another example of how hot IVY runs

[vargis14]

http://www.tweaktown.com/reviews/4663/a ... dex11.html

This is a day old tweaktown review showing the 3770k hitting 98c at 4700 or so MHZ and 1.35 or so volts......its starting to look like intels 22nm process has some leakage.Even stock temps are higher then a comparable sandy bridge cpu.

I for one will not be upgrading to ivy bridge the performance gains are minimal compared to my 2600k plus i can overclock without needing a car radiator to cool the cpu.Perhaps a overclocked Ivy cpu will help heat a room with inadequate heating.
Humm its chilly in here let me run prime 95 and warm up this room
Once again Anyone who upgraded to sandy bridge over a year ago made the best decision possible for futureproofing.
I mean when is the last time anyone can remember that a cpu over a year old is still in the top dog 5% of any cpu out there including cpus yet to be released.
It is amazing in my eyes,back in 2005 and before if you purchased the best cpu you would be very lucky to have a top 5% performing cpu 3-5 months later.
I feel there is no reason whatsoever to upgrade to ivy if you are currently using a sandy cpu.......i would love to grab a few 200$ 2600ks from my microcenter then sell then on ebay once supplies dry out.
Looks like we will have to wait for haswell to see what it brings to the table before i even think of upgrading.If pci-E 2.0 at 8x would bottleneck a gtx 680 or 7970 i could see upgrading to ivy for pci 3.0 but until pci2.0 turns into a bottleneck there is absolutely no reason to upgrade.
That my opinion feel free to disagree, agree and add your thoughts.
Peace

[Flying Fox]

You should have posted the temperature topic in your other thread, not spawn another one in another forum. But sure, you are trying to justify not needing to upgrade, will let you run a bit here for now.

BTW, the temps are just for OCed, how about stock? Is it also hotter than SB? This may be the reason why Intel is delyaing the notebook chips to fit the thermals in another revision/stepping.

I am thinking about an IB HTPC build, we will see how much more power and heat for the lower end models.

[Voldenuit]

Meanwhile, the stock 3770K draws 22W less under load and 34W less under idle. It's still hotter than the 2600K by 9°C under load at 10°C at idle, but it's hard to separate the cause because the benchmarks were run on different motherboards - fan profiles, voltage profiles, etc may be different between the ASRock and ASUS boards, for starters. The higher temperatures could also be due to a higher power density because there is less surface area to transfer heat to the cooler.

EDIT: Oh, and are they reporting absolute temperature or temperature delta? Are they using the on-chip sensor (which will give calibration errors between chips) or a dedicated thermal probe? They also only record a single core's temperature - differences in how IB loads cores vs SB and BD would influence that - for instance, the IB system might be running a single core at higher clocks at idle while the rest of the cores are in low power state as a hypothetical situation.

In short, there are still a lot of unanswered questions, and I don't think the tweaktown article answers them. It does raise some questions, though, and it would be worth other sites pursuing these questions in their reviews when IB comes off NDA.

[End User]

I look forward to the TR review. I'm in dire need of transferring my current i7-920 rig over to Linux duties after my current Q6600 Linux box died. The 3770K was my first choice.

[mako]

This is an engineering sample, right? While the thermals probably won't change much for the final silicon, I still would wait and see.

[Ifalna]

I believe so ES have E0 /8 Steppings, but there are already processors out there with E1 / 9 Stepping which is the "release Stepping" and these get hot when OCd too.

I just link the threads of peeps toying around with them:
http://www.overclock.net/t/1242313/more ... pdated/160
http://forums.anandtech.com/showthread. ... 05&page=17

Fun fact: You can already buy Ivy i5 3570K processors. Kinda makes me wonder whether it's supposed to be that way to fuel the rumor mills or if vendors use illegal tactics to get an advantage.

Personally I'll wait for TRs Benchmarks before I decide whether to plug a SB or an IB into my Z77 MoBo.

[flip-mode]

I for one will not be upgrading to ivy bridge the performance gains are minimal compared to my 2600k

You deserve this: Duh!

[Derfer]

So if it's more power efficient, but super leaky, then we could get another huge power efficiency boost by fixing the leaks? Presumably in haswell.

[vargis14]

I cannot remember where but just this morning i read that they just might change the tdp from 77 watts to 95 like the current sandy k series.
Seems like the switch to 22nm does slightly lower wattage but tdp must include heat output as well.Looks like the 22nm process will just make intel more chips per wafer,improving profit margins,along with the native usb3 ,PCIe 3.0 and improved on die graphics of course.

And to flip was a DUH really necessary.Its does not really bother me you can remark all you want if you think i am a not a smart person.I do not disrespect anyone on the forums or any blog.I would just expect the same respect i show everyone else.
And all because i said something most of us would understand without saying, but not everyone reading this forum knows everything you do.

[TurtlePerson2]

This topic is kind of strange. I don't think people remember their high school (or college) physics.

Heat is energy. Power is the integral of energy with respect to time. Therefore, if two processors run for the same amount of time and one has a lower average power dissipation during that time then it will dissipate less heat.

By the way, leakage in CPUs isn't something that you just patch up like a leaking pipe. Leakage will not be getting much better in Haswell. It's the same 22 nm process, so any gains in leakage will have to come from the circuits themselves. Large gains in leakage reduction generally come from process technology (e.g. HKMG).

Seems like the switch to 22nm does slightly lower wattage but tdp must include heat output as well.Looks like the 22nm process will just make intel more chips per wafer,improving profit margins,along with the native usb3 ,PCIe 3.0 and improved on die graphics of course.

No. Wattage is power. TDP is the exact same power. All good temperature measurements are taken after the chip has been running at 100% utilization for a long time. The maximum temperature is a function of two things: the chip's power and the cooling system.

Please stop making stuff up.

[Voldenuit]

Seems like the switch to 22nm does slightly lower wattage but tdp must include heat output as well.

Temperature != heat.

A smaller chip that puts out less heat might end up at a higher temperature because there is less surface area for heat transfer/dissipation.

Tweaktown's test showing Ivy drawing 22W less power at the wall than Sandy supports Ivy's TDP being 77W instead of 95W.

[integer]

I cannot remember where but just this morning i read that they just might change the tdp from 77 watts to 95 like the current sandy k series.

Was it this NordicHardware article?

[vargis14]

Nope ,but a good find there.What i read did not have pictures:)

[novv]

This is exactly the same old story with Pentium 4 Prescott vs Northwood. As everyone remembers Northwood was much cooler and enjoyed a greater OC too. So nothing new...

[TurtlePerson2]

My VLSI design professor lectured about FinFETs a few days ago. A quote from the slides, "Confined geometries, multiple tightly packed parallel fins hinder heat removal." Apparently this can cause the temperature of the devices to be hotter than usual, especially for large devices such as clock drivers. According to the professor's slides, "Can raise temperature by 10 to 15 ºC for clock and I/O devices." This could explain why Ivy Bridge might not overclock as well as Sandy Bridge. While the overall TDP of Ivy Bridge could still be lower, localized heating could cause certain devices to cease functioning at higher frequencies.

[Corrado]

No. Wattage is power. TDP is the exact same power. All good temperature measurements are taken after the chip has been running at 100% utilization for a long time. The maximum temperature is a function of two things: the chip's power and the cooling system.

Please stop making stuff up.

Also efficiency, (I believe). I can't speak from a processor perspective, but physics in general need to take efficiency into account. Heat is nothing more than power that was not used to complete work. If somehow, you have a 100% efficient chip (not possible, but theoretically), you wouldn't need a heatsink as 100% of the electricity going in would be converted into work.

However, in this case, I believe that the efficiency of the architecture is roughly equal.

[Ifalna]

While the overall TDP of Ivy Bridge could still be lower, localized heating could cause certain devices to cease functioning at higher frequencies.

Now combine that with a much smaller Chip-Area, thus less Area for heat transfer to the cooler and you know why Ivy runs hotter than sandy.

Maybe Intel needs to improve heat transfer DIE->Spreader and combine that with the carbon/copper sink tech we've read about.

you wouldn't need a heatsink as 100% of the electricity going in would be converted into work.

Nope. Not unless you find a friction free conductive material, which does not exist (at normal temperatures). I'm not really sure how one can speak about efficiency in that context the first place, because when you put a current though a thin wire you WILL get heat. There's no architecture that can avoid that.

Now the architecture can be optimized that less transistors need to switch states to accomplish the mathematical operation, thus consuming less power (per operation), and they can make the transistors smaller so they need to transport fewer electrons/holes to form a conductive/resistive layer (Difficult, electrical field strength gets very large, that's why they have to reduce the voltages) but that's pretty much it.

[chuckula]

OK, from the reports I've seen online here's an initial take:

1. Ivy Bridge does consume less power than Sandy Bridge at stock speeds.
2. BUT: The chips themselves tend to heat up more. Since IB is much smaller than SB, the die can heat up to a higher temperature while still consuming power at a lower rate.
3. The 95 watt TDP labeling is a little confusing but may be due more to the *cooling* requirements than the actual power consumption. Basically, you need a beefier cooler to keep the chip within the specified temperature range even though the actual power being consumed by the chip is in fact lower.
4. Overclocking is not that great on IB due to the temperature issue but not really due to power consumption. Interestingly enough, an IB overclock to 4.8 Ghz uses a a little less power than a stock-clocked Bulldozer 8150 (according to the Tweaktown article that has been posted). The IB temperature, however, shoots up to unacceptable levels as you go above 4.5Ghz and approach 5Ghz. Basically, you can overclock IB to about 4.5Ghz without too much difficulty, but it will not do as well as the more mature Sandy Bridge samples out there that can hit 5 Ghz without too much trouble.
5. Keep your Sandy Bridge chip if it overclocks well. If you are interested in a lower-power mobile device, then Ivy Bridge should be good since it *does* consume less power than SB.
6. Intel will likely be able to improve these numbers over time as the 22nm process matures. Remember that when they first launched the 32nm process, the only chips that came out were the huge Gulftowns and later on the lower-end 2-core Clarkdales. The launch at 22nm, while not without issues, is actually much more ambitious in attacking a lot of different market segments quickly.

[Corrado]

Nope. Not unless you find a friction free conductive material, which does not exist (at normal temperatures). I'm not really sure how one can speak about efficiency in that context the first place, because when you put a current though a thin wire you WILL get heat. There's no architecture that can avoid that.

I know that, thats why I said its NOT physically possible. In theory though, if something is 100% efficient, there is no energy lost through heat. Anyone with basic physics knows that you can NEVER have something that is 100% efficient though.

However some metals have more or less resistance and will he up more or less when running current through them. Also, some architectures are more efficient than others, otherwise every Xghz chip with Y number of transistors on the same lithography process at Z load would produce the exact same heat, but they don't.

[Ifalna]

I know that, thats why I said its NOT physically possible.

Not quite correct on that one. It's impossible at normal temperatures, once you approach a low enough temperature it's another matter (sadly only for exotic materials).

Not exactly "on topic" but it might be an interesting read.

http://en.wikipedia.org/wiki/Superconductivity

[flip-mode]

OK, from the reports I've seen online here's an initial take:

1. Ivy Bridge does consume less power than Sandy Bridge at stock speeds.
2. BUT: The chips themselves tend to heat up more. Since IB is much smaller than SB, the die can heat up to a higher temperature while still consuming power at a lower rate.
3. The 95 watt TDP labeling is a little confusing but may be due more to the *cooling* requirements than the actual power consumption. Basically, you need a beefier cooler to keep the chip within the specified temperature range even though the actual power being consumed by the chip is in fact lower.
4. Overclocking is not that great on IB due to the temperature issue but not really due to power consumption. Interestingly enough, an IB overclock to 4.8 Ghz uses a a little less power than a stock-clocked Bulldozer 8150 (according to the Tweaktown article that has been posted). The IB temperature, however, shoots up to unacceptable levels as you go above 4.5Ghz and approach 5Ghz. Basically, you can overclock IB to about 4.5Ghz without too much difficulty, but it will not do as well as the more mature Sandy Bridge samples out there that can hit 5 Ghz without too much trouble.
5. Keep your Sandy Bridge chip if it overclocks well. If you are interested in a lower-power mobile device, then Ivy Bridge should be good since it *does* consume less power than SB.
6. Intel will likely be able to improve these numbers over time as the 22nm process matures. Remember that when they first launched the 32nm process, the only chips that came out were the huge Gulftowns and later on the lower-end 2-core Clarkdales. The launch at 22nm, while not without issues, is actually much more ambitious in attacking a lot of different market segments quickly.

That is all very interesting. Based upon that I might even prefer Sandy over Ivy in many cases. Very interesting. Ivy is certainly no Penryn.

I'm starting to get the itch and the Haswell scratch seems a long way off. Haswell could bring a much improved 22nm process along with a much improved architecture all at once. Could be big.

[flip-mode]

I know that, thats why I said its NOT physically possible.

Not quite correct on that one. It's impossible at normal temperatures, once you approach a low enough temperature it's another matter (sadly only for exotic materials).

Not exactly "on topic" but it might be an interesting read.

http://en.wikipedia.org/wiki/Superconductivity

Quit trying so hard to miss the point he's making.

[Flying Fox]

OK, from the reports I've seen online here's an initial take:

1. Ivy Bridge does consume less power than Sandy Bridge at stock speeds.
2. BUT: The chips themselves tend to heat up more. Since IB is much smaller than SB, the die can heat up to a higher temperature while still consuming power at a lower rate.
3. The 95 watt TDP labeling is a little confusing but may be due more to the *cooling* requirements than the actual power consumption. Basically, you need a beefier cooler to keep the chip within the specified temperature range even though the actual power being consumed by the chip is in fact lower.
4. Overclocking is not that great on IB due to the temperature issue but not really due to power consumption. Interestingly enough, an IB overclock to 4.8 Ghz uses a a little less power than a stock-clocked Bulldozer 8150 (according to the Tweaktown article that has been posted). The IB temperature, however, shoots up to unacceptable levels as you go above 4.5Ghz and approach 5Ghz. Basically, you can overclock IB to about 4.5Ghz without too much difficulty, but it will not do as well as the more mature Sandy Bridge samples out there that can hit 5 Ghz without too much trouble.
5. Keep your Sandy Bridge chip if it overclocks well. If you are interested in a lower-power mobile device, then Ivy Bridge should be good since it *does* consume less power than SB.
6. Intel will likely be able to improve these numbers over time as the 22nm process matures. Remember that when they first launched the 32nm process, the only chips that came out were the huge Gulftowns and later on the lower-end 2-core Clarkdales. The launch at 22nm, while not without issues, is actually much more ambitious in attacking a lot of different market segments quickly.

That is all very interesting. Based upon that I might even prefer Sandy over Ivy in many cases. Very interesting. Ivy is certainly no Penryn.

I'm starting to get the itch and the Haswell scratch seems a long way off. Haswell could bring a much improved 22nm process along with a much improved architecture all at once. Could be big.

Simply put, with Intel's tick-tock strategy there is usually little to gain if you upgrade from the tock (new architecture) to the tick (the shrink). In fact, of most hardware, going from 1 gen to the next is going to have small incrementatl differences. The value buyer usually skips 1 or 2 generations for his/her upgrade. It is nothing new. Heck, my 875K is still ticking strong and not much incentive for me to upgrade. Now if I want to hand it down to my brother and then upgrade myself to scratch that itch, that's a different story of course.

My problem with this thread is, the OP seemed to be trying to publicly say how hot and stuff to justify his not wanting to go from upgrading his SB. There really is no need to do that. Certainly it is not for our amusement.

[flip-mode]

Simply put, with Intel's tick-tock strategy there is usually little to gain if you upgrade from the tock (new architecture) to the tick (the shrink). In fact, of most hardware, going from 1 gen to the next is going to have small incrementatl differences. The value buyer usually skips 1 or 2 generations for his/her upgrade. It is nothing new. Heck, my 875K is still ticking strong and not much incentive for me to upgrade. Now if I want to hand it down to my brother and then upgrade myself to scratch that itch, that's a different story of course.

My problem with this thread is, the OP seemed to be trying to publicly say how hot and stuff to justify his not wanting to go from upgrading his SB. There really is no need to do that. Certainly it is not for our amusement.

That was annoyance as well and was why I said the OP deserved an eye roll At home an X4 955 is doing all I need but I'm still on DDR2. I have 8 GB but that runs out quick when I start experimenting with virtual machines. Upgrading to 16 GB or even 12 GB of DDR 2 is a non-starter - DDR2 is crazy expensive. So my conundrum is how to get to DDR3. I don't really want to put money into replacing my AM2+ motherboard with an AM3+ motherboard - that feels like a crummy investment. So it's either Sandy/Ivy or wait for Haswell. I was pretty happy waiting for Haswell until I started messing with virtual machines more and more and getting annoyed with running out of RAM. When I saw that deal for the $200 i7 2600K I almost cried - it really made me want to jump. I even made a passing mention of it to the wife. Unfortunately, our home laptop is also getting pretty crusty.

[SPOOFE]

The value buyer usually skips 1 or 2 generations for his/her upgrade.

'Tis what I was going to say. It astounds me to see people talking about owning Gen X processor and lamenting how Gen X+1 isn't so great in comparison... forgetting that there are still people out there running on Gen X-2 or -3.

I don't even overclock and I don't anticipate upgrading my SB i7 for five years or so. Maybe four if my near-term future is more lucrative than I expect.

[clone]

Meanwhile, the stock 3770K draws 22W less under load and 34W less under idle. It's still hotter than the 2600K by 9°C under load at 10°C at idle, but it's hard to separate the cause because the benchmarks were run on different motherboards - fan profiles, voltage profiles, etc may be different between the ASRock and ASUS boards, for starters. The higher temperatures could also be due to a higher power density because there is less surface area to transfer heat to the cooler.

I always wondered if surface area was going to become a problem, assumed if it became an issue they'd start using fewer and fewer layers but as transistor count increased the surface area didn't change much.

today is different though, I wonder if localized heating is too blame as cores go unused?

[Ifalna]

Quit trying so hard to miss the point he's making.

There is no point.

Since normal users (and i'm sure most tech labs) don't have a means to measure the actual amount energy that goes out as heat we cannot say anything about efficiency by using the temperature. Just because a CPU 1 gets hotter than CPU 2 does not mean that processor 1 actually radiates more energy than the other. If the Area for heat transfer gets smaller and the dissipation of heat gets more difficult you need LESS energy to achieve higher temperatures.

All we can do is measure the energy that goes IN during a given time period in which we let it perform a representative task, that gives us a "energy/task" figure. If that figure is lower on CPU 1 it's more efficient that CPU 2. If CPU 1 gets 100K hotter (arbitrary number ) it tells us only one thing: Dissipation of the heat is too difficult, we either need to work on reducing the resistance for heat or need more aggressive cooling.

[Arclight]

http://www.tweaktown.com/reviews/4663/asrock_z77_extreme6_intel_z77_with_ivy_bridge_motherboard_review/index11.html

This is a day old tweaktown review showing the 3770k hitting 98c at 4700 or so MHZ and 1.35 or so volts......its starting to look like intels 22nm process has some leakage.Even stock temps are higher then a comparable sandy bridge cpu.[...]

That can't be an indicative of the final product. He most probably has an engineering sample.

Also, with all due respect towards TT, but look at FX 8150 sporting 8 degrees in idle......Is he testing in a freezer? And even if those values were the delta from the ambient temperature, it's still impossible cause that would make the Ivy's temp go way past 100 degrees Celsius which would most proabably never happen since the chip would throttle down.

[Waco]

That can't be an indicative of the final product. He most probably has an engineering sample.

Also, with all due respect towards TT, but look at FX 8150 sporting 8 degrees in idle......Is he testing in a freezer? And even if those values were the delta from the ambient temperature, it's still impossible cause that would make the Ivy's temp go way past 100 degrees Celsius which would most proabably never happen since the chip would throttle down.

All FX chips have messed up temperature readings at idle. My 8120 when I had it would report 13 C idles when running at 4.5 GHz. My water cooling is badass...but not that badass.

[TurtlePerson2]

Quit trying so hard to miss the point he's making.

There is no point.

Since normal users (and i'm sure most tech labs) don't have a means to measure the actual amount energy that goes out as heat we cannot say anything about efficiency by using the temperature. Just because a CPU 1 gets hotter than CPU 2 does not mean that processor 1 actually radiates more energy than the other. If the Area for heat transfer gets smaller and the dissipation of heat gets more difficult you need LESS energy to achieve higher temperatures.

All we can do is measure the energy that goes IN during a given time period in which we let it perform a representative task, that gives us a "energy/task" figure. If that figure is lower on CPU 1 it's more efficient that CPU 2. If CPU 1 gets 100K hotter (arbitrary number ) it tells us only one thing: Dissipation of the heat is too difficult, we either need to work on reducing the resistance for heat or need more aggressive cooling.

The electrical power that is dissipated in the computer is equal to the heat dissipation of the machine. That's how thermodynamics works. Any equipment that measures the current and voltage that goes into the machine will tell you how much heat the system dissipates.