Intel is in major trouble!!!

[NetRyder]

I meant their current generation of CPUs was in trouble of being out paced by AMD's current offerings.

And I completely agree with that statement.

 

[ElementalDragon]

well.... from the fingertips of neither an AMD or an Intel Fanboy (since my desktop is a P4 2.4GHz w/hyper threading and my laptop is an Athlon XP), unless software starts to severely back up the Athlon 64's 64-bit'ness, i don't think there will be any overkill between the two. to me, what's the point of buying a 64-bit processor when the only OS that can utilize the major portion of the CPU is still pretty much a public beta release. I don't even think manufactured computers that ship with an Athlon 64 come with Windows XP 64-bit edition.

 

[Sazar]

elemental.. you miss the big picture... its not an out and out 64bit processor.. its an x86-64 processor ie it does BOTH 32bit AND 64bit... and it does 32bit FASTER than todays regular 32bit processors...

so regardless of whether there are 64bit apps to take advantage of the architecture or not... the processors are simply faster than intel's offerings...

 

[Sazar]

I didn't mean that Intel was in trouble as a company, I was so Naive to think that they would go bankrupt, I meant their current generation of CPUs was in trouble of being out paced by AMD's current offerings.

their future offerings might be too since prescott IS their 'next-gen' cpu...

intel may want to rethink its mhz is GOD strategy... their M's are a fine product and they don't advertise em based on speed because clock for clock they do more work than the northy's and consequently their prescott's and eventually tejas...

 

[Goatman]

I know if I buy a laptop it'll be either a iBook, or something with a Pentium-M

 

[LeeJend]

Intels heat problem is due to dielectric loss (leakage) because the silicon on silicon transistor design did not work out at expected at the small geometry. The parasitic capacitance between the junctions and substrate is directly proportional to frequency so at the high clcok speed Intels uses it suffers from a bad process choice. Using silicon on insulator solves that problem.

AMD chose the silicon on insulator approach because they saw what happened with Tbred. Intel had to make their own mistake before they will switch processes. Figure 12 months to retool and develop their in house process steps, 18 if they wait until the 60nm process and effectively bypass 90nm. Then Intels heat problem will be history.

Intels next step will be to put 2 parallel processors on one die at 4 gig clock speed. Probably when they go to 60nm. That will give something like an XP7000 rating. Throw in an x86 instruction set and 64 bit OS (someday) and you're looking at an XP10000 chip rating. AMD has no clear path to those kinds of speeds.

Besides my premise was that the "AMD killer" would be a killer based on price. Performance will be about the same. AMD can not compete in a price war anymore. Their 3 Billion dollar cash reserves were almost completely depleted and the tiny profit they are now earning after 3 years of billion dollar losses is too small to replace their capital. AMD was within 6 months of bankruptcy, selling at a couple bucks a share. The only thing that saved them was flash memory sales. The flash market is now saturated and prices are dropping horribly.

Performance wise AMD is still saddled with their clunky 333-400 mHz FSB. I just switched to dual DDR and have seen little improvement in memory bandwidth due to that slow FSB. (Yes, I confirmed it is running in dual DDR mode.) AMD needs to find a way to double FSB speed. That path is not obvious using Complex Instruction Set Architectures.

AMD can't die out? Remember the quotes form the last 5 years:
Voodoo will always be king.
Nvidia will always be king.

Big companies are cursed by the inability to change once committed to a path. It would require people who make millions a year saying they screwed up. If they said that they would be replaced. So the lies go on, until the company folds or the board of directors steps in and starts firing people.

 

[NetRyder]

elemental.. you miss the big picture... its not an out and out 64bit processor.. its an x86-64 processor ie it does BOTH 32bit AND 64bit... and it does 32bit FASTER than todays regular 32bit processors...

so regardless of whether there are 64bit apps to take advantage of the architecture or not... the processors are simply faster than intel's offerings...

Well, the question is does it really matter in the real world if one processor is slightly faster than the other? Don't you think CPU speeds are already high enough for most purposes already? How often does one make use of their processor's full potential anyway? The real world performance gains today with the a64 are really not that noticeable to justify the upgrade from other processors running at similar clock speeds. Same is the case with HT. They're simply not being used to their full potential.

There are other bottlenecks that need to be addressed before we can see a truly significant performance leap. Disk I/O is one big factor. No matter how fast processors get, there is always going to be that upper bound which won't be broken until these other factors are taken care of.

 

[LeeJend]

Quote
Intel may want to rethink its mhz is GOD strategy... their M's are a fine product and they don't advertise em based on speed because clock for clock they do more work than the northy's and consequently their prescott's and eventually tejas...

One of my other points. Intel already has anything they need already developed and sitting in the closet ready to roll out. The reason they don't is because the industry business model says they will make more profit on the path they are on.

 

[Sazar]

amd has a worse dual channel implemention than intel wrt socket A.. so going dual channel you will likely not see any improvements if any @ all...

a64 has on-die memory controller which is where most of the boost in performance comes from... therefore the whole memory clunky thinger is a thing of the past... the only problem for amd is how they will be able to ramp it up

currently @ 800mhz... will improve to 1000mhz with future iterations end of year/early next year... perhaps earlier...

amd's has new deals in germany and they are GETTING money to stay there... they have debts but they have changed around their business policy big time... remember the highest margins are in the server business where amd has made massive inroads offering a faster/better/cheaper solution that is better in upto 4 way configs than intel's solutions for the most part... THAT is where amd is going to get its money from mainly...

 

[LeeJend]

Are comps fast enough?

-Gaming PC's are finally adequate right now XP3000 and a 9800 PRO/XT.
-My PC at work (P4-2.6 gig) is way too slow for the analysis work I do.

Modern development is based on simulation to reduce cost. The models are having to get bigger and bigger which chokes the life out of the computers. Most engineers, programmers, biotech people spend 80% of there time sitting around waiting for a simulation to run. Look at the protein folding project we're sponsoring here at NTFS. Those tiny sections of the overall project take hours or days to run on our souped up gaming machines.

An XP10000 would be a good start!

 

[LeeJend]

Woah. AMD is running 800 mHz FSB on the 64 bit proc's? I missed a press release somewhere. Got a link?

Or is that 800 only on the, now internal, memory interface?

 

[Sazar]

hypertransport == 800... internal... bypassing lots of little log-jams that existed before...

 

[LeeJend]

Or is the 800 Mhz just taking credit for doubling the 400 MHz 32 bit path to 64? In which case Intels x86 chip will have a memory FSB of 1600-2000 Mhz.

Mark Twain said there's three kinds of lies:
Lies.
Damn lies!
And statistics!!!

Time for me to reread hypertransport specs.

This has been a fun thread.

 

[Sazar]

Or is the 800 Mhz just taking credit for doubling the 400 MHz 32 bit path to 64? In which case Intels x86 chip will have a memory FSB of 1600-2000 Mhz.

Mark Twain said there's three kinds of lies:
Lies.
Damn lies!
And statistics!!!

Time for me to reread hypertransport specs.

This has been a fun thread.

intel is already claiming qdr 200 == 800... I fail to see how this can get any higher

 

[LeeJend]

When its 32 bits wide quad pumped 200 mhz you call it 800. If you double the data path to 64 bit then you move twice the data and call it 1600. The pin count is a killer though.

The next gen AMD sockets are almost a thousand pins. That's scary!

 

[MargaritaIce]

and btw the Prescott core makes big improvements to HT-ing.

Im gonna get the 3.4 Extreme Edition and you all can tell me how crappy it is, then

 

[Sazar]

the improvements to HT will still have to be programmed for m8

multi-threaded apps == boosted performance with HT... but day to day apps ? office applications and the like are MUCH faster on an amd a64 system though the extra l3 cache for the EE will improve performance for you...

still.. I am quite happy with my 800mhz beatie

 

[Goatman]

Prescott is currently on a 90nm process, and it runs hotter than Northy

There is no evidence of any of Intel's future. This is all speculation, Hyper-Transport, Athlon 64 speed bumps, and what not are all there. They have it at FX-53 and still no plan on it slowing down.

The EE is a huge waste of money, especially when you can buy a FX-51 for half the price. and the Athlon outperforms it.

 

[ming]

Just one sentence.
Intel needs more cycles to complete a process than AMD. That's why you may have noticed that an AMD 3200+ (2.4GHz) is equivalent to an Intel P4 3.2GHz.

 

[Sazar]

the intel 800mhz cpus actually do pretty well v/s amd's socket A based solutions that are expected to compete against them... with the B processors it was a tougher job but the C's perform better mainly due to improved fsb... IF you compare the same northy's v.s one of the 64bit solutions though you will find the scenario to be different..