Sign in with
Sign up | Sign in
Your question
Closed

Report: Intel Broadwell-K CPU to Launch in Q4 2014

Tags:
  • Intel
  • CPUs
Last response: in News comments
Share
October 22, 2013 4:21:16 PM

So 12 months to wait for a 5% bump in speed.... Oh but the gpu you won't use is twice as powerful. Good times....
Score
22
a b å Intel
a c 146 à CPUs
October 22, 2013 4:30:46 PM

So, if there is an LGA1150 Broadwell, it will be coming out roughly a whole year after the mobile/AIO/etc. variants and not be compatible with current motherboards... and the roadmap naming seems to imply LGA1150 models will only be K-models, which means $220+ parts.

This is more or less what I was expecting.
Score
0
Related resources
October 22, 2013 4:37:49 PM

Where's the SATA Express and DDR4 support?
I'm waiting to push my SSD to 2GB/s.
Score
2
October 22, 2013 4:50:47 PM

Why is a rumor being brought up, that Intel already debunked? Intel stated that mainstream and enthusiast parts would remain on LGA. In fact, Tom's ran an article about it themselves...

I don't see why a "defect density issue" would delay a processor launch by a year....unless, of course, the engineers are just completely clueless on how to fix it. Which would be rather usual for Intel.
Score
-4
a b à CPUs
October 22, 2013 5:04:32 PM

natoco said:
So 12 months to wait for a 5% bump in speed.... Oh but the gpu you won't use is twice as powerful. Good times....


nobody upgrades for the CPU anymore, it is all about the chipsets these days. This one is obviously a minor refresh, but adding stream support for USB3, and moving RST to support PCIe Storage (which will be the precursor to SATA Express in the following gen) are worth getting at least a little excited about.

Also, with the TDP going up while overall platform continues to go down means that there is more and more motherboard stuffs being shoved onto the CPU die, but I guess we will have to wait and see what exactly that is.

But you are right. At the end of the day the focus here is to get X86 into more and more mobile devices rather than pushing the high performance devices. With new consoles coming out in a month, and true 'next gen' games starting to pop up on PC next year, I would bet that the late 2015 or early 2016 Skylake/mont refresh will have something special for those of us who upgraded with Sandy Bridge. At that point we are expecting to see full DDR4 support, SATA Express, the possibility of most of the northbridge moving to the CPU die, PCIe4, and possibly the introduction of 10,000 Ethernet for home users.
And on top of that we are expecting a renewed (for at least 1 generation) attention to high end performance rather than the little 5-10% bumps we have seen the last few generations.
Score
-2
a b à CPUs
October 22, 2013 5:10:05 PM

sykozis said:
Why is a rumor being brought up, that Intel already debunked? Intel stated that mainstream and enthusiast parts would remain on LGA. In fact, Tom's ran an article about it themselves...

I don't see why a "defect density issue" would delay a processor launch by a year....unless, of course, the engineers are just completely clueless on how to fix it. Which would be rather usual for Intel.


Right, ENTHUSIAST parts will remain on LGA. This means that if you are building a system then you are either stuck with a K series part which will run on LGA, or else you will get a motherboard with a BGA chip pre-installed. Presumably these BGA chips will have a pay-to-unlock scheme if you want extra cores, HT, upgraded graphics, etc.

People like myself who typically build higher end systems under a budget will be forced into a more expensive K sku rather than the non-k unit we would normally purchase. Not the end of the world, but it is a bit annoying. Thankfully the next gen will supposedly go back to normal again... but after that I am afraid the nonK chips will be locked to BGA forever.
Score
0
a c 127 à CPUs
October 22, 2013 5:45:23 PM

natoco said:
So 12 months to wait for a 5% bump in speed.... Oh but the gpu you won't use is twice as powerful. Good times....


If you want no GPU, then you go with the xX9 chipsets, not theZX8 chipsets. That's been pretty common since LGA1156.

milktea said:
Where's the SATA Express and DDR4 support?
I'm waiting to push my SSD to 2GB/s.


The first platform to have DDR4 is Haswell-EP, might not even make it to the Haswell-E platform but its hard to say.

For us it will probably come in the chips after Broadwell.

I am also excited for SATA Express but honestly, you can't tell the difference for the most part. The main benefit I can see is going to be a increase in IOPS.

CaedenV said:
natoco said:
So 12 months to wait for a 5% bump in speed.... Oh but the gpu you won't use is twice as powerful. Good times....


nobody upgrades for the CPU anymore, it is all about the chipsets these days. This one is obviously a minor refresh, but adding stream support for USB3, and moving RST to support PCIe Storage (which will be the precursor to SATA Express in the following gen) are worth getting at least a little excited about.

Also, with the TDP going up while overall platform continues to go down means that there is more and more motherboard stuffs being shoved onto the CPU die, but I guess we will have to wait and see what exactly that is.

But you are right. At the end of the day the focus here is to get X86 into more and more mobile devices rather than pushing the high performance devices. With new consoles coming out in a month, and true 'next gen' games starting to pop up on PC next year, I would bet that the late 2015 or early 2016 Skylake/mont refresh will have something special for those of us who upgraded with Sandy Bridge. At that point we are expecting to see full DDR4 support, SATA Express, the possibility of most of the northbridge moving to the CPU die, PCIe4, and possibly the introduction of 10,000 Ethernet for home users.
And on top of that we are expecting a renewed (for at least 1 generation) attention to high end performance rather than the little 5-10% bumps we have seen the last few generations.


10Gbe probably wont come to home users for a few years TBH. Its still pretty expensive and there isn't even a decently priced home router that can use it so having it on your NIC would be pointless as the router will be limited to 1Gbe.

CaedenV said:
sykozis said:
Why is a rumor being brought up, that Intel already debunked? Intel stated that mainstream and enthusiast parts would remain on LGA. In fact, Tom's ran an article about it themselves...

I don't see why a "defect density issue" would delay a processor launch by a year....unless, of course, the engineers are just completely clueless on how to fix it. Which would be rather usual for Intel.


Right, ENTHUSIAST parts will remain on LGA. This means that if you are building a system then you are either stuck with a K series part which will run on LGA, or else you will get a motherboard with a BGA chip pre-installed. Presumably these BGA chips will have a pay-to-unlock scheme if you want extra cores, HT, upgraded graphics, etc.

People like myself who typically build higher end systems under a budget will be forced into a more expensive K sku rather than the non-k unit we would normally purchase. Not the end of the world, but it is a bit annoying. Thankfully the next gen will supposedly go back to normal again... but after that I am afraid the nonK chips will be locked to BGA forever.


Again, Intel has already displaced this. There is nothing pointing towards this as the only thing with extra cores we have seen is a article showing how Intel plans to possibly add extra cores in the CPU as backups because as you get smaller, the chance of failure increases.

This is why rumors are so bad. Rumors spin out of control and keep getting pushed until the actual release.

As for the slides, anyone else notice it says PCIe 2.0 5GT/s? Why would the 9 series chipset (the next in line) go from PCIe 3.0 to PCIe 2.0?

Maybe a misprint or this is just fud.
Score
0
a b å Intel
a c 146 à CPUs
October 22, 2013 7:02:40 PM

sykozis said:
I don't see why a "defect density issue" would delay a processor launch by a year....unless, of course, the engineers are just completely clueless on how to fix it.

Broadwell for mobile/laptop/AIO/etc. only got delayed by a few months from late-2013 to early-2014 due to yield problems. LGA Broadwell-K (presumably K-only models) is the new chip on Intel's roadmap.

By the time Broadwell-K comes out, Skylake will be just around the corner unless Intel changes their mind about not delaying Skylake.
Score
0
a b å Intel
a c 146 à CPUs
October 22, 2013 7:09:24 PM

jimmysmitty said:
As for the slides, anyone else notice it says PCIe 2.0 5GT/s? Why would the 9 series chipset (the next in line) go from PCIe 3.0 to PCIe 2.0?

Maybe a misprint or this is just fud.

Probably no misprint there. With the DMI interface's limited bandwidth, the IO hub does not have sufficient CPU-chipset bandwidth to support PCIe 3.0 on the IO hub on top of all other built-in IO anyway.

On Intel systems, the Ivy Bridge or newer i5/i7 CPUs provide 16x PCIe 3.0 lanes (or 40 for SB-E/IB-E on LGA2011) while the IO hub provides PCIe 2.0. Older and lower-end Intel CPUs only support PCIe 2.0.
Score
0
October 22, 2013 7:58:04 PM

So is Nvidia's Maxwell, also Q4 2014, 4*2-1-4=3 Half life 3 confirmed!
Score
1
October 22, 2013 10:16:30 PM

Broadwell will be getting the Southbridge onto the CPU die.
Nehalem was the NB and the memory controller, Sandy Bridge was the GPU, Haswell was the VRMs, I believe.
Score
0
a b à CPUs
October 23, 2013 1:33:25 AM

InvalidError said:
jimmysmitty said:
As for the slides, anyone else notice it says PCIe 2.0 5GT/s? Why would the 9 series chipset (the next in line) go from PCIe 3.0 to PCIe 2.0?

Maybe a misprint or this is just fud.

Probably no misprint there. With the DMI interface's limited bandwidth, the IO hub does not have sufficient CPU-chipset bandwidth to support PCIe 3.0 on the IO hub on top of all other built-in IO anyway.

On Intel systems, the Ivy Bridge or newer i5/i7 CPUs provide 16x PCIe 3.0 lanes (or 40 for SB-E/IB-E on LGA2011) while the IO hub provides PCIe 2.0. Older and lower-end Intel CPUs only support PCIe 2.0.


But in the middle diagram it is written 2.0 above the PCIe and 3.0 below the PCIe and the 3.0 can't be the SATA value because the chipset supports SATA 6. So which is it? Maybe it supports both?
Score
0
October 23, 2013 5:31:12 AM

What's the difference between the 2014 Haswell refresh and Broadwell-K. Are both 14nm?
I'm guessing the 2014 Refresh is exactly the same as 2013 Haswell - except being 14nm. The Broadwell-K has minimal architectural improvement over Haswell - better iGPU, etc.
Correct?
Both the Refresh and Broadwell-K use the 9 series CS?
Score
0
a b å Intel
a c 146 à CPUs
October 23, 2013 5:56:55 AM

FunSurfer said:
3.0 can't be the SATA value because the chipset supports SATA 6. So which is it?

SATA v3.0 is 6Gbps.
Score
2
October 23, 2013 7:35:54 AM

BGA Broadwell will be great for tablets, laptops and NUC's. Plug and Play LGA K variants should keep desktop enthusiasts happy (or at least as happy as they are with Haswell). I'm not worried about Intel's strategy or timing.
Score
0
October 23, 2013 7:36:42 AM

(double post)
Score
0
October 23, 2013 8:24:36 AM

Am i the only one who thinks lately it would be better to put out a new product once in 2 years that trying to rush out every year? This goes the same for GPUs
Score
0
a b å Intel
a c 146 à CPUs
October 23, 2013 8:32:40 AM

ohim said:
Am i the only one who thinks lately it would be better to put out a new product once in 2 years that trying to rush out every year?

This is pretty much what Intel wants to start doing now and with only 5-7% gains between incremental generations, alternating between desktop and laptop to skip a generation on each complete cycle probably would not hurt. It is not like a 10-15% performance increase every two years instead of 5-7% per year would make much of a difference for most people either.

That would be fine with me... if Intel would pass some of the savings they are going to make off of that back to consumers. Trends from the past few years however indicate this is highly unlikely.
Score
0
October 23, 2013 11:56:02 AM

I don't work in the business, but I gather that there are a couple of changes to the annual tick-tock cycle.
1. Between Skylake (Jan 2016?) and Sandy Bridge 1-9-11, there would have been 5 years but only 4 chip series. Ivy Bridge took nearly 16 months, etc.
2. The 9 series chipsets will be compatible with Haswell, but the 8 series chipsets won't be compatible with Broadwell, even with a BIOS flash - unlike previous generations.
Score
0
October 23, 2013 12:42:49 PM

Haswell refress is suposed to be 22nm version with some upgrades maybe from Broadwell architecture. Could this be mobile K-series broadwell announcement? But if it is, why so much later than normat mobile broadwell, so not likely quess...
Score
0
October 29, 2013 2:29:03 AM

Hope they will remove FIVR from die , that cause temp skyrocket when OC even delidded it's still hot.
Score
0
a b å Intel
a c 146 à CPUs
October 29, 2013 7:47:26 AM

enterprise24 said:
Hope they will remove FIVR from die , that cause temp skyrocket when OC even delidded it's still hot.

Haswell Refresh reuses LGA1150 so FIVR will stick around for that. Broadwell for mobiles and SoC has FIVR too so it seems unlikely that Intel will take it out for Broadwell-K which is also supposed to reuse the LGA1150 socket, albeit without 8x series chipset compatibility.

If Intel does decide to take FIVR back, it most likely won't be until Skylake and since Intel does not officially condone overclocking even on K/X-chips, loss of overclocking due to extra heat likely won't be the main reason for doing so if they do do it.

Also, since 14nm chips will likely operate at even lower voltages, maintaining proper voltage regulation with off-chip regulators is going to become that much more challenging since voltage drop across the motherboard, socket and LGA substrate wiring is going to account for a larger chunk of total voltage.
Score
0
October 29, 2013 9:35:07 AM

InvalidError said:

Also, since 14nm chips will likely operate at even lower voltages, maintaining proper voltage regulation with off-chip regulators is going to become that much more challenging since voltage drop across the motherboard, socket and LGA substrate wiring is going to account for a larger chunk of total voltage.


Yep, there is not going back. You are right on target in that!
Score
0
December 4, 2013 10:22:48 AM

Quote:
I don't work in the business, but I gather that there are a couple of changes to the annual tick-tock cycle.
1. Between Skylake (Jan 2016?) and Sandy Bridge 1-9-11, there would have been 5 years but only 4 chip series. Ivy Bridge took nearly 16 months, etc.
2. The 9 series chipsets will be compatible with Haswell, but the 8 series chipsets won't be compatible with Broadwell, even with a BIOS flash - unlike previous generations.


we don't know that yet. keep the rumors to a minimum. already SO much misinformation spewed as fact its becoming hard to use the internet as a reliable resource when it comes to sensitive stuff.

we will know soon enough.
Score
0
a b å Intel
a c 146 à CPUs
December 4, 2013 11:27:44 AM

klepp0906 said:
Quote:
2. The 9 series chipsets will be compatible with Haswell, but the 8 series chipsets won't be compatible with Broadwell, even with a BIOS flash - unlike previous generations.


we don't know that yet. keep the rumors to a minimum.

It isn't like Intel has a particularly good track record with maintaining socket backwards compatibility and with Broadwell being intended mainly for the laptop market, rumors saying that it will require a new chipset due to changes in power specifications sound quite plausible.

If what Intel said about the Q1-2014 delay for mobile Broadwell not affecting Skylake's roadmap then we probably can still expect the LGA1151 Skylake in 2015, which could make Broadwell-K from Q4-2014 one of Intel's shortest-lived mainstream CPU lines. Makes me wonder why they would even bother. Intel's server roadmaps also indicate that everything is still on-schedule.

Looks like the 9x series chipsets will be mostly a Haswell-Refresh thing.

http://arstechnica.com/gadgets/2013/10/intels-next-gene...
Score
0
!