• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Intel Admits That Ivy Bridge Runs Hotter Because Of 22nm Shrink

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
pelov said:
Don, I think his point was that IB-E would have a larger die thus better heat dissipation. The issue with Ivy isn't the voltage but rather that it can't dissipate the heat with that added voltage. A larger die could potentially alleviate some of those issues and we may see some higher clocks.

We saw the same with SB vs SB-E where the larger SB-E chips would idle and load at lower temperatures given the other conditions are equal. It's different this time around --as you've already mentioned the prime operating voltage with that chart -- so it remains to be seen how much or if a larger die would benefit.
Click to expand...

I'm actually looking very forward to IBE's chip design. I have a feeling Intel is going to spread out the cores to allow for more heat dissipation.

I'll probably be going back to X79 because of it
 
What i mean is that SB was designed for the 32nm process. IB is "just" an SB shrink with a better iGPU manufactured in a 22nm tri-gate process.
Intel would have had to change the SB architecture (core/cache layout and topography, execution units etc) of SB in order not to have that high power density, hot spotting and nonuniform power map with IB at 22nm Tri-Gate.

On the other hand, Haswell is being designed for the 22nm tri-gate process and i believe it will not have the same shortcomings.
 
Wouldn't Ivy Bridge run cooler if they simply deleted the on-chip GPU?
Then arrange the 4 cores into 4 symmetrical quadrants of the square CPU, to optimize cooling.
Who really needs 8 cores of an Ivy Bridge-E, anyway?
Why does Intel insist on force goose-feeding us the on-chip GPU?
We, the customers, want something better.
 
vailr said:
Wouldn't Ivy Bridge run cooler if they simply deleted the on-chip GPU?
Then arrange the 4 cores into 4 symmetrical quadrants of the square CPU, to optimize cooling.
Who really needs 8 cores of an Ivy Bridge-E, anyway?
Why does Intel insist on force goose-feeding us the on-chip GPU?
We, the customers, want something better.
Click to expand...

1. Disabling the IGP is negligible
2. Intel engineers design these chips like this for a reason
3. IGP is great for laptops. The bulk of Intels sales
 
On the new system that I just build using the Intel® Core™ i5-3570K, Intel Desktop board DZ77GA-70K and the Corsair H100 I was hitting temps in the high 80s to low 90s at 4.6GHz. Which I do believe to be a little higher than what I would like to stay at. So my experience is that it is running hot so far.
 
The GPU acts as dark silicon as well, actually aiding in the cooling of the chip.

Furthermore, you can't just "delete" the GPU. You'd have an even smaller chip with exacerbated cooling issues.
 
pelov said:
The GPU acts as dark silicon as well, actually aiding in the cooling of the chip.

Furthermore, you can't just "delete" the GPU. You'd have an even smaller chip with exacerbated cooling issues.
Click to expand...

What I meant was: leave the overall CPU dimensions about the same, omitting the on-chip GPU. If it would help, add in cooling channels made of carbon fiber tubes filled with a non-volatile liquid antifreeze (or whatever) for improved cooling. Then call for a re-designed Z78 motherboard with no video out ports.
In other words: sort of a 22 nm die-shrink of the socket 1366 I7 CPU, but re-engineered to fit in an 1155 socket.
 
Last edited:
vailr said:
What I meant was: leave the overall CPU dimensions about the same, omitting the on-chip GPU. If it would help, add in cooling channels made of carbon fiber tubes filled with a non-volatile liquid antifreeze (or whatever) for improved cooling. Then call for a re-designed Z78 motherboard with no video out ports.
In other words: sort of a 22 nm die-shrink of the socket 1366 I7 CPU, but re-engineered to fit in an 1155 socket.
Click to expand...

They already have that. It's called Socket 2011
 
Don Karnage said:
They already have that. It's called Socket 2011
Click to expand...

Too expensive and too many cores for my needs.
What's wrong with a good quality yet <$150 up-to-date & current technology motherboard, using a (low voltage & cool running) 22 nm, 4 core (hyperthreaded) CPU, designed specifically for desktop machines using only a discrete video card?
 
Last edited:
Can someone clarify - is the issue that there is more heat than before, so use a bigger heatsink/cooler/fan to strip off the excess heat?

Or it is more insidious, that there is a "heat bottleneck" such that the chip can't get rid of the excess heat, no matter how big of a heatsink you use, due to the small footprint of the chip?

I would think it's the first issue, as the 2nd seems to defy thermodynamics unless there is a bizarre heat blocking agent at work inside the chip?
 
KingFatty said:
Can someone clarify - is the issue that there is more heat than before, so use a bigger heatsink/cooler/fan to strip off the excess heat?

Or it is more insidious, that there is a "heat bottleneck" such that the chip can't get rid of the excess heat, no matter how big of a heatsink you use, due to the small footprint of the chip?

I would think it's the first issue, as the 2nd seems to defy thermodynamics unless there is a bizarre heat blocking agent at work inside the chip?
Click to expand...

The issue is the chip is alot smaller. So you need to remove alot more heat per mm2. Even tho you produce less overall heat.

[
pd.jpg
 
KingFatty said:
Can someone clarify - is the issue that there is more heat than before, so use a bigger heatsink/cooler/fan to strip off the excess heat?

Or it is more insidious, that there is a "heat bottleneck" such that the chip can't get rid of the excess heat, no matter how big of a heatsink you use, due to the small footprint of the chip?

I would think it's the first issue, as the 2nd seems to defy thermodynamics unless there is a bizarre heat blocking agent at work inside the chip?
Click to expand...

In laments terms the die is so small the heatsink cannot remove the heat fast enough. It'll only get worse with 14nm
 
"as the 2nd seems to defy thermodynamics unless there is a bizarre heat blocking agent at work inside the chip? "

it's a 3D chip so more filler I would think ,so the heat has more insulation to travel though to get to the heat spreader.
-if sb did not run relatively cool @ eg.4.8 ghz, ib temps would not stand out as being high. IMO
 
Just to make sure, did Intel "cheap" out by using the paste instead of the solder or are they roughly the same? And if they did cheap out, isn't that good for us as lower prices? Well, besides overclockers, I mean.

(I will admit, I was looking forward to 5.2 Ghz on air-cooled Ivy like everyone else.)
 
KingFatty said:
Can someone clarify - is the issue that there is more heat than before, so use a bigger heatsink/cooler/fan to strip off the excess heat?

Or it is more insidious, that there is a "heat bottleneck" such that the chip can't get rid of the excess heat, no matter how big of a heatsink you use, due to the small footprint of the chip?

I would think it's the first issue, as the 2nd seems to defy thermodynamics unless there is a bizarre heat blocking agent at work inside the chip?
Click to expand...



Revolution 11 said:
Just to make sure, did Intel "cheap" out by using the paste instead of the solder or are they roughly the same? And if they did cheap out, isn't that good for us as lower prices? Well, besides overclockers, I mean.

(I will admit, I was looking forward to 5.2 Ghz on air-cooled Ivy like everyone else.)
Click to expand...

This is in my estimation what happened.

First, 22nm is harder to cool because of thermal density. Even though it generates fewer watts of heat to dissipate, those watts are more concentrated in a smaller package.

Second, Intel "cheaped out" going with thermal paste instead of a much more conductive solder. Meaning that the heat takes longer to transfer out of the cores to the IHS to be whisked away by your HSF/water/LN2/whatever.

So in essence although there is less heat to dump it takes longer to actually get rid of it and therefore the cpu runs hotter than the previous generation.

EDIT: And this bodes poorly for future chips, if they continue to use thermal paste instead of solder, we're likely to continue to see these higher temps going forward. As it was likely done for cost reasons - and works just fine for stock performance - they're unlikely to revert back to solder. Except perhaps in the '-E' series of chips (IB-E, Haswell-E) where they expect higher heat generation to begin with.
 
Last edited:
I expected the IB being 22nm it would run cooler. I guess I was wrong again because I also expected bulldozer 32nm to run cooler than phenom II. 🙁
 
Meh, it just requires better cooling, at least it responds to it unlike SB which is a dog with chilled+.

It's also reportedly able to take voltage better than SB.

Perhaps SB was boring, though it's "better" for the majority of the minority users, IB destroys SB for the small minority of the minority users.

Proof of concept success, fab experience gained, win overall for intel.
 
nemesismk2 said:
I expected the IB being 22nm it would run cooler. I guess I was wrong again because I also expected bulldozer 32nm to run cooler than phenom II. 🙁
Click to expand...

IB does run cooler just not overclocked.

Many have stated this run of 22nm was all about lower power. By the time haswell is out you will really see how far you can push the chips at 22nm.

And amd never learned from the Pentium 4, I don't believe intel will make that mistake again.
 
Does anyone here think IB-e was delayed because of this? If so maybe they will go back to planar designs for their enthusiast SKU's?
 
SHAQ said:
Does anyone here think IB-e was delayed because of this? If so maybe they will go back to planar designs for their enthusiast SKU's?
Click to expand...

Intel only makes 22nm with trigates. So any planar design would have to be 32nm or higher.

Also...who says IB-E is delayed. Or it even gets released. Or was it just an assumption that there had to be an IB-E?
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top