http://forum.lowyat.net/index.php?showtopi...post&p=50703693

Yes, Intel uses cheapo thermal interface material aka TIM which has a horrible heat conductivity between die and integrated heatspreader aka IHS.

Die density or vast amount of components being packed in a smaller die size is another culprit. This has admitted by Intel themselves and it looks like there's no action taken by Intel uptil now

Thanks for the link.It seems like 10'c drop replacing the tim thing.
Guess IB is double combo with poor tim and pack die size.

But worth getting a old processor 2500k than 3570k ? the only advantage is watt usage.15 watts ok idle or overclock ?

yeah i also got roughly 10C drop when replacing TIM and running chip topless (w/o IHS) my 3770K is now benchable at 5GHz. Although it's hotter than SB, IB saves power, faster and has extra features over the SB.

When they claimed 15w off means this apply for both idle and load states. But hey who cares about idle shit.

waloa eh, so hardcore replacing TIM.

just in here....got an i5 3570k..and what? intel giving sxxt to us inside it?

omg... the steps there by kelvin seems hardcore to replace the thing inside...

huh , ? you also replace the tim grease ?
Wow , really , warranty void ? at the time you replace , are you worry the processor cannot boot up ?
I saw step , really dangerous
btw ,did you try on IB 2500k replace the tim? maybe getting a better result ?

I'm pretty sure that Intel will have their own reason to switch to regular TIM instead of solder. Besides, it only affects a small minority of users (enthusiasts) and it doesn't make a huge difference for regular mainstream users who will never overclock or will never know or find out the inner workings of the box they call a PC. Still no answer as to whether TIM will be used again in Haswell, if it did then the temps would still be a little bit high.

With the proper techniques and tools, it won't be more difficult than assembling your own rig unless of course you're this guy. A thin razor is all that is needed actually.

SB processors uses solder instead of TIM on IB. Replacing it will have disastrous results with you ripping the die off the substrate and the die stuck to the bottom of the IHS. There is no proper way of removing the IHS on SB without factory controlled equipments.

It is more likely the "fear" of Ivy cannibalize SB-E than "it only affects a small minority of users(enthusiasts)". Not affecting normal user is a BS reason. Intel know/learned very well how SB killed gulftown.

All Celeron + Pentium CPU since Core 2 Duo are TIM applied, all of them share the same SKU vs their bigger brother like Core 2 Duo + i3(Nehelem) which use soldered IHS. That means Intel already able to binned the chip b4 applying the IHS.

The K series is intended for Enthusiast. Normal user wont buy K series. Intel could have just do the solder for K series alone. There is little selling price diff between the non-K and K series. I am pretty sure a lot of Enthusiast wont mind pay another RM50 extra for their CPU get soldered professionally by Intel.

But we all know, A soldered nicely done Ivy can kill SB-E, if it wasnt for Bulldozer fail, Intel wouldnt cut corner on their Ivy.

right. they did that on purpose. if 2600K can beat nehalem's E, then 3770K can beat 2600K or even SB E. thats why they must "reduce" the power by making it "more hot".

It all boils down to cost of the product and being the minority group, we'll have to go along with what the mainstream users are going for. For all we know, Intel might just disable overclocking altogether for their entire range of processors like what they did with their Xeon lately.

SB didn't kill Gulftown because SB is cheaper and more affordable to majority of users which is an obvious choice and SB didn't excel in every benchmark out there, ST and MT. If it did then it did kill Gulftown but from reviews, you can see that Gulftown still holds its ground pretty well in a highly threaded program that could utilize 6 cores/12 threads. Gulftown and SB-E was never a high volume production to begin with since there are only so many enthusiasts with that much cash in hand and cheaper mainstream products is easier to sell and more profit.

Intel could do a solder for K series processors but they won't. You're creating 2 separate assembly lines for 2 processors that requires different packaging methods and you need 2 different machines to do the job as well which just adds cost. The use of solder adds a lot of cost to the end product as the solder itself is composed of rare metals and definitely more costly than cheaper TIM.

IB won't kill SB-E even if it had solder in it as once you got rid of the thermal limitation, it overclocks about the same as SB but needs more voltage which is to say that IB works best when the voltage is low and it is purposely made that way. This is not to say that I support Intel for what they are doing, they might change back to solder with Haswell and everything is back to where it should be but even if that didn't happen, they could've used a better quality TIM. Delidding and applying aftermarket TIM has given a huge improvement in temps, it suggests that it is not necessary for Intel to go back with solder, just use better quality TIM.

Fixing Ivy Bridge CPU temps: IHS removal

http://www.youtube.com/watch?v=XXs0I5kuoX4...player_embedded

Nice thread " 3770k IHS Removals" :
http://www.xtremesystems.org/forums/showth...C-to-71C/page13

This post has been edited by stasio: Aug 4 2012, 05:15 PM

SB does cannibalize the sales of gulftown since most of the app we use are <4 threads & performing much better in SB. I do not think Intel have any reason to repeat this again on Ivy. the recent leak slides shows that Haswell Mainstream is going to launch b4 the Ivy-E. I dont think Haswell will offer any substantial performance upgrades. As it is much more closer to the mobile chip than a desktop chip, Intel is pushing Ultrabooks against tablet now. The significant TDP improvement + GPU in Haswell is likely to be their main focus. Then they just port the same SKU for desktop with higher TDP.

For K series, Because Intel had does 2 assembly lines all the way back from core 2 duo. The Pentium/Celeron series as we know are TIM base, it is a same chip vs their bigger brother. if they done it on the value segment, cost wont be a biggest issue on expensive K series. I dont think the cost is going to affect much to Intel especially they are charging a premium on K models that easily cover it. It is more like Intel wanted to do or not.

This post has been edited by Boldnut: Aug 4 2012, 06:49 PM

You cannot consider sales cannibalization outside of its target market. SB is targeted at mainstream, Gulftown and SB-E is targeted at enthusiasts or people who really needs more threads on their consumer level machine. There is a clear separation of the target market so even if SB does sell more than SB-E, it is already obvious that SB is targeted at the price range where many can afford.

Sales cannibalization can happen if Intel suddenly releases a Core i3 3xxxK. Because Core i3 and Core i7 exists in the same mainstream category, you can expect that it will affect the sales of Core i7 3770K due to the fact that both are in the <1k range where many can afford them. Intel very well knows that if they do release an overclockable Core i3, it will be a best seller and affect the sales of the higher end products. As it is right now, Core i5 2500K/3570K is a best seller as it overclocks, cheaper than the Core i7 by a huge margin, and has about 80% of the performance of a Core i7.

Gulftown, SB-E, IB-E is just Intel's pet project and only accounts for a small percentage of their revenue so they don't really give a lot of priority to developing it faster. SB, IB and Haswell are mainstream products and it needs to stick to the schedule because we're not talking about desktop products only, this also includes mobile products which accounts for a huge percentage of Intel's revenue. Enthusiast level products don't rake in a lot of money, add also that there are more mainstream users that are just not savvy or interested in overclocking their CPU.

The factory equipment from C2D days are old, you can bet that it is way better for them to buy newer machines with lower % of packaging errors. Setting up 2 assembly lines is tougher than it is to set up 1 assembly line that does everything. There is more reasons actually as to why Intel might go with TIM for their future products beyond IB, but my post is getting too long.

Lets just say the most of Casual user even gamers are not so interest in 6 cores due to the slow 6 thread adoption on consumer apps. Performance user(not enthusiasts) are not going for 6 core due to the good SB performance. They are there to pay for P4 extreme, Core 2 quad extreme not anymore on SB-E.

my 2500K isnt my most exp CPU I paid, just fyi.

This post has been edited by Boldnut: Aug 4 2012, 10:30 PM

We've come to a point where most consumer applications are still stagnant with 4 cores but soon we'll need it. A dual HD7970 can be bottlenecked with current mainstream CPU setups, we will need more CPU performance to ensure that such bottlenecks won't happen. When will that happen is up to Intel, as of now 6 cores is just not going to happen on the mainstream side. Non enthusiasts are not going for 6 cores because Intel doesn't price them according to the mainstream price bracket. Price is often a limiting factor instead of performance.

The market shifted, there used to be a lot more enthusiasts back then but the numbers are shrinking so there are less people willing to buy an expensive CPU that does almost the same as a cheap one in daily task. Definitely, Core i5 2500K pales in comparison to the price of CPUs of the past. Based on the price that most enthusiasts pay back then, I would say that even the high end CPUs now are actually very affordable.

heck the 4 threaded Apps are not even fully optimized for the 4 cores, u mostly see the "common" Apps claim to support 4 thread are poorly optimized on the 3rd/4th core.

Core 2 quad is like 5yrs old already. I think we need another 5yrs to fully adopt/optimized the 6 cores. I wonder who is the guy behind AMD to come up the idea to design bulldozer for "special" 8 core + expecting software developer to adopt it. *good luck with that. I think Intel high IPC are more feasible for now.

This post has been edited by Boldnut: Aug 4 2012, 11:23 PM

sorry if this has been asked before but guys, is it really necessary to run Prime95 for 8 hours? It has been running almost to 2hours now and so far no errors and can i just assume my newly-OCed clock stable?

My setting's currently @4.2GHz 1.152Vcore (offset -0.065V). And any advices on improvement of my setting?

Thanks in advance all sifus here.

There is definitely some problem with the current MT applications not supporting more than 4 cores but I can say the same when they first introduced dual cores. Back then there was nothing MT about applications and a few years later it has never occurred to CPU manufacturers to go back to a single powerful core which like multi core approaches, has its drawbacks once a certain limit is reached. It is difficult to code something that wants it done sequentially to something parallel but multi core is the de facto standard of future CPUs and we'll just have to find ways to make it better.

There is nothing wrong with BD's approach but they aimed it at every single market instead of focusing them on the server market alone. Intel has a huge surplus of funds to throw at different projects and one of them is Knights Corner, aim for higher core density instead of higher core performance. Luckily Intel only sells them to a niche market which would benefit from having more cores.

2 hours of Prime95 is considered minimum, you need 12 hours or more. Even so, it won't guarantee you 100% stability for every programs you run. Prime95 simply tests a small portion of the CPU although you're looking at 100% load in Task Manager.

You can run other stress testing programs like OCCT, IntelBurnTest, LinX and HCI memtest to ensure that you have covered more areas that might cause instability. Even then, you're still not guaranteed 100% stability but your chances are only higher. Only Intel knows whether the CPU is 100% stable overclocked but they definitely wouldn't publish their custom written stress testing software to the public.

4.2GHz is pretty mild, you shouldn't encounter any issues. If there is any instability and it BSODs, write down the BSOD code and check for it. Most of the time it needs you to increase the voltage for more stability.

it is much harder to adopt the quad core over the dual core MT coding, just the dual core alone it took these software developers long enough. Quad core is out about 5yrs already, we still barely able to fully optimized it, unless the software developers find a way to do it easier or they put more resources to it, it is going to take as long to adopt 6 MT apps. I am not saying BD is bad, but with AMD putting hope on consumer software adopting 8 core MT, it is going to take a lot more longer. There is a reason why Intel still retaining 4 core on Haswell

My usual procedure for testing an OCed system:
1.)Set Overclock Setting to the desired setting
2.)Use IBT 3 Loops,set to (Total memory GB-1GB,eg:4GB-1GB=3GB)
3.)Make sure it pass
4.)Than lower vcore
5.)Final setting run 20 loops IBT with same memory setting as 3 loops

If this pass,the system is quite stable already