What if the Hammer fizzles?

[Quetzacoatl]

As much as we try to avoid it, I think we need to take some time to consider that AMD could fail with the Hammer...As much as I hate to admit it, even being a fan of AMD in general, if their product fails, politically and economically, they are going to inevitably be pushed into a corner by Intel (not in a round room...). Everyone seems to be thinking that the Hammer will blow away everything, but i'm starting to think otherwise, and getting a bit more sucpicious of AMD's claims and dates.

Starting out, we have the .13um T-bred late arrival fiasco, and then some poor/decent yields...not a good sign, also leading to the dropping of SOI on the Barton. Additionally, the only working hammer *so far* is 800Mhz with limited benchmarks.

Now, in retalation, Intel is supposedly pushing for an early release of the Prescott and Yamhill with IA-64..I think (not x86-64 like AMD, right?* Sounds like stiff competition from a much more mature Pentium 4.

Now, on the other hand, what if the AMD Hammer delivers, and Intel get's knocked down another rung, would that be a serious coup? Probably not, but I doubt they would like another Athlon vs Katmai ordeal. Any opinions or thoughts on this?

"When there's a will, there's a way."

 

[chuck232]

Well I think AMD's kinda stuck in a corner already. First off, Intel procs are outperforming Athlons and are starting to really become somewhat competitive in pricing. For example a XP2200+ is about $220-230USD and a P4 2.26B which performs about the same is about $260USD.

Then the hype about Hammer has gotten so high, with everyone expecting phenominal performance, and maybe Hammer just isn't up to it. They've already pushed Hammer's launch to December of this year and Intel seems to be accelerating Prescott to 1H03 in order to stem the flow of AMD procs.

On the other hand if AMD delivers, Intel wil lose some market share, maybe to 60% and AMD gets to 30%, Intel would strike back using its almost endlessly deep pockets of cash, physical ability, and technical ability and tear AMD a new one. I'm pretty sure everyone would agree with me that Intel could easily take out AMD. The only reason they don't is cause:

1. They're comfortable with their 75%+ market share
2. AMD as competition makes Intel and unlikely suspect of monopolism..(dunno if that's a word)
3. They're making loads of money

So why should Intel take out AMD and get made into a NS. Plus the money needed for R&D to creat products to kill AMD, it's just not worth it for Intel. They'd be giving themselves more a lot more trouble than it's worth.

Falling down stairs saves time

 

[lhgpoobaa]

indeed
we all know that intel is greedy, but even they know that its in their best intrest to have some competition. lawsuits against monopolystic tendencies are bad news. and costly.

and your right about the hammer. sure the performance per clock will be better, but it wont be this fantastic superchip that can do everything before u even turn it on. LoL
2 more pipelines will help with speed ramping, extra cache will be good and the onboard memory controller a wellcome step... but to expect more than 25% extra performance @ same clock is pushing it.


<font color=green>Proud member of THG's</font color=green> <font color=blue>Den Of Thieves</font color=blue> :lol:

 

[SammyBoy]

Well, according to all that I've heard, and no doubt a rehash for many people on the boards, is that the T-bred had issues going to .13µ due to AMD's inability to properly redo the Palomino core into a fashion that would make the addition of 256KB cache to the Barton simple and not take away from the Hammer budget. There were timing issues everywhere, and AMD had no extra R&D budget or man-hours to invest into the redesign. The mobile version of the T-bred were out so much earlier because the timing issues only seem to become problems at higher speeds, as shown by the horrible overclocking ability of the T-bred. Rumor has it that Barton samples are already out, so it's quite possible, due to production time, that the timing issues were solved 3 months ago, and that right now, the Barton is finalized and being mass produced to fill the OEM and retail channels. AMD would be hard-pressed, it seems, to get the T-bred up to 2400+ speeds, so I'm thinking that is where the Barton will take over, maybe even at 2300+. T-bred was more of a transition product than an actual core. Much like the Northwood A... I'm sure that Intel didn't expect the tremendous yields they got from that core, and it probably allowed them time to stretch out the NWA core, and redesign the NWB, which, if I remember correctly, is s'posed to eventually have a 10% smaller die, due to redesign alone, no process shrinks.

As to the disapearence of SOI from the Barton core, it has been rumored that much like the T-bred core was a test of the .13µ process, the Barton core would have been a test of SOI wafers. With the appearence of the A0 stepping of the Hammer back at IDF, AMD proved that it was able to create a working (though very raw) SOI processor on the .13µ process. Hammer shouldn't have a problem with .13µ because it has both the budget and the ground-up design for the .13µ process. That said, those issues that AMD admitted were happening to Opteron might be true for the Claw' as well (though it was said to be a MP problem, not a processor problem...), hence the hard-locked processors at Computex and the pushing back of the release rumors.

Will the Hammer flop? Quite possible. It also might succeed beyond AMD's wildest dream. Most likely, though, it will fall somewhere in between. It will probably be a competative processor to not only the NWB, but the Prescott as well, but it probably won't "destroy" the competition, as AMD hinted at in the past.

Couple more things. An <A HREF="http://www.xbitlabs.com/news/story.html?id=1025349410" target="_new">x-bit</A> report seems to show that the Barton's PR rating will jump 600 points when compared to a T-bred of the same clockspeed. This could also be a reason why AMD pushed back the Hammer, as it would seem to indicate, if true, that the K7 core has plenty of life left in it, and shouldn't be relegated to the bargin bin quite yet. To be honest, I think the AMD engineers are very attached to the K7 core, beacause it was that baby that propelled AMD to the top of the enthusiast's world back in 1999 through 2001. They don't want to see it go, and may be a little leary yet of the K8's chances. Therefore, while the budget may not be there, there is plenty of spirit, which is probably why the Barton still exists, instead of just limping along on the T-bred until the Hammer made an appearence. Also in the article, there is indication that all Barton processors will be made at the UMC fab, the foundry that AMD made a deal with a few months back. This would, in essence, open the whole floor of Dresden to Clawhammer and Sledgehammer processors, and also would mean that Dresden would be a pure .13µ SOI fab, while the UMC fab would handle the .13µ bulk silicon (AthlonXP, Duron?).

-SammyBoy

 

[baldurga]

Imo we are responsibles for the high expectatives on Hammer. As times goes on, the "myth" goes high. This is not good, cause people will be disappointed and disatisfied with the Hammer performance when hits the street. The problem won't be K8 but the fan's overoptimistic hopes.

Tbred has been good in lowering this expectations and to get the brain cooler, be more realistic and wait for the new CPU.

DIY: read, buy, test, learn, reward yourself!

 

[Matisaro]

Starting out, we have the .13um T-bred late arrival fiasco, and then some poor/decent yields...not a good sign, also leading to the dropping of SOI on the Barton. Additionally, the only working hammer *so far* is 800Mhz with limited benchmarks.

Locked at 800mhz to prevent overclocking according to amd, take it with a grain of salt. And that 800mhz did beat a 1.6ghz p4.

The Cash Left In My Pocket,The BEST Benchmark

 

[Quetzacoatl]

I'm still not that impressed yet, it was a Willy 1.6 Pentium4, which the Palamino handily defeated anyway. AMD needs to pump out some more benchmarks and higher clocked Hammer's to satisfy the hunger for more proof of it's ability.

"When there's a will, there's a way."

 

[ritesh_laud]

And that 800mhz did beat a 1.6ghz p4.

Well it was an unverified claim by a single web site that supposedly got a chance to sit down with a Hammer and run Quake 3 on it. I doubt that AMD would have allowed actual Hammer benches to be disclosed, so I personally don't believe that that web site was actually testing a Hammer, it probably made the scores up.

Ritesh

 

[Kemche]

800mhz did beat a 1.6ghz p4.[\quote]

And don't hope the preformance will be linear with the frequency increase.

KG

"Artificial intelligence is no match for natural stupidity." - Sarah Chambers

 

[IIB]

all the samples who made it OUT OF AMD run 800Mhz. and tthose were sent on Q1. by now there must be a new batch of samples out there (1.6Ghz?) and inside AMD the Hammer must already run at lunch clockspeeds.


This post is best viewed with common sense enabled

 

[Hoolio]

Does anyone mind if I just chip in?
(no pun intended)

I thought that there was little information on the hammer due to the fact AMD were keeping it all very quiet. THG would have reported if AMD had published a clock speed but I don't think they have so where has this 800Mhz come from?

Alot of the opteron stuff is speculation, there is still very little information out and I still find it exciting but I do hope this processor goes well from AMD. I personally will not believe any benchmarks unless they are from a reptuable source.

As we are now going to 64 bit processors are they now x87 chips not x86, It is difficult to compare a chip that can perform 64 bit instructions to one that can perform only 32 bit instructions. In theory (remember in theory) if the chip is well built it can perform two 32 bit instructions in one 64bit instruction cycle.

Also the intel chip is a pure 64 bit chip, is needs emulation to run 32 bit code, whereas the AMD supports 32/64. I remember a while ago reading that the performance through the 32bit emulation was slow. Therefore I predict unless intel have changed their plan the AMD chip will run better until pure 64bit applications are built. I know there is already a 64bit version of windows XP(i wish there was a 64 bit version of win 2k) otherwise intel would be screwed.

I believe that AMD will pull away at the beginning until 64 apps are built, I believe the AMD plan is to release a pure 64 bit porcessor once 64 bit apps are more common.



<P ID="edit"><FONT SIZE=-1><EM>Edited by Hoolio on 07/01/02 07:51 PM.</EM></FONT></P>

 

[eden]

And it was a 256K L2 CH btw...not the 512K intended.

--
Intel and AMD sitting under a tree, P-R-O-C-E-S-S-I-N-G!

 

[Guest]

Now, in retalation, Intel is supposedly pushing for an early release of the Prescott and Yamhill with IA-64..I think (not x86-64 like AMD, right?* Sounds like stiff competition from a much more mature Pentium 4.

Intel IA-64 is not backward compatable, AMD x86-64 is. this gives another level for the competition.



<i>if you know you don't know, the way could be more easy ...</i>

 

[Kemche]

I Don't know if 512K is intended. What I have heard so far is it's going to be 256Kb at launch. Again only the time will tell.

KG

"Artificial intelligence is no match for natural stupidity." - Sarah Chambers

 

[Kemche]

Imo we are responsibles for the high expectatives on Hammer. As times goes on, the "myth" goes high.

I totally agree. But sometimes this is good. This will make Intel and AMD work harder and deliver what we expect.

KG

"Artificial intelligence is no match for natural stupidity." - Sarah Chambers

 

[texas_techie]

Will the hammer be "a flop". Interesting question.

Sammy sumamrized the situation nicely (almost exactly). But as you all may know, the Hammer is up to at least 1.6 gigs ( as of a 2 months ago anyway).
I dont think Hammer will flop, but I dont think its gonna be the greatest thing since penicilin either. Why? Cuz Intel is putting out some nice silicon. By the time Hammer is released it will likely be 20% faster than Palamino but only 10% (at most) faster than Intel's best. With Prescott, that margin will probably be lost.
But im gonna buy it and the mobo no matter what. Why? The Hammer reference chipset will be the last one designed in Austin. So that board is gonna have sentimental value to me. Dresden will be designing the desktop chipsets in the future. Them Germans make a good car, lets hope they can make a good chipset too.


Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.

 

[eden]

How else would there be a 1MB Opteron?
They said CH will have 512K L2 for sure, so that the other has twice, keeping a good level of performance between each, and not a castrated Duron-like CH.

--
Intel and AMD sitting under a tree, P-R-O-C-E-S-S-I-N-G!

 

[imgod2u]

I thought that there was little information on the hammer due to the fact AMD were keeping it all very quiet. THG would have reported if AMD had published a clock speed but I don't think they have so where has this 800Mhz come from?

The alpha chip of Hammer demonstrated at Computex I think it was, was running at 800MHz. It was benchmarked briefly (we don't even know the system specs asside from the Hammer) against a 1.6 GHz Willamette machine in Q3A and it ran as fast as the Willamette.

Alot of the opteron stuff is speculation, there is still very little information out and I still find it exciting but I do hope this processor goes well from AMD. I personally will not believe any benchmarks unless they are from a reptuable source.

Well, almost all of the performance-related things are speculation asside from that one benchmark.

As we are now going to 64 bit processors are they now x87 chips not x86, It is difficult to compare a chip that can perform 64 bit instructions to one that can perform only 32 bit instructions. In theory (remember in theory) if the chip is well built it can perform two 32 bit instructions in one 64bit instruction cycle.

Not a bit of this is correct. First off, x87 is the FPU specification capable of 64-bit and 80-bit floating point math. It has existed since the Pentium 2 (and even before I think). All modern FPU's are x87 FPU's.
Secondly, 1 64-bit operation does not equal 2 32-bit operations. One of the biggest myths of 64-bit computing. Read the <A HREF="http://www.anandtech.com/guides/viewfaq.html?i=112" target="_new">Anandtech FAQ</A>

Also the intel chip is a pure 64 bit chip, is needs emulation to run 32 bit code, whereas the AMD supports 32/64. I remember a while ago reading that the performance through the 32bit emulation was slow. Therefore I predict unless intel have changed their plan the AMD chip will run better until pure 64bit applications are built. I know there is already a 64bit version of windows XP(i wish there was a 64 bit version of win 2k) otherwise intel would be screwed.

As it said in the FAQ, all 64-bit computing means is the ability to handle 64-bit integers and also the ability to address 64-bit of memory address. With Intel's 4-bit extension, systems can have up to 64GB of memory. So unless you need more memory in that, 64-bit computing doesn't help. The other thing is 64-bit integers. To date I cannot recall one single consumer-level application that either uses or would benefit at all from using 64-bit integers. There's a reason 64-bit hasn't been brought to the consumer market, it's useless in the consumer market as of right now.

I believe that AMD will pull away at the beginning until 64 apps are built, I believe the AMD plan is to release a pure 64 bit porcessor once 64 bit apps are more common.

It doesn't work that way. x86-64 is x86 and that's why it's not compatible with IA-64. Just because a software is 64-bit doesn't mean it'll work on just any 64-bit processor. Does stuff on a Mac, which uses a 32-bit processor work on an Athlon or P4? Things will still need to be written specifically for a "pure 64-bit" processor.

 

[slvr_phoenix]

The alpha chip of Hammer demonstrated at Computex I think it was, was running at 800MHz. It was benchmarked briefly (we don't even know the system specs asside from the Hammer) against a 1.6 GHz Willamette machine in Q3A and it ran as fast as the Willamette.

This still cracks me up to no end. A <b>1.6GHz Willamette</b> is horrible performance in and of itself.

Worse, someone could <i>severely</i> castrate the Willy by pairing it with CAS3 PC100 SDRAM. Why, I'd bet even an 800MHz P3 with CAS2 PC133 could nearly match if not flat out beat such a PoS Willamette system.

That whole thing was just a big PR stunt, worthy only of <b>Apple</b>. AMD is sinking to all new lows.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.

 

[chuck232]

And still I ask myself, how do we even know that these numbers are true? For all we know, it could be made up numbers. I mean I can make a website and say that my confidential CPU Intel gave me is a Prescott2 and can outperform a XP2200+ at 100MHz? For all we know, this could be a load of BS that someone put there for a joke. I mean, you guys don't even trust some pretty reputable websites, but when this popped up, everyone seemed to flock to it as though it were our savior or something. AMD lovers are quick to believe, but someone with a bit of a brain would state the obvious, how do we know that this is real.

Falling down stairs saves time

 

[Siddhartha]

I did not know that the Tbred was late or a fiasco. It may not meet overclocker's expectations but this is a moot point because most consumers do not and will never overclock.

"Just the facts ma'am"

 

[imgod2u]

One of the definitions of being a fanboy is that you put as much bias as possible for the company you have. You will bash any site that speaks something unfavorable towards whoever you're a fanboy for and you will praise any site that says something favorable towards your company of preference. It's not about credibility or truth, it's about being a diluded fanboy.

Secondly, AMD's main market is the enthusiate market, as they don't have very heavy OEM ties to appeal to the mass consumers. So overclocking ability is actually a very big part of AMD's market share.

 

[mbetea]

what if the sky wasn't blue?
what if ants were 3' long?
what if michael jackson never mutilated himself?
what if you were ahmad rashad's son?
what if you see what i'm getting at?


the world looks a lot better using a glidecam

 

[lhgpoobaa]

i dont think it was amazingly late or anything... just people expected it much earlier... besides, they can change their roadmaps any time they want.

OC wise of course the initial batches were dissapointing, but subsequent ones show much better overclocks, indicating that the 0.13 process will mature nicely over the following months.

<font color=purple>I am the Motherboard Destroyer!
Hear Me Roar!</font color=purple> <font color=orange>*SQUEAK*</font color=orange>

 

[Siddhartha]

"One of the definitions of being a fanboy is that you put as much bias as possible for the company you have. You will bash any site that speaks something unfavorable towards whoever you're a fanboy for and you will praise any site that says something favorable towards your company of preference. It's not about credibility or truth, it's about being a diluded fanboy.

Secondly, AMD's main market is the enthusiate market, as they don't have very heavy OEM ties to appeal to the mass consumers. So overclocking ability is actually a very big part of AMD's market share."

Are you calling me a fan boy?


"Just the facts ma'am"