETs are great because they are faster and use less power, but they are manufactured in a way that makes it more difficult to dissipate heat.

    I could send you guys some technical papers that may be able to explain this better if you’re interested.

      • Duck
      • 7 years ago

      That’s not really relevant to this article. The TIM instead of solder is the biggest change regarding high CPU temperatures under overclocking. Followed by the smaller die size (increased power density).

        • superjawes
        • 7 years ago

        How is that not relevant? TurtlePerson pointed out how the technology would make it more difficult to dissapate heat. Overclocking results in more heat being generated than under normal conditions, therefore making the technology used a possible issue.

          • FranzVonPapen
          • 7 years ago

          Just because TurtlePerson said it doesn’t make it so. Do you believe everything you read?

          I find the explanation offered in the article (which Duck echoed) to be much more plausible and meaningful than TP’s claim.

            • Sam125
            • 7 years ago

            Well, FinFET transistors are denser and have some other added benefits which are pretty much required to continue shrinking transistors at their current rate. So they have a higher heat density which would in turn lead to chips that’re harder to cool. It really isn’t rocket science (although I really love materials science). ; )

            So, in essence they both said the same thing.

            • Duck
            • 7 years ago

            Yes, that’s all I meant by it.

            It’s not really relevant because it’s It’s so far down the list, it will probably make more of a difference if you lap the IHS and heatsink base to a mirror shine.

            • superjawes
            • 7 years ago

            Without checking out the literature myself, I find myself more likely to trust someone who uses the term “technical papers” in reference to evidence 😉

            And I wasn’t really taking a stand either way other than pointing out that, without digging into all the details and figuring out how much each factor contributes to heat, TP’s comment about FinFETs is relevant to the conversation.

            PS
            I did take a peek at what I could Google, and on the substance of the original argument, I could definitely see FinFETs causing some dissipation issues. In fact: [url<]http://www.tauworkshop.com/PREVIOUS/05_Slides/tau05-3.1.pdf[/url<] Slide 14 and 15 paint a very interesting picture (especially for someone with a EE degree). Again, without doing a full analysis, I couldn't tell you how significant each individual piece of the puzzle is, but the FinFET technology could certainly have an effect on heat dissipation.

            • NeelyCam
            • 7 years ago

            It’s neat stuff, but I don’t think it applies to Intel’s finfets. The slide deck is focused on SOI processes; the oxide prevents effective heat transfer “down” from the transistor. Note the model on slide 10: there is no thermal path to the “bulk”.

            In contrast, Intel’s process is finfet on bulk silicon, where the silicon substrate conducts heat away from the transistor channel. And in flip-chip packaging, it’s this substrate that’s connected to the heat spreader with some ‘goo’.

            Yes, the fin structure could make the top of the fin heat up a bit more than in planar structures, but because the bottom of the channel is well cooled, I doubt it’s that much of an issue… So, unless TurtlePerson2 sends me those technical papers pointing otherwise, I will assume that this is not the reason why chip temperature monitors say IB is superhot.

            I think the small die size and the poor thermal contact to the heat spreader are more likely to be the culprits.

            • TurtlePerson2
            • 7 years ago

            You are correct that SOI causes similar problems, but you are incorrect to think that bulk FinFETs don’t have the thermal problems. Take a look at the link I posted below if you don’t believe me.

            • NeelyCam
            • 7 years ago

            Yeah, I already did, and I agreed with you.

            But you also said you doubt that the temperatures in the fin and in the drain/source regions would be different. That’s incorrect in bulk silicon – drain/source are well cooled because of a thermal connection to the bulk. If there is any fin structure related overheating happening, it needs to be focused on the top of the fin itself.

        • FranzVonPapen
        • 7 years ago

        I imagine you are correct and will stick to that until proven otherwise.

        • TurtlePerson2
        • 7 years ago

        The smaller die size argument/increased power density is kind of strange. Google “power density cpu” and you will see that power density has essentially leveled off.

        Sandy Bridge Die Size = 216 mm^2
        Ivy Bridge Die Size = 160 mm^2
        Sandy Bridge TDP = 95 W
        Ivy Bridge TDP = 77 W
        Sandy Bridge Power Density = 0.43 W/mm^2
        Ivy Bridge Power Density = 0.48 W/mm^2

        That’s a fairly small increase in power density.

        The argument about TIM and solder is more interesting. I don’t know much about that stuff, so I can’t really comment. If the problem were that simple, then I don’t think Intel would have it right now though.

      • Damage
      • 7 years ago

      Feel free.

        • Ragnar Dan
        • 7 years ago

        It might better illustrate the issue if you found a way to measure the temps other than only using the CPU’s temperature data. That way we can be sure the reported temps are valid, and how much of it is making its way to the HSF.

        And I for one would like to see how things work w/ the heatspreader removed on the IB to see whether that would make a big difference.

        And it’d probably help to recount the extra output per clock of the IB in this piece, too.

          • TurtlePerson2
          • 7 years ago

          I’ve done this sort of thing before and it’s really expensive. You need an IR camera to get a good idea of what’s going on.

          Of course this is all moot because a 65 watt CPU would fry in an instant if it were run without its heat spreader. Even if it were idle it would still burn up.

          It’s also moot because these things are done with flip-chip packaging, so it would be very difficult to get more accurate data than the CPU temperature sensors.

            • Ragnar Dan
            • 7 years ago

            I meant that they would remove the ‘spreader’s TIM, rather. Sorry about that. But I’m sure you’re right, anyway, but it would be an interesting exercise and would help show what’s really going on even if they just got a couple of sensors and fixed them in two spots on the HS.

            I also wonder about the voltage figures printed above, since the mobo should be in charge, and can be measured elsewhere manually, too. (Typing on a screen kbd is a PITA).

            • Sam125
            • 7 years ago

            It really shouldn’t be that expensive. Borrowing an IR camera from someone and bam, instant heat tests. You’d just need to know a Fire Marshall or a scientist who’d have one lying around. 😛

            • NeelyCam
            • 7 years ago

            Or an energy auditor. And all of those are perfectly accurate with infinite resolution

            • NeelyCam
            • 7 years ago

            [quote<]It's also moot because these things are done with flip-chip packaging, so it would be very difficult to get more accurate data than the CPU temperature sensors.[/quote<] Could you please tell us how CPU temperature sensors work?

            • TurtlePerson2
            • 7 years ago

            I don’t know exactly how it is done in a modern CPU, that’s outside my expertise. It could be done using the thermal voltage of a diode, or the frequency of an oscillator on the chip. I would assume that it’s integrated directly into the silicon, otherwise it wouldn’t be a very accurate temperature at all.

            Here’s some info on flip-chip packaging:
            [url<]http://en.wikipedia.org/wiki/Flip_chip[/url<]

            • NeelyCam
            • 7 years ago

            Using a diode for thermal detection could work, but I think that would be close (or in) the substrate, so it wouldn’t be measuring the temperate in the fin. Using a ring oscillator to detect the temperature, though, could work as fin channel temperature would affect the oscillation frequency.. hmm, that’s pretty clever!

            I was only thinking of diode-based temperature detection, so that’s why I didn’t think fin temperature could be detected with these chip monitors… Ok, so it’s possible that the fin heating could cause the chip temp monitors to show high temperature.

            The next question is: does it matter if the fin channel temperature is high, while the bulk/drain/source temperatures are low. I think it does – the hot electron effects will certainly cause bad stuff in the channel/oxide.

            I’m convinced. I agree with you that this is partly the reason (the smaller die size still plays a role, though..)

            • TurtlePerson2
            • 7 years ago

            I don’t really know enough about HCI to give you a good answer to your question. My understanding is that it only matters what the temperature in the channel is, but there shouldn’t be a very large difference between the channel, source and drain in terms of temperature. They’re less than 100 nm apart.

            It’s an interesting point that the higher temperatures could lead to increased long term breakdown and degradation. Historically NBTI, HCI, and TDDB have been fairly rare, but with shrinking process technology they’re becoming more common. All of these are affected by temperature, though I doubt that a few degrees will make any noticeable difference.

          • Corrado
          • 7 years ago

          Just do as I suggested below. Use the same cooler and take the temp from a fin in the cooler, or multiple places on the cooler. If the IB gets hotter than the SB, its transferring heat just fine and just makes more of it. If its cooler than SB, than the IB CPU is clearly not dissipating the heat properly through the heatspreader and to the HSF.

            • Ragnar Dan
            • 7 years ago

            Yep, your post wasn’t there when I last read before entering mine.

        • TurtlePerson2
        • 7 years ago

        Sent.

          • NeelyCam
          • 7 years ago

          I’d be interested, too. Could you send me a message with links?

            • TurtlePerson2
            • 7 years ago

            A good paper to look at would be:

            “Thermal analysis of ultra-thin body device scaling”

            The link below will take you to that paper if you don’t have access to something like IEEE.

            [url<]http://171.64.99.46/tcad/pubs/device/iedm03_pop.pdf[/url<] Search for FinFET spot heating or something similar and you should find several papers mentioning it.

            • NeelyCam
            • 7 years ago

            Good stuff – thanks!

      • utgiannini
      • 7 years ago

      TurtlePerson2,

      I would like to read about the heat issues if you could send me a copy or post where I can find that information. I appreciate the insights! Thanks,

    • geekl33tgamer
    • 7 years ago

    [quote<]Our Ivy CPU flirts with 100°C at those settings[/quote<] As a reference point, my FX-8120 [s<]8-Core[/s<] Quad Module CPU runs all of it's *whatever AMD's calling them these days* at 5003 Mhz. It peaks at approx 70-75C loaded. How does an additional 5W of power draw (and a negligible voltage increase) translate into 20C difference? I can't figure it out, as it defies all logic? Is the GPU component a power supping pig when the CPU's overlocked, is the cooler a worse design (I would assume you used the same cooler, or it's the same as SB's anyway)? Take a blunt edge to one and pop it's protecting cap off and take a look 😀

      • Airmantharp
      • 7 years ago

      I’m not sure how an increase in thermal density and a change in TIM for IvB resulting in a higher measured temperature at the die defies logic. With both CPUs using the same amount of power at the same voltages and same clockspeeds, it makes perfect sense. Measured temperature is not power draw nor thermal emission, which are equal.

        • geekl33tgamer
        • 7 years ago

        I’m sorry, but the temps are just TO different for it to be just that… Different TIM compounds will affect it maybe 5C either way from a baseline reference, and thermal density even less so. The latter can also be recovered with a good cooler/heatpipe combo.

        That of course assumes the chip is making good contact with that cap for maximum heat transfer, but no one knows for sure.

          • Visigoth
          • 7 years ago

          I enjoy reading these very accurate technical papers by our armchair CPU architects/scientists. I would really like to see the scientific data to back up your apparently incredibly thorough research into the field of thermodynamics.

            • ImSpartacus
            • 7 years ago

            And I enjoy reading smartass comments. In fact, I’m kind of an expert on the subject…

            • Bensam123
            • 7 years ago

            Why don’t you just take what he says with a grain of salt? Not everyone needs to release a peer reviewed paper with statistical analysis showing their personal results. Be happy he’s giving you a good idea and go write the paper and take credit for it.

          • Airmantharp
          • 7 years ago

          Temperature is just one dimension of a multidimensional system, and can’t be directly compared!

          A direct comparison would be one that compares the temperatures of IvB and SnB, all other things being equal, divided into their respective die area measurements.

      • NeelyCam
      • 7 years ago

      [quote<]As a reference point, my FX-8120 8-Core Quad Module CPU runs all of it's *whatever AMD's calling them these days* at 5003 Mhz. It peaks at approx 70-75C loaded[/quote<] Unrelated. [quote<]How does an additional 5W of power draw (and a negligible voltage increase) translate into 20C difference?[/quote<] Different process, different transistors. Different CPU. [quote<]I can't figure it out, as it defies all logic?[/quote<] It doesn't defy logic.

      • Bensam123
      • 7 years ago

      You’re saying something that a lot of people don’t want to hear. Performance wont be the same, but mhz to mhz this represents quite a bit of difference in temperature. IF and only IF you aren’t fudging numbers to make a BD look a lot better.

      I would recommend Geoff and/or Scott take a look at this.

      • NeelyCam
      • 7 years ago

      I’ll repeat myself just because I like my own… voice?

      [quote<]Let me try to take an honest stab at explaining it with an analogy. (Apologies to science purists - this is not a perfect analogy..) Think of the CPU chip as a deep lake. There is an equally deep river (thermal cement between the chip and the heat spreader) flowing from that lake to a dam (thermal paste+cooler+fan). In this analogy, power consumption is the same as somebody pouring water into the lake. If you overvolt/overclock the chip, that means more water is pouring into the lake, or if you're idling, almost no additional water is going into the lake. The water level represents chip temperature. When the water level rises, it starts flowing through the river to the dam. The dam passes the water through at some flow rate, keeping the water level from rising too much. Now, if you have a large lake (=large chip), somebody can pour a huge amount of water (power consumption) and the water level (temperature) doesn't rise that fast. It will rise, though, and eventually the dam has to pass the water through... The amount of water that needs to pass through is pretty massive, so you need a big dam. If you have a small lake (=small chip), the water level rises much faster for the same amount of water being poured in than with the large lake. Also, the river is unfortunately smaller (because the chip area with the thermal cement touching the heat spreader is so small). Even the hugest dam doesn't help that much because the river is so damn small - the water levels are still higher. This is what's happening in IB. The chip is small (small lake) but the power consumption is still large, especially overclocked/volted (water pouring into the lake). Even a very large dam that's fully open (large cooler) doesn't keep the water level (temperature) that low. Maybe you need pumps (water/LN2) to pump the water out right at the mouth of the river, to keep the lake from reaching the lakefront houses (temp going over Tjmax),,? The mouth of the river is still pretty small - you can't fit too many pumps there.. Things are much easier with a larger lake (die), say, Lake BullDozer. You can pour more water in before the water level rises too much, and the river is wider so it's easier to keep the level low. The only way Lake IvyBridge residents can handle this is by building their houses to higher elevation (higher Tjmax).. The weather is much sunnier at Lake IvyBridge, though, so it's all worth it.[/quote<]

    • Duck
    • 7 years ago

    Did you know that power draw will increase not just with increases in frequency and voltage, but with increased temperature too?

    It’s not too noticeable though. Especially if you use more powerful fans to hit lower temperatures.

      • UberGerbil
      • 7 years ago

      Thermal runaway is well-understood by just about everybody who has paid any attention to overclocking over the past decade, and I’m pretty confident that includes the entire TR crew.

        • NeelyCam
        • 7 years ago

        To add to that, thermal runaway doesn’t happen much anymore because thermal protection schemes (thermal throttling).

        • willmore
        • 7 years ago

        Not just runaway. CMOS devices run better at lower temperatures. The Rds(on) of MOSFETS is lower at lower temperatures.

        • Duck
        • 7 years ago

        Thermal runaway is more something that happens with BJT class-A power amps. I didn’t think it was that well known that CPU power consumption will increase if you let it get hot.

    • gmskking
    • 7 years ago

    So essentially there is no advantage in going with the Ivy over Sandy. This is not good, the reason why most people waited for Ivy was better temps when overclocking. This completely nulls that. Glad I went with Sandy.

      • Airmantharp
      • 7 years ago

      There are plenty of advantages, though they are more tilted to the mobile side of things.

      Still, on the desktop, you get faster Quicksync, chipsets that have a 4xUSB 3.0 controller, and PCIe 3.0 for multi-GPU rigs, to name a few.

      In the end, the 100MHz or so slower that IvB will run next to an SnB CPU will be mitigated by the marginally higher IPC.

      • absurdity
      • 7 years ago

      Hasn’t it been pretty well understood for a while that there’s little advantage to it? It is still a lower power draw (non-overclocked, anyways), which is appealing to me.

        • UberGerbil
        • 7 years ago

        It certainly seemed well-understood to me when I bought my 2500K last autumn. Of course I wasn’t planning to be a crazy overclocker (the only reason I bought the K was because it was actually cheaper than the stock 2500 during the NCIX sale), but even so I was puzzled by all the folks here who said they were waiting for Ivy Bridge, which afterall was just a tock (with some GPU tweaks, which also shouldn’t matter to most of those same folks). It seemed to me that unless you were addicted to having the very latest new hotness (in which case you already had an SB), an upgrade to SB then got you almost all of what you’d get from an upgrade to IB now… plus many more months in which to enjoy it.

        I even thought about posting a poll on the subject, because I was wondering if the enthusiasm was just a product of timing or economics — enough people had skipped a generation or two that IB just happened to be coming along at the point where people were ready to upgrade or could afford it.

        (That said, I think the argument may be very different when mobile IB comes out, and I’ve actually been resisting buying a new laptop for that reason)

          • Airmantharp
          • 7 years ago

          Well, anyone within range of Ivy most likely waited-

          but most of us just picked up Sandy, and yeah, we’re all still happy as can be :).

          I personally can’t justify an immediate upgrade to Ivy with a 2500k at 4.8GHz under the hood, running cool and quiet to boot.

          Upgrading to newer GPUs that would take advantage of the faster PCIe bus (being divided into two 8x 3.0 connections) might be a compelling reason in the future.

          • Airmantharp
          • 7 years ago

          Also, Mobile Ivy with Mobile Kepler, Optimus in tow, seems like the perfect mobile gaming solution to me. I’m just waiting for HP to shove them in an Envy with a good Nvidia GPU (660M+, please).

    • BIF
    • 7 years ago

    I didn’t mention it in the other article’s comments, but I need performance. But heat and noise are bothersome to me. This has always been a catch-33 problem for me; not unlike a three-legged stool.

    1. Not transferring heat can cause failures or can shorten component life expectancies
    2. Transferring heat can create noise, which can be fatiguing and/or can interfere with audio recording
    3. Transferring heat increases temperature in the room which is uncomfortable for human beings

    So it may well be the best decision for me to wait for Haswell to see if the heat problems have been mitigated at least a bit.

      • Firestarter
      • 7 years ago

      clocked at 4.5ghz or something like that, a normal tower cooler should handle it quietly enough unless you’re in a studio or something

      • Duck
      • 7 years ago

      You are confusing temperature with heat (power output).

      Haswell will likely increase performance, transistors, heat output. The next tick after haswell will shrink it down, lowering die size and power output.

      Therefore, If you want low power (heat) for noise and to keep your human beings feeling confortable, you buy one of Intel’s ‘ticks’ such as Ivy Bridge.

        • xeridea
        • 7 years ago

        Ivy doesn’t use any less power when overclocked. One could say that it uses more power, due to having to run your fans faster. All that talk of 3D transistors was hype.

          • bcronce
          • 7 years ago

          It uses drastically less power in “normal” operations and about the same or slightly more in extreme overclocking.

          “I thought the Toyota Prius was supposed to get good gas mileage, but I have a lead foot and slam on the breaks and the gas mileage is horrible. All hype.”

          • Duck
          • 7 years ago

          It does use less. 4.4-4.5GHz on stock voltage will have a lower power consumption than with SB at that speed and stock voltage.

          Just because you can drive it so hard the cooling can’t cope with what you are asking it to deal with doesn’t mean it always uses more power. It means it’s easier to take the overclocking too far because 22nm needs lower voltages to operate.

        • kamikaziechameleon
        • 7 years ago

        wait are you posting this after reading the article or not???

          • Goty
          • 7 years ago

          Now why would he do that and risk damaging the glorious image he has of IVB in his head!?

          😉

      • willmore
      • 7 years ago

      Duck is partly right–untill he gets of on that Haswell tangent.

      Heat and temperature are different things. Temperature increases as the amount of heat in an a given mass increases. So:

      1) you’re right, not transfering heat will cause a component to heat up–which usually reduces its lifetime
      2) Transferring heat does no create noise (unless you’re talking some something esoteric like phonons in a crystaline matrix). Actually, transferring heat well lowers the temerature difference over a thermal junction and allows the rest of the cooling solution to run slower. I can show you with number is if you like.
      3) Transfering heat will increase the temperature and that can be uncomfortable to humans, but you’re missing an important point. You are going to come to a thermal balance. All that changing the efficacy of thermal transfer does is decrease the temperature of the item being cooled. Again, I’d be glad to show you this with numbers if it would help.

      The ‘solution’ here is to either run IVB at stock clocks and voltages or wait to see how this whole TIM/soldering issue resolves.

        • NeelyCam
        • 7 years ago

        [quote<]Duck is partly right--untill he gets of on that Haswell tangent.[/quote<] Yeah - that's why I had trouble deciding if I should thumb up or down. Decided to do neither

          • Duck
          • 7 years ago

          I’m right though. What is not right about it?

            • willmore
            • 7 years ago

            Your Haswell comments are speculation. I don’t see a reason to address that.

            • Duck
            • 7 years ago

            I said it was likely. It was a prediction. It is not pure speculation because it is based on the trend that CPUs get more powerful over time, and that they tend not to increase in performance on all of Intel’s ‘ticks’.

            • Anomymous Gerbil
            • 7 years ago

            Read the article.

        • BIF
        • 7 years ago

        “Actually, transferring heat well lowers the temerature difference over a thermal junction and allows the rest of the cooling solution to run slower. I can show you with number is if you like.”

        Yep, upon reflection, I agree with this. My current Q6600 transfers heat very well with the Corsair CPU water cooler, so the corsair radiator fan and the big case fans can and do run very slowly. At these slow speeds, they produce no discernable fan noise. But my concern is that Ivy Bridge being “even warmer” might require the fans to run faster than on my current system, thereby making audible noise and potentially ruining my peace and quiet.

        I have observed that my office/studio already gets warm quickly when my Q6600 system is up and running, even if it’s at idle. The room temps are warm enough to notice and I don’t like being hot and sweaty.

        “The ‘solution’ here is to either run IVB at stock clocks and voltages or wait to see how this whole TIM/soldering issue resolves.”

        I should have said in my post that I never overclock. I just never feel the need to mess with it. So that’s one less thing I will need to worry about, and it’s one less heat-generating factor.

        Actually, I doubt that I would be pushing the system hard under most circumstances…virtual instruments can do this if you use those that require a lot of processing, but knowing my usage patterns, I’m not too worried about it, and indeed CPU is not the primary reason for my upgrade. The primary reason is to prepare to retire my current 5-year-old components in an orderly manner before they pop a bad surprise on me. I can wait another year if necessary, but I would rather not wait another two years.

        I hope this additional info helps clear some things up.

    • Firestarter
    • 7 years ago

    Welp, better break out the razor and decap that processor. I’m curious how it looks underneath and how much cooler it can run!

      • codedivine
      • 7 years ago

      Given your username, I am sure you are liking IB’s increased temparatures 😀

      • nico1982
      • 7 years ago

      Seconded.

      • willmore
      • 7 years ago

      They did decap a processor and found that the TIM was different–not soldered. It’s in TFA:
      [url<]http://www.overclockers.com/ivy-bridge-temperatures[/url<]

Pin It on Pinterest

Share This