Hammer, The Chip To Break or Make AMD?

[YossI]

As I see it the K-8 is AMDs Chance to get the approval it deserves and for ever leaving the image of being second place and a follower to Intel.

Performance was never the problem for AMD since the Athlon.

Poor Media coverage & public awareness, notorious VIA chip-sets complains (some legit some not), the Mhz issue with the uneducated crowds and HEAT were clouding the undeniable high performance of the K-7 series.

Now with Hammer AMD gets another shot! A shot to gain the Media & public attention to the MHz vs IPC issue and a chance to departure from the OLD AMD image as being second to Intel.

A chance to tackle the issues clouding the athlon, Like Heat and lack of thermal protection.

And gain a Brand name associated not with a follower but as a leader!

Will the K-8 be the processor to MAKE AMD?

 

[dullard]

<< Will the K-8 be the processor to MAKE AMD? >>



It depends on 3 things: performance, price, and industry acceptance. At the moment we really don't know the answer to any of them.

Performance. There was one leaked benchmark around here about 2 weeks ago. Everyone knows that one benchmark is meaningless (kinda like Apple only showing a photoshop benchmark and nothing else). We also have no way of knowing if the benchmark is true or not. However, lets pretend it is. It showed the Hammer to perform about the same as a 3.4 GHz P4. The Hammer will be released at about the same time that Intel releases its 3.0 GHz P4. Thus the most we can assume is that the Hammer will be roughly 10% faster than the competition - not a major increase in most people's opinion.

Price. I'm willing to bet the price will be higher than AMD customers are used to. Remember just 2 years ago, AMD was selling processors for $800+ each. AMD is running at a loss currently and thus needs a higher priced processor. I'd expect them to go up to $500 easily. Plus we have no idea how much extra costs there will be on a 64-bit processor. Intel's 64-bit processor costs $3000...

Industry acceptance. This part no one knows. However I have a few thoughts. 1) The vast majority of us don't need a 64-bit processor. 2) I assume the 64 bit Windows is extremely expensive. Does AMD have enough clout to force Microsoft to make a cheap consumer 64-bit Windows? I have no idea.

The statements above are my best guesses. If they are true, I don't think Hammer will be an initial success. Its up to AMD. Can they convince us to pay more for a 64 bit processor that we do not need, cannot fully use (no 64 bit Windows), and has only a 10% speed boost?



<< A shot to gain the Media & public attention >>



This depends on the sales people at Best Buy and other computer stores. I know many, many computer buyers think that Intel is better simply since a Best Buy salesman said so 4 years ago. Remember that they work on commision and thus earn more money if they convince a shopper to buy the more expensive Intel computer. That image still hasn't changed, since they liked their Intel computer. AMD needs something extremely drastic to switch the Best Buy salespeoples minds.

 

[Novgrod]

I don't know how to measure public opinion on cpus, so I'll stick by what I have nominally followed: semicondoctor analysts. First, if Enron and the burst of the .com bubble have proven anything, it's that analysts don't know what they're saying. Same thing with the public

When the Athlon came out, it "caught intel by surprise." At the 1 ghz level, it succeeded because "intel can't sell its chips fast enough." The tbird was supposed to "compete with the p3 in the value segment" once the p4 was released. Now the northwood is supposed to "close the door on AMD."

So AMD is an upstart, and nobody quite knows (aside from those people who buy their cpus) why they've been moderately successful. I don't think that anything will change this perception until AMD's market cap is much larger--if it is ever much larger. We can always hope i suppose

I suppose this sounds like I like AMD and i'm playing the persecuted zealot role, but I've heard all this poo on CNBC.

 

[AGodspeed]

Performance. There was one leaked benchmark around here about 2 weeks ago. Everyone knows that one benchmark is meaningless (kinda like Apple only showing a photoshop benchmark and nothing else). We also have no way of knowing if the benchmark is true or not. However, lets pretend it is. It showed the Hammer to perform about the same as a 3.4 GHz P4. The Hammer will be released at about the same time that Intel releases its 3.0 GHz P4. Thus the most we can assume is that the Hammer will be roughly 10% faster than the competition - not a major increase in most people's opinion.

Assuming Intel can get out a 3GHz Pentium 4 by the end of the year (yeah, I know I have to assume the same for AMD). In SpecInt 2000, the ClawHammer 3400+ scored 26% higher than the 3GHz Pentium 4.

If we assume that a ClawHammer 3400+ is equal to a 3.4GHz Pentium 4, then the ClawHammer 3400+ will be 12% faster than the 3GHz Pentium 4 in 32-bit code right off the bat (the difference between 3400 and 3000 is 12%, not 10%). Of course, I make no mention of 64-bit at all, which will be a very big deal to those that have anything to do with Linux. In addition, I'm not mentioning the fact that AMD's model ratings as they stand are fairly conservative. For example, an Athlon XP 2000+ is more or less equal in performance to a 2.2GHz P4 than a 2.0GHz P4, making the 2000+ roughly 10% faster than the 2GHz P4 in widely used computer applications (which has been confirmed by every hardware review site that I've read).

Also, AMD has mentioned serveral times that ClawHammer will scale better as process technologies shrink from .13u to .09u to .065u and so on. And also remember that the P4 won't have an advantage with SSE2, as AMD just recently confirmed that the first Hammers will have full SSE2 support.

I agree with your pricing scenario though. I wouldn't be surprised in the least to see $500 ClawHammers at debut. In fact, I'm betting more like $700 or even $800 ClawHammers. AMD will charge a few extra bucks here and there for that rather large model rating.

I also agree with your "industry acceptance" viewpoint. I don't see MS fully supporting Hammer at launch (of course, I could be dead wrong). However, other OS's (like Linux) have been jumping on x86-64 like it was some hot chick. A bunch of Linux variations have already been ported over to x86-64.

Anyway, we'll see how it all turns out by the end of this year.

2003 is looking like a cool year.

 

[andreasl]

<< Assuming Intel can get out a 3GHz Pentium 4 by the end of the year (yeah, I know I have to assume the same for AMD). In SpecInt 2000, the ClawHammer 3400+ scored 26% higher than the 3GHz Pentium 4. >>



Got a link for this? I seriously hope you're not basing this from the Jerry quote from a previous CC where he said the performance would be twice that of the current SPECINT top processors.

 

[AGodspeed]

<<

<< Assuming Intel can get out a 3GHz Pentium 4 by the end of the year (yeah, I know I have to assume the same for AMD). In SpecInt 2000, the ClawHammer 3400+ scored 26% higher than the 3GHz Pentium 4. >>



Got a link for this? I seriously hope you're not basing this from the Jerry quote from a previous CC where he said the performance would be twice that of the current SPECINT top processors. >>



Here ya go.

In fact, here's the thread I started.

 

[andreasl]

To make a long story short. That 1350 SPECINT number for the Hammer is based on a quote by Fred Weber at MPF last year. Where he said it would be twice as fast as current leading processors. AMD DID NOT RELEASE ANY ACTUAL FIGURES. Everything else is based on the imagination of some creative journalists and wishful AMD supporters. And guess what, I am sure that this is exactly what AMD wanted and they achieved it by being as vague as they could possibly be. Think about it..

Sorry I said Jerry in my earlier post when it was actually Fred Weber who made that statement. But I just think it's funny that things like these get converted into "facts" so soon on messageboards

 

[AA0]

I dunno about you but 10% performance boost seems rather large to me. But I guess your one of those people that gets the video card 5% slower because its cheaper, and a slower chipset, then the slower board with that chipset because its only 5% right? When you start adding things up they get to be big. 10% is todays computers is a lot, think 200Mhz or more.

 

[AGodspeed]

<< To make a long story short. That 1350 SPECINT number for the Hammer is based on a quote by Fred Weber at MPF last year. Where he said it would be twice as fast as current leading processors. AMD DID NOT RELEASE ANY ACTUAL FIGURES. Everything else is based on the imagination of some creative journalists and wishful AMD supporters. And guess what, I am sure that this is exactly what AMD wanted and they achieved it by being as vague as they could possibly be. Think about it..

Sorry I said Jerry in my earlier post when it was actually Fred Weber who made that statement. But I just think it's funny that things like these get converted into "facts" so soon on messageboards >>



Lol, you are certainly correct. I also doubt the validity of those scores. But who knows...

 

[dullard]

<< I dunno about you but 10% performance boost seems rather large to me. >>



Like I said, we have no idea if the Hammer will perform as it claims. 10% sounds like a good improvement. Then you realize that computer chips come out about 8% faster about every 2 months. So if the 3400+ Hammer comes out near the end of the year (about when the 3.0 GHz P4 should come out) - then you find that the P4 is about 2.5 months behind the Hammer in speed. That really isn't much of a difference in time - especially if the Hammer happens to be delayed. If the Hammer came out sooner (like tomorrow) then that is a different story as it is roughly 50% faster than the fastest P4. Again, this assumes the unofficial claims of 3400+ speed is true.

 

[MasterHoss]

I firmly believe that AMD's next generation chips will intensify the competition between Intel and AMD but won't push Intel down to the pits.

 

[Novgrod]

ten percent is not a big figure, nor will it soon be a big figure in computers.

10% is a big change in gnp in a year because the average is some 2%. Where the average with a cpu is some 80% increase annually, 10% is, in fact, not that big a deal.

Furthermore, the premium for that 10% in performance--the difference between the 100% best system and the 90% best system is sharp. A .07 ghz drop in performance (from the xp 2000 to 1900) will save 31% in money, or $88.

With intel, the savings are closer to 35%, or $221.

Now unless you have a quantity of money such as to make $88 or $221 inconsequential, you pay a heavy price premium for that ~5% (in the case of AMD) or ~10% (intel) performance boost.

So you see, 10% doesn't really matter that much, and the only reason to buy it is for conspicuous consumption or irrationality.

 

[BD231]

I dont really want to comment on how fast this new Hammer really is, because, like the rest of you, I have no clue. But you all can be sure that AMD will have something just as fast or faster than what Intel has to offer, thats plain and simply a givin. AMD has come to far to just let the ball drop and say f*ck it, Intel win's.

YossI- What kind of question is this anyway?

Will the K-8 be the processor to MAKE AMD?

There is no possible way ANYONE knows the answer to that Q.

 

[Texmaster]

<< To make a long story short. That 1350 SPECINT number for the Hammer is based on a quote by Fred Weber at MPF last year. Where he said it would be twice as fast as current leading processors. AMD DID NOT RELEASE ANY ACTUAL FIGURES. Everything else is based on the imagination of some creative journalists and wishful AMD supporters. And guess what, I am sure that this is exactly what AMD wanted and they achieved it by being as vague as they could possibly be. Think about it..

Sorry I said Jerry in my earlier post when it was actually Fred Weber who made that statement. But I just think it's funny that things like these get converted into "facts" so soon on messageboards >>



I was curious about their origination. Thanks andreasl!

 

[Remnant2]

My guess -- and we're all guessing right now -- is that the K8 is going to be quite a chip, and probably will be what allows AMD to survive back into profitability.

I'm basing this on:

1) A fairly promising architecture. Not hyper-pipelined, but from what info has been released already, it looks interesting..and an integrated memory controller should boost performance quite a bit.
2) It leverages SSE2. All the work Intel is putting into getting software developers to optimize their apps for SSE2 automatically benefits AMD once Hammer is released. This levels the playing field again with the P4.
3) x86-64. Won't matter much for the desktop market ( I doubt we'll see gamers/productivity apps compiled for 64bit computing for along time), but the benefits of the increased address space and reduced register contention (x-64 also doubles the # of registers) should improve performance on server/workstation software.
4) Small size. It is projected to be a fair percentage smaller than the P4, and ridiculously smaller than anything EPIC. This translates into cost savings and increased # fabbed -- allowing AMD to price them cheaply if need be.
4) rumored performance. bah, not a strong point at all, but what the hell. Since we're all just making guesses, this is as valid a reason as any.

 

[x86]

<< I dont really want to comment on how fast this new Hammer really is, because, like the rest of you, I have no clue. But you all can be sure that AMD will have something just as fast or faster than what Intel has to offer, thats plain and simply a givin. AMD has come to far to just let the ball drop and say f*ck it, Intel win's. >>



I must agree that as long as Jerry Sanders is alive, AMD will never give up. As for the cost of the CPU, I am estimating that it will cost 2X as much as the fastest AMD CPU because it integrates 2 CPU cores (32-bit & 64-bit), and a chipset.

-x86

 

[ST4RCUTTER]

Hammer, The Chip To Break or Make AMD?


That's a tough one to call. We have yet to see a running sample, much less benchmarks that aren't based off software simulation. Does the architecture look promising? Absolutely. With features like an 800Mhz FSB (6.4GB/s bandwidth), better branch prediction (comparable to the P4), on-die memory controller, SOI, SMT, and SSE2 it's definitely a departure from AMD's Intel-modified predecessors. Time will tell. I do think AMD made a wise move in the way they've implemented Hammer. This is a chip that can span every market segment they participate in. Mobile, high end desktop, low end server, and mid to highend (4-chip/8-core) systems. This should cut down on the costs along with the small die sizes. SOI may be a nice little bonus on the desktop, but in the mobile markets it could be critical in allowing AMD to cut down on the core voltages while maintaining performance. All this looks good on paper, but in the end all that matters is delivery of the product.

I think another overlooked key for AMD's success is the return of "some kind" of profitability in the flash market. Many people forget that AMD was making almost as much on flash at the end of 2000 as they were CPU's.

 

[Midnight Rambler]

<< I must agree that as long as Jerry Sanders is alive, AMD will never give up >>

Then AMD is history ... 'cause ol' Jerry is gone come April, I do believe. Maybe there's hope for AMD afterall.

 

[poopaskoopa]

<< Poor Media coverage & public awareness >>



I believe this would be considered poor marketing in every other business, but that'd sound like we're criticizing AMD and that's-just-not-cool. Giving away 50 CPUs at sunrise didn't put these guys on anyone's Hot List, I'd imagine. Still bitter about missing out on a free CPU...

 

[AGodspeed]

dullard, it looks like the fastest P4 this year will be 2.8GHz, not 3GHz. So if AMD releases their ClawHammer 3400+ on time (when 2.8GHz P4 is released) it's performance lead should be 18%, not taking into account the conservative nature of ClawHammer and it's 64-bit capabilities.

 

[Texmaster]

<< dullard, it looks like the fastest P4 this year will be 2.8GHz, not 3GHz. So if AMD releases their ClawHammer 3400+ on time (when 2.8GHz P4 is released) it's performance lead should be 18%, not taking into account the conservative nature of ClawHammer and it's 64-bit capabilities. >>



I want benchies!

 

[KenAF]

AMD does not intend to release Clawhammer this year. I wish people would quit claiming otherwise.

AMD hopes to have the first shipment of the processor by the end of the year. Hope does not equal "will," and "first shipment" does not equal release. It is unrealistic to expect the Hammer to ship this year, or even in January or February of next year. IMO, a more realistic scenario:

January 6: Intel releases 3.0GHz P4 (AMD's top processor at the time is the XP 2800+)
March 28: AMD announces Clawhammer 3400+ and 3000+
July 26: AMD announces Clawhammer 3600+ or 3800+
July 28: Intel introduces its first P4 processors on .11m process, improved P4 design @ 3.4GHz and 3.8GHz
October 26: AMD announces Clawhammer 4000+ based on .13 process
October 28: Intel announces P4 4.0GHz
January 4: AMD announces Clawhammer 4400+ based on .11 process
January 6: Intel announces P4 >=4.4GHz

 

[AGodspeed]

<< AMD does not intend to release Clawhammer this year. I wish people would quit claiming otherwise.

AMD hopes to have the first shipment of the processor by the end of the year. Hope does not equal "will," and "first shipment" does not equal release. It is unrealistic to expect the Hammer to ship this year, or even in January or February of next year. IMO, a more realistic scenario:

January 6: Intel releases 3.0GHz P4 (AMD's top processor at the time is the XP 2800+)
March 28: AMD announces Clawhammer 3400+ and 3000+
July 26: AMD announces Clawhammer 3600+ or 3800+
July 28: Intel introduces its first P4 processors on .11m process, improved P4 design @ 3.4GHz and 3.8GHz
October 26: AMD announces Clawhammer 4000+ based on .13 process
October 28: Intel announces P4 4.0GHz
January 4: AMD announces Clawhammer 4400+ based on .11 process
January 6: Intel announces P4 >=4.4GHz >>



AMD's roadmap clearly states that a ClawHammer 3400+ will be "released" during the 4th quarter. Whether that means they'll only be shipping or actually in retail channels is not known at this time. However, I must say that your timeline above for 2003 is not realistic at all IMO.

Saying that the ClawHammer 3400+ will be delayed a full 3 months and claiming that Intel will have somehow converted to a different process by the end of July in 2003 (.11u, where'd you hear that) is not realistic IMO. But it doesn't matter, you nor I won't know for sure until the time comes.

 

[KenAF]

You post it and show me where it says released. I think you are allowing your enthusiasm for AMD to cloud your judgement.

I suggest you go back and listen to the 4Q report and Banc of America presentation. In one or both, they clearly state that they hope and expect to ship the first Clawhammers by the end of the year. In the latter presentation, they say they are on schedule to ship by the end of the year. IIRC, in these presentations, they actually say "end of the year" several times. I don't recall them ever saying "4th quarter," although that is certainly what is shown on their shipment roadmap. I don't recall them saying "release" a single time during either of those presentations. If you look back to the past roadmaps, AMD has the 2000+ listed for 4Q 2001, the 1800+ for 2Q 2000...and when did we see those processors?

Intel's process roadmap shows the conversion to .11 in the first half of 2003, AMD's roadmap shows conversion to .11 in the second half (around mid-year) of 2003. Of course, arguing about something 18 months into the future...would seem rather pointless.


Ken
KT7A w/ Thunderbird 1.4 @ ~1540MHz; A7M266-D w/ dual 1900+ @ 1.70GHz; KR7A w/ 1900+ @ ~1840MHz

 

[AGodspeed]

If you look back to the past roadmaps, AMD has the 2000+ listed for 4Q 2001, the 1800+ for 2Q 2000...and when did we see those processors?

So wouldn't that mean that AMD would instead release the ClawHammer 3400+ at the beginning of January 2003. Judging by your above comment, that's exactly what it would mean. Yet for some reason you come to the conclusion that the ClawHammer 3400+ will debut in late March of 2003. A little contradicting me thinks....

Intel's process roadmap shows the conversion to .11 in the first half of 2003, AMD's roadmap shows conversion to .11 in the second half (around mid-year) of 2003.

AMD's Processor Roadmap clearly shows .09u conversation sometime in the 2nd half of 2003. Where are you getting .11u from?

I suggest you go back and listen to the 4Q report and Banc of America presentation. In one or both, they clearly state that they hope and expect to ship the first Clawhammers by the end of the year.

I read through the whole slide presentation and have listened to most (but not all) of the audio part. I heard no such thing, however I will listen to the presentation again when I get time.