Sledge before Claw??!?!

[ritesh_laud]

Looks like AMD is releasing Opteron before the desktop Hammer! Check out <A HREF="http://www.amdzone.com/articleview.cfm?articleid=1210&page=2" target="_new">AMDZone's</A> notes on today's conference call, scroll down to Hans Mosselman's question near the bottom (search for Mosselman in your browser).

"Q: On hammer it seems strategy has changed to address the server market first. Will you focus on Opteron first?

A: First product coming to market will be based on Sledgehammer die. Barton will ship Q1 will compete performance desktop and mobile. Most focus on Hammer will be server and workstation. Will become player in volume desktop in the 2nd half of next year."

Ritesh

<P ID="edit"><FONT SIZE=-1><EM>Edited by ritesh_laud on 10/16/02 07:05 PM.</EM></FONT></P>

 

[juin]

LOL good luck

Now what to do??

 

[ritesh_laud]

Beats me how they expect the T-Bred Rev B core (even w/ 512K L2) to compete with the P4 for at least the next nine months. I predict a major shift of enthusiasts over to Intel over the coming months.

Ritesh

<P ID="edit"><FONT SIZE=-1><EM>Edited by ritesh_laud on 10/17/02 08:13 AM.</EM></FONT></P>

 

[zengeos]

Well...With 512kb L2 cache, Barton should be competitive with P4. Only Hyperthreading on P4 will ikely keep it faster overal than Barton, and how many people will make use of the hyperthreading?

A few percent maybe?

It seems that Athlon (K7) going into next year will likely be 1-2 speed grades, performance-wise, behind Intel. This will definitely hurt, but if AMD improves their financial situation this quarter it should be survivable.

Looking at the rumored chip production roadmaps that have been floating around the last week or so, it seems that AMD might be making a strategic move to ensure the first production run of chips (500-1000 wafers) is profitable, while they ramp up production further to provide enough supply to meet projected demand.

A disappointment for enthusiasts, for sure, BUT, at least 2/4P systems will be available for those olks wanting them and, since the MP systems were actually relatively good sellers among enthusiasts, the first Opteron systems should also sell (relatively) well.

ClawHammer Clawhamer...wherefor art thou ClawHammer?

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>

 

[vacs]

It's almost frightening to see how many people haven't a clue about how Hyper-Threading works. Most people, as it seems, still believe that HT is some sort of API where the OS and the software have to be specially written and optimized for it to be able to gain some benenfit.

Well, that's exactly the big advantage of HT: you don't need to optimize anything. The OS must support it, that's right (winxp does), but then the OS decides how the programs running on that system can fill up most effectively the execution slots. Every application can immediate take advantages of this technology, it's just depends on how well a program can make use of it. Speed gains can therefore be anything between 0% and 70%, again, without any specific HT optimization necessary for all the different games and applications.

Of course, apps can be further optimized for HT through future patches but you don't need to do that initially, HT works out of the box with WinXP.

Hope that makes it somewhat clear...

Concerning: Sledge before Claw...
The question is not which CPU gets out first (since both won't be available for the massmarket till late summer 2003 but the real questions are:
1. How good will it perform when comparing to the P4s available at that time?
2. How much will it costs?
3. Will be software available to take advantage of this 64-bit extensions?

The answer to the third question I can give straight away: No, there won't be any software available specifically for the x86-64 mode until mid 2004...

 

[eden]

Ok dude, speed ramps are from -10% to 30% on DESKTOP as far as I am concerned, and according to Anandtech. Please don't stretch without proof. Most daily programs will stay at 0, or even decrease performance, while some will give about 10-30.
HT for desktop, in conclusion, is NOT where it's at the most, it's when you use SMP-supported apps, where Xeons and AthlonMPs would be used, THAT is an optimum start, but for daily home use, it can do some wonders perhaps when you are editing a video and wanna hear music as well.

There are already compilers for x86-64, naturally we can expect companies who are doing them to prepare SOME programs at least.



--
"Let Go." -Avril Lavigne

 

[ritesh_laud]

Ok dude, speed ramps are from -10% to 30% on DESKTOP as far as I am concerned, and according to Anandtech.

Well I guess we'll find out on November 14 what improvements HT Rev 2 has over the current HT in the Xeon that Anandtech tested. I'm guessing that they've at least eliminated the *decrease* in performance.

Ritesh

 

[Schmide]

I'm guessing these speed penalties will persist. Sometimes the sum of the parts can't equal the whole. Nothing can make up for good programming and conscious cache usage.

EDIT added: HT is still a very cool technology IMO.

Complicated proofs are proofs of confusion.<P ID="edit"><FONT SIZE=-1><EM>Edited by Schmide on 10/18/02 02:03 PM.</EM></FONT></P>

 

[vacs]

well buddy eden,
At this year's IDF Intel showed a application running 70% faster with HT enabled. So proof is there, you just have to know about it Of course, that was only one program running that much faster, most programs in deed only run 10%-25% faster.

I ask you what good is a 64-bit compiler if there is no 64-bit OS for that platform? Although MS has successfully ported winxp to x86-64, they still haven't officially announced if they will support it! ok, there is also linux, but Torvald himself said that he prefers ia64 over AMD's 64-bit extensions...

Maybe you have forgotten, but it took almost 2 years after releasing the first P4 until the first P4 compilers and optimized apps were available. It's just not going to happen that fast with AMD's x86-64...

Update: yes, those speed penalities will unfortunately persist within the new P4s. But the speed decrease is almost minimal and just happening with some rare applications. I guess everything has its advantages and disadvantages...<P ID="edit"><FONT SIZE=-1><EM>Edited by vacs on 10/18/02 05:03 PM.</EM></FONT></P>

 

[imgod2u]

There have been several hints by Intel interviews (including one Anand did) that HT v.2 will incorporate a kind of psuedo-hyperthreading, which is basically generating yet another thread to help single-threaded programs. This virtual thread will not be part of the program per say, but will use idle processor resources to look ahead in the program, and prefetching information as needed.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

 

[Bahumut]

The question is not which CPU gets out first (since both won't be available for the massmarket till late summer 2003 but the real questions are:
1. How good will it perform when comparing to the P4s available at that time?
2. How much will it costs?
3. Will be software available to take advantage of this 64-bit extensions?
<font color=red>
The answer to the third question I can give straight away: No, there won't be any software available specifically for the x86-64 mode until mid 2004...</font color=red>

Not entirely sure, but was Windows XP-64 a 64 bit version of windows. Whoops I forgot the Microsoft part, guess it's only a fancy 32 bit version.


Pain is the realization of your own weakness.

 

[vacs]

jup, but unfortunately is Intel's 64-bit architecture somewhat quite different than AMD's 64-bit extensions, speak, both are NOT compatible in any way.

Windows XP 64-bit exists but only for intel architecture and this version won't run on any Hammer CPU...

 

[zengeos]

Hyperthreading, similar to multithreading is similar to multitasking in that several threads of one program, or one thread of several programs can be processed simultaneously. Most Wndows programs are single threaded. However, with multiple programs running at the same time, several threads can be processed at the same time. BUT how many people will really benefit from this in a Windows environment. Most people don't really *run several programs simultaneously. They are single tasking when it comes to computers. Will there be some largely intangible benefit? Sure. A few things will seem more responsive, but most tasks will hardly change at all and not noticeably.

It's not a matter of optimizing or not optimizing (needed or not); it's a matter of tangible/noticeable diference in performance. I personally doubt there will be much noticeable difference in the real world, of Hyperthreading...at least initially. So how many people will make use of it? 95%+ likely won't really notice it.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>

 

[eden]

It obviously has improved, but I just wanted to clarify what you said, as it seemed very biased and what would an Intel marketting expert do, I didn't want that to slip by just like that.

HT is nice and all, but don't go advertising that it will let you listen to 5 Britney Spears musics while editing your niece's birthday video!

--
"Let Go." -Avril Lavigne

 

[eden]

Interesting, you could almost say it's an "Optional Prefetching" method which in any case can help, and that it is free to be thrown anytime during processing!

I still wonder at what Intel means by "revised HT".

I'd like to add though, expect higher P4 temps and wattage usage starting on the 3.06GHZ. Could be why they augmented mobo speccs. Obviously making the core more busy would put the clock gating techniques less frequent, therefore an increase of heat output. Not by much I'd say since the trace cache is already feeding half the exec units in theoretical and optimal moments, so a small 10-20% increase should be expected.

--
"Let Go." -Avril Lavigne<P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 10/18/02 09:40 PM.</EM></FONT></P>

 

[Dark_Archonis]

By "revised", Intel says it means that they have improved the performance you get from HT, and Intel has stated that the performance decreases and disadvantages of HT v1 are gone in HT v2. I mean Intel wouldn't enable HT in <i>desktops</i> unless they knew it would be a useful thing.

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! </font color=green>