Hardly. Those are too expensive and Intel is aware of the general public, I am, so they must be too. Only hardcore enthusiasts with money to throw away are gonna be ready for that. It's not a major market.

I think we are talking simple cost saving here, with a bit of background thinking.

 

I don't know, the 3820 is cheaper than a 3770k and the 3930k is only a couple hundred more... Certainly the 3960x is in a differently league, though.

This isn't an architectural problem inherent to Ivy Bridge. it is a manufacturing issue which is easy to solve. Even if Intel chooses not to solve it, the user can solve it on their own by de-lidding the CPU.

 

Until some IVB owners actually do delid their chips it's only speculation that the thermal paste interface is adding significantly to temperatures. It might make a big difference or it might make a small difference, we can't know for sure until people are game enough to try alternatives. Personally I reckon it's a combination of the new process and the TIM. But mostly the 22nm process. Not that I wouldn't like to be proven wrong in this instance...

 

Cheers,

 

Mike

I don't know, the 3820 is cheaper than a 3770k and the 3930k is only a couple hundred more... Certainly the 3960x is in a differently league, though.

 

Not even thinking 3820 lol. 3930K is anything I would be considering, and its some double price, and x is a hardcore, as always. So, yes, quite more expensive.

Until some IVB owners actually do delid their chips it's only speculation that the thermal paste interface is adding significantly to temperatures. It might make a big difference or it might make a small difference, we can't know for sure until people are game enough to try alternatives. Personally I reckon it's a combination of the new process and the TIM. But mostly the 22nm process. Not that I wouldn't like to be proven wrong in this instance...

 

Cheers,

 

Mike

 

I'm not sure why the site that posted this finding didn't post any test results, but I fully believe that bog-standard TIM has nowhere near the thermal conductivity of the flux-less solder previously used. I'm with you though, definitely want to see some results.

Someone did de-lid an IB CPU and the temps didn't drop.

 

http://www.overclock...red-without-ihs

 

I still haven't heard an explanation of how heat is transferred out of the center of the 3D tri-gate matrix. I have long wondered if limitations in thermal conduction from the interior regions of the 3D "brick" would prove to be the limiting factor for tri-gate processors. And I'm still wondering...

Someone did de-lid an IB CPU and the temps didn't drop.

 

Thanks Bob,

 

Interesting read. I certainly was one of the SBers mentioned in the thread that was hoping for extremely high OCs.

 

Kind regards,

Someone did de-lid an IB CPU and the temps didn't drop.

 

http://www.overclock...red-without-ihs

 

I still haven't heard an explanation of how heat is transferred out of the center of the 3D tri-gate matrix. I have long wondered if limitations in thermal conduction from the interior regions of the 3D "brick" would prove to be the limiting factor for tri-gate processors. And I'm still wondering...

 

Two problems:

 

1) no shim (not enough pressure can be applied when mounting the HSF so proper thermal transfer can't occur)

2) the HSF contact plate hasn't been lapped (same problem as above, but also not enough contact area)

 

We're not quite there yet on this one... Someone needs to lap an HSF and shim their IB, then test again.

 

It simply isn't reasonable to expect power consumption to have dropped significantly while temperatures have gone up significantly, except in the case of a thermal dissipation issue - physics doesn't work like this.

 

Furthermore, when overclockers report high core temps but low heatsink temps, the thermal transfer issue is illustrated further.

Two problems:

 

1) no shim (not enough pressure can be applied when mounting the HSF so proper thermal transfer can't occur)

2) the HSF contact plate hasn't been lapped (same problem as above, but also not enough contact area)

 

We're not quite there yet on this one... Someone needs to lap an HSF and shim their IB, then test again.

 

It simply isn't reasonable to expect power consumption to have dropped significantly while temperatures have gone up significantly, except in the case of a thermal dissipation issue - physics doesn't work like this.

 

Furthermore, when overclockers report high core temps but low heatsink temps, the thermal transfer issue is illustrated further.

 

I think thermal dissipation is likely to be precisely the issue...but not heat transfer to the heatsink, but rather internal heat dissipation within the 3D structure of the tri-gate transistor matrix. Nothing yet has assuaged my concerns about this potential problem. I have yet to see a plausible exposition on how the physics of heat transfer out of the 3D tri-gate core might work.

The 22nm chips will be limited in number of volts compared to larger processes.

 

Isn't it more about relative voltage tolerance? SB certainly held its own against larger processors vis a vis voltage. One could argue it theoretically should take less absolute volts to over volt the smaller process compared to the larger process, so your statement above doesn't really address the issue specifically. I think another important question is how tri-gate will compare to non-tri-gate as far as relative voltage tolerance is concerned.

 

I think thermal dissipation is likely to be precisely the issue...but not heat transfer to the heatsink, but rather internal heat dissipation within the 3D structure of the tri-gate transistor matrix. Nothing yet has assuaged my concerns about this potential problem. I have yet to see a plausible exposition on how the physics of heat transfer out of the 3D tri-gate core might work.

 

I have wondered about this all along as well. On the other hand, it's a shorter distance to the heat exchange medium, but there also is more heat generation density, so you could be right. It sure appears the direction Intel has decided to move towards is for mobile applications. The demand for high performing desktops keeps shrinking, right along w/ the die size. Their heavy emphasis on the GPU, built in memory controller, and finally SoC betrays this. The only real solution for FS enthusiasts lies in new software that can utilize more cores, DX11, etc, i.e., less reliance on raw clock speed. This is the only direction compatible w/ the apparent current trajectory of hardware development. Can't see how Haswell will change this, unless of course the thermal issue is hugely resolved, in which case their might be another upgrade cycle left for FSX users.

 

Noel

Ivy Bridge's poor voltage scaling and high temperatures aren't because of the thermal paste solution with core/HS contact. This has been also proven cooling the processor without HS using shim and results didn't improve as it was also pretty much expected. This speculation about poor thermal conductivity has no basis in reality whatsoever. Processor has too much leakage current in its new trigate transistors causing poor voltage scaling and small surface area further worsens ability to remove heat from the core. No paste or soldering is going to change that.

 

It's true, of course, that earlier solution to use soldering offers probably better conductivity between core and HS, but it has no true effect on overclocking ability in general: difference is too small and it can't overcome the effects of the processor architecture. This is also probably the reason why Intel chose TIM instead of soldering and using TIM they could probably cut costs a bit. Ivy Bridge has had its own problems during the developement, such as bad yields due the new architecture and Intel probably welcomes every possibility to cut corners where things actually don't matter.

 

Some enthusiasts seem to be grasping straws and hoping some miracle that Ivys will clock at or better than Sandys. And frankly, I don't see the point. Ivy Bridge is still damn good processor and you should be able to push it around 4.5-4.6GHz, where it can still compete in many applications with Sandys clocked near or at 5GHz. That is not bad and in the end just MHzs don't count, at least I don't care how the performance comes, with IPC improvement or higher clocks. And we also have to remember that not every Sandy will go 5GHz at least easily.

 

Ivy Bridge is just Intel's tick-phase die shrink update to older architecture and as such it isn't bad. This overclocking/heat buzz is just getting too big attention in internet forums nowadays IMO.

Ivy Bridge is still damn good processor and you should be able to push it around 4.5-4.6GHz

 

I have to say that I agree. The new IB processors are still excellent over-clockers. They just can't take as much voltage as the older Sandybridge chips.

 

From what I understand, the temps are okay as long as you stay below 1.3volts. And for most IB chips, 1.3volts is within 4.5-4.7GHZ.