Yonah (dual core Dothan) thread - formerly Yonah clues

[K. Eng]

http://www.anandtech.com/cpuchipsets/sh ... spx?i=2388

The green bar showing Yonah starts well before the 2006 and Beyond part of the chart. I'm hoping that this implies a 'Q3 launch.

[Batuta]

...

[K. Eng]

My guess is that Windows XP and other SMP aware operating systems won''t need any changes to recognize a dual processor. However, I am sure that Intel will release a new power management driver to control Yonah's ability to shut down the core that isn't being used.

Some dual core processors are no different than SMP. For example, the upcoming Pentium D and Pentium XE (dual core Prescott), contains no arbitration logic connecting the two cores. The chipset sees the processors as separate entities on the frontside bus and the cores cannot talk to each other without using the bus. Yonah has arbitration logic, but I do not know how advanced it is.

Costs on Pentium D and Pentium XE will probably be high due to the factors you listed, though Intel will continue to ship single core products for some time. Competition with AMD should also work to keep costs down, and 65nm fabbing is just around the corner.

[Batuta]

...

[gazingwa]

MHZ and performance are 2 different things, there will always be a performance race. but it will no longer be only by increasing clock speed

[K. Eng]

The 'speed race' referred pretty much to the MHz rating of a CPU and not the actual performance. The Pentium 4 was the ultimate speedracer; Intel designed to clock astronomically high (and it did, relative to AMD's K7 family).

While shrinking the fab process generally allows CPUs to achieve higher clock rates, remember that it also makes the CPU smaller. Thus, even if the move to 65nm does not offer a lot of clockspeed headroom, it will allow Intel and AMD to get more CPUs out of a wafer. The benefits are clear when one considers that on a 90nm process, dual core Pentium D (2x Prescott) are in the 200 mm2 range, compared to just 112 mm2 for a single core Prescott (with 1MB L2). Going to 65nm should make that dual core chip more economical to manufacture.

K. Eng:

nd 65nm fabbing is just around the corner.

But since higher integration is associated with higher speeds and the article mentioned that both AMD and Intel "have given up on the speed race", that is somewhat contradictionary.

[Batuta]

...

[K. Eng]

It all depends on how you define performance.

In terms of single thread, integer bound tasks (MS Office, email, basic stuff), performance will probably not increase all that much, and it may not matter, since even a 750 MHz Pentium III is capable of running Windows XP and Microsoft Office XP very well, so long as it has enough RAM.

Gaming performance right now is more dependent on GPU rather than CPU, and the GPU performance race shows no signs of slowing down.

I do see clock speed as less of a selling factor, and the ability to handle more tasks concurrently becoming more important. Imagine, for example, a home server that delivers digital content of several types (games, digital tv, audio) to multiple users in different parts of a house. A computer that can do a ton of things at once at good speed is more useful than a computer that can do only one thing at a time at the best speed possible. Here's another example. I have two friends who both ran FTP servers all day and also played games once in awhile. Friend A had a dual Pentium III system at 1 GHz, while friend B had a Pentium 4 system at 2.13 GHz (a 1.6 GHz overclocked, no HT). Friend A could run PlayStation emulation and the server at the same time without any problems, while Friend B often had stuttering problems in the game because the CPU had to keep switching tasks. So Friend A's system was better performing overall, even though Friend B's system was much much faster for any given single task.

But w/o the clock speed as the main factor of pushing CPU performance, there will still be a hard time in achieving the doubling of computer performance every 2-3 years that our entire IT industry has come to rely on.

[Batuta]

...

[K. Eng]

The Inquirer just put out some new figures on expected Yonah clock speeds: http://www.theinquirer.net/?article=22500

That's according to recent road maps seen by the INQUIRER, which show that the dual core models will clock at 2.17GHz, 2GHz, 1.83GHz and 1.67GHz. Each of these processors will include 667MHz buses, have 2MB of L2 cache, and support other desktop features such as VT.

It doesn't look like Yonah will initially clock much higher than Dothan, so hopefully the improvements to the execution cores (FP improvements called "Digital Media Boost") will bring performance benefits. I'm looking forward to VT - being able to run different OSes concurently on the same hardware would be great for people developing for different platforms (e.g. Windows and Linux at the same time).

I think we might have reached the end of stellar growth for the IT tech sector for good and we'll see a rather flat line in that sector from now on.

I think that the computing sector is relatively mature in the U.S. at this time, but markets in India and China present many new opportunities, so I am hopeful.

[K. Eng]

http://news.com.com/Intel+spills+beans+ ... g=nefd.top

The chip, which will be made on the 65-nanometer process, will come with a number of enhancements over the current Pentium M line of notebook chips, Mooly Eden, vice president of the mobility group, said at a briefing here.

"When I speak of dual cores, I am not talking about a 10 percent to 20 percent improvement in performance. I am speaking about something crazy," Eden said.

But no 64-bit. Although to be honest, x64 compatibility seems to give 5% performance boost (at most) in typical situations.

One thing Yonah won't have, at least initially, is the ability to run 64-bit applications.

"We made a conscious decision not to include it" because of the impact on battery life, Eden said.

He further added that the Yonah chip has been in the works for years. "I believe we were designing it before anyone knew how to spell dual core," he said.

Yonah is the real deal as far as dual cores go. The Smithfield (Pentium D, aka Smithfield core) is nothing but a lousy hack compared to Yonah.

[wolfman]

Or compared to dual core opterons...

[K. Eng]

There was another Yonah demonstration recently:

http://www.tomshardware.com/hardnews/20 ... 23943.html

Intel's Mobility Group VP, Mooly Eden, came to the Mobile and Wireless World Conference to bring IT managers up to speed on Intel's Yonah processor, but also showed the firm's Active Management Technology (AMT) that promises to become a "virtual IT department" on a chip.

AMT, wholly integrated in silicon, basically gives IT managers remote out of band access to laptops and desktop to enable repairs on such devices. Out of band access bypasses the computer's operating system and other loaded software and therefore reaches computers that otherwise would be unreachable. As long as a computer's hardware is functional, IT staff can access the computer on a network and fix software related problems - even when a damaged device is turned off.

This is going to do wonders for IT management. People on the road could have instant access to their corporate IT dept for help if their notebook gets disabled by a virus or malware I just wonder how this works over the Internet.

[K. Eng]

More news from x86-secret, which I consider to be a reliable source. Translated through Google:

Good news, those[Yonah CPUs] functions without any problem with 2.5 GHz and do not present that very few bugs. The yields are also excellent and INTEL should be prét with launching the mass production as of the re-entry.

http://babelfish.altavista.com/babelfis ... 898&intl=1

Looks like it's ready to go. Too bad Intel won't launch this for another 6 months The Stable Platform Image is great for companies, but I have to say that it makes me impatient.

[K. Eng]

http://www.tmcnet.com/usubmit/2005/Jul/1164898.htm

Yonah has started to sample:

Intel continued the development of its forthcoming 65nm process technology and began providing customers with samples of microprocessors codenamed Yonah, Presler and Dempsey, the company's first 65nm dual-core microprocessors for notebook, desktop and server platforms, respectively. Intel's industry-leading 65nm process technology will enable cost-effective production of dual-core processors, with platforms ramping into high volume in 2006.

[K. Eng]

The folks over at Laptoplogic.com have obtained what appear to be Japanese roadmaps for the mobile Intel dual cores. However, note that these are not official Intel slides.

http://www.laptoplogic.com/news/08/08/2005/410/0/

A few interesting things to note: Yonah will almost completely displace Dothan in Q1 2006. Merom will launch in Q3 2006, although Yonah will occupy almost all of the market segments except the high-performance professional segment.

You can read the stuff in the original Japanese: http://pc.watch.impress.co.jp/docs/2005 ... gai201.htm

This site is generally reliable in its predictions.

[HamTard]

Does that mean Q1 2006, all Dothan based notebooks will get a big price drop to be cleared out?

[K. Eng]

Probably yes. Manufactuerers are going to want to clear out old stock ASAP.

Does that mean Q1 2006, all Dothan based notebooks will get a big price drop to be cleared out?

[jasondavis]

Care to venture a guess as to how much of a price drop should be expected? I realize that's pretty much impossible to do accurately... but if you feel like ballparking it...

[K. Eng]

I have no idea.

Even the manufacturers probably do not know at this point how much they will discount in order to move old stock.

Care to venture a guess as to how much of a price drop should be expected? I realize that's pretty much impossible to do accurately... but if you feel like ballparking it...

[tom2517]

But as I understand it, Yonahs are not 64 bit processors, at least at first. Personally I would wait till Yonahs becomes 64 bit processors and pair it with Windows Vista to really benefit from 64 bit technology.

[K. Eng]

64 bit is overhyped. There are 2 benefits to x64 processors (1) ability to address more than 4GB of RAM (2) Double the number of GPRs.

The first advantage doesn't really help most people. The vast majority of users are not going to be using more than 4GB of memory. Most mobile chipsets don't support more than 2GB. Most people won't ever force their computer to go near 4GB, even in virtual memory.

The second advantage is worth a typical performance boost of 5%, or about the same as doubling the L2 cache.

64 bit provides advantages, but they are not as great as people make them out to be.

Windows Vista will have x64, 32-bit x86, and Itanium builds, so anyone with even somewhat modern computers should be able to run it.

http://www.pcworld.com/news/article/0,aid,121435,00.asp

But as I understand it, Yonahs are not 64 bit processors, at least at first. Personally I would wait till Yonahs becomes 64 bit processors and pair it with Windows Vista to really benefit from 64 bit technology.

[K. Eng]

Some architectural details have emerged concerning Yonah:

http://www.theinquirer.net/?article=25349

You know that it will be about a good chunk faster clock for clock, per core, than Dothan, especially on multimedia apps. Numbers like 30% are being passed around for encoding. Why? Two main reason, the main one is a real SSE1/2/3 engine instead of the somewhat hacked version in the older cores. Intel did this one really right.

The other big one is an early exit for some instructions, and if memory serves, idiv is one. What this means is that an instruction can be retired after eight or so stages of the 13 in the pipeline. This can make a huge difference for things like encoding movies and the like.

[K. Eng]

There are rumors of a set launch date now:

http://www.x86-secret.com/?option=newsd&nid=909

The article is mostly about Conroe and Merom, but it mentions at the bottom that Intel is looking for an official Napa platform announcement on January 5, 2006.

[Aroc]

Does that mean Q1 2006, all Dothan based notebooks will get a big price drop to be cleared out?

I wouldn't think so. Since the majority of users won't need dual-core. Although better chips for EVERYONE *is* desirable. Dual core included. And this is just more speculation, but I imagine, at least initially, demand might outstrip supply, especially with the initial pre-orders of the dual-cire thinkpads. So I'm speculating that dual-core, at least initially *may* be high-end, with single core dothans being mainstream, at least for the first half of 2006, if not for the whole year. The Celeron M might disappear, but I think single core Pentium M won't die that quick of a death. But that's just speculation. Maybe there will be a glut of mobile dual core chips? No one has a crystal ball.

Put me down for a dual-core Txx with UXGA or better screen.

[K. Eng]

http://pc.watch.impress.co.jp/docs/2005 ... biq121.htm

There are some interesting slides on this page with some details on the optimizations done to floating point and streaming instructions on Yonah.

Also note the dynamic allocation of the L2 cache. The cores use only as much or as little cache as they need.

[K. Eng]

http://www.theinquirer.net/?article=27770

Intel supposedly released Yonah clockspeeds, model numbers, and prices to OEMs

[K. Eng]

http://www.anandtech.com/systems/showdo ... i=2611&p=4

AnandTech apparently has a Yonah sample in their labs. In the above link, they compare the size of the Yonah core to the tri-core Xenon in the XBOX360.