Sign in with
Sign up | Sign in
Your question

AMD delays

Last response: in CPUs
Share
September 13, 2002 12:33:32 AM

<A HREF="http://news.com.com/2100-1001-957757.html?tag=fd_top" target="_new">Here</A>

<P ID="edit"><FONT SIZE=-1><EM>Edited by a2112a on 09/13/02 01:37 AM.</EM></FONT></P>

More about : amd delays

September 13, 2002 1:25:44 AM

Acocording to lemming it was suppost to be release in august.AMD have trouble with 0.13 micron

At the end i have speak with a horny lady
September 13, 2002 1:51:44 AM

Several sites are reporting the delay now. I was hoping I was inccorect about that rumor - crap. Well, they are claiming the movement to a 333mhx bus is the trouble. Most sane people are calling bullsh*t.
AMD and many other experts have said the change to that bus speed has a minimal impact on performance. So why would AMD delay an extra quarter for minimal gain? Besides, hammer doesnt really have a "bus". So WTF is real reason?
I have a few speculations:

a. They saw their initial release speeds and performance would not be competitive with the latest Norhtwoods. So they need to make some improvements in order to make it competitive.

b. AMD probably has a good idea of what Prescott can do. And seeing the initial hammer couldnt compete (i have proof of this btw) they needed to make some improvements.

c. Since prescott and Hammer would be rivals. That leaves Northwood and Barton for the "value" market. I doubt the Barton could handle a highend Northy - so they wanna make some improvements there.

d. TSMC announced they arent gonna make near as many wafers as projected - and since TSMC was gonna help AMD, they need more time to get volume shipments ready.

e. Also, AMD might be playing around with putting DDR 2 on their memory controllers. I dont have any proof of that though.


Anyone else wanna guess whats behind this delay?

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
Related resources
September 13, 2002 2:37:56 AM

looks like amd fans won't have any presents this christmas. alot of ppl knew this was gonna happen.

prescott will kill hammer. northwood will kill barton. make room for TEJAS sucka!

<font color=green> there's more to life than increasing its speed -Ghandi</font color=green>
September 13, 2002 4:35:40 AM

Quote:
Anyone else wanna guess whats behind this delay?

Looks to me like a problem with their 0.13 micron process. They must be getting piss poor yields, else why would Barton also be delayed? If only Hammer was delayed then we could say it was SOI issues or performance re-evaluation or chipsets or a number of other things, but Barton getting pushed to Q1 '03 means one of two things:

1. 0.13 micron troubles preventing volume shipments (where are those 2400+ AXPs??)

2. AMD can't afford to relegate Dresden fab space to Barton because they need it for Hammer, and UMC can't produce volume shipments of Barton yet.

Anyway since flash has also been in the crapper for the last year, this may be the closest AMD has been to real trouble since the early '90s.

Ritesh
September 13, 2002 12:00:51 PM

Quote:
d. TSMC announced they arent gonna make near as many wafers as projected - and since TSMC was gonna help AMD, they need more time to get volume shipments ready.


While your other points may have some validity, this one likely doesn't. AMD is partnered with UMC, not TSMC. UMC is on schedule with their process apparently.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>
September 13, 2002 1:10:13 PM

I figured it wasn't going to be out on time like 4 months ago.

<font color=blue>Unofficial Forum Cop</font color=blue>
September 13, 2002 2:26:14 PM

Quote:
a. They saw their initial release speeds and performance would not be competitive with the latest Norhtwoods. So they need to make some improvements in order to make it competitive.

b. AMD probably has a good idea of what Prescott can do. And seeing the initial hammer couldnt compete (i have proof of this btw) they needed to make some improvements.

Hell, I expressed concern over this a good while ago and everyone called me nuts. It marks a dark day for AMD when you can say it and you don't have trolls left, right, and center jumping on you thrirty seconds later.

Quote:
c. Since prescott and Hammer would be rivals. That leaves Northwood and Barton for the "value" market. I doubt the Barton could handle a highend Northy - so they wanna make some improvements there.

Eh. I think AMD has almost entirely given up on K7 by now. Look how long it even took them just to finally squeeze Thoroughbred out. And most of us knew then and know now that it was a problamatic core with little room for the expansion that it <i>should</i> have provided. Then ThoroughbredB supposedly gets released which fixes these problems so that AMD can even release the AXP 2400+ and 2600+, only where are they? More delays, and this time over a product that they supposedly have working correctly in the first place.

My opinion is that Barton's delay has nothing to do with its ability to compete with a Northy. Barton's delay is simply because AMD has royally FUBARed their K7 production because they're putting way too many resources into the constantly pushed back Hammer. They neglected putting resources into K7 work for so long that now they're suffering badly because Hammer isn't there to rescue them like they thought it would be.

Quote:
e. Also, AMD might be playing around with putting DDR 2 on their memory controllers. I dont have any proof of that though.

If AMD <i>doesn't</i> do this, I'm going to be greatly disappointed with AMD because it will validate every single concern that I've had over the use of an on-die memory controller not being a flexible enough solution to meet the needs of consumers.

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 13, 2002 2:26:42 PM

I am here to announce the release of my anti-gravity boots.

<font color=red>I'd like to dedicate this post to all my friends, family, and fans. Without them this post would never have been possible. Thank you!</font color=red>
September 13, 2002 2:28:30 PM

Quote:
I figured it wasn't going to be out on time like 4 months ago.

With as many times as Hammer has been pushed back, it's a surprise that AMD CEOs aren't required to wear face paint, because they've definately got some clowns over there. ;) 

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 13, 2002 2:38:00 PM

Quote:
I am here to announce the release of my anti-gravity boots.

And I am announcing the release of my new vita-drink that restores virginity. :) 

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 13, 2002 2:40:17 PM

What is your release date?

<font color=blue>Unofficial Forum Cop</font color=blue>
September 13, 2002 2:40:35 PM

That would be really weird.

<font color=red>I'd like to dedicate this post to all my friends, family, and fans. Without them this post would never have been possible. Thank you!</font color=red>
September 13, 2002 2:42:06 PM

I have announced it today.

<font color=red>I'd like to dedicate this post to all my friends, family, and fans. Without them this post would never have been possible. Thank you!</font color=red>
September 13, 2002 2:47:17 PM

So you are shipping today?

<font color=blue>Unofficial Forum Cop</font color=blue>
September 13, 2002 2:54:01 PM

Yes, I did announce it today.

<font color=red>I'd like to dedicate this post to all my friends, family, and fans. Without them this post would never have been possible. Thank you!</font color=red>
September 13, 2002 3:07:21 PM

Quote:
That would be really weird.

In clinical tests conducted by the nation's leading experts, it has been proven to restore virginity in 98.7% of the women tested and 100.0% of the men tested. And in the survey conducted, 9 out of 10 religious figures cannot identify any discernable difference between a true virgin and our test subjects who drank 'Virgin-Ice'.

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 13, 2002 3:15:42 PM

Excerpt from <A HREF="http://amdzone.com" target="_new">AMDZone.</A>
----------------------------------------------------
<font color=red>
martin2000 on the BBS has posted what he says are the reasons for the Hammer delay. He says his source is someone from AMD, and the reaons do seem reasonable.
(1) Hammer's layout need to redesign. The old design cannot reach competitive speed at introduction. Need to add one extra layer of metal, just like Tbred-B. A new layout revision takes time.

(2) SOI process is not as stable as expected. The yield is still poor and volume production is still a very challeging task. AMD process people is improving SOI process but this is not easy. Even IBM still has problems with SOI but IBM can afford to have a low yield with high-end Power4 unlike AMD where cost is a major concern. The nice thing is AMD already has a non-SOI backup plan in case SOI doesn't deliver the promise.

(3) Integrated memory controller has proven to be a two-edge sword. It does improve performance by 10-20% compared with the same frequency CPU without independent memory controller. However, the integrated memory controller make the CPU less scalable (in term of frequency). Also current design have some stabability issues when performing at high frequency, but new a revision is in progress and hopefully can fix these issues.

(4) realistically Hammer will not become widely available until early or mid Q2. If non-SOI version is required then it will be further delayed until late Q2.

Now I can not vouch for the truth of this information, but again it seems possible.</font color=red>

--------------------------------------
Ritesh
September 13, 2002 3:41:10 PM

The hammer has been pushed back since '98, perhaps '97. Either way, what you call the hammer today is not what the hammer was originally designed to be.
September 13, 2002 4:27:07 PM

Where do I need to order them from?

<font color=blue>Unofficial Forum Cop</font color=blue>
September 13, 2002 4:32:57 PM

i'm noticing that people who've commented on this topic that neither attacks or defends amd's hammer delay is worthless... err, is providing worthless info, and trying to make this major news (however unsurprising to many) irrelevant.

the first sign is denial.

<font color=green> there's more to life than increasing its speed -Ghandi</font color=green>
September 13, 2002 4:34:06 PM

yep yep, it's all of the above. AMD is having problems. Which is very bad for AMD as it could lead them out of business. And the delay truely does hurt cos now everyone is buying pentium 4's. Even my friend bought a pentium 4. It's a lot more flexible in upgradabilty compared to the athlon, which is crap anyway.

<A HREF="http://www.anandtech.com/mysystemrig.html?id=9933" target="_new"> My Rig </A>
September 13, 2002 4:35:19 PM

Nobody is suprised. I expected it. Am I supposed to troll and say AMD now sucks?

<font color=blue>Unofficial Forum Cop</font color=blue>
September 13, 2002 5:20:49 PM

you two are funny..

i like to hear matasiro comment on this though.

<font color=green> there's more to life than increasing its speed -Ghandi</font color=green>
September 14, 2002 12:59:40 AM

Ahhh, what a powerful thing respect is, isn't it? Texas, this isn't directed at you, but just a rant in general. When I said certian things before about AMD, everyone said I was spouting FUD and Hammer would "blow away the northwood" according to many AMD fans. Even alot of respected members in these forums were expecting the Hammer to be released this year. Now, when you (texas) say things which I have already said before a while ago in other posts, everyone is speechless now and no one is saying that you're spouting FUD. Of course, someone as respected as you texas, doesn't get flamed much since you have AMD sources (according to you). Even someone like slvr gets flamed for saying certain things, and it then happens to be true, and all the flamers are speechless. This just goes to show that newer forum members are not as respected, even if they provide high quality information, facts, and most of their information/speculation is usually correct.

Now, anyways, on to the topic:

Quote:
<i>Written by Texas_techie</i>
a. They saw their initial release speeds and performance would not be competitive with the latest Norhtwoods. So they need to make some improvements in order to make it competitive.

b. AMD probably has a good idea of what Prescott can do. And seeing the initial hammer couldnt compete (i have proof of this btw) they needed to make some improvements.

c. Since prescott and Hammer would be rivals. That leaves Northwood and Barton for the "value" market. I doubt the Barton could handle a highend Northy - so they wanna make some improvements there.

d. TSMC announced they arent gonna make near as many wafers as projected - and since TSMC was gonna help AMD, they need more time to get volume shipments ready.

e. Also, AMD might be playing around with putting DDR 2 on their memory controllers. I dont have any proof of that though.

Now, just to let everyone know, these are all educated guesses based on everything I know.

For a.

I knew it! I said the same thing like a few weeks ago in another post, and everyone just said Iw as an Intel fanboy, and said I was "stretching the truth". That's what I believe. I think AMD knows what kind of performance Intel has up their sleeve with the 3.06G P4, not to mention Banias, which will indirectly compete with hammer in the mobile market. Primarily because of hyper-threading improvements, The 3.06G P4 will provide more competition for hammer then expected, which is why they are probably
revising it.

For b.

Also correct. I mean, of course Prescott would perform better than the initial hammers, but now that Prescott has been moved to H2 of next year, that buys AMD some time to further revise hammer.

For c.

I agree with you there.

For d.

I think you meant UMC

For e.

That would be a very wise idea for AMD.

Also, I believe that AMD still does have a few 0.13 micron issues (when i stated this 2 months ago, everyone called me crazy, and Mat said that AMD's 0.13 process was fine). The 0.13 problems are probably causing the delay of hammer and Barton. On top of that, I believe there are in fact issues with SOI, because it's been stated by IBM itself that SOI is challenging to implement and the yields for it are very low. On the plus side, SOI does allow to ramp the clock speed of a CPU by 35%, and significantly lowers power usage. SOI is also expensive. I believe AMD needed SOI for Hammer since it would allow it to ramp to higher clock speeds. If AMD uses the backup-no SOI plan (if they do have one) it will probably affect the hammer's ability to ramp. Also, as several people have mentioned before, the on-die mem controller limits somewhat the ability for the CPU clock speed to ramp high.

Intel orgininally was opposed to SOI, but recently they said it would be needed to overcome several physical limitations in the manufacturing process. Intel, though, won't be using SOI until 2005, since they'll spend these 3 years trying to improve and perfect on SOI.

Quote:
<i>Written by ritesh_laud</i>
Looks to me like a problem with their 0.13 micron process. They must be getting piss poor yields, else why would Barton also be delayed? If only Hammer was delayed then we could say it was SOI issues or performance re-evaluation or chipsets or a number of other things, but Barton getting pushed to Q1 '03 means one of two things:

1. 0.13 micron troubles preventing volume shipments (where are those 2400+ AXPs??)

2. AMD can't afford to relegate Dresden fab space to Barton because they need it for Hammer, and UMC can't produce volume shipments of Barton yet.

Anyway since flash has also been in the crapper for the last year, this may be the closest AMD has been to real trouble since the early '90s.

Correct, AMD is indeed having alot of problems, I believe. Thus, Hammer and Barton have been delayed to 2003. And if Intel holds true to their promise of releasing the 3.06 in november, things will only get worse for AMD.

Also, that excerpt from AMDZone seems pretty accurate according to my estimates, and others'.

Quote:
<i>Written by slvr_phoenix</i>
Hell, I expressed concern over this a good while ago and everyone called me nuts. It marks a dark day for AMD when you can say it and you don't have trolls left, right, and center jumping on you thrirty seconds later.

Exactly! the same thing happened to me when AMD fans went all mad because of the statements I made.

Quote:
<i>Written by slvr_phoenix</i>
Eh. I think AMD has almost entirely given up on K7 by now. Look how long it even took them just to finally squeeze Thoroughbred out. And most of us knew then and know now that it was a problamatic core with little room for the expansion that it should have provided. Then ThoroughbredB supposedly gets released which fixes these problems so that AMD can even release the AXP 2400+ and 2600+, only where are they? More delays, and this time over a product that they supposedly have working correctly in the first place.

My opinion is that Barton's delay has nothing to do with its ability to compete with a Northy. Barton's delay is simply because AMD has royally FUBARed their K7 production because they're putting way too many resources into the constantly pushed back Hammer. They neglected putting resources into K7 work for so long that now they're suffering badly because Hammer isn't there to rescue them like they thought it would be.

That's an interesting point. It actually might be true. Hammer was intended to blow away Intel's flagship, and it's unknown if that will happen. AMD continued working on the K7 so they could keep Intel at bay until Hammer arrived. If too little resources were put into extending the K7, that could be a reason for delays.

Also, I believe AMD didn't really expect to have alot of problems with hammer.

Quote:
<i>Written Dragon9137</i>
The hammer has been pushed back since '98, perhaps '97. Either way, what you call the hammer today is not what the hammer was originally designed to be.

Wow, I'm suprised someone actually has knowledge of this. it is indeed true. AMD has been working on Hammer for at least 5 years, even though it's not a new core; it's based off of the K7.

Shallowbaby, I also would be very interedting to see if Mat has anything to say.

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
September 14, 2002 2:05:00 AM

Its cool,
I post what I know when I know it. I dont recall trolling or jumping on anyone.
I know the current Hammer can beat the best Intel has - but I also know that Prescott is gonna beat Hammer. But only by a VERY small margin. But when prescott starts to ramp up, then the problems start. I dont see Hammer ramping as quickly as Prescott = trouble for AMD.
The fact Hammer can beat Intel now is moot - since they cant get volume shipments. I honestly dont know whats up with the manufacturing process.
My guess, along with everyone elses, is that March had better see some damn impressive stuff by AMD. Or I seriously think AMD will have to fold within a few years. Also - lest we forget - the real money is in the server arena. AMDs future depends more on SledgeHammer than anything. And I havnt heard a single peep about Sledge.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
September 14, 2002 2:50:28 AM

Perhaps Sledge IS the secret weapon for AMD?
In a way they want it to be their main focus for now, therefore Claw will be held without much repercussions in contrast to their new goal of server targetting. Rather, they are focusing much more on getting Sledge in time as promised, so as to not once again look like they cannot supply what companies demanded. (Just like K7 and the commercial which had OEMs waiting with limited supplies)

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A>
September 14, 2002 3:03:03 AM

Quote:
Ahhh, what a powerful thing respect is, isn't it? Texas, this isn't directed at you, but just a rant in general. When I said certian things before about AMD, everyone said I was spouting FUD and Hammer would "blow away the northwood" according to many AMD fans. Even alot of respected members in these forums were expecting the Hammer to be released this year. Now, when you (texas) say things which I have already said before a while ago in other posts, everyone is speechless now and no one is saying that you're spouting FUD. Of course, someone as respected as you texas, doesn't get flamed much since you have AMD sources (according to you). Even someone like slvr gets flamed for saying certain things, and it then happens to be true, and all the flamers are speechless. This just goes to show that newer forum members are not as respected, even if they provide high quality information, facts, and most of their information/speculation is usually correct.


Again those are people who don't have open minds. Fanboys are to be ignored if you are to pursue your life here!
BTW from what Texas said, yes the Hammer CAN destroy NW, it's the Prescott which is what AMD is revising against. Perhaps they are planning indeed to improve it to a point Prescott will need more than an extra FSB and cache.

Quote:
which will indirectly compete with hammer in the mobile market.

Link me to any place where mobile K8 has been mentioned. The way I see it, AMD will probably not do one for a while, seeing as it is going to do nothing but hurt them, given the fact the current problems arising, limited fab space (not to mention UMC having also problems), AND unworthy chipset performance.
BTW with all due honesty, Hammer against Banias, IMO Hammer would definitly win. Banias is supposed to have an IPC close to the P3M and Athlon 4, in fact I heard less than P4M, so I doubt it can possibly compete Hammer Mobiles (they have up to 20-30% better IPC, 20% guaranteed from on-die IMC), IF they were to be made.

Quote:
I mean, of course Prescott would perform better than the initial hammers

While there is probably no doubt it SHOULD, if we know Intel right, they will probably not put THIS much IPC improvements, and also start with less than 800MHZ FSB. I have my doubts about them making a Prescott so 'fantabulastic', it makes the people at Discrete cry for not optimizing for the P4 so well! It will most definitly be a significant improvement over the original P4 IPC, but if you hype it as much as Hammer is being hyped, you will no doubt end up with disappointments. That's why I've given up on hyping etiher Hammers or Prescott.
Quote:
And if Intel holds true to their promise of releasing the 3.06 in november, things will only get worse for AMD.

I think that was the whole point of what we are discussing here, why things are delayed. You honestly can't say they would have delayed Barton if say, the 3.06GHZ came out in January 2003?

Quote:
AMDZone seems pretty accurate according to my estimates

I am surprised the ultimate AMD fanboy refuge had facts to backup the bad stuff coming in AMD's way these days!

Quote:
Wow, I'm suprised someone actually has knowledge of this. it is indeed true. AMD has been working on Hammer for at least 5 years, even though it's not a new core; it's based off of the K7.

I have MAJOR doubts that AMD was thinking of Hammer since 1997. That is so far away, two years later K7 came out, I just don't beleive this unless there is major backed proof. I don't know why, but 5 years ago, it is a bit hard to beleive AMD was thinking of Hammer, if it isn't a NetBurst-like total core redesign, which btw also took 5 years, or so I've heard.

Overall yes AMD is in deep bullcrap, they kicked themselves into the pit, now let's see if the brilliant minds who brought us K7, can dig themselves back up, or else let the market dudes and Mr. Anderson do the auditting firm stuff!

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A>
September 14, 2002 3:36:25 AM

Just hit reply for the last post, so this isn't just for you Eden. When I took a stroll over to amd.com, I had a peep at the roadmap. Lo and behold, it had been <A HREF="http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,..." target="_new">updated</A>. And guess what, it corresponds completely with the cnet.com report of the Hammer being pushed back. As you can see in the roadmap, the Hammer is no longer in both 2H02 and 1H03, but fully in 1H03. Also, you'll notice, if you've looked at AMD roadmaps for the last year, things are no longer seperated in quarters, but halves... which pretty much says to me that they are trying to cover their asses in cases of product slips (well, we said 1H03, and it was released in June, so technically, we are on time...). Also, the Barton core takes the place of the Hammer, in terms of position, meaning that there might be a possible December release, but they have enough time covered to push it out to March 03 and still be in the timeframe. And, there looks to be a quick transition from the .13 process to the .09 process, at least for the Claw... the mobile and sever versions get nothing, at least until '04, it seems.

-SammyBoy
September 14, 2002 4:51:17 AM

Quote:
I have MAJOR doubts that AMD was thinking of Hammer since 1997. That is so far away, two years later K7 came out


I know for a fact AMD has worked on this since 97'
I saw a webcast seminar where someone in R&D from AMD Sunnyvale described the Hammer right down to the registers. And that webcast lecture was given at Stanford.

Here is a link:

http://murl.microsoft.com/videos/stanford/ee380b/000927...

That lecture was given around 97' or so. BTW - im not basing Prescott's performance on what I think. Im basing it on what AMD thinks it can do.




Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
September 14, 2002 4:59:55 AM

LOL, what would happen if you overdose on that, would be tighter than a grasshoppers twat.

I CANT TAKE IT ANYMORE,,<--- comes out of the closet screaming i am an AMD bieocth fan boy..
September 14, 2002 7:00:00 AM

Quote:
Of course, someone as respected as you texas, doesn't get flamed much since you have AMD sources (according to you).


Just chiming in, texas DOES have amd sources, I have pics to prove it, ask him for copies if youd like.

Secondly, when texas says something hes believed, you had nothing to back up your "not out this year claim" when you made it, so then it fit the DEFFINTION of fud.

To sum it up, even meltdown is right sometimes simply because if you always pick black(amd messes up) eventually you will get a result which is black.

:wink: Heatsinks, if you dont overclock, use the <b>STOCK!</b> :wink:
September 14, 2002 7:01:02 AM

Quote:
Also, I believe that AMD still does have a few 0.13 micron issues (when i stated this 2 months ago, everyone called me crazy, and Mat said that AMD's 0.13 process was fine). The 0.13 problems are probably causing the delay of hammer and Barton.


And amd got a 2.8ghz tbred off of a flawwed .13 process....suuure.

:wink: Heatsinks, if you dont overclock, use the <b>STOCK!</b> :wink:
September 14, 2002 5:17:42 PM

The interesting thing is that Sledge is keeping to it's original launch date of H1 2003. It has not been delayed. Hmm, you say the current Hammer can beat a 2.8P4? That seems pretty accurate. Also, do you know if AMD is designing the clawhammer to compete keeping hyperthreading in mind? I agree that AMD has to get their act together or else they will be in big trouble.

Quote:
<i>Written by Eden</i>
Again those are people who don't have open minds. Fanboys are to be ignored if you are to pursue your life here!
BTW from what Texas said, yes the Hammer CAN destroy NW, it's the Prescott which is what AMD is revising against. Perhaps they are planning indeed to improve it to a point Prescott will need more than an extra FSB and cache.


I'm still uncertain that they would revise the current Hammers THAT MUCH, so much that they could beat Prescott when it comes out.

Quote:
<i>Written by Eden</i>
Link me to any place where mobile K8 has been mentioned. The way I see it, AMD will probably not do one for a while, seeing as it is going to do nothing but hurt them, given the fact the current problems arising, limited fab space (not to mention UMC having also problems), AND unworthy chipset performance.
BTW with all due honesty, Hammer against Banias, IMO Hammer would definitly win. Banias is supposed to have an IPC close to the P3M and Athlon 4, in fact I heard less than P4M, so I doubt it can possibly compete Hammer Mobiles (they have up to 20-30% better IPC, 20% guaranteed from on-die IMC), IF they were to be made.

I don't know why you haven't heard of the mobile K8 yet. I mean, it's on the roadmap on AMD's own site <A HREF="http://www.amd.com/us-en/Processors/TechnicalResources/..." target="_new">here</A>. It shows that the mobile Claw will debut in the middle of next year, but because of current Hammer delays, I'm betting the mobile version will be delayed to H2.

I would like to know, Eden, where you "hear" about these things. That is complete FUD; Banias having an IPC <b>less</b> than that of a P4M? Come on, how could you even mention that? Intel has blatantly stated at IDF a few days ago that <b>clock for clock, Banias WILL HAVE better performance than P4M, which therefore means it will have a higher IPC</b>. Banias will <b>definitely</b> have a higher IPC than a P3-M or Athlon 4, because the P3-M Tuallie already is roughly equal in terms of clock for clock performance with the Athlon4.

In regards to the Hammer, though, it will probably perform better than the best Banias. I agree on that. But, even the mobile K8 won't be able to touch Banias in one area: power. At full power, Intel demonstrated that Banias running at 1.6Ghz outputs only 7W of heat. Sitting idle, it drops down to about 1W. Hammer WILL NOT be able to touch Banias in that area, which will be a major seeling point of Banias.

Quote:
<i>Written by Eden</i>
While there is probably no doubt it SHOULD, if we know Intel right, they will probably not put THIS much IPC improvements, and also start with less than 800MHZ FSB. I have my doubts about them making a Prescott so 'fantabulastic', it makes the people at Discrete cry for not optimizing for the P4 so well! It will most definitly be a significant improvement over the original P4 IPC, but if you hype it as much as Hammer is being hyped, you will no doubt end up with disappointments. That's why I've given up on hyping etiher Hammers or Prescott.


I'm not hyping Prescott, I'm just stating the facts. I mean, you do agree that intial hammers won't be able to compete with Prescott.

Quote:
<i>Written by Eden</i>
I have MAJOR doubts that AMD was thinking of Hammer since 1997. That is so far away, two years later K7 came out, I just don't beleive this unless there is major backed proof. I don't know why, but 5 years ago, it is a bit hard to beleive AMD was thinking of Hammer, if it isn't a NetBurst-like total core redesign, which btw also took 5 years, or so I've heard.


Texas backs up my statement that AMD has been working on Hammer since 1997. Also, it's true that the Netburst (P7) core took 5 years to make. Of course, it was made from scratch, and was a radically different design, whereby back in 1997, the K7 was in the process of being completed, and AMD was starting to think about the hammer. But, I'm not too sure whether the original Hammer was supposed to be based off of the K7. Maybe Texas can confirm whether that's true.

Quote:
<i>Written by Matisaro</i>
Just chiming in, texas DOES have amd sources, I have pics to prove it, ask him for copies if youd like.

Secondly, when texas says something hes believed, you had nothing to back up your "not out this year claim" when you made it, so then it fit the DEFFINTION of fud.

To sum it up, even meltdown is right sometimes simply because if you always pick black(amd messes up) eventually you will get a result which is black.


Ok, I agree that Texas has AMD sources, I wasn't saying he doesn't. I'm cool with that. I did in fact have something to back up my claim. It was the same evidence that is now being rumoured to be the problem. Back when I made that claim, I mentioned that AMD would have problems with SOI, and they <i>might</i> have problems ramping Hammer to high clock speeds. Back then, I also mentioned that if AMD had these problems, Hammer would be delayed to Q1 2003. I won't argue that it seemed like FUD to you, but all I'm saying is I try and support my claims with as much evidence/information as possible. IMHO, my claims are not bad, considering I never worked in the tech industry, and don't have any connections at all. Plus, add to the fact I'm only 16. So, I hope that you understand me Mat, and that you realize that I don't get my information out of the black, or make it up, like meltdown. I don't make blind guesses, I base all the rumours, claims, and speculations of mine on evidence, knowledge, and looking at all these things from several different perspectives and thinking of the most "realistic" outcome.

Quote:
<i>Written by Matisaro</i>
And amd got a 2.8ghz tbred off of a flawwed .13 process....suuure.


I meant that AMD may be having 0.13 problems that are specific to Hammer and SOI. With the initial Hammers, they may be having the same problems as Tbred-A, and that may be the cause of the delay, because they are probably revising the core to get rid of those problems. That would back up a claim someone else made that AMD is adding another layer to Hammer.

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
<P ID="edit"><FONT SIZE=-1><EM>Edited by Dark_Archonis on 09/14/02 02:55 PM.</EM></FONT></P>
September 14, 2002 6:38:10 PM

It's worth mentioning that the K7 is pretty much the Next Gen core, but packaged in a Socket 7 form by AMD when they acquired Next Gen. It is very possible that they started their own core based on this shortly after they acquired Next Gen and have been working on it this entire time (i.e. the K8) while still capturing the marketing using the K7 core. Despite the revisions, which by themselves are pretty significant, AMD has never had to make that drastic a change to the interior core of the K7. It's possible their engineers have been working on something else. I.e. the K8.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 14, 2002 6:55:18 PM

Quote:
Just chiming in,

only lemmings chime in, Fugger was right!

"<b>AMD/VIA!</b>...you are <i>still</i> the weakest link, good bye!"
September 14, 2002 7:09:29 PM

Who cares, by the time the chip gets out, its time will be said and done. My money is on that consumers won’t care either. Since they are perfectly happy with the equivalent Intel CPU they bought 4 to 6 months previous.

The CPU could have been AMD's saving grace, but its too late, the company is bleeding out ,and losing market share faster than formula 1 cars go, this so called chip war is over and done with.

Also please remeber general consumers don’t give a [-peep-] about performance they want something cheap and something fast. Something that AMD is having trouble offering. Plus Intel is where they want to be market wise, and AMD just can’t muster it up to give the giant some real competition. I personally would have given them a fair chance if the chip was due when slated. But now all I see is yet another late release for AMD, and yet another disappointing show for them as it has been in the past. But hey AMD has pulled rabbits from their hat’s before, they just might amaze me yet… but I doubt it this time.

-Jeremy


<font color=blue>Just some advice from your friendly neighborhood blue man </font color=blue> :smile: <P ID="edit"><FONT SIZE=-1><EM>Edited by spud on 09/14/02 03:16 PM.</EM></FONT></P>
September 14, 2002 7:40:34 PM

Quote:
Hmm, you say the current Hammer can beat a 2.8P4?

I don't think anybody can argue that Sledge would rape a P4, but it's the Clawhammer we are worrying now, sadly. (Sledge also uses Dual Channel like P4, along with many other IPC boosting techniques)

Quote:
I'm still uncertain that they would revise the current Hammers THAT MUCH, so much that they could beat Prescott when it comes out.

Which is why I am also uncertain to how far will Intel improve Prescott. You know, it may run at 1.2V and 0.09m, but if Intel crams it with cache that heats the CPU dramatically, and more IPC improvements, it can only go so far before it becomes like the 1st NW's thermal measures, which means they will balance the thermal capabilities and the IPC. I know Intel goes very frugal on their chips, so I doubt Prescott is any different, if they are still as devoted to OEM suppliers.
Quote:
I don't know why you haven't heard of the mobile K8 yet. I mean, it's on the roadmap on AMD's own site here. It shows that the mobile Claw will debut in the middle of next year, but because of current Hammer delays, I'm betting the mobile version will be delayed to H2.

Sorry, I claimed something without backing it up, my bad to you and texas!

Quote:
I would like to know, Eden, where you "hear" about these things. That is complete FUD; Banias having an IPC less than that of a P4M? Come on, how could you even mention that? Intel has blatantly stated at IDF a few days ago that clock for clock, Banias WILL HAVE better performance than P4M, which therefore means it will have a higher IPC. Banias will definitely have a higher IPC than a P3-M or Athlon 4, because the P3-M Tuallie already is roughly equal in terms of clock for clock performance with the Athlon4.

I could've sworn THG's Hard News said that, but anyway my bad again, it has to be P4M's performance. Still that means it will be a tad better than P3M's if so, because P3 Tuallies in Mobiles are definitly better performing IPC-wise against P4s, if I recall the previous benches. So that still backs my claim about Hammer Mobile being better.

Quote:
In regards to the Hammer, though, it will probably perform better than the best Banias. I agree on that. But, even the mobile K8 won't be able to touch Banias in one area: power. At full power, Intel demonstrated that Banias running at 1.6Ghz outputs only 7W of heat. Sitting idle, it drops down to about 1W. Hammer WILL NOT be able to touch Banias in that area, which will be a major seeling point of Banias

Hmm I didn't see any IDF articles showing Banias specs, I thought it was disclosed since it was still in dev.
BTW while that can get this much low power consumation, Intel and THG STATED that they will not be able to attain 8 hour durations yet. So whether it's less power or not, it all depends on how it is configured and paired, but it will not be the 8-hour dream we thought it would automatically have.

Quote:
I'm not hyping Prescott, I'm just stating the facts. I mean, you do agree that intial hammers won't be able to compete with Prescott.

Facts or not, it all depends again on Intel's engineering. If they go for a small die like AMD, it may have problems with cooling just like Tbred's small surface. But then again the IHS may help.

Quote:
that intial hammers won't be able to compete with Prescott

Clawhammer or Sledgehammer, you are talking about?
Claw, as I stated for the 3rd time now, it depends on two factors:
1)What AMD is revising against.
2)Intel's goals for Prescott, in thermal and core performance.
Sledge, I still think not. Sledge ALSO will have 1MB Cache IIRC, plus Dual Channel, (more than easily removes that DDR II need for it), SSE2, more core improvements. That is more than enough to compete and beat Prescott. If AMD wants Sledge to be accepted, it HAS to offer very nice server performance.

Quote:
Plus, add to the fact I'm only 16

I must admit that came as a shock!
I thought I and AMD_Man, and a few younger people here were the younguns. Didn't think you were also, I imagined you were above 20, but in a way that explains a lot of your discussions, but it also shows your mature view when it came to discussing money handling in Canada.




--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 09/14/02 03:58 PM.</EM></FONT></P>
September 14, 2002 7:59:24 PM

Hee Hee Hee...all this speculation as to way AMD have put back thier release of Hammer by 1Q. All I can say is that there are a number of reasons why AMD have done this, and all the people jumping on that its a flaw with the Hammer processor could be wrong.

For example it might be more to do with problems tranistioning their FAB to SOI, or UMC aren't ready to take on production of the exisiting Athlon range, or problems with the 8000 series chipset or maybe they still haven't finalised the motherboard specfications (dont forget this is a totally new platform we are talking about)....yes it could also be problems with the Hammer processor itself.

Personally I think AMD are still trying to finalise some niggley problems with transition to SOI and are still yet to finalise the motherboard/platform specifications with third party manufacturers. A more worrying thought I have in the back of my mind it that Hammer does not scale too well and the integrated memory controller is proving to be more difficult that previously thought.

As for AMD delaying Hammer to compete with Prescott...AMD haven't got a friggin clue as to how Prescott will perform..hell Intel haven't even talked about it in any detail all we know is that it will use 1MB of L2 cache and will feature a 800Mhz bus and *may* have some sort of SSE3 enhancements.

Keep guessing people..whatever AMD's problems I hope its not too serious.

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple>
September 14, 2002 8:41:03 PM

Quote:
<i>Written by Eden</i>
Which is why I am also uncertain to how far will Intel improve Prescott. You know, it may run at 1.2V and 0.09m, but if Intel crams it with cache that heats the CPU dramatically, and more IPC improvements, it can only go so far before it becomes like the 1st NW's thermal measures, which means they will balance the thermal capabilities and the IPC. I know Intel goes very frugal on their chips, so I doubt Prescott is any different, if they are still as devoted to OEM suppliers.

I agree. Intel, as you know, is adding a 7th layer to Prescott, compared with NW's 6. Add that with 0.09, low-k dielectric, and of course the low power it consumes and it should rune fine.

Quote:
<i>Written by Eden</i>
Facts or not, it all depends again on Intel's engineering. If they go for a small die like AMD, it may have problems with cooling just like Tbred's small surface. But then again the IHS may help.

Nah. Intel won't go for a REALLY small die, because they know they'll have the same problems AMD has. They also won't make the die too big for Prescott. I forget exactly, but I heard Prescott's dies size is about 80mm^2.

Quote:
<i>Written by Eden</i>
Clawhammer or Sledgehammer, you are talking about?
Claw, as I stated for the 3rd time now, it depends on two factors:
1)What AMD is revising against.
2)Intel's goals for Prescott, in thermal and core performance.
Sledge, I still think not. Sledge ALSO will have 1MB Cache IIRC, plus Dual Channel, (more than easily removes that DDR II need for it), SSE2, more core improvements. That is more than enough to compete and beat Prescott. If AMD wants Sledge to be accepted, it HAS to offer very nice server performance.

Sorry about that, I meant Clawhammer, not sledge. The only thing stopping Intel really will be the cost of Prescott, not the thermal properties. If Prescott is supposed to output less than 60W of heat (I heard about 55W), then heat won't be much of a problem, since NW's already go up to 60W. Plus, Intel is making a new radial high-performance heatsink for Prescott, or so I heard.

Sledge will no doubt outclass Prescott, but it's unfair comparing the 2. The real challenge for sledge would be Nocona (Prescott Xeon).

Quote:
<i>Written by Eden</i>
I must admit that came as a shock!
I thought I and AMD_Man, and a few younger people here were the younguns. Didn't think you were also, I imagined you were above 20, but in a way that explains a lot of your discussions, but it also shows your mature view when it came to discussing money handling in Canada.

I believe it comes as a shock to quite a few people in these forums. Alot of people who don't know me are suprised as to how much I know about computers. Of course, I'm not bragging or anything.



- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
September 14, 2002 8:50:22 PM

"Sledge also uses Dual Channel like P4"

The P4's that use DDR or SDRAM memory don't use dual channel. The Only ones that do are the ones using the old 16 bit versions of RAMBUS. I do not know about the 32 bit ones.

"Which is why I am also uncertain to how far will Intel improve Prescott. You know, it may run at 1.2V and 0.09m, but if Intel crams it with cache that heats the CPU dramatically, and more IPC improvements"

I think that Intel will probably focus on IPC improvements. Hyperthreading is a start and there's also a few neat tricks that they will be using in Banias that will improve it (let's hope they port those over to the desktop P4s).

"Still that means it will be a tad better than P3M's if so, because P3 Tuallies in Mobiles are definitly better performing IPC-wise against P4s, if I recall the previous benches."

The P4s initially were designed mostly for scalability, not IPC rates. Banias will have better IPC than the current P4Ms.

"Intel and THG STATED that they will not be able to attain 8 hour durations yet."

There's more than just the CPU drawing power. Amoung the most power hungry parts are the LCD screens and your drives (especially CD burners). Lowering CPU power consumption DOES help improve battery life.

"If they go for a small die like AMD, it may have problems with cooling just like Tbred's small surface."

The current socket 478 P4's and Celerons are actually very small. They're only about 50% than a post stamp and yet they produce a great deal of heat. Intel came up with a rather interesting heat sink retension mechanism that makes it so that the heat sink is resting on surrounding supports, not entirely the CPU (though they do have a high contact pressure). If I misread your statement, please elaborate more.

"SSE2,"

Where'd you hear that it would have SSE2? I'm just curious.

"If AMD wants Sledge to be accepted, it HAS to offer very nice server performance."

Yes, it had better. :) 

Knowledge is the key to understanding
September 14, 2002 9:09:39 PM

Quote:
The P4's that use DDR or SDRAM memory don't use dual channel. The Only ones that do are the ones using the old 16 bit versions of RAMBUS. I do not know about the 32 bit ones.

I was saying it in a general context, not applying to all, but that Dual Channel IS available on P4s, and in fact DDR will use it soon.

Quote:
I think that Intel will probably focus on IPC improvements. Hyperthreading is a start and there's also a few neat tricks that they will be using in Banias that will improve it (let's hope they port those over to the desktop P4s).

There's something we haven't discussed yet, will HyperThreading add too much heat? If simulating dual cores does have near Dual performance, and knowing that Dual Core setups are heat generators, it may impose a problem for the 3GHZ P4s, and my theory is that a part of the reason for better mobo specifications is to accomodate to Hyper Threading enabled thermal situations.

Quote:
The P4s initially were designed mostly for scalability, not IPC rates. Banias will have better IPC than the current P4Ms.

Yes it's true, but P4M's are still hotter than all laptop CPUs so far, which is quite a suprise considering it uses a much less IPC than others. Also, higher than P4M IPC is what backs my claim, that means it should reach and be a bit better than P3M and Athlon's 4's IPC performance.

Quote:
There's more than just the CPU drawing power. Amoung the most power hungry parts are the LCD screens and your drives (especially CD burners). Lowering CPU power consumption DOES help improve battery life.

A lot of what you're countering me in this post is something I had in mind before stating it, so you were stating the obvious!
However I was saying this because many were thinking Banias systems will be running with 8 hour specs, which I found absurd in current situations, but won't be in the near future.

Quote:
Where'd you hear that it would have SSE2? I'm just curious.

I thought it became known everywhere, SSE2 will be on Hammers. That will most likely prove whether or not Lightwave 7B is still Intel biased, or that in fact SSE2 is propelling P4s so far. SSE2 will definitly stop one of P4's biggest advantages, so that will provide some competition finally, in the multimedia sector, which once used to be the Athlon's biggest strengh.

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A>
September 14, 2002 9:20:01 PM

Quote:
I agree. Intel, as you know, is adding a 7th layer to Prescott, compared with NW's 6. Add that with 0.09, low-k dielectric, and of course the low power it consumes and it should rune fine

Yes I forgot about the new silicon enhancements Intel researched and developped for 0.09m. Those will surely help. However it still leaves my mouth sour as to if the extra 512KB cache may heat a big part of the die and affect overall thermal performance, not to mention the FSB.

Quote:
Nah. Intel won't go for a REALLY small die, because they know they'll have the same problems AMD has. They also won't make the die too big for Prescott. I forget exactly, but I heard Prescott's dies size is about 80mm^2.

You kind of contradicted yourself in a blind way. Tbred B is 84mm^2!
So yes they are going for a small die if so, so they better have good cooling expectations.
Quote:
I believe it comes as a shock to quite a few people in these forums. Alot of people who don't know me are suprised as to how much I know about computers. Of course, I'm not bragging or anything.

It's not so much the knowledge, (we got AMD_Man with that already) but it's the way you write. It just seemed that it couldn't be someone too old, I was suspecting you might be young, but I never gave it guarantee to my thought. And many times, like discussing future cores nobody here talks about, it seemed you were old, cuz that was like juin's job, to talk about never mentioned before cores from Intel!

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A>
September 14, 2002 11:24:18 PM

"There's something we haven't discussed yet, will HyperThreading add too much heat? If simulating dual cores does have near Dual performance, and knowing that Dual Core setups are heat generators, it may impose a problem for the 3GHZ P4s, and my theory is that a part of the reason for better mobo specifications is to accomodate to Hyper Threading enabled thermal situations."

The P4 is a very hot chip even without the hyperthreading. I do not know if the hyperthreading causes any additional increase in temperature, but it's mostly the increase in the clock speed that causes the temperature to rise.

Knowledge is the key to understanding
September 15, 2002 12:35:45 AM

Quote:
I do not know if the hyperthreading causes any additional increase in temperature

I read somewhere that Intel confirmed that hyperthreading is why the 3.06G P4 may have new power requirements. The reason for this is simply that hyperthreading reduces the number of idle transistors by allowing multiple threads to be processed at the same time. Since there are fewer idle transistors, the processor will use more power and dissipate more heat.

From what I remember, the P4 is very aggressive in "clock gating" unused parts of the CPU in order to save power. That's why there such a big difference between the P4's "maximum" power consumption (70W or whatever) and the typical load (significantly less than an Athlon's, hence lower temps). Because during most tasks, much of the P4 is just sitting there not being used and those transistors go into standby and stop switching. Hyperthreading will reduce the amount of clock gating, hence increasing power usage and heat.

Ritesh

<P ID="edit"><FONT SIZE=-1><EM>Edited by ritesh_laud on 09/14/02 07:38 PM.</EM></FONT></P>
September 15, 2002 1:27:41 AM

Wow, good debate here. I've learned mucho about Hammer/Prescott.

By the sounds of it, Hammer wouldn't have been all that great if AMD just launched it as it was. Am I the only one here that's afraid that AMD's going the way of the crapper? Hammer was so hyped and everything and seems like it wouldn't have been able to beat even a NW, much less a Prescott. I mean if they had of released it as it is, I'm pretty sure most people wouldn't have been all that happy.

I don't know much about CPU architecture and stuff, but ritesh, what you said sounds very believable. Now whether its true or not, I don't know, but it sounds very convinceing and makes a lot of sense.

I really hoped that Hammer would be somewhat of a revolutionary product, but maybe it was AMD trying to do too much that set them behind this far. Now with so much effort and money spent in this project, if they pull it through with a lot of sales, I could see AMD fall way back. Sorry to say it, but that's what I think could very easily happen.

P.S. <b>Eden:</b> Do I sound as young as I am? :wink:

...And all the King's horses and all the King's men couldn't put my computer back together again...
September 15, 2002 5:58:12 AM

"There's something we haven't discussed yet, will HyperThreading add too much heat? If simulating dual cores does have near Dual performance, and knowing that Dual Core setups are heat generators, it may impose a problem for the 3GHZ P4s"

Hyperthreading does not use dual cores - it merely seeks to use the existing core more efficiently. (I did noticethat you wrote "simulating dual cores".) I just don't want people to start thinking hyperthreaded CPU's are dual core designs, when they're not.

Your concerns are reasonable, though. Having the ALU and/or FPU doing work a greater % of the time will no doubt generate a little extra heat. I don't think it'll be very drastic though. And you can be sure the folks in product engineering have an acceptable thermal solution in hand. They've also got some very interesting ideas for future thermal solutions, but I can't talk about that. ;) 

* Not speaking for Intel Corp *
!