AMD, Intel Socket History Question

[patrickjp93]

Back in the early days, Intel and AMD CPUs fit into the same sockets. Why is this no longer the case? Wouldn't it benefit both companies to co-develop new sockets together by giving consumers more incentives to upgrade chips without having to move to a new board? I know board makers pay licenses to implement the sockets, but there's no way this outpaces the profits both companies reap from chip sales.

 

If anything, the numerous sockets (and chipsets) benefit the board makers by giving them new products ever so often. Obviously R&D must go into the chipsets too, so why have so many? I can understand having 1 set aside for corporate environments and one for consumers, but do we really need 5+ chipsets (on the Intel side) going at once? Again I know these provide royalties to both companies, but vs. the profit margin on chips, is this really non-negligible?

 

Why did we depart from having both companies fit in the same socket? At least on the surface it just makes no sense except for board makers.

 

[Katsu-K1L]

I Actually Didn't know about that but yea it would make upgrading soooooo much easier because no need to buy a new motherboard lol

[TheRetiredSlave]

Because the CPUs are built differently and have different strengths/weaknesses. They are specifically built for a wide range of uses and are made to fit into a specific set of criteria depending on what the application calls for.. It will always be true that a machine built to do everything will be great at nothing, and a machine built to be great for one thing will not be good for another.

 

Different chipsets are necessary for different customers. There's absolutely no reason for an Ultrabook to have advanced SLI support or a ton of PCI lanes, just like a high-end workstation chipset doesn't really need to focus on power/thermal efficiency.. 

 

 

For example, if you want a laptop to be completely fanless, you need components that create less heat than what can passively dispersed by the chassis.. If you want 8 cores at 5 Ghz, you'll need to throw 200W into a chip. it's impossible to build one platform that can do both of these things at the same time.

[patrickjp93]

 It will always be true that a machine built to do everything will be great at nothing, 

Very wrong. One machine built out of the best components for individual tasks can be great at all of them. That is the inevitability of digital logic. In mechanical systems you'd be right, but not in digital.

 

 

 

Different chipsets are necessary for different customers. There's absolutely no reason for an Ultrabook to have advanced SLI support or a ton of PCI lanes, just like a high-end workstation chipset doesn't really need to focus on power/thermal efficiency.. 

You have a point regarding mobile vs. desktop platform, but not for workstation. There's no reason the workstation chipset can't be the only desktop chipset developed and allow the user free reign to tune settings. And all the workstation chipsets really do is add more storage support, better error correction, and more PCIe lanes. In the case of the last, there is no reason to do R&D for lesser chipsets to support less if you can build it all once and be done while allowing the user to use as much or as little as they wish.

 

This would actually force MOBO prices down and force more MOBO competition. There's no downside to minimizing the number of chipsets running per platform.

[WinNut]

I Actually Didn't know about that but yea it would make upgrading soooooo much easier because no need to buy a new motherboard lol

I worked with the Asus SP97-V about a month ago, the CPU socket (Socket 7) accepted either an AMD K6 or a Pentium MMX. It also accepted Cyrix processors.

[Katsu-K1L]

I worked with the Asus SP97-V about a month ago, the CPU socket (Socket 7) accepted either an AMD K6 or a Pentium MMX. It also accepted Cyrix processors.

 

i never even heard of that o_o

[WinNut]

i never even heard of that o_o

I'm 4 years younger than you.

[Katsu-K1L]

I'm 4 years younger than you.

What does that have to do with anything....?

[WinNut]

What does that have to do with anything....?

The fact that you are older than me and I have had experience with older technology than you. Oh yes, the heatsink on that thing was 1.5cm tall and the only fan was the CPU fan connected over Molex. I accidentally killed that machine when I installed a second floppy drive.

[Katsu-K1L]

The fact that you are older than me and I have had experience with older technology than you. Oh yes, the heatsink on that thing was 1.5cm tall and the only fan was the CPU fan connected over Molex. I accidentally killed that machine when I installed a second floppy drive.

Well where i live sadly i have no access to older tech and no here is interested in anything but getting "Apple" so i have no way of knowing about older sockets/Tech unless i research them online....

[patrickjp93]

Well where i live sadly i have no access to older tech and no here is interested in anything but getting "Apple" so i have no way of knowing about older sockets/Tech unless i research them online....

Apple: the evil, overpriced bane of programmers' existence (mine included), with a huge market share in general purpose laptops for old people and professional applications for artists, and a strong phone following which makes it worthwhile to pay the licensing fee to put an app on the Apple app store.

[Katsu-K1L]

Apple: the evil, overpriced bane of programmers' existence (mine included), with a huge market share in general purpose laptops for old people and professional applications for artists, and a strong phone following.

yea overpriced as fuck but well....they make good quality shit, my ipad has been with me for 3 years now and looks like its brand new but well i take good care of my shit but well lol the conclusion is that i'm awesome and apple is evil .

[patrickjp93]

yea overpriced as fuck but well....they make good quality shit, my ipad has been with me for 3 years now and looks like its brand new but well i take good care of my shit but well lol the conclusion is that i'm awesome and apple is evil .

They need the 17" models of the MBP back, they need more tactile keys (Razer Blade for example of good tactile mobile keyboard), they need to use the extreme series mobile chips at that price, they need to make RAM upgrade-able, they need to make their audio headset port be bi-directional (it can only send or receive data at once, it cannot be running audio out and in at the same time), and they need to stay on top of the most recent GPU generations. The GTX 800m series has been out forever now. If I could have waited for the Broadwell release to buy my MBP, I would have because the 800 series cards are freaking amazing.

 

Also, wtf is with only 2 USB 3.0 ports? On a full laptop 3 is the minimum, end of discussion.

[TheRetiredSlave]

Very wrong. One machine built out of the best components for individual tasks can be great at all of them. That is the inevitability of digital logic. In mechanical systems you'd be right, but not in digital.

Actually, I'm the exact opposite of "very wrong".. Have you heard of ASICs or FPGAs?

 

Let me ask you this: if you were to fabricate a custom chip for the sole purpose of hashing SHA-256, how would it perform per-watt when compared to a general purpose GPU? For Bitcoin mining, obviously the ASIC specifically designed for mining bitcoin will be significantly more efficient because it doesn't need any of the additional circuitry that a general purpose GPU has.. While general purpose GPUs have the flexibility of being able to do anything (gaming, accelerate video editing, general purpose computing, or even cryptographic calculations) their efficiency will always be limited by the fact that they are bloated by the requirement of being able to do anything.

 

Computers are machines. Just because something is microscopic and embedded into a piece of silicon doesn't change the inherent principle that a machine/circuit designed for general tasks will always be less efficient at one task than a custom machine/circuit designed for that one task.. 

 

And yes, computers are "fast enough" nowadays that a machine with good specs will be excellent at any tasks you could feasibly do. However, this doesn't change the fact that custom hardware designed for one specific task will outperform even the fastest general purpose hardware you can get.. 

 

You have a point regarding mobile vs. desktop platform, but not for workstation. There's no reason the workstation chipset can't be the only desktop chipset developed and allow the user free reign to tune settings. And all the workstation chipsets really do is add more storage support, better error correction, and more PCIe lanes. In the case of the last, there is no reason to do R&D for lesser chipsets to support less if you can build it all once and be done while allowing the user to use as much or as little as they wish.

 

This would actually force MOBO prices down and force more MOBO competition. There's no downside to minimizing the number of chipsets running per platform.

 

The same principles hold true for chipsets as well.. Adding functionality requires additional circuits that can perform that functionality. Additional circuits take up extra space and increase R&D costs. Larger chips are more costly to manufacture (less chips per wafer, lower yields, etc).. These costs are either taken out of their profit margin (lol, that's not going to happen) or handed down to the consumer.

There's absolutely no reason for your grandma to pay extra for chipsets that can do advanced tasks, whether they're enabled or not, if all she's going to do is check email and play solitaire. It makes more sense to design a chipset to have basic functionality and focus on making it low-cost for everyone who doesn't need any advanced features.. If they can sell a high-end workstation chip at a high cost, they'd only be losing money by nerfing its capabilities and selling it as a lower-tier chipset (granted, all the chips get binned according to their performance, but binning high-tier chips into low tiers isn't as cost effective as designing a cheap chipset to saturate the affordable market)

 

 

Anywhoo, to get back to your original question, it sounds you're making the assumption that Intel is devoting substantial amounts of money toward R&D on a low-end chipsets. I would seriously doubt this to be the case. It would most likely be the case that products like the H81 are just the result of taking a higher end chipset (like Z87) and removing any unnecessary circuitry.. R&D costs realistically wouldn't be anywhere near as high as the high grade server stuff, and thus they could sell it as a more affordable option.

 

 

Edit: if this either doesn't make sense or I've repeated myself a bajillion times, I apologize.. it's 3AM and I can't sleep..

[Faa]

He just makes a thread to argue with people.

[patrickjp93]

The same principles hold true for chipsets as well.. Adding functionality requires additional circuits that can perform that functionality. Additional circuits take up extra space and increase R&D costs. Larger chips are more costly to manufacture (less chips per wafer, lower yields, etc).. These costs are either taken out of their profit margin (lol, that's not going to happen) or handed down to the consumer.

There's absolutely no reason for your grandma to pay extra for chipsets that can do advanced tasks, whether they're enabled or not, if all she's going to do is check email and play solitaire. It makes more sense to design a chipset to have basic functionality and focus on making it low-cost for everyone who doesn't need any advanced features.. If they can sell a high-end workstation chip at a high cost, they'd only be losing money by nerfing its capabilities and selling it as a lower-tier chipset (granted, all the chips get binned according to their performance, but binning high-tier chips into low tiers isn't as cost effective as designing a cheap chipset to saturate the affordable market)

 

 

Anywhoo, to get back to your original question, it sounds you're making the assumption that Intel is devoting substantial amounts of money toward R&D on a low-end chipsets. I would seriously doubt this to be the case. It would most likely be the case that products like the H81 are just the result of taking a higher end chipset (like Z87) and removing any unnecessary circuitry.. R&D costs realistically wouldn't be anywhere near as high as the high grade server stuff, and thus they could sell it as a more affordable option.

 

 

Edit: if this either doesn't make sense or I've repeated myself a bajillion times, I apologize.. it's 3AM and I can't sleep..

My point about chipsets is this: do the R&D once and sell 1 or 2 chipsets vs. doing R&D on 5 different chipsets and selling such a wide array of them while getting less profit from each respectively. This unifies the support needs (more money saved by Intel), unifies the cost (good for gamers, personal server builders, and workstation people who are Intel's consumer bread and butter), and requires there be only 1 or 2 manufacturing processes running simultaneously instead of 4 or 5.

 

From a cost-benefit standpoint, it just makes no sense to have so many.

[TheRetiredSlave]

My point about chipsets is this: do the R&D once and sell 1 or 2 chipsets vs. doing R&D on 5 different chipsets and selling such a wide array of them while getting less profit from each respectively. This unifies the support needs (more money saved by Intel), unifies the cost (good for gamers, personal server builders, and workstation people who are Intel's consumer bread and butter), and requires there be only 1 or 2 manufacturing processes running simultaneously instead of 4 or 5.

From a cost-benefit standpoint, it just makes no sense to have so many.

Though like I just said, it sounds like you're assuming that they spend tons of money on R&D costs on the low end chips.

I would assume that they actually only manufacture 2-3, because like you say, it doesn't make sense to manufacture 6 completely different chips on the consumer side. Though what I was saying is that the Q87 supports so much more than the H81 that it most likely requires different R&D, so it probably doesn't make sense to just sort through them just with binning. It's probably more economical to do R&D on a lower end chipset and run them with a cheaper manufacturing process with lower grade materials (...in a fabrication plant in a country that pays lower wages, etc, etc)..

Having different products in different tiers allows them to have more flexibility in what markets they can satisfy.. If you're just looking for an ultra cheap board with basic functionality for a budget htpc, it doesn't make sense to pay a ton on a chipset with the same technology as a high-grade workstation board. Like I said, adding features only increases costs, which only decreases sales..

Of course, accurately calculating how much money would be saved or lost by each situation would require information that we don't have access to.

There are advantes and disadvantages to any system. I would assume they do what they do because they have accountants who have already done a detailed analysis of their strategies and product lineup, and those accountants probably found that it would be beneficial to have 6 different chipsets on the consumer side and many more in the enterprise side.