Socketed LGA1150 Intel Broadwell-K to be released in Q4 2014

[Fjodor2001]

http://www.techspot.com/news/54404-...broadwell-k-chips-to-arrive-in-late-2014.html

The chip in question is expected to be based on the LGA 1150 socket although it may still require a new motherboard due to a change in Intel’s power specifications for the socket. Specifically, we are told the new requirements involve a different power supply for VCCST and a V_PROC_IO for 1.05V and a new chip topology that requires a modified THRMTRIP output buffer.

http://www.legitreviews.com/intel-2014-desktop-processor-roadmap-leaked-broadwell-k-socketed_126888

The most interesting this on the slide is the introduction of Broadwell-K processors in late 2014. Details are sketchy, but from what we gather Intel’s 9-series chipsets will support both Broadwell (BRW) and Haswell Refresh (HSW-R) microprocessors, but will lack support of current-generation Haswell processors. This is due to the fact that the 4th Generation Core i processors have an integrated voltage regulator (Intel calls this FIVR: Haswell’s Fully Integrated Voltage Regulator.) It appears that Intel has moved this back off-die and back onto the motherboards, in an attempt to reduce the heat produced by the processor. This means that those building Intel Haswell platforms today with the 8 series chipset could end up with no upgrade path when it comes to the processor. This is certainly not good for enthusiasts that always like to have an upgrade path, especially on a brand new socket.​

NOTE: Parts of this info has already been mentioned at the end of this thread , but I figure the subject is important enough to mandate its own thread. Also, the links above contain some additional info compared what is already published in the other thread.​

 

[Khato]

I find the speculation that Haswell Refresh and Broadwell-K would remove the FIVR quite difficult to believe. I could maybe see it if there was a socket change, but without one? No, that just doesn't make any sense whatsoever. Now a power change as mentioned in the techspot article wouldn't be quite as surprising - it's typically the case that there are subtle changes that require new motherboards to support, but they've always been backwards compatible. (As in, new motherboards would still work with original Haswell.)

 

[Vesku]

If they are only releasing the K models as LGA and not anything else then I think that is probably the last hurrah or getting perilously close (1 more LGA K series after Broadwell perhaps). Those low core count E series will then lose their oddball status.

 

[Pheesh]

does sound like it would be a non backwards compatible socket though- they would have to make it different to avoid mixing. That's really interesting that they would be moving it off-die already.

 

[Techhog]

If they are only releasing the K models as LGA and not anything else then I think that is probably the last hurrah or getting perilously close (1 more LGA K series after Broadwell perhaps). Those low core count E series will then lose their oddball status.

Don't jump the gun. I doubt that they'll give up desktop quite that soon. It seems more likely that the 14nm delay is the cause of this. There's no way that Intel's desktop sales have already dropped so much that they've decided to completely exit the market in 1-2 years. They'd at least need to make a dent in the tablet market before that's even worth considering.

 

[ShintaiDK]

Don't jump the gun. I doubt that they'll give up desktop quite that soon. It seems more likely that the 14nm delay is the cause of this. There's no way that Intel's desktop sales have already dropped so much that they've decided to completely exit the market in 1-2 years. They'd at least need to make a dent in the tablet market before that's even worth considering.

Besides the Broadwell-K and E3 Broadwell Xeons, the next 14nm product for desktop is 14nm Skylake. Desktop is simply last in the chain for a new process node. All mobile and servers comes first.

 

[mrmt]

Don't jump the gun. I doubt that they'll give up desktop quite that soon. It seems more likely that the 14nm delay is the cause of this. There's no way that Intel's desktop sales have already dropped so much that they've decided to completely exit the market in 1-2 years. They'd at least need to make a dent in the tablet market before that's even worth considering.

They will not exit the desktop market anytime soon, if anything because this is a market that nobody else can fill. That said, it's obvious that the desktop will be trailing the mobile line for the times to come.

No wonder Rory Read pointed desktop as a growth opportunity for AMD. On that market, not only AMD handicaps in power management are less pronounced, it will get less attention from Intel.

 

[sm625]

Why would they move the FIVR off die. Talk about stupid.

 

[_Rick_]

Why would they move the FIVR off die. Talk about stupid.

Thermal limitations kind of makes sense. Considering the high thermal density of even 22nm CPUs, there's plenty of reason to reduce heat (which promotes further heat buildup) by separating two hotspots.

A sensible choice for the desktop and the 1150 Xeons would be to place the voltage regulation hardware on the CPU package, but not onto the die - that way it would remain an "easy" update.

I suppose especially outside of mobile, where more heat is being generated, the integrated VR may not be working as well as intended.

 

[Techhog]

Besides the Broadwell-K and E3 Broadwell Xeons, the next 14nm product for desktop is 14nm Skylake. Desktop is simply last in the chain for a new process node. All mobile and servers comes first.

Sounds about right to me.

 

[ShintaiDK]

Why would they move the FIVR off die. Talk about stupid.

Thermal limitations kind of makes sense. Considering the high thermal density of even 22nm CPUs, there's plenty of reason to reduce heat (which promotes further heat buildup) by separating two hotspots.

A sensible choice for the desktop and the 1150 Xeons would be to place the voltage regulation hardware on the CPU package, but not onto the die - that way it would remain an "easy" update.

I suppose especially outside of mobile, where more heat is being generated, the integrated VR may not be working as well as intended.

They wont. The article is simply wrong and shows the usual lack of any kind of research.

To add to the stupidity. Haswell Refresh is a platform update only. And you can use your 4770K in a 9 series board. just like Broadwell-K. So if the FVIR was removed..how would that work? It wouldnt!

But again, the Broadwell motherboard routing/implementation guides should already have removed any doubts in the first place.

The only changes with Broadwell besides IGP and the shrink is these instructions:
ADOX/ADCX/MULX for improving performance of arbitrary-precisionn integer operations
RDSEED to generate 16-, 32- or 64-bit random numbers according to NIST SP 800-90B and 800-90C
PREFETCHW instruction

 

[StrangerGuy]

Prediction: another tiny IPC and clock increase, another tiny drop in wattage and yet another AVX set that next to nothing would use. IHS solder remains unfixed. Yaaaaaaawn for desktops even without a mandatory mobo change.

While the rest of the world will be much more interested in the 14nm Atoms.

 

[Fjodor2001]

Prediction: another tiny IPC and clock increase, another tiny drop in wattage and yet another AVX set that next to nothing would use. IHS solder remains unfixed. Yaaaaaaawn for desktops even without a mandatory mobo change.

While the rest of the world will be much more interested in the 14nm Atoms.

Did you notice the 95 W TDP? If Broadwell-K was just a node shrink of a 4770K it would only have around 65 W TDP or so at 14 nm (down from 84 W TDP @ 22 nm). So if it's at 95 W then they have ~30 W more to play with. Possibly that means more CPU cores, a seriously beefed up iGPU / eDRAM, or something like that.

 

[ShintaiDK]

Did you notice the 95 W TDP? If Broadwell-K was just a node shrink of a 4770K it would only have around 65 W TDP or so at 14 nm (down from 84 W TDP @ 22 nm). So if it's at 95 W then they have ~30 W more to play with. Possibly that means more CPU cores, a seriously beefed up iGPU / eDRAM, or something like that.

Those 95W wont get more right the more you post it. Its simply a platform design TDP. And note the 1 next to 95W? Guess those notes below would be good to have now...

 

[Fjodor2001]

Those 95W wont get more right the more you post it. Its simply a platform design TDP. And note the 1 next to 95W? Guess those notes below would be good to have now...

I agree that those notes would be good to have, but we don't.

However in previous similar slides published for IB/H, the TDP specified has meant the TDP of the top end CPU model (e.g. 4770K). Is there any reason to believe that should not be the case this time?

 

[Khato]

However in previous similar slides published for IB/H, the TDP specified has meant the TDP of the top end CPU model (e.g. 4770K). Is there any reason to believe that should not be the case this time?

I wasn't able to find similar slides for IVB/HSW with a quick search, mind sharing some?

Only similar slide I found was the previous desktop roadmap which lacked BDW altogether - http://www.dailytech.com/Report+Int...les+Haswell+Refresh+for+2014/article31770.htm - and it still had the same 95W-35W range for both consumer and business segments on Haswell and Haswell Refresh. Such means that even if the 95W top end does mean something it doesn't necessarily apply to BDW.

 

[Fjodor2001]

I wasn't able to find similar slides for IVB/HSW with a quick search, mind sharing some?

Only similar slide I found was the previous desktop roadmap which lacked BDW altogether - http://www.dailytech.com/Report+Int...les+Haswell+Refresh+for+2014/article31770.htm - and it still had the same 95W-35W range for both consumer and business segments on Haswell and Haswell Refresh. Such means that even if the 95W top end does mean something it doesn't necessarily apply to BDW.

All the TDP numbers specified in that slide are max TDP for the top end CPUs. E.g. for Ivy Bridge-E it's 130-140W, which matches the actual TDP of the Ivy Bridge-E 4960X.

Possibly the 1) in the slides is used to indicate that the 95 W TDP is only needed for Haswell Refresh LGA1150 motherboards that want to be forward compatible with Broadwell-K CPUs, but it's not really needed for Haswell/HaswellRefresh CPUs. But nobody can be sure of course without more specific info.

 

[Khato]

Possibly the 1) in the slides is used to indicate that the 95 W TDP is only needed for Haswell Refresh LGA1150 motherboards that want to be forward compatible with Broadwell-K CPUs, but it's not really needed for Haswell/HaswellRefresh CPUs. But nobody can be sure of course without more specific info.

While I'd like to say why that isn't the case it isn't exactly possible with information that's been released (either intentionally or not) to the public domain. That previous roadmap is as close as I can get, but that's merely circumstantial evidence that Broadwell wasn't planned for a desktop release due to its lack of presence on the roadmap.

 

[Fjodor2001]

While I'd like to say why that isn't the case it isn't exactly possible with information that's been released (either intentionally or not) to the public domain. That previous roadmap is as close as I can get, but that's merely circumstantial evidence that Broadwell wasn't planned for a desktop release due to its lack of presence on the roadmap.

Hard to say why Broadwell-K was not on that roadmap. Maybe that roadmap had an intentionally restricted timeline length. I'm sure Intel have plans/roadmaps for Skylake too, but it does not show up in the latest roadmap in the OP either.

But if you have info saying otherwise, please share it.

 

[Techhog]

I hope it's not a 95W TDP. If it is, you'll need a golden chip and enthusiast class cooling to get past 4GHz most likely. :/

 

[FlanK3r]

What about Haswell refresh? Some improvements for cold OC?

 

[mikk]

Is there any reason to believe that should not be the case this time?

It's meaningless. TDP isn't specified yet, at least for Broadwell and furthermore in early Roadmaps for Haswell and Ivy Bridge 95W was inaccurate.

 

[ShintaiDK]

What about Haswell refresh? Some improvements for cold OC?

Its a platform refresh. Not a CPU refresh. If lucky we just get a speedbin.

 

[Fjodor2001]

It's meaningless. TDP isn't specified yet, at least for Broadwell

You say the TDP isn't specified, yet it's specified in the slides?

If you say the TDP is not for Broadwell, then I guess you mean it's for Haswell Refresh? So the Haswell Refresh CPUs would not be possible to use on existing 84 W TDP Haswell motherboards. I.e. the 95 W TDP motherboards would only be needed for the Haswell Refresh CPUs sold in the short timeframe between the release of Haswell Refresh and Broadwell-K?

 

[ShintaiDK]

You say the TDP isn't specified, yet it's specified in the slides?

If you say the TDP is not for Broadwell, then I guess you mean it's for Haswell Refresh? So the Haswell Refresh CPUs would not be possible to use on existing 84 W TDP Haswell motherboards. I.e. the 95 W TDP motherboards would only be needed for the Haswell Refresh CPUs sold in the short timeframe between the release of Haswell Refresh and Broadwell-K?

There is a note you admit not knowing what is. Haswell and Ivy bridge also showed as a 95W max platform, yet it was not 95W TDP CPUs.

There is no Haswell Refresh CPUs either, its Haswell CPUs.