IB ES Overclocking!!!

i sure hope that's due to the wrong TJ Max like a few are saying in that thread

it probably actually has to do with the leakage from those trigate trannies though...

from what benches were in the thread, the difference between similarly clocked SB and IB is quite small
the main benefit would be power consumption at this point

interested in some retail chip reports though :)
 
Those temps make no sense. Fail heatsink/PEBCAK error, or the temps are getting reported incorrectly IMO.
 
The temps, IBT results, and voltage used don't match up. I think that Intel ES 3770k was designed to fail. I'd wait on official release before putting to much trust in experiences like the ones found in that thread.
 
SonDa5 said:
The temps, IBT results, and voltage used don't match up. I think that Intel ES 3770k was designed to fail. I'd wait on official release before putting to much trust in experiences like the ones found in that thread.
Click to expand...

You could say the same thing about kepler 0_o
 
Yeah, there could be any number of issues.

However, when running at a similar power when overclocked as say some Sandy Bridge processor (likely different speeds to get the same power consumption), IB will certainly run at a higher temperature given similar conditions like same ambient temp and same cooling.

The reason is the same as previous chips that underwent a die shrink. A 216mm^2 chip such as 4c Sandy Bridge can dissipate to the heat spreader better than a ~161mm^2 4c IB due to the 25% decrease in area on IB.
 
The thing runs hot as hell, we're not imagining it. It must be the tri-gate 3d resistors and lower nm process.
 
pxc said:
Yeah, there could be any number of issues.

However, when running at a similar power when overclocked as say some Sandy Bridge processor (likely different speeds to get the same power consumption), IB will certainly run at a higher temperature given similar conditions like same ambient temp and same cooling.

The reason is the same as previous chips that underwent a die shrink. A 216mm^2 chip such as 4c Sandy Bridge can dissipate to the heat spreader better than a ~161mm^2 4c IB due to the 25% decrease in area on IB.
Click to expand...

Yea, that shrink in die size looks like it's also hurt heat dissipation. I guess we'll be waiting until IB-E to see just how effective Tri-gate is (or like the the person above referred to it as leaking "trigate trannies." Thank you for the mental image). Nearly 1.4vcore to get to 5ghz is pretty brutal and it shows. A stepping would likely be months away -- unless they've had one in the works -- which would put a stepping somewhere between IB and Haswell release.

Should still make great laptop chips :)
 
pxc said:
Yeah, there could be any number of issues.

However, when running at a similar power when overclocked as say some Sandy Bridge processor (likely different speeds to get the same power consumption), IB will certainly run at a higher temperature given similar conditions like same ambient temp and same cooling.

The reason is the same as previous chips that underwent a die shrink. A 216mm^2 chip such as 4c Sandy Bridge can dissipate to the heat spreader better than a ~161mm^2 4c IB due to the 25% decrease in area on IB.
Click to expand...

this sums it up well, the chip running hotter doesn't mean it's using up more power,, it just has more Heat to dissipate per square mm vs sandy bridge, at the same exact TDP, IB will run hotter than SB, but both should produce the same amount of heat.
 
Last edited:
Total heat dissipation for the entire chip should be lower than 32nm, correct? Even if it's being reported at 80 C. directly on the die, the chip as a whole shouldn't be getting any hotter on the surface where the heat sink does its work.

The Tri-gate transistors are supposed to be more efficient than the planar ones. I'm not entirely convinced the measurements are correct. I've never seen a greater higher heat output for the same speed chip with a smaller fab process. The maximum TWP of these chips is supposed to be less than Sandy Bridge as well -- so with all things considered, if the chip is doing the same amount of work as a Sandy Bridge, how can it be putting out more heat with a lower TWP?
 
aphexcoil said:
so with all things considered, if the chip is doing the same amount of work as a Sandy Bridge, how can it be putting out more heat with a lower TWP?
Click to expand...

Temperature and power/heat flux are not the same thing. As an example, if you're pushing 100W through a 1in^2 area, vs 100W through a 1cm^2 area, which is going to run hotter? Obviously the 1cm^2 area, but they are still doing the same amount of work.
 
aphexcoil said:
Total heat dissipation for the entire chip should be lower than 32nm, correct? Even if it's being reported at 80 C. directly on the die, the chip as a whole shouldn't be getting any hotter on the surface where the heat sink does its work.

The Tri-gate transistors are supposed to be more efficient than the planar ones. I'm not entirely convinced the measurements are correct. I've never seen a greater higher heat output for the same speed chip with a smaller fab process. The maximum TWP of these chips is supposed to be less than Sandy Bridge as well -- so with all things considered, if the chip is doing the same amount of work as a Sandy Bridge, how can it be putting out more heat with a lower TWP?
Click to expand...

it's not putting out more heat, it's having a harder time moving the heat it does produce off of the chip, and into the heatsink.
 
mzs_biteme said:
I little bit better result here... 5Ghz with under 80deg temps, cooled with fan-less heatsink...!
Click to expand...

That looks like much more better. The kind of results I have been anticipating.

Looks like this cpu is going to be very fast.
 
Probably the die just isn't making good contact with the IHS. This is an engineering sample.
 
Nah these guys know they're shit, they mounted and re mounted the heatsink and im sure if the surface was un level they would've lapped it.
 
ShuttleLuv said:
Nah these guys know they're shit, they mounted and re mounted the heatsink and im sure if the surface was un level they would've lapped it.
Click to expand...

It was a loaner ES that needed to be returned to Intel. My guess is that Intel had it designed to only operate at stock clocks and not for over clocking.
 
The page I read the guy explicity stated he mounted and re mounted the hs at xs.
 
ShuttleLuv said:
Nah these guys know they're shit, they mounted and re mounted the heatsink and im sure if the surface was un level they would've lapped it.
Click to expand...

I was talking about inside the chip case, between the processor die and the Integrated Heat Spreader (the top of the CPU case), not the CPU cooler.
 
Last edited:
extide said:
Looks like that thread got blackholed. What was the skinny?
Click to expand...

Guy worked for marketting and had an engineering sample that he was able to bench and OC. It was reporting some serious heat and no one knew why. It was able to hit 5ghz on air but temps were over 100c. The guy didn't really get to test it properly since he lacked the hardware so it was just a few benchmarks and heat tests.
 
GeneO said:
Marketing don't know jack about hardware, no surprise.
Click to expand...
You'd probably be surprised to learn that marketing in large tech companies often has very technical groups/departments. ;) I used to work for one and had access to all kinds of unreleased computer hardware, available for check out like in a library.

The IB ES poster is just showing what software is reporting, and as pointed out earlier in the thread, there could be many issues to cause the readings shown. I wouldn't take it as gospel. I do expect to see the regular whining about temps when they do come out though, regardless of whether it matters or not. People seem to like the below ambient temps reported with air cooling, so we're usually not dealing with reality, just what people want to believe. ;)
 
No I wouldn't be surprised - you know I was kidding right? I followed the thread before it was shut down. he was using a good air cooler and knew what he was doing. Something wrong in the sample or the reporting.
 
SonDa5 said:
It was a loaner ES that needed to be returned to Intel. My guess is that Intel had it designed to only operate at stock clocks and not for over clocking.
Click to expand...

all depends on the sample they got. intel releases samples that are meant to be run at higher clocks, some have lower latency L1/2/3 cache, some are for low voltage, some are for low temps. they are all essentially the same processor they just cherry pick through X amount of processors, label them and then thats how they are sent out for specific testing. they should all be overclockable, just some may not overclock higher then others and some may not perform as well as others at the same clocks.



GeneO said:
No I wouldn't be surprised - you know I was kidding right? I followed the thread before it was shut down. he was using a good air cooler and knew what he was doing. Something wrong in the sample or the reporting.
Click to expand...


or intel is just having trouble getting the heat to transfer from the chip to the actual heat spreader. which really wouldn't surprise me given how small the cores are. it could explain why the i7 2500k replacement is only a whopping 100mhz faster, but all we can do is wait and see what the temps are with retail chips. but i'm not ready to write it off as a reporting error/bad sample just yet.
 
Last edited:
Intel wouldn't dish out a loaner ES that misrepresents the lot to review sites and certainly not if it ran that hot. Ivy will almost certainly run hotter than SB despite the decrease in TDP due to IB's shrink in die size, allowing for less surface area to transfer that heat. I think any hopes of achieving 5ghz+ clocks on IB will require some serious water cooling and, by the looks of the reviews, quite a bit of voltage as well.

Anand's review of Kepler looks to have the same issue: small powerful die = lots of localized heat. And much like IB, great perf-per-watt but heat... oh boy the heat...
 
Pelo - thats to be expected though, you just gotta get around that heat somehow.
 
Probly because "really fine" 22nm process of the CPU die, the cores internally not being able to dissipate the heat quick enough. That's why @ 5Ghz with added Vcore running prime 95 most IB chip will throttle down quick even when the heatsink cooler is barely warm.

Unless peep go xtreme sub ambient cooling:)
 
Last edited:
Monitoring software has been able to read TJ Max directly from each core of the CPU since the first Core i arrived. The sky high core temperatures from the next generation have nothing to do with the wrong TJ Max being used. The smaller CPUs get, the more difficult it becomes to transfer that heat away from the core.
 
You must log in or register to reply here.
Back
Top