AMD's Next Generation Processor Technology

Esekla writes "AMD has released info about their upcoming processor technology. The press release claims that they're producing circuits that run 30% faster than any other published benchmarks using "Fully Depleted" Silicon-on-Insulator and AMD's metal gating technology and actually has a good bit of technical detail for a press release."

Comments

Metal gates?

[sleepingsquirrel]

I'm not a process guy, so could someone explain why they're claiming metal gates are better? I was under the impresson that metal gates were more compatable with high-k gate oxides, but I didn't see any mention of non-SiO2 dielectrics. And on that note, does anyone know if AMD is trying out any low-k dielectrics for the interconnect?
I also noticed that one of the lines in the slide said something to the effect of, "Mesa isolation was used to keep things simple". Does this mean that they just did that for the one test wafer to keep things easy, but it'll be no problem once we get the process into production? Or are we talking about something that's still many years in the future?

Re:Metal gates?

[FuzzyDaddy]

I'm no longer in the CMOS biz, but let me take a stab at it.

Polysilicon has been the gate material of choice because it is much easier to process. However, metal would reduce the resistance of the gate. (The gate acts like a little capacitor, and the resistance of the gate affects the amount of time it takes to charge up and discharge, which affects the switching time.) I think the processing ease of Polysilicon is lost when you don't use Silicon dioxide as the gate material - for example, if you used a high-K dielectric. I don't know if metal is inherently more compatible with high-k materials, just that it's less compatible with SiO.

They also mention the metal gives a "tunable work function" (probably by adjusting the silicon/nickel ratio), which I would guess would change the turn on voltage of the transistor. Tuning the turn on voltage could certainly tweak up the speed a bit.

Re:Metal gates?

[sarpedon77]

Metal gates have 4 main advantages in advanced CMOS transistors:
(1) The gate resistance is reduced. This lowers the switching delay in some cases. Remember that the delay is proportional to the product of the resistance and capacitance (the 'RC' product).

(2) In polysilicon gates, the free carrier density is very high (1E20 carriers per cubic cm). Even so, under high electric fields that are needed to switch a transistor, there is a small depleted layer created right at the interface of the gate and the dielectric. This effectively acts as a capacitor in series with the dielectric and increases what is called the "effective oxide thickness". This is very bad, especially when process engineers are trying extremely hard to reduce the oxide thickness. At the scales we are at now, every Angstrom counts. In metal gates, the carrier density is 1000X higher. This makes it much harder to deplete and you regain the 4 angstroms. This means either higher performance with the same gate dielectric thickness, or you can get the same performance and increase the dielectric thickness by 4A, thereby reducing the gate tunneling leakage current (and hence power) by an order of magnitude. This is a big deal.

(3) Some high dielectric constant materials (that are candidates to replace silicon dioxide) are not very compatible with polysilicon. This could mean either thermodynamic instability or interfacial charge created that "pins" the workfunction (and affects the switching threshold voltage of the transistor)

(4) In fully-depleted silicon on insulator (FD-SOI, or "depleted substrate transistor" in Intel parlance) transistors, the threshold voltage comes out wrong when using doped polysilicon gates. It makes the transistor either too slow or too leaky. There is a desperate need for tuning the threshold voltage by using a different workfunction which can be found in some metal gates.

Of course, metal gates aren't without their problems. (the predecessors of today's transistors had metal gates - hence the 'M' in CMOS - Complementary METAL Oxide Semiconductor - which were replaced by polysilicon gates for processing ease.) Inability to be easily patterned, withstand high processing temperatures, reliability issues are just a few of them.

depleted silicon?

[konichiwa]

Wow, they must have circumvented UN Resolution 1441 in buying that depleted silicon from the depths of Niger's black market.

WAR AGAINST AMD

Re:Will anyone notice the speed?

[Cruel Angel]

I know I will. Not so much for myself, but in the fact that as the technology speeds up, I see more of some of my friends.

Compile times for programs, and render times for graphics are steadily getting better, which means they finish projects faster, and have more developed social lives.

Which brings me to an interesting question. Is this true:
Faster CPU's = More free time for 'Working' Nerds?

it seems to work in my circle of friends, but is it a 'universal' truth?

Not at all...

[gfxguy]

It's been my experience that people expect you to be able to do more. Work twice as fast? They want twice the output.

I work in the 3D department of a television production studio, and the better the equipment we get, the more demanding the clients are. Often enough it's even worse - we might show a new feature we couldn't do before because the rendering times would be too long, but instead of taking 3 or 4 times the amount it would have, the new hardware brings it to 1.5 or 2 - it still takes longer, it's just that now we can do it.

Re:I/O Speed Please

[Jeffrey Baker]

Opteron already has an excellent memory subsystem and fast paths to PCI-X peripherals. Aggregate I/O and memory bandwidth in 4-way Opterons is pretty sick, and although it won't compete with insane systems from IBM and SGI, it is a lot better than anything else you can get in an $8000 box. What were you hoping for?

All well and good, but...

[Aardpig]

...nowadays I think that the last component of a PC which needs speeding up is the CPU. Many other components act as a brake on the real-world efficiency of systems; one particularly close to my heart is the cache size. Most computational problems which I come across are too large to fit in less than 2 Mb; therefore, on processors which have a much lower clock speed than x86 offerings, but a much larger cache, I get much better results. The Sparc III series is a good example; the clock speed is around 500Mhz (maybe higher on more recent versions), but the 4 Mb instruction cache & 4 Mb data cache (IIRC) mean that the sort of numerical problems I solve can fly. Of course, it could be argued that this is due to the superiority of the SPARC architecture over x86, but you get my point.

I'd be interested to try out one of the new Pentium M processors (as found on Centrino platforms); I understand they have 1 Mb caches, and this may give them quite a performance boost for numerically-intenstive stuff.

Re:Will anyone notice the speed?

[Junks Jerzey]

In all honesty I believe the Slashdot whining is because a lot of posters are poor college students or jobless teenagers. This means they generally cannot afford the shiny stuff. About this time last year I was running a PII-233 myself. By denouncing the great you can make the not-so-great seem better.

I have said this before, and I will said it again. I'm a professional software developer. I work on high-end 3D games, and I have a penchant for working with large, high-level languages that so many programmers put down as "too slow," such as Lisp, when I can. When I had an 866MHz Pentium III, wow, that was my dream machine. It felt like I had infinite processor cycles. If something ever felt a little sluggish, it was because I did something dumb and a little algorithmic tweaking made it go away. I never felt the need for more speed. Ever. Seriously. And now I have a P4 with 3x the clock speed (which I have for reasons other than the old PC not being fast enough).

The "gotta have more speed" issues come down to three major things:

1. Certain very specific tasks eat up all the processor power you can throw at them, such as high-end scientific numerical work (think: systems of tens of thousands of equations) and video compression. Both of these are specific enough that they shouldn't be driving general, across-the-board, desktop CPU development. Ideally, video compression should be done via coprocessor, just as drawing texture mapped triangles is. If we didn't have GPUs like those from nVidia and ATI, we'd need CPUs clocked at 100GHz in order to achieve the same results.

2. Some things are slow, but they often come down to really poor design or have nothing to do with processor speed. Boot time, for example. Or sometimes you hit Help in a giant program like Quark or Maya and there's a substantially long period before the help shows up. That's not a processor bottleneck; that's another program being paged in, maybe even the Java runtime stuff to support it, and then a monstrous index of data being loaded. But people see things like this and immediately think the processor is too slow.

3. There are certain outdated--IMO--activities that some people engage in which are fundamentally flawed, and hence slow. A good example is building monstrous applications using C++. C++ doesn't have formal support for separately compiled modules, so each one is compiled independently, you need an ugly make system to sort out the dependencies, and then they all get thrown into a massive link step at the end. People who write code with Delphi don't have this problem; compile time is effectively zero for most projects. Ditto for Lisp or Python. C++ is a necessary language, but again it shouldn't be the impetus for processor upgrades.

Thanks for reading.

Intel vs AMD

[nepheles]

It's about time that AMD got some recognition for their work, and, more specifically, their R&D. 3DNow! was miles ahead of MMX, and the Athlon was vastly superior to the P3. The AthlonXP in turn beats the P4, Mhz for Mhz. The widespread opinion is that AMD processors are the poor-man's Intel. "Good, but not as good". Hopefully the new Opertron (it will be amazing if the Itanium does nearly as well in the 64-bit marked) and announcements like this will help redress the balance. And show that marketing budget isn't a measure of CPU quality.

Re:Will anyone notice the speed?

[arjun]

you are not running gentoo. now are you ?

Faster CPUs are a huge benefit.

[Andy Dodd]

Without modern CPUs, home video editing would not be practical (and hence the market for DV camcorders would be much smaller.)

You obviously haven't tried compressing 2 hours of video into DVD-quality MPEG-2, let alone trying to compress it into DivX to send home videos to some relatives.

Would we really need more than 800 MHz on a home computer? I have a 1.7 GHz P4 laptop, and a 1.1 GHz Athlon. Upgrading to a Barton 3000+ (2 GHz or so actual clockrate, but much more efficient per clock than my current TBird) would take my 14-hour encoding jobs down to 7 hours. A difference between taking most of the day and running while I sleep.

And reencoding 1080i HDTV recordings into a more managable size... yikes... I've had 24 hour encoding jobs before.

So my suggestion: Go buy a DV camcorder, or an HDTV tuner card. I guarantee you you'll be desperate to upgrade that poke-ass 800 MHz machine in under two weeks.