Now that Haswell is close to release, can we start talking about Broadwell?

Intel DOES want to move to all-BGA production, as to lock-down the market: however with AMD still offering even SOMEWHAT competitive chips, they need the illusion of freedom. If they went ALL-BGA now, a lot of intel fans may move to AMD and even cause a rise in AMD ownership. As soon as AMD is out of the desktop picture entirely, intel will go all-BGA: I bet my nuts on it.

I don't see the feasibility in doing so..especially from the enterprise and OEM segments. That would increase the cost and time to repair a system in the field. Plus it would compound a backorder issue...if the motherboard was backordered, it now creates a backorder of the CPU, and vice-versa.

AMD has such a small market share right now, Intel doesn't see them as a threat when it comes to the DIY segment, which is small potatoes. If Intel did move to BGA only, they only have to fear the massive supply scale of their OEM customers moving to AMD..which is not feasible because AMD doesn't have the manufacturing and supply capacity. At which point, enterprise customers would just simply hang on to their existing systems longer.
 
Laptops already operate on BGA platform and have all of the negative attributes you mentioned, but are selling like crazy and getting expensive repairs by the millions. As you probably know, if a laptop requires a new CPU, sound chip, chipset, etc. It really needs the whole lot replaced: most of the time: people just buy a new laptop: win for intel. If desktops worked the same way, intel wins even more. The problem is that AMD still have a sliver of market share and are still sold at brick-and-mortar PC shops, so the desktop customer still has a choice at this point. The average consumer may be swayed if the salesperson told the customer "you can't upgrade an intel system, but you can upgrade an AMD system." As soon as AMD is no longer sold at brick-and-mortar PC shops, Intel will go BGA. It only means win for them at that point.
 
That is precisely why the company I work for went with laptops containing AMD processors in an LGA package.
 
There is no confirmation regarding that. All we have is some random rumor..

I believe all ultrabooks and notebooks are BGA.

I think a few computers in the entry-level will be BGA, while low end motherboards will still have sockets. I would not be surprised all-in-one computers are being targeted here and small form factor PCs from the major OEMs. Why? How many people you know that buy these $300 to $500 PCs from Walmart or the $800 all-in-one know how to change a processor or let alone a motherboard? Keep it simple for the normal folks, but let the tinkerers have their fun in the higher end. There is still money to be made from both ends of the PC market spectrum-- the non-tech savvy and the do-it-yourselfers.

Anything higher than entry-level will still have sockets on motherboards from ASUS and so on. (I bet some of you haven't seen the new 8-series boards in prototype form from MSI and a few others.)
 
I don't know a whole lot about programming, I'm just an amateur one, but I think the problem isn't "lazy programmers" writing the software and games, it is the large companies that build the framework for those programmers that need to step it up. DirectX needs to find creative ways to translate API calls to multiple processors, so when a programmer calls a function it doesn't just run on one core, but many. The frameworks need to find ways to use the cores and let the programmers access the functions that utilize multiple cores. At least this is my opinion. (I guess the problem still is "lazy programmers", but they aren't the ones making the software, they are the ones who program the frameworks and publish the standards that programmers should adhere to and use.)
 
I don't know a whole lot about programming, I'm just an amateur one, but I think the problem isn't "lazy programmers" writing the software and games, it is the large companies that build the framework for those programmers that need to step it up. DirectX needs to find creative ways to translate API calls to multiple processors, so when a programmer calls a function it doesn't just run on one core, but many. The frameworks need to find ways to use the cores and let the programmers access the functions that utilize multiple cores. At least this is my opinion. (I guess the problem still is "lazy programmers", but they aren't the ones making the software, they are the ones who program the frameworks and publish the standards that programmers should adhere to and use.)

DirectX 11 spreads calls to multiple cores. It's the central game logic that is extremely hard to multithread, because you need to know what is happening now in the game before you can determine what happens next.
 
I am not sure the "hard" part is even making a multi-threaded engine.

I think the hard part is making an engine that works on 1 core, 2 core and 8 cores dynamically.

A programmer could easily say, 1 core for audio engine, 1 core for input, 1 core for game engine, 1 core for AI. Already up to 4 cores without any fancy stuff.

The hard part is saying...okay this works on 4 cores and technically 8 cores thou under utilized...so how is this going to work on a dual core?

That is how I understand it.
 
This is just my two cents, but why can't CPUs just have a hardware dispatcher built in to recieve the instructions and send them to cores dynamically? Just make the cpu appear as one unit and not have to code per core?
 
This is just my two cents, but why can't CPUs just have a hardware dispatcher built in to recieve the instructions and send them to cores dynamically? Just make the cpu appear as one unit and not have to code per core?

Each core already does that internally nearly to the maximum that can be achieved since typical code has branches and dependencies every few instructions that force a more serial execution.
 
Each core already does that internally nearly to the maximum that can be achieved since typical code has branches and dependencies every few instructions that force a more serial execution.

Oh i see.. so it would end up being insanely complex for it to do "more" than it already does...
 
There will be little to gain unless you weakened the individual cores first and that would just be a game of redefining what a core is.
 
I am very happy with the direction Intel and others are going.

AMD is not a big threat to Intel, Intel's real threat is ARM. And ARM forced Intel to think of efficiency.

It's like how EU forced European car manufacturers to make their cars move efficient, economic, environment friendly and safe, while at the same time they have to make their cars faster and better handling if they want to sell them. So it's like a double attack, more efficient and economic, yet faster. The same thing is happening in the IT world.

Instead of just throwing more brute horse power and watts at CPUs, Intel, starting with Haswell will focus on battery life and temperature, while increasing speed. CPUs are already incredibly fast, much, much faster than what an average person needs. When a few years ago, I'd recommend an office user a medium range laptop, nowadays, I just tell them to get the cheapest i series laptop they can find, cause it'll run the email, office and the browser just fine. The hardware has left the software behind and Windows is getting lighter too, Windows 8 runs super fast on old hardware, so like again, like with cars, double attack, hardware gets faster, software gets lighter and more optimized. As for the power, even the new, next gen consoles coming out won't match powerful desktops and laptops that are available now.

You have to realize, the world is moving to mobile. Laptops, tablets and phones and I'm am more excited with the idea of 24 hour battery life laptops and tablets, fanless thin laptops than with the idea of powerful, unoptimized, power sapping CPUs.

Plus, I don't want X86 to lose out to ARM, that Windows application library is just too valuable.
 
Back
Top