Any benefits to going BGA only with Broadwell?

Fjodor2001

Diamond Member
Feb 6, 2010
3,705
182
106
Hi,

So we all know Broadwell will be BGA only. And there's been a lot of discussions in other threads regarding the drawbacks that this will have for the enthusiast crowd.

But what are the benefits?

And what is the reason Intel is going down this route?

1. Will it be cheaper? In that case why?

2. Will it enable smaller motherboards (and computers)? Bear in mind that the VRM and PCH is likely to also be integrated, so what is actually left on the motherboard? Can we go below the Mini-ITX form factor? And in that case how small dimensions would be reasonable to expect?

3. Will it enable fanless Mac Mini type desktop computers, that still have 2500K/2600K/3570K/3770K levels of performance?

4. Any other benefits or reasons for Intel deciding to go BGA only?
 

Concillian

Diamond Member
May 26, 2004
3,751
8
81
Hi,

So we all know Broadwell will be BGA only. And there's been a lot of discussions in other threads regarding the drawbacks that this will have for the enthusiast crowd.

But what are the benefits?

And what is the reason Intel is going down this route?

1. Will it be cheaper? In that case why?

2. Will it enable smaller motherboards (and computers)? Bear in mind that the VRM and PCH is likely to also be integrated, so what is actually left on the motherboard? Can we go below the Mini-ITX form factor? And in that case how small dimensions would be reasonable to expect?

3. Will it enable fanless Mac Mini type desktop computers, that still have 2500K/2600K/3570K/3770K levels of performance?

4. Any other benefits or reasons for Intel deciding to go BGA only?

1) It will be cheaper for Intel since Laptop and desktop SKUs will likely be the same now... It will also improve Intel revenue, because more people will use Intel boards instead of outside OEM boards.

2) smaller than mini ITX on the desktop? Not likely. The size of the connectors, slots, and such are the main factor there.

3) BGA does nothing for thermals. Why would it enable that? Perhaps design changes will result in thermals that allow fanless PC use, but that's independent of LGA vs. BGA.

4) Money, money, and more money (for Intel)
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,705
182
106
1) It will be cheaper for Intel since Laptop and desktop SKUs will likely be the same now... It will also improve Intel revenue, because more people will use Intel boards instead of outside OEM boards.

2) smaller than mini ITX on the desktop? Not likely. The size of the connectors, slots, and such are the main factor there.

3) BGA does nothing for thermals. Why would it enable that? Perhaps design changes will result in thermals that allow fanless PC use, but that's independent of LGA vs. BGA.

4) Money, money, and more money (for Intel)

1) Ok. But will it be cheaper for end users buying desktop PCs? Remember that there might be additional costs pushed to the motherboard manufacturers (e.g. for having to keep more SKUs), and then onwards to the consumers.

2)
a) Well, the I/O connector block only has to be around 10-15x3x3 cm. And since Mini-ITX motherboards are 17x17xHeight cm there should be room for size reduction, right? After all, moving the VRM and PCH off the motherboard should enable smaller motherboards.
b) What components will actually be left on the motherboard if the VRM and PCH is integrated in the CPU?

3. True. But with 14 nm comes lower TDP. And moving the VRM and PCH on die will also lower TDP. So it's not really an effect of going BGA only, but an effect of what I just mentioned. The BGA might enable smaller computers though, since you get rid of the socket. So the question is whether all of this will be enough to enable small fanless small computers of the type I mentioned in my OP?

4. Yeah, probably correct. Is that the main driving force for Intel going BGA only? And if so, why didn't it happen earlier, before Broadwell?
 

Torn Mind

Lifer
Nov 25, 2012
11,567
2,626
136
Make people buy Haswell? Or make them go buy AMD chip so the latter doesn't die on them?
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
Hi,

So we all know Broadwell will be BGA only.

4. Any other benefits or reasons for Intel deciding to go BGA only?

4. Yeah, probably correct. Is that the main driving force for Intel going BGA only? And if so, why didn't it happen earlier, before Broadwell?

The fact that Haswell is not BGA-only is what makes me believe that Intel is not going to go a BGA-only route with Broadwell.

The timing doesn't makes sense from a project platform management standpoint.

It only makes sense to make a transition like this when the entire platform is ready for it.

If the rumor was that Skylake was going to BGA-only then I'd put some credence into it. That the rumor has Broadwell going BGA-only despite Intel having already seeded the market with Haswell-supporting mobos and platforms by that time makes the rumor have zero credibility for me.
 

grimpr

Golden Member
Aug 21, 2007
1,095
7
81
I think i'll wait about 8 months with Haswell to see whats coming out of Intels plans, strategy and AMD affairs, i dont like the rumours of it and it stinks like a very shortlived platform, like 1156.
 

Khato

Golden Member
Jul 15, 2001
1,202
240
106
I'll preface my thoughts with the fact that I don't yet believe that the rumors are true - my guess is that unlike current Intel will be offering BGA versions of desktop processors instead of forcing OEMs to use laptop parts if they want non-socketed.

1. Yes, it's cheaper. From the manufacturing standpoint you eliminate the socket, which digikey lists as an approximately $12 part. Sure it's cheaper for the actual bulk purchases that an OEM makes (edit, looks like it drops down to around $7 in large quantities), but it's still an unnecessary cost. As for having to keep more SKUs on hand, well, it depends in large part upon how Intel approaches it. It could be that Intel would only have two SKUs of each physical die - one that can be unlocked to their highest speed offering while the other has a limit... and in order to unlock it from a limited functionality mode you have to buy the microcode upgrade.

2. As stated in my preface, I believe this is the reason why Intel would offer BGA versions of their desktop processor SKUs. Right now most all SFF/unique designs have to use mobile processors because there aren't BGA SKUs of the desktop models. (Yes heat is also a concern, but the vastly lower power usage of desktop Broadwell is likely why they'd offer the option then.)

3. Potentially. Really more a matter of power usage than anything else.

4. My first thought on the matter was the fact that it could allow for marginal improvements in the signalling electrical characteristics. LGA is definitely good, but soldered BGA is better... and then it gets even more interesting if you start talking about having the memory chips mounted directly to the motherboard as well. That's one of the reasons why graphics cards can obtain far higher frequencies on their memory after all.
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
If the rumor was that Skylake was going to BGA-only then I'd put some credence into it. That the rumor has Broadwell going BGA-only despite Intel having already seeded the market with Haswell-supporting mobos and platforms by that time makes the rumor have zero credibility for me.

The obvious answer to me is that the plan changed. It does work with my idea that Intel is intending to make desktops "laptops without a screen". Perhaps the plan is accelerating.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
The obvious answer to me is that the plan changed. It does work with my idea that Intel is intending to make desktops "laptops without a screen". Perhaps the plan is accelerating.

But why? From Intel's perspective what value does it bring to Intel in Intel opting to not package and sell LGA-style Broadwells?

Look at their current k-branded processors. They cost maybe $15 more than their non-k twins.

That $15 is the premium Intel makes in selling a k-processor now. Why would Intel opt to not only reap the profits of premiums like that but also turn down the opportunity to sell LGA-style cpus at a $5-$10 premium over their BGA siblings?

It makes sense that Intel would be rolling out more BGA-style offerings for the desktop, but it makes no sense whatsoever that Intel would be killing off all LGA-type processors for the mainstream desktop. There is money to be made in offering both, as the k-processors prove, and Intel isn't about to turn down extra money.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Well the ideas that came to me are:
- They are pessimistic about Broadwell's ability to hit the clock speeds high enough to be faster than Haswell.
- They pulled resources to get the Smartphone/tablet Atom project accelerated.
- They are transitioning the performance desktop to the Extreme line.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
Well the ideas that came to me are:
- They are pessimistic about Broadwell's ability to hit the clock speeds high enough to be faster than Haswell.
- They pulled resources to get the Smartphone/tablet Atom project accelerated.
- They are transitioning the performance desktop to the Extreme line.

Good points, definitely.

The first bullet happened with Ivy over Sandy too. My Ivy barely touched the same clocks as my Sandy, albeit at much lower power usage.

Maybe Intel saw something in the demographics of IB buyers that made them decide not to do that again?

Bullets 2 and 3 are definite possibilities too. But bullet 2 would be a killer for margins, Intel is known to be more measured than that.

Bullet 3 makes sense until you factor in the timeline. Broadwell-E will come out when, 2025? (a joke, but you get my point - the lag between mainstream and the extreme variants based on server parts is bordering on the ridiculous already)

Seriously though Broadwell will debute before Haswell-E, and you are thinking Intel would have Haswell mainstream adopters wait until Broadwell-E debuts before upgrading? Surely Intel wants those mainstream peep's broadwell upgrade monies sooner than that?
 

pablo87

Senior member
Nov 5, 2012
374
0
0
For OEMs, the cost of BGA is much lower - mfg (socket is hand inserted still probably; then processor installed, then heatsink fan; vs bga machine mounted plus heat sink fan); purchasing (no socket) & logistics cost are less, probably less shrinkage too.

Also BGA is lower profile so helpful for those applications that need this.

The biggest cost back in the day was inventory write downs/price protection. But the channel business is much less these days, and presumably BGA is targeted at value parts, not high end.

There is an interesting business dynamic with the mb makers: if they receive Intel's full support, they make good money. if they don't, they're out of that business.
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,248
136
2) smaller than mini ITX on the desktop? Not likely. The size of the connectors, slots, and such are the main factor there.

They are already working on it :)

NUC1-462x346.jpg


http://www.pcpro.co.uk/blogs/2012/09/13/intel-nuc-mini-pc-review-first-look/

Of course it's not gonna be expandable....But it's gonna be tiny.
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
2)
a) Well, the I/O connector block only has to be around 10-15x3x3 cm. And since Mini-ITX motherboards are 17x17xHeight cm there should be room for size reduction, right? After all, moving the VRM and PCH off the motherboard should enable smaller motherboards.
What? It needs to be around 11 inches, or 25-30cm. How is Mini-ITX going to get you more slots to work with? Ironically, of course, today, expansion slots get used as break-out plates, with no cards, quite commonly. But, the standard still necessitates enough room for 7 slots, in addition to the integrated IO plate, and we want the option.

b) What components will actually be left on the motherboard if the VRM and PCH is integrated in the CPU?
RAM slots, power connectors, PCI-e, SATA, serial and parallel headers, USB headers, front panel headers, switches and LEDs...

We can put 1155 CPUs in Mini-ITX form factors today. Here. Now. You can go buy the boards. They aren't too expensive. If you want expensive, smaller SBCs are out there and are not likely to become less common, as low-end TDP drops. But, you either pay much more, or give up expansion options (or both, in the case of standard expansion slots).

Whether they go BGA-only (unlikely, IMO) or not--at least this soon--it changes cost structure for Intel and mainboard manufacturing, more than anything else.

They are already working on it :)

NUC1-462x346.jpg


http://www.pcpro.co.uk/blogs/2012/09/13/intel-nuc-mini-pc-review-first-look/

Of course it's not gonna be expandable....But it's gonna be tiny.
Why do they make such otherwise neat little boxes, and then screw them over? The rear needs to have 6-8 USBs, and the front 2-4. Those two side by side need to be a quad stack, as a start.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Why do they make such otherwise neat little boxes, and then screw them over? The rear needs to have 6-8 USBs, and the front 2-4. Those two side by side need to be a quad stack, as a start.

My first thought was "well I'll just buy a port multiplier", but if you're going to clutter your desk up with expansions then why bother to make the PC so small?

I might be able to get by with just the extra USB ports on my keyboard and monitor, granted, but I agree that 2 USB ports is suboptimal.
 

dma0991

Platinum Member
Mar 17, 2011
2,723
1
0
Why do they make such otherwise neat little boxes, and then screw them over? The rear needs to have 6-8 USBs, and the front 2-4. Those two side by side need to be a quad stack, as a start.
Exactly, that what came to my mind when I first saw it. It would need a minimum of 4 to actually be of any use. 2 USB is occupied by a mouse and keyboard, 1 USB if you're using a wireless keyboard and mouse combo.

I would really like to plug in a large external HDD to it semi-permanently but without additional USB 2/3, its going to cost me more for a Thunderbolt capable external HDD. Or maybe that was Intel's plan all along, push users to use Thunderbolt. :hmm:
 

pablo87

Senior member
Nov 5, 2012
374
0
0
more thoughts:

- BGA has a market segmentation drawback: if you do both, you can price BGA more aggressively and socket at a premium which should yield higher ASP.

- Right now, Intel is staring at a unit cost increase of $15-20 because of increased depreciation; that's why GS has a $16 target for Intel, IMO and that's probably where I'll take a position (though pls Intel if you're going to cut the dividend, the earlier the better). it's possible that the BGA savings will be used by Intel to raise prices to compensate for the higher depreciation.

- the other major factor, the elephant in the room so to speak wr to BGA is what will be Broadwell's TDP axis or schwerpunkt be? In my view - and hindsight is always 20-20, it was time a while back to split the cores and focus BW on the 2W-18W TDP range (and of course BGA only), with the higher core staying put where IB was ie 18W - 77W (and socket).

Because I for one have little confidence in Atom and it's follow ons because it's playing catch up on 2 fronts, whereas BW PRESUMABLY would have the performance crown guaranteed, and then it's just a matter of reducing power draw...
 
Last edited:

VirtualLarry

No Lifer
Aug 25, 2001
56,227
9,990
126
It is pretty ridiculous, when you consider that the Foxconn E-350 bookPC has two USB3.0 on the front, and four USB2.0 on the back.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,705
182
106
What? It needs to be around 11 inches, or 25-30cm. How is Mini-ITX going to get you more slots to work with? Ironically, of course, today, expansion slots get used as break-out plates, with no cards, quite commonly. But, the standard still necessitates enough room for 7 slots, in addition to the integrated IO plate, and we want the option.
When I said the I/O connector block only needs 10-15x3x3 cm, I meant the rear connector block on a typical Mini-ITX motherboard. In fact it can be made even smaller, depending on what I/O connectors you have.
RAM slots, power connectors, PCI-e, SATA, serial and parallel headers, USB headers, front panel headers, switches and LEDs...

We're talking about building a very small desktop computer here. Not some ATX based monster with 8 HDDs, 2 full size discrete GFX cards @ 500 W, 6 memory slots, and whatever.

Instead this is a more likely setup for the onboard/internal connectors on a very tiny desktop computer:

* No need for RAM slots, if RAM is soldered onto motherboard.
* No LGA socket, since the CPU is soldered onto the motherboard.
* No PCI-E connector, only use IGP.
* 1-4 SATA connectors. Possibly none at all will be needed, if the SSD flash memory is soldered onto the motherboard, and there is no HDD or optical drive.
* 1-4 internal USB connectors (of which 1 or 2 go to the front panel).
* Power on/off connectors, leds and so on only need a very small pin I/O block.
* Regarding the power connector, it depends on what PSU solution is selected. It could be some picoPSU-like solution, and the power connector could be made physically smaller.

Also, as mentioned previously the VRM and PCH will also be integrated in the CPU, so it will no longer be on the motherboard.

Do you really think such a motherboard would still require the same size as the 17x17xHeight cm Mini-ITX motherboards we have today? I don't. The question is just how small it can get? :confused:
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
1) It will be cheaper for Intel since Laptop and desktop SKUs will likely be the same now... It will also improve Intel revenue, because more people will use Intel boards instead of outside OEM boards.


Once again Broadwell won't come for desktops. It will stay on Haswell if the story is true.
 

Aikouka

Lifer
Nov 27, 2001
30,383
912
126
It makes sense that Intel would be rolling out more BGA-style offerings for the desktop, but it makes no sense whatsoever that Intel would be killing off all LGA-type processors for the mainstream desktop. There is money to be made in offering both, as the k-processors prove, and Intel isn't about to turn down extra money.

Do I think Intel will push BGA with Broadwell? Yes. However, I don't think it's nearly as doom-and-gloom as everyone is suggesting. Let's take a step back and look at this. Where did this whole news story originate from? The OEM system builders. Why is that a big deal? The way I see it, if I was an OEM system builder, I would prefer BGA packaging as it's cheaper, and I'm already using it on my laptops anyway.

In other words, I'm sayin' that system builders are probably pushing to start using the same parts among their desktops, laptops and x86-based tablets.

Would Intel save money if they just manufactured everything as BGA? Sure. Would I suspect it? No. Why? Because I couldn't imagine a single end-user-motherboard manufacturer that would be on board with the change. Imagine if you were ASUS or Gigabyte and you had to spend extra money to purchase CPUs ahead of time. These CPUs can potentially double to triple the original cost of your motherboard, and who has to eat that cost initially? The motherboard manufacturer. If I were one, I would also be pushing for significantly less product differentiation. Unlike now where you may have a half-dozen different motherboards from a single manufacturer, it would probably be too expensive to outfit all of those boards that may not even sell with pricey CPUs.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
Because I couldn't imagine a single end-user-motherboard manufacturer that would be on board with the change. Imagine if you were ASUS or Gigabyte and you had to spend extra money to purchase CPUs ahead of time. These CPUs can potentially double to triple the original cost of your motherboard, and who has to eat that cost initially? The motherboard manufacturer. If I were one, I would also be pushing for significantly less product differentiation. Unlike now where you may have a half-dozen different motherboards from a single manufacturer, it would probably be too expensive to outfit all of those boards that may not even sell with pricey CPUs.

Let me preface my post with firstly saying I agree with your synopsis, don't take the following comments as me being opposed to the rational and practical argumentation you lay out above.

That said, on the topic of ASUS and Gigabyte - if the BGA thing was true, from their perspective how would be any different from how they currently manage their AIB business units which are responsible for the discrete video card products?

From that perspective this would just be more of something they already know how to deftly manage (and profit from). In fact it may be to their advantage over that of their mobo competitors who currently aren't GPU AIB resellers as well because those are the guys who have a steep learning curve ahead of them (ASUS and Gigabyte having already climbed that learning curve with their GPU AIB business).

I agree it adds product mix complexity, inventory management complexity, distribution complexity...etc etc...but the fact that the GPU AIB business actually works makes me think a CPU AIB business might be viable too.

And no more a concern for the enthusiast than it is a concern for the enthusiast now when it comes to their selection and variety of discrete GPU products.

That said, I bring this up only to make the point that while I too do not see the mobo makers "jonesing" for this to happen, I see the ones who already handle GPU AIB business as being just fine with it happening if it were to actually happen.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,227
9,990
126
Would Intel save money if they just manufactured everything as BGA? Sure. Would I suspect it? No. Why? Because I couldn't imagine a single end-user-motherboard manufacturer that would be on board with the change. Imagine if you were ASUS or Gigabyte and you had to spend extra money to purchase CPUs ahead of time. These CPUs can potentially double to triple the original cost of your motherboard, and who has to eat that cost initially? The motherboard manufacturer. If I were one, I would also be pushing for significantly less product differentiation. Unlike now where you may have a half-dozen different motherboards from a single manufacturer, it would probably be too expensive to outfit all of those boards that may not even sell with pricey CPUs.

I think that this is a major sticking point. Generally speaking, when mobos are introduced, they may carry somewhat of a premium, depending, but over the lifespan of the mobo, they don't often decrease in price all that much.

Yet, Intel CPUs get price cuts regularly. Who is going to eat the cost of the inventory of CPUs, if they are already soldered to the boards, and waiting to be sold? It would seem to greatly increase the cost of inventory, and inventory write-downs, to the mobo mfgs.

If you thought Abit going out of business was bad, wait until the fallout from a few years of this BGA soldered-on CPU nonsense.

But perhaps that's Intel's long-term plan - to eliminate as many 3rd-party mobo mfgs as possible from the market, so that they can sell their own 1st-party mobos for a premium, especially to large OEM contracts.
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
* No need for RAM slots, if RAM is soldered onto motherboard.
Yes, there is a need for RAM slots. That's a very common thing to upgrade, as a given density of RAM gets cheaper, and you start wishing you had more. CPUs are not commonly upgraded. RAM and nonvolatile storage are. Leave that crap to phones and tablets. It's not like we don't have small RAM slots, either (MiniDIMM and SODIMM).

* 1-4 internal USB connectors (of which 1 or 2 go to the front panel).
Make that 8-12 and you're getting somewhere. I've actually had use for as many as 8 at once, and use 4-5 normally (discrepancy due to monitor hub).

* Regarding the power connector, it depends on what PSU solution is selected. It could be some picoPSU-like solution, and the power connector could be made physically smaller.
PicoPSU uses the same connector as every other. Some way or another, much more power needs to be handles, for other devices.

Also, as mentioned previously the VRM and PCH will also be integrated in the CPU, so it will no longer be on the motherboard.
Meh. CPUs without integrated voltage regulation can be had on very small SBCs, already. It can drop the board maker's costs, but they can fit this stuff in tiny packages, already. The cost and thermals have been keeping it out of consumer devices, not inability to make them (would you want a 50dBA@1m book PC?).

Do you really think such a motherboard would still require the same size as the 17x17xHeight cm Mini-ITX motherboards we have today? I don't. The question is just how small it can get? :confused:
It can get about as small as a MicroATX SFF. Past that, it should have some major incentives, likely of the kinds Intel is not fond of, to be worth buying (in particular, it needs to very cheap, to offset the reduction in long-term flexibility). MiniITX and smaller aren't useful enough to make up for their shortcomings, outside of grandma's email box, and kiosk-style uses. Maybe one day, but not one day soon.

The size of the motherboard has to do with the standard, and a desire to not be stuck throwing the whole PC away, if it could have been improved upon. The ATX/BTX PC is the last platform with anywhere near its flexibility, and there are people that rely on that, and people that wish they had it when they get an uber-cheap PC and find they can't fix it or upgrade it.