Dirk Meyer helped rescue Advanced Micro Devices from a tight spot once before. Now the company is hoping he can do the trick again. In his interview to ZDNet, he talks about the Opteron version of Windows, the Linux support for it, Athlon64 and more.
AMD Bets the House on Opteron
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Though I’m not in the server business, I can’t wait for these x86-64 chips to come out. AMD has never disappointed. Even when then were basically the cheaper also-rans (along with Cyrix), and they really couldn’t keep up completely with the comprable Pentium-based system, they were still a damn good bang for the buck (in fact, I’m writing this on my trusty old K6-2/500 based system and loving it). Now, AMD is clearly not an also-ran; if they promise any particular performance, they deliver!
It’s going to be interesting to see how big a punch AMD will throw at Intel. I think that until more 64 bit s/ware starts coming our way they have to have some other feature(s) (which one(s) I’d like to know) to beat processors cranked out of Intel’s shops. Nobody is mentioning the price scheme and how they plan to compete with Intel in this front, though this was more of a technological interview.
http://www.outwar.com/page.php?x=695874
Note, Opteron is aimed at the server. The server world isn’t a closed hell like the desktop market. There is lot’s of competition. If Opteron has good performance, then commercial UNIX vendors will simply recompile their apps and ship them. Since a lot of server apps (Apache, Oracle, etc) already run on 64-bit hardware, this transition will be a whole lot easier for them than it is going to be for the Microsoft crowd.
Although the Opteron is AMD’s server chip and athlon 64 the desktop chip that doesn’t mean there will not be desktops with Opterons in them. Mobo makers and barebones companies are lining up hardware for this. If you want your opteron desktop running windows you very well will have to wait for the fall when MS releases windows for athlon 64 and AMD releases the athlon 64’s . Supposedly Shuttle will have Opteron XPC’s out. They have shown them as shows. I would hardly call a SFF (small form factor) computer like a shuttle (picture a 8 inch aluminum cube) a server (though the do have models semi-targeted at that).
Unlike previous server chips I think Opterons will be in a good many peoples desktops. People have dual Athlons, and Xeon desktops but not to common. This i think will change with the Opteron. Especially if they are in the sub 500 bucks a chip range a few months after launch like it sounds like they might. If the Opteron is expensive come fall the athlon 64 will be well priced for sure (since thats what it is for)
I wish AMD the best of luck with 64bit, I think they will do very well. A good ad campaign and they are good to go. By fall if the economy picks up any people who haven’t bought computers for a few years will be out buying and AMD showing 64bit verses Intels goods will definitly cause a good time for AMD. AMD will find themselves in a great position no matter what and make big server inroads. Possibly carry Freebsd and linux with them some.
On a completely differant note I wonder what the OS picture in the coming year will be like, with windows on differant CPU’s and maybe various versions of software out there to take advantage of differant cpu’s. Apple *MIGHT* be split among two differant CPU’s . Amiga and Beos (zeta) might make a flicker someplace in the market. Maybe things will be slightly diverse for a little bit. Or life moves on with Windows the same, Apple the same, and everything else well, still basicly no place.
On a completely differant note I wonder what the OS picture in the coming year will be like, with windows on differant CPU’s and maybe various versions of software out there to take advantage of differant cpu’s. Apple *MIGHT* be split among two differant CPU’s .
Apple is absolute, positively *NOT* going x86.
Why? PPC970. Changing ISAs is a MAJOR hassle, one Apple can’t possibly take on amidst a shift between OS 9 and OS X. PPC970 will provide them the power they need without having to change ISAs.
I think one thing that people tend to forget is the Opteron is backwards compatible. All your existing 32bit software will run on it. So unlike the Itanium, you won’t need to wait for specific software but current software will run. This is a big bonus. From early benchmarks on sites like http://www.aceshardware.com , even when running 32 bit software, the Opteron is significantly faster than an equivalently clocked Athlon Palomino.
As for 64 bit software support, http://www.xbitlabs.com/news/cpu/display/20030407193326.html is a comprehensive list of vendors who have already pledged x86-64 support. Sweet list
I too like everyone else can’t wait for the Hammer.
I said that Aceshardware had the Opteron benchmarks. My bad. It is actually XBitlabs. The article is http://www.xbitlabs.com/news/cpu/display/news6338.html
“Apple is absolute, positively *NOT* going x86. ”
It would make sense for the server market! Why not, only their server apps would require porting.
Dan
Even if Apple doesn’t go x86 (and the latest stuff on the rumors sites indicates that the PPC 970 is ahead of schedule enough that it won’t) it’ll still be split between CPUs. You’ve the POWER-derived PPC 970, and the PowerPC-derived G3 processors. It’ll be similar to todays G3/G4 split, but the PPC 970 is a fundementally different CPU, and a 64-bit one at that.
It would make sense for the server market! Why not, only their server apps would require porting.
Why do they need to go x86 on their servers? Who wouldn’t prefer a dual PPC970-based Xserve? The processors will have lower power dissipation (not to mention that they’re clocked lower)
There’s no need for Apple to change horses in midstream.
Even if Apple doesn’t go x86 (and the latest stuff on the rumors sites indicates that the PPC 970 is ahead of schedule enough that it won’t) it’ll still be split between CPUs. You’ve the POWER-derived PPC 970, and the PowerPC-derived G3 processors. It’ll be similar to todays G3/G4 split, but the PPC 970 is a fundementally different CPU, and a 64-bit one at that.
As long as the ISA is the same, there aren’t any problems, even between 32/64-bit CPUs.
Sun successfully made the transition between sparcv8 and sparcv9 simply by packaging both 32-bit and 64-bit builds of all OS components. 32-bit applications still run on a 64-bit kernel. Software vendors can include 32-bit and 64-bit builds of their applications on a single disc, if they decide a 64-bit version is warranted. This can all happen 100% transparently to the user. (although I’d assume that companies providing 64-bit builds might place a big “G5 optimized” notice on the front of their box, or whatever name Apple decides to market PPC970 as)
Seeing as Intel is primarily focusing the I2 on the server market and NOT the desktop market all the while still producing x86 CPU’s, what is the point? Until you have every day software that is 64bit then it makes little sense to use a 64bit CPU (of course I am referring to the PC side of the world). If you use 32bit apps with the opteron, it’s not going to make it faster, infact the Opteron will use it’s 32bit arch to process the 32bit app instructions bypassing the 64bit part of the CPU entirely.
Don’t go buying the Opteron because you think it will beef up your 32 bit app speeds. It won’t.
Wait until the desktop software market starts to saturate with 64 bit apps before you run off and buy something just because you have to have it.
.02
I think I’ll still prefer a dual Althon64 or Opteron system for around $1000 than a dual PPC970 for $4500 before the end of next year. PPC will probably scream, but I’m sure Apple won’t make it scream in my price range.
“Apple is absolute, positively *NOT* going x86.”
I knew I should have maybe added more to my original state ment. I in no way said x86. By 2 i was implying the PPC970 and the current PPC since they may have differances between them and not be a no hassel upgrade. I think Apple is going PPC970 as well.
The Opteron is faster than the current crop of Athlons when running 32 bit applications. The integrated memory controller (which runs at CPU clock speed), coupled with the better cache schemes and larger cache make it faster.
Regardless of whether desktop apps get ported to x86-64, getting an Opteron will speed things up.
Can somebody answer the following question…
Consider a linux kernel that is capable of running on the Opteron, 64 bit mode that is. Combine that with a gcc 3.2
compiler which is designed for the 64 bit instruction set.
Then why is it a big deal to port a program which is written in C/C++? Isn’t it a simple recompile.
My point is that, if the kernel and compler can handle the 64 bit instructions, all the system libraries written in C or a higher level language should be trivially ported, and then the apps can be trivially ported. It’s just a simple recompile.
I think the main problem is the size of primitives. On a 32 bit system, and integer (int) is 32 bits wide. But on a 64 bit system, an int would be 64 bits wide.
The most obvious problem is File IO, with binary file formats. You’d be writing twice as much data as expected. Text files won’t shouldn’t have problems, another reason to keep on plugging XML
.
The ‘endian problem’ is how bits are represented in memory. x86 uses a scheme called Least significant bit first (LSB), where 6 in binary would be 011. IIRC, MACs use Most Significant Bit first (MSB), where 6 in binary would be 110. But because the Opteron is x86, this problem doesn’t occur.
Aside from that, there are other problem as well, but seeing as I don’t port code much, I won’t know. That class on computer architecture was so very long ago
You’re right..I should have added the term “respectively”. Meaning that yes, the Opteron is more efficient in that it’s arch is more advanced and it will improve speed, however it will NOT run double quick. It won’t use the “64bit part” to run 32 bit code. I’ve heard people discussing how it will run so much faster because: “can you imagine 32 bit code running on a 64 bit chip?” .
I concede that my previous comment was poor and hastily written.
// I think the main problem is the size of primitives. On a 32 bit system, and integer (int) is 32 bits wide. But on a 64 bit system, an int would be 64 bits wide.//
This should’t be a problem with integer computations in the program. The C compiler should make this transparent to the programmer, i.e., the programmer shouldn’t worry about the size of an int when used for computations. Pointers will be the size of integers, so this should not be a problem as well. I really can’t see bit manipulations going bad either,
since the endian is not changing. I also see no problems with in memory data structures, as long as the data elements are being accessed according to the C standard.
//
The most obvious problem is File IO, with binary file formats. You’d be writing twice as much data as expected. Text files won’t should’t have problems, another reason to keep on plugging XML
. //
I agree with this.