x86-64 is not needed.

We all should buy new software for RISC CPUs.
Posted on 2004-05-17 16:19:31 by bitRAKE
No OS for x86-64 I meant. There's Windows for Itanium ofcourse.
But linux for x86-64 is about as good as linux (or any other *nix flavour) for any other 64 bit architecture. You have to build your 64 bit applications from sourcecode. Whether you do that on x86-64 or on an Itanium, doesn't matter (apart from the fact that the Itanium is much faster).

x86-64 is not needed.


Indeed.

We all should buy new software for RISC CPUs.


Why would we want to do that? We should run our current software on the new system, and if we have to upgrade our software again (which we will anyway), we shall upgrade to new native versions (or .NET ofcourse).

If in a few years most software is .NET anyway, it will be very easy to switch from one architecture to the next. You can finally pick the right CPU for the right job. A freedom that is priceless, and often overlooked. Intel probably wanted to wait for that, and if we're lucky, x86-64 won't change anything here. .NET may actually be an advantage for x86-64. Now you only need to develop one version, not separate 32 and 64 bit versions. And certainly not separate Itanium and x86-64 versions.
Posted on 2004-05-17 16:27:38 by Scali
x86-64 might not be needed, but neither is Itanium. :)
Posted on 2004-05-17 16:57:44 by bitRAKE
Itanium is needed, obviously.
It's aimed at big corporate servers, like other big 64 bit CPUs, such as the aforementioned POWER4.
Its performance is in a different class from x86-64.
...
Posted on 2004-05-17 17:06:49 by Scali

I generally need a single CPU with more grunt, rather than multiple CPUs. Multi-CPU systems only work when your program is parallelizable.


Except that when you're running multiple processes, they tend to keep their CPU to themselves for quite some time and you don't get the usual "multitasking slowdown" you get with a single CPU. E.g., I can compile in the background and continue to edit without any noticeable delay. Generally, after four CPUs you don't notice anything, but there is a *big* difference between single and dual CPU systems, and a noticeable, though not huge, difference between dual and quad systems. Of course, it all depends on what you're normally running. It is nice, however, to have the OS, device drivers, and all that junk running on one CPU so your processes get the other CPU to themselves.
Cheers,
Randy Hyde
Posted on 2004-05-17 17:24:09 by rhyde

Itanium is needed, obviously.
It's aimed at big corporate servers, like other big 64 bit CPUs, such as the aforementioned POWER4.
Its performance is in a different class from x86-64.
But you probably have never even seen a non-x86 CPU up close, so I don't expect you to understand.
...
FYI: this is a direct personal attack.
Posted on 2004-05-17 17:31:54 by bitRAKE
E.g., I can compile in the background and continue to edit without any noticeable delay. Generally, after four CPUs you don't notice anything, but there is a *big* difference between single and dual CPU systems, and a noticeable, though not huge, difference between dual and quad systems.


There WAS a big difference. But perhaps you have forgotten about Hyperthreading already? HTT CPUs are nice and responsive aswell, even under heavy load. Also, your example of editing while compiling is nice, but the editor is hardly making use of that second CPU. Therefore it is not very cost-effective. Another reason to prefer HTT ;)
Posted on 2004-05-18 02:39:34 by Scali
HAY man,

I have this blazingly fast 8086 that runs at two million cycles a second and it can address up to 64 thousand bytes of memory, who will ever need more computing power than this ?

Of course 64 bit will come, AMD have one running with the Opteron, the flopped Itanium has seen its share of mileage and some of the Intel cores at the moment have disabled 64 bit code. Apple twin G5 hardware is supposed to be nearly amazingly fast and it can handle a lot more than 4 gig of memory and its with us here today.

Intel appear to be retiring the PIV core due to heat problems and are working on the PIV-M which is apparently based on the PIII core so I would not look at a PIV as the bee all and end all, even though they do perform OK with Hyperthreading. Shows up as 2 processors on win2k.

Take 2.

HAY man,

I have a blazingly fast 80286 that screams along at 16 million cycles a second and has 4 million bytes of available memory, can task switch in protected mode.

Muhahahaha.
Posted on 2004-05-18 11:47:51 by hutch--
...and hutch valiantly tries to save the thread?
Posted on 2004-05-18 12:28:56 by f0dder
Of course 64 bit will come, AMD have one running with the Opteron, the flopped Itanium has seen its share of mileage and some of the Intel cores at the moment have disabled 64 bit code. Apple twin G5 hardware is supposed to be nearly amazingly fast and it can handle a lot more than 4 gig of memory and its with us here today.


64 bit has been with us for ages, long before Opterons, Itaniums or G5s. Just never in the mainstream, because there was no need for it. There still isn't a mainstream need, since there is no mainstream software in 64 bit. And there is no direct reason for going 64 bit with most mainstream software. What good would a 64 bit version of eg IE, Outlook or Office do? (This situation is very different from the move from 8->16 bit, or 16->32 bit, since those platforms had serious limitations (both mathematically and with addressing) that were alleviated by adding more bits. 32 bit isn't actually a limiting factor for most applications, since they neither need more than 32 bits for calculations nor for addressing).

Intel appear to be retiring the PIV core due to heat problems and are working on the PIV-M which is apparently based on the PIII core so I would not look at a PIV as the bee all and end all, even though they do perform OK with Hyperthreading. Shows up as 2 processors on win2k.


Wrong. First of all, the P4-M is a mobile version of the P4. The Pentium-M is a different CPU, and combines elements of P3 and P4. Secondly, the P4 is still performing a lot better than the Pentium-M, and there is no guarantee that the Pentium-M core will ever scale to competitive clockspeeds, after all, it was not designed to do so. Thirdly, Intel never announced that they would replace the P4 with the P-M. They have not said a thing about the replacements at all, other than that they would be dual-core solutions.
Finally, Windows XP is the only Windows optimized for Hyperthreading, afaik. Needless to say that Hyperthreading requires a slightly different approach to thread scheduling than a regular dual-CPU system. So Win2k would probably not make the most of HTT, and may actually do worse in some cases.

For now, P4 with HTT seems to deliver the most performance in many cases, such as photorealistic rendering and multimedia encoding/decoding, and it does this without the help of 64 bit. Why would I want 64 bit then? I am not interested in less performance. Hence, I have no need to go 64 bit at this time.
Posted on 2004-05-18 12:30:18 by Scali
My own personal experience has always been very good with AMD CPU's. They are always been low cost compared to much more expensive CPU's from Intel. From an architecture standpoint I have always been able to squeeze more out of them then the Intel CPU's. Not that I have had problems with Intel's CPU's (they are still running good), but they were always more expensive. YMMV.
Posted on 2004-05-18 12:58:12 by bitRAKE
...
I generally had little trouble beating AMD chips with Intel ones, because when AMD had 3dnow!, Intel offered SSE, and when AMD had SSE, Intel had SSE2.
Using the new features of the Intel chips, it was easy to beat the AMDs with their more dated technology. Even my Celeron 1.6 laptop beats my XP1800+ quite easily when SSE2 is involved, while that is a low-budget Intel chip competing with a high-end AMD one (well, it was once high end, and I paid almost as much for it as I would for a similar Intel. The motherboard was even more expensive than Intel ones, as was the cooling required).
Posted on 2004-05-18 13:06:27 by Scali
...
Posted on 2004-05-18 13:11:19 by bitRAKE
...
Posted on 2004-05-18 13:13:30 by Scali
Sorry, I don't recall. You are a good coder.
Posted on 2004-05-18 13:20:41 by bitRAKE
I think it was the countzero-problem, or was it the mul-problem? F0dder might remember. We wrote many kinds of routines anyway, ANSI C, vanilla x86, MMX, SSE2... and optimized for different architectures. I provided the fastest Athlon version anyway.
Posted on 2004-05-18 13:23:44 by Scali
Sorry, I've been really into some other work. Didn't mean to be slacking over here.
Posted on 2004-05-18 13:26:21 by bitRAKE
It was quite some time ago, and you provided a number of routines aswell. ...
Posted on 2004-05-18 13:33:31 by Scali
Is an excuse needed? Like I said, I can't recall even enough to provide an excuse if one was needed.
Posted on 2004-05-18 13:36:34 by bitRAKE
Anyway, let's conclude that I do about equally well with Intel and AMD when it comes to optimizing. But as I said, as long as Intel keeps offering extra features over AMD, it's quite easy to use them and get more performance.
Posted on 2004-05-18 13:53:26 by Scali