Linux has not caught on in the *desktop* space. Linux has made good headway in the server space. As people advocating Linux on the desktop, you have to objectively questions the reasons for that.
After all, rational people would rather not have to pay Microsoft or Apple for software that provides an equal amount of service. If a free operating system cannot compete with commercial counterparts that cost infinitely more ($100/$0 = infinity), then you need question why.
There are outliers, but people on *average*, behave rationally - after all, rational behavor is what all of economics is predicated on.
If a business owner or even a family using a computer could cut costs and not lose anything, then why wouldn't they?
Microsoft asserts its monopoly power to force computer vendors into pre-installing windows onto most PCs. One can also claim that Apple has the same issue in that all of their computers come installed with an Apple operating system.
Since PC vendors operate on very thin margins, one would assume that cutting out the cost of the OS would be a tremendous competitive advantage for a prospective PC vendor. If Dell, IBM, HP etc... all need to pay the "Microsoft tax", a competitor should be able to undercut them by offering "equivalent" Linux based machines. However, this has not been a successful strategy.
If Linux can provide an adequate substitution for Windows, then Microsoft really doesn't have any more monopoly pricing power. Therefore, another rational strategy is for a large computer vendor such as Dell to completely drop Microsoft. After all, if Linux can provide the same amount of service that Windows can, why wouldn't Michael Dell just give the finger to Microsoft and drop them completely? If would save them literally *billions* of dollars. In fact, there is a tremendous economic incentive for computer vendors to drop Microsoft. Yet they do not comply. Why?
Perhaps the *assumption* that Linux provides an equal amount of service compared to Windows needs to be re-examined. In fact, by proof of existence, users would rather pay $100 (or whatever Windows costs these days) NOT to use Linux (or BSD or any other free OS). They are acting rationally. If they weren't, some keen business person would be able to *game* that differential and make money (Like the almost non-existent linux-computer vendors).
So the question remains. Why?
And don't say that people are "stupid". They aren't. They behave rationally on average. People choosing servers for example, tend to use Linux without batting an eye.
Ok, even if you use FlOPs you'd still be inaccurate.
The number of instructions used to execute a program on an Intel machine may be very different from the number of instructions used to execute the same program on an AMD machine.
This is because the compiler may choose to use a different set of instructions based on the properties of the underyling CPU. If you have a very large instruction cache, you may use a different set of instructions since you can fit more bytes in the cache - for example. A more complicated example was given with my other post about predicated VLIW.
Let's say your computable task was a matrix multiply of a large matrix. Depending on bandwidth of the memory system and execution core, the compiler will generate very different code. To use the SAME code on different machine architectures is to give an unfair advantage to the machine with the more optimal code.
Even if you had the SAME code, some machines do an internal translation (to micro-ops in the case of x86) which means that underneath the ISA, they are executing a very different program than the one the programmer told it to (but still functionally correct). Some machines may optimize instructions away, or add instructions to the stream in order to speed up overall execution time.
That's why when you run SPECfp or SPECint, you choose the best compiler for your machine. And to measure performance, you measure time and not instructions/second.
You'd do the MUL and ADD on every iteration an predicte the result based on the test. So you'd execute MUL and ADD every iteration but throw away some computes because you only take one mul or add depending on the test.
You'd execute half as many floating point ops. If you have a good branch predictor (say test_func() was very predictable) you might be just as fast as the predicated machine with half as many FLOPS.
FLOPS don't include integer ops. Also, compilers may compile things differently for different machines. For example - SSE3 or Itanium. 64 bit vs. 32 bit. etc...
Intel would clearly win the FLOPs because they have higher peak floating point execution bandwidth - due to higher frequency (on P4). Or more FP execution units on Itaniums.
Performance is measured in seconds, not flops or mhz. Pick a computable task - measure how long it takes each machine to complete it.
People are well aware of the scaling limits and have been for years.
There is a fundamental physical limit that puts a cap on the amount of heat that can be removed from a solid per unit time.
We are fundamentally power limited. Moore's law says the transistor density increases exponentially, but we can't switch those transistors faster because the chip gets too hot and we can't remove that heat fast enough - FUNDAMENTALLY due to the laws of physics.
So there's a tradeoff. Either put more transistors on the chip and reduce speed, or put fewer transistors on the chip and increase speed.
This is very different from the past, when we had the luxury of BOTH increased transistor speed and increased density. The total power was not yet high enough to cause a problem.
For those interested in the details, I refer them to the following paper:
Transmeta was an overhyped technology. YOU are the one that has bought into THEIR marketing. They were supposed to be a performance leader when they got their VC money. When they failed at that, they "transformed" their goals into a power efficiency story.
Ultra low voltage Pentium-M's deliver far higher performance per watt than any Transmeta part.
Guess what - most x86 chips do "code morphing" but in hardware. It was foolish to think that a software solution could be more efficient.
Looking at last years P/E ratio is not the way to value Google or any stock. You need to estimate their growth rate.
Hypothetically, if the earnings increase by 10x and the stock stays at current levels, you'd have a P/E of 8.235
When the stock went IPO last year, they released information about what they earned in the previous year and everyone thought that the $85 IPO price was expensive. Given what they have earned this year - their P/E at IPO time was cheaper than Coca-cola - hardly a growth company.
Growth rate is everything. If you think Google is done growing at 30% a year compounded - sell. If not and you think there is more growth ahead that is sustainable for say 5 years - buy.
Intle NEEDS to prove that they can still make a good x86 chip from "scratch".
I guess that making 32 billion dollars last year MAINLY from x86 sales isn't proof enough.
They invented the microprocessor.
People said it was impossible, but they made the first pipelined x86 (486) People said it was impossible, but they made the first superscalar x86 (Pentium) People said it was impossible, but they made the first out of order x86 (PPro) They made the first multithreaded x86 (P4) They made a dual core x86 with plans for more in the future. The innovations in circuit design are unreal. The ALUs in your 3.0GHz P4 are running at 6.0GHz. Anybody with a degree in electrical engineering can appreciate what an acheivement that is in and of itself is.
All the while maintaining a 6-12 month process step lead over all of the competition (look for 65nm late this year, early next year).
How much more proof do you need?
They are the largest chip company on the planet. Their market cap exceeds their biggest compeitor by about 18x.
Perhaps they pushed too far with the P4 architecture, but that's what innovators do - they push limits and take risks. As far as I can tell, the Athlon/Opteron were 'safer' designs, which in the end, turned out to be pretty competitive. But you don't get to the top or stay there by playing it safe...
Despite all that AMD has accomplished, their stock value is at EXACTLY the same level as it was in 1985.
'maximizing profits' is a nebulous term, and often not making the maximal profit available today makes sure I can continue making profits tommorow.
How about "Maximizing profits over the lifetime of the corporation"?
A share of stock entitles you to a share of all future earnings. So you value the stock by estimating all future earnings and then discounting that number back to the present value using the prevailing interest/inflation rates.
You wouldn't get into toilet paper sales if you were Apple because some giant consumer product corporation would crush you - you have no toilet paper competitive advantage - you'd probably not be able to secure any reasonable paper pricing and you wouldn't know the first thing about manufacturing toilet paper in an efficient manner. If you did, and had a reasonable business case for making that advantage profitable, then of course you would go into toilet paper - why wouldn't you? GE used to make only lightbulbs 100 years ago, but now they make nuclear reactors and provide financial services. If you can make a buck, you do it. Only the most adaptable businesses survive.
The parent poster is only pointing out that not being the market leader tommorow is often a safer strategy for shareholders
Maybe. But that's debatable. In the technology business, you often see a "winner take all" pattern. See Microsoft, Intel, Dell, Ebay, Adobe, Ipod, etc...
#1's take the prize, #2's still exist, but they stagnate - #3's are rarer still.
The Ipod is the only reason Apple's share price has multiplied over the past few years - in fact in the latest quarter, Ipod sales accounted for near 50% of total profits. It's getting to the point where computer profits will soon be secondary. Another interesting stat from the 10Q is that notebook profits have only grown 7% (vs about 333% for Ipods). In other words, the P/E multiple on Apple (about 36x last years earnings) is based entirely on Ipods.
What would happen to the stock price if Apple didn't have such a dominant position in Ipods? What happens when everybody and their mother has an Ipod? How is deccelerating Ipod growth and stagnant computer growth going to sustain a 36x multiple on the stock? Shareholders want to know...
I don't think that saying - "we're happy being a niche player" is going to cut it on Wall Street.
Not every company needs to (or should) try to maximize sales and market penetration like Microsoft - just like every person doesn't need to try to be as rich as Bill Gates, as musical as Mozart, as tall as Shaq, etc.
If you are a private company - fine. You can do that.
But apple is a publicly traded company that has a fiduciary duty to MAXIMIZE profits for their shareholders. Apple would be negligent in this if they did anything less. The shareholders (the board) could have Jobs fired if they felt he wasn't doing everything possible to maximize profit.
If profits and sales stagnate, so will the stock price which will upset shareholders.
1) Fast installs - Ummm, did you buy your software on DRAM? It has to get off the disk or network first which is obviously going to be your bottleneck.
2) Memory swap file - again - dumb. We have 64 bit computers now people. Just get a motherboard and throw in 4GB or more of RAM and you'll get a fast-path (high bandwidth) to that memory instead of some weak I/O connection (SATA vs. FSB you decide) that is entirely managed by the OS's virtual memory manager. Still swapping? Add more ram...
3) Again - just add more ram to the motherboard and let the OS take care of managing shit for you.
4) Many other things? Like what?
This is just stupid. This is for computers that cannot accept large amounts of ram on the motherboard.
Just buy a motherboard that can accept a lot of RAM and your shit will just get cached into ordinary DRAM which has much more bandwidth than some I/O device.
Arguments that I have seen that don't make sense:
1) Applications will startup faster
- Why? They still need to load into RAM from disk the first time. Why cripple yourself with an I/O DRAM card when you can just throw DRAM onto the high memory bandwidth memory channel on the motherboard?
2) Lower latency - Uhmm - refer to above - your shit is persistent on DISK, so it still has to load into RAM and you still need to wait.
3) Use as a write buffer. Again - use DRAM - the OS takes care of this - there's this thing called virtual memory... It was invented in the 60s or 70s...real cutting edge technology.
If you could create a consioucness with an algorithm, then you could run it on a regular PC - it would still be 'aware' but would just perceive time on a much slower scale than we do due to it running so slow.
Technology has nothing to do with it (PHP, Perl, Apache, blah blah blah whatever...). I mean Ebay is one of the most valuable internet sites out there and their interface is pretty piss poor.
What matters is that a LARGE and YOUNG group of people visit the site and therefore you can sell advertisors to that audience. Fox believes that this audience is worth $580 million. If you can attract a similar audience with your website, you could retire very rich.
No it doesn't. The wall street opinion of a company has a pretty limited perspective.
The shareholders are the OWNERS of the company. It is McNealy's JOB to increase shareholder value. He has no other function. Therefore, he (management) has failed miserably. That's the metric that is used in corporate America, Europe, Japan etc... Because the stock price has not gone up - investors are skeptical of Sun's future. Do you have your hard earned money in SUNW? The board of directors, which represent the shareholder interests, could have him fired.
It's scalability at the high end, especially for things like networking performance.
So you believe that the benefit of networking performance justifies the cost (several billions) of SPARC? I'm confused - how exactly does SPARC excel at "networking performance"? I thought that was a "system" thing, not a "processor" thing. You can buy TCP/IP offload engine chips for far cheaper than funding a 600 person engineering team for several years.
This happens - there is nothing new about late processors.
Why have late processors at all when there are perfectly good ones available from Intel/AMD?
Furthermore, Niagra = Cell so it's not exactly a special Sun-only technology.
How is this relevant?
You were saying that the performance of 8 cores (in-order) with 4 threads per core was going to be a competitive advantage for Sun. I am saying that IBM already has similar technology with CELL which it is already ramping in volume. Again - Sun's processor development is late and redundant.
They are still selling the stuff!
But for how long? Their cost structure is totally out of whack with the industry. They're getting priced into oblivion for the reasons I've been trying to state. Sun needs to cut cut cut - to save costs and focus on things that they are strong at.
No - it is about competing with those in the areas where those are weak: Linux in terms of high-scalability, Dell in terms of high-performance servers, Intel in terms of core performance and high-end-scalability.
They can compete with Linux in terms of scalability. They cannot compete with Dell on the low end - too much pricing pressure. They cannot compete with IBM on the high end - their hardware isn't competitive enough. They certainly cannot compete with Intel in processor design - they've been behind for about 8 years now. They've cancelled their giant UltraSPARC-V effort. Intel/AMD have scalable processors, it's just that nobody has built systems to take advantage of that yet. Again - what exactly is their strategy?
I'd be interested in how a 8-core processor running at high GHz is going to be slower than 2-core processors at the same speed.
Niagra is 1.4GHz. The competition is twice that. Niagra is using SINGLE-SCALAR IN-ORDER processors. I'm guessing you're not a processor designer (I am), but this is a huge performance disadvantage. How else are you going to pack 8 processors onto a die of that size? - they needed to make them simple and slow. You get a 2x performance from superscalar/out-of-order, and another 2x from frequency. That's 4x per core. So 8/4 = 2 which is dual core. IE - they have equivalent throughput to current dual core offerings.
Sun are changing, and some of the markets you mention are shrinking, but it would be a sad day if everything was stock-market driven and large sections of a company that included huge amounts of expertise and IP were dumped just to please short-term interests of investors.
If the IP is dumped then it had no value. Ultimately, the only reason someone is going to give you money to build a computer is to see a positive return on that money. My belief is that Sun does have valuable IP - but it's not in processors. It's in the software.
I often hear this, and to me it seems a little... premature. The only people who are in a position to judge this in reality are the management themselves
You mean like Ed Zander, who is a great manager, who left to go to Motorola? There's a pretty handy metric by which you can measure the performance of management. It's the stock price. I think that speaks for itself.
My impression is that these are pretty good ways forward. There is about to be a phenomenal demand for storage because of the growth of networked multimedia and because of new regulations in the USA and elsewhere for businesses to keep all information about everything. Storage and archiving are going to be big, and if you can get a complete systems integrated with your servers from one source, this is a real advantage.
But they want to compete with EMC which uses Dell for distribution? Hmm, good luck with that!
As for the processors, these tend to leap-frog each other in cycles. Intel and AMD are fast now, but they have hit a problem in terms of heat and are moving to multi-core. The new SPARC is going to be way ahead terms of performance
The point of ditching SPARC is to reduce costs. If they leap-frog each other (which is a dubious assumption - SPARC has been way behind for a while, which is why they completely lost the workstation market to Dell), that would suggest that Intel/AMD are doing just fine bringing capable processors to market. What exactly is the competitive advantage of using SPARC that would justify its enormous investment? At some point investors want to see a return on that investment. They've been waiting a while... Ultrasparc V was cancelled. Sun has never built an out-of-order processor. Niagra is late. Furthermore, Niagra = Cell so it's not exactly a special Sun-only technology.
My main point is this. Sun delivers mediocre hardware, but they can't even deliver that at a price the market wants because competitors offer solutions that are far cheaper and still get the job done. The management strategy seems to be this: Let's compete with Linux, EMC, Dell, IBM and Intel all at once! This is madness - you can't be good at everything. Vertically integrated computer companies are going the way of the dodo. Pick some key areas where you have a competitive advantage (Like software and systems) and focus there while outsourcing the rest. Your costs will come down and you can differentiate yourself from the competition. What's next? Is Sun going to fab their own custom DRAM? Does management even know why their computers are expensive relative to the competition?
open-sourcing of Solaris, and running it non-SPARC was a great idea. It means that developers can use Solaris for free, and then easily migrate to very fast SPARC-based systems when demand requires this.
Fast SPARC is an oxymoron. I'll believe it when I see it.
It certainly is dependent to some extent, but if you look at the details of these services and contracts on their reports you will see that they mention both hardware and software. This is a significant change on what they were doing years ago.
I would also like to see Sun move away from the hardware business and start to focus on software. Solaris is solid and java is good for the enterprise. Sun's strength is in software and systems engineering. I'd like to see them push standardization of commodity hardware up the enterprise stack. Let them spec out the enterprise systems that Dell can go off and build. McNealy...You DO NOT want to compete with Dell. You DO NOT want to compete with Intel (kill SPARC already).
If they can get their software stack to run on this commodity hardware, then they have a survivable future. They can become the "Microsoft" of the server world. Cleary they can differentiate themselves from Linux and Microsoft with better features in Solaris that are more suited towards the enterprise. Otherwise, I just see Linux based systems becoming more and more capable as IBM continues to add more enterprise features year after year.
Sun management seems clueless. Witness the latest aquisition of a storage company. They still want to sell hardware? They still pour billions into SPARC development? Why? Intel and AMD do a perfectly good job of making fast scalable processors. It's just that nobody builds systems that take advantage of that yet. Here's your chance Sun...
Irrelevant in the "makes money" sense of the word. The C language has a huge and lasting impact on the computing world, but nobody cares that it was invented at AT&T.
I like to buy stocks on the cheap, so I have been investigating Sun as a potential investment opportunity. Sometimes the market overly discounts bad news. In the case of Sun, it doesn't look good. I wouldn't put my hard earned money into the stock.
Here is a direct quote from their latest 10Q SEC filing:
Our Products net revenue was unfavorably impacted by competition and a continuing market shift in overall computer system demand away from our data center servers towards the usage of enterprise and entry level servers
and...
services revenue consists primarily of maintenance contract revenue, which is recognized ratably over the contractual period
Hmm...maintenance contracts that will run out once the customers move away to cheaper/more-capable hardware solutions from IBM/Dell and the likes. So, contrary to your assertation, service revenue is hugely dependent on your hardware lineup.
The outlook is pretty grim... Ultimately, you need a tangible hardware product to sell in order to add the lucrative services to later on. Sun just doesn't have anything that compelling to offer in their product pipeline.
Slashdot Mirror
Linux has not caught on in the *desktop* space. Linux has made good headway in the server space. As people advocating Linux on the desktop, you have to objectively questions the reasons for that.
After all, rational people would rather not have to pay Microsoft or Apple for software that provides an equal amount of service. If a free operating system cannot compete with commercial counterparts that cost infinitely more ($100/$0 = infinity), then you need question why.
There are outliers, but people on *average*, behave rationally - after all, rational behavor is what all of economics is predicated on.
If a business owner or even a family using a computer could cut costs and not lose anything, then why wouldn't they?
Microsoft asserts its monopoly power to force computer vendors into pre-installing windows onto most PCs. One can also claim that Apple has the same issue in that all of their computers come installed with an Apple operating system.
Since PC vendors operate on very thin margins, one would assume that cutting out the cost of the OS would be a tremendous competitive advantage for a prospective PC vendor. If Dell, IBM, HP etc... all need to pay the "Microsoft tax", a competitor should be able to undercut them by offering "equivalent" Linux based machines. However, this has not been a successful strategy.
If Linux can provide an adequate substitution for Windows, then Microsoft really doesn't have any more monopoly pricing power. Therefore, another rational strategy is for a large computer vendor such as Dell to completely drop Microsoft. After all, if Linux can provide the same amount of service that Windows can, why wouldn't Michael Dell just give the finger to Microsoft and drop them completely? If would save them literally *billions* of dollars. In fact, there is a tremendous economic incentive for computer vendors to drop Microsoft. Yet they do not comply. Why?
Perhaps the *assumption* that Linux provides an equal amount of service compared to Windows needs to be re-examined. In fact, by proof of existence, users would rather pay $100 (or whatever Windows costs these days) NOT to use Linux (or BSD or any other free OS). They are acting rationally. If they weren't, some keen business person would be able to *game* that differential and make money (Like the almost non-existent linux-computer vendors).
So the question remains. Why?
And don't say that people are "stupid". They aren't. They behave rationally on average. People choosing servers for example, tend to use Linux without batting an eye.
Ok, even if you use FlOPs you'd still be inaccurate.
The number of instructions used to execute a program on an Intel machine may be very different from the number of instructions used to execute the same program on an AMD machine.
This is because the compiler may choose to use a different set of instructions based on the properties of the underyling CPU. If you have a very large instruction cache, you may use a different set of instructions since you can fit more bytes in the cache - for example. A more complicated example was given with my other post about predicated VLIW.
Let's say your computable task was a matrix multiply of a large matrix. Depending on bandwidth of the memory system and execution core, the compiler will generate very different code. To use the SAME code on different machine architectures is to give an unfair advantage to the machine with the more optimal code.
Even if you had the SAME code, some machines do an internal translation (to micro-ops in the case of x86) which means that underneath the ISA, they are executing a very different program than the one the programmer told it to (but still functionally correct). Some machines may optimize instructions away, or add instructions to the stream in order to speed up overall execution time.
That's why when you run SPECfp or SPECint, you choose the best compiler for your machine. And to measure performance, you measure time and not instructions/second.
Just to elaborate:
Let's say I had some pseudo-code:
for (1 to 100 million) {
if (test_function() == true) {
A = A * B;
} else {
A = A + B
}
}
A VLIW might compile to
LOOP:
PREDICATE = CALL TEST_FUNC
MUL, ADD, PREDICATE
JMP LOOP (exit when done)
You'd do the MUL and ADD on every iteration an predicte the result based on the test. So you'd execute MUL and ADD every iteration but throw away some computes because you only take one mul or add depending on the test.
A non-predicted machine might do this:
LOOP:
CONDITION = CALL TEST_FUNC
JMP CONDITION, non-mul
MUL
jmp LOOP
non-mul:
ADD
jmp LOOP
You'd execute half as many floating point ops. If you have a good branch predictor (say test_func() was very predictable) you might be just as fast as the predicated machine with half as many FLOPS.
FLOPS don't include integer ops.
Also, compilers may compile things differently for different machines. For example - SSE3 or Itanium. 64 bit vs. 32 bit. etc...
Uhh...flops is just as stupid as Mhz.
Intel would clearly win the FLOPs because they have higher peak floating point execution bandwidth - due to higher frequency (on P4). Or more FP execution units on Itaniums.
Performance is measured in seconds, not flops or mhz. Pick a computable task - measure how long it takes each machine to complete it.
People are well aware of the scaling limits and have been for years.
f -Proc-IEEE-Limits.pdf
There is a fundamental physical limit that puts a cap on the amount of heat that can be removed from a solid per unit time.
We are fundamentally power limited. Moore's law says the transistor density increases exponentially, but we can't switch those transistors faster because the chip gets too hot and we can't remove that heat fast enough - FUNDAMENTALLY due to the laws of physics.
So there's a tradeoff. Either put more transistors on the chip and reduce speed, or put fewer transistors on the chip and increase speed.
This is very different from the past, when we had the luxury of BOTH increased transistor speed and increased density. The total power was not yet high enough to cause a problem.
For those interested in the details, I refer them to the following paper:
http://www.intel.com/research/documents/Bourianof
Give me a break.
Transmeta was an overhyped technology. YOU are the one that has bought into THEIR marketing. They were supposed to be a performance leader when they got their VC money. When they failed at that, they "transformed" their goals into a power efficiency story.
Ultra low voltage Pentium-M's deliver far higher performance per watt than any Transmeta part.
Guess what - most x86 chips do "code morphing" but in hardware. It was foolish to think that a software solution could be more efficient.
P/E Ratio: 82.35. Any other questions?
Looking at last years P/E ratio is not the way to value Google or any stock. You need to estimate their growth rate.
Hypothetically, if the earnings increase by 10x and the stock stays at current levels, you'd have a P/E of 8.235
When the stock went IPO last year, they released information about what they earned in the previous year and everyone thought that the $85 IPO price was expensive. Given what they have earned this year - their P/E at IPO time was cheaper than Coca-cola - hardly a growth company.
Growth rate is everything. If you think Google is done growing at 30% a year compounded - sell. If not and you think there is more growth ahead that is sustainable for say 5 years - buy.
Sounds like x86 :)
Crap, crap, crap, loaded with more crap. But 99% of the world uses it due to market forces.
Intel's marketing prowess is much better than its competition
I think you mean *manufacturing*
When AMD can pony up 10 billion dollars a year to invest in fab capacity then Intel has something to worry about.
Intle NEEDS to prove that they can still make a good x86 chip from "scratch".
I guess that making 32 billion dollars last year MAINLY from x86 sales isn't proof enough.
They invented the microprocessor.
People said it was impossible, but they made the first pipelined x86 (486)
People said it was impossible, but they made the first superscalar x86 (Pentium)
People said it was impossible, but they made the first out of order x86 (PPro)
They made the first multithreaded x86 (P4)
They made a dual core x86 with plans for more in the future.
The innovations in circuit design are unreal. The ALUs in your 3.0GHz P4 are running at 6.0GHz. Anybody with a degree in electrical engineering can appreciate what an acheivement that is in and of itself is.
All the while maintaining a 6-12 month process step lead over all of the competition (look for 65nm late this year, early next year).
How much more proof do you need?
They are the largest chip company on the planet.
Their market cap exceeds their biggest compeitor by about 18x.
Perhaps they pushed too far with the P4 architecture, but that's what innovators do - they push limits and take risks. As far as I can tell, the Athlon/Opteron were 'safer' designs, which in the end, turned out to be pretty competitive. But you don't get to the top or stay there by playing it safe...
Despite all that AMD has accomplished, their stock value is at EXACTLY the same level as it was in 1985.
'maximizing profits' is a nebulous term, and often not making the maximal profit available today makes sure I can continue making profits tommorow.
How about "Maximizing profits over the lifetime of the corporation"?
A share of stock entitles you to a share of all future earnings. So you value the stock by estimating all future earnings and then discounting that number back to the present value using the prevailing interest/inflation rates.
You wouldn't get into toilet paper sales if you were Apple because some giant consumer product corporation would crush you - you have no toilet paper competitive advantage - you'd probably not be able to secure any reasonable paper pricing and you wouldn't know the first thing about manufacturing toilet paper in an efficient manner. If you did, and had a reasonable business case for making that advantage profitable, then of course you would go into toilet paper - why wouldn't you? GE used to make only lightbulbs 100 years ago, but now they make nuclear reactors and provide financial services. If you can make a buck, you do it. Only the most adaptable businesses survive.
The parent poster is only pointing out that not being the market leader tommorow is often a safer strategy for shareholders
Maybe. But that's debatable. In the technology business, you often see a "winner take all" pattern. See Microsoft, Intel, Dell, Ebay, Adobe, Ipod, etc...
#1's take the prize, #2's still exist, but they stagnate - #3's are rarer still.
The Ipod is the only reason Apple's share price has multiplied over the past few years - in fact in the latest quarter, Ipod sales accounted for near 50% of total profits. It's getting to the point where computer profits will soon be secondary. Another interesting stat from the 10Q is that notebook profits have only grown 7% (vs about 333% for Ipods). In other words, the P/E multiple on Apple (about 36x last years earnings) is based entirely on Ipods.
What would happen to the stock price if Apple didn't have such a dominant position in Ipods? What happens when everybody and their mother has an Ipod? How is deccelerating Ipod growth and stagnant computer growth going to sustain a 36x multiple on the stock? Shareholders want to know...
I don't think that saying - "we're happy being a niche player" is going to cut it on Wall Street.
It'd be far harder for a hacker to find a way to optimize the binary than change some constant
Then it's time to bust out the dynamic recompiler
Not every company needs to (or should) try to maximize sales and market penetration like Microsoft - just like every person doesn't need to try to be as rich as Bill Gates, as musical as Mozart, as tall as Shaq, etc.
If you are a private company - fine. You can do that.
But apple is a publicly traded company that has a fiduciary duty to MAXIMIZE profits for their shareholders. Apple would be negligent in this if they did anything less. The shareholders (the board) could have Jobs fired if they felt he wasn't doing everything possible to maximize profit.
If profits and sales stagnate, so will the stock price which will upset shareholders.
Yes, you're technically savvy...
1) Fast installs - Ummm, did you buy your software on DRAM? It has to get off the disk or network first which is obviously going to be your bottleneck.
2) Memory swap file - again - dumb. We have 64 bit computers now people. Just get a motherboard and throw in 4GB or more of RAM and you'll get a fast-path (high bandwidth) to that memory instead of some weak I/O connection (SATA vs. FSB you decide) that is entirely managed by the OS's virtual memory manager. Still swapping? Add more ram...
3) Again - just add more ram to the motherboard and let the OS take care of managing shit for you.
4) Many other things? Like what?
This is just stupid. This is for computers that cannot accept large amounts of ram on the motherboard.
Perhaps people don't understand virtual memory?
We now have 64 bit computers people.
Just buy a motherboard that can accept a lot of RAM and your shit will just get cached into ordinary DRAM which has much more bandwidth than some I/O device.
Arguments that I have seen that don't make sense:
1) Applications will startup faster
- Why? They still need to load into RAM from disk the first time. Why cripple yourself with an I/O DRAM card when you can just throw DRAM onto the high memory bandwidth memory channel on the motherboard?
2) Lower latency - Uhmm - refer to above - your shit is persistent on DISK, so it still has to load into RAM and you still need to wait.
3) Use as a write buffer. Again - use DRAM - the OS takes care of this - there's this thing called virtual memory... It was invented in the 60s or 70s...real cutting edge technology.
4) Until they start selling software on DRAM
The problem is not the hardware.
It's the software.
If you could create a consioucness with an algorithm, then you could run it on a regular PC - it would still be 'aware' but would just perceive time on a much slower scale than we do due to it running so slow.
Wow, a 4-issue in order MIPS? That's adorable :)
Technology has nothing to do with it (PHP, Perl, Apache, blah blah blah whatever...). I mean Ebay is one of the most valuable internet sites out there and their interface is pretty piss poor.
What matters is that a LARGE and YOUNG group of people visit the site and therefore you can sell advertisors to that audience. Fox believes that this audience is worth $580 million. If you can attract a similar audience with your website, you could retire very rich.
No it doesn't. The wall street opinion of a company has a pretty limited perspective.
The shareholders are the OWNERS of the company. It is McNealy's JOB to increase shareholder value. He has no other function. Therefore, he (management) has failed miserably. That's the metric that is used in corporate America, Europe, Japan etc... Because the stock price has not gone up - investors are skeptical of Sun's future. Do you have your hard earned money in SUNW? The board of directors, which represent the shareholder interests, could have him fired.
It's scalability at the high end, especially for things like networking performance.
So you believe that the benefit of networking performance justifies the cost (several billions) of SPARC? I'm confused - how exactly does SPARC excel at "networking performance"? I thought that was a "system" thing, not a "processor" thing. You can buy TCP/IP offload engine chips for far cheaper than funding a 600 person engineering team for several years.
This happens - there is nothing new about late processors.
Why have late processors at all when there are perfectly good ones available from Intel/AMD?
Furthermore, Niagra = Cell so it's not exactly a special Sun-only technology.
How is this relevant?
You were saying that the performance of 8 cores (in-order) with 4 threads per core was going to be a competitive advantage for Sun. I am saying that IBM already has similar technology with CELL which it is already ramping in volume. Again - Sun's processor development is late and redundant.
They are still selling the stuff!
But for how long? Their cost structure is totally out of whack with the industry. They're getting priced into oblivion for the reasons I've been trying to state. Sun needs to cut cut cut - to save costs and focus on things that they are strong at.
No - it is about competing with those in the areas where those are weak: Linux in terms of high-scalability, Dell in terms of high-performance servers, Intel in terms of core performance and high-end-scalability.
They can compete with Linux in terms of scalability. They cannot compete with Dell on the low end - too much pricing pressure. They cannot compete with IBM on the high end - their hardware isn't competitive enough. They certainly cannot compete with Intel in processor design - they've been behind for about 8 years now. They've cancelled their giant UltraSPARC-V effort. Intel/AMD have scalable processors, it's just that nobody has built systems to take advantage of that yet. Again - what exactly is their strategy?
I'd be interested in how a 8-core processor running at high GHz is going to be slower than 2-core processors at the same speed.
Niagra is 1.4GHz. The competition is twice that. Niagra is using SINGLE-SCALAR IN-ORDER processors. I'm guessing you're not a processor designer (I am), but this is a huge performance disadvantage. How else are you going to pack 8 processors onto a die of that size? - they needed to make them simple and slow. You get a 2x performance from superscalar/out-of-order, and another 2x from frequency. That's 4x per core. So 8/4 = 2 which is dual core. IE - they have equivalent throughput to current dual core offerings.
Sun are changing, and some of the markets you mention are shrinking, but it would be a sad day if everything was stock-market driven and large sections of a company that included huge amounts of expertise and IP were dumped just to please short-term interests of investors.
If the IP is dumped then it had no value. Ultimately, the only reason someone is going to give you money to build a computer is to see a positive return on that money. My belief is that Sun does have valuable IP - but it's not in processors. It's in the software.
Genius post. Thanks - I enjoyed reading it.
Slashdot is libertarian? Bwahahahaha.
You're new here aren't you?
I often hear this, and to me it seems a little ... premature. The only people who are in a position to judge this in reality are the management themselves
You mean like Ed Zander, who is a great manager, who left to go to Motorola? There's a pretty handy metric by which you can measure the performance of management. It's the stock price. I think that speaks for itself.
My impression is that these are pretty good ways forward. There is about to be a phenomenal demand for storage because of the growth of networked multimedia and because of new regulations in the USA and elsewhere for businesses to keep all information about everything. Storage and archiving are going to be big, and if you can get a complete systems integrated with your servers from one source, this is a real advantage.
But they want to compete with EMC which uses Dell for distribution? Hmm, good luck with that!
As for the processors, these tend to leap-frog each other in cycles. Intel and AMD are fast now, but they have hit a problem in terms of heat and are moving to multi-core. The new SPARC is going to be way ahead terms of performance
The point of ditching SPARC is to reduce costs. If they leap-frog each other (which is a dubious assumption - SPARC has been way behind for a while, which is why they completely lost the workstation market to Dell), that would suggest that Intel/AMD are doing just fine bringing capable processors to market. What exactly is the competitive advantage of using SPARC that would justify its enormous investment? At some point investors want to see a return on that investment. They've been waiting a while... Ultrasparc V was cancelled. Sun has never built an out-of-order processor. Niagra is late. Furthermore, Niagra = Cell so it's not exactly a special Sun-only technology.
My main point is this. Sun delivers mediocre hardware, but they can't even deliver that at a price the market wants because competitors offer solutions that are far cheaper and still get the job done. The management strategy seems to be this: Let's compete with Linux, EMC, Dell, IBM and Intel all at once! This is madness - you can't be good at everything. Vertically integrated computer companies are going the way of the dodo. Pick some key areas where you have a competitive advantage (Like software and systems) and focus there while outsourcing the rest. Your costs will come down and you can differentiate yourself from the competition. What's next? Is Sun going to fab their own custom DRAM? Does management even know why their computers are expensive relative to the competition?
open-sourcing of Solaris, and running it non-SPARC was a great idea. It means that developers can use Solaris for free, and then easily migrate to very fast SPARC-based systems when demand requires this.
Fast SPARC is an oxymoron. I'll believe it when I see it.
It certainly is dependent to some extent, but if you look at the details of these services and contracts on their reports you will see that they mention both hardware and software. This is a significant change on what they were doing years ago.
I would also like to see Sun move away from the hardware business and start to focus on software. Solaris is solid and java is good for the enterprise. Sun's strength is in software and systems engineering. I'd like to see them push standardization of commodity hardware up the enterprise stack. Let them spec out the enterprise systems that Dell can go off and build. McNealy...You DO NOT want to compete with Dell. You DO NOT want to compete with Intel (kill SPARC already).
If they can get their software stack to run on this commodity hardware, then they have a survivable future. They can become the "Microsoft" of the server world. Cleary they can differentiate themselves from Linux and Microsoft with better features in Solaris that are more suited towards the enterprise. Otherwise, I just see Linux based systems becoming more and more capable as IBM continues to add more enterprise features year after year.
Sun management seems clueless. Witness the latest aquisition of a storage company. They still want to sell hardware? They still pour billions into SPARC development? Why? Intel and AMD do a perfectly good job of making fast scalable processors. It's just that nobody builds systems that take advantage of that yet. Here's your chance Sun...
McNealy has to go... This isn't 1999 anymore.
Irrelevant in the "makes money" sense of the word. The C language has a huge and lasting impact on the computing world, but nobody cares that it was invented at AT&T.
I like to buy stocks on the cheap, so I have been investigating Sun as a potential investment opportunity. Sometimes the market overly discounts bad news. In the case of Sun, it doesn't look good. I wouldn't put my hard earned money into the stock.
Here is a direct quote from their latest 10Q SEC filing:
Our Products net revenue was unfavorably impacted by competition and a continuing market shift in overall computer system demand away from our data center servers towards the usage of enterprise and entry level servers
and...
services revenue consists primarily of maintenance contract revenue, which is recognized ratably over the contractual period
Hmm...maintenance contracts that will run out once the customers move away to cheaper/more-capable hardware solutions from IBM/Dell and the likes. So, contrary to your assertation, service revenue is hugely dependent on your hardware lineup.
The outlook is pretty grim... Ultimately, you need a tangible hardware product to sell in order to add the lucrative services to later on. Sun just doesn't have anything that compelling to offer in their product pipeline.