AMD Magny Cours CPU-Z Validation

[btarunr]

Here's the first CPU-Z validation of AMD's 12-core Magny Cours processor. Whatever details the existing version of CPU-Z does read, perfectly matches the specifications of the processor sketched out so far. Firstly, it's based on AMD's upcoming socket G3 package that marks Opteron's transition to high-level integration within a single package. With 1,974 pins, socket G3 is able to provide as many as six 16-bit HyperTransport 3.1 links, and four DDR3 memory channels. The package is one of AMD's first multi-chip modules, that houses two six-core dies (dubbed "nodes"), onto one package, and connects the two using a HyperTransport link.

Each node has 6 x 512 KB of L2 cache and 6 MB L3 cache shared between the six cores. Out of 6 MB, 1 MB of the cache is reserved for low-level system operations, namely the HT Assist (probe filter) that aims to lower memory subsystem latencies, reduces queuing delays due to lower HyperTransport traffic overhead, and minimizes probe traffic to increase system bandwidth. The CPU-Z reading of 10 MB total chip L3 cache is spot-on. Also seen on the validation page are details on the reference motherboard, called "AMD Dinar", that uses SR5690 (same chip as 890FX) + SB750 chipset. The CPU-Z validation can be found here.



View at TechPowerUp Main Site

 

[btarunr]

Many Thanks to KitKat for sending this in.

 

[FlanK3r]

And what do u think about this

If u not know, what is it, it is too Magny Course ES in benchmark...

 

[DarkOCean]

What a monster,also very low voltage!

 

[FlanK3r]

What a monster,also very low voltage!

and score....more better then 2x Nehalem at the same clock!

 

[FreedomEclipse]

so it uses DDR3 but its still under 200FSB???

 

[thezorro]

wow this is a real monster.
intel is finished.

just my two cents.

 

[btarunr]

so it uses DDR3 but its still under 200FSB???

That 200 MHz is base frequency. This one's system interface is HyperTransport 3.1 6400 MT/s.

 

[wolf]

and score....more better then 2x Nehalem at the same clock!

Could that maybe be due to 12 cores instead of 4 with HT?

 

[btarunr]

Could that maybe be due to 12 cores instead of 4 with HT?

(If and) when it is able to sell for the same prices as competing Nehalem chips, it becomes irrelevant how many AMD cores are pitted against Intel cores.

 

[thezorro]

Could that maybe be due to 12 cores instead of 4 with HT?

no, 2x nehalem are 8 intel cores with hyperthreading

 

[Assimilator]

So after spending a year blathering about how Intel's quads aren't "true" quad-core designs, AMD goes and pulls the same trick with Magny-Cours. GG AMD, keep innovating!

As for the benchmarks, AMD's ES chips always do well while their production parts consistently fail to live up to the same level, and I doubt it'll be much different here. Besides, wPrime is one benchmark.

 

[InTeL-iNsIdE]

its a server chip anyway, this isnt a desktop processor, nehalem is enthusiast but still desktop, although there are the server variants aswell in the zeon series

Im betting this will be 4x the price of a nehalem anyway

Any ideas whether these can be used in multi socket ?

 

[FordGT90Concept]

Only 800 MHz? That's lousy. :|

 

[[I.R.A]_FBi]

 

[btarunr]

So after spending a year blathering about how Intel's quads aren't "true" quad-core designs, AMD goes and pulls the same trick with Magny-Cours. GG AMD, keep innovating!

Yeah, try making a 700 mm², 1808 million transistor die on 45 nm SOI. Back when Intel made its quad-core chips, something twice as big as Conroe was possible on 65 nm /45 nm HKMG. Intel's quad-core design ensured it just has to make one kind of dies (Conroe/Penryn) and then package them differently for dual-core/quad-core chips. With Magny Cours, it's the constraints of the manufacturing technology that's making 12-core a monolithic die impossible.

 

[wolf]

no, 2x nehalem are 8 intel cores with hyperthreading

still only 8 cores + HT still at a core disadvantage.

and I don't believe that screen shot at all.

 

[thezorro]

12 cores magny cours at 3.2 ghz
intel is finished.

just my two cents.

http://www.tweaktown.com/news/13115/another_cpuz_of_magny_cours/index.html

 

[Airbrushkid]

So when are we going to see 3.0 gig per core and like a board that's supports 94 gig that's 8.0 per core and one last thing 10 pci xpress slots so we can put 10 GT300 video cards in it. Now that be a monster!

 

[wolf]

Sell a retail 12 core chip and we'll talk. two 6 cores just isn't the same.

 

[mdm-adph]

Only 800 MHz? That's lousy. :|

It's obviously running on Cool'n'Quiet, silly.

12 cores magny cours at 3.2 ghz
intel is finished.

just my two cents.


http://forums.techpowerup.com/attachment.php?attachmentid=28296&stc=1&d=1252414083

http://www.tweaktown.com/news/13115/another_cpuz_of_magny_cours/index.html

No, Intel is not finished, nor would I want them to be. If this makes AMD competitive on the high-end again, that would be nice.

 

[Polarman]

AMD - INTEL - AMD - INTEL

Funny..

I do not hear anyone here talking about IBM's upcoming POWER7 CPU.

 

[wolf]

I do not hear anyone here talking about IBM's upcoming POWER7 CPU.

And is it being released at consumer prices with matching motherboards?

 

[newtekie1]

12 cores magny cours at 3.2 ghz
intel is finished.

just my two cents.


http://forums.techpowerup.com/attachment.php?attachmentid=28296&stc=1&d=1252414083

http://www.tweaktown.com/news/13115/another_cpuz_of_magny_cours/index.html

And you think by the time this is released, Intel will not have continued to improve their products? As it stands right now AMD is still behind Intel, so I find it hard to believe that, even if this is better than what Intel has available, Intel is finished...

 

[PVTCaboose1337]

If Intel is finished why do they have a better processor than AMD selling cheaper than AMD can sell theirs?

http://forums.techpowerup.com/showthread.php?t=103335

The new i5 seems to destroy the WHOLE line of current AMD chips. Whoops.