AMD prepares three-core processors

[hollo]

i wouldn't go so far as to call an AMD tri-core damage control - that's like saying selling a lower speed CPU is damage control, when really it's an expected result of the randomness of the manufacturing process - some processors will clock a lot further than others, so you have a mixture of higher and lower clocked parts.

similarly if you have a quad core die with a defect on one core, disable that core and sell the other three, its good business sense. one core is only 25% of the total yield (well less than 25% if you take the l3 cache into consideration), which is less than the clockspeed difference between the lowest and highest binned parts in a typical processor series (eg 36% speed difference between e6320 and x6800).

AMD are gonna produce a lot of quad cores, so i'm glad they won't discard 3 perfectly good cores because of a random manufacturing defect in the 4th one. if it was up to me the ability to disable a core would be a BIOS option, cos a quad core will only overclock as fast as its weakest core,

[terrace215]

The number of desktop apps for which 3 cores is optimal has to be very small.

If the intention is a higher-clocking, lower-power part (than a quad core) because App X doesn't use all 4 cores, chances are that a DUAL core part, which can clock even higher @ lower power would be even better than these hypothetical Tri-cores. The only case in which this isn't true is that of applications that really split into 3 parallel workloads.

As for a new layout vs. disabled QC parts, they surely have a lot of QCs for which one core is bad / power-hog / low-clocker. Seems odd that they would junk them, and if they are using a new layout for a 'native' 3-core part, they would then be selling the imperfect QCs as DCs if they aren't using them for TC and aren't throwing them away. Seems unlikely.

So, in summary, I don't think there's much of a viable Tri-core segment in between DC and QC, BUT if they ARE doing this, I would expect them to use the imperfect QCs to do it, not make a new part.

If K10's individual core speed/power bin variation (incl flat-out defective cores) is bad enough, they need to do something, but I'm surprised they don't simply turn them into DC parts, rather than try for this strange middle segment that isn't optimal for much. At first glance, one might say, "well you can get more $ for a tri than a dual", but at what costs to QC & DC pricing and clarity of segmentation?

[Donnie27]

Seriously, if you are to troll. Do it better. Making a tricore when you got a quad is the easiest task in the world. But why, it makes no sense when you got the MCM. yield are massively higher on dualcores than on quads. And since Intels CPUs consist of much more cache thats more resiliant due to extra cache cells can make up for cache defects. Its just not logical.

Tricores is only a desperate move to salvage parts of the huge mistake with the native quad.

So perhaps you should cut the crap and just realize AMD is up to crap to its neck with this native BS.

Absol-phookin-lutly QFT! LOL! Its a reach at best on either company's part to down bin failed products. With PR trying to say its good is a Joke.

I got two of Intel's BS "Intel Web Outfitter" tool kits discs, oh brother Liars. I saved them to prove a point to my Intel only Buds. These folks (AMD and Intel) make Used Car Salespersons look Honest I say Persons because one of our worse ones here locally is female. On Intel's part, an even worse one was the whole; For Gamers it's so Fast it's unfair BS when Athlon64 was much faster and was easily the best processor for Games and or Gaming rigs.

Sad thing with Trolls though. You can post 9 things Intel did wrong, one they did good and you're still flamed for the one and called a Fanboy.

[safan80]

New board rule....

If you want to see AMD threads without the crap, just add a handful of people to your ignore list. The threads look absolutely amazing without the BS in them.

that is soo true.

[Donnie27]

i wouldn't go so far as to call an AMD tri-core damage control - that's like saying selling a lower speed CPU is damage control, when really its an expected result of the randomness of the manufacturing process - some processors will clock a lot further than others, so you have a mixture of higher and lower clocked parts.

similarly if you have a quad core die with a defect on one core, disable that core and sell the other three, its good business sense. one core is only 25% of the total yield (well less than 25% if you take the l3 cache into consideration), which is less than the clockspeed difference between the lowest and highest binned parts in a typical processor series (eg 36% speed difference between e6320 and x6800).

AMD are gonna produce a lot of quad cores. if it was up to me the ability to disable a core would be a BIOS option, cos a quad core will only overclock as fast as its weakest core,

Actually that's called speed binning. Intel speed bins two Dual cores before there's a Quad Core before they make a MCM. Now it would be pretty stupid to use one with one of the Cores dead, wouldn't it? They throw it in the pile meant for Core Solo and sell it as such. One bad core means AMD is now selling 3 instead of 4. Everyone knows this is bound to happen (any one in the Fab business has failed parts=P). That AMD's selling them is not drawing evil looks, it's their trying to say Hey WOW here's something great Spin!

Intel has far more control over MCM than AMD does with its Monolithic design, just common sense here. Even AMD acknowledged as much.

[hollo]

AMD's selling them is not drawing evil looks, it's their trying to say Hey WOW here's something great Spin!

agree with you, but where does that ^ come from?

[adamsleath]

Gotta love PR these days.

yes it's all bs...

but it is possible nonetheless.

i doubt there will be a tricore....

but if programmers actually utilised cores properly 3 should be betta than 2.

...and if priced properly i still reckon youd sell some - let's see

Intel AUD:
q6600 = $330-350ish
e6600 = ~$300ish (lol)
e6420 = $240ish
e4400 = $150ish
e2160 = $100ish

AMD AUD:
3600 sempron = $80
5000+ = $160ish
5200+ = $165
5600+ = $190
6000+ = $200ish
.....................................->look at this gap: between 200 and 330aud <-
6400+ = $330ish
Phenom quad..... $330+?????????????? no idea

not having seen any sales figures but arent the majority of sales coming from dual core chippies?

absolutely no way are quad cores mainstream and they wont be until there are more models available at lower prices....and spare me the q6600 sales pitch.
http://www.channelregister.co.uk/200...on_quad-cores/ - July...amd sure are behind...

http://www.efluxmedia.com/news_AMD_L...ast_08489.html

http://www.reghardware.co.uk/2007/04...ream_until_09/ blah just forecasts cant find real figures regarding quads - to - dual &#37;ages...

Mainstream quad-core boxes will move more slowly, accounting for just five per cent of the systems shipped in Q4

multi-threaded games

That's one reason why, come Q4, quad-core systems will account for 33 per cent of PCs sold during that quarter, according to iSuppli.

[[XC] gomeler]

Betting this will just be a Phenom x4 with a core disabled, that way they can salvage some of the cost of building up that silicon. Makes perfect sense, would be even better if they completely knocked off their dual-core lines and replaced them with tri-cores, could brag about a greater computing density versus core 2 duo at lower clocks and equivalent TDP(maybe?).

[awdrifter]

They should keep the dual-cores for the lower end to compete with the E4xxx and E2xxx series though.

[Donnie27]

agree with you, but where does that ^ come from?

Just happened

[lapdog]

If these are native three cores a MCM may fit in the socket. Six core would be a nice first.
If these are just one core with a defect disabled it's still all good, AMD can sell all the more for higher yields.

[clayton]

If its lower than Quadcore, expensive than Dualcore by not much. All I can say is.... hell yeah!

[Trouffman]

tri-core can really be interesting oin desktop configuration but I don't understand well why they made this :/

[Thesavage]

Some more info here,
http://www.planet3dnow.de/vbulletin/...11#post2893311

[MAS]

Some more info here,
http://www.planet3dnow.de/vbulletin/...11#post2893311

what is he speaking about?

[MAS]

though the main idea is:

27,3&#37; defekt (55/Wafer)
0,4% Single Core (fast 1/Wafer)
4,6% Dual Core (9/Wafer)
23,2% Triple Core (47/Wafer)
44,5% Quad Core (90/Wafer)

47 Phenom X3 chips per 1 wafer! it's enough for mass production/

[STaRGaZeR]

though the main idea is:



47 Phenom X3 chips per 1 wafer! it's enough for mass production/

And enough to see that AMD&#180;s native quad-core manufacture process is broken. If they don&#180;t sale those faulty quad cores as triple, they will loose even more money than now.

[[XC] gomeler]

Oh god I hope their yields aren't really that low. The number of completely defective chips is disheartening, the fact that they can't even salvage a single core out of the chips does not speak well of their 65nm process combined with the new core.

[Cooper]

Oh god I hope their yields aren't really that low. The number of completely defective chips is disheartening, the fact that they can't even salvage a single core out of the chips does not speak well of their 65nm process combined with the new core.

The manufacturing process is not tweaked to get lots of single-cores

[alayashu]

i'm sure the yields are better now than how they were last year.

[Face]

AMD to launch triple core on Monday

San Francisco: A day before IDF 2007

Once San Francisco and New York wakes up AMD on Monday, AMD will announce its triple core X3 Phenom CPU. It will be the first triple core CPU for X86 but we've seen this concept with Xbox 360 IBM chip. We wrote about this chips just hours ago here and in the meantime we managed to confirm that they are real.

FudZilla

Well.. that should be in about........................ now. :p:

Just few more hours to see if the rumor is true or just the usual *.

:shrug:

[[XC] gomeler]

The manufacturing process is not tweaked to get lots of single-cores

true but notice the enormous (1/4) of the chips that can't even be salvaged for a single core, that is what I was pointing out.

[Cooper]

true but notice the enormous (1/4) of the chips that can't even be salvaged for a single core, that is what I was pointing out.

Agreed, but than again this 25% part may have memory controller, crossbar or HT defects.

[Cooper]

FudZilla

Well.. that should be in about........................ now.

Just few more hours to see if the rumor is true or just the usual *.

There is gonna be Press Release regarding triple-core CPUs in Q1 2008.

FUD as usual

[Shintai]

Those stats dont look right. Specially with the fully defect compared with the others. And there seems to be artificially high number of tricores compared to fully defect and the others.