65nm Athlon 64 may be coming sooner than expected...

[stupid]

It takes researchers a minimum of 9 months for lithography tools to be built by tool manufacturers before they can be shipped to and used by AMD, and considering AMD's expectations for manufacturing in 65nm to ramp quickly, availability of parts may happen sooner than later. AMD also recently stated that it expects to be announcing quad-core processors at the tail end of the year right into 2007, and 65nm technology would definitely play a big role in delivering on that goal.

http://www.dailytech.com/article.aspx?newsid=418

No hard dates yet, and no mention of when or if S939 will transition to 65nm.

[ronrem]

Quad core figures to be far beyond my budget,for a long time. The move from 130nm to 90 nm meant MUCH cooler chips,making much quieter puters possible. Another step toward cooler running will further expand the threshold for passive HS's,and low rpm case fans. Gotta love that.

[rpsgc]

No hard dates yet, and no mention of when or if S939 will transition to 65nm.

I guess it depends on how long they will keep making them!

[frostedflakes]

Good to hear. I think AMD is really feeling pressure from Intel on the 65nm shrink. IIRC 65nm used to be slated for 2007, but from the sounds of it we may see the parts mid/late-2006. I've even heard rumors that the dual core Turion64 will be manufactured using a 65nm tech., and I think they're supposed to be available mid-2006. Whether there's any truth to this, I do not know.

[stupid]

No hard dates yet, and no mention of when or if S939 will transition to 65nm.

I guess it depends on how long they will keep making them!

I believe S939 will still be in production as the "new" Sempron series in 2007.

[jaganath]

AMD also recently stated that it expects to be announcing quad-core processors at the tail end of the year right into 2007

What kind of performance uplift can the average user expect from going quad-core? As we have seen with the AMD X2's, dual cores does not necessarily mean double the heat, so if that trend continues with the 4X cores, the performance envelope for near-silent computers could become very impressive indeed.

[stupid]

AMD also recently stated that it expects to be announcing quad-core processors at the tail end of the year right into 2007

What kind of performance uplift can the average user expect from going quad-core? As we have seen with the AMD X2's, dual cores does not necessarily mean double the heat, so if that trend continues with the 4X cores, the performance envelope for near-silent computers could become very impressive indeed.

Kinda difficult to gauge since what "average user" would consider going for a quad core? Games are only just starting to take advantage of dual core; I think you can count all the one that are currently avaible on one hand. The Quads are more for servers than for the home, besides how much will a Quad core cost; $1,500?

The average user will gain more from the single and dual cores. The smaller 65nm process will allow for less power hungry CPUs which means smaller/better performing laptops, and easier to cool desktops.

Based on an article from either Anandtech or Tomshardware, a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+. That is not bad for AMD since the Merom was specifically design for the laptop. I would say it is a safe bet to assume that if the X2 3800+ is built using the 65nm process the gap would drop to 10v. Therefore, one could also assume that AMD's Turion dual cores will run cooler than the preliminary Merom.

[jaganath]

a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+. That is not bad for AMD since the Merom was specifically designed for the laptop. I would say it is a safe bet to assume that if the X2 3800+ is built using the 65nm process the gap would drop to 10v.

Did you mean to write watts there instead of volts? AFAIAA both the Merom and the X2 are/will be powered by a regular 12V line.

[qviri]

Based on an article from either Anandtech or Tomshardware, a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+.

So it's running at approximately -18.7 volts?

Edit: damn, beaten.

[Erssa]

Based on an article from either Anandtech or Tomshardware, a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+. That is not bad for AMD since the Merom was specifically design for the laptop. I would say it is a safe bet to assume that if the X2 3800+ is built using the 65nm process the gap would drop to 10v. Therefore, one could also assume that AMD's Turion dual cores will run cooler than the preliminary Merom.

Volts? I know X2s use around 1.4 volts, so Merom must really go off the scale .
More seriously, I am sure merom will have significantly lower power consumtion then X2s. Merom isn't coming out until 3rd or 4th quarter and yes, it is based on the Pentium M, which has been far less power hungry then X2s, and Merom cores will be manufactured with a better process, so I would believe that systems based on meroms would drain a lot less watts per day then their X2 counterparts. As far as I know, neither Anandtech or THG have had preliminary tests on Merom core. I read both sites regularly. However they both have had tests on Yonah aka Core Duo. Even their Pentium 4 based processors could close the cap to some extend. Presler is already a bit cooler then Prescott and the rumors say that Conroe cores should be even cooler (but hardly on level with amd, just my guess).

[frostedflakes]

AMD also recently stated that it expects to be announcing quad-core processors at the tail end of the year right into 2007

What kind of performance uplift can the average user expect from going quad-core? As we have seen with the AMD X2's, dual cores does not necessarily mean double the heat, so if that trend continues with the 4X cores, the performance envelope for near-silent computers could become very impressive indeed.

Kinda difficult to gauge since what "average user" would consider going for a quad core? Games are only just starting to take advantage of dual core; I think you can count all the one that are currently avaible on one hand. The Quads are more for servers than for the home, besides how much will a Quad core cost; $1,500?

The average user will gain more from the single and dual cores. The smaller 65nm process will allow for less power hungry CPUs which means smaller/better performing laptops, and easier to cool desktops.

Based on an article from either Anandtech or Tomshardware, a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+. That is not bad for AMD since the Merom was specifically design for the laptop. I would say it is a safe bet to assume that if the X2 3800+ is built using the 65nm process the gap would drop to 10v. Therefore, one could also assume that AMD's Turion dual cores will run cooler than the preliminary Merom.

Actually, I doubt 65nm quad core would be insanely priced. It'd probably be similar to current dual core manufactured on a 90nm process. The dies are pretty big, but not big enough that getting decent yields is impossible.

However, quad core probably won't be very affordable until the next die shrink (45nm I believe).

[stupid]

a preliminary dual core Pentium M (Merom if I remember correctly) uses about 20 volts less than the Althon 64 X2 3800+. That is not bad for AMD since the Merom was specifically designed for the laptop. I would say it is a safe bet to assume that if the X2 3800+ is built using the 65nm process the gap would drop to 10v.

Did you mean to write watts there instead of volts? AFAIAA both the Merom and the X2 are/will be powered by a regular 12V line.

Sorry, about that. Yeah, I meant watts instead of volts.

[~El~Jefe~]

Merom will have a slightly longer pipeline than Dothan... same with yonah, also 64 bit instructions finally, (intel is so retarded for not having it on 2 years ago).

Also, I believe more transistors.... Currently, Core Duo is drawing a lot more power than notebook people expected.

I have no doubt it will be less but I think not much if at all when you consider the external memory controller.

[Erssa]

65nm AMDs may come later then you expect according to the Inquirer

[stupid]

65nm AMDs may come later then you expect according to the Inquirer

Well, I prety much given up on holding out for a 65nm S939 dual core CPU. I've already delayed my upgrade for about 6 months. So I'm pushing forward to simply pickup a 90nm CPU.

[rpsgc]

65nm AMDs may come later then you expect according to the Inquirer

Let's just hope that they're wrong...

[jaganath]

quad core probably won't be very affordable until the next die shrink (45nm I believe).

Just a question; how much further can they shrink die sizes before they start coming up against physical constraints like leakage current and electron tunnelling?

Moore's Law

Gordon Moore himself stated in an interview that the law may not hold valid for too long, since transistors may reach the limits of miniaturization at atomic levels.

In terms of size [of transistor] you can see that we're approaching the size of atoms which is a fundamental barrier, but it'll be two or three generations before we get that far—but that's as far out as we've ever been able to see. We have another 10 to 20 years before we reach a fundamental limit. By then they'll be able to make bigger chips and have transistor budgets in the billions.
—(techworld)

While this time horizon for Moore's law scaling is possible, it does not come without underlying engineering challenges. One of the major challenges in integrated circuits that use nanoscale transistors is increase in parameter variation and leakage currents. As a result of variation and leakage, the design margins available to do predictive design is becoming harder and additionally such systems dissipate considerable power even when not switching. Adaptive and statisical design along with leakage power reduction is critical to sustain scaling of CMOS. A good treatment of these topics is covered in Leakage in Nanometer CMOS Technologies. Other scaling challenges include (1) the ability to control parasitic resistance and capacitance in transistors, (2) the ability to maintain proper transistor electrostatics that allow the gate terminal to control the ON/OFF behavior, (3) system level power delivery, (4) thermal design to effectively handle the dissipation of delivered power, and (5) solve all these challenges with ever reducing cost of manufactuing of the over all system.

[rpsgc]

32nm?

[perplex]

45nm, 32nm and 22nm is expected. After that, who knows? I think the industry will shift from Silicon to something else after that.

[CA_Steve]

re: how much further

<shrugs> every barrier seems to bring a new solution...deep UV lithography, deep UV in liquid lithography, bonded wafers, strained silicon, etc...

[ryboto]

i read somewhere...and don't ask me to link it..cause it was like 2 years ago, but, i read that around 5nm is where electrons begin to jump, and all the bad stuff happens. But, there was mention in whatever it was i read, that increasing the power might circumvent the noise, but then, if we have to eat more power at a smaller process, whats the point?

[Erssa]

but then, if we have to eat more power at a smaller process, whats the point?

Well that has been the trend for a long time now. The points are e.g increased processing power and smaller cost per chip.

[jaganath]

i read that around 5nm is where electrons begin to jump, and all the bad stuff happens.

Well this Intel document states that the gate dielectric in a 65nm transistor (which is where we are at now) is only 1.2nm thick (about 5 atomic layers or 5 atoms); so you may have been thinking of that rather than 5nm.

Intel PDF: Addressing Power at the Transistor Level-page 4

However good news for SPCR types:

Discovery Reduces Leakage Current

University of Kentucky researchers have discovered a means of reducing gate leakage current of transistors in computer chips that will permit chip producers to continue developing more efficient and powerful chips with reduced power consumption.

Zhi Chen, associate professor of electrical and computer engineering, found that applying rapid thermal processing directly on gate insulators – used to control current flow of transistors in computer chips – can dramatically reduce the chips' leakage current and correspondingly the power consumption. In fact, the technique can improve the insulating qualities of gate insulators so that their direct tunneling current is reduced by 10,000 to 100,000 times. No effect was found if rapid thermal processing was not directly applied on the gate insulators.

In order to improve computer chips' performance, transistors' size and gate insulators have to be continuously shrunken so that more components can be packed into a single chip. Computer chip producers were hitting a wall in downscaling the transistors and gate insulators because of their inability to reduce the leakage current of the existing gate insulators. This new technique will help the chip producers to develop more powerful chips with low power consumption.

Chen and his team will present their findings in a paper to be presented Dec. 7-9 at the 2005 International Semiconductor Device Research Symposium in Bethesda, Md.

http://www.uky.edu/

Posted 7th December 2005

[stupid]

I am difinitely not in the computer or engineering field, but aren't CPUs expected to move away from silicon and gravitate towards silicon germanide alloy or germanium within less than a decade?

[jaganath]

I am difinitely not in the computer or engineering field, but aren't CPUs expected to move away from silicon and gravitate towards silicon germanide alloy or germanium within less than a decade?

Are you referring to the use of SiGe in strained silicon? This is already implemented in all of AMD's 90nm chips, and some Intels (Prescott etc).

[stupid]

I am difinitely not in the computer or engineering field, but aren't CPUs expected to move away from silicon and gravitate towards silicon germanide alloy or germanium within less than a decade?

Are you referring to the use of SiGe in strained silicon? This is already implemented in all of AMD's 90nm chips, and some Intels (Prescott etc).

Similar. From what I can recall in an article I've read, SiGe or Ge is supposed to totally replace silicon, not just SiGe in strained silicon. I could be wrong though.