Has anyone heard anything about the 65NM process yet? I haven't seen anything significant yet about when it supposedly to come out.
Printable View
Has anyone heard anything about the 65NM process yet? I haven't seen anything significant yet about when it supposedly to come out.
Intel is converting one of their fabs to 65, it should be completed by next year.
AMD's new 65nm facility (Fab36) is already up and running but production isn't yet in commercial scale. It takes some time to get the machines tuned. At it's worst, 65nm A64 cpu's will kick in soon after Intel goes 65nm late this year. I'd love to see those little green 65nm stones before the blue ones...
The worst thing people at AMD can imagine is that they are overrun by Intel - again - like it was when Intel got 90nm out first.
But WORSE :devil:Quote:
Originally Posted by largon
Intel is always first to a new process. They spend enough on fabs to make sure no one even has a chance at it. Currently Intel is saying first 65nm chips will be Q1 2006, with febuary being tossed around as the launch date for Yonah (which will be the cheapest/simplist 65nm core).Quote:
Originally Posted by largon
This is the driving force for revision of all current textbooks in microelectronics fabrication.
Do you think the socket M2 will be out before AMD hits 65nm?
yes
I wonder if AMD will create a FX 59 in the 90 and 65 nm process flavors.
FX-59 will strictly be a San Diego core ;)
It won't be San Diego in any way, because it will have a new DDR-2 memory controller.
FX-59 will be San Diego ;)
could you explain why you think that, or give the source of your "conclusion"Quote:
Originally Posted by Vapor
Thanks
Instead of SOI, 65nm will use silicone. *cough* :banana:
Vapor
Why would AMD release a new chip before M2 comes out? Single-core leadership was gained with FX-55 and FX-57 is, by far, the fastest single-core chip on the market right now in the target area - games.
AMD could probably produce an FX-59 on 0.09 San Diego core but why bother.
jahjahbinks
silicone's for boobs, silicon's for chips.
65nm would be nice for 939..
PS it is lowercase n not N...
U didn't get my joke *cough*, did u? :D
http://www.xtremesystems.org/forums/...&postcount=451
That was after a lot of coughing, now I need some water. :cool:
I did I just didn't think it was funny...
Sorry if I offended anyone, priv msg me and I will be glad to take it out.
It's pretty much been confirmed that FX-59 will be a San Diego by an XIP who knows his stuff and M2 won't be around for awhile, you better believe that there's gonna be another single core product between now and then.
The FX59 will be out well before 65nm. You don't have to have industry contacts to know that. They'll launch it no later then Intel launches their 65nm chips to steal some of the thunder.
and if we are lucky AMD will beat Intel to the punch with superior process....
I read somewhere that when choosing a smaller manufacturing process like the upcoming 65nm, the production costs actually goes down... how can this be? The error margin must be bigger, development costs goes up etc. Does the smaller technology require less raw material or something? ;) Like I said, I just read it somewhere and got curious...
lets just say you can go from a 90nm process to a 45nm process... each chip takes up 1/4 the space meaning that you can make atleast 4 times as many chips from the same wafer... even if you lose twice as many chips... The cost of production is 1/2 as much as to make a 90nm chip.
Plus the smaller the process the less heat it puts out unless you are intel....
Which is what Overclockers like... Hence you can sell it at a highier price..
They win both ways...
ok, thanks for the explanation
your welcome
so it will be the FX 61 before they come out with the M2? I see no new core being worth buying before the M2 currently
S939 will never become 65 nm, just like S462 never used 90 nm. I think 65 nm M2 will show up 12 months from now, when nobody want's a 939 except for budget systems.
I wonder what will happen with HTT and RAM speeds. They will use 667 MHz DDR2, does it mean that the CPU will run at (333 x multiplier)? If so, then I guess there will be a long wait for both mature CPU's and mobos that can handle at least 500 MHz bus speeds for some decent overclocking. I don't know if you need new chipsets for the new socket. If not, then mobo manufacturers will probably try to sell mobos with old chipsets in the beginning. A 400 MHz capable chipset is the best for today, but will only give you a 20 % overclock with M2... :stick:
HT will stay the same, but the memory divider will change.Quote:
Originally Posted by Mats
How do you know that? Won't they run at 3x333 instead of 5x200 (HTT)?Quote:
Originally Posted by saratoga
Because that would be both expensive and pointless.Quote:
Originally Posted by Mats
Speculations...Quote:
Originally Posted by saratoga
However, I hope you're right, otherwise we'll get lower max multipliers and a demand for faster chipsets like I said. I hope we'll find out long before they show up.
may I suggest an X2?Quote:
Originally Posted by eliminate
I think that should be highier up on your list than the M2
The FX59 WILL be a .09 3.0Ghz San Diego core cpu. This most likely will be the last single core processor introduced by AMD. After it's release (which by the way should be expected in the next 2-3 months) AMD will shift it's focus to dual core cpu's. IIRC, ddr2 cpu's will be released sometime during Q2 2006, starting with the Opterons.......
Hehe...hopefully the 1200+ pin DDR2 Opties will do the same :p:Quote:
Originally Posted by Troman
EDIT: THen again, that'll be impossible running reg ecc ddr2 on server boards......
are they definately going to be ECC only like the 940's?
Yeah the Opterons made for multiple CPU implementations will remain with ECC only from what I understand. It just makes so much sense in a server workstation environment... more so than overclockability (I never overclock a serious work computer myself)
Yup, right on point there. The main reason why the s940 single Opteron cpu was moved to s939 wasn't because AMD wanted to create a workstation solution which utilized non-Ecc DDR. Rather it had much to do with the cost of manufacturing for s940 motherboards in general. All current s940 motherboards are manufactured with a 6-layer pcb. This results in a much higher cost for the end-user and therefore very little demand for a single cpu s940 workstation solution. By shifting the socket to 939, the single cpu Opteron workstation becomes a much more feasible solution for those who are on a budget ;)Quote:
Originally Posted by Ugly n Grey
my thought is if you want more than 1 Cpu on a mobo you can afford a $200+ mobo
I think there's at least 6 months until the next FX chip so that puts us right where M2 should be announced. I'll be really surprised if there is a S939 FX-59.
Quote:
Originally Posted by nn_step
100% of the cost isn't materials though is it? Factor in the cost of building the fab, additional work, testing and packaging costs. You aren't exactly at 50% cost anymore. In fact for AMD, you're probably losing money for the first couple years anyway.
Well I guess your gonna end up really surprised. All the roadmaps I've seen put the FX-59 as a 3.0ghz san diego core on socket 939.Quote:
Originally Posted by IvanAndreevich
I based my estimate on pure manufacturing costs aloneQuote:
Originally Posted by Lithan
Is it the Socket F/1207 that will have integrated PCIE controller? Better 3D performance, worse/no overclocking because of ECC RAM, DUH!!! :stick:
I want both!! Will AMD really stop themself from making THE gaming machine by not using the integrated PCIE controller in A64? :confused:
You guys forgot about R&D cost, each fab lab can cost up to several billion dollars.
But do you really want to lock your self into exactly one type of Graphics card connection?Quote:
Originally Posted by Mats
besides the latencies for graphics aren't that bad...
What's wrong with PCIE? Unless you have X850 AGP lying around which I don't.Quote:
Originally Posted by nn_step
Well, I was just speculating, or dreaming.. we don't even know what that PCIE controller can do in reality. As usual, wait and see... :slap:
Lets see, Oh I know the fact that is going to be replaced within the next 18 months :slap: PCI-E2Quote:
Originally Posted by Mats
Well, good for you then. :toast:Quote:
Originally Posted by nn_step
you don't plan on jumping on the bandwagon too?
already?Quote:
Originally Posted by nn_step
graphics cards don't need it. most video cards don't even use the full bandwith of AGP 4x let alone 8x
Why can't we overclock ecc ram? It's the same except for the ecc chip on it, ocz can make us a ecc chip with headroom :D right ocz.
I think they mean buffered memory, not ECC.Quote:
Originally Posted by 4rory
We can but not as much, but we can NOT overclock registered RAM...Quote:
Originally Posted by 4rory
About AMD Fab36 (only 65nm nodes):
Quote:
CEO Hector Ruiz said that AMD will start production in the first quarter
There will be no 90nm procs for socket M2?
My theory: They'll start with 65 nm for (dual core) mobile chips. Why? Well, because of Intel's 65 nm PM. AMD already have dual core chips, but they need to cut down in power consumption to make it competitive for laptops now that Yonah will show up.
Here they're mentioning 333 MHz bus speed, I think it's true. I wouldn't expect to see a new socket coming out running with a divider (ya know what I mean) as standard.
No way. Until the introduction of 65nm (That will be on the least 4 month or more after the introduction of Socket M2), AMD don't got enough capacity to shift everything to Dual Core based products. Neither Toledo or a Manchester are cheap enough to manufacure for AMD to make a full transition to Dual Core based products after the A64 FX-59/60 or whatever it is. Expect a nice Single Core Athlon 64 lineup in Socket M2, because until 65nm, I don't think that AMD will massively manufacture such big Cores, it will limit its capacity even further that it is today. With the exeption that Fab 36 is already doing wonders in 90nm...Quote:
Originally Posted by s7e9h3n
Not sure about the Base Clock. If I recall correctly, Winchester was gonna be the first K8 to support Hyper Transport 2.0 (Hyper Transport 1.0 was 800 MHz (200 MHz * 4) and Hyper Transport 1.1 was 1 GHz (200 MHz * 5)), but I am completely pointless about what were its specifications and why it was never advertised, or if it did maked it to Winchester. Anyone got more info about that? It should be closely related to the future Base Clock in K8s...
Got a link confirming that the Opterons will have thier own integrated PCIe controller as last I saw this was speculation/rumour...Quote:
Originally Posted by Mats
AFAIK AMD won't have volume .65 chips until late 2nd half of 2006 with some early production .65 in very low volume at or just before the beginning of the 2nd half of the year. Until then FAB 36 will be .90 production with socket M2 coming out in the first half of 2006 according to AMD's latest road maps I think we'll be seeing the latest FX chip introduced on socket M2 with a DDR1 version coming out later for socket 939.
Ummm, my guess is that you haven't seen an E6 "San Diego" core yet ;) . Well, I'll put it to you this way - AMD will soon start transitioning all the current A64 s939 E4 San Diego's (eg. 3700+, 4000+, etc) to this type of core and it definately isn't anything like the current Diegos......Quote:
Originally Posted by zir_blazer
Rev. E6? JH-E6 maybe? I think that soon we will know what on the earth AMD does with the Toledos and Manchesters that got an unworking Core but does fine with the other Core. Its a fairly better idea that throw the entire Core in the trash can when you can still sell it for a profit, but then, AMD is recycling Cores and San Diegos didn't exactly getted a new Revision because those are just Toledos. I am speculating, but the argument is pretty well sustained (Athlons 64 3500+ with Clawhammers or San Diego Cores with half of its Cache L2 disabled and others are examples of this, but now with an entire Core disabled or even a Core and part of the Cache L2 of the other Core. Its possible).Quote:
Originally Posted by s7e9h3n
http://img373.imageshack.us/img373/1229/infolv7hl.jpg http://img327.imageshack.us/img327/4026/3505cpuz5cm.jpgQuote:
Originally Posted by zir_blazer
;)