Ivy Bridge Running Hotter Than Intel's Last-gen CPU 182
crookedvulture writes "The launch of Intel's Ivy Bridge CPUs made headlines earlier this week, but the next-gen processor's story is still being told. When overclocked, Ivy Bridge runs as much as 20C hotter than its Sandy Bridge predecessor at the same speed, despite the fact that the two chips have comparable power consumption. There are several reasons for these toasty tendencies. The new 22-nm process used to fabricate the CPU produces a smaller die with less surface area to dissipate heat. Intel has changed the thermal interface material between the CPU die and its heat spreader. Ivy also requires a much bigger step up in voltage to hit the same speeds as Sandy Bridge."
Good! (Score:5, Funny)
Re: (Score:2, Offtopic)
Wait, you were in an office not lit with horrible fluorescent strip lighting?
Re: (Score:2)
And now I'm gonna try to make a cookie on my old Athlon, just to prove that I can.
(10 minutes later)
Mmm, pretty good, even if the bottom does taste like graphite!
Re: (Score:2)
They use a good fraction of the amount of power that a CFL does, and CFLs do get hot when running. I have one LED bulb (so far) and it gets pretty warm after its been on a while.
Re: (Score:3)
Do you think it's possible to switch out some of those bulbs for the tanning ones? I've heard a lot of complaints about my skin color recently (pasty white), but I can't see myself bringing a laptop with me into one of those beds at the local tanning salon.
Re: (Score:2)
This reminds me of something that happened in one of our old computer labs, one which we affectionately called "The Ice Box". As the story goes the A/C was originally designed on the assumption that the room would contain a few dozen computers, which it did, and an equal number of CRTs, which it did not. I suppose it's because the new building took so long going up that they missed the big switch to LCDs.
The whole thing is probably apocryphal but I found it quite amusing all the same; there's something quit
Re: (Score:2)
Obviously, if it's twenty degrees outside you need more heat, not an air conditioner. ;)
Re: (Score:2)
Re:Good! (Score:5, Funny)
That's twenty degrees celsius, which is quite warm where I'm from.
That's twenty degrees of *WOOSH*, which is quite embarrassing where I'm from. ;)
Re: (Score:2)
I've had people tell me that LCD's actually generate close to the same amount of heat, but when in University I worked in a computer lab which switched and it was noticeably cooler. This was however a state school with a shared heating system between multiple buildings which meant that they did some silly things like turning the heat on and off based on a calendar not the weather, so I can't attest scientifically that this was based entirely on the LCD switch over.
Re: (Score:2)
Not to mention the invention of the thermostat.
Hotter != more heat (Score:5, Informative)
After switching all my lights to LED bulbs, its a bit cold in my office. A new, hotter CPU could be just what I need.
You're confusing temperature and heat. A candle burns hotter than a person, but a person puts out more heat (100W) than a candle (80W). Likewise, Ivy Bridge puts out less heat than Sandy Bridge, even though it's hotter.
Re: (Score:2)
Re: (Score:3)
(In heat production) I stand between the candle and the star.
Re: (Score:2)
A candle burns hotter than a person? How did they find that out? Burn the test subject in a calorimeter?
Re: (Score:2)
The take away comment here is if you want to heat a room forget about lighting a fire, just go invite some hot bodies over.
Re: (Score:2)
Can I restate what you said but a little more clearly?
The candle has higher power density than a person.
Or more specifically...
IB power output (or heat output, same thing) is LESS than SB, but it fluxes out of a smaller die area, so the junction temperature on the lid of the package is higher. The package psi-JC (junction to case heat transfer rate, a physical property) has not improved significantly, so the same heat flux rate from the die, through the lid (and through the heatsink, and through the chassi
Re: (Score:2)
>Can I restate what you said but a little more clearly?
No, apparently not.
Re: (Score:2)
LOL
Re: (Score:2)
Re: (Score:2)
The amount of heat produced is the power going in. The power going in is lower. So the heat produced will be lower. The temperature inside will rise until it reaches a point where the convection rate matches the input power. The more "insulated" the CPU die is from the outside, the hotter it will get to reach equilibrium. Think of the temperature as a form of heat movement pressure.
Re: (Score:3)
The two processors are NOT the same mass. The Ivy Bridge has less surface area which translates into higher temperatures while dissipating an equal amount of heat. If you were to be speaking of two identical metal blocks, then yes.. temperature would directly correlate to the heat output.
Why is this funny? (Score:2)
I partially heat my home office with my Pentium 4.
Really. Sometimes I boot my old box just because the room is chilly.
Re: (Score:2)
Do some BitCoin mining on your GPU, with aggression set to maximum.
It'll be like a tropical island within the hour, except without the wonderful ocean breeze that normally accompanies it.
Re: (Score:2)
A bit cold?
I once worked at a company that moved in to an office space before heating was installed. It was the middle of winter, in Canada, and all the equipment was brought up via crane through an opening in the side of the building.
I used a computer running a busy loop program to heat my office. It worked well as a space heater.
notice the "when overclocked" caveat (Score:5, Informative)
It's clear in the article, but the headline here sort of implies that the chips run hotter in general, whereas this test is only saying the new chips run hotter when overclocked. From what I can find, when run at the rated voltages/speeds, Ivy Bridge CPUs run at about the same temperature as last gen's CPUs.
Re:notice the "when overclocked" caveat (Score:5, Informative)
It's clear in the article, but the headline here sort of implies that the chips run hotter in general, whereas this test is only saying the new chips run hotter when overclocked. From what I can find, when run at the rated voltages/speeds, Ivy Bridge CPUs run at about the same temperature as last gen's CPUs.
Seems like that would make sense if at normal 'voltage/speed' the Ivy Bridge is using less power. Based on the the numbers in the link the Ivy Bridge has a higher overall thermal resistance, junction-to-air, of roughly 30% [=((100C-20C)/(80C-20C))*(231W/236W)]. Based on other reviews the Ivy Bridge processors uses less power at stock frequency/voltage so that may be offsetting much of the temperature rise due to an increase in package resistance and heatsink interface resistance, under normal conditions.
Power dissipation increases exponential with increases in frequency/voltage and it appears to rise faster with the Ivy Bridge processors. So as the power dissipation approaches or exceeds that of the Sandy Bridge processor much higher processor temperatures will be measured in the Ivy Bridge because of the higher thermal resistances.
I think this is a non-issue for the average consumer. However, overclockers would probably be better off with the Sandy Bridge hardware.
Heat transfer scales with area (Score:2)
Re: (Score:2)
Ivy Bridge is smaller in area than Sandy Bridge. Assuming I got the right numbers from Wikipedia, 160 mm^2 vs 216. That's 74% the area for heat transfer.
Agreed, but it doesn't necessarily scale linearly when including spreading affects. Using the numbers you provided one would estimate a ~35% increase in temperature at a given power dissipation, when comparing the Ivy Bridge to the Sandy Bridge. Based on the linked article the increase was only ~30%, which seems reasonable if expecting slightly improved heater transfer performance due to spreading across an oversized heatsink.
With the limited amount of hardware information in this thread, one could probab
Re: (Score:2)
Clearly, you failed to read. The measurements are in mm^2, meaning they're area measurements, not linear. The GP is correct.
Re: (Score:2)
However, overclockers would probably be better off with the Sandy Bridge hardware.
At the same clock, probably, but with the die shrink, Ivy Bridge should get better performance per watt than the Sandy Bridge OC'd to the same power. I may be making the silly assumption that OC is still about getting more performance and not just pushing hardware beyond its limits.
How Intel prices those isn't directly relevant to their performance, of course, since Ivy Bridge is the "new hotness".
Re:notice the "when overclocked" caveat (Score:5, Interesting)
Re:notice the "when overclocked" caveat (Score:5, Insightful)
Re: (Score:2, Funny)
I also care deeply about how other people use their hardware, and would like to subscribe to your newsletter.
Re: (Score:2)
As a card-carrying dyke who's fully paid up on her membership dues, I couldn't care less about e-peen.
My computer is overclocked because I do a lot of video encoding. For gaming it doesn't make a difference (and if you check the logs, most of the time in most games the CPU underclocks itself to 1.6GHz anyway), but when you're doing a video encode, particularly a large video encode or a lot of transcoding (the kind of operation that will keep your CPU pegged for 3 days in a row, and I'm talking about a Core
Re: (Score:2)
Re: (Score:2)
Not to mention how many actually DO OC their chips now? 2%? 3%? if it reached even 5% frankly I would be amazed.
Intel wouldn't be actively fighting OC in his products if it was only 2-3%. 775 socket was the last one that allowed you to OC the shit out of every cpu you put in it (Im typing this from C2D 2.4@4GHz, Celeron 420 1.6@3GHz before that). Basically almost 100% more cpu with few bios changes, a lot more fps in games (especially on weaker cpus).
It used to be possible to buy $25-40 cpu and OC to $100-150 performance level.
This time has passed with new sockets. Now Intel has OC tax, OC'able CPUs start at $200 and
Re: (Score:2)
Re: (Score:2)
If you think Intel don't let you overclock so they can upsell to some single-digit (being _very_ generous) percentage of customers (who are nearly entirely budget-driven in their purchasing), you're off in la-la land.
Re: (Score:2)
Re: (Score:2)
I'd be astounded if any more than a vanishingly tiny proportion of people overclocking did this.
It's like going to a $CAR forum and concluding
Re: (Score:2)
Re: (Score:2)
Which pretty much proves my point.
Unless you think every Tom, Dick & Harry who chips his car's engine or puts on a high-flow exhaust is a street racer ?
Re: (Score:2)
Re: (Score:2)
Intel don't block overclocking because of some insignificant number of computer enthusiasts, they do it to help prevent widespread fraud by unscrupulous operators who pass off overclocked chips as genuine.
Re: (Score:2)
When the 300A was released, you could already buy a full Pentium 2 at 450Mhz.
Re: (Score:2)
Depending on what you were doing, the full speed on-die 128k L2 cache on the overclocked 300A could be faster than the half-speed off-die 512k cache on the P2.
Re: (Score:3)
I dont know. I got a core i7 950 @ 3ghz for my new workstation at home, and with a corsair bolt on water cooler I was able to easily get it to 4.2ghz stable. It runs cool and only uses a couple hundred watts. It crushes anything I throw at it, so why not overclock?
If you overclock because you enjoy tinkering with your hardware or if you actually need every little clock increase for whatever it is you're using your computer for, more power to you. But I think he has a point - overclocking is not as necessary anymore for "standard" users/gamers as it was a few years ago.
I, too, overclocked everything back in the days of the 486, Pentium, P2, P3 (plus the various AMD alternatives). But that was mostly because back then the clock increase actually made a huge difference
Re: (Score:3)
Re: (Score:3)
The general benchmarking consensus seems to be that IB is ~10% faster than the comparable SB chips, without OCing. Throw that "invisible" 10% onto the OCed clock speed, and IB should still be coming out ahead, at a SB-style 5.1 GHz.
(I haven't seen any OC benchmark comparisons yet, to see if this is actually true)
Re:notice the "when overclocked" caveat (Score:5, Interesting)
This is true as far as it goes, but the behavior when overclocked is telling for more than how well you can overclock: At the risk of stating the obvious, the chips the overclockers are having heat issues with are the ones Intel is manufacturing. That means Intel isn't going to be able to ramp the clock speed very easily for the same reasons that the overclockers are running into trouble, unless there is some significant and avoidable flaw in the chip or the process that they can remove in future revisions.
On the plus side, this gives AMD a little breathing room to try to catch up a little.
Re: (Score:2)
Even though TDP has gone down from 95 to 77W the die size has shrunk from 216 to 160 mm^2, so energy density is up from 0.44 to 0.48 W/mm^2. It's probably getting harder and harder to make heat sinks to spread it effectively enough, particularly with overclocking. For the non-overclocker I'd say the new chips are clearly better though as they're fan noise, battery life and electricity-bill friendly with a small boost in performance and $5-10 cheaper than the equivalent SB. And a better IGP if you'll ever us
Re:notice the "when overclocked" caveat (Score:5, Insightful)
Also, if a large part of the reason why the Ivy Bridge CPU runs hotter is the smaller area of the chip and the changed thermal interface materials, this means that while the new CPU chip might run hotter than the previous one, it doesn't put out more heat. ;-)
The CPU is hotter but the heat sink is cooler since the energy can't be transferred from the chip to the heat sink fast enough.
If this is the case, then Intel need to do something about the CPU package before going to higher frequencies.
It also means that people needing the extra heat in their cold rooms would be disappointed since the heat output would be lower, not higher.
Re: (Score:2)
Intel need to do something about the CPU package before going to higher frequencies
This is the story of die shrink - more performance per area, less heat per performance, but more heat per area.
I remember when my 486's ran without any passive (much less active) cooling at all. Today even my Atoms struggle with passive cooling solutions.
Re: (Score:3)
Speed? (Score:2, Insightful)
Ivy also requires a much bigger step up in voltage to hit the same speeds as Sandy Bridge.
I get the feeling that they have very weird notions about what constitutes CPU Speed...
Re: (Score:3)
They omitted the word clock, but I'd still say their meaning was clear.
3d tri-gate not as good as promised (Score:4, Interesting)
Remember a year ago Intel was bragging about their new 3d tri-gate process would be 50% more power efficient: http://www.intel.com/content/www/us/en/silicon-innovations/standards-22nm-3d-tri-gate-transistors-presentation.html [intel.com].
Comparing the i7 3770K against the 2600K, which is clocked at the same frequency it's only 17% more power efficient: http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review/20 [anandtech.com]
Also you have to bare in mind some of the power saving is due to the DDR controller power gating
Re:3d tri-gate not as good as promised (Score:5, Informative)
Read the Anandtech review, that's total system power consumption. If you compare just the CPU power consumption it's ~33% more power efficient (66W increase from idle to load for Ivy Bridge vs 98W increase for SB). And if you look at the GPU intensive comparisons, IB is ~20% more power efficient, but that's including a ~33% increase in GPU cores and an increase in GPU clock, for an ~40% increase in performance while using 20% less power. For the first generation chips on a brand new production process, those are very good results. I expect to see them improve as their 22nm tri-gate process matures.
Re: (Score:2)
People are just whiny these days. They expect every new product to completely blow away the old one in all respects, and cry when that doesn't happen.
Same shit with the nVidia GTX 860. It is by all accounts a great card. The fastest single consumer GPU ever, much more efficient power wise per work done, emits less heat, and so on. However it doesn't completely crush the previous generation of hardware. It is only faster, not crushingly faster. So people get all mad about it as though if evil nVidia had just
Re: (Score:2)
Remember a year ago Intel was bragging about their new 3d tri-gate process would be 50% more power efficient: http://www.intel.com/content/www/us/en/silicon-innovations/standards-22nm-3d-tri-gate-transistors-presentation.html [intel.com].
Yes, the slides are unambiguous: "Greater than 50% reduction in active power going from 32nm to 22nm". Now, Intel tells us they have been predicting modest efficiency gains all along. For the last few months mabe. The truth is, Intel realized months ago the process would not meet expectations and already fired up the spin machine back then.
Subject needs "overclocking" (Score:5, Insightful)
Re: (Score:2)
Re: (Score:2)
If you have a sandybridge CPU you overclock by default. The CPU supports dynamic OCing to increase performance based on available cooling capacity, and CPU/GPU demand.
Intel TurboBoost (Score:2)
I think these days technically ALL of us do overclocking thanks to Turbo Boost and similar such technologies which up the frequency of the processor when only some cores are loaded. [intel.com]
You're slipping, intel. (Score:5, Interesting)
So lemme see here...Intel's new CPU dies are now smaller (good), which makes them less dissipative of heat (bad), so they decide to use worse thermal paste stuff?
Seems legit.
Re: (Score:2)
so they decide to use worse thermal paste stuff?
I don't think I've used Intel or AMD-supplied thermal paste in 10 years - I haven't done the precise math, but I assume I make it back on the electric bill over time by using less active cooling energy. And I rarely overclock anything.
Comment removed (Score:4, Insightful)
Re: (Score:2)
Its an ENGINEERING SAMPLE (Score:2)
If the retail releases also have this issue, then its newsworthy
CARE FACTOR ZERO - don't over clock problem solved (Score:2)
The majority of people do not overclock their CPUs so this is not an issue for the majority.
Easily explainable (Score:2)
Power consumption varies with the square of the voltage (p=v^2/r) while the power consumption varies linearly with the frequency, if it takes signicantly more voltage to over clock then it's no wonder the power usage is so high.
Re: (Score:2)
RTFA, it doesn't take more voltage (absolute), it requires a higher boost above "stock" voltage.
There are several reasons for this:
1. lower thermal conductivity of the packaging. Higher temperatures can increase resistance, and therefore, voltage requirements.
2. 22nm process is brand new. Neither the process nor the CPU die have been optimized yet. Both will mature over the next 12 months.
3. SB is a second generation 32nm design on a fully mature process, it's nearly as good as it will get.
4. At 22nm, there
Heat Spreader (Score:2)
Looking at the pictures in the Overclockers.com link, you'd probably get better thermal dissipation if had Intel left the heat spreader off, with nothing expect the protective overcoat on the back of the chip.
Actually, I bet modders are going to start cracking the IHS off for that very purpose, in order to directly contact their heatsinks.
This is a Feature (Score:4, Insightful)
It's hotter when overclocked. Overclockers love having to run pipes and submerge things. How are you going to justify hauling out the liquid nitrogen if it's running cool?
Meanwhile everyone else is happier that it runs cooler, takes less power, is faster, and even costs less than Sandy Bridge.
This is Win Win, people.
Why are you OCing Ivy? (Score:2)
Seriously. What kind of quantum chromodynamics calculations are you simply not getting done today?
Heh heh heh (Score:3)
Re: (Score:2)
>wouldn't have it been possible for the engineers to figure out some sort of solution
They did. The solution is "Don't overclock your processor".
Re: (Score:2)
Or "Overclock it less, since Sandy Bridge spoiled everyone."
Re: (Score:2)
Depends on the CPU, but most (~90%) Sandy Bridge i5 CPU's will take 4GHz without any problem, and about 50% can hit 4.5GHz without breaking a sweat. This is for an i5 2500k, which comes with a stock clock speed of 3.2GHz. My 2500k is currently running at 4.7GHz and it doesn't have any stability problems at all, even when I run video encodes that keep the CPU at 100% occupancy 24/7... when I push it to 4.8GHz (change the multiplier from 46x to 47x, with system bus at 103MHz), it starts to crash. I suspect th
Re: (Score:2)
That's coming in q4.
Re: (Score:2)
... a processor without a on die gpu, if I'm going to have more heat I want more performance.
Yes, as I read the article, it basically says "New CPU processing speeds, New CPU heat levels ... with overclocked OLD cpu".
Re: (Score:3)
Uhhh, you realise it turns itself off if you don't use it...
Re: (Score:2)
Re: (Score:3)
Ivy Bridge does not use more power than Sandy Bridge and it's as expensive/cheaper than Sandy Bridge.
The only thing that isn't as good is overclocking headroom. Hardly sounds like a bad compromise, especially considering the much-improved GPU.
Re: (Score:2)
People paying for these CPU are gamers and enthusiasms anyway, they have a discrete GPU and don't use the built-in one. I'm pretty sure less than 1% of the SandyBridge/IvyBridge sold use the built-in GPU since it's an horror compared to real graphic cards.
You'd be wrong, especially in laptops. I have a Sandy Bridge Celeron U3600 in my laptop, and I'm using the integrated graphics. They're plenty powerful enough for desktop compositing, and I get respectable framerates playing the occasional game.
Re: (Score:2)
My laptop has manually-switchable graphics (Sandy Bridge or extra-low-end AMD chip), and I've noticed that Sandy Bridge is actually a bit better at hardware acceleration (In IE9) than the AMD chip.
Re: (Score:2)
P.S. You should check out the Ivy Bridge pricing before opening your stupid mouth.
Re: (Score:2)
When you say "marginally", can you find a single source / release from the past year indicating more than a 10-20% speed boost for IB? Because this is what was predicted, and what has been delivered.
Re: (Score:2)
Re: (Score:2)
As for being a "partner" a lot of muslims consider dog to be a dirty unclean animal.
Yes, Muslims are hateful bigots. Your point?
Irony meter overload.
Re: (Score:3, Funny)
You think that's bad?
He was born in Hawaii. Odds are, at some point, he has actually eaten Spam.
*shudder*
Re: (Score:3, Funny)
Dog is tasty. I just can't vote for someone who puts pineapple on his pizza.
Re: (Score:2)
Dog is tasty. I just can't vote for someone who puts pineapple on his pizza.
Practically nobody in Hawaii puts pineapple on pizza. Hardly anyone in Hawaii even cooks pineapple, we eat it fresh or at worst canned. I never even heard of "hawaiian pizza" until I moved to the mainland. Wikipedia says it was first created in Canada.
Re: (Score:2)
Practically nobody in Hawaii puts pineapple on pizza
I sure did when I lived there. So did other people. Called it pineapple and ham though, not 'hawaiian.' The Dominos Chicken Pestoza was good too, but I don't know if they serve it anymore over there.
And spam musubi is the best bra.
Re: (Score:2)
It's "brah." We don't call people boobslings.
Oh I know. I would never mock Hawaiianspeak on purpose.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
its actually a bit cooler, intel does not give a shit about thermal performance outside of its defined operating parameters, and I dont blame them
Re: (Score:2)