 | |
Is Itanium the first 64-bit casualty?
Hardware Chips Processor, cache, memory chips, etc. (comp.sys.ibm.pc.hardware.chips)
 |
| | Thread Tools | Search this Thread | Display Modes |
#331
 
| |||
| |||
|
Re: Is Itanium the first 64-bit casualty? -
06-12-2009
, 10:37 PM
|
Robert Myers wrote: On Jun 11, 3:12 am, Terje Mathisen <"terje.mathisen at tmsw.no" wrote: Robert Myers wrote: It's no less a problem for Via than it is for anyone else. Power consumption, not x86, ate Itanium, and it could eat x86, as well. Power consumption did _not_ eat Itanium, at that time (10+ years ago) almost nobody even cared about it, except to stay within a given maximum power budget. It's been my assumption all along, and, again, I hope someone knowledgeable will correct me, that, were there no power budget constraint, "fixing" both P4 and Itanium would have been much easier and perhaps even relatively straightforward. That's a fantasy universe, of course, because the power budget is a hard constraint for reasons having nothing to do with the cost of energy. You see your hypothesis that power constraints killed Itanic is already falsified. |
Robert.
------------------------------
Read a Bill Todd post and "lose attitude"?? That some sort of
oxymoron attempt? If there are posts on comp.arch with more attitude
than his I would be surprised. Although I might have someone kill
filed.
del
#332
| |||
| |||
|
|
The real questions are a) when will the market for low power x86 be big enough b) will x86 bring enough value to outweigh ARM's incumbency. I am not as sure as Mitch may be that x86 is guaranteed to wun in the low power market. I think that it is entirely possible that ARM may have established itself well enough as to not be displaced. And I also worry that ARM may then start migrating up the food chain. I have read "The Innovator's Dilemma". I know that attacks from below are more likely to succeed than coming diown from above, in both power and cost. But I don't think that there are any overwhelming technical obstacles to low power x86. Just business and organizational ones. |
I think ARM is as entrenched in the embedded market as x86 is in the PC
market. I don't think there will be enough meaningful takeover of market
segments from either side into the other's territory -- ever.
I those areas that have some overlap between the PC and cellphone market
will automatically segment themselves off towards one side or the other.
For example, a smart phone will never be quite as capable on the
Internet as a PC, when it comes to multimedia-type stuff, so if you want
multimedia stuff in a handheld device, it's going to be something that
is always slightly higher than a smart phone. It might have phone
capabilities of its own, but it might be too annoying to use as a phone
most of the time.
Yousuf Khan
#333
| |||
| |||
|
|
"Yousuf Khan" <bbbl67 (AT) yahoo (DOT) com> wrote ARM and Linux are fine, there should be little effort required to port to that. My doubts are that any x86 Windows program can be even half as easily be ported to any non-x86 Windows. Significant differences always exist in x86 vs. non-x86 Windows implementations. Like for example, it's unlikely that you'll see a non-x86 version of DirectX. What makes you say that? It's an API like OpenGL and others. |
Quote:
|
So why don't we see this more often? If it's as simple as selecting a different compiler switch, then it shouldn't cost too much money for an application developer to run the source through a few different switches. Compiling is easy, the expensive part is being able to run and test the software. Going back to my remark at the top, how many people do have access to anything but an x86 box? |
lot of them don't bother as they don't expect any sales in other platforms.
Quote:
|
In theory, yes. In practice, more open-source is more easily ported. That's your opinion. In my experience, proprietary software is often written with portability in mind. At a previous company we wrote software for many different architectures. We bought various software packages from others that had to work on those architectures (and did with minimal porting). None of this was open source. On the other hand, much of the open-source software I have used was extremely hard to port, due to using many GCC-isms, not adhering to the C standard, and using UNIX library functions such as open() when there are portable alternatives. Ie. unportable unless your definition of portable means "compiles with GCC"... |
You're looking at portability from a different perspective than I am. I
don't care how easy it is to port between one compiler and another, I
only care if it's available on platform or another. If that means using
GCC to compile all your stuff across platforms, then GCC it is. There's
plenty of good programs that run compiled with GCC, I don't care if it
can run 10% faster with some other compiler.
Yousuf Khan
#334
| |||
| |||
|
|
Read Bill Todd's post and lose your attitude. |
Which Bill Todd post would that be?
Yousuf Khan
#335
| |||
| |||
|
|
Robert Myers wrote: Read Bill Todd's post and lose your attitude. Which Bill Todd post would that be? |
Quote:
|
That's likely in the case of Itanic, anyway, since Montecito was reportedly designed to run at nearly twice the clock rate that it actually shipped with had the Foxton power-management technology actually worked (and at nearly twice the clock rate it would certainly have made many people stand up and take notice). |
test.
As to power not mattering to P4 and people using it anyway, I'll
assume that you are being deliberately obtuse. Power not being
important is one of the reasons all those 4GHz P4's hit the market as
Intel was on it's way to 5GHz. Good thing Intel never had to come to
its senses and abandon the architecture. That's why AMD practically
owns the x86 market now.
Robert.
#336
| |||
| |||
|
|
In fact, web based apps are one of the things that reduce legacy sensitivity, not just in the phone space, but all across PCs. I use Gmail. Now I no longer care about legacy email apps. I use Google docs. Now I no longer care (so much) about legacy Microsoft office. [...] Note that legacy data compatibility is becoming important for web apps. Who has not lost data from a website that has disappeared? How do I back up my LinkedIn and FaceBook data easily and automatically, against the day that these services die or charge too much? |
saying that web apps decouple the legacy app/data problem from
processor/OS/default display hardware and couple it instead to browsers
(which may of course have a strong CPU/OS coupling)? They also leave the
coupling to corporate survival and consistent business model in place --
if Google suddenly switched to a business model where revenue came from
customers using the current -- and only the current -- version of their
apps and APIs, you could quite easily start worrying about legacy
app/data problems for your email and documents.
The legacy data (aka data portability) issue for complex
social/collaborative web apps is even worse because it's not just a
matter of coding but of settling the underlying intellectual property
issues, or attribution issues if you prefer. Just what is "your"
Facebook" or LinkedIn data? Your friends/contacts? Their
friends/contacts too? Your outgoing feed? Your incoming feed? If you
recommend someone's work on LinkedIn, is that your data because you
wrote it, or their data because it's about them? If either one of you
leaves, it becomes virtually useless, but in different ways. Blah blah
blah...
paul
#337
| |||
| |||
|
|
Wilco Dijkstra wrote: "Yousuf Khan" <bbbl67 (AT) yahoo (DOT) com> wrote ARM and Linux are fine, there should be little effort required to port to that. My doubts are that any x86 Windows program can be even half as easily be ported to any non-x86 Windows. Significant differences always exist in x86 vs. non-x86 Windows implementations. Like for example, it's unlikely that you'll see a non-x86 version of DirectX. What makes you say that? It's an API like OpenGL and others. So where is it? Stating a bunch of theoretical truisms is great. |
non-x86 and they were proven wrong. I don't think DirectX is that important as
there are good alternatives, but I don't see a reason why it couldn't be ported.
Quote:
|
So why don't we see this more often? If it's as simple as selecting a different compiler switch, then it shouldn't cost too much money for an application developer to run the source through a few different switches. Compiling is easy, the expensive part is being able to run and test the software. Going back to my remark at the top, how many people do have access to anything but an x86 box? A large company should be able to get access to more than an x86 box. A lot of them don't bother as they don't expect any sales in other platforms. |
If we get $200 non-x86 laptops things are going to be very different.
Quote:
|
In theory, yes. In practice, more open-source is more easily ported. That's your opinion. In my experience, proprietary software is often written with portability in mind. At a previous company we wrote software for many different architectures. We bought various software packages from others that had to work on those architectures (and did with minimal porting). None of this was open source. On the other hand, much of the open-source software I have used was extremely hard to port, due to using many GCC-isms, not adhering to the C standard, and using UNIX library functions such as open() when there are portable alternatives. Ie. unportable unless your definition of portable means "compiles with GCC"... You're looking at portability from a different perspective than I am. I don't care how easy it is to port between one compiler and another, I only care if it's available on platform or another. If that means using GCC to compile all your stuff across platforms, then GCC it is. There's plenty of good programs that run compiled with GCC, I don't care if it can run 10% faster with some other compiler. |
we produced was also performance critical so we chose the best compiler on each
platform.
Wilco
#338
| |||
| |||
|
|
On Jun 13, 4:47 pm, Yousuf Khan <bbb... (AT) yahoo (DOT) com> wrote: Robert Myers wrote: Read Bill Todd's post and lose your attitude. Which Bill Todd post would that be? Bill Todd wrote: That's likely in the case of Itanic, anyway, since Montecito was reportedly designed to run at nearly twice the clock rate that it actually shipped with had the Foxton power-management technology actually worked (and at nearly twice the clock rate it would certainly have made many people stand up and take notice). Your claim that no one cared about power doesn't even pass the laugh test. |
that into a full-blown "that's definitely what happened" statement?
No, my statement didn't pass the laugh test, because quite obviously
it's your job to come up with belly-rippers. You grasp at straws like no
farmer I've ever known.
Quote:
|
As to power not mattering to P4 and people using it anyway, I'll assume that you are being deliberately obtuse. Power not being important is one of the reasons all those 4GHz P4's hit the market as Intel was on it's way to 5GHz. Good thing Intel never had to come to its senses and abandon the architecture. That's why AMD practically owns the x86 market now. |
Power consumption obviously mattered enough so that P4 couldn't ever go
towards the 4.0Ghz mark, but that didn't prevent people from using P4 at
the lower Ghz marks. Certainly there was no dearth of sales of P4's at
the lower Ghz like there was a dearth of sales of Itanic at *any* Ghz.
Yousuf Khan
#339
| |||
| |||
|
|
Robert Myers wrote: On Jun 13, 4:47 pm, Yousuf Khan <bbb... (AT) yahoo (DOT) com> wrote: Robert Myers wrote: Read Bill Todd's post and lose your attitude. Which Bill Todd post would that be? Bill Todd wrote: That's likely in the case of Itanic, anyway, since Montecito was reportedly designed to run at nearly twice the clock rate that it actually shipped with had the Foxton power-management technology actually worked (and at nearly twice the clock rate it would certainly have made many people stand up and take notice). Your claim that no one cared about power doesn't even pass the laugh test. Uh-huh, so you take a mild "it's a possibility" sentiment and convert that into a full-blown "that's definitely what happened" statement? No, my statement didn't pass the laugh test, because quite obviously it's your job to come up with belly-rippers. You grasp at straws like no farmer I've ever known. I don't want to discuss Intel products with you because you apparently |
consumed by Itanium was a *big* problem, Terje's advice
notwithstanding. He does follow Intel products, and I'm not sure why
his memory of history is so skewed.
As to Bill Todd's comment and "it's a possibility," everything that
we say as humans is an opinion. It is my opinion that "no one cared
about power" is so untrue as to be funny--belly-laugh material. As is
often the case, your opinion differs from mine. If previous posting
history is any guide, this could go on forever. I've finally learned
that the only way to end it is to let you have the last word, which I
will.
In any case, Bill didn't say "that's a possibility," he said "it's
likely." Del took umbrage at Bill's posting style. I can cope with
people whose self-assurance is justified.
Quote:
|
As to power not mattering to P4 and people using it anyway, I'll assume that you are being deliberately obtuse. *Power not being important is one of the reasons all those 4GHz P4's hit the market as Intel was on it's way to 5GHz. *Good thing Intel never had to come to its senses and abandon the architecture. *That's why AMD practically owns the x86 market now. Power consumption obviously mattered enough so that P4 couldn't ever go towards the 4.0Ghz mark, but that didn't prevent people from using P4 at the lower Ghz marks. Certainly there was no dearth of sales of P4's at the lower Ghz like there was a dearth of sales of Itanic at *any* Ghz. |
always do. Your arguments are slam dunks to you because whatever you
happen to think of defines all possibilities and interpretations.
Power consumption killed P4, even if you don't (or didn't) follow the
logic.
Intel had something in the back room *only* because power consumption
was so important for notebooks. That's why it was developing Banias.
Just because *you've* woken up to power consumption ten years into the
new millennium doesn't mean that everyone did.
Robert.
#340
| |||
| |||
|
|
Robert Myers wrote: It's no less a problem for Via than it is for anyone else. *Power consumption, not x86, ate Itanium, and it could eat x86, as well. Power consumption did _not_ eat Itanium, at that time (10+ years ago) almost nobody even cared about it, except to stay within a given maximum power budget. Complexity, hardware and software, is what caused the critical delays. As many have noted here (c.arch) previously, if Itanium had been delivered when originally promised, with the same performance as the actual first generation chips had, it would have blown everything else away. It certainly was jammed up against what was the maximum tolerable |
consumption not been a hard constraint, it would have been clocked
faster and/or had more active transistors to get better performance.
Owners of corporate data centers may not have cared about the cost of
electricity, but it's been a serious consideration for Beowulf
clusters since well before Opteron.
Robert.
|
« Previous Thread
|
Next Thread »
| Thread Tools | Search this Thread |
| Display Modes | |
Linear Mode
| |
Powered by vBulletin Version 3.5.3
Copyright ©2000 - 2010, Jelsoft Enterprises Ltd.
Copyright ©2000 - 2010, Jelsoft Enterprises Ltd.