PDA

View Full Version : AMD (REAL) Quad, MORE details


Pages : [1] 2 3

Heinz68
11-28-06, 07:51 AM
theINQUIRER was able to dig few more DETAILS :) about the AMD's Altair FX real Quad, 65 nanometre silicon
The article states that each CPU is going to have 2MB of dedicated L2 cache memory, total of 4MB of L2 memory plus 2MB of shared L3, total 6MB.
I don't get it, since it has 4 cores shouldn't it be 4X2MB L2 plus 2MB L3, total 10MB, probably some mistake or I'm wrong.

It should have a new Hypertransport Bus working at 4000MHz or more (another fine detail digged :) )
It will continue to use 1207 pin socket F 2P same as for the 4X4 QuadFather (http://www.nvnews.net/vbulletin/showthread.php?t=80113)
The 4X4 should be available in few days and the Altair FX real Quad should arrive Q3 2007
Intel and the fanboys better watch out :D

Here is theINQUIRER (http://www.theinquirer.net/default.aspx?article=35209) link

Badboy_12345
11-28-06, 08:14 AM
I see no benchmark numbers :bleh:

Heinz68
11-28-06, 08:42 AM
I see no benchmark numbers :bleh:Neither do I see any benchmark numbers for Intel REAL Quad.

Intel17
11-28-06, 09:02 AM
Neither do I see any benchmark numbers for Intel REAL Quad.

I don't know why people such as yourself keep bashing Kentsfield simply because it's two dies on one chip rather than one big integrated piece of silicon.

For all intents and purposes, it's a quad core, meaning that on one chip, there are four discrete processing cores.

retsam
11-28-06, 09:08 AM
you guys should look into how mainferame processors are set up on die....i would love for intel or amd to call IBM's multicore solution not real....they would laugh right in there faces.

nekrosoft13
11-28-06, 09:30 AM
amd "real" quad

haha

for me if i can see it, touch it, then its real. not something on a paper

Galactic
11-28-06, 10:21 AM
amd "real" quad

haha

for me if i can see it, touch it, then its real. not something on a paper
The 'real' part refers to it being an actual quadcore, not two dualcores glued together.

nekrosoft13
11-28-06, 10:26 AM
2x2 = 4

1+1+1+1 = 4

so whats the difference

Marvel_us
11-28-06, 10:44 AM
2x2 = 4

1+1+1+1 = 4

so whats the difference

Difference is in the company's definition. Like a marketing thing by AMD.

Core 2 Duo and Kentsfield are handing it to AMD and all they can really do now is say, "Just wait for our REAL quad core solution!".

If AMD's "real" quadcore method is better then it will wipe the floor with Kentsfield.

If it only shows minimal gains then people with a Kentsfield won't really care.

nekrosoft13
11-28-06, 10:53 AM
plus by seeing how hell C2D overlocks, intel could release couple much higher clock C2D right now and hurt amd even more

i'm not loyal to either intel or amd, i had intel, now i have amd, my next step will be intel again.

Heinz68
11-28-06, 10:55 AM
I don't know why people such as yourself keep bashing Kentsfield simply because it's two dies on one chip rather than one big integrated piece of silicon.

For all intents and purposes, it's a quad core, meaning that on one chip, there are four discrete processing cores.I chalendge you to find one post of mine where I was bashing Kentsfield, before you call me (people such as yourself).
Or there is never late to apologies.

I was just thinking since "Badboy_12345" was smart ass reply it deserved nothing better than smart ass reply from me.
Or did "Badboy_12345" had some problem to understand "Hardware Rumor Mill" or Q3 2007.

In the meantime I see Intel Yorkfield (native quad-core) posted in "CPUs, Motherboards and Memory (http://www.nvnews.net/vbulletin/showthread.php?t=81112)" forum and nobody ask for benchmark numbers there, sure more proper place to ask.

On top of it according to theINQUIRER (http://www.theinquirer.net/default.aspx?article=36010) the rumor about the Yorkfield is wrong.
There are a few things that you won't see this time around, real quad cores and higher FSBs. The Yorktown rumours that were flying around at IDF were and are wrong, there is no converged quad core parts until Nehalem. There are also no plans for a 1600FSB unless something bad happens to Nehalem/CSI, or AMD pressures Intel hard.
I know some will say but it's theINQUIRER, I say rumor is a rumor doesn't matter where it comes from, that's why we have rumor forum.
VERY FEW people at this forum very selectively believe theINQUIRER only when it says something negative about AMD/ATI

nemecb
11-28-06, 11:49 AM
Q3 2007:thumbdwn:

Bearclaw
11-28-06, 04:17 PM
Q3 2007:thumbdwn:
I saw that, what a bummer.

Although the 4x4 is comign out in a few days according to the inquirer...

Lfctony
11-28-06, 05:24 PM
It's the whole 7950GX2 story all over again. It's one card, not it's 2 cards, it's 1 PCIe slot, no it's 2 PCBs. Blah blah blah yadda yadda yadda. By definition, I think it's stupid to argue about this.

TiKiMaN1
11-28-06, 06:30 PM
From what I read in Maximum PC you can expect Intels "True" (although I personally consider the QX6700 true quad anyway) out around Q3 2007 as well.

DarkJedi664
11-28-06, 06:39 PM
Out of curiosity, how is the QX6700 true Quad Core? Anyways, the Opteron version of K8L (which I think this chip is), is due in March/June time, so we'll know performance then :)

Dazz
11-28-06, 07:07 PM
What they mean by true quad core they mean all the CPU's can compuicate with each other directly either by HT or interconnect. While Kentsfield uses the FSB as such it's not a true quad core CPU but considering nothing on the market can do so much in so little time it will do.

superklye
11-28-06, 07:52 PM
It IS true/real quad...it's just not NATIVE quad.

radekhulan
11-28-06, 08:49 PM
What they mean by true quad core they mean all the CPU's can compuicate with each other directly either by HT or interconnect. While Kentsfield uses the FSB as such it's not a true quad core CPU but considering nothing on the market can do so much in so little time it will do.

FSB is not inferior to HT by any means ;-) 1066MHz vs 1000MHz..

Heinz68
11-28-06, 08:59 PM
It IS true/real quad...it's just not NATIVE quad.
Some technology articles when describing the differences in Quad architecture are using "REAL" and others talk about "NATIVE" but basically they talk about the same thing.
If the REAL (NATIVE) architecture was not better, Intel would not be developing one for the future.

Anyway the bottom line for people upgrading right NOW and wanting to have Quad there is only one choice and that is Intel.
Well maybe the AMD 4x4 QuadFather which according to many rumors should be available in few days might be an alternative, provided the performance/value ratio is going to be close to the Intel Quad.

superklye
11-29-06, 01:24 AM
Some technology articles when describing the differences in Quad architecture are using "REAL" and others talk about "NATIVE" but basically they talk about the same thing.
If the REAL (NATIVE) architecture was not better, Intel would not be developing one for the future.

Anyway the bottom line for people upgrading right NOW and wanting to have Quad there is only one choice and that is Intel.
Well maybe the AMD 4x4 QuadFather which according to many rumors should be available in few days might be an alternative, provided the performance/value ratio is going to be close to the Intel Quad.
Well, I don't care what the articles are saying because they're wrong.

Real means it exists. And soon enough, there will be quad-core Intel chips. Sure, it'll just be two Core 2 Duos slapped together on the same die, but it's still FOUR CORES. That's a REAL quad core chip.

Native quad core, on the other hand, is also real and is something totally different from non-native. Those CPUs will have four cores that were designed from the ground up to work together and talk to one another via other means that just the FSB. Those are also REAL quad core chips.

Heinz68
11-29-06, 01:35 AM
Well, I don't care what the articles are saying because they're wrong. I give up.

superklye
11-29-06, 01:59 AM
They're using a word incorrectly and subsequently people are, too.

Do you know that "impeach" just means "bring to trial." 99% of people think it means to kick someone out of office and this is the same situation.

Heinz68
11-29-06, 08:27 AM
More rumors about the AMD Quad core Altair FX dug by theINQUIRER

LINK (http://www.theinquirer.net/default.aspx?article=36017) for this article
Quad core Altair FX needs an F1207+ board to really fly

and the Altair FX - the big picture :) http://img153.imageshack.us/img153/5558/newaltairgn0.jpg
EDITED:
Dang it! I just noticed the dangINQUIRER using again the prohibited 4 letter word (R**L) when describing QUAD :)
Please do not flame me I'm just the messenger, go flame The Author he doesn't mind.
You'll find the link at the bottom of the article

Dazz
11-29-06, 11:14 AM
FSB is not inferior to HT by any means ;-) 1066MHz vs 1000MHz..
HT runs at 1000MHz and it is bi directional so effectively it's 2000MHz like DDR Vs SDR. Also think
of it this way, the FSB is shared with all other devices, south bridge, north bridge data from memory
to CPU's and then you have the CPU's comunicating via the FSB this will satuate the FSB. HT works
independtly current desktop CPU's have have 1, one to the South bridge and the other for the memory
(not refering to memory controller BTW. Now you have Opterons which have extra channels (2x for dual
skt, 4x for quad etc.) and you will see memory peformance increase as each core can access memory.
I recall seeing people with DDR 400 getting benchmarks of over 10GB/sec for bandwith on opertons as
each CPU/core this is not noticed as benchmarks testing memory only only use the 1 CPU per socket. As
they all have their own lanes to use rather then sharing. But yeah rather then real use native which ever
floats your boat. This was also why AMD X2 CPU's peformed better then the P4 dual cores, % wise i mean
not over all. The more cores using the FSB the lower the efficenty.