Anandtech Review of Ivy Bridge is Up now! Other sites added

[blckgrffn]

In laments terms this is "Cheaper" then using intel's specialized solder. This is the reason why IB temps are higher. By using TIM, they have effectively turned the "heatspreader" into a heat barrier.

Lots of info over here

http://www.overclockers.com/forums/showthread.php?t=705620&page=9

That's hilarious! :thumbsup: Intel!

Jeeze, save a few bucks on each CPU and punish us (here) all... Too bad they didn't go with the premium stuff with at least the K CPUs where it makes sense too. I suppose that would have really been asking a lot, given how tight their production process is.

Perhaps they'll spend the coin on IVB-E, which would make it pretty phenomenal chip if the thread @ overclockers is true.

It looks like BFG was right to insist on an extra factor changing how things were working vs SB.

 

[Nemesis 1]

When People compare IVB to Prescott...then you know the thread has left the realm of reality and gone either viral...or very stupid.

Leave it alone man . Enjoy the ignorance and let them rave. Its all they have for stones and there only pebbles that won't even crack a window grains of sand is all they be. Dying from AMD performance thrist.

 

[BFG10K]

They've put up a full article about the TIM issue, very interesting read: http://www.overclockers.com/ivy-bridge-temperatures

Here are some snippets:

Intel is using TIM paste between the Integrated Heat Spreader (IHS) and the CPU die on Ivy Bridge chips, instead of fluxless solder.

Most importantly here, if Intel is using TIM paste between the IHS and CPU die, the IHS effectively becomes a heat barrier rather than a heat spreader.

 

[dma0991]

Must be Intel's way of gimping IB to make Haswell look better. But considering some IB chips with high end watercooling are voltage limited rather than temp limited, there may not even be an improvement if Intel decides to use fluxless solder instead.

 

[oceanside]

considering some IB chips with high end watercooling are voltage limited rather than temp limited, there may not even be an improvement if Intel decides to use fluxless solder instead.

Perhaps the reasoning behind the move to TIM. Still, I hope we get some adventuresome delidders running direct mounts in the next couple of weeks to share some data. If the benefits are tangible, I'll be pretty disappointed Intel chose this route. That paste imprint on the IHS in the article's picture looks pretty bad.

 

[Absolute0]

Leave it alone man . Enjoy the ignorance and let them rave. Its all they have for stones and there only pebbles that won't even crack a window grains of sand is all they be. Dying from AMD performance thrist.

Whoa man, I don't know what you're about. I saw a die shrink that was accompanied by a heat increase, I was reminded of Prescott. I was surprised that everyone acted as if this phenomenon hadn't occurred before.

"dying of AMD performance thirst?" ... I don't own an AMD.

Sometimes this forum gets so polarizing. I'm not trying to take a side.

 

[Destiny]

Whoa man, I don't know what you're about. I saw a die shrink that was accompanied by a heat increase, I was reminded of Prescott. I was surprised that everyone acted as if this phenomenon hadn't occurred before.

"dying of AMD performance thirst?" ... I don't own an AMD.

Sometimes this forum gets so polarizing. I'm not trying to take a side.

I owned a Prescott before building my AMD rig last year...with the Prescott I kept getting BSOD or non-boots because of heat issues even with mild overclocks....

I was thinking about scrapping the prescott for gold - but I might keep it for old times sake...

 

[PreferLinux]

Whoa man, I don't know what you're about. I saw a die shrink that was accompanied by a heat increase, I was reminded of Prescott. I was surprised that everyone acted as if this phenomenon hadn't occurred before.

"dying of AMD performance thirst?" ... I don't own an AMD.

Sometimes this forum gets so polarizing. I'm not trying to take a side.

I will just point out that Prescott was not just a die shrink, but also increased the pipeline depth significantly.

 

[Lonbjerg]

I will just point out that Prescott was not just a die shrink, but also increased the pipeline depth significantly.

Yup:
http://en.wikipedia.org/wiki/Pentium_4#Prescott

Why the comparsion is even more stupid...

 

[AtenRa]

They've put up a full article about the TIM issue, very interesting read: http://www.overclockers.com/ivy-bridge-temperatures

Here are some snippets:

Was that CPU a retail or an engineering sample ??

 

[Quantos]

Wow, interesting read. I didn't know TIM was that bad compared to solder joint. It's a completely different league, it seems. D:

I wonder how much it costs Intel to use one instead of the other. Definitely the 22nm process / 3dgate technology is probably more expensive to manufacture, so I guess they had to cut corners somewhere. The thing is that it also makes virtually no difference between a certain voltage. Who cares if the chip runs at 40c or 50c at stock clocks and voltages?

 

[Alaa]

Anand should investigate this TIM stuff.

 

[dma0991]

Perhaps the reasoning behind the move to TIM. Still, I hope we get some adventuresome delidders running direct mounts in the next couple of weeks to share some data. If the benefits are tangible, I'll be pretty disappointed Intel chose this route. That paste imprint on the IHS in the article's picture looks pretty bad.

I'm still waiting for the person who delidded his IB processor to post some results without any IHS or the TIM that Intel applied. Wonder whats taking him so long.

 

[ShintaiDK]

Or it could be because 77W/160mm2 = 0.48W/mm2 and 95W/216mm2 = 0.44W/mm2.

Not to mention the cores themselves in Ivy is alot smaller compared to the area of the chip with Sandy. A direct shrink would have made Ivy 125mm2 or so. Or 77W/125mm2 = 0.62W/mm2.

I dont think the TIM matters...

 

[Quantos]

Or it could be because 77W/160mm2 = 0.48W/mm2 and 95W/216mm2 = 0.44W/mm2.

Not to mention the cores themselves in Ivy is alot smaller compared to the area of the chip with Sandy. A direct shrink would have made Ivy 125mm2 or so. Or 77W/125mm2 = 0.62W/mm2.

I dont think the TIM matters...

Apparently it matters. A lot. The thermal conductivity of TIM compared to a solder joint is really, really low.

Have a look at BFG10K's link.

 

[ShintaiDK]

Apparently it matters. A lot. The thermal conductivity of TIM compared to a solder joint is really, really low.

Well, we need to see a Sandy and Ivy without IHS compared to one of each that does. Then we can argue about if the TIM compound is an issue or not. It would be silly to go bananas over the TIM compund, if the same results was still valid without IHS and TIM to interfere.

Until then its pure speculation.

 

[Quantos]

Well, we need to see a Sandy and Ivy without IHS compared to one of each that does. Then we can argue about if the TIM compound is an issue or not. It would be silly to go bananas over the TIM compund, if the same results was still valid without IHS and TIM to interfere.

Until then its pure speculation.

Agreed. I'm not sure that can be done, though. The solder joint is patented by Intel, and I would assume this means it's not necessarily easy to acquire and/or install.

 

[podspi]

I am sure Intel had a good reason to do what they did. It doesn't make sense for them to 'gimp' anything.

Less impressive refresh = less sales from people who already own Intel products.

I do agree, though. Once they are out in number, here is hoping Anand takes a look into this (even if a few IB must be sacrificed for the cause...)

 

[dma0991]

Agreed. I'm not sure that can be done, though. The solder joint is patented by Intel, and I would assume this means it's not necessarily easy to acquire and/or install.

I don't see the need to acquire the solder joint. If the person could do a run without the IHS and it showed a tremendous difference in temp reduction, we've already proved that the TIM is part of the problem.

 

[Quantos]

Well my understanding is that IB was never meant to convince SB owners to switch. Basically, the biggest advantage IB has over SB is lower power consumption, which doesn't mean much to overclockers. Thus, IB's main target was anything that could benefit from lower power consumption. With no intended OC, the TIM is plenty to keep temperatures low enough. Considering it's cheaper (probably), the margin is probably what allows Intel to sell IB at the same price, or cheaper than SB.

 

[Lonbjerg]

You know we are having good hardware times, when the main grief agianst a CPU boils down to overclocking...make me smile....luxury problems in the developed world indeed.

And a useless "metric" to evalute a CPU on...when som many ovther factors are more important.

 

[Absolute0]

Yup:
http://en.wikipedia.org/wiki/Pentium_4#Prescott

Why the comparsion is even more stupid...

Yes I am quite aware of this. My point was the fact that 130nm --> 90nm got hotter. 32nm to 22nm got hotter. One person said "this has never happened before" so I pointed out Prescott.

I didn't know the "P" word would immediately get me flamed for being a performance thirsting AMD fanboy and making stupid comparisons! LOL... IVB is not a Prescott that is for sure, but it does run hotter than the chip it replaces, so it has that in common.

Now, to go into even more detail, I was further reminded of Prescott because of a couple other things, one being extreme overclocking. Prescott couldn't outclock the P4 Northwood C on air (usually do to thermals) but the extreme overclockers with the liquid nitrogen were able to take it to higher frequencies than Northwood and topple WR benchmarks with it. Which reminded me of IB compared to SB, not gaining much on air, but in the world of Xtremely cold overclocking its an improvement.


I think it's funny how I make the connection that Prescott was also a process shrink that ended up hotter to inform someone who said this hasn't happened, and saying this is enough to get me flamed as an "AMD performance thirsting" fanboy who is "making stupid comparisons." This forum is so caught up in the idea of fanboyism and having to battle one side versus the other. It's not a good thing.

 

[LBXAC20]

You know we are having good hardware times, when the main grief agianst a CPU boils down to overclocking...make me smile....luxury problems in the developed world indeed.
And a useless "metric" to evalute a CPU on...when som many ovther factors are more important.

This metric you are referring to can't be performance because clock for clock, its only 3.25% faster compared to sandy according to http://www.hardwarecanucks.com/foru...4-intel-i7-3770k-ivy-bridge-cpu-review-6.html.

Power consumption and GPU performance is much improved, but those are metrics that most enthusiasts posting on Anandtech hardly care about.

So the most important metric Ivy was going to make an impression on was through overclocking and yet it overclocks worse than Sandy (on air).
So thats why people evaluate Ivy on overclocking and thats why its a disappointment to many. Simple really...

I'm not even going to address the whole 'developed world' spiel.

 

[RussianSensation]

Leave it alone man . Enjoy the ignorance and let them rave. Its all they have for stones and there only pebbles that won't even crack a window grains of sand is all they be. Dying from AMD performance thrist.

It's obvious that IVB is PresHOT reincarnate. It's slower than the competition, consumes far more power and is a lousy overclocker. Those are facts.....:sneaky:

I think people really don't understand Intel's Tick-Tock strategy. Historically, even when Intel went from Conroe to Penryn, the performance increase in IPC was not that great either. Even with slightly higher IPC and more overclocking headroom, it still wasn't even worth upgrading from 65nm Q6600 G0 to a 45nm Q9550. The first legitimate upgrade path presented for Kentsfield owners were Nehalem/Lynnfield chips. Some even skipped those. However, the 45nm stop-gap Penryn chips were put to shame by i7 920.

It seems more than anything that some people are upset they don't have a valid reason to throw their $ away. If enthusiasts want to spend $ that was allocated towards a CPU upgrade, there are plenty of alternative ways to do so with much higher benefits - one can purchase multiple GPUs, SSDs in RAID/PCIe SSD cards, 120Hz 3D gaming setups, etc. IVB was never meant to be a significant improvement over SB since Intel's primary design focus was reducing power consumption, improving IGP and making IVB a better performance/watt chip for the mobile segment. Even if IVB could clock to 5.3ghz on air, that still would have been a lousy upgrade for 4.8-4.9ghz SB owners and inevitably be convincingly trumped by Haswell in 2013.

 

[Absolute0]

lol, now I really regret mentioning PrescHOT! IVB is hot that is for sure but it's not BAD the way Prescott was. I am considering buying an IVB for my next build! So when I mentioned Prescott it was because of the temperature increase... Not because I am trying to say that Ivy Bridge is bad, or the "reincarnation" or prescott.

I am debating between getting a 3770K on a Z77 mobo or going S2011 with a 3820 and having a better CPU upgrade path.

What's cool about Ivy Bridge is the ability to achieve a good OC on low volts or a very small overclock while undervolting. So if you don't need bleeding edge performance, you can get yourself a very fast, efficient computer.

Anyone else on this fence? Going S2011 for the upgrade path (I like the idea of getting a 3820 now with 2*8gb ram and in 2-3 years dropping in something like an IB-E 8-core or something with 2 more sticks of ram) or springing for the 3770K.
IDK I will probably decide based on whatever Microcenter decides to do