In Part 1 I discussed how the software development world is about to be turned on it's head. Now in Part 2 I look at how the hardware world may be about to undergo even bigger changes and why it wont be a hardware manufacturer leading the way.
Post a Comment
"if AMD was to include "code-morphing" software which recompiled the 32 bit x86 code to use these extra registers, would 32 bit performance be as good as it is in the hardware compatibility mode? Could performance even go up?"
HP's Dyanamo project did precisely this. It translated PA-RISC code to PA-RISC code and achieved an average speedup of about 20% over the output of the optimizing C compiler. A native-to-native dynamic translator like Dynamo has the advantage that it does not have to do extensive translation (the native code can run as-is) and so can focus on making dynamic optimizations (eliminating branches are runtime, etc). I wouldn't be surprised if some form of Dynamo resurfaced to get better performance out of the Itanium.
The Itanium is interesting for other reasons. Basically, Itanium has more functional units than code can use effectively. Getting 3-way parallelism out of C code is tough, and the Itanium2 has 6 (!) integer units. One can easily see a good compiler taking advantage of those extra units to do stuff like range-checking, dynamic type-checks, etc. It also has a short 8-stage (!) pipeline that minimizes the cost of branches. Combined with something like Dynamo to help optimize-out the cost of dynamic dispatch, and and Itanium becomes one hell of an attractive CPU for dynamic/polymorphic languages.
The author tries to paint a picture made of self speculation about a computer armageddon, streight out of a movie! We have plot twists, bad guys, good guys, and the final saviors coming in just in the nick of time (Will the linux desktop be ready in time?)
Sadly, things in real life are much more boring. Microsoft are not going to be selling 'Office Boxs' due to it not being worth the price. With XBox, they can use the games for the profits, but with this supposidly 'Office Box' it would have to be considerably cheaper then what someone can get elsewhere.
From dell, one can get the lowest end system with windows for $200. For companies buying in bulk, they can probably get such prices for less.
So where will the profit from these office boxs come from? Hardware? They gotta be cheaper then the competition of a full fledged computer, and so just like with the XBox, they will lose money on the hardware.
Software? The minute you actually have to pay money for software, you lose the vendor lockin, and people will start looking for alteernatives (i.e. SUN will push star office as a replacement etc...)
well, if MS sold those boxes for 150 bucks, then had a subscription service for businesses for their software, that would be a good deal.
"A multi-CPU design allow Microsoft to run not only applications at speed but also removes the need for various hardware devices (sound chip etc.) This is useful in a consumer games machine but this sheer power will also enable emulated x86 applications to be run alongside native apps at speed, a multi-PowerPC X-Box 2 makes a powerful - and low cost - competitor to a PC. "
This does not make much sense to me... Why dedicate an entire powerPC cpu to doing sound when you could use a dedicated dsp like an Atmel for much cheaper?
Other dedicated hardware like GPUs are also much cheaper and faster at what they do than a general purpose CPU.
What other hardware do you need a PowerPC dedicated processor for? The joystick?
I guess running x86 apps alongside native ones would be cool, but I don't see Microsoft paying for extra CPUs just to make this work a little smoother.
I bet that if the other processors in the XBox2 are at all related to PowerPC, they will not be general purpose, or able to run code efficiently enough for emulation of X86, but rather highly modified versions, more like a normal DSP or GPU.
I don't see how it would be wise for Microsoft to enter the hardware market. Right now their OS and Office business units are the only ones that make money. If they enter the hardware market they are then competing againt those same companies that currently buy products from these two units.
Think about it, if I was Dell and all of a sudden Microsoft was comepeting against me (whether their product was cheaper or not), I would very quickly be looking for an alternative to be installing on my PCs. I think the old term for this is "biting the hand that feeds you".
It's just not feasible. The whole world -- big companies, big countries -- is arrayed against anything like this happening. I can't tell you how this fails, or how MS's current dominance comes undone. But one way or the other, MS will fail. They got something almost by accident - this ridiculous monopoly. It's an accident of history, a blip. It'll be gone. The only question is how long it lasts. I expect a lot of thrashing about from MS to prolong it, but so what.
While I don't agree with it all this is another good article in the main. Particually MS trying to use Palladium to extend their computer tax to hardware even if it doesn't come with there OS, or from a company that has to charge you for their OS even if you don't want it.
I don't beleive that the Office Box thinish client idea will pan out. At the moment the XBox hardware looses them money, which they recoupe in lincencing development kits (the same as the rest of the Console industry, the hardware is a loss leader for getting users that will get you developers who you get your money off). So an Office box would not get them any real revenue when they can have the same type of thing produced and have the OEM foot the hardware bill, remember most OEM's have razor thin margins, while they reap the larger profits.
Let's say that IBM, Intel, and AMD won't or can't hold off the ridiculous scenario written up here. How long do you think one of the Korean or other CPU manufacturers would take to make a total non MS computer for installing Linux on? A few days? Maybe a few weeks? Actually they are already out there. Linux users would just change to those and bypass MS.
But what is proposed isn't going to happen. IBM, Intel, AMD, Dell, etc., aren't going to let it happen.
Microsoft building a computer... Honestly, I do actually like the sound of that. I have yet to run into any major problems with Windows or any Microsoft hardware (mouse, keyboard, etc.), so if MS went into the computer building business, I would certainly take a look at what they had to offer.
You can say a lot of things about these articles, but the author sure knows how to write a great cliff hanger.
Tax.Net :
it the old dream of MS to sell serices via the net on certificated Hardware. From Born to Death, you pay to MS not less....
it's so easy : don't buy it.
it's so easy : would you buy socks which sues you if you use them with the wrong shoes ?
Hopefully, Europe is far away from the US-Law and, hopefully, the EU will resist.
krgds,
Frank
What does the Xbox have to do with a hill of beans? Just because MS is able to choose what processor they want for the Xbox(duhh, they make it), what does that have to do with the PC? Microsoft can't just choose what processor they want(well, they can, but that would be utterly stupid). What about all the x86 code out there? No, you can't just throw .NET out there yet. It's going to take years and years and years for .NET to come out in shrink-wrap, and even then there will be billions of lines of x86 code out there. I guess they could throw some kind of emulation layer into the OS, but I still don't see it happening anytime soon. People have been predicting for years that some kind of hybrid game-console, web browser thingie will take over the computing needs of the majority of the public. I just don't see that happening.
Like someone else pointed out, this guy seems to want to spin this into some kind of mystery novel.
Three points. 1) Rumors as fact and lots of speculation aren't going to lead to much. 2) hate to burst your bubble but PowerPC isn't going to be taking over the world. 3) You say that MS can force Intel, AMD, and IBM to do what it wants. Sorry but that is simply NOT possible.
try reading what he said after the comment.
he is claiming that PPC will take back the speed crown because of IBMs ne compiler, the fact that Intell is having a huge leakage problem, and a few other things...all of which make sense to me....infact, I think that intel is in for a huge fall if it can not get the power consumption problem solved soon.
Well, let me guess at Part 3... not only will the concepts behind the cell processor change the nature of computing "power" and where it is located, but the cell phone will change the nature of location based computing. This would both make a company such as, but not necessarily, Microsoft more like a Bell company, and allow the subscription like charges to be spread over the lifetime of devices (12-18 months generally). Not only is this an issue, but DRM should appropriately be placed at the purchasing point, and personal technology allows for both DRM and smart money-like capabilities. Fun!
Depending how far in the future you look, controlling the desktop top might mean controling mini-nets.
Wireless technology provides interesting capabilities. In theory you could have a PDA or cell phone as a remote interface to a server or two. An example I am thinking about is for home use but could be applied elsewhere. A typical home could have a wireless box controlling various media such as movies and music. The PDA or cellphone would have the GUI interface to select what movies to watch or what music to play. The wireless server wouldn't need a monitor. Wireless keyboards could be used to enter long strings of data. If TVs have high resolution, people could surf the net. You could perform a Google search on a PDA and the result could be piped to the nearest video display.
Some day, you might be able to gather data off your home server remotely through the use of either PDA or cellphone and it wouldn't matter where you were located. If you are in a grocery store and need your grocery list, use you cellphone to grab the list off your home server. Some kind of wireless VPN connection could be made. I am not good with technical details but hopefully you get the idea.
Full time connections to the internet, wireless technology, and the proliferation of cellphones and PDAs, computing may not be sitting in front of a desktop computer.
It might be hard for one company to control cellphones, PDAs, and servers/desktops.
Will this come to be? The technology is either here or almost here.
Itanium was obviously an awful business mistake. AMD has made the right decision by producing IP which they can leverage across both commodity desktop and server lines (Athlon64/Opteron). In this way the high volume commodity market helps offset the cost of a server processor.
Unlike the Athlon 64, Intel has almost no means of cutting the Itanium down into a processor for use in the commodity (x86) market, which was certainly never a design requirement for the Itanium but may be the only way for Intel to recoup their losses from a business perspective. Unfortunately, given the added R&D that would be required to repurpose the Itanium as a commodity chip, it's an investment Intel is unlikely to make.
Article: "...it looks like the Itanium and it's 13 year, multi-billion development cost could potentially turn out to be the biggest mistake in commercial history."
Not that it's of any consequence, but I can think of several business ventures which were considerably worse. The $8 billion Iridium satellite cluster which is in the process of burning up in Earth's atmosphere comes to mind. So does VeriSign's purchase of NSI for $21 billion, especially when coupled with the recent $100 million sale of NSI.
cheezwog: This does not make much sense to me... Why dedicate an entire powerPC cpu to doing sound when you could use a dedicated dsp like an Atmel for much cheaper?
Other dedicated hardware like GPUs are also much cheaper and faster at what they do than a general purpose CPU.
What other hardware do you need a PowerPC dedicated processor for? The joystick?
Okay, first, let's get the rumors straight. Despite claims that the XBox Next will have "four or more IBM processors", it's much more likely (and insider knowledge has suggested) that the XBox Next will instead contain four chips manufactured by IBM. As to whether this will be something like two PPC9*0 processors and two HyperTransport bridges for the processors is uncertain.
Given that Microsoft will be targeting HDTVs with resolutions of up to 1920x1080, it doesn't seem unreasonable that two processors would be advantageous. The Sega Saturn and 32X both shipped with dual processors (an omen of dismal failure for the XBox Next?) One processor could be dedicated to managing the virtual world being presented, and another could handle the actual presentation.
Although given that the XBox Next will supposedly be smaller than the current XBox, and that IBM will be providing custom fabrication for Microsoft, it seems odd that Microsoft simply wouldn't use a multicore processor, which in the case of a Power5 derived architecture would support up to 4 hardware threads.
I wouldn't take the rumors of an SMP XBox too seriously at this point.
x86 will die
x86 won't die. IA32 will die... eventually, once both Intel and AMD are pushing 64-bit chips backwards compatible with IA32. x86 will evolve.
Inted doesn't hold any sort of speed crown right now. Everything I've seen shows the Opteron/AthlonFX series blowing the crap out of Intel's stuff for every day usage.
No matter how much the guy might hope for software to become more and more hardware independent, there would have to be some kind of massive acceleration of Moore's law to make any sort of instruction set emulation patalable to any gamers. Why would a gamer buy a PPC based machine that emulates his current games way slower than an x86 currently plays them? I would think that someone would have to release a non x86 processor with 5x or more the speed of the current fastest x86 to be able to emulate it properly. Not to mention be cheaper.
And gamers are the ones who spend the big bucks which currently drive the vendor high end products.
About his PPC claims: what's more likely (but still unlikely) to happen is that the PPC will ultimately go the way of the DEC Alpha. It's a crying shame that architecture was killed off. It was a damn good processor.
PowerPC is a major technological cornerstone for IBM, its not just for Apple.
It is significant to IBM's Linux corporate desktop initiative. A processor platform and OS that IBM controls fully that is cost effective to deploy in low and midrange enterprise applications. I'm not saying that IBM is going to forego Intel or AMD solutions, just that it will favor deploying PowerPC anywhere it can. I can see PowerPC being around for a long time.
My computing days are over if I have to pay MS to run Linux. Period.
I believe Microsoft is in fact singular.
Interesting article.
Xbit labs has an early sample of hte Tejas where Dissapates 150W @ 2.8GHz.
http://www.xbitlabs.com/news/cpu/display/20040111115528.html
I don't take sides, but Intel has some serious work to do to Prescott and Tejas to make them worth while upgrade options. Who knows what all the extra transistors are for but the way things look now it could be a good year for AMD.
BTW, with the current trend of Intel and AMD moving into the server side of the market I expect that we will see more R&D from both companies in the server area. With the introduction of 64 bit desktop processors means an entirely new class of low end servers. It will be a sad day at Sun Microsystems when Dell starts selling Linux and windows servers on 64 bit Xeon processors. It will be difficult to get into a price war with AMD/Intel when the R&D for their server chips are masked by their massive desktop sales.
Oh how many times have articles like this been published... "this will be the future!" and it never is. Historically, things have ended up significantly less sensational than commentators prophesize. We're not all using VR, are we? We still have film cameras, right (and professionals still use them extensively)? We're still not renting software, are we (for the most part)? All of these things were predicted to have not only happened but been totally ubiquitous by now by writers like this in the past.
The future of hardware is much less fantastic than this writer imagines. Already technology has slowed down drastically from its boom period in the 90s and even up until this time last year. No new amazing video card technologies every three months, no new motherboard chipsets every month and a half, and processor speeds may still be climbing but their performance gains between clock speed increases are diminishing rapidly. Prescott has an enormous advantage over Northwood when dealing with large blocks of data in multitasking environments -- the author doesn't seem to have done his research in this matter at all, beyond reading the first paragraph of some TH or other typically uninformative processor review and paraphrasing it. The Prescott core is up to 35% faster than an equivalent-speed Northwood in business environments using real-world software.
PPC is not going to forge ahead in the desktop world because it's too expensive and has virtually no hardware support compared to IA32. Apple and IBM have their niche markets, and they will not expand significantly in the near future even if they offered twice the performance of IA32 (which is almost dead anyway, to be replaced by AMD64 or some other extended architecture with full backwards compatibility) because the simple fact of the matter is that practically everyone is broke and no one really needs a new computer to accomodate new software. Software drives hardware upgrades, not vice-versa. Proprietary hardware is dying, not advancing. It is far more likely that the PPC will die in the future because of its high price and poor hardware and software support when compared to cheaper desktop IA32 and AMD64 hardware. The only thing keeping SPARC, MIPS, IA64, and PPC hardware alive right now is proprietary software on the high end (and the fanaticism of Apple people). If the same engineering, visualization and virtualization software was available for GNU/Linux and Windows, that would be the end of proprietary designs by the end of the decade. Who would pay $10,000 for an IntelliStation POWER when you could buy several dual Opterons for the same price -- and the AMD machines would be faster and more easily and cheaply fixed.
I don't know why people think that closed designs are the wave of the future. The only way that could happen is if they were so cheap that you'd think little of throwing them away when they break.
Besides, for many years the 68k and PPC Apple machines were faster and easier to use than x86 PCs were, yet they have consistently failed to dominate any markets. If it didn't happen before when conditions were much more favorable, there is no reason to believe that this industry shift to PPC will take place in the future.
-Jem
If Microsoft actually tried this would it work? This is a difficult question but Microsoft are probably the one company who could do it. ---> how? when? all they have is a crap OS..
Regardless of all the personal prognistication presented here, I implore of the author to PLEASE write correctly. It's much harder to read an article with so many grammar mistakes as found here.
I'm also of the opinion that in many ways, the author is smoking pot in terms of just how much power Microsoft will be able to employ. Why? All the open source zealots will get out the word to the common people, regardless of whether they want to hear it or not. Also, Microsoft is only one potential vendor of hardware, and when people learn of just how restrictive the terms proposed by this article are, there will be a huge backlash by the consumers. How will Microsoft truly pressure people into buying their hardware? I know of people still using ancient versions of Windows, even DOS, and being sufficiently satisfied with what they've currently got that they wouldn't upgrade their system unless and until it dies.
No, the only way Microsoft would be able to command such things from consumers with any significant penetration would require Microsoft to figure a way for countries to pass laws that require something that only Microsoft provides. Now, how likely is THAT concept?
Prescott has an enormous advantage over Northwood when dealing with large blocks of data in multitasking environments -- the author doesn't seem to have done his research in this matter at all, beyond reading the first paragraph of some TH or other typically uninformative processor review and paraphrasing it. The Prescott core is up to 35% faster than an equivalent-speed Northwood in business environments using real-world software.
I beg to differ. The current Prescott chips do not have “enormous” advantage over the Northwood ones. If anything, the new Prescott chips are inefficient compared to their Northwood cousins. No Prescott chip in the market performs better than a comparable Northwood chip at the same clock speed.
In fact, Intel's fastest processor is their new 3.4GHz Pentium 4 Extreme Edition processor which is built upon a Northwood architecture. These chips marginally outperform their Prescott cousins at the same clock rate.
However, Prescott does have subtle advantages over Northwood, which will become more conspicuous at much higher clock speeds, analyst suggest 3.6GHz/4.0GHz and above. Prescott has a better branch predictor, a revamped scheduler, improvements on integer multiplication(I think they added a new ALU), a much larger cache, introduction of SSE3(13.new instructions), to mention a few.
Should you run out to buy a Prescott processor tomorrow? Absolutely not! The first few Prescott processors from Intel will suffer severe penalties as a result of Intel increase the Pentium 4 pipeline from 20 stages to over/approximately 31 stages, beginning with the Prescotts.
In fact, I won't be purchasing a Prescott processor until they hit the 5GHz barrier. Then, I'd wager that they have an enormous advantage over their Northwood predecessors. Today, however, they don't. And they won't for a while(sometime in 2005/6).
With regards to the article, I think it is poorly researched and based on unfounded speculation, hearsay and bias.
http://www.anandtech.com/cpu/showdoc.html?i=1956
More worthless semantic bickering...
By Bascule (IP: ---.atmos.colostate.edu) - Posted on 2004-02-10 00:41:21
x86 will die
x86 won't die. IA32 will die... eventually, once both Intel and AMD are pushing 64-bit chips backwards compatible with IA32. x86 will evolve.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
IA32 is x86 emulation... so how can that be "Backwards compatible... x86 itself is the compatability goal (for AMD Intel HAS to follow in it's footsteps!) (GO AMD)
n fact, I won't be purchasing a Prescott processor until they hit the 5GHz barrier. Then, I'd wager that they have an enormous advantage over their Northwood predecessors. Today, however, they don't. And they won't for a while(sometime in 2005/6).
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Somehow I think Presscott (yes I HISSED the snake-ish name out)is a mistake... too many problems.. and ppl are the BETA testers, most of the presssscott instructions (and more) will be implemented in TEJES which will certainly clock over 5GHz (but be no better than AMD'S offerings)
Microsoft... is singular,
Intel... is singular,
AMD... is singular.
Individual corporations/companies are *not* groups.
Not if you are from the UK. In British english, collective nouns are plural. Ie: "The government are passing new laws."
Juding from Nicholas Blachford's name, I had a hunch that he's European, which is confirmed by his webpage (blachford.info).
IA32 is x86 emulation... so how can that be "Backwards compatible... x86 itself is the compatability goal (for AMD Intel HAS to follow in it's footsteps!) (GO AMD)
IA32 refers to the 386 instruction set, the last time the x86 architecture saw a significant change.
A32 refers to the 386 instruction set, the last time the x86 architecture saw a significant change.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Don't fully know what I was Thinking.... my apologies
x86 is much more dependant of the desktop computer ( and low end workstations ) market than the PowerPC is. Actually, PowerPC would survive even without Apple as they are much used in embedded ( industrial & others ) platforms, alongside with ARMs and others. That market uses from 100MHz to 600MHz PPC's.
Without PCs, current x86s wouldn't survive enconomically as it cannot stand comparison with others families ( for example, TODAY, the fastest industrial temperature grade x86 is a 166MHz Pentium MMX ). Nevertheless, 386 and 486 are still used TODAY for these applications.
For example of multi-PPC components, watch Xilinx Virtex II Pro ( www.xilinx.com ) programmable gate arrays. The biggest part have 4 embedded PPC cores. They are IBM manufactured low end PPC cores ( 400MHz, no FPU ) embedded in the programmable logic fabric. With a few thousand $ or euros, you could build your own multi ppc cores chip.
If the better architecture would win, we would have seen 68000 based PC replaced by Alphas, They would have run CP/M initially and then somemthing similar to AmigaOS ... PowerPCs have been superiors platforms for more than 10 years by now ...
And I thought that the king of this kind of me-no-understand-markets/cheeky/say-whatever/throw-caution-in-the-wind speculation was mr. John C. Dvorak.
Long live the new king.
Wrong, wrong, wrong. The P4 Extreme Edition is not based on Northwood, it's based on Gallatin, which is the Xeon core.
If anyone finds that Northwood is the same performance level as Prescott then that only attests to their poor benchmarking skills. In a *real world* multitasking environment, the Prescott is faster than the Northwood. When using software from Adobe, Microsoft and McAffee at the same time (Photoshop, Word, VirusScan) the Prescott is 35% faster than Northwood. If Anand or anyone else didn't find that or know how to elicit the better performance of the Prescott then he didn't read his press kit carefully enough or design his benchmarking procedure with accuracy in mind.
-Jem
If Anand or anyone else didn't find that or know how to elicit the better performance of the Prescott then he didn't read his press kit carefully enough or design his benchmarking procedure with accuracy in mind.
Oh I see, it's faster if I'm careful about selecting tests that it's fast in? I didn't realise!
You on Intel payroll? :p
While P4EE is a Xeon part, the Xeon is based on Northwood, so you can hardly repeat wrong three times. I'll bet you enjoyed it so though.
When you say DRM, I think "Digital Radio Mondial", so, don't use the term DRM unless you mean Digital Radio. Digital Rights management is either nose dive, or nosebleed. I leave that to you to decide. The future is bleak, actually. With isa addressing gone- you have no base structure you can run to when you need to. everyone hates "windmodems", not because they don't use isa adressing, but because driver sources are protected from anyone but microsoft. I don't think pci on macs means pci anywhere, and to me, that's more conflicted that you can ever imagine. Granted, the Power processor finding it's way onto desktops will be a boon, it will just mean apple will lose some of it's protective padding that microsoft uses it for.
It won't be until quantum optical channel processing comes around, until you begin to see revoloutionary things occur.
Industry chip makers are die cut competeting, mm thickness, and not really rethinking the technology. The pentium HT offers some hope, as really, they can add as many extra instruction set banks as they want, for paralell virtualization. but, dont' get me wrong- I love the power processor. It will change the playing field, but it'll change the die hard apple lovers forever, because apple won't be allowed to be the only ones to have it. That's where it's going. And where it should go.
Microsoft... is singular,
Intel... is singular,
AMD... is singular.
Individual corporations/companies are *not* groups.
They are if you live just about anywhere except North America.
It has to do with culture. Whereas US culture perceives a corporation as simply a legal entity to hide its members behind, other countries see a corporation as a reference to a group of people working together (note that I'm talking from a cultural, not legal, perspective).
If you go to England, Australia, New Zealand or, as I said, just about any other part of the English speaking world, corporations are referred to as a group of individuals, not an individual entity.
Oh I see, it's faster if I'm careful about selecting tests that it's fast in? I didn't realise!
It's pretty trivial to design a benchmark that can give similar - if not identical - results on platforms with vastly different levels of overall performance.
The point is that, yes, you *do* need to be careful about selecting tests if you want to throughly test performance. A single threaded high CPU load test isn't going to show much of a difference on a multi- vs single-CPU machine, even though other benchmarks could show nearly a 100% performance difference.
If you didn't realise this, you should probably go and research a bit more about the theory and practice of benchmarking, because you certainly aren't able to objectively assess benchmark results.
One thing is for sure, Intel is loosing the crown more and more.
Within a few months intel processors will be completely out of the price/performance charts. The Athlon64/Opteron are a winning team.
But Intel has a weapon under the coat... the Pentium M, if Intel expands the "M" with a licence of the AMD 64bit extensions, and spend time and money on it it can result in a nice desktop/server processor.
From dell, one can get the lowest end system with windows for $200. For companies buying in bulk, they can probably get such prices for less.
I saw this and just laughed. If you think businesses are buying such boxes (and for less than $200) you're nuts.
That said, the Microsoft hardware angle is about control and control is about money.
While mastering the skill of patronisation you seem to have no grasp of sarcasm. Congratulations.
