Dell Set to Introduce AMD's Triple-core Phenom CPU

[safan80]

http://hardware.slashdot.org/hardware/08/02/17/0149207.shtml


AMD is set to launch what is considered its most important product against Intel's Core 2 Duo processors next week. TG Daily reports that the triple-core Phenoms — quad-core CPUs with one disabled core — will be launching on February 19. Oddly enough, the first company expected to announce systems with triple-core Phenoms will be Dell. Yes, that is the same company that was rumored to be dropping AMD just a few weeks ago. Now we are waiting for the hardware review sites to tell us whether three cores are actually better than two in real world applications and not just in marketing

[spicypixel]

Guess this is a good thing, at the end of the day I think dell disables most overclocking options in the bios anyway so to the end user who in all fairness is buying dell - wont be overclocking much. 2.2 ghz tricore might sway someone from a 2.4/2.6ghz dual core with some persuasive advertising, and it should be priced around the £80-90 mark so it might just pay off.

[eric66]

'Yes, that is the same company that was rumored to be dropping AMD just a few weeks ago'

:ROTF:

[FischOderAal]

http://www.tgdaily.com/content/view/36092/135/

more interesting :p

[onewingedangel]

so the dies costs the same to make as a quad (granted many are probably failed quads), and their priced up against and will perform on par with intel's duallies, which cost less than half to manufacture.

And you wonder why AMD are in trouble.

[squilliam]

not necessarily. These are probably all bad quads. It gives them a chance to recycle, and spend less, because they don't have to figure out how to make a new processor. I'm sure Intel would love to sell all the quad cores that only have 2/4 cores working. I'm sure they'd sell it as a special edition or something.

[Shintai]

not necessarily. These are probably all bad quads. It gives them a chance to recycle, and spend less, because they don't have to figure out how to make a new processor. I'm sure Intel would love to sell all the quad cores that only have 2/4 cores working. I'm sure they'd sell it as a special edition or something.

Intel only have 100% working quads. I´m not sure if you are familiar with the MCM design. Intels quads are made afterwards of 2 working dualcores put together.

[Jowy Atreides]

so the dies costs the same to make as a quad (granted many are probably failed quads), and their priced up against and will perform on par with intel's duallies, which cost less than half to manufacture.

And you wonder why AMD are in trouble.


not necessarily. These are probably all bad quads. It gives them a chance to recycle, and spend less, because they don't have to figure out how to make a new processor. I'm sure Intel would love to sell all the quad cores that only have 2/4 cores working. I'm sure they'd sell it as a special edition or something.

he said that they were failed quads,
failed or not it still costs the same to produce

and that's why amd need help, if they could reliably manufacture quads then they wouldn't need to sell the failure for less profit

[[XC] Lead Head]

All semi conductor manufacturing have failure rates. AMD has also been known to have fairly good yield rates as well. Having the tri-core CPUs allows them use CPUs that would normally be thrown out. Making them MORE money. No matter which way you try and spin it, reselling broken quad cores as tri cores will result in more money.

Anyways, AMD has done this before when they were back in the green as well. With dual core die's having one core disabled, and being sold as single cores

[RaZz!]

so the dies costs the same to make as a quad (granted many are probably failed quads), and their priced up against and will perform on par with intel's duallies, which cost less than half to manufacture.

And you wonder why AMD are in trouble.

no big deal, as i think the tri-core cpus are quads with a bad core disabled. so they actually save money, because they use failed quads (which else would be trashed...).

nothing changes for amd: two production lines (duals & quads) which would exist with or without tricores. amd just adds a manufacturing step to put quads with one bad core in a tricore package :p:

[v_rr]

he said that they were failed quads,
failed or not it still costs the same to produce

and that's why amd need help, if they could reliably manufacture quads then they wouldn't need to sell the failure for less profit

Well Nvidia and ATI do this for a long long time and it´s money that doesn´t go to the garbage. (8800GT is a failed 8800GTS)

Making tri-core from a failed quad core means money in the pocket and in market means a new segment that can became very profitable to AMD.

You know that by Q4 2008 only 15-25% of all CPU´s will be quad-core? All others are dual, tri, and single core.

This tri-core is a god way to open a new segment and sell lots of CPU´s. And specialy to bring dual-core users to future tri-core users with AMD processors.

[Shintai]

Lead Head;2777408']All semi conductor manufacturing have failure rates. AMD has also been known to have fairly good yield rates as well. Having the tri-core CPUs allows them use CPUs that would normally be thrown out. Making them MORE money. No matter which way you try and spin it, reselling broken quad cores as tri cores will result in more money.

Anyways, AMD has done this before when they were back in the green as well. With dual core die's having one core disabled, and being sold as single cores

What happens when AMD run out of tricores and Dell wants more to satisfy demand?

AMD need a real tricore aswell if they dont already have it. Else its logistic suicide. And thats something they would lose Dell for.

It is fine to mix some failed quads into the product mix to help on the losses. But to ahve a product solely dependent on the failure of another is not good. So I am sure we will soon if not already see the "true" tricore.

[Vapor]

Lead Head and RaZz, quads with one bad core would not just be trashed....just disable another core and bam, dual core.

Tri-cores are a way to get $15 more for those CPUs.

[v_rr]

What happens when AMD run out of tricores and Dell wants more to satisfy demand?

AMD need a real tricore aswell if they dont already have it. Else its logistic suicide. And thats something they would lose Dell for.

It is fine to mix some failed quads into the product mix to help on the losses. But to ahve a product solely dependent on the failure of another is not good. So I am sure we will soon if not already see the "true" tricore.

8800GT is a failed 8800GTS, and 8800GT is in a market of price that sells much more than 8800GTS market price.
And now what? Is Nvidia suiciding?

No... AMD just need to have their fabrics on top of production. If market don´t want more Phenom X4, they sell Phenom X3. Simple. Makes money and doesnt have machines stoped.

The key here is increase market share and open new segment with tri-core. You have "n" examples in the past of products with parts desactivated.
Remember X800 PRO mod to have XT pipes;
Athlon XP mod to double the cache to Barton, etc etc
In Nvidia you have also many examples of unlocking mods.

[Shintai]

8800GT is a failed 8800GTS, and 8800GT is in a market of price that sells much more than 8800GTS market price.
And now what? Is Nvidia suiciding?

No... they just need to have their fabrics on top of production. If market don´t want more Phenom X4, they sell Phenom X3. Simple. Makes money and doesnt have machines stoped.

8800GT is not just failed GTS as such. They are also different speedbin. So you can easily have GTS class GTs, just with a lasercut. It would compare to AMD intentionally destroying 1 core on a quad to make a tricore.

The difference between 8800GT and 8800GTS is very small. Its 1 shadercluster out of 10 Thats maybe 2-3% of the chip. Not 15-20%.

8800GT vs 8800GTS is more like a 2.2 vs 2.3Ghz Quad.

But maybe AMD is willing to sacrifice working quads to forfill the tricore demand. I dont know.

[Ugly n Grey]

What happens when AMD run out of tricores and Dell wants more to satisfy demand?

AMD need a real tricore aswell if they dont already have it. Else its logistic suicide. And thats something they would lose Dell for.

It is fine to mix some failed quads into the product mix to help on the losses. But to ahve a product solely dependent on the failure of another is not good. So I am sure we will soon if not already see the "true" tricore.

I get what you're thinking here Shintai, but the agreements are usually structured around fairly fixed numbers and I doubt Dell is putting themselves in a position of demanding more than AMD can supply. Dell is probably the best in the world at supplier management and they have the track record to back that up.

As to your other thought, AMD can make a tri core as a non failed quad fairly easily, in fact from a design process, nothing is stopping them from etching three cores into the same space they normally etch four. The question is, why? I imagine it depends on the failure rate of quads and I have no knowledge there. I guess they could make a true Tri-core that utilizes only a three core footprint, but I imagine things like thermal management post packing might get tricky. I'm not convinced they will do this personally and I've seen nothing.

Mind you dollar is king and if tri-cores sell, I bet they make one on 45nm ASAP.

[v_rr]

8800GT is not just failed GTS as such. They are also different speedbin. So you can easily have GTS class GTs, just with a lasercut. It would compare to AMD intentionally destroying 1 core on a quad to make a tricore.

The difference between 8800GT and 8800GTS is very small. Its 1 shadercluster out of 10 Thats maybe 2-3% of the chip. Not 15-20%.

8800GT vs 8800GTS is more like a 2.2 vs 2.3Ghz Quad.

But maybe AMD is willing to sacrifice working quads to forfill the tricore demand. I dont know.

What about 8800GS? Is in a segment that sell lot more then 8800GT and 8800GTS and have lots of parts desactivated.

Nvidia suiciding again? They are using G92 from 350$ to 150$. Drastic isn´t it ;)
RV670 is being used from 450$ (3870X2) to 130$ (HD 3850).

Well his is only a very simple and actual example. You have lots of examples like this in history.

Again the Mods to grafics and CPU´s that unlock new functions. X800, Athlon XP, etc
The objective is always sell the best part. If the best part don´t have more demand pass to lower parts.
And all in all have always the fabric working at 100% capacity.

[Shintai]

What about 8800GS? Is in a segment that sell lot more then 8800GT and 8800GTS and have lots of parts desactivated.

Nvidia suiciding again? They are using G92 from 350$ to 150$. Drastic isn´t it ;)
RV670 is being used from 450$ (3870X2) to 130$ (HD 3850).

Well his is only a very simple and actual example. You have lots of examples like this in history.

Again the Mods to grafics and CPU´s that unlock new functions. X800, Athlon XP, etc
The objective is always sell the best part. If the best part don´t have more demand pass to lower parts.
And all in all have always the fabric working at 100% capacity.

You do know you are talking about different cores? And 2 diesizes? There is also 2 versions of GS.

Also your comparision is fataly flawed since these are usually done in speedbins and with INTENTIONALLY lasercuts. 2-3% is one thing, 15-20% loss is a whole other. Also nVidia earns ALOT more than AMD does on their cores.

[Shintai]

I get what you're thinking here Shintai, but the agreements are usually structured around fairly fixed numbers and I doubt Dell is putting themselves in a position of demanding more than AMD can supply. Dell is probably the best in the world at supplier management and they have the track record to back that up.

As to your other thought, AMD can make a tri core as a non failed quad fairly easily, in fact from a design process, nothing is stopping them from etching three cores into the same space they normally etch four. The question is, why? I imagine it depends on the failure rate of quads and I have no knowledge there. I guess they could make a true Tri-core that utilizes only a three core footprint, but I imagine things like thermal management post packing might get tricky. I'm not convinced they will do this personally and I've seen nothing.

Mind you dollar is king and if tri-cores sell, I bet they make one on 45nm ASAP.


En, if I am not mistaken AMD got a real tricore die. And also a dualcore without L3. Its simply a too big market since tricores would quickly outsell quads.

[v_rr]

You do know you are talking about different cores? And 2 diesizes? There is also 2 versions of GS.
Also your comparision is fataly flawed since these are usually done in speedbins and with INTENTIONALLY lasercuts. 2-3% is one thing, 15-20% loss is a whole other.

8800GS is a G2 cuted all arround. IMO.
First review of G92 core 8800GS
http://en.hardspell.com/doc/showcont.asp?news_id=2159

The 8800GS is a 15-20% cuted full G92. cuted shaders, less bus, cuted ROP and TMU´s.

G92 is sold from prices 350$ to 150$, and the 150$ price segment sell much much more witch mean that many many cores are proposly desactivated witch is a loss and suicide by your prespective.

This only prove that your teory is simply wrong.
AMD can pick up a good quad-core and desactivate 1 core like Nvidia does with G92 (8800GTS, 8800GT, 8800GS).
And still win lot of money because volume.


Also nVidia earns ALOT more than AMD does on their cores.

Realy? Last time I checked in Q4 2007 Nvidia margins where 45% and AMD margins 44%. Wow, thats realy ALOT :ROTF:

[Shintai]

8800GS is a G2 cuted all arround. IMO.
G92 is sold from prices 350$ to 150$, and the 150$ price segment sell much much more witch mean that many many cores are proposly desactivated witch is a loss and suicide by your prespective.

This only prove that your teory is simply wrong.
AMD can pick up a good quad-core and desactivate 1 core like Nvidia does with G92 (8800GTS, 8800GT, 8800GS).
And still win lot of money because volume.



Realy? Last time I checked in Q4 2007 Nvidia margins where 45% and AMD margins 44%. Wow, thats realy ALOT :ROTF:

Jeebus...Its like comapring a QX9770 with a Q9450 in your scenario..seriously.

Also GS is not the same, GS is G92-100 cores. And compare the disabling of 1 shader cluster with one core is nuts. If they removed 25% of the ROP, TMU and shadercluster..then sure

AMD made how much money? While NVidia earned 257million in profit for alittle less revenue in Q4.

nVidia just earns more money, period.

[v_rr]

Jeebus...Its like comapring a QX9770 with a Q9450 in your scenario..seriously.

Also GS is not the same, GS is G92-100 cores. And compare the disabling of 1 shader cluster with one core is nuts. If they removed 25% of the ROP, TMU and shadercluster..then sure
8800GS is one G2 with less shaders, less ROPs and less TMU´s. So yes in the final you will have more then 25% of the chip desactivated.



AMD made how much money? While NVidia earned 257million in profit for alittle less revenue in Q4.

nVidia just earns more money, period.
Now you change to talking about profit/loss because you notice that margins are equal :rolleyes:

Nvidia don´t earn more mony from each chip. It earns 1% more. That´s realy alot :yepp:

Profit/losses is another cenario that don´t have nothing to do with this and don´t interest to bring here.

[mnewxcv]

question is are the tri cores almost as good as quad? should be pretty close i would think, especially in programs optimized for dual core cpus.

[Shintai]

8800GS is one G2 with less shaders, less ROPs and less TMU´s. So yes in the final you will have more then 25% of the chip desactivated.


Now you change to talking about profit/loss because you notice that margins are equal :rolleyes:

Nvidia don´t earn more mony from each chip. It earns 1% more. That´s realy alot :yepp:

Profit/losses is another cenario that don´t have nothing to do with this and don´t interest to bring here.

Still 2 chips, not one :rolleyes:

Margin is only calculated on chips sold. In short, nVidia dont waste as much as AMD. That again leads us back to the 2-3% vs 15-20% :yepp:

1% is still alot if its what makes you lose or earn money when you dont have any ;)

[Nedjo]

I dont know.

OMG universe is about to colapse! :eek:

[KoHaN69]

So what happens to Intel's failed dual-cores? :shrug:

[zerazax]

become Celeron's?

[zerazax]

Anyways, AMD really needs the OEM contract to get things rolling again, and if Tri-Core performance can beat out a dual core, then it will help competition certainly. Has anyone actually tested/benched these yet or are we all waiting for them to hit with Dell first?

[KoHaN69]

become Celeron's?

seems so

stupid fud (http://www.fudzilla.com/index.php?option=com_content&task=view&id=3089&Itemid=1)

[Donnie27]

So what happens to Intel's failed dual-cores? :shrug:

I don't know, why would Intel use a Single Core plus a Dual core on an MCM anyway?

IMHO, this question will be answered once Nehalem ships. As it stands right now, MCM is far more efficient to produce and has a larger profit margin. A failed core becomes a Single Core. If part of the single fails it becomes a Celeron.


Lead Head and RaZz, quads with one bad core would not just be trashed....just disable another core and bam, dual core.

Thanks Vapor and 3 bad cores becomes a single. The only thing trash are dead useless CPUs. Also note that that 4th core don't have to truly be bad. It could be 2 or 3 hundred MHz slower than the other 3.

Last be not least, *needing 3 to beat 2 will be a marketing nightmare. All Intel has to do is stick a couple of Budget Quads in that segment and yikes!

[initialised]

he said that they were failed quads,
failed or not it still costs the same to produce

and that's why amd need help, if they could reliably manufacture quads then they wouldn't need to sell the failure for less profitI don't think you understand the distribution of defects across a silicon wafer. The larger the die the higher the probability that any given die on the wafer will have a killer defect or fall off the edge of the wafer. But if your die is divided into equivalent cores which can be disabled independently then instead of being a failed die it is simply a failed core so you lose less. As far as I can tell this has always been part of AMD's strategy with Phenom and it makes perfect sense in an homogeneous multiprocessor die as it allows a profit to be made from what would otherwise have been a failed die. I suspect other manufacturers do the same (G80s and G92s with 128, 112, 96, 64, 32 Stream processors where one or more bank or 16 has failed?)