Rewind back to the launch of Intel’s 3rd generation processor, Ivy Bridge. Enthusiasts and gamers who overclock the top end model were a little disappointed with the maximum CPU frequency that could be achieved. What used to be a 5.0-5.2 GHz ceiling on Sandy Bridge became 4.8-5.0 GHz on Ivy Bridge, and there was more of a silicon lottery as to whether the user got a good overclocking chip or not. Then move forward to Haswell, and the situation compounded in a similar way. 4.6 GHz to 4.8 GHz became the norm on Haswell, a 200 MHz drop which absorbed any IPC increase the new architecture had.

One of the reasons that enthusiasts attributed to this decrease in overclocking was the thermal interface material between the die on the package and the heatspreader. Various experiments by enthusiasts with removing the heatspreader and replacing the TIM inside the CPU to provide better contact and heat transfer characteristics. The adventurous users even used metallic binding material, but the benefits were published for all to see – depending on how good (or bad) the CPU was in the first place, up to a 20ºC drop was seen for overclocked processors at full load. 

As part of the Intel press conference call this week, it was announced that a new Haswell processor will come to market to help adjust some of these issues. This new processor will be a fully unlocked Haswell core, codename ‘Devil’s Canyon’ and should be out mid-year (note, Computex is in June). The features that Intel is promoting are the improved thermal interface material, an updated packaging and 9-series chipset support.

‘Improved TIM’ is a little ambiguous – whether this means a more substantial amount of goop under the heatspreader or a proper metalling binding to the heatspreader I do not know, although I have asked Intel. Updated packaging could lead to a thicker/thinner heatspreader, focusing more on the hot areas of the die, or a different binding agent between the PCB and the heatspreader. It may turn out easier to remove the heatspreader if needed, or harder. The final comment had me a little confused, saying 9-series support but not saying anything about the 8-series. I would assume that as a Haswell processor this new SKU would have 8-series support, although Intel have told us that this new CPU is not targeted at 8-series.

There is no word if this is going to be placed at the same frequency as the i7-4770K or the i7-4771/4790, the name or the rated TDP. Intel will let us know nearer to release, and as soon as we get this information and are able to share with you, we will.



Comments Locked


View All Comments

  • Ryan Smith - Wednesday, March 19, 2014 - link

    It'll be interesting to see if Intel goes with the solder based TIM here, as the information implies. The paste has been fine for stock CPUs and I don't really expect to see solder TIM CPUs to overclock much better (see our IVB-E), but having lower temperatures would still be nice.
  • Homeles - Thursday, March 20, 2014 - link

    Haswell appeared to suffer from both the die's distance from the IHS, and the continued use of thermal paste instead of solder. Perhaps Intel worked on both issues.
  • chizow - Thursday, March 20, 2014 - link

    Exactly right, I have two 4770K rigs and never bothered to de-lid either of them because the results I have seen around the web post-delid seem to indicate no change in max overclock, just a pronounced 15-20C drop in operating temps under load. The main problem though with Haswell's high operating temps with the standard TIM is that it prevents you from really stress testing an overclock because the chip will start throttling.

    In any case, I am happy with both my 4770Ks, especially the 2nd one from Costa Rica. Hits 4.4GHz at 1.18V while the 1st one I bought at launch from Malay only does 4.2 at 1.24V. It can go up to 4.3 but requires 1.28V and temps jump up a ton, so not worth it. Didn't want to wait for this updated Haswell-K because of the new chipset requirement that isn't much of an upgrade and I needed a 2nd machine to replace my wife's aging X58 rig that was starting to have some EOL type issues.
  • munim - Wednesday, March 19, 2014 - link

    Knowing Intel's R&D budget, and considering I have already spend around $400 CAD on a 4770K, I'm irked. However, I'm happy that we made enough noise about it so as to lead to this development.
  • MikeMurphy - Wednesday, March 19, 2014 - link

    Poor TIM requires beefier (more expensive) cooling systems which is a cost burden to the OEMs. In some cases, like many ultrabooks, the CPU speeds are even throttled despite such cooling. It's unforgivable that Intel has been doing this since IVB.
  • extide - Wednesday, March 19, 2014 - link

    In laptops/ultrabooks there is no heatspreader, so this issue doesn't really apply to them.
  • Homeles - Thursday, March 20, 2014 - link

    Laptops don't have heatspreaders.
  • willis936 - Wednesday, March 19, 2014 - link

    The TIM quality myth was busted last year. There's a post floating around where someone did some science to show that IHS-die clearance is a much bigger factor than TIM choice. Delidding does increase OC headroom quite a bit. I got an extra 300MHz out of my 4770k. The heatsink is still cool to the touch even when the CPU is 70 degrees but you can't change the fact that dies are getting smaller and power consumption is staying roughly the same.

    A soldered die would be good to see because people started complaining when they moved away from that. Though even with a delidded CPU I only get 4.5GHz out of my 4770k. 4.5 without delidding is lucky with 4.8 being a good standard deviation above. Your haswell estimates are a little optimistic but engineering samples are always good chips so if that's all you have to go by that's fine. Also the OC headroom variance seems a lot bigger than ever with haswell. Hopefully these chips are all good picks.
  • V3ctorPT - Thursday, March 20, 2014 - link

    My 4770K can get up to 4.8Ghz until he starts to throttle down... I use it at 4.6Ghz, I never delidded it, with afraid of ruin it. And it's faster than my old 2600k@5Ghz.
    What I'm little upset is that the "new" cpu that has good TIM is only supported by the 9 series chipset. Are they kidding? It's the same socket for the Z87 and Z97, why does that cpu only work in a Z9 series?? $$$$ grabers...
  • willis936 - Thursday, March 20, 2014 - link

    Yeah that's a little sad. I might flip a 4770k to upgrade to one of these chips for a few bucks but I'm definitely not flipping a motherboard. But hey if they don't want my money I'm sure they're fine with that.

Log in

Don't have an account? Sign up now