Future plans for the Duron

[blakerwry]

Anybody know if AMD will continue their Duron line of CPU's? I remember it was originally speculated on some hardware review sites that Durons would move to the 133mHz FSB hat the Athlon XP's enjoyed... but I wonder if that will happen.

Additionally.. do you think it's even neccessary to continue the Duron line cosidering that lower end T-breds(1.3-1.47 gHz) are ~$50 when compared to the ~$39 1.3ghz duron?

Personally I would get the T-bred even for a low cost computer, mainly because it offers a cooler running CPU meaning a quieter computer... the performance increases may or may not be noticable depending on the computer's use (keep in mind that a low cost (<$500) computer was probably not designed to be a gaming machine)

 

[Tea]

In a word, nothing at all. The Duron is EOL. That's official. AMD announced it quite some time ago. The current Athlon line will take over the entry level role (in fact the price difference is such that we stopped selling Durons several months ago - Athlon 2000 is our entry level these days) and the high-end will be filled by the 166 FSB versionz of Toohardtospellbread for a few more months, then the 512k cache Barton, and finally, toward the end of the year, the Athlon 64.

That'z the official word from AMD, and I've not seen any evidence anywhere that they are going to do something else. It was a great little chip, but it's time has come.

 

[Mercutio]

One of the local distributors here had a XP1400 - 1600 CPUs priced LESS than the Duron 1.3 for about two weeks, until they were sold out.

Make no mistake, the Duron is still a good choice for easy upgrades. That's what I've been buying those AK32s for: a $50 board, $30 chip and $5 fan makes for a great upgrade for folk still stuck on early P2s, and the DDR sockets and support for the XP2600 on those boards means there's still room for another upgrade.

Tannin, you're using the 2000 at the low-end? I stick with the 1700+ for budget machines and the 2100+ for anyone with an actual budget. I think there's maybe $10 difference between the 2000 and the 2100 (er, in the US, at least). Even the 2400 is looking widely appealing at this point.

 

[Handruin]

Chech out this roadmap:

 

[blakerwry]

hmm.. looks like the Paris and the Victoria will take the Duron's role in the future(both having 1/4th the L2 cache)...


Question here.... since the new hammer chips will be capable of 64 bit operation while still using x86 instructions.... will software designed for them be able to run on 32bit chips such as the AthlonXP?

is it possible to run something like a virtual machine to do the conversions that might be neccesary?

if we can't run 64 bit programs on our old 32 bit chips... does that mean that we'll see some of the same things that we saw when we went from win3.1 to win32 (16 bit and 32 bit versions of each program.. only now it will be 32bit and 64 bit)?

Or will we see something more similar to current programs that support using SSE/3Dnow... basically if your hardware supports it it will be used.. otherwise it wont... the program doesn't require the "extra" features.

Or will we just not see much software designed specifically for 64 bit Athlon operation? I imagine some things will get a real boost.. while others might not (ex: 3d rendering vs word processing)

 

[Mercutio]

Yep. We have that situation now, in fact. There are IA64 versions of Windows 2000 and Linux, and software must be specially compiled for that chip, exactly like software for an entirely different architecture (e.g. Alpha, PPC, UltraSparc).
x86-64 will run current software, but current chips won't run software compiled for x64-64.

As to why we NEED x86-64, well, there are a few reasons (better precision for things like color depth, addressing for staggering amounts of RAM, that sort of thing), but in general I wouldn't expect it to make much difference.

The "real world" impact of 32-bit software, if you remember back that far, wasn't that great, either. Everyone thought their PCs would get faster and of course they didn't, even if there was an enhancement in capability.

 

[blakerwry]

so do you expect a real push for 64 bit computing in the PC market or is this going to be more of a server/workstation thing?

 

[cas]

does that mean that we'll see some of the same things that we saw when we went from win3.1 to win32 (16 bit and 32 bit versions of each program.. only now it will be 32bit and 64 bit)?

Sort of.

Like Mercutio says, new x86-64 OS binaries and device drivers will be required for 64bit operation(what AMD calls Long Mode). Like the transition from 16 to 32 bit, older application software will work, and will execute without emulation.

The fundamental difference this time, is that a new operating system is not required. Both Linux and Windows for x86-64 are built from the same code base as their 32bit versions. Obviously, there have been some changes to handle the larger processor state, and changes to the page table hierarchy, but most of the code remains unchanged. Both Windows and Linux have been 64bit clean for a while.

For third party programmers (particularly application programmers), it’s just a question of avoiding nasty assumptions, and recompiling your program.

As I work my way through AMD’s 5 volume documentation set, it is interesting to note the various tweaks they have made to the architecture. Some instructions have been dropped, registers have been widened and increased, and the protection scheme has been substantially simplified. Ultimately though, x86-64 is just what AMD promised: a wider x86.

 

[Handruin]

so do you expect a real push for 64 bit computing in the PC market or is this going to be more of a server/workstation thing?

My impression is that AMD plans to target home and mobile users for 64-bit computing. If they can keep the price reasonable for the Athlon 64, I don't see why it can't eventually become main stream. Software support becomes another issue, but I thought the Athlon 64 was backward compatible, even if you sacrific some performance.