Intel 22nm 'Ivy Bridge' Core i7 3770K Reviews

[rge]

Part of the heat issue is intel looks like did NOT use solder on ivy for die/IHS attach. IMOG posted pic here, showing paste on delided ivy cpu. Some of the cheaper cpus in past have had paste like E7200 paste tim vs E8400 solder tim.

Since not using solder, should get better temps removing IHS, and even better if use high thermal conductance tim (indigo but problematic install) directy to waterblock. Significant portion gradient core/IHS (in past articles describing thermals for IC) was attributed to tim interface. Only takes a minute with razor blade to delid non-soldered IHS...but the real issues are having to mod cpu socket, risk killing board/cpu from shorts if not secured properly.

I assume intel also using paste for xeon counterparts, but wonder...

I guess for running at stock settings, made financial sense to use paste, just less beneficial for overclocking. Wish new thermal conductance, and exactly what they were using though, for comparison sake.

[Hondacity]

basically youre saying intel doesn't know how to make cpus. that is too funny.

[LightSpeed]

basically youre saying intel doesn't know how to make cpus. that is too funny.

I dont think he meant that. What he meant is intel is using paste between the core and IHS instead of soldering it, and that reduces thermal conductivity but is cheaper. I cant verify the points though

[Piledriver]

that's my point.. i dont buy what everyone says oh it's such a smaller die size that hsf cant cool it blah blah.. we had/have much smaller and similar power draw consumption numbers various chips that this wasnt the case until ivb but now everyone says that..

And how many transistors both have? The real problem is that there are just more things producing heat for the same die size. I may be wrong but probably that's the reason right there.

[hokiealumnus]

Ivy Bridge Temperatures – It’s Gettin’ Hot in Here, courtesy IMOG. We also have a feeler out to Intel for comment. They already commented, but with humor (the interface is "secret sauce"). We hope to get a bit more detail the second time.

[Manicdan]

i say someone cut open a cheap IB, scrape that paste off and use it on a test bench and see how it compares to other pastes

also the layer of paste i see in their image shows that its really thick, pointing toward bad pressure

for the argument of how bad paste is, there was a nice little debate about adding a copper plate to the 7970 core so the universal waterblocks could fit, and we were told that it would be only a few C and quite insignificant. so if the IHS is mounted properly, it should contribute to just a few C increase, not 20C

and to the point about density, all math i see sofar points to total chip size, but didnt they increase the gpu size? id rather see core vs core density instead of chip vs chip (it might not change the numbers by much though)

[hokiealumnus]

and to the point about density, all math i see sofar points to total chip size, but didnt they increase the gpu size? id rather see core vs core density instead of chip vs chip (it might not change the numbers by much though)

Sandy Bridge die:



Ivy Bridge die:



They appear to be very close to the same amount of die space per core. Ivy may have lost a little bit due to the GPU increase, or they may have simply expanded the CPU in that direction and kept the cores proportionally the same (which looks to be the case based on the die elongation). Either way, the die dedicated to cores is very close to the same.

EDIT - Hahaha...until right now I never noticed they have the same text under the image. That's amusing.

[Cold Fussion]

Part of the heat issue is intel looks like did NOT use solder on ivy for die/IHS attach.

It has a temperature issue but it doesn't look like it has a heat issue.

[rge]

basically youre saying intel doesn't know how to make cpus. that is too funny.

No, what I said was intel is using paste instead of solder for die attach per IMOG's pic. Solder has higher thermal conductance. Not so much difference at low power and stock, but larger difference at higher power/Oced. Then I inquired as to what the thermal conductance of intels non-solder paste was.

But you must know the conductance of paste/die attach that intel is using to make that comment. Can you list the thermal conductance of both solder and intels current tim, the difference it makes at stock vs overclock on Ivy die, I am genuinely interested in learning?

@Cold Fussion, I think you were trying to point out that the overall wattage or heat output wasnt the problem, and I never made that claim anyways. I am sure you know what I meant, and I would argue about your nitpicking but after reviewing reading comprehension on other post, clearly you are correct for nitpicking.

[Mats]

Given the stretched, rectangular design of the die, how much does the orientation of a heatpipe cooler matter?

Does turning a tower cooler 90° make any difference? It's not an optimal solution in terms of case airflow or video card clearance, but I'm just curious..
Not that water cooled CPU's runs that much cooler tho.

Edit: I realized that dual towers like D14 already have the heatpipe orientation the other way than regular tower coolers.

[BeepBeep2]

Given the stretched, rectangular design of the die, how much does the orientation of a heatpipe cooler matter?

Does turning a tower cooler 90° make any difference? It's not an optimal solution in terms of case airflow or video card clearance, but I'm just curious..
Not that water cooled CPU's runs that much cooler tho.

It definitely could depending on if you use a Heatpipe-Direct Touch cooler (middle heatpipe over length of core) or where the heatpipes are placed...

I doubt very much.

I also doubt that the difference in temp between TIM and solder accounts for up to 20c. Maybe 5-10 at most... as dies shrink so does the surface area, one reason we have heatspreaders in the first place is to disperse that heat before it gets to the heatsink.

[kuroikenshi]

It'd be cool if the whole void could be filled in with a di-electric heat conductor, its maybe far from feasible, but that might help to dissipate further heat. Anyone willing to try?

[Monstru]

I see you already linked out graphics study, here is the CPU review and the mobile review:

IVY Bridge – Partea I – Intel Core i5 3570K si Core i7 3770K | lab501

IVY Bridge – Partea III – ASUS K55J | next.lab501

[Leeghoofd]

Try your air cooler passively, so without a fan spinning, does it get hot ? doesn't get hot here so conductivity is the issue...

[dasa]

i say someone cut open a cheap IB, scrape that paste off and use it on a test bench and see how it compares to other pastes

also the layer of paste i see in their image shows that its really thick, pointing toward bad pressure

for the argument of how bad paste is, there was a nice little debate about adding a copper plate to the 7970 core so the universal waterblocks could fit, and we were told that it would be only a few C and quite insignificant. so if the IHS is mounted properly, it should contribute to just a few C increase, not 20C

and to the point about density, all math i see sofar points to total chip size, but didnt they increase the gpu size? id rather see core vs core density instead of chip vs chip (it might not change the numbers by much though)

yep many variables that people are guessing about
as for the 7970 shim i made one for my gpu block out of 1mm copper sheet and with 2 globs of arctic ceramic it still beats most full cover blocks temps
20c ambient water 25c peak temp
1125mhz 25c idle 34c max during stalker cop benchmark

[BlindFreddie]

@ Fugger:
You know that IB CPU that Mike's going to cough up ...
Got a razor blade & a TRUE, mate?
You know, in the interests of Xtreme research ...
Pretty please?
Think of the qdos there'll be for XS ... Headlines like "XS Overclocking God Sets WR IB Clocks on Air"

(NB: not an IB processor)
================================================== ========================================
[EDIT]:
korrrhonen has posted his experience with removing an IB IHS at http://www.sf3d.fi/forum/viewtopic.p...&t=50&start=10 (in Finnish)
1. Without the IHS, the small CPU PC board flexes and the pads don't contact the socket pins properly.
2. He replaced the Intel TI paste with some high-end stuff and put the IHS back on.
The difference was not worthwhile for achieving higher clocks - only gained about 50MHz. Temp changes not stated.

He gave no estimate of the thickness of the layer of TI paste.
Someone else has suggested sanding down the underside of the rim of the IHS to reduce the thickness of the TIM layer.
Another idea (mine) would be to leave the IHS off and put a shim layer - thin dense cardboard might do - around where the edge of the IHS went, to prevent the CPU PC board flexing away from the socket pins.
Go for it ...