Why is Ivy Bridge HOT, power hunger and bad clocker? - immatured 22nm process ...

I bet many people will be wondering why Ivy Bridge is not the 77W it's supposed to be, but 95W instead. The retail CPUs clearly show 77W in CPU-Z, but the box states 95W.

Maybe you know, Ivy Bridge are around the corner. Maybe you knows too, they are worst clocker then Sandy Bridge. We know why ...

As you know, Intel is with the 22nm production late. Production is not good, and there are big problems with the chips. The original "on paper" concept of 22nm chip with Tri-gate transistors is extremely low supply voltage. But, with current revisions Intel can not keep voltage in planned values. This is a problem. The chips have a higher voltage than planned, broadly comparable with Sandy Bridge. And that's wrong.

Tri-gate transistor needs to switch to a lower voltage. But for a correct recognition of the I/O status needs more current than planar transistor. Three-gate area is greater than one-gate and the current is several times higher than in Sandy Bridge chips. When Intel reach a planned low voltage, everything will be fine. Lower voltage means acceptable currents, less leakage and a great consumption. Unfortunately, it does not meet the current "E1" revision.

Current 22nm chips have high voltage, higher than they should have. The values are similar to Sandy Bridge chips. Properly should be the default voltage below 1V and it is not now. But Ivy Bridge needs a lower voltage, at the same voltage as Sandy Bridge consumption and temperature is significantly higher due to higher currents in the chip.

Basic Ivy Bridge idle voltage is above 1V, higher than Sandy Bridge. The load voltage is lower than that of Sandy Bridge and consumption is lower, but temperatures are higher. If the Ivy Bridge voltage increases, consumption and temperatures extremely jumps up. This problem can be solved only by improving the production, so maybe its time for another revision. Indeed it may be a potential problem in laptops with the highest third-generation Core i7 models.

In the desktop this problem occur with less overclocking than Sandy Bridge and significantly higher power consumption and temperatures. If you have a nice 5GHz + Sandy Bridge, keep it for now. Ivy Bridge ends with overclocking on the air somewhere around 4.6 to 4.7 GHz. But slightly lower overclocking then Sandies compensates higher performance per clock, so it is not a major problem.


10

Belgium Massman says:

Power consumption is going to be very interesting on Ivy Bridge. I wonder what the reviews will say about this. Remember what OBR/RBOX2 posted on the forums a while back?




http://hwbot.org/forum/showthread.php?t=44030

Norway knopflerbruce says:

A similar graph for SB would be great.

K404 says:

The graph only looks skewed because at 4.8GHz the chip starts to fall over.

My SB is the same, more or less. Smooth up to around 4.9-5GHz then each 100MHz needs a disproportionate voltage increase.....same as ALL CPUs as they head towards their limit.

The test sample is not great. That's all that can be said from this graph.



United Kingdom borandi says:

Interesting set of results, but that graph is more like a series of varied points than anything consistent. It is combining certain things and making assumptions that aren't exactly clear at all. An increase in voltage and frequency increases power consumption and temperature?!? Whodathunk it. It wouldn't pass peer review in a research journal, that's for sure. Be ready for some results soon :)

United States sin0822 says:

from my testing at stock the CPU can easily fit into a 77W TDP at stock. Perhaps OCed over its turbo it starts to eat power, otherwise id still classify it as 77W.

However I don't measure system power consumption only that of the CPU with a 1-3% margin of error, i could measure total and subtract my CPU. for more accurate results.

so a 130W difference in 1ghz of your testing OBR that isn't that bad(it is bad, but not Sbe Bad), especially with such a high vcore, haha looks like my E0 CPU right there, but I know you used an E1. thanks for your findings however more data is good!

IMO SB at 1.5v+ is a monster CPU heat provider as well.

IMO Ivy is a loop of heat and hate, if you get the CPU warmer it demands more vcore, you give it more vcore it gets hotter, until you go and the CPU does around 1.75v and dies(that is if your CPU does die). ONly way around this is to cool the CPU down, either through better ambient cooling IMO a good air cooler, as Ivy isn't about total heat capacity it is about thermal conductivity. IDK i tested leakage scaling with temp, the results are very interesting once you cool the CPu down to a certain temperature, the power consumption is greatly reduced.

You think you should have your results borandi published in a journal? Watch out guys andatech reviewer is going to publish is an article lol. :D

Czech Republic OBR says:

after nda you will see better version of graph, sandy + ivy + e0 ivy in one graph and its is very interesting comparision ...

United States sin0822 says:

yea I believe you, i want to see what boradi will give us tho.

Denmark M.Beier says:

sin0822 said: yea I believe you, i want to see what boradi will give us tho.

Borandi ?

This is a nice change, you will actually read his opinion and recognize it? I saw a few scenarios of which one thought "get a room".

Please do not take offence OBR... But the source that I am really looking forward to reading in about 5 days, is Lab501, their approach covers everything I would like to know + alot of useless stuff /Useless to me, but not many others, and after all, though I dont like to admit it, there are other people too :p:

FlanK3r says:

OBR said: after nda you will see better version of graph, sandy + ivy + e0 ivy in one graph and its is very interesting comparision ...


there is big diference between IB E1 and E0 with the same voltage? Example at 4.6 GHz?

United States sin0822 says:

M.Beier said: Borandi ?

This is a nice change, you will actually read his opinion and recognize it? I saw a few scenarios of which one thought "get a room".

Please do not take offence OBR... But the source that I am really looking forward to reading in about 5 days, is Lab501, their approach covers everything I would like to know + alot of useless stuff /Useless to me, but not many others, and after all, though I dont like to admit it, there are other people too :p:


i will read it sure.

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)