Anyone thought of getting a quad Opteron system? :)

http://www.tyan.com/products/html/thunderk8qs.html
Posted on 2004-05-14 18:14:41 by bitRAKE
Why is that raw maths calcculations means more to me than framerates ever will. Give me a number cruncher and I won't have time for games.
Posted on 2004-05-14 20:19:59 by Eóin
One of the big pluses for me is being able to start with a dual CPU setup and then upgrade to quad CPUs at a later date. This makes the entry level cost less than $2K USD more than a comparible dual Opteron. So, then the question becomes do you need the power of two dual Opterons in one box, and is it worth $2K to do that? This $2K includes the additional cost of eight way CPUs and the motherboard.
Posted on 2004-05-14 21:51:26 by bitRAKE
I generally need a single CPU with more grunt, rather than multiple CPUs. Multi-CPU systems only work when your program is parallelizable. My stuff is usually pipelinable at best.
So a dual Opteron system may be more powerful than a single P4 in theory, but in practice the P4 might be faster, because its single pipeline can easily beat a single pipeline of an Opteron, especially with SSE/SSE2/SSE3 and HTT.

If I would go dual at all, I'd probably choose Xeons, because then I'd still have the advantages of the P4 pipeline.
Quad is really out of the question for me, because it's already unlikely that I can use a second CPU often to speed up calculations. More than 2 CPUs would probably be near 100% waste.
Posted on 2004-05-15 05:20:41 by Scali
I personally only use singlke CPU systems, I havent found the need for a dual CPU system.
Posted on 2004-05-15 21:39:49 by x86asm
It all depends on what you are doing. I have some genetic algorithms I'm playing with and I could really use a few GB of memory and as much processing power as I can throw at it. It is multi-threadable and would fully take advantage of all CPUs. Fortunately, I do have the money as well.
Posted on 2004-05-15 21:58:20 by bitRAKE
I'd like a dual or HT system... every once in a while I have 100% peak CPU usage in a non-preemtible thread, which can cause things like winamp skipping. Or, sustained heavy 100% CPU usage in a single thread can cause the rest of the system to become unresponsive.

But wouldn't have use for more than that, most stuff I write aren't easy to split into MT code. As bitRAKE says, it all depends on what you're doing.
Posted on 2004-05-15 23:43:25 by f0dder
I'm seriously considering to buy a dual Xeon after this summer. I want to get experience with dual-chip and Hyper-Threading before it becomes the norm. Processors can't increase clock frequency forever (gate oxide is six atoms thick at the moment) and multi-core processors will be released in a year or two. It's inevitable. :cool:

By the way, it's not terribly expensive. An Asus PC-DL Deluxe costs 200 $ and a Xeon 2.4 costs only 200 $ as well. Thank you AMD! ;)
Posted on 2004-05-16 01:07:46 by C0D1F1ED
I will never thank AMD. I find it unforgivable that they extended x86 to 64 bit, meaning that we'll be stuck with it for at least another decade, rather than migrating to a modern, efficient 64 bit architecture at this time.

What I wonder though... Where is AMD going next? Prior to Athlon, they never had a decent CPU. The Athlon was basically a clone of the Pentium Pro core, built with technology that they bought from the Alpha division. Now they extended this to 64 bit... But what is next? I haven't heard a thing about a next-gen architecture from them. Perhaps they have no idea, because they'd have to design a CPU from scratch again, and this time there is no example from Intel, and no technology to buy?
Posted on 2004-05-16 05:25:14 by Scali

I will never thank AMD. I find it unforgivable that they extended x86 to 64 bit, meaning that we'll be stuck with it for at least another decade, rather than migrating to a modern, efficient 64 bit architecture at this time.

That part I totally agree with!
What I wonder though... Where is AMD going next? Prior to Athlon, they never had a decent CPU. The Athlon was basically a clone of the Pentium Pro core, built with technology that they bought from the Alpha division. Now they extended this to 64 bit... But what is next? I haven't heard a thing about a next-gen architecture from them. Perhaps they have no idea, because they'd have to design a CPU from scratch again, and this time there is no example from Intel, and no technology to buy?

I've read that they'll bring dual-core processors. Before Intel does... But it's still a Pentium Pro architecture. :rolleyes:
Posted on 2004-05-16 08:35:09 by C0D1F1ED
Sadly, as I said above, multi-CPU/core is not very useful unless you have software that can take advantage of it (is parallelizable).
So if we are stuck at one core-design, and the only thing to get more processing power is to throw more cores at it, I think we have a problem.
The logical thing to do, the way I see it, is to try and optimize the core to make it as efficient as possible, if the silicon technology reaches its limits. Naturally x86 is the first thing to go. You can use the transistors that are wasted on the decoder/tracecache/etc in a much better way.
Parallel to that development, I suppose the software will also have to be optimized more, to make the most of the cores.
Posted on 2004-05-16 08:47:05 by Scali

I will never thank AMD. I find it unforgivable that they extended x86 to 64 bit, meaning that we'll be stuck with it for at least another decade, rather than migrating to a modern, efficient 64 bit architecture at this time.
AMD merely created a viable migration path which Intel was failing to produce. The world was not going to jump to another plateform and continued to pressure for faster x86-32 CPUs -- this is the whole reason Intel produced a product like the P4 - which they are now shifting to a backburner. AMD is helping to push us out of a local maximum, and I thank them for it.
Posted on 2004-05-17 10:32:50 by bitRAKE
Anyway, Itanium was around for ages before the aMD64 and no one bought it cause it just cost too much. Whos fault was that?
Posted on 2004-05-17 11:08:38 by Eóin
It is Intel's and not just because of the price of the CPU. Intel expected the customer to bare the burdon of migration cost. It is an amazing architecture, but cost of ownership is huge. Intel basically said, "if you can afford to migrate to this CPU then you can afford to pay us big money for it." Few people were willing to pay for tickets on a boat they have to row and Intel rides for free. That might have worked when there wasn't a software base, or if code was truely portable.
Posted on 2004-05-17 11:16:01 by bitRAKE
I think you're turning things around, bitRAKE.
You claim that AMD offers a viable migration path that Intel failed to produce. I think AMD creates a migration path that nobody really needs yet, and in the process, shuts Intel out from their own planned migration path. Virtually nobody needs 64 bit yet, which is part of the reason why there aren't any desktop Itanium models yet.
Claiming that the Itanium is too expensive is therefore silly. It's not aimed at the same market as AMD64. It's aimed at big servers that need to process huge amounts of data. It does very well there (competes head-on with monsters like the POWER4, which is not exactly cheap either).

As for migrating... Itanium runs Windows and x86 programs, so migration is just the cost of hardware and Windows. And you can get reasonably cheap Itanium workstations these days, since Intel made some 'low budget' models (< $1000) with less cache and such.
Other than that, an Opteron is just as useless. There are about as little 64 bit programs for Opteron as for Itanium.
And P4s are still faster in most cases, and not many people require 64 bit addressing yet. So the 64 bit thing is mostly marketing hype from AMD at this time.

I would rather go proper 64 bit than to get stuck for some more years with a patched-up Pentium Pro, which really doesn't get us anywhere.
Posted on 2004-05-17 12:19:54 by Scali
another_old_member, if we didn't need the migration path then we would have already migrated away from x86 silly. :) Shutting Intel out of their own migration path is just business - haven't you ever played chess - you promote your pieces while blocking your opponents. :)

What you would rather do has no berring on what the industry can manage in reality. The capital investment in x86 forces a migration path and Itaniums poor performance on x86 has not been the answer.
Posted on 2004-05-17 15:24:58 by bitRAKE
I don't think you understand what I meant.
I meant to say that there wasn't really a need for 64 bit CPUs yet (not for desktops anyway), but AMD tried to create that need.
And that is obviously why we haven't migrated away from x86 to other 64 bit architectures yet, there was no need.

Itanium's poor performance on x86 was not an issue in the market that it is aimed at, and at the time it would move towards the desktop, it would no longer be an issue there either (x86 performance is now fixed by developing a JIT in cooperation with MS, and MS had already developed .NET, which is a better long-term solution). AMD has now spoilt that.

You try to reverse it. You conclude that there is a need for x86-64 because AMD provides the CPUs. This is not true. AMD tries to create a need for them. Intel worked from the assumption that the need is not there yet. Intel is actually right, if you look at the sales of x86-64 CPUs vs x86-32 CPUs (and the lack of 64-bit software).

You might want to open your eyes and look beyond the AMD marketing talk.
Posted on 2004-05-17 15:43:40 by Scali
another_old_member, I am not arguing anything.

I agree there is no common need for 64-bit CPUs, currently.

AMD is positioning - that is obvious.

Itaniums poor performance will be an issue when the time comes.

I conclude there is a need for x86-64 because Itanium does not fill the same role - this is obvious as well.

Intel played conservative because they are in a controlling position.

I have NEVER read any AMD marketing and like Itanium.
Posted on 2004-05-17 15:58:49 by bitRAKE
I agree there is no common need for 64-bit CPUs, currently.


Doesn't sound like it, you claim that AMD is offering a migration path that is REQUIRED.

Itaniums poor performance will be an issue when the time comes.


What poor performance? Itaniums eat Opterons for breakfast. As for x86, firstly, as I said in my last post, that is already fixed (read!!!), secondly, that will not be important. Only native code matters in the long run.

I conclude there is a need for x86-64 because Itanium does not fill the same role - this is obvious as well.


Not obvious at all. Backwards logic.
You say Itanium fills a different role. This does not guarantee that the other role is a required one at all. So far, we have 64 bit CPUs on the market, no OS, no software. I conclude that there is no need.

Intel played conservative because they are in a controlling position.


I would call the Itanium revolutionary rather than conservative. x86-64, now THAT is conservative.

I have NEVER read any AMD marketing and like Itanium.


You sound like you are just reiterating AMD marketing talk anyway, and you are not informed about recent Itanium developments.
Posted on 2004-05-17 16:08:21 by Scali
No OS? Isn't there WinXP for itanium, as well as some linux crap for amd's x86-64?
Posted on 2004-05-17 16:18:40 by f0dder