Any benefits to going BGA only with Broadwell?

[Charles Kozierok]

Reminds me of those TVs with DVD players built into them. One breaks, both get tossed out. And then Blu-Ray comes out and you're stuck with a player you never use.

This graphic posted in an older thread here nails the issue for me. If people care more about "style" than functionality, fine, they have that option. But don't take away ours.

 

[Joje]

cool. technologies nowadays r rly amazing

 

[NTMBK]

How about all the complaining about the parade of sockets on motherboards lately? That is largely because of further and further integration of services and features into the CPU. But integrating video and memory controller functions actually has a purpose. What is the purpose of forcing people to buy CPUs and motherboards together?

Think about it as a "compute card". Your CPU, RAM sockets are all on one daughterboard. But what if that daughterboard used a standard socket across multiple generations? What if you could upgrade your entire all-in-one PC by pulling out your Broadwell compute card and installed a Skymont compute card, complete with DDR4? Is that not an improvement?

 

[pablo87]

BGA saves money and in my opinion, part of a broader initiative (eg. chipset integration) to reduce the total cost (BOM, logistics, channels) of other components so there can be more left for Intel/processor. This is nothing new: twas the initial thrust behind integrated graphics back in the late 90's.

While Upgradeability has been a key feature of PCs and more or less a sacred cow, its also the biggest and perhaps the only remaining area where significant cost can be taken out of the PC eco-system.

Moreover, if you believe the vast majority of PC customers today value small form factor/integration/power savings >>> upgradeability, then its a win win.

Make no mistake: Intel's profitability level and cash flow (already: market has to be a little shocked that they raised $6B debt this week - I would be) is in jeopardy because of higher capex and resultant depreciation and something has to be done about protecting and perhaps raising ASP and this is not only the low lying fruit, its also nothing new for Intel.

Personally, I like it. There's a lot of unnecessary BOM, logistics and channel costs in PCs today, this will deliver more value to customers and make PCs more competitive with the Apple/Chrome/ARM tidal wave - intel needs this badly esp. given Microsoft's lacklustre operating system...talk about a perfect storm.

 

[taltamir]

People are buying Apple stuff for a whole lot of reasons

1. There are similar products by non apple companies, I just don't remember their names.
2. I compared 2 apple products, so if someone is already buying apple, its a choice of which.

Reminds me of those TVs with DVD players built into them. One breaks, both get tossed out. And then Blu-Ray comes out and you're stuck with a player you never use.

This graphic posted in an older thread here nails the issue for me. If people care more about "style" than functionality, fine, they have that option. But don't take away ours.

I agree with all of it but that doesn't mean there is no market for such devices.

 

[Charles Kozierok]

Think about it as a "compute card". Your CPU, RAM sockets are all on one daughterboard. But what if that daughterboard used a standard socket across multiple generations? What if you could upgrade your entire all-in-one PC by pulling out your Broadwell compute card and installed a Skymont compute card, complete with DDR4? Is that not an improvement?

I suppose it depends on how it's done. But I'm very highly skeptical. I think the entire point of this is to force people to toss out old PCs and buy new ones, rather than upgrading or repairing them.

Moreover, if you believe the vast majority of PC customers today value small form factor/integration/power savings >>> upgradeability, then its a win win.

Nobody is saying they shouldn't have BGA chips. Just that there should be options. More choices is better. Fewer choices is worse.

 

[Fjodor2001]

To expound upon this - what if television sets, and encrypted cable boxes became integrated? And then you would have to replace your whole television set, if you changed cable tv providers?

That's the kind of rather pointless integration that we're talking about here.

<somewhat_ot>

What if the encrypted cable box was integrated in te TV, but allowed you to install any SW you wanted? I.e. switch to a new cable provider, and the integrated box automatically downloads the new SW from your new provider over the Internet and you're done.

Or it could be some generic SW that could handle any compatible cable provider. I.e. establish a common communication protocol, and download video streams over the Internet from all providers you have a subscription with.

The way it is today with separate boxes for each provider is crap. And not having it integrated in the TV is also crap. I'd like to have as few boxes and cables as possible messing up my living room.

</somewhat_ot>

 

[taltamir]

Nobody is saying they shouldn't have BGA chips. Just that there should be options. More choices is better. Fewer choices is worse.

If that is the case then nobody here disagrees on anything. Because anyone I have seen here that didn't make proclamations like "this is just intel out to exploit the customers" has stated that there is reasons to have both.

 

[cbn]

I suppose it depends on how it's done. But I'm very highly skeptical. I think the entire point of this is to force people to toss out old PCs and buy new ones, rather than upgrading or repairing them.

I don't think that is Intel's plan.

A major difference between Apple and Intel is that Apple makes money on each device sold (All-in-One PC, Laptop, tablet, smartphone, etc). Intel only makes money on the chip (the major exception being the Intel smartphone).

 

[Insert_Nickname]

It's not just being able to repair things that is at stake here. It's also being able to customize your machine. Right now if you need a lot of video power but not much CPU, or if you want a lot of CPU but don't need fancy motherboard features, or if you're okay with integrated video but want the latest motherboard gadgets, you can get any combination you want. The more that components are integrated, the more those possibilities go away. They are not going to pair a low-end processor with a high-end motherboard or vice-versa. It will all be uniformly, linearly scaled, and you'll have to take whatever they have on offer.

^^ This

If for some (strange and slightly terrifying... ) reason I want to pair a Celeron G465 and an ASUS P8Z77V Deluxe, why shouldn't I have that option...?

(Not that I'd ever want to of course, but I -could- see a use for such a combination...)

 

[cbn]

Nobody is saying they shouldn't have BGA chips. Just that there should be options. More choices is better. Fewer choices is worse.

Certainly the plug in modules could have sockets.

Another option would be "software unlockable" CPUs (mentioned in the Anandtech podcast) for the BGA mainboards or BGA plug in modules. This would also increase choices.

 

[Khato]

Another option would be "software unlockable" CPUs (mentioned in the Anandtech podcast) for the BGA mainboards or BGA plug in modules. This would also increase choices.

How 'bout the option of Intel not offering any desktop Broadwell SKUs?

Removing the desktop, aka high power, market would allow them to tune it exclusively for the intended mobile applications... which may well allow it to benefit from the process shrink that much more.

 

[cbn]

How 'bout the option of Intel not offering any desktop Broadwell SKUs?

Removing the desktop, aka high power, market would allow them to tune it exclusively for the intended mobile applications... which may well allow it to benefit from the process shrink that much more.

I'm sure Broadwell will still be used for OEM desktop, but I wonder about 14nm FinFET.

If we look at Intel's 22nm FinFET, the max OC on Ivy Bridge (~4.9 Ghz, air) appears to be slightly lower than Sandy Bridge (~5.0 Ghz, air). Granted this is comparing 22nm in its first year to a mature 32nm process, but it makes me wonder. (Heat density, etc.)

14nm may very well be leakier and/or more tuned to mobile. If that happens, where will the max stock clocks of Broadwell lie? Max overclocks?

Maybe Haswell (on the mature 22nm process) becomes the go to chip for DIY desktop builders till Intel figures something out.

 

[Khato]

14nm may very well be leakier and/or more tuned to mobile. If that happens, where will the max stock clocks of Broadwell lie? Max overclocks?

Maybe Haswell (on the mature 22nm process) becomes the go to chip for DIY desktop builders till Intel figures something out.

I'd guess we'll have a better idea about where Broadwell might be on frequencies once we see how Haswell compares to IvyBridge. It may well be the case that Intel expects a repeat of IvyBridge with Broadwell, where it's easy to extra the power savings of the new process but difficult to maintain high clock frequencies? So why even attempt to keep it as a high-speed desktop part when the market is moving towards low power mobile SKUs? Especially if Haswell doesn't have any competition in the desktop market.

 

[cbn]

I'd guess we'll have a better idea about where Broadwell might be on frequencies once we see how Haswell compares to IvyBridge. It may well be the case that Intel expects a repeat of IvyBridge with Broadwell, where it's easy to extra the power savings of the new process but difficult to maintain high clock frequencies? So why even attempt to keep it as a high-speed desktop part when the market is moving towards low power mobile SKUs? Especially if Haswell doesn't have any competition in the desktop market.

Let's say for the sake of argument that successive nodes are more heat dense (despite improvements in xtor design and material science).

Maybe Intel can develop a wider Core architecture? This wider Core architecture could increase single thread performance even if clocks stagnated or even decreased.

The silicon cost for this larger Intel core wouldn't be much at 10nm, 7nm or 5nm. Even the power would probably be acceptable for desktop on these small nodes too.

 

[cbn]

While Upgradeability has been a key feature of PCs and more or less a sacred cow, its also the biggest and perhaps the only remaining area where significant cost can be taken out of the PC eco-system.

Moreover, if you believe the vast majority of PC customers today value small form factor/integration/power savings >>> upgradeability, then its a win win.

With regard to desktop, I don't really see small form factor as being more desirable (once a certain point has been reached.)

Granted, all things being equal, I would prefer smaller over larger. But this is only because the current cpu tech doesn't require larger.

 

[blastingcap]

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.

Desktop PC enthusiasts are a dying breed. At some point it will not be worth it to even have -K as an option anymore, if the costs of doing that exceed the potential extra profit from doing that. Will we reach that point with Broadwell? Skylake? After? I don't know. All I know is that desktop PC enthusiasts are a dying breed.

 

[taltamir]

Desktop PC enthusiasts are a dying breed. At some point it will not be worth it to even have -K as an option anymore

Yet the K option is a fairly recent addition to intel indicating the opposite.
But even if what you say is true, intel will only terminate those kind of products when they are no longer profitable. There is no deadline... it will happen naturally when consumer demand becomes low enough.

I think we are a long way off from that day

 

[VirtualLarry]

^^ This

If for some (strange and slightly terrifying... ) reason I want to pair a Celeron G465 and an ASUS P8Z77V Deluxe, why shouldn't I have that option...?

(Not that I'd ever want to of course, but I -could- see a use for such a combination...)

You're right. This move to BGA CPU+mobo SKUs, might reduce overall upgrades. Some people buy a cheap low-end CPU to stick into a high-end mobo, to do a cheap platform upgrade, with a plan to upgrade the CPU to something high-end at some later time and date.

 

[pablo87]

With regard to desktop, I don't really see small form factor as being more desirable (once a certain point has been reached.)

Granted, all things being equal, I would prefer smaller over larger. But this is only because the current cpu tech doesn't require larger.

I was referring to value/mainstream users not power users and gamers, apologize for the lack of clarity.

 

[pablo87]

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.

This is also the way I see it - a two pronged strategy - BGA et al. to reduce cost and maintain/increase value/mainstream ASP and socket to maintain/raise ASP in the high end segment. It dovetails with further processor segmentation as well insofar as tablets/low power is concerned.