The Real Reason Behind AMD's Downfall

sedaine

Distinguished
Sep 10, 2007
282
0
18,790
For the past year we have been bombarded with news of AMD's Phenom and Barcelona Quad-core processors. This seems to only trouble one person. ME!

The question everyone should be asking is why didn't AMD release the K10 dual core first? Instead they chose to chase a more ambitious dream, albeit difficult to bring to reality - even Intel said it would have difficulty making such a chip.

Further more, Intel had done all the research for AMD and suggested that the market for Quads was not as great as that for the dual cores. AMD still went guned for the Quad.

One thing is clear, had AMD made the K10 dual core (Kuma), it would be very close in performance to the core 2 duo, and it would not have required the much talked about L3 cache with the dreaded bug.

One must ask - if AMD had gone ahead and made the K10 dual core, instead of playing "I have bigger balls than you" against Intel, would they be in a different position right now?

The answer is yes. AMD would be producing Kuma and bringing in revenue, that revenue would then be channeled to producing and fine-tuning the defunct Phenom/Barcelona.

So who is to blame for the downfall of AMD?

Hector Ruiz! Why - because he promised a core 2 quad killer - yet these are not mainstream CPU's but more enthusiast. He went ahead and put most of AMD's finances into the Barcelona/Phenom project and ignored the Kuma CPU. All along they should have been working on releasing a safe bet Kuma, not the high risk Barcelona.

Hector took the AMD ship in the wrong direction - he took it into the core wars with a great looking gun - but no cash to buy the ammo.

 

dragonsprayer

Splendid
Jan 3, 2007
3,809
0
22,780
Intel built the 900 and 800 sereies cpus while they stole the core 2 basic designf rom amd and did the R&D on it.

They combined the 8oo and core 2 to make the qx6700 sprinkled in some tech from p4 mess

amd tried to jump to quad cores when os is only ready for 2 threads - thread rotation may be the issue?

yes I agree with your thread - amd did something many do - they projected publicly. who knows why?

if they had kept a low profile then we be paying alot more for cpu's
 
We don't really know what the Kuma core could have done. I know why AMD went for Quad core first though.

Due to their delays they had in Phenom they had to try. I give them the credit for trying but they still screwed themselves over since they somehow could not get a dual core version out first.

I am not sure why they couldn't since it was basically just a K8 with enhancements, not a new arch like Core 2(technically Core 2 is a new arch from the P4/PD series just based on PIII) so it should have been very easy to update K8 to K10 and release it.

But I think the problems and delays stem from the L3 cache itself as this is AMDs first try for a L3. And with their direct connect architecture might not work right with it without any sort of major overhaul. I know Intel had a L3 with some of its P4 EE's but since it used the FSB and was a single core it might have had less problems. All AMD really needed to do was pump up the L2 like Intel did and make that a shared cache instead of trying to add the L3 as a shared chache.

I just wounder where Kuma is now. You would think that after the Phenom flop they would have rushed Kuma out as fast as possible.
 

MarkG

Distinguished
Oct 13, 2004
841
0
19,010


Surely tri-core would just be a defective quad-core with one failed core? If three cores work, it's much better to sell it as a tri-core chip than throw it away.
 
I have also heard that what AMD had planned was to make only Quad cores and if one had a bad core or two they would market them a tri/dual cores. Even single cores are supposed to be just quad cores with only one core working.

I would hope not. Although that may be effective to AMD price wise, only to make one type of cored CPU, it might not be as that might cause some problems.
 

sedaine

Distinguished
Sep 10, 2007
282
0
18,790
Thing is, AMD plans to release some Kuma's without L3 cache but bigger L2 cache.

This is something they should have and COULD have done 6-9 months ago.

One thing that has to be made clear - Intel makes it's money from the Duo's not the Quads. They only reached 1 million quads 2 months back. Everything else was a Duo. AMD is working on a Quad mobile CPU - again another mistake!

IF AMD had made the Kuma, which would have been a much, much, much easier chip to make - even with the L3 cache - they would be selling millions of Kuma's this Christmas. In fact, it would have been easier for AMD to hit the 3.0 Ghz mark with Kuma.

Since they had been making x2 at 65nm for about a year now, it would have made logical sense to start with Kuma - make revenue - then go for th experimental. At least they would have a product on the market that could compete with the Core 2 Duo's which are still very relevant.

Instead, AMD will release Kuma in 6 months time to battle 45nm Intel chips.

They are a year late with Kuma and they have no excuse for that.

All this because they were trying to live up to their wishful CEO's dreams bringing Intel down instead of strategically planning their poduct line whilst minimizing risk.

I'll be frank - many people would have bought Kuma 6-9 months ago - perhaps even me if it came in at 3.0Ghz. What is the reason for Kuma Q2 2008? Can anyone explain?
 

jkflipflop98

Distinguished



Totally incorrect. Core and Core2's basic design was taken from the Pentium 3 and Pentium M lines. AMD had jack squat to do with the part's design.
 

sailer

Splendid


If Kuma had come out last June like it was originally slated, I probably would have bought one myself. There isn't much real reason for quads now, and far less then. But I think a real reason for the delay of Kuma was teh troubles with K10 in general and the Phenom quad in particular. AMD only has a relatively small amount of money for R&D, so it put that money into trying to get the quad right. Now that has failed, so its dropping back to try to get Kuma out.
 

dragonsprayer

Splendid
Jan 3, 2007
3,809
0
22,780


the 800 series is the 2 die in a package technology
then the 900 single die due core netburst

the core 2 came after the 900 series - the 900 was dual core pentium, shrinks

the qx6700 is the core 2 dual core combined with the 800 series dual chip

the point is intel stepped and tested each way or tick tock before they even called it that


as far as amd - the m was similar pipe line to amd - yes they did not buy or steal tech from amd





 

quantumsheep

Distinguished
Dec 10, 2005
2,341
0
19,790
I reckon IBM should buy AMD as a gateway into the x86 market. If IBM could use it's funding to finance AMD's research we might end up with a total duopoly, which'd be brilliant.

Oh, and they might have enough sense to sack our "friend" Hector.
 

OlSkoolChopper

Distinguished
Dec 15, 2007
564
0
18,980


What amazes me is that you'll find poeple on this forum and elsewhere who are arguing with a strait face that the tricore is a valid chip because of blah blah blah. They don't realize that its as if Honda was selling a Civic with a three cylinder engine because they couldn't make the fourht one work. If that hapened there would be screaming and lawsuits all the way back to Japan. But since it's AMD, the people's darling, the completely blind suporters stay up at nites coming up with reasons why buying a chip with one quarter of it turned off because it is just plain faulty is a good thing! Now they're talking about selling chips where half is faulty! How well does that augre for the other half? What did AMD do for you people? Have great sex with you and you bonded for life like some whiny needy b****? I'm no Intel fanboi and I'd be more than hapy to buy a sizzling AMD Phenom or even two on a QFX II. But right now, AMD is nothing more than an embarasement.
 

WR

Distinguished
Jul 18, 2006
603
0
18,980
What amazes me is that you'll find poeple on this forum and elsewhere who are arguing with a strait face that the tricore is a valid chip because of blah blah blah. They don't realize that its as if Honda was selling a Civic with a three cylinder engine because they couldn't make the fourht one work. If that hapened there would be screaming and lawsuits all the way back to Japan.
Nah, fusing off a defective part of the chip is common in the industry and has no measurable impact on performance or power consumption compared to a part designed without the defective units at all. And unlike "fusing off" a heavy cylinder, no one's going to complain if the CPU comes at 21.99g instead of 21.3g.
 

cnumartyr

Distinguished
Nov 3, 2007
2,287
0
19,780



Tri-Core is a good idea to help yields.

It's like the E21x0 line or the E63/6400. Were they bad chips because they had half of their L2 Cache disabled becuase it was either defective or ran too hot?

Same with the Q9300 Yorkie.

That being said, I wouldn't be suprised is Dual Core Nehalems were just Quad core rejects.
 

sailer

Splendid


Besides that, Intel pulled something similar back in the days of the 386 and 486. They'd sell the good chips as a DX version and chips that didn't have a functioning math co-processor as a SX version. Then they came out with an idea of selling an add on chip which added the math co-processor and doubled the mhz speed. When Intel pulled that stuff, lots of people turned to AMD. Of course, the original defective Pentiums caused no small ruckus and a bunch more people to buy AMD chips. It was following the Pentium disaster that I made the switch to AMD.

Olskool- You mentioned a QFX II. I wondered for a while why a version of the 5000+ BE couldn't be made that would work on the old QFX platform, basically just a 65nm version of the old chip, but most likely AMD just wants to forget that platform and wishes everyone else would forget it too.
 

OlSkoolChopper

Distinguished
Dec 15, 2007
564
0
18,980
I can't stres enough how much of an AMD or Intel fanboi I'm not. I'm sure that each one of the companies had done their dirty deeds at points in the past, but that doesnt efect me at all. I'm interested in making the most out of my next build, not silicon history. I would be happy to buy a Kraft Peanut Buter CPU as long as the price/perf was good. When AMD has a, say, 2.5 Quad at $300 and a 2.5 Dual at $240, is anyone going to care about the 2.5 Tri at $270? Booooooooooring! Its just anothre desparate marketing ploy from a company which has long lost any shred of integrety and trying to cover up the fact that their manufactering processes are substandard with hype hype hype. The old QFX was a misbegoten POS but I am convinced that there is a signifciant market for a dual socket mobo to handle a quad in each side and which doesnt use FB-DIMMs. AMD screwed the pooch with its first try and now a perfectly good concept has gone the way of the dodobird.
 

sailer

Splendid
Writing during a break in the football game. Both companies surely have their faults. I turned to AMD because of some things Intel did years ago. But Intel's management has changed for the better and AMD's has turned to the worse. Right now, I'm looking toward an X48 motherboard with an Intel quad for my next build, as soon as the hardware gets released and in stock at Newegg. I'm willing to pay for performance, but when I do, I want just that, performance and no excuses.
 

OlSkoolChopper

Distinguished
Dec 15, 2007
564
0
18,980


Right on! Go Giants. Not that I give a damn about Eli and Plaxico but I want to see someone anyone cut the Pats (lying cheating videotapers...) down to size. Hold onto that lead NYG!

I've been reading up on X48 and it doesn't seem to realy offer anything too revolutionary, and I cerrtainly am not interested in sheling out for current DDR3 prices. I'm going to be perfectly hapy with a P35 as long as I can stick a 45nm anything in the socket! I'm stil holding onto the scant hope that ncix and other retailers are right and the Wolfies and Yorkies are coming along in three weeks! Let's pray that the Giants beat the Pats and that my Q9550 will be in my paws by the end of Jan! :sol:
 

niz

Distinguished
Feb 5, 2003
903
0
18,980
I'm waiting for Nehalem before I upgrade, its due 2nd half of next year sometime... Aparently that thing will seriously rock. According to some Intel veep the performance gain Nehalem has over Core2 is actually more than Core2 had over Netburst. It does completely away with FSB (finally) which is a real performance bottleneck on multicore systems.