Are Intel's major income m/b chipsets and not cpu's?

[minim3]

Think about it. Yes, intel produced a great cpu. But intel raises the fsb, first to 1066 and next to 1333. With 1333 via and sis can't produce a cheap alternative since they don't have license. Ati's license is about to expire as well. So that leaves m/b manufacturers with intel or nvidia chipsets. Although usually sis,via chipsets sucked, they served a nice role, lowering the prices.
Lately with all the noise about 680i and sata problems (data corruption etc) that leaves us with really no alternative! I know that 680i is high end, but for the ~150$ market only 1 chipset is descent, the P965.

I'm speculating that the 965 chipsets are sold with a generous profit margin, unlike cpu's. (say they cost 10$ to make and been sold for 40$ to the mobo manufacturer). No choice is given, buy it at the fixed price or go out of business. *If anyone has some real info in how much these chipsets actually cost, that would be very helpful.

Are we really gaining anything (performance wise) from a 1066fsb instead of 800? And since there are DDR2-6400 available, why not bump the fsb (officially) into 1600 and run ram-fsb in synch like we used to in p4? (ratio 1:1, 800fsb>DDR-3200).

 

[heartview]

Total sales, I'm not sure. But for profit? CPU's are FAR FAR FAR more profitable for Intel and AMD than any chipsets could be. Think about, the CPU costs typically twice or more what your motherboard does (which includes the chipset and other required components).

 

[minim3]

NO
but if someone wants a feature on an intel board then they have to use an intel proc

hmm, one sentence answers don't tell me anything. I also ask about fsb freq. Why must we go to 1066 and then 1333 and not straight to 1600? The ram is right there! Pls try to expand in your laconian-no answer.

 

[heartview]

hmm, one sentence answers don't tell me anything. I also ask about fsb freq. Why must we go to 1066 and then 1333 and not straight to 1600? The ram is right there! Pls try to expand in your laconian-no answer.

The incremental jumps have more to do with multipliers than anything, as I understand it. They COULD make much bigger jumps than they do, but that would typically increase engineering costs for everyone else involved in making the whole system. With the current smaller step approach they can spread the re-engineering costs over several products and product lines.

Think of it this way. You make car engine and rate its gas mileage based on a speed of 50 MPH. You could probably predict its efficiency at 55 MPH fairly easily and relatively accurately, but what about at 100 MPH? Since most engines lose efficiency the faster they go, it becomes increasingly difficult to judge by how much with bigger jumps in speed.

This same concept can be used with computer hardware. It is typically a lot harder to figure out how a given architecture will perform with a 1 GHz jump in speed versus a 100 MHz jump in speed. Since there are several parts of a system tied to a given clock and its multipliers there is a lot more that can go wrong with big jumps than with little jumps. The current approach is just a lot easier on the other people.

After all, it wouldn't do AMD or Intel any good to make a processor that nobody can engineer a chipset or motherboard for.

 

[minim3]

if intel makes a motherboard with features other boards dont have some people will want to buy the motherboard. if they buy the motherboard they will need to buy a processor too.
with the ram and stuff you add features at different stages to broaden you reach into the the market share

me thinks that instead of have a new m/b with 1600fsb, we get to have these interim steps, so we gonna have to buy 2 or 3 m/b until we get to the 1600fsb for no other reason but greed. I'm wondering how much is the performance gain of 1066 fsb? 5% over 800fsb?

 

[sandmanwn]

as a publicly traded stock intel is required to release this info.
http://www.intel.com/intel/finance/earnings/2006.htm

the break down on page 10 on this pdf doc shows that cpu's gained nearly 3x the revenue of the chipsets and motherboards designed for that specific processor market.
http://www.intel.com/intel/finance/earnings/IntelQ32006EarningsRelease.pdf

you can also see what intel has lost to its competitors from the previous 2005 earnings report.

 

[minim3]

hmm, one sentence answers don't tell me anything. I also ask about fsb freq. Why must we go to 1066 and then 1333 and not straight to 1600? The ram is right there! Pls try to expand in your laconian-no answer.

The incremental jumps have more to do with multipliers than anything, as I understand it. They COULD make much bigger jumps than they do, but that would typically increase engineering costs for everyone else involved in making the whole system. With the current smaller step approach they can spread the re-engineering costs over several products and product lines.

Think of it this way. You make car engine and rate its gas mileage based on a speed of 50 MPH. You could probably predict its efficiency at 55 MPH fairly easily and relatively accurately, but what about at 100 MPH? Since most engines lose efficiency the faster they go, it becomes increasingly difficult to judge by how much with bigger jumps in speed.

This same concept can be used with computer hardware. It is typically a lot harder to figure out how a given architecture will perform with a 1 GHz jump in speed versus a 100 MHz jump in speed. Since there are several parts of a system tied to a given clock and its multipliers there is a lot more that can go wrong with big jumps than with little jumps. The current approach is just a lot easier on the other people.

After all, it wouldn't do AMD or Intel any good to make a processor that nobody can engineer a chipset or motherboard for.

And yet we have people with fsb of more than ~450Mhz * 4. Running rock solid 24/7, with existing hardware. So your example, although a good one, doesn't really apply to this particular case. P4's always worked best when in 1:1 ratio with ram. Is C2D any different? The first thing you do when you buy dd2-800 is drop it down to 1:1 (533) ratio and o/c like mad.

 

[minim3]

as a publicly traded stock intel is required to release this info.
http://www.intel.com/intel/finance/earnings/2006.htm

the break down on page 10 on this pdf doc shows that cpu's gained nearly 3x the profit of the chipsets and motherboards designed for that specific processor market.
http://www.intel.com/intel/finance/earnings/IntelQ32006EarningsRelease.pdf

you can also see what intel has lost to its competitors from the previous 2005 earnings report.

Net output is called "net" because it is "netted" of the value of intermediate goods and services used up. Included in net output may also be inventories of unsold outputs, valued at average market prices. Net output is sometimes also calculated in terms of physical product units which can be sold. *from wikipedia

Net isn't the clean profit one. It is logical to have a 3 times greater net in the cpu's since they costs many times over that of chipsets. True profit margin isn't shown in the pdf.

 

[sandmanwn]

Are Intel's major income m/b chipsets and not cpu's?

eh, it answers your question. cpu's earn intel the most income. that was my only point.

 

[minim3]

Are Intel's major income m/b chipsets and not cpu's?

eh, it answers your question. cpu's earn intel the most income. that was my only point.

Pls tell me how it is the same when say a C2D costs over 100$ to manufactured and sold at 200-300 (profit margin 50-75% for intel?) I assume that higher binned c2d cost more to produce (QoS etc).

But in chipsets (which aren't really cutting edge) manufactured still@130nm cost 10-15$ and sold 40-50$ directly to m/b manufacturers. Cutting out the middle man. Profit margin over 300%.

yeah it's the same. /sarcasm

 

[sandmanwn]

until you can prove what intel sells their chipsets for then i cannot verify your argument.

im guessing you just want to start an argument seeing as you asked the question and then you are now giving us your own answer. :roll:

 

[minim3]

until you can prove what intel sells their chipsets for then i cannot verify your argument.

im guessing you just want to start an argument seeing as you asked the question and then you are now giving us your own answer. :roll:

Well hopefully someone with greater knowledge of the subject will enlighten us. The thing is that until we can figure out which has the greater profit margin (cpus or chipsets) for intel, theres no real way of really knowing. And you guessed wrong! You mistaken the net value for pure profit, when it's clearly not the case. I'm just saying my opinion. It can be changed, but not with false information.

 

[nevesis]

And yet we have people with fsb of more than ~450Mhz * 4. Running rock solid 24/7, with existing hardware. So your example, although a good one, doesn't really apply to this particular case. P4's always worked best when in 1:1 ratio with ram. Is C2D any different? The first thing you do when you buy dd2-800 is drop it down to 1:1 (533) ratio and o/c like mad.

so rock solid @ 1.8 ghz? wow color me impressed.
actually c2d is different, depending on the chipset I know on p965 4:5 is faster in benchmarks

 

[kmjohnso]

You have no idea how much they cost... The cost is not the cost of just producing the product but the capital,materials, design, and RD to develop the processor. How did you come up with $10-15??

 

[sandmanwn]

if i remember correctly the chipsets were moved to 300mm wafers starting about 3rd quarter of last year which puts them on equal footing with the 300mm wafers for processors.

it comes down to 130nm chipsets vs 65nm processors. $40-50 chipsets vs $300 processors. it looks like the processors definitely make more income.

 

[kmjohnso]

That just will effect the production. But the development of the processor and 65nm process and combined costs of the equipment to produce it may offset the per wafer advantage. Chipsets are made using old tech and old equipment. I think you'd have to really know more about operating, liscensing agreements, R/D, etc costs to make such a conclusion. Processors are probably still much more profitable as the competition is lower.

 

[baddad]

Lets look at it this way, Intel makes 1000 chipsets for a 1000 boards at $15 and sells for $50 each, then makes a 1000 CPU's for $100 and sells for $200 each. Which product did they make the most money on?

 

[sandmanwn]

right on

 

[kmjohnso]

That still very very incomplete. If the processor true margins are even remotely close they will make more on the processors. But how do you know the margin? What are you accounting for? If we assume the chips to be sold for 200 and the chipsets sold for 50 then we could obviously solve for what they would need to be to make more for processors:

Profit/Chipset = 50 - Cost/Chipset
Profit/CPU = 200 -Cost/CPU

For Profit/Chipset > Profit/CPU Cost/CPU-Cost/Chipset > 150. So even if they could make chipsets for free. A cpu would have to cost >150 to produce for it to be less profitable. We don't know the costs though??? Unless you work at Intel or have financial records your "proof" is nothing.

This whole argument is pointless. Do you honestly care how much intel is make where?

 

[pausert20]

Pls tell me how it is the same when say a C2D costs over 100$ to manufacturer and sold at 200-300 (profit margin 50-75% for intel?)

It costs considerably less to manufacture a CPU. From what I heard costs for the 130nm process was about $45 to produce a processor. Each decrease in process technology gives about a 30% reduction in cost to manufacture a processor. So now if you do the math a Conroe should be costing Intel about $23 to produce. Remember that cost of fabs are included in the cost and they have been going up. The new 45nm Fabs are costing Intel ~4 to 4.5 Billion.

I assume that higher binned c2d cost more to produce (QoS etc).

No, the cost to produce is the same. Intel just does a more thorough job of bin testing the parts.

But in chipsets (which aren't really cutting edge) manufactured still@130nm

All 965 and the new G33 (Bearlake) chipsets are being produced on a single 90nm 300mm fab.