Is IntEl's 22nm and ivy bridge a colossal failure?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Pure idiocy.

The IVB chips draw less power at a given frequency, and therefore make less heat. If the sensors say they run warmer, it's because they're not running the same coolers, they have a smaller surface area to conduct heat away from, or have the temperature sensor in a different part of the die.

Less watts is less watts.
 

Arzachel

Senior member
Apr 7, 2011
903
76
91
Pure idiocy.

The IVB chips draw less power at a given frequency, and therefore make less heat. If the sensors say they run warmer, it's because they're not running the same coolers, they have a smaller surface area to conduct heat away from, or have the temperature sensor in a different part of the die.

Less watts is less watts.

Ding ding ding!

Intel's 22nm process seems to be fine, it's a design "issue" if you could call it that.
 

Boulard83

Member
Apr 13, 2012
82
0
0
Pure idiocy.

The IVB chips draw less power at a given frequency, and therefore make less heat. If the sensors say they run warmer, it's because they're not running the same coolers, they have a smaller surface area to conduct heat away from, or have the temperature sensor in a different part of the die.

Less watts is less watts.

Your right about the smaller area to dicipate the heat.

But about the watts, stock VS stock your right. but when OCing, these new Ivy seems to draw more current. A 1,3v Ivy can draw more watts than a 1,3v SB since the law apply to the Amp and the V its drawing. Ivy need more Amp to stay stable at a X clock with Y volts when OCed. The point seems to be ~1.3v and 4.6-4.8ghz with non-exotic cooling.
 

Absolute0

Senior member
Nov 9, 2005
714
21
81
Mighta been mentioned somewhere before but power consumption increases with temperature. So OC'd IVB definitely not looking like the power savings everyone was expecting.

"Idontcare" did substantial testing to demonstrate this.*


Forgot to weigh in on this thread topic

Definitely not a failure. IVB brings a significant IPC increase and a few additions like the integrated graphics. Really all-in-all it's an amazing processor! Sandy Bridge has us spoiled... So a little overclock headroom is gone and that sucks, I mean I wanted to do 5.2 Ghz just like everyone else! It's a failure to the .01% of users who constantly upgrade and overclock so that they can have cutting-edge performance. But to Everyone else IVB is great.
 
Last edited:

nenforcer

Golden Member
Aug 26, 2008
1,775
14
81
77w @ 160mm^2 = 0.48w/mm^2 (Ivybridge)
95w @ 216mm^2 = 0.44w/mm^2 (Sandybridge)

That's where thermal conductivity becomes important.

And in reality it's probably not evenly distributed like that and more concentrated near the cache and cores while the edges are cooler. With the new first time 3D TriGate there could very well be some bad hotspots in IB that didn't exist in SB.

I'm still curious about the GPU improvements in the HD 4000 and if the GPU is also using these new transistors or if its tacked on using the old process. Mixing / matching isn't usually a good design decision.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Even with those issues, I still think it's far from a failure and it will be a great thing for laptops. IGP performance will rise by a significant amount, CPU performance will get a small boost, and power consumption will go down a bit.
 

GammaLaser

Member
May 31, 2011
173
0
0
And in reality it's probably not evenly distributed like that and more concentrated near the cache and cores while the edges are cooler. With the new first time 3D TriGate there could very well be some bad hotspots in IB that didn't exist in SB.

I'm still curious about the GPU improvements in the HD 4000 and if the GPU is also using these new transistors or if its tacked on using the old process. Mixing / matching isn't usually a good design decision.

Yes, I bet the heat density delta would be much greater if only the cores/cache area are considered without the IGP, especially as the IGP-to-Core area ratio increased in IVB.
 

Dufus

Senior member
Sep 20, 2010
675
119
101
Pure idiocy.

The IVB chips draw less power at a given frequency, and therefore make less heat. If the sensors say they run warmer, it's because they're not running the same coolers, they have a smaller surface area to conduct heat away from, or have the temperature sensor in a different part of the die.

Less watts is less watts.

Please show link(s) to substantiate your claim that Ivy uses less power, particularly for frequencies above 5GHz. So far all I have found is this. It would be nice to actually see some results also showing power consumption for a particular benchmark with the higher clocks or even running linpack c/w AVX at highest stock frequency.
 

amenx

Diamond Member
Dec 17, 2004
4,418
2,740
136
IB looks like a 'colossal failure' only to overclockers who strive past the 4.5ghz mark (and even then that is not absolutely ascertained). For modest OC'ers and for non-OC'ers (probably 99% of IBs intended market) - and who are not coming from SB - it looks like a huge win. Certainly more attractive than what else is out there (BD or any AMD offerings).
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Pure idiocy.

The IVB chips draw less power at a given frequency, and therefore make less heat. If the sensors say they run warmer, it's because they're not running the same coolers, they have a smaller surface area to conduct heat away from, or have the temperature sensor in a different part of the die.

Less watts is less watts.

This analysis is quite wrong. When you're packing more transistors in a smaller space, with greater leakage, there will be more heat.
 

Concillian

Diamond Member
May 26, 2004
3,751
8
81
Guys, the link in the original post... the first graph that comes up when you click that link is power consumption for various CPUs at different frequencies.
This first graph clearly shows IVB at it's max stable OC (4.8 GHz) drawing less than SB @ 5.2 GHz. Both are at their max stable overclocks.
Then the very next graph shows IVB at 20C higher temperature.

EDIT: TT appears to have changed their graph to remove the SB 2600k @ overclocked speed, when it was there before, it showed higher power consumption at significantly lower temperature. I'm not sure why they removed it, but it used to be there. You'll note that the link to the picture in the OP appears to be broken, this is because they changed the picture to remove the 2600k @ overclocked. Something fishy is going on at TT.

IVB is getting to a significantly higher temperature while producing less heat.

Read that last sentence again and again, and you will come to the same conclusion I did. The cause of the high temperature cannot possibly be leaky gates. Leaky gates waste electrical energy as heat... power consumption would be high, but it is not high.


There are three possible explanations for this:

1) thermal gradient is worse. Material between the heatsink and the source of the heat is less conductive than the SB design. This is possible if the fin-fet transistors are somehow deeper into the filmstack of the material on top is less conductive than the SB design. I don't pretend to have any clue as to the design here, but fin-fet is a major departure from conventional processes, and they may have had to alter materials throughout the stack to pull it off.

In layman's terms, this potential cause is much like a person in a cool house in shorts and a t-shirt vs. the same person in a bed with a pile of sheets and blankets covering them. The body is putting out the same amount of heat energy, the air (heatsink) is the same temperature, but in one case the body is cold and in the other it is comfortably warm.

2) Density is significantly higher. This is also possible because we know:
- The process is smaller
- The IGP has been given a higher percentage of total die area than on SB
- The die is smaller than SB (according to Wikipedia 216mm2 vs ~160mm2 fo IVB)
These could all result in the heat being dispersed over less area.

The obvious example here is 10 people spread comfortably throughout a 2500 square foot house vs. 10 people in a closet. 10 people put out a reasonably similar amount of heat in either case, but 10 people in a closet will get warmer than 10 people spread throughout the entire house.

3) It's possible that the thermal sensor has been put closer to the source or has been redesigned to be more sensitive than the sensor in SB.
- I find this one not so likely, but it's technically a possible reason. If it's closer to the source, then the same thermal gradient would produce higher temperatures.

In any of these cases, the available data does NOT suggest that IVB is leaky. Leaky would shoot power consumption and heat up. Power consumption is not high, only temperature.

This suggests there is an impediment to heat being removed rather than IVB being a major source of heat energy like some other CPU designs (Bulldozer, netburst P4) which DID have major power consumption to go along with their higher temperatures.
 
Last edited:

gevorg

Diamond Member
Nov 3, 2004
5,070
1
0
Some comments here are giving me flashbacks about speculation right before Sandy Bridge release. Folks, ignore TweakTown, and wait for a reputable source to review IVB like Anandtech or Xbitlabs.

Here is excerpt from Anand's IVB preview, stock clocks, but a very good starting point.

Under load however the power savings are significant. The Core i7 3770K pulls 27 fewer watts while delivering better performance than the 2600K.
http://www.anandtech.com/show/5626/ivy-bridge-preview-core-i7-3770k/10
 

Reikon

Senior member
May 25, 2003
693
0
0
Some comments here are giving me flashbacks about speculation right before Sandy Bridge release. Folks, ignore TweakTown, and wait for a reputable source to review IVB like Anandtech or Xbitlabs.

Here is excerpt from Anand's IVB preview, stock clocks, but a very good starting point.


http://www.anandtech.com/show/5626/ivy-bridge-preview-core-i7-3770k/10

Since when is Tweaktown not a reputable source?

And nobody is arguing that Ivy Bridge uses less power at stock, but the problem is heat when scaling up, which you can't infer from stock performance.
 

Gikaseixas

Platinum Member
Jul 1, 2004
2,836
218
106
Not a colossal failure at all, that would be BD
Could the extra heat come from the more powerful HD4000 graphics chip?
 

Dufus

Senior member
Sep 20, 2010
675
119
101
1) thermal gradient is worse.
This has already been shown very likely by the higher Tjmax and possibly lower TDP and given that if Tcase max will be similar to SNB. IVB packaging has shown 95W while it seems the cpu register is 77W, which is right?

IMO it's not just that thermal resistance is higher between core and IHS producing higher temps but whether it is compounded by poor scaling with higher OC. Where's all the RAPL and Icc measurements?

I'm not a fan of power of the wall measurements but looking at that chart the difference between the 2600k at idle and load is 200W and for the 3770k 212W, would have been nicer if it were the same board used.

 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
Since when is Tweaktown not a reputable source?

1. No 2600K gets that high of a score in Cinebench R11.5.

http://www.tweaktown.com/articles/4...e_i7_3770k_and_core_i5_3570k_cpus/index6.html

Other sites 2600K at stock gets 6.8-6.9, here it gets more than 7.5.

2. iGPU scores are way off. HD 3000 is getting 25% higher in 3DMark Vantage while Far Cry 2 is getting almost 70% greater: http://www.lostcircuits.com/mambo//...ask=view&id=104&Itemid=1&limit=1&limitstart=6
http://en.expreview.com/2012/02/19/ivy-bridge-core-i5-3570k-engineering-sample-test/21214.html

3. Idle power results show great difference between Sandy Bridge platforms(whether -E or non E), while other sites don't show that.

http://www.anandtech.com/show/5626/ivy-bridge-preview-core-i7-3770k/10
http://www.tomshardware.com/reviews/core-i7-3960x-x79-sandy-bridge-e,3071-19.html
http://techreport.com/articles.x/21987/15
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
After disabling the IGP on my 3570K load temps dropped 4C

That's very strange, especially if you are using discrete graphics. Not using the iGPU should power gate it off completely. That's not a question off whether there's leakage in idle or not, its off.
 

Don Karnage

Platinum Member
Oct 11, 2011
2,865
0
0
That's very strange, especially if you are using discrete graphics. Not using the iGPU should power gate it off completely. That's not a question off whether there's leakage in idle or not, its off.

Thought the same thing. Reason i disabled it is because the lucid mvp allows you to use discreet graphics and the igp in tandem and having the IGP enabled was throwing off the temps.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Is intel's entire 22nm line expected to exhibit the same problems? intel might have a real problem here, there is absolutely no compeling reason to upgrade. 10% max performance increase, far worse overclocking, they run much hotter, and intel will want to charge a premium for a worse product. People that have used the thing are warning users to be very careful with voltage adjustment also. Apple skipped it and rumors are that they're going with Trinity, it makes sense now that numbers are starting to surface.
 

Puppies04

Diamond Member
Apr 25, 2011
5,909
17
76
Is intel's entire 22nm line expected to exhibit the same problems? intel might have a real problem here, there is absolutely no compeling reason to upgrade. 10% max performance increase, far worse overclocking, they run much hotter, and intel will want to charge a premium for a worse product. People that have used the thing are warning users to be very careful with voltage adjustment also. Apple skipped it and rumors are that they're going with Trinity, it makes sense now that numbers are starting to surface.

Oh noes, a few thousand "enthusiasts" are going to skip this generation of chip!!!111!!11

Intel might aswell shut its fabs and call it a day.


* As already stated numerous times in this and many other threads I am waiting for IBs actual release and a proper review before I make any judgements.

** and to all the people claiming IB is a failure because their SB chip is stable at 5ghz @ 1.3v please understand one of 2 things happened....

1. You purposely sought out a golden chip
2. You got very lucky

I would guess that 99% of 2500k/2600k/2700K will not hit that performance point, comparing the few IB chips that are out in the wild to the top 1% of SB chips is stupid and counter productive..