Mobile Ivy Bridge Power Consumption Disappointment

So, I read a few reviews of the new mobile Ivy Bridge quad-cores. The overall consensus seems to be that there is not a significant improvement in power consumption compared to Sandy Bridge chips. Intel publicized charts showing that at lower voltages, such as idle voltages, 22nm "3D" transistors could consume up to 50% less power. Notebookcheck.net is reporting that at idle, Ivy is clocked at 1200MHz as opposed to the 800MHz idle of Sandy Bridge chips. This seems to indicate that they are running at the same voltage at idle. The i7-2760QM consumes 2W less than the i7-3610QM at idle. The notebookcheck reviewer did mention that they were using pre-production CPUs which could be the cause, but Intel would most likely try to show off every new feature of the new chips at launch. Plus, none of the other sites using the Intel and Asus provided test laptop were reporting any significant gains either. This is a huge disappointment because the integrated graphics improvement would not be a huge selling point to a large portion of the people in the market for an Intel quad-core laptop.

Does anyone know what the idle clocks are supposed to be for mobile Ivy Bridge quad-cores? Does anyone expect a fix to this issue? Thanks.

 

Idle power consumption is identical to Sandy Bridge -- I would guess that they've already gated everything that can be gated and the laptops that have been reviewed so far are not running at a lower voltage.

If these were the CPUs Intel was trying to show off, they would not be pre-production. They come from some manufacturer that sent the reviewer a laptop with a chip Intel gave them for testing. My workplace gets a lot of these test chips (albeit for servers, not for laptops) and they are never the same as the final version because first, there are bugs that are fixed and second, Intel doesn't want everyone to know the performance ahead of time.

Also, the laptop reviews so far have not been very convincing. They don't compare CPUs to CPUs, just laptops to laptops. The CPU is not the only part of a laptop that draws power, sometimes it's not even the dominant part. This is why I have been looking at desktop reviews -- they actually got a retail chip and they are much more professional.

 

Well, they really can't go much lower at idle - the idle cores have been almost completely shut down since Core 2 Duo, with i7 being much better at this. It's power consumption (and heat) under load that should be lower - also allowing for higher overclocks...

 

Trying to gather in as much info as possible. Can anyone explain what this quote means in comparison to SNB:

Confused on why IVB would "give up" a couple MHz on the OC clock speed.

 

Ivy doesn't like being overvolted. Take a look at the bottom of this page -- the load power consumption nearly doubles when the voltage is increased to 1.35V.

 

Thanks. So, the same may also be true with IVB-M.

How big of a problem is this? Do people usually stay away from increasing CPU voltages within lappies anyway?

 

It is not a problem for laptops at all. Ivy is oriented towards the lower voltages found in mobile devices. Even on desktops, when you run at stock speeds and voltages, the power consumption is substantially lower. Laptops should be even better off, but we haven't seen a review of a retail one yet so it's hard to be sure. The problem is that when you take this low-voltage oriented architecture and crank up the voltage to overclock, it doesn't work very well.

 

I was talking about the mobile 3920xm - technically, it should overclock better than the 2920xm at the same voltage. You can't change the voltage on either, but I'm assuming the highest voltage (set by the processor's voltage regulator) is the same for both chips...

 

Could it only be QS chips that are out for testing aren't fully matured or what? Intel sent us QS desktop Ivy Bridge chips last week for our display systems to show off for customers.

 

Not surprised. What they claim pre launch > what they claim at launch > real world performance has alot of the results filtered to much less then the original claim. Happens all the time and I wouldnt be surprised when the ivy chips ship that it might get a bit less then what they are claiming now.

 

Intel are master marketers, but I wonder how long it will be before more people start to wise up to their antics. It's the same crap with every platform, lots and lots of promises/features/improvements, and they rarely deliver.

 

Nah, Sandy Bridge was pretty good. Intel usually has a pretty good track record, which is why people were disappointed when 22nm wasn't what it was cracked up to be.

 

Well, if not delivering means beating AMD for years upon years in a row in not only features but also performance... which is really their only competition.

Intel needs to start under-promising and over-delivering. IVB is a great evolutionary jump IMO. It has performance improvements all around. Past a certain voltage plane they may get exponentially hotter, but that only applies to people who overvolt their CPU, which is NOT something you want to do... especially in a notebook. Heat may kill things in the long run, but overvolting will do real damage if you don't know what you're doing.

All I'm saying is that people are whining and complaining about nothing, it seems. Yes, IVB is a smaller die on a smaller process. Yes, they're using a whole new transistor design. But if I also remember correctly, people whined and moaned about Arrandale and Clarksfield when they came out too... "too hot"... "not enough of a performance bump"... but then they got their hands on final shipping products which were pretty awesome and performed just fine. There's enough performance in Arrandale to satisfy 75% of the population. Those that are whining now haven't even used IVB yet...

 

Notebookcheck.net did do a really good job compared to most sites in their test setup, using a barebone laptop to swap out multiple CPUs. This reduced outside variances from other components. But like you are saying, they were pre-production parts, not retail. The review may not be a full indication of what to expect.

For the IB laptops that Intel and Asus sent out, their battery life not being spectacular seems to be the flip-side of the mobile Sandy Bridge launch. The test laptops sent out for review of mobile Sandy Bridge had a better performance/watt-hour rating than any production quad-core sandy bridge laptop reviewed since, according to Anandtech.

So far, 22nm FinFETs are not looking nearly as good as Intel pumped them up to be.

 

^it is the FIRST line of chips made on the process. just how good do you expect it to be? so good that theres no room for improvement in the future? what would that even look like?

 

Judging all of Ivy Bridge by the 3720QM is a mistake.

It has no adjustable TDP, and it's the first part to be released.

I have yet to see an apples-to-apples review, which would be hard to do because there is no 2.6GHz sandy bridge part to compare with, but the reviews out there don't even attempt reasonable comparison.

 

What do you mean apples to apples? Notebookcheck did a pretty thorough review and the i7-3610qm (2.3-3.3ghz) performs around 10-15% faster and uses like 5% less power.

Adjustable TDP has nothing to do with it, it's a limitation of Intel's current 22nm process. Maybe they'll make it significantly more efficient later but right now it's mediocre at best.

 

Yes, compared it with what? A 2.4GHz part. It is not a clock-for-clock comparison like we can get in good reviews of the desktop chips.

My point about adjustable TDP is that is the "killer feature" for Ivy Bridge re: power consumption, and we have yet to see it.

 

That's simple. I expect it to be better than Sandy Bridge. I'm not expecting leaps and bounds of improved power consumption. I'm just expecting a small but noticeable improvement. If the chip causes the entire notebook to consume 10% more power than it's predecessor at idle, that is a step in the wrong direction. If they moved backwards this time, then there is a lot of room for improvement with Haswell on this new process.

IMO, Intel is also leaving room for AMD to move further in on the ultraportable space. If Intel's quad-cores consume more power than their predecessors, the dual-cores likely will as well. AMD will have a better IGP and they are claiming their new APU's double the performance per watt of their predecessor. That would make a formidable opponent to Intel's Low TDP mobile chips. AMD would offer better battery life, graphics, and prices, while offering enough CPU performance for the average consumer.

If Ivy Bridge could clock down to 800 MHz like Sandy Bridge, it would be a significant improvement. The question is whether or not such a low voltage would meet the threshold voltage requirements of the transistors.

Variable TDP only affects the top-end power consumption. It would have no affect on how much power you consume while web browsing or watching a movie. It would only affect the maximum power the CPU could draw. Most people don't reach that limit for significant periods of time. Thus, it would offer very limited gains in battery life. This is my understanding anyway, if you know something different, please share.

 

@HAL; I was only saying that Intel makes a lot of promises that they can't seem to deliver on. Just marketing fluff, but vastly different in the real world. I have come to expect it from Intel any more.
We'll just have to see in the coming weeks what IVB will be like in the real world, apart from the limited tests and reviews out now.

 

Like I said before... look at Arrandale and Clarksfield. It's not supossed to be significantly more efficient right now. It's a brand new process with a brand new transistor design. Mastering a process takes time. Sandy Bridge was a prime example of how to master a process technology...

And it is... just not by leaps and bounds.

Who's to say it can't? We've seen exactly two CPU's based on IVB for mobile. The 3720QM and the 3610QM. Wait until there's more to go on before making assessments of the entire fab process and it's limits.

 

@crazy1, considering the definition of TDP, I think that can be considered obvious. It's like saying a car has a variable redline but it only affects people who race. I see you are more concerned about idle power consumption.

I am waiting to see what the low voltage parts have to offer before rendering judgement on Ivy Bridge, since they are the ones with adjustable TDP (that and the Extreme parts, but those are less interesting since they are power hogs and IVB overclocks badly on air). It should make for some interesting ultrabooks and ultraportables using SKUs with larger cache, hyperthreading, etc. than would normally get stuffed in there.

 

What about Arrandale and Clarksfield? They were both significantly faster than their predecessors.

I understand that new processes need time to mature but if you look at Intel's pre-release slides and PR material it was hyped up to be a lot better than what we have now.

They compared it to a i7-2670qm. Which is like clocked like 100-200mhz lower.

I'm sure adjustable TDP is great but what does that have to do with what we are talking about?

 

It depends. If the topic is complaining about preproduction BIOSes managing ES CPUs, not much. If the topic is what IVB is potentially good at wrt power and efficiency, see my previous post.

 

Yes, because that was a completely new architecture. Core 2 to Core-i. This isn't that radical of a shift.

You're still basing opinions on a minuscule bit of information... What ALLurGroceries is saying is 100% correct. BIOS features are still being revised and finalized, and there's really nothing out there to form solid opinions on yet.

 

The post was about disappointing power consumption numbers at launch. I tried to make it clear in the original post that there is limited reliable data. The data available is not very promising, though. I don't expect to see another objective comparison of mobile Ivy Bridge parts like notebookcheck's review, until the dual-core launch in over a month. From what I've read so far, it seems like that will be the time to make some final judgements on mobile Ivy Bridge parts.

 

Yeah maybe they retest it when the BIOS of Clevo works 100% with Ivy and when they got production Samples.
But if not, when Intel refresh their chips, then they are going to do another test.