AMD Socket "?" 65nm | DDR III | PCI-E II

[The Green Bean]

Multiple cores :shocked:

http://www.c627627.com/AMD/Athlon64/Athlon64.gif

 

[CheesePoofs]

PCIE II ?? Wouldn't that be PCIe X32?

Shouldn't their next die shrink be coming Q4 2006 instead of Q4 2007? I was under the impression that die shrinks occured about every two years, and AMD switched to 90nm Q4 2004.

 

[jiffylube1024]

Originally posted by: hatim
Multiple cores :shocked:

http://www.c627627.com/AMD/Athlon64/Athlon64.gif

Q4 2007. This is close to 2008 we're talking about here!

Also, next year looks unexciting so far. A transition to DDR2 looks to be the main change, with a 2.667 GHz part on 90nm. Considering many people have DDR1 at 500-600 MHz already, this means our systems should still be plenty fast for another year .

The big question will be: how will these suckers overclock?

 

[zorrt]

good! the longer it takes for technology to come out, the less money I will have to spend to upgrade :]

 

[MBrown]

That is a Loooooooooooongg time.

 

[n7]

I find it rather disappointing to see how little things are actually improving.

It's going to be another few years at the earliest before multithreaded software will even be mainstream.

And during that few years, there will be bascially no reason to to upgrade our CPUs if we already have X2s, since actual performance improvements will be very small.

Sure, DDR2 will help a bit, since that will mean a much higher HTT also, but look at that X2 5000+.
8x multi ?!?
Unless things drastically change, i'm guessing OCing headroom is going to be extremely small.

Ah well...

 

[DrMrLordX]

n7, the only good thing I can see about moving to DDR2 is that the modules run at a lower vdimm than DDR1. This should reduce overall power consumption, and depending on how well DDR2 modules handle overvolting, it might open up some interesting budget-OC possibilities.

Also, if we have a technological stagnation, some weird stuff might happen in the price/availability department if lots of buyers just hold on to old parts instead of buying new ones.

 

[Leper Messiah]

Originally posted by: DrMrLordX
n7, the only good thing I can see about moving to DDR2 is that the modules run at a lower vdimm than DDR1. This should reduce overall power consumption, and depending on how well DDR2 modules handle overvolting, it might open up some interesting budget-OC possibilities.

Also, if we have a technological stagnation, some weird stuff might happen in the price/availability department if lots of buyers just hold on to old parts instead of buying new ones.

lower volts is actually a bad thing. i would guess that DDR2 needs as much power as DDR, but since its a lowe voltage, it needs more amps to get that energy.

 

[Furen]

The fact that the Q4 2007 parts will be 65nm does not mean that they will be AMD's first 65nm parts. 65nm parts should come out next year and (hopefully) will allow amd to do at least 2-3 clock speed increases (I know, a couple of years back 2-3 clock increases was horrible and now we're HOPING to at least get that). That would leave X2s at 3GHz stock at least, and probably quite a bit higher overclocked.

 

[RichUK]

Originally posted by: CheesePoofs
PCIE II ?? Wouldn't that be PCIe X32?

Shouldn't their next die shrink be coming Q4 2006 instead of Q4 2007? I was under the impression that die shrinks occured about every two years, and AMD switched to 90nm Q4 2004.

Yeah thats right. Pci-e Specifications does in it's initial release support 32 lanes, but for some odd reasons everyone went with 20 lanes (Mobo Manufacturers) i suppose this was because the PCI slots were still installed on the mobo and there was no need for more lanes, as no more PCI-e slot can be installed.

I would have thought that the new chipsets like from nvidia will support x32 lanes (NF5 i presume), as nvidia at the moment are using an MCP chipset in addition to the current NF4 chipset for their new SLI x16 for the extra lanes (16 lanes instead of the current x8 lanes for both GFX cards in SLI). Then after the original Pci-e spec has been exceeded then maybe Pci-e 2 (or the successor) will just support more lanes/bandwidth.

I just cant wait until server motherboards are using Pci-e instead of PCI-X, aka TEH SCSI card goodness. Because on a standalone PC motherboard you would have the ability to utilise full bandwidth on Pci-e instead of the crappy PCI SCSI cards now. As now you have to buy a PCI-X 64bit native motherboard to support the full SCSI bandwidth, where as if the motherboard was using Pci-e you could interchange the SCSI cards for use on a standalone motherboard (NF4) as well as Server mobo?s.

 

[DrMrLordX]

Originally posted by: Leper Messiah

lower volts is actually a bad thing. i would guess that DDR2 needs as much power as DDR, but since its a lowe voltage, it needs more amps to get that energy.

Not true, at least not according to Samsung. Check out this pdf:

http://www.samsung.com/Products/Semicon...Info/10182003_appl_note_ddr2_lp_kh.pdf


from the PDF itself:

DDR2 SDRAM consume less standby powers and operating
powers than DDR SDRAM since the power supply of DDR2 scales
down to 1.8V from 2.5V. At the worst operating condition, DDR2
SDARM saves 65% power consumptions over DDR SDRAM.

 

[exar333]

wrong, DDR2 uses less power than DDR

 

[Viditor]

Originally posted by: CheesePoofs
Shouldn't their next die shrink be coming Q4 2006 instead of Q4 2007? I was under the impression that die shrinks occured about every two years, and AMD switched to 90nm Q4 2004.

The 65nm die shrink will be even sooner...they have samples already, and should be shipping in volume well before next summer.

 

[Budman]

Still shows Socket 939 Semprons but you cant buy them anywhere.....

 

[Mojoed]

Originally posted by: MBrown
That is a Loooooooooooongg time.

Yea really.

Too far out for me to care right now.

 

[Viditor]

I wonder who made up the chart? They appear to have made a few mistakes...