The end of AMD

[gorbs]

jumpingjack, thanks for your reply.

i will wait until cyber Monday rolls around then make my decision. i think that is a great deal too but i have absolutely no experience with an AMD setup. thats the main reason i asked plus my eyes are running red from all the reading on-line about the two. all of my pc's are intel based, work and home.

so if i do an AMD rig it will be a new experience for me lol

 

[Acanthus]

Originally posted by: Viditor

Originally posted by: SickBeast
The Phenom is a worse engineering failure than the P4 was IMO. The P4 at least won a good chunk of the benchmarks upon its release. That said, it is interesting to look back upon how much better the P4C was compared to the original. Perhaps the Phenom can mature to that point as well. The only problem is that the P4C made the P4 suck less. It still sucked compared to the A64 (for most usage patterns).

Actually, the P4 had terrible sales in it's first few months...Willamette was even getting beat up by Durons for the first 5 months of it's life.

I can, however, understand why people are comparing the Phenom to the P4. Both were considered somewhat radical designs, albeit for different reasons. Both underperformed at launch, and had many people calling them 'engineering failures'.

Here's the main problem for AMD...intel was the market leader by a country mile when they released the P4. They were guaranteed to sell the processors simply due to the overall demand in the market for chips. AMD could not produce enough CPUs to compensate for their increased demand. AMD is in no position to lose 10-20% marketshare to intel. This would be catastrophic (and many signs point to this being a very real possibilty during 2008).

I have been hearing exactly that statement since mid 2006...it hasn't happened, nor do I expect it to. AMD's marketshare continues to increase, even with the K8...
Since the K10 is obviously superior to the K8, why would the Phenom X2s do worse than current X2s?

Its not $60?

 

[dmens]

Originally posted by: Viditor
I can promise you that this is completely wrong...AMD didn't underestimate anything or get comfy, or fall down on the job.
AMD was limited by it's resource constraints and vastly reduced sales from the recession in 2002...

oh yeah, and how would you know? my school friends at amd say C2D shook their management to the core, and no pun intended.

 

[Evleos]

You guys are aware that the market needs AMD, right?
Us the consumers will have to pay the price if AMD fails. It just seems that some intel fanboys is enjoying AMD's suffering (schadenfreud?)

However, it luckily seems that AMD is going to bump the clock pr clock performance of Phenom, I guess by bumping northbridge speed - which now is locked at 2ghz, making the CPUs scale very poorly.

http://www.amd.com/us-en/Proce..._15332%5E15348,00.html

And as some Anandtech writer pointed out earlier, it is only beyond 2,5ghz we will see it perform.

I'm definately going to migrate to AM2+ as soon as i find a good motherboard.. Now where is SB700...

 

[SickBeast]

Originally posted by: Viditor
Fair enough...but at it's launch, P4 was also (wrongly) thought to be an engineering failure.

I realize that the P4 was conceived by intel's marketing department...so I suppose the engineers were simply following orders. That said, the processor was inefficient...this is why many saw it as a failure (myself included). I suppose 'misguided engineering' may be a better way to put it, but regardless of your terminology, the P4 had some serious issues.

Since Intel cut their prices equally, it should have nothing to do with gaining marketshare.
What is probably more important is that AMD's platforms are far more cost effective in the low end.

Until the E2140 came along, AMD's X2s were significantly cheaper than the cheapest C2Ds, which at the time were the 4400's (costing $120 or so). At that time, you could get an X2 3600+ for around $60, which annihiliated the dual-P4s in their price range.

At this point I would imagine that 90% of people are buying the E2140/60/80 instead of the X2 seeing as they have reached cost-pairity. I'm pretty sure we haven't seen current marketshare numbers.

I think you're forgetting that the huge die is only on the X4s...and Kentsfield is actually larger.
To compare to the X2, we will need to know what the Phenom X2 die size is going to be...

Fair enough. Are you sure they're not just disabling some of the cores?

Kentsfield may be larger, but think about it: if the Phenom die has a single manufacturing error it has to be thrown out. Kentsfield has double the margin of error because they're using two dies.

 

[coldpower27]

Originally posted by: SickBeast
Kentsfield may be larger, but think about it: if the Phenom die has a single manufacturing error it has to be thrown out. Kentsfield has double the margin of error because they're using two dies.

The difference in die size between Kentsfield and Agena is pretty much negligible in terms of the total surface area.

Agena = 285mm2
Kentsfield = 2x143mm2 = 286mm2

Intel is definitely the one who has the advantage right now with a much more mature 65nm process, cheaper without SOI, as well MCM approach.

In reference to Phenom X2 if they are going to call it that anymore as all I have heard it's going to be the Athlon x2 6000 Series, were going to likely get die sizes in the 180-200mm2 range which is pretty much a safe bet. Remember your still keeping all the LV3 Cache this is not 1/2 like the Conroe is to Kentsfield.

 

[SickBeast]

Originally posted by: coldpower27

Originally posted by: SickBeast
Kentsfield may be larger, but think about it: if the Phenom die has a single manufacturing error it has to be thrown out. Kentsfield has double the margin of error because they're using two dies.

The difference in die size between Kentsfield and Agena is pretty much negligible in terms of the total surface area.

Agena = 285mm2
Kentsfield = 2x143mm2 = 286mm2

Intel is definitely the one who has the advantage right now with a much more mature 65nm process, cheaper without SOI, as well MCM approach.

In reference to Phenom X2 if they are going to call it that anymore as all I have heard it's going to be the Athlon x2 6000 Series, were going to likely get die sizes in the 180-200mm2 range which is pretty much a safe bet. Remember your still keeping all the LV3 Cache this is not 1/2 like the Conroe is to Kentsfield.

That's interesting. With the additional cache per core, that chip may be able to match or even beat the C2D clock-for-clock, especially if they can work the bugs out. Also, heat will be less of an issue, which should help overclocking.

Nice info. :thumbsup:

 

[coldpower27]

Originally posted by: SickBeast

Originally posted by: coldpower27

Originally posted by: SickBeast
Kentsfield may be larger, but think about it: if the Phenom die has a single manufacturing error it has to be thrown out. Kentsfield has double the margin of error because they're using two dies.

The difference in die size between Kentsfield and Agena is pretty much negligible in terms of the total surface area.

Agena = 285mm2
Kentsfield = 2x143mm2 = 286mm2

Intel is definitely the one who has the advantage right now with a much more mature 65nm process, cheaper without SOI, as well MCM approach.

In reference to Phenom X2 if they are going to call it that anymore as all I have heard it's going to be the Athlon x2 6000 Series, were going to likely get die sizes in the 180-200mm2 range which is pretty much a safe bet. Remember your still keeping all the LV3 Cache this is not 1/2 like the Conroe is to Kentsfield.

That's interesting. With the additional cache per core, that chip may be able to match or even beat the C2D clock-for-clock, especially if they can work the bugs out. Also, heat will be less of an issue, which should help overclocking.

Nice info. :thumbsup:

Thanks by the way!

I don't see why it would, in Single Threaded scenarios on Quad Core Agena a Single Core has access to all the cache already when the other 3 aren't doing anything, and if your talking about 2 thread scenarios then 2 cores have access to the cache which is the same as Kuma essentially anyway.

 

[CTho9305]

Originally posted by: SickBeast

Originally posted by: Viditor
Fair enough...but at it's launch, P4 was also (wrongly) thought to be an engineering failure.

I realize that the P4 was conceived by intel's marketing department...so I suppose the engineers were simply following orders. That said, the processor was inefficient...this is why many saw it as a failure (myself included). I suppose 'misguided engineering' may be a better way to put it, but regardless of your terminology, the P4 had some serious issues.

I dunno... have you ever read Doug Carmean and Eric Sprangle's paper which makes engineering arguments for super-deep pipelines (50-70 stages)?

 

[Acanthus]

Originally posted by: Evleos
You guys are aware that the market needs AMD, right?
Us the consumers will have to pay the price if AMD fails. It just seems that some intel fanboys is enjoying AMD's suffering (schadenfreud?)

However, it luckily seems that AMD is going to bump the clock pr clock performance of Phenom, I guess by bumping northbridge speed - which now is locked at 2ghz, making the CPUs scale very poorly.

http://www.amd.com/us-en/Proce..._15332%5E15348,00.html

And as some Anandtech writer pointed out earlier, it is only beyond 2,5ghz we will see it perform.

I'm definately going to migrate to AM2+ as soon as i find a good motherboard.. Now where is SB700...

Not wanting AMD to fail or wanting AMD to fail is not what causes AMD to ultimately fail...

I keep seeing this argument and it makes no sense at all.

Yes, competition is good.

AMD is failing to compete on all fronts and bleeding money like a seive. No economics spin is going to fix that. Companies that dont make money eventually fail.

End of story.

As for the Phenom comments, we all know how much HTT speed has affected the architechture in the past (hint: 500mhz and 2000mhz are about 1% apart).

 

[SickBeast]

Originally posted by: CTho9305

Originally posted by: SickBeast

Originally posted by: Viditor
Fair enough...but at it's launch, P4 was also (wrongly) thought to be an engineering failure.

I realize that the P4 was conceived by intel's marketing department...so I suppose the engineers were simply following orders. That said, the processor was inefficient...this is why many saw it as a failure (myself included). I suppose 'misguided engineering' may be a better way to put it, but regardless of your terminology, the P4 had some serious issues.

I dunno... have you ever read Doug Carmean and Eric Sprangle's paper which makes engineering arguments for super-deep pipelines (50-70 stages)?

I haven't.

Every real-world example of that type of chip has proven to be slower than other options. Look at the AMD 2900XT GPU. It has a 'longer pipeline' (in GPU terms) than the G80-nVidia cards, and it gets trounced in most benchmarks. It requires a higher clockspeed to compete, which makes it a hotter running and less efficient chip. Same goes for the P4. It just doesn't make sense to me. What have AMD and intel both done wrong?

 

[CTho9305]

Originally posted by: SickBeast

Originally posted by: CTho9305

Originally posted by: SickBeast

Originally posted by: Viditor
Fair enough...but at it's launch, P4 was also (wrongly) thought to be an engineering failure.

I realize that the P4 was conceived by intel's marketing department...so I suppose the engineers were simply following orders. That said, the processor was inefficient...this is why many saw it as a failure (myself included). I suppose 'misguided engineering' may be a better way to put it, but regardless of your terminology, the P4 had some serious issues.

I dunno... have you ever read Doug Carmean and Eric Sprangle's paper which makes engineering arguments for super-deep pipelines (50-70 stages)?

I haven't.

Every real-world example of that type of chip has proven to be slower than other options. Look at the AMD 2900XT GPU. It has a 'longer pipeline' (in GPU terms) than the G80-nVidia cards, and it gets trounced in most benchmarks. It requires a higher clockspeed to compete, which makes it a hotter running and less efficient chip. Same goes for the P4. It just doesn't make sense to me. What have AMD and intel both done wrong?

I'd be skeptical about that GPU part. In GPUs, many of the issues that arise in deeply-pipelined general-purpose processors (load to use latency, back to back dependent instruction issue, branch mispredict latency, etc) are non-issues complexity- and performance-wise because of the throughput-oriented latency-insensitive nature of graphics calculations. Other factors probably have more significance in the 2900xt-vs-8800GT performance differences. Unfortunately I know almost nothing about the ATI GPU architecture (nvidia has some really good presentations on their GPUs). I know enough to be dangerous (enough to draw incorrect conclusions), so take this with a grain of salt .

For the P4, keep in mind that it wasn't just a deep pipeline - it was narrower than P3 / K7, had a miniscule L1 data cache, and a trace cache rather than an ordinary L1 instruction cache. I wouldn't doubt that some of these (and other) differences from contemporary architectures had signiificant roles in P4's suckage.

 

[Viditor]

Originally posted by: dmens

Originally posted by: Viditor
I can promise you that this is completely wrong...AMD didn't underestimate anything or get comfy, or fall down on the job.
AMD was limited by it's resource constraints and vastly reduced sales from the recession in 2002...

oh yeah, and how would you know? my school friends at amd say C2D shook their management to the core, and no pun intended.

I also have friends at AMD, and I agree that C2D really depressed them...but that wasn't my point. The point is that there was absolutely NOTHING more that AMD could have done given their limited resources.
It's not a question of a lack of will or foresight, it's a question of resources and the amount of time it takes to design and build a new chip.

 

[VirtualLarry]

Originally posted by: SickBeast
They have Fusion and maybe 45nm to look forward to, but intel will surely have an answer to both. Not only that, but I would rather a 'C2D' derivative in my future CPU instead of a 'Phenom' derivative.

Why would you say that? There's nothing inherently inferior, that I can see, about the Phenom design. The primary problem is the yields and the Ghz that they can clock to. If they were competitive in that dept, their design is definately also competetive.

 

[w00t]

I'll believe AMD is done when they are

AMD is working on fusion to incorporate a CPU and GPU into one chip increasing mass production and decreasing cost.

 

[Phynaz]

Originally posted by: w00t
I'll believe AMD is done when they are

AMD is working on fusion to incorporate a CPU and GPU into one chip increasing mass production and decreasing cost.

Which is 3 years out, according to recent history.
What keeps the company fed until then?

 

[DrMrLordX]

http://www.asiarooms.com/trave...f-uae/uae-cuisine.html

Arabic Mezze?

 

[Cookie Monster]

Originally posted by: w00t
I'll believe AMD is done when they are

AMD is working on fusion to incorporate a CPU and GPU into one chip increasing mass production and decreasing cost.

Yes for the low end market and it wont be here til maybe late 2008 or 2009.

 

[Viditor]

Originally posted by: Phynaz

Originally posted by: w00t
I'll believe AMD is done when they are

AMD is working on fusion to incorporate a CPU and GPU into one chip increasing mass production and decreasing cost.

Which is 3 years out, according to recent history.
What keeps the company fed until then?

I see...then let's base Nehalem's release on the actual release vs the proposed release of Itanium. Are you saying we will see Nehalem in 2011 then?

It's a nonsensical argument...

 

[SX2012]

I dont mean to appear like some AMD fan-boy, but almost everyone(except for you enlightened people reading this) seem to think the flagship parts makes or breaks the whole company. When in fact..... you may be forgetting that the average person is totally oblivious to whatever they are purchasing. The typical mom n' pop and non-gamers are the people buying most of the computers and only know what a salesmen tells them. AMD can be slower across the board indefinitely still be a healthy company if they manage it right.

What I still don't understand to this day is the switch from 939 to AM2 for the x2's which gave like a 2 percent performance boost and renedered my DFI ultra-d obsolete somehow. Just an example of BAD MANAGEMENT. Thats why a lot of gamers went from 939 to Conroe because AMD screwed them over for nothing and Intel walked right in and offered a no-brainer deal.

ohh, and the 2900XT is the only reason i would need to fire everyone on the top and start from scratch.

 

[Aluvus]

Originally posted by: SX2012
What I still don't understand to this day is the switch from 939 to AM2 for the x2's which gave like a 2 percent performance boost and renedered my DFI ultra-d obsolete somehow. Just an example of BAD MANAGEMENT. Thats why a lot of people just skipped AM2 or went to Core Duo.

The purpose of going to Socket AM2 was to move to DDR2 memory. Due to the integrated memory controller on the Athlon 64s (and X2s), a new socket was required to change memory types. AMD delayed this transition by quite a lot, and made the move long after Intel did, because they wanted to wait for performance of DDR2 to catch up.

 

[hans007]

i would say that amd is maybe in about as bad shape as they were during the k6 days. that said, i woudlnt be surprised if they turned it around with an infusion of cash. they probably will have to remake themselves into some other sort of company, which they already are with the amd fusion project etc. i think they gave it one more shot to beat intel at intel's own game but they probabyl have figured that they cant now.

that said, i doubt they are dead.

 

[SX2012]

this i know, from this here article

http://www.anandtech.com/showdoc.aspx?i=2738&p=4

Okay so AMD had DDR-I and Intel had DDR-II.....big deal. From those old benchmarks i can see they hadnt even saturated DDR yet. Sure the bandwidth went up but the chips never took advantage of it. They should have spent that R&D they blew on nothing and used it to get Phenom out 6 months earlier, because that whole time they bleed themselves to death offering old cpu's at higher frequencies.

 

[JumpingJack]

Originally posted by: gorbs
jumpingjack, thanks for your reply.

i will wait until cyber Monday rolls around then make my decision. i think that is a great deal too but i have absolutely no experience with an AMD setup. thats the main reason i asked plus my eyes are running red from all the reading on-line about the two. all of my pc's are intel based, work and home.

so if i do an AMD rig it will be a new experience for me lol

Not a problem ... in terms of working with AMD equipment, it is no different really (unless you are trying to overclock, then there are some other 'tricks' to squeeze more out). It is the same kind of build (case then PSU/drives, socket CPU, memory, add MB, cable up, power up and install OS)... Intel or AMD it is the same procedure.

Netburst has, rightfully so, soured many tastes to Intel ... the C2D introduction, and subsequent revisions simply out class anything (Athlon or Pentium 4) on the market right now ... you may want to consider reading up on the reviews, you can skim them by just looking at the Anandtech review.

For the cheapest possible build though AMD is a good build, I have different philosophies around computer building... I like to game as much as the next guy, but I also use my computer for a host of many different things so I tend toward what is faster even if it is 50 to 100 bucks more expensive. If gaming is all you want to do... the XBOX 360 is a good way to get a gaming platform, media, and such for much cheaper than a full computer build.

If you have any questions... just ask. Many a person around here will likely be happy to answer.

 

[clarkey01]

We must congratulate AMD for giving it's best. If anyone on this board feels they could of done better in the last 3 years then it's time for a reality check.

It's very well someone on a board saying " AMD sux, they did nothing for 4 years, borrrring" . I fully believe they have tried there best, why wouldn't they ? If they come short then why is there an outcry and teasing? Intel fans do know there beloved company will only raises prices if AMD failed?

What AMD need is someone with big pockets and knowledge of the industry....IBM