An article on Japanese site PCWatch writes about a briefing Microsoft gave journalists earlier this week about 64-bit products it is proposing. One 64-bit OS is conspicuous by its absence, it would appear. Microsoft has outlined, according to the article, four different types of .NET server – Web, Standard, Enterprise and Datacenter – and the article gives an interesting table describing the differences between these products. In particular, the table outlines IA-64 support, memory requirements, Numa support and also whether the products will support multiple CPUs. But Opteron support is still a mystery.
funny, AMD on servers! what’s next? Servers runing windows?
Microsoft’s .NET framework uses something called the CLR to execute hardware-independent code. I suspect x86-64 support will be enabled via the CLR rather than native compilers.
First off, the OS isn’t running on the CLR. Having a .NET runtime for x86-64 isn’t going to make .NET Server run Opteron-ized.
Second, Win32 is a hardware independent API. They don’t need .NET to support a different processor architecture; they have had that ability for many years now. Remember NT for PowerPC? NT for Alpha?
I thought Opteron can run 32-bit applications without problems. Wouldn’t that be true for 32-bit OSs as well? Maybe that’s why it wasn’t mentioned because at first you could still run the 32-bit version of Windows .NET Server. Am I wrong?
>>NT for Alpha?<<
this is kind of an oxymoron… remember VMS for Alpha?
VMS -> OS2 -> NT -> 2000 -> XP (with a little 95 mix to make it sweet)!
When i was 15 doing work expereince (about 10 years ago) .. walk up to a box bring up dos window.. type dir.. hmm long file names cool ol NT. DOs command window looks a little strange… “Ooohh its a Dec alpha” COOL!
🙂
Glenn
The word is that Opteron will run 32bit code faster too. So, Microsoft doesn’t _need_ to have an x86-64 version to “support” the chip. Of course, to get the full advantages, the OS really should understand the new 64bit extensions; 32bit code with 64bit OS on Opteron runs 32bit code faster than 32bit code in 32bit mode.
But, I run Linux and (Open)BeOS, so I couldn’t really care less except that I very much would to be assured of Opteron’s success if only because Itanic is so slow and expensive.
Programs from Microsoft and other vendors that wanted to use the 64-bit features of Opteron would need to be recompiled specifically for x86-64 if Microsoft were to release such an OS. However, since Microsoft already has a grand scheme to move everything to .NET anyway, now is not the time to introduce a short lived category of 64-bit programs that don’t use .NET, except on ia64 where Microsoft has preexisting commitments.
My impression is that a 64-bit .NET program could run on both Itanium, x86-64, and perhaps other platforms if Microsoft releases the CLR for them. There is only one chance to go through this hardware and software transition, so Microsoft probably wants to make it count.
what’s next? Servers runing windows?
Next is a Sun Blade running Apache software (or Sun Linux); slowly …
>>Next is a Sun Blade running Apache software (or Sun Linux); slowly …<<
Sun, slow?
sounds like another oxymoron! At least with Apache, security is taken into account!!
TheIsraeli: funny, AMD on servers! what’s next? Servers runing windows?
Uhmm, most of the web servers out there uses some version of Windows, Linux only comes #2.
CDN: I thought Opteron can run 32-bit applications without problems. Wouldn’t that be true for 32-bit OSs as well? Maybe that’s why it wasn’t mentioned because at first you could still run the 32-bit version of Windows .NET Server. Am I wrong?
Yes you can, but to take advantage of Opteron, the app must be ported to x86-64. A lot of features in Opteron are 64-bit only.
Anonymous: But, I run Linux and (Open)BeOS, so I couldn’t really care less except that I very much would to be assured of Opteron’s success if only because Itanic is so slow and expensive.
The Itanic you speak of is faster than POWER4. (Itanium 2). The problem is there isn’t much native applications, so most of the things are ran under software emulation.
Besides, there isn’t any prove either than Opteron would beat Itanium in terms of speed.
CattBeMac: Sun, slow?
Notice he was sacarstic?
Besides, Solaris 8 for Intel was slow, at least compared with Linux.
—
However, maybe Microsoft didn’t want to give AMD support because of the bad job they did in defending Microsoft in the antitrust case 🙂
>>Besides, Solaris 8 for Intel was slow, at least compared with Linux.<<
I hate to say it, but I can imagine that Solaris 8 for Intel is an afterthought compared to the SPARC version. I have only used Solaris sparingly on Intel in the past. We use Solaris only on Suns here.
>>The Itanic you speak of is faster than POWER4. (Itanium 2).<<
Actually it’s a toss up at the moment, both are faster than the other on certain benchmarks!
What makes you say that?
http://www.netcraft.com/survey/
>> Uhmm, most of the web servers out there uses some version of Windows, Linux only comes #2.
>> Uhmm, most of the web servers out there uses some version of Windows, Linux only comes #2.<<
I must have missed that… actually UNIX/Linux has about 70% of the webserver market, while Windows has 30%. There are conflicting numbers though, but to some it up, Microsoft likes to count anything that host a webpage and I mean mom and pop PCs included! That is about as sad as Apple bragging about being the #1 Unix vendor!!
Dear Microsoft Friends,
Opteron support is coming along well. It’s simply a big project.
Don’t quote me on this, but we love Opteron. It’s so much cleaner and elegant than that pile of VLIW shit that Intel gave us. Like we want to actually have SSL run fast? Duh, hardware dudes, slow SSL = sell more computers = sell more OS.
Ever since Andy stepped down, those boys at Intel have been kinda slow. When our tech guys got out of hand (Cairo), we shitcanned the lot of them. Without Andy, our pals forgot it’s all about the money. Ooops. That’s Steve’s job now. I better send him a memo.
Expect to see some cool x86-64 demos later this year.
Cordially,
Bill
apple does ship the most unix systems now.
What makes you say that?
http://www.netcraft.com/survey/
This is a survey of HTTP servers. Apache could run on Windows NT, and it runs just fine (in fact under pressure, it runs better tha IIS on Windows NT). Face the facts.
CattBeMac: I must have missed that… actually UNIX/Linux has about 70% of the webserver market, while Windows has 30%. There are conflicting numbers though, but to some it up, Microsoft likes to count anything that host a webpage and I mean mom and pop PCs included!
You actually combined all the UNIX flavours together, I didn’t. (Because many of this flavours are as different from rainforest to the antartica).
CattBeMac: That is about as sad as Apple bragging about being the #1 Unix vendor!!
Sun is the #1 UNIX vendor.
Bill Gates: Don’t quote me on this, but we love Opteron. It’s so much cleaner and elegant than that pile of VLIW shit that Intel gave us
VLIW is much cleaner than x86-64. VLIW is just new, and therefore lacks any good compilers. It is the same thing that plague RISC when it first came out.
Besides, Eugenia, why isn’t that post (fro “Bill Gates”) moderated down?
Dear Rajan,
Apparently you haven’t talked to any of the folks who have done compiler work for VLIW. Not a walk in the park. VLIW goes beyond even the mental complexity of programming a DSP chip.
Even that old warhorse, Cutler, chimed in to support x86-64.
As you know with Alpha, the emphasis was keeping the architecture simple enough to ensure massive throughput. Itanium is a very complex ISA.
We’re not saying Intel did a bad job with Itanium. Just that it is needlessly complex. They needed the Alpha team sooner than they got them.
If Itanium had offered the market leading x86 support that AMD does, the whole world would be behind their push to 64 bits. Instead, the support is tepid at best.
Intel’s strategy with their next generation CPU has relegated the chip to high-end servers and workstations. This strategy, which Intel has never used before, is because Itanium needs to cook a while before being ready for the mainstream market. Look at the Intel roadmaps.
There is no affordable Itanium in the market until 2005. Which is when Longhorn will be shipping, with great support for Itanium.
We are the world’s best software company. We know Itanium.
Bill
While there does exist a few differences here and there in each version of UNIX and related Operating Systems. It is, in most cases, easier to go from one version of a UNIX to a different version of a UNIX, then it was for many people to go from Windows 3.x to Windows 9x, to Windows NT 4.0, to Windows 2000 and now to Windows XP.
With each release of Windows, Microsoft drastically changes the way nearly everything is done. This requires the user/administrator to learn an entire new set of skills in order to operate that platform.
In the UNIX world, it is exceptionally easy to translate the skills learned one on UNIX or UNIX-like platform to other UNIX, or like, platforms. Sure, there might be a few differences in how services are started, how run levels are determined. However, the vast majority of skills learned on one UNIX easily transfer over to another UNIX.
One more thing, when a new version of a UNIX is released, you already know how to use it and do everything that you did before. Most of the time, when an update to a UNIX or UNIX-like OS occurs, it may add a few pieces of technology, but most of the time it adds hardware support. Rarely do the vendors drastically alter the OS to require one to relearn how to do everything.
>>CattBeMac: That is about as sad as Apple bragging about being the #1 Unix vendor!!
Sun is the #1 UNIX vendor.<<
see my point 🙂