*AMD Delays Thoroughbred. Fully Updates Roadmap for Next 2 Years.*

[manwithplan]

anyone got any ideas on the actual clock speeds of the clawhammer chips?

It would be interesting to see if the gap between actual Ghz and the PR rating was getting bigger or smaller with the new architecture...

 

[KenAF]

manwithaplan,

As reported by other sites like Xbit labs, the Hammer series will achieve substantially higher IPC, and thus attain much higher performance at equivalent clock speeds. According to Xbit labs, the initial Clawhammer 3400+ should ship at about 2.0GHz (give or take ~100MHz) late this year, with 30+% higher performance per clock than the Athlon. Of course, it's not yet clear whether these ratings are for 32-bit code or x86-64 bit code.

 

[Mavrick007]

AMD's roadmap doesn't look that far off.. so they aren't going to release the "faster" cpu at the designated time, but they are still releasing the TBred and Barton still according to the roadmap here.AMD Roadmap

At least they're not like Microsoft who says they're releasing their new OS at a certain time and end up changing the name cause it conflicts the time released. When have you seen even the product not in beta come out on time? What about Diablo II, it was delayed how long? Haha Warcraft III, I was supposed to have played it by like a year or more ago and it's still not going to be out for awhile. I would say that AMD is still pretty much on track.



<<

<< << I'm sure this question has been asked a million times, will the new AMD chips be compatible with the current socket A form factor and mobos? >>
Not a chance in the world. The Athlon XP utilizes a 462-pin socket, the ClawHammer utilizes a 754-pin socket. >>

>>



The current mobos should work for the Tbred, but I don't expect that they will work for the Bartons, and certainly not for the Hammers coming out. The Tbred is just a supped up Athlon XP and the Bartons are again not alot different than the Tbred other than using the SOI process to reduce heat. The Hammers will be based on the x86-64bit extensions and be the successor to the XP varieties(including XP, TBred and Barton), but they will be significantly different in pin arrangement on the socket like AGodspeed said. I don't know for sure if the current mobos will work for the Bartons, but we shall see hopefully around the end of this year.



<< I presume the Barton core will last until MS has a 64 bit OS for the clawhammer. >>


MS has their 64bit XP server OS (Windows.NET Server) coming out later this year but this is not going to be the consumer based 64bit OS that we will expect. I would say that AMD isn't stupid, they have to be developing with some OS supplier(which will most likely be MS) to develop an OS based on the x86-64Bit variety cause they wouldn't want to lose market share by going with a small company that noone knows about. As much as I don't like it, people have made MS the most consumer used OS around and that's not going to change for quite some time. Why change to a 64bit cpu if no "widely-used" OS didn't offer functionality for that system? I bet they are building their cpu according to specs to the new 64bit OS's being released. We're looking at a new Windows Longhorn(upgrade to XP) coming out probably the end of this year or early next, and then another "major" realease of Windows to follow around 2004 sometime.

I think the Barton will last alot longer than expected since the Hammers will be mostly used in server environments for a long while, and the Barton will take over as being the main consumer rig in AMD's arsenal. But we also have to have a "cheaper" alternative out there too (aka Celeron or Duron) so we'll probably see a crippled version of the Barton too.



<< anyone got any ideas on the actual clock speeds of the clawhammer chips?
It would be interesting to see if the gap between actual Ghz and the PR rating was getting bigger or smaller with the new architecture... >>


I think it's safe to say that they should be rated around what Intel releases around that time. That's what the PR rating is basing off of now in comparison to Intel. The Clawhammer 3400+(64bit) should be out around Xmas and I don't have specs but wouldn't it be safe to say that it would be close to 2.6Ghz? It depends on how AMD sizes up to Intel. If they suddenly start to lose the race with Intel when Intel puts out a faster cpu, they will have to change the PR rating cause it won't match up. I don't like the PR rating cause it is misleading and it only matters to people who don't know much about computers and want to get something new cause of the faster speed or "higher PR" numbers. We rate our processors by Ghz or Mhz and that's not supposed to be the "speed" (although it does have some correlation), it's just the frequency and because of that AMD and Intel aren't going to do the same tasks exactly the same way.

Hehe Just my little rant for the day..

 

[Armitage]

On the OS front ...
As anandtech article shows, the existing 32 bit operating systems will work on the hammer processors. And if you look historically, the OS support tends to lag the chips significantly, we used 16 bit OS on 32 bit CPUs for quite awhile, I expect it will be the same with the 32 to 64 bit switch.

It has also been announced that Suse will be producing a 64 bit linux version for the hammer processors, although I'm suprised they're setting their sights on the 2.6 kernel.

If MS wants to blow off x86-64, they do so at considerable risk considering the market share linux has in the server market.

 

[AGodspeed]

On the OS front ...
As anandtech article shows, the existing 32 bit operating systems will work on the hammer processors. And if you look historically, the OS support tends to lag the chips significantly, we used 16 bit OS on 32 bit CPUs for quite awhile, I expect it will be the same with the 32 to 64 bit switch.

It has also been announced that Suse will be producing a 64 bit linux version for the hammer processors, although I'm suprised they're setting their sights on the 2.6 kernel.

If MS wants to blow off x86-64, they do so at considerable risk considering the market share linux has in the server market.

Yup, I agree. Also see this thread.

 

[Semimaker]

Just to dispel the myth about SOI processors that appears to have grown up around here.

The fabricators use SOI so the transistors they can produce will change faster. It's all in the name the silicon is on an insulator which will not allow the electrons through it so keeping the junction closer to the surface, therefore faster.

Cheers.

 

[Soccerman]

why does no-one think that AMD will be upgrading their K7 line of CPU's besides die shrinks and the like? I can think of at least one thing they could do to speed things up when it comes to mhz-mhz performance.. it only has a 64 bit bus to it's L2 cache, compared to the P3's 256 (I think it was 256, it's been a while) bit wide bus to the L2 cache. but this is only one thing, an improved branch prediction unit (like the one found on the K6-2 line) would also help things along. or if they wanted to go really far, they could impliment some form of SMT that's invisible to the program (ie doesn't require programs to be rewritten like the P4's SMT does).

don't just give up on the K7 yet. AMD has not yet maxed it's performance out yet.

 

[Athlon4all]

<< why does no-one think that AMD will be upgrading their K7 line of CPU's besides die shrinks and the like? I can think of at least one thing they could do to speed things up when it comes to mhz-mhz performance.. it only has a 64 bit bus to it's L2 cache, compared to the P3's 256 (I think it was 256, it's been a while) bit wide bus to the L2 cache. but this is only one thing, an improved branch prediction unit (like the one found on the K6-2 line) would also help things along. or if they wanted to go really far, they could impliment some form of SMT that's invisible to the program (ie doesn't require programs to be rewritten like the P4's SMT does). >>

I agree that there is plenty aside from Die Shrinks that AMD can do to max out the Athlon (all the reasons you mentioned are very good ones, and I wish they would implement them), but it's all about wether AMD thinks it's necessary or not. I guess time will tell, but right now, all indication points to there be no significant architechural improvements being make to the K7 core before Hammer is released. Maybe with Barton we will see these improvements because Barton will most likely eventually reach the same speeds as the Athlon XP and take over the value sector like Intel is doing with the Tualeron

 

[KenAF]

<< why does no-one think that AMD will be upgrading their K7 line of CPU's besides die shrinks and the like? I can think of at least one thing they could do to speed things up when it comes to mhz-mhz performance.. it only has a 64 bit bus to it's L2 cache, compared to the P3's 256 (I think it was 256, it's been a while) bit wide bus to the L2 cache. but this is only one thing, an improved branch prediction unit (like the one found on the K6-2 line) would also help things along. or if they wanted to go really far, they could impliment some form of SMT that's invisible to the program (ie doesn't require programs to be rewritten like the P4's SMT does).

don't just give up on the K7 yet. AMD has not yet maxed it's performance out yet. >>

Yes they have (or will by the end of 1Q 2003 for the desktop). If you listen to AMD's presentations, they say on several occasions that Hammer builds on the "proven core" of the Athlon. If you look at Paul Demone's article, with the AMD's diagrams of Hammer, you see that Hammer is an improved Athlon core with new head-end supporting x86-64 and an integrated memory controller. The Hammer is AMD's upgrade to the K7 line. AMD indicated this week that Hammer will completely replace Athlon on the desktop.

SMT is a lot of work, and is expected for K9 in 2006.

 

[usual_suspect]

Does anyone know for sure the date the xp 2100+ will be released. I called a couple of online stores and no one seems to know. I'm more interested in it because it will mean a drop in 2000+ prices. I know AGodspeed said it was available for pre-order on pricewatch.com, ETA March 14 but I can't find that. Thanks.