Combined roadmaps for Broadwell and Haswell Refresh

[ShintaiDK]

Both Broadwell and Haswell Refresh will be updated to the 9 series chipsets with SATA Express support among other things.

 

[NTMBK]

Nice find. I wish that people wouldn't cut off the bottom and top of these slides though!

Interesting that there is not only an i7-4960HQ, but also >=4960HQ. Are we going to see even faster Iris Pro graphics in notebooks?

(And Broadwell Iris Pro is going to be delicious.)

 

[ShintaiDK]

Remember having >= doesnt mean there will be a product. Just that there might be an option for one.

 

[SiliconWars]

Erm, Broadwell 2H 2014?

 

[Soulkeeper]

this sata express stuff should be interesting
Any specific mention of it being based off pcie 2.0 or 3.0 ?

 

[SOFTengCOMPelec]

Any mention yet on when mainstream 6 core Intel processors will show the light of day ?

 

[jpiniero]

Erm, Broadwell 2H 2014?

I did see an article suggesting that Intel is still planning on releasing 14 nm Core models in the first half. That could be the missing Broadwell desktop BGA models, which makes sense considering they would be very low volume.

Any specific mention of it being based off pcie 2.0 or 3.0 ?

It's 3.0 with up to support of two lanes (2 GB/sec)

 

[jpiniero]

Any mention yet on when mainstream 6 core Intel processors will show the light of day ?

I could see Intel doing a 6 Core Atom processor eventually.

 

[SOFTengCOMPelec]

I could see Intel doing a 6 Core Atom processor eventually.

Ok, you got me there!

And it would be almost as fast as a dual core G1610, as long as you can use all 6 cores effectively.

 

[USER8000]

Erm, Broadwell 2H 2014?

It appears so. The new roadmap was originally mentioned here:

http://chinese.vr-zone.com/75335/in...2014-but-diy-pc-get-haswell-refresh-07252013/

It does make sense. Haswell has only been recently launched,and it would make no sense for it to be replaced in under a year.None of the recent Intel refreshes have been replaced in under a year.

 

[ShintaiDK]

Add 13 months to Haswell, and you are in H2 2014

 

[SOFTengCOMPelec]

It does make sense. Haswell has only been recently launched,and it would make no sense for it to be replaced in under a year.None of the recent Intel refreshes have been replaced in under a year.

It's expected to be just a slightly improved Haswell, to keep people happy until Skylake/skymont, because broadwell may not be coming to the desktop (unless that has changed, since I last heard about it).

 

[meloz]

Thank you ShintaiDK for sharing this.

The way these slides are designed, it looks like we will certainly get the Haswell Refresh before Broadwell. I have now officially given up any hope of buying a Broadwell desktop in 2014.

This "refresh" better bring with it some serious gains in iGPU, if nothing else. Otherwise many consumers would have waited a year to upgrade for no reason. Maybe the guys who bought Haswell this year are the smarter ones.

Erm, Broadwell 2H 2014?

Could mean two things:

1. Intel are experiencing some manufacturing issues with 14nm.
   
2. The focus is on lower power SoC process and Airmont. Intel are willing to delay notebook bound Broadwell to push Airmont out first. Although all information until now has been that Broadwell will be the first 14nm product, so who knows.

2H 2014 could mean anything, a December launch even. Or July. They have given themselves some timetable flexibility rather than nailing the launch down to a specific quarter.

 

[NTMBK]

The way these slides are designed, it looks like we will certainly get the Haswell Refresh before Broadwell. I have now officially given up any hope of buying a Broadwell desktop in 2014.

"Before Broadwell"? Try instead of Broadwell. Expect the LGA desktop to go Haswell->Haswell Refresh->Skylake.

Look at Ivy Bridge, and the amount that people moaned that on the desktop it wasn't worth upgrading over Sandy Bridge because it only saved a bit of power and didn't overclock as well. Why would Intel bother? They may as well keep their 22nm fabs active for longer, and dedicate 14nm capacity to the mobile products where power savings actually translate into better sales.

This "refresh" better bring with it some serious gains in iGPU, if nothing else.

That's what I'm hoping. I want Intel to release the SKU that they never did with Haswell round 1- LGA, fully unlocked overclocking, with Crystalwell L4$ and GT3 graphics. That massive cache alone would make it a pretty decent improvement over the 4770k- there are already a handful of benchmarks where the mobile quad-core with Iris Pro beats the 4770k due to the massive cache.

http://techreport.com/review/24879/intel-core-i7-4770k-and-4950hq-haswell-processors-reviewed/13

(Note- though this says "OpenCL", it is OpenCL which is running on the CPU cores and not the IGP.)

 

[USER8000]

"Before Broadwell"? Try instead of Broadwell. Expect the LGA desktop to go Haswell->Haswell Refresh->Skylake.

http://wccftech.com/xeon-phi-hpc-codenamed-knights-corner-revealed/

It looks like Broadwell(I assume for server) is arriving in 2015.

 

[ShintaiDK]

http://wccftech.com/xeon-phi-hpc-codenamed-knights-corner-revealed/

It looks like Broadwell(I assume for server) is arriving in 2015.

Yep, its all Xeons.

Skylake in 2015 for desktop:

 

[meloz]

So they will make a BGA notebook oriented Broadwell, and eventually a large-socket version for servers / workstations. But they will not make it for desktop. Us desktop guys will have to make do with a Haswell refresh before straight jump to Skylake.

I do not know if desktop has truly become this irrelevant as yet, but Intel are certainly trying their best to make it happen. This kind of inequitable attitude towards desktop products is unlikely to motivate consumers and future PC sales. Thus it only further serves to 'prove' the insignificance of desktop. What a wonderful trap of confirmatory bias and self-fulfilling prophecy Intel have built themselves here.

 

[NTMBK]

So they will make a BGA notebook oriented Broadwell, and eventually a large-socket version for servers / workstations. But they will not make it for desktop. Us desktop guys will have to make do with a Haswell refresh before straight jump to Skylake.

I do not know if desktop has truly become this irrelevant as yet, but Intel are certainly trying their best to make it happen. This kind of inequitable attitude towards desktop products is unlikely to motivate consumers and future PC sales. Thus it only further serves to 'prove' the insignificance of desktop. What a wonderful trap of confirmatory bias and self-fulfilling prophecy Intel have built themselves here.

Intel's big plan seems to be something like this:

-move all low-end desktops to BGA chips, preferably in something compact like an AIO, or a NUC.
-move the high-end desktops to the "enthusiast" platform derived from Xeons

It means your ma-and-pa PCs get more compact and power efficient, while we're forced to wait an extra year for the new hotness.

 

[ViRGE]

Aww crud. I don't want to go back to ribbon cables.

 

[Sweepr]

So Haswell-U/Y got pushed from (probably) late H1 to (probably) early H2 just like its other mobile cousins were planned all along. Huh big deal.

 

[ShintaiDK]

Aww crud. I don't want to go back to ribbon cables.

I am not exactly thrilled either to put it mildly. I wonder how the SATA org came to this conclusion.

"Huge" socket for either 2 SATA or a single SATA Express that can be x1 or x2.

 

[ViRGE]

I am not exactly thrilled either to put it mildly. I wonder how the SATA org came to this conclusion.

"Huge" socket for either 2 SATA or a single SATA Express that can be x1 or x2.

Obviously it was based on backwards compatibility. You can't roll out SATAe and immediately discontinue SATA support at the same time. And motherboards certainly aren't going to put both connectors on the board at a time when motherboards keep on getting smaller.

 

[24601]

2x PCIe 3.0 is pretty worthless (Even worse if it's 2x PCIe 2.0)

y no 16x (or at least 4x) PCIe 3.0 for cheaper superfast SSDs? (And finally have a real alternative to ram-cacheing)

1x PCIe 2.0 = 500 MB/s (5 GT/s)
1x PCIe 3.0 = 985 MB/s (8 GT/s)

 

[Idontcare]

So they will make a BGA notebook oriented Broadwell, and eventually a large-socket version for servers / workstations. But they will not make it for desktop. Us desktop guys will have to make do with a Haswell refresh before straight jump to Skylake.

I do not know if desktop has truly become this irrelevant as yet, but Intel are certainly trying their best to make it happen. This kind of inequitable attitude towards desktop products is unlikely to motivate consumers and future PC sales. Thus it only further serves to 'prove' the insignificance of desktop. What a wonderful trap of confirmatory bias and self-fulfilling prophecy Intel have built themselves here.

Its starting to look that way.

 

[blackened23]

So they will make a BGA notebook oriented Broadwell, and eventually a large-socket version for servers / workstations. But they will not make it for desktop. Us desktop guys will have to make do with a Haswell refresh before straight jump to Skylake.

I do not know if desktop has truly become this irrelevant as yet, but Intel are certainly trying their best to make it happen. This kind of inequitable attitude towards desktop products is unlikely to motivate consumers and future PC sales. Thus it only further serves to 'prove' the insignificance of desktop. What a wonderful trap of confirmatory bias and self-fulfilling prophecy Intel have built themselves here.

Oh, please. What's the alternative here? They could release a desktop broadwell only for this forum to be filled with whiners (and/or die hard AMD fans who would never buy intel anyway ) complaining of a mere 10% ipc increase? Yeah that sounds like a much better option, it will be Haswell all over again. The fact of the matter is that efficiency/mobility has replaced all-out IPC as the metrics that are most important to the market, and that is what Broadwell is focused on. If they released Broadwell for desktop, it would not be a dramatic step forward in terms of IPC. Granted, it would be a sizable increase - but desktop IPC is just not going to be a focus because desktop isn't selling. Period. The largest markets for these chips will be enterprise/data centers and mobile - and efficiency is of utmost importance for both of these areas.

Besides which, who cares? Haswell-E is being released next year with 8C as a baseline, and Intel will also be shuffling more "desktop" users to the enthusiast platform. This makes sense considering the direction of the market towards mobile devices. If you look closely at Intel's roadmaps, you'll find that fewer chips will be released on the "mainstream" platform while the price points of entry for the enthusiast platform will be lowering - consider that the 4820 hexa core CPU will be a fully unlocked "K" SKU and has a price range under 300$. Intel will likely be moving more of us over to the E platform and as far as i'm concerned, i'm okay with that - especially considering the prices for it will be lowering.