IVY E uses solder tim as expected, see delidded (and killed) 4960X with solder tim

[rge2]

IVY E uses solder tim as expected, see delidded (and killed) 4960X with solder tim
Link below, Toppc (same guy that typically posts engineering sample specs long before NDA are up, ie like he did with 4770K, etc), delidded his Ivy E eng sample 4960x, killing cpu in process as he had to pry solder apart (he believed the nonsensical internet rumor of cant solder 22nm, ie thought it was paste, even though after cutting adhesive he said he had to really pry to get it apart). You can clearly see the solder, amongst his ranting and cussing that someone spreading rumors about paste tim owes him another cpu.


http://translate.google.com/transla...E8%84%82-%E8%B6%95%E7%BE%9A%E7%BE%8A%E5%8B%92

 

[Pheesh]

The die surface area is notably bigger on IVB-E vs the 4 c die lines. Given it's on the same process as the high end xeon cpu it's not surprising it's soldered. Nice to get confirmation, but kind of silly he's blaming others for his own carelessness and stupidity.

 

[WhoBeDaPlaya]

Bloody Intel. The least they could give us is a soldered-on IHS.
Instead, we get DPed - no solder + artificial gimping of the K SKUs.

 

[Vesku]

Well there were people saying TIM wasn't so much cost cutting as the 22nm transistors causing more heat stress on the solder. ROFL at those people, if retail Ivy-E is soldered than there is no technical reason the K series couldn't be as well.

 

[FwFred]

Bloody Intel. The least they could give us is a soldered-on IHS.
Instead, we get DPed - no solder + artificial gimping of the K SKUs.

They are adding "K" to the cheapest 2011 part (i7-4820K), so that's helpful. Of course there is no $230 i5 and the platform is more expensive.

 

[Idontcare]

So much repressed anger and frustration writhing beneath the surface in this thread.

I don't get it, people take business too personal and then emotions come into play.

As to "technical reason versus cost cutting"...uh, is there a difference? I never came across a technical challenge in engineering that didn't come down to a cost-benefits justification. The two are inseparable, this should be obvious.

 

[WhoBeDaPlaya]

So much repressed anger and frustration writhing beneath the surface in this thread.

No worries mate, no aneurysms here, just kicking myself for waiting this long and finally getting what is essentially 5GHz SB level of performance, with bonus thermal problems.

 

[Vesku]

So much repressed anger and frustration writhing beneath the surface in this thread.

I don't get it, people take business too personal and then emotions come into play.

As to "technical reason versus cost cutting"...uh, is there a difference? I never came across a technical challenge in engineering that didn't come down to a cost-benefits justification. The two are inseparable, this should be obvious.

I'm more referring to people who make claims about 22nm suddenly making solder unusable. I've encountered it here and at other tech enthusiast sites. Why do people need to make up reasons, just acknowledge it was a decision Intel made for mostly cost reasons and move on. They obviously would rather you buy an -E chip than a K chip, that's business.

 

[crashtech]

This is great news!

 

[Idontcare]

I'm more referring to people who make claims about 22nm suddenly making solder unusable. I've encountered it here and at other tech enthusiast sites. Why do people need to make up reasons, just acknowledge it was a decision Intel made for mostly cost reasons and move on. They obviously would rather you buy an -E chip than a K chip, that's business.

Well I can understand your frustration with that, I see where you are coming from there.

I know not everyone is going to appreciate the hobby-angle as I do, but I actually enjoyed my 3770k more than my 2600k because I got to experiment and play with the whole delidding and bare-die aspects.

On the flip side, it stinks that Intel would intentionally cripple the thermal conductivity of their K-class processors (knowing full well that enthusiasts buy them for OC'ing) just to build in more upsell opportunity for their premium S2011 platform

Intel is on a roll lately in terms of making sure everyone knows they are resting on their laurels, milking the market and their customers, and once again giving their competitors the opportunity to one-up them again.

Humans, so predictable. :|

 

[BallaTheFeared]

Doesn't seem legit to me.

When I delidded my dead i5-2500k I had to rip the die out through the PCB, leaving a gaping hole in said PCB...

Also as IDC has pointed out many times, it's not tim vs solder, it's the distance that is the problem.

 

[Idontcare]

Doesn't seem legit to me.

When I delidded my dead i5-2500k I had to rip the die out through the PCB, leaving a gaping hole in said PCB...

Also as IDC has pointed out many times, it's not tim vs solder, it's the distance that is the problem.

D: That is a crime in at least 23 states

 

[AdamK47]

Doesn't seem legit to me.

When I delidded my dead i5-2500k I had to rip the die out through the PCB, leaving a gaping hole in said PCB...

Also as IDC has pointed out many times, it's not tim vs solder, it's the distance that is the problem.

It does appear as though heat was applied to separate the heat spreader from the die. You can see the solder bubbled up along the side of the die in the shots.

 

[BallaTheFeared]

Yes it looks like he says he used a heat gun, I think there is a good chance Ivy-E will be soldered. What isn't legit to me is the rest of it, strikes me as trolling.

 

[24601]

[redacted]

Commie, if we wanted your meta commentary on other posters we'd let you know. Otherwise knock it off.

-ViRGE

 

[SOFTengCOMPelec]

Well I can understand your frustration with that, I see where you are coming from there.

I know not everyone is going to appreciate the hobby-angle as I do, but I actually enjoyed my 3770k more than my 2600k because I got to experiment and play with the whole delidding and bare-die aspects.

On the flip side, it stinks that Intel would intentionally cripple the thermal conductivity of their K-class processors (knowing full well that enthusiasts buy them for OC'ing) just to build in more upsell opportunity for their premium S2011 platform

Intel is on a roll lately in terms of making sure everyone knows they are resting on their laurels, milking the market and their customers, and once again giving their competitors the opportunity to one-up them again.

Humans, so predictable. :|

Today I saw (video, another thread on here) an Intel executive, boasting about giant power savings on their present/future cpu chips.
But I'm convinced that when they go from soldered to TIM, the (approx) 20 degree temperature rise for a typical user (ok, heavy user, but NO overclocking), will add something like 10 watts to the computer power consumption, as higher temps usually increases power consumption.

Over an ownership period of many years, and if the computer sees medium to high cpu/IGP usage, could add a small fortune to the electricity bills, way in excess of what it probably saves Intel in manufacturing costs.

E.g. 10 Watts x 4 (years) x 8000 hours (left on most of time) = 320,000 extra watts (per accumulated hours over 4 years, = 320 KWh), which in expensive electricity countries, would be about (320 * $0.235~ =) $75 over the life of the computer. (way less if computer is switched off and/or lightly loaded).

--------------------------------------------------------------------

Also, I'm amazed at people destroying expensive cpu's by de-lidding them clumsily. Heat guns and expensive cpu's do not mix!

-----------

EDIT: I have noticed a mistake in my post. As the cpu temp rises, the fan will typically auto-speed-up, completely messing up my calculations. I still think it will run hotter for a typical user (doing medium to lots of stuff on the cpu/IGP, but NOT overclocking).

 

[rge2]

Doesn't seem legit to me.

When I delidded my dead i5-2500k I had to rip the die out through the PCB, leaving a gaping hole in said PCB...


Also as IDC has pointed out many times, it's not tim vs solder, it's the distance that is the problem.

Toppc is reliable, he posted accurate performance of 4770k and 3770k eng samples prior to NDA. His results are usually on many sites...here is his comparison of engineering sample Ivy E vs Sandy E at tech power up.
http://www.techpowerup.com/183176/c...ghly-10-faster-than-i7-3970x-early-tests.html

What matters for tim is bond line thickness (gap as you call it), thermal conductivity, and interface resistance. All 3 matter.

Intel could decrease bondline to 10 microns for tim and improve temps, which isnt likely to happen, and there may be a reason for thicker bondline.
Or intel could increase thermal conductivity since best pastes are in 5-8 w/mk range when applied, and intel indium solder has higher thermal conductivity by 10X (87 w/mk) and lower contact resistance.

Solder vs tim matters even more given intels thick bondline.

 

[BallaTheFeared]

Definity, I didn't see he used a heat gun at first.

Though I still stand by what I said, anyone "first" to explore excepts the risks involved.

No doubt it isn't his only sample, and most likely he didn't pay a dime for them.


Hence my feeling while Ivy-E is most likely soldered, the rest of it like blaming everyone is just trolling. Someone presumably that competent can't claim ignorance. There was no care taken in the delidding process, it was just to prove what I feel, someone already knew.

 

[Pheesh]

Well there were people saying TIM wasn't so much cost cutting as the 22nm transistors causing more heat stress on the solder. ROFL at those people, if retail Ivy-E is soldered than there is no technical reason the K series couldn't be as well.

Really? The IVB-E die has larger minimum dimensions than than sandy bridge 1155 chips. Most speculation was that IVB-E WOULD be soldered, as opposed to the K series because of this. (and also because it derives it's assembly process with the high end server die lines)

There is no technical reason they CANT be soldered, but it's not outrageous to suspect failure rates from drop tests/thermal cycling were higher than they wanted for the 2c/4c die's.

 

[sxr7171]

Today I saw (video, another thread on here) an Intel executive, boasting about giant power savings on their present/future cpu chips.
But I'm convinced that when they go from soldered to TIM, the (approx) 20 degree temperature rise for a typical user (ok, heavy user, but NO overclocking), will add something like 10 watts to the computer power consumption, as higher temps usually increases power consumption.

Over an ownership period of many years, and if the computer sees medium to high cpu/IGP usage, could add a small fortune to the electricity bills, way in excess of what it probably saves Intel in manufacturing costs.

E.g. 10 Watts x 4 (years) x 8000 hours (left on most of time) = 320,000 extra watts (per accumulated hours over 4 years, = 320 KWh), which in expensive electricity countries, would be about (320 * $0.235~ =) $75 over the life of the computer. (way less if computer is switched off and/or lightly loaded).

--------------------------------------------------------------------

Also, I'm amazed at people destroying expensive cpu's by de-lidding them clumsily. Heat guns and expensive cpu's do not mix!

-----------

EDIT: I have noticed a mistake in my post. As the cpu temp rises, the fan will typically auto-speed-up, completely messing up my calculations. I still think it will run hotter for a typical user (doing medium to lots of stuff on the cpu/IGP, but NOT overclocking).

I don't get that either. All for a small sum in savings upfront. How they can say in good conscience they are doing everything to lower power consumption is beyond me.

 

[Vesku]

There is no technical reason they CANT be soldered, but it's not outrageous to suspect failure rates from drop tests/thermal cycling were higher than they wanted for the 2c/4c die's.

Within the 3 year retail warranty period? Highly doubtful.

Within the 1 year OEM warranty period? Unbelievable.

Then there is the fact that the total die area is a minimal aspect to claims of transistor thermal behavior being the primary reason for the switch to thermal interface material. As the purported behavior would mainly involve vertical stress.

 

[Concillian]

Today I saw (video, another thread on here) an Intel executive, boasting about giant power savings on their present/future cpu chips.
But I'm convinced that when they go from soldered to TIM, the (approx) 20 degree temperature rise for a typical user (ok, heavy user, but NO overclocking), will add something like 10 watts to the computer power consumption, as higher temps usually increases power consumption.

Umm... IDC has this data, I hope he doesn't mind me re-posting it, it's buried in the delid thread, but doesn't have a direct link in the first post of that thread, I think it's a pretty interesting chart and warrants some publicity:

So on this CPU it's 7 watts from ~30C difference that's attributable to a combination of the H100 (vs stock cooler), the reduced voltage, and de-lidding / TIM. So I think you're looking at maybe 1-2 watts, since I think the H100 vs. stock cooler and the voltage difference are much bigger factors than de-lidding / TIM at stock clocks. And that's with IBT, one of the heaviest loads imagineable. As loads and heat decreases, any consumption difference also decreases.

 

[JimmiG]

 

[Shmee]

Things will get worse before they get better I guess. We will need drastic change.

 

[Arachnotronic]

Intel is on a roll lately in terms of making sure everyone knows they are resting on their laurels, milking the market and their customers, and once again giving their competitors the opportunity to one-up them again.

Humans, so predictable. :|

Heh...given Intel's (relatively) precarious financial situation I'd say they are just focusing their efforts on the really sore spots where they need to break in. Once phones/tablets aren't so much of a damn sore spot, we enthusiasts may get some love again