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

[Phynaz]

I mean CPU thermals obviously. And they only care about system thermals because of laptops and tablets.

No.

 

[Ajay]

Sure, Intel can make a 65 W 14 nm node shrink "copy" of the 4770K with some more & redesigned EUs and 30% increased iGPU performance perhaps.

But really, most people sitting on a SB 2500K or later has no reason to upgrade to that. If they want to give those people a reason to upgrade they'll have to provide something more than that, like 6 CPU cores and/or a seriously beefed up iGPU / eDRAM. And such a change would likely require around 95 W TDP as in the OP slides, despite the node shrink to 14 nm.

Then they can have low power BGA Broadwell CPUs for the mainstream AIO PC desktop PCs and similar.

Since the desktop plans for Broadwell has shifted back and forth from LGA, to BGA only, and now back to LGA (and BGA), it looks like Intel has been reconsidering their plans a couple of times. So I guess anything is possible.

More eDRAM and beefier iGPUs are coming for sure (plus integrated IOH). I don't see six core desktops coming though. Intel is dropping four core CPUs from their HEDT lineup starting with HW-E, so the baseline will be six cores. I imagine that will be the plan at least for HW-E, Skylake-E and Skymont-E; we'll see what happens after that. But it doesn't make sense for Intel to offer six cores on standard desktop when they can make themselves and their partners a whole lot more money on the HEDT platform.

 

[Fjodor2001]

But it doesn't make sense for Intel to offer six cores on standard desktop when they can make themselves and their partners a whole lot more money on the HEDT platform.

The margins may be higher on the HEDT platform, but the volumes are also much less. So the total profit from that platform may not be as much.

If they can increase desktop sales by e.g. 20% by delivering a 6 core Broadwell-K, then they may make more total profit from that, despite the profit margin being lower per unit sold.

 

[ShintaiDK]

The margins may be higher on the HEDT platform, but the volumes are also much less. So the total profit from that platform may not be as much.

If they can increase desktop sales by e.g. 20% by delivering a 6 core Broadwell-K, then they may make more total profit from that, despite the profit margin being lower per unit sold.

You are not getting a 6 core Broadwell-K, because it wouldnt sell in a volume needed. And the only reason you got the E series is due to Xeons. Without Xeons there wouldnt be an E series on the desktop.

 

[Fjodor2001]

You are not getting a 6 core Broadwell-K, because it wouldnt sell in a volume needed.

How do you know that? And how do you know what volumes are would be needed, and that a 6 core Broadwell-K would not sell at those volumes?

Also, any reason a Broadwell-K that you are expecting, which basically is a copy of a 4770K but with better iGPU, would sell better? The HD4600 iGPU is already good enough for most people, and the rest will buy a discrete GFX card anyway. I.e. no reason to upgrade from a 2500K or later for most people.

With a 6 core CPU it would be something else. Especially since we're now entering a next-gen gaming era with the PS4/XBONE making use of 6/8 cores.

 

[ShintaiDK]

How do you know that? And how do you know what volumes are would be needed, and that a 6 core Broadwell-K would not sell at those volumes?

Also, any reason a Broadwell-K that you are expecting, which basically is a copy of a 4770K but with better iGPU, would sell better? The HD4600 iGPU is already good enough for most people, and the rest will buy a discrete GFX card anyway. I.e. no reason to upgrade from a 2500K or later for most people.

With a 6 core CPU it would be something else. Especially since we're now entering a next-gen gaming era with the PS4/XBONE making use of 6/8 cores.

You are not getting a 6/8 core mainstream. Nomatter how many threads you keep making about it.

You need to buy the platform for it if you want one. And its not LGA1150 or FM2+.

 

[Fjodor2001]

You are not getting a 6/8 core mainstream. Nomatter how many threads you keep making about it.

You need to buy the platform for it if you want one.

You would be more credible if you actually tried to answer the questions and had some facts or reasoning to back up your statements.

I could turn your way of arguing around and say "Do you think there will be less chance of a 6/8 core mainstream CPU just because you keep repeating "it will not happen" in post after post?"

Just like you kept repeating "There will be no APU in PS4"?

 

[ShintaiDK]

I know how your argumentation works. Just like in the other thread where you claim the FX8350 is the best selling. Only to ask others for documentation when they say its not.

You can just buy IB-E today if you want 6 cores, or Haswell-E next year if you want 8 cores.

 

[Fjodor2001]

You can just buy IB-E today if you want 6 cores, or Haswell-E next year if you want 8 cores.

And just what does that have to do with anything? You can also just buy a 4770K + discrete GFX card today if you want better GPU performance. I.e. no chance/need for a better iGPU in Broadwell-K?

You make it sound like a 6 core Broadwell-K would be some kind of unachievable luxury item. But in fact the iGPU part of the Broadwell-K will likely occupy more die area than those 2 extra cores. Also, you're on 14 nm so you can fit twice the amount of transistors per die area compared to at 22 nm which the 4770K is on.

With that being said I do think it's likely that we'll not see a mainstream 6/8 core CPU until Skylake or beyond, but I wouldn't rule it out since the plans for Broadwell seems to have been shifting back and forth several times (LGA vs BGA, and so on).

 

[ShintaiDK]

By that splendid logic we should have 8 core IBs on LGA1155. Since there is no product, then there is no market big enough to substain it outside the Xeon reuse. Aka E series.

Your logic is based around you want the product, then you work backwards on how its possible. Instead of the other way around. Its a classic for phantom products.

Broadwell have always been LGA1150 for E3 Xeons.

 

[Fjodor2001]

By that splendid logic we should have 8 core IBs on LGA1155. Since there is no product, then there is no market big enough to substain it outside the Xeon reuse. Aka E series.

And by your splendid logic we would always be stuck with 1 core mainstream CPUs. Because before 2 core mainstream CPUs had been introduced, there obviously were no 2 core mainstream CPUs, which you would take as proof of that there is no market for it.

Your logic is based on only analyzing the past and present, and then making an assumption that things remain static going forward. It's a classic mistake.

 

[Hulk]

I don't know about the reality of a 6 core mainstream part but I do know a 6 core mainstream Broadwell would be pretty much be the only thing that would make me upgrade my system instantly. Assuming I can get the chip for <$300.

 

[NTMBK]

I don't know about the reality of a 6 core mainstream part but I do know a 6 core mainstream Broadwell would be pretty much be the only thing that would make me upgrade my system instantly. Assuming I can get the chip for <$300.

What about Haswell-E? Hopefully the 6 core part will get a little cheaper with that.

 

[Insert_Nickname]

What about Haswell-E? Hopefully the 6 core part will get a little cheaper with that.

Wouldn't that be nice? As with LGA-2011, I have a feeling that its the platform itself that will be relatively expensive. Not the CPU itself. Quad channel DDR4 sounds relatively expensive.

Oh, well. Wasn't really planning on upgrading before Skylake, but if the X99 platform is good I can always change my mind.

 

[Ajay]

I don't know about the reality of a 6 core mainstream part but I do know a 6 core mainstream Broadwell would be pretty much be the only thing that would make me upgrade my system instantly. Assuming I can get the chip for <$300.

Like ShintaiDK said, we have 6 core HEDT part because of Xeons. Maybe by Skymont, with the development of more thread heavy games (driven in part by X One and PS4) Intel might find it's time to had a top DT part w/6 cores - we'll just have to see. Since Skymont is supposed to be 10nm and some version of BGA mounting (non-upgradable), that also might incline Intel to offer a slightly higher end offering, then again, maybe not since they might prefer a higher churn rate instead).

 

[jpiniero]

6 core still isn't happening on mainstream unless it can be done in a laptop. Which could be feasible with Skymont but I wonder if they will keep it at 4 and either continue to add more GPU cores or increase/maintain margins as the fab costs skyrocket.

 

[beginner99]

Like ShintaiDK said, we have 6 core HEDT part because of Xeons. Maybe by Skymont, with the development of more thread heavy games (driven in part by X One and PS4).

Consoles are irrelevant because even a current dual-core haswell with HT can match the console CPUs in terms of CPU performance, even with 8 threads.

 

[Khato]

Like ShintaiDK said, we have 6 core HEDT part because of Xeons. Maybe by Skymont, with the development of more thread heavy games (driven in part by X One and PS4) Intel might find it's time to had a top DT part w/6 cores - we'll just have to see. Since Skymont is supposed to be 10nm and some version of BGA mounting (non-upgradable), that also might incline Intel to offer a slightly higher end offering, then again, maybe not since they might prefer a higher churn rate instead).

While I do laugh every time I see someone call Skylake's successor Skymont such is incorrect - it's Cannonlake. If I recall correctly, the only reference to Skymont was a 'roadmap' that someone made up which also had an outdated code name for Broadwell. It never made sense to me that everyone latched onto it either when the *mont naming is quite clearly reserved for the atom line (Silvermont, Airmont, Goldmont.)

Anyway, I'm quite curious to see what Broadwell-K ends up being - I'd currently give it a 50/50 chance of having a 6 core SKU. (Have to remember that Haswell-E is reported to bump the core count for that platform up to 8/6 instead of the current 6/4 and it will likely launch around the same time or before Broadwell-K.)

 

[Ajay]

Consoles are irrelevant because even a current dual-core haswell with HT can match the console CPUs in terms of CPU performance, even with 8 threads.

Not when a given console game is scaled up properly. Of course, many are not, but the good ones are and then they bottleneck a PC just as much as they console.

 

[Ajay]

While I do laugh every time I see someone call Skylake's successor Skymont such is incorrect - it's Cannonlake. If I recall correctly, the only reference to Skymont was a 'roadmap' that someone made up which also had an outdated code name for Broadwell. It never made sense to me that everyone latched onto it either when the *mont naming is quite clearly reserved for the atom line (Silvermont, Airmont, Goldmont.)

Thanks for the correction.

 

[Fjodor2001]

Not when a given console game is scaled up properly. Of course, many are not, but the good ones are and then they bottleneck a PC just as much as they console.

Yes, and thanks to the PS4 and XBONE which both have 8 CPU cores the game developers will now be forced to code for parallelism to actually make use of the hardware, or they will be overrun by those who do.

Moving such a game over to a PC environment it will then of course automatically also make use of 6/8 cores. With each CPU core being more powerful on a PC, the games will be able to surpass the console versions.

Welcome next-gen gaming... :awe:

 

[NTMBK]

Wouldn't that be nice? As with LGA-2011, I have a feeling that its the platform itself that will be relatively expensive. Not the CPU itself. Quad channel DDR4 sounds relatively expensive.

Oh, well. Wasn't really planning on upgrading before Skylake, but if the X99 platform is good I can always change my mind.

X79 motherboards don't actually seem as insanely priced as I'd feared. The cheapest one I've seen is a Biostar one, which is fairly reasonably priced. http://www.ebuyer.com/341239-biosta...11-8-channel-audio-atx-motherboard-tpower-x79 Yes, DDR4 may prove to be pretty expensive- but on the other hand, you're getting DDR4!

 

[CHADBOGA]

X79 motherboards don't actually seem as insanely priced as I'd feared. The cheapest one I've seen is a Biostar one, which is fairly reasonably priced. http://www.ebuyer.com/341239-biosta...11-8-channel-audio-atx-motherboard-tpower-x79 Yes, DDR4 may prove to be pretty expensive- but on the other hand, you're getting DDR4!

Hasn't it been the case for the last few generations of new ram types, that the first release of a new generation shows very little, if any, overall improvement to total system performance?

Whether that is because they start with modest ram speeds or the first memory controllers for the new generation of ram are far from optimal?

 

[Ajay]

Hasn't it been the case for the last few generations of new ram types, that the first release of a new generation shows very little, if any, overall improvement to total system performance?

Whether that is because they start with modest ram speeds or the first memory controllers for the new generation of ram are far from optimal?

Yeah, DDR4 latencies will be higher, IIRC, than DDR3 2133. From what I've read it won't be till sometime in 2015 that faster speeds are available - but that will be a mixed bag. One the one hand, DDR4 2133 will be more affordable, maybe even the same cost as DDR3 depending on how the RAM market is doing, but the faster speeds will come at a a steep premium.

 

[Insert_Nickname]

X79 motherboards don't actually seem as insanely priced as I'd feared. The cheapest one I've seen is a Biostar one, which is fairly reasonably priced. http://www.ebuyer.com/341239-biosta...11-8-channel-audio-atx-motherboard-tpower-x79 Yes, DDR4 may prove to be pretty expensive- but on the other hand, you're getting DDR4!

That actually seems a decent, if a little spartan, board. If I had need of an LGA-2011 system, I'd at least consider it. You can always add controllers, if you need them.

Yeah, DDR4 latencies will be higher, IIRC, than DDR3 2133. From what I've read it won't be till sometime in 2015 that faster speed are available - but that will be a mixed bag. One the one hand, DDR4 2133 will be more affordable, maybe even the same cost as DDR3 depending on how the RAM market is doing, but the faster speeds will come at a a steep premium.

From what I'm seeing timing for DDR4-2133 should start at 11-11-11-30 or 12-12-12-30. Not too bad compared to DDR3-2133. With "new" RAM there is always the possibility that you can reuse it down the road, so a small price premium compared to DDR3 is not out of line I think. Should also run cooler too, due to newer manufacturing process and lower voltage. DDR3-2133 and higher can get a little toasty and power hungry...