Intel Increases Overclocking Potential in Haswell Processors

[tweakboy]

Overclocking has become increasingly difficult ever since Sandy Bridge and it is made a bit more difficult with Ivy Bridge mainly because it runs hotter because Intel decided to use a substandard thermal paste instead of fluxless solder. This could very well change with Intels Haswell platform. As reported by Anton Shilov of XBit Labs, Intel will offer more overclocking flexibility with the new platform, and here's why.
Read more

 

[Puppies04]

because it runs hotter because Intel decided to use a substandard thermal paste instead of fluxless solder.

Intel fully explained why they felt that fluxless solder wouldn't have worked long term with IB. Show me a chip that has been delidded and had aftermarket paste installed in 5 years time still running at decent temps and I will change my mind until then people need to stop making up BS and talking about things they obviously don't understand.

 

[Smoblikat]

Awesome, if they do this then never again will I look to AMD for any CPU solution. I was quite angry that they locked down overclocking so much. if they allow this on non-K processors it will be almost as good as the LGa775 days of grabbing a 60$ E5200 and OCing the hell out of it.

 

[BenchPress]

Due to being a wider architecture, I expect Haswell will have a lower base frequency. It should also be optimized for lower voltages, to make it better for the mobile market. In return it should have a higher IPC though, and good overclocking potential if you increase the voltage and keep it cool enough to not destroy those sensitive transistors.

 

[Smoblikat]

Due to being a wider architecture, I expect Haswell will have a lower base frequency. It should also be optimized for lower voltages, to make it better for the mobile market. In return it should have a higher IPC though, and good overclocking potential if you increase the voltage and keep it cool enough to not destroy those sensitive transistors.

The article said it has higher frequencies, so I assume the base will be the same. I bet the BCLK is set to 100 by default and the multi is set to 33, or whatever it is at.

 

[Idontcare]

Intel fully explained why they felt that fluxless solder wouldn't have worked long term with IB. Show me a chip that has been delidded and had aftermarket paste installed in 5 years time still running at decent temps and I will change my mind until then people need to stop making up BS and talking about things they obviously don't understand.

I agree on the latter, but I have not had the pleasure of acquainting myself on what you mentioned in the former (that Intel explained themselves).

I'm not above doing my own heavy-lifting with google to track it down, but if you happen to have the link handy and don't mind posting it up then I'd really enjoy reading up on Intel's reasoning.

 

[BD231]

Intel never said a thing about their new thermal solution until people started complaining and it's obvious they'd have never released their CPU's with unlocked multi's if they weren't listening. Overclocking potential was limited by a poor quality thermal option, shame on Intel but whatever, it's not like Ivy offers anything over it's predecessor anyway.

 

[MLSCrow]

Intel fully explained why they felt that fluxless solder wouldn't have worked long term with IB. Show me a chip that has been delidded and had aftermarket paste installed in 5 years time still running at decent temps and I will change my mind until then people need to stop making up BS and talking about things they obviously don't understand.

This has to be a joke, right?

-Intel made a bad decision to use TIM instead of their patented application of fluxless solder between die and IHS in an attempt to decrease coast and improve profitability.
-Fluxless solder would have most definitely worked long term with IB (please link their explanation as to how it wouldn't have worked long term, as simply stating so is laughable).
-Intel used an extremely cheap and poor quality TIM which is much to blame for high temperatures in Ivy Bridge chips.
-It was because using cheap TIM instead of solder that has lead to the need/desire to delid IB chips in the first place.
-No paste or grease based TIM is going to be as efficient after 5 years, which is why solder would have been the ideal method (no one will ever need to re-solder their sandy bridge chips, just reapply the TIM between IHS and cooler).
-Phase change or liquid metal alloys are a decent alternative to other forms of TIM as they will last much longer and have improved thermal conductivity.

 

[MLSCrow]

it's not like Ivy offers anything over it's predecessor anyway.

It did offer native PCIe3.0, ability to use love voltage DDR3, HD4000 instead of 3000, and some added security features, but otherwise I agree...nothing that would really stand out or be of much use to most of us at least at this point in time.

 

[Puppies04]

I agree on the latter, but I have not had the pleasure of acquainting myself on what you mentioned in the former (that Intel explained themselves).

I'm not above doing my own heavy-lifting with google to track it down, but if you happen to have the link handy and don't mind posting it up then I'd really enjoy reading up on Intel's reasoning.

Hi idc,

I'm on my phone at the moment so I will have to wait until I get home to find the announcement but I was sure it had been linked in several threads on this forum. They basically said that the increased thermal cycling from the 22nm design meant that solder would have been unsuitable over the full lifetime they expect the chips to run for. I can't comment on the truth behind that statement but it doesn't seem to have stopped various other people grabbing their tin foil hats and claiming that Intel intentionally gimped themselves for various ridiculous reasons.

I will see if I get chance to find the link later.

 

[bunnyfubbles]

Hi idc,

I'm on my phone at the moment so I will have to wait until I get home to find the announcement but I was sure it had been linked in several threads on this forum. They basically said that the increased thermal cycling from the 22nm design meant that solder would have been unsuitable over the full lifetime they expect the chips to run for. I can't comment on the truth behind that statement but it doesn't seem to have stopped various other people grabbing their tin foil hats and claiming that Intel intentionally gimped themselves for various ridiculous reasons.

I will see if I get chance to find the link later.

using a cheap paste instead of fluxless solder to increase profit margins isn't exactly a "ridiculous" reason...nor does intel have to worry about "intentionally gimping" their parts when they lack competition

 

[NTMBK]

Given that Haswell with have VRMs on the die as opposed to the motherboard, I'm a bit dubious about how well its going to overclock. No more motherboards with overspec VRMs to give you more headroom.

 

[LOL_Wut_Axel]

Given that Haswell with have VRMs on the die as opposed to the motherboard, I'm a bit dubious about how well its going to overclock. No more motherboards with overspec VRMs to give you more headroom.

Overspec VRMs on SB/IB were nothing more than marketing bullets, anyway. A 6+2 or 8+2 power phase design is well more than enough for heavy overclocking due to the paltry power consumption of the CPUs.

 

[BenchPress]

The article said it has higher frequencies, so I assume the base will be the same. I bet the BCLK is set to 100 by default and the multi is set to 33, or whatever it is at.

I didn't mean BCLK. I meant stock frequency. Haswell might sacrifice frequency for IPC. If not, that probably means they ate into the overclocking headroom.

That would explain why they're trying to satisfy the overclocking community with more options to tweak...

 

[Puppies04]

using a cheap paste instead of fluxless solder to increase profit margins isn't exactly a "ridiculous" reason...nor does intel have to worry about "intentionally gimping" their parts when they lack competition

Intel have plenty of competition, from a company called intel.

 

[Puppies04]

Hi idc,

I'm on my phone at the moment so I will have to wait until I get home to find the announcement but I was sure it had been linked in several threads on this forum. They basically said that the increased thermal cycling from the 22nm design meant that solder would have been unsuitable over the full lifetime they expect the chips to run for. I can't comment on the truth behind that statement but it doesn't seem to have stopped various other people grabbing their tin foil hats and claiming that Intel intentionally gimped themselves for various ridiculous reasons.

I will see if I get chance to find the link later.

Ok, I am not finding a link to what I thought I had read as a direct release from intel so I may have confused conjecture from a third party with an intel statement. Perhaps we can get a comment from our resident intel employee????

 

[Borealis7]

they'll never admit to that. they are not fools.

 

[rge2]

From intels point of view, seem norms for them not to use solder on IVY, given stock settings specs. Most intels cpus on low TDP curve use PTIM (polymer/paste tim). Only the high TDP cpus need or use the more expensive, laborious STIM (solder tim). That info is readily available from reading intel white papers, like link where it describes the reason for developing solder in first place.
http://download.intel.com/technolog...ials_Technology_for_Environmentally_Green.pdf

Sandy 2600K is 95W TDP. Since development of STIM, cpus 95W+ range TDP have solder. The thermal design of say 55W does not call for more expensive/laborious solder, as difference AT STOCK between stim/ptim would be marginal, hence those have PTIM.

And the surface/material of IVY is same as surface of Sandy bridge, and no reason cant use solder, thermal cycling of IVY being more than 130W cpu is nonsense. Solder per intels own white papers has higher not lower resistance to thermal cycling....that is just supplanting one rumor with another.

>99% cpus are run at stock. IVY's TDP and temps at stock fall in range of using PTIM on intels past cpus.

The only issue here is the K cpus, when overclocked, the TDP calculated for OCed would fall in range of using solder TIM, hence OCes <1% got the short end of the stick. No question if intel sold those overclocked with TDP of 130W, they would have solder.

But to have just K's soldered, since it is likely post IHS where they are programmed with thermals and what cpu, would be an ordeal.

I just hope intel heard enough complaints from the <1% for intel to find a way to pass on any expensive of using both solder for OCed ones and STIM rest, to at least give next generation Haswell STIM for ones we intend to OC.

No way to prove why intel did it, but thermal cycling difference issue makes no sense at all. And to me, looking at specs of stock IVY (temps and TDP) historically vs like E7200 (65W, PTIM), it shouldnt be that much of a surprise.

Just no love for those that bought the K cpus.

 

[bunnyfubbles]

From intels point of view, seem norms for them not to use solder on IVY, given stock settings specs. Most intels cpus on low TDP curve use PTIM (polymer/paste tim). Only the high TDP cpus need or use the more expensive, laborious STIM (solder tim). That info is readily available from reading intel white papers, like link where it describes the reason for developing solder in first place.
http://download.intel.com/technolog...ials_Technology_for_Environmentally_Green.pdf

Sandy 2600K is 95W TDP. Since development of STIM, cpus 95W+ range TDP have solder. The thermal design of say 55W does not call for more expensive/laborious solder, as difference AT STOCK between stim/ptim would be marginal, hence those have PTIM.

And the surface/material of IVY is same as surface of Sandy bridge, and no reason cant use solder, thermal cycling of IVY being more than 130W cpu is nonsense. Solder per intels own white papers has higher not lower resistance to thermal cycling....that is just supplanting one rumor with another.

>99% cpus are run at stock. IVY's TDP and temps at stock fall in range of using PTIM on intels past cpus.

The only issue here is the K cpus, when overclocked, the TDP calculated for OCed would fall in range of using solder TIM, hence OCes <1% got the short end of the stick. No question if intel sold those overclocked with TDP of 130W, they would have solder.

But to have just K's soldered, since it is likely post IHS where they are programmed with thermals and what cpu, would be an ordeal.

I just hope intel heard enough complaints from the <1% for intel to find a way to pass on any expensive of using both solder for OCed ones and STIM rest, to at least give next generation Haswell STIM for ones we intend to OC.

No way to prove why intel did it, but thermal cycling difference issue makes no sense at all. And to me, looking at specs of stock IVY (temps and TDP) historically vs like E7200 (65W, PTIM), it shouldnt be that much of a surprise.

Just no love for those that bought the K cpus.

well if that's the case it bodes very well for s2011 and future IvyBridge-E chips

although hopefully they bring the solder back to at least the K series Haswell, it would definitely help to differentiate them from the other CPUs which would be able to overclock via BCLK straps yet be held back by thermal limitations, wouldn't hurt to charge a bit more for the solder chips either, as enthusiasts would have more options altogether win-win-win

 

[Ajay]

Due to being a wider architecture, I expect Haswell will have a lower base frequency. It should also be optimized for lower voltages, to make it better for the mobile market. In return it should have a higher IPC though, and good overclocking potential if you increase the voltage and keep it cool enough to not destroy those sensitive transistors.

From what I've read so far, Intel is expecting to keep frequencies around the same as IVB. The only real issue with more execution resources seems to be presumed problems with port forwarding with the four ALUs (such has been argued over at the RWT forums), something that apparently has been worked out by Intel supposedly without increasing latencies either. We'll have to wait for further disclosures to see how they've managed that. But generally, it appears that Intel had a lot of headroom with IB and hence was able toss in the kitchen sink as well. That, or they are blowing smoke up our butts ^_^.

As far as overclocking really goes, nice to have variable BCLK, but I'd reserve judgement till the chips are out there (especially after the 'disappointment' over IB).

 

[Phynaz]

-Intel used an extremely cheap and poor quality TIM which is much to blame for high temperatures in Ivy Bridge chips.

This much of your post as been proven false by a member here (IDC) who replaced the Intel stock TIM with IC Diamond with no change in temps.

 

[piesquared]

Why is this considered an increase? It's just back to where it should have been originally.

 

[Thor86]

Why is this considered an increase? It's just back to where it should have been originally.

Agreed, BLCK overclocking replaced FSB overclocking. Now Intel is back to BLCK overclocking like the Nehlem days.

 

[HexiumVII]

When will Haswell launch? I'm thinking of upgrading my shop Nehalem i7 920 @ 3.2GHz to Sandy (i have a 2700K lying around ) and just did an Ivy of my home machine. But i'm thinking of just waiting for a Haswell

 

[beginner99]

Whats the take away message? link is blocked here at work...