Comments Locked

59 Comments

Back to Article

  • quidpro - Monday, February 26, 2007 - link

    What is Glaze? google is not helping.
  • Hoser McMoose - Wednesday, January 31, 2007 - link

    The calculation for the number of transistors for cache is off by a fairly significant amount in this article! The article states 288M transistors for 6MB of cache, which is FAR too low!

    Actual transistors for 6MB cache, if we assume ZERO redundancy is:

    6MB * 1024KB/MB * 1024B/KB = 6,291,456 bytes

    6,291,456 bytes * 9 bits/byte = 56,623,104 bits (remember ECC makes it 9 bits, not 8)

    56,623,104 bits * 6 bits/transistor ~= 340M transistors

    In reality it is actually probably more like 350 or 360M transistors required for 6MB of L2 cache since you always want a bit of redundancy. It's a bad thing if you have to throw out an entire chip due to a single bit error in your 6MB of cache!

    Now, taking those numbers, if we assume that the 410M Transistor figure is accurate, the change in transistors from Conroe to Penryn is a bit different.

    Conroe has 291M transistors of which at least 226M are L2 cache. That leaves only 65M assuming no redundancy, or maybe more like 50M transistors with a few redundant cache cells. According to www.sandpile.org (and I have no idea what their source is) each core of the Core 2 Duo has only 19M logic transistors. If this is accurate then it's actually HUGELY impressive, since that is WAY less then either the Pentium 4 or the Athlon64.
  • Murst - Monday, January 29, 2007 - link

    "Intel isn't disclosing how thick the Hafnium gate dielectric layer is in its 45nm transistors, but we do know that it is thicker than the 1.2nm SiO2 gate dielectric used in previous 90nm and 65nm transistors."

    Why would intel make the gate larger in a shrink? Is this a typo, or is there some huge benefit to this?
  • IntelUser2000 - Tuesday, January 30, 2007 - link

    quote:

    Why would intel make the gate larger in a shrink? Is this a typo, or is there some huge benefit to this?


    The reason is that making the gate dielectric thinner than 1.2nm is not a good idea. 1.2nm dielectric is already equivalent to 5 atoms in thickness. Electrons can literally squeeze through the dielectric "insulator" because the dielectric is so thin. Because the material used for the dielectric layer is different from the 65nm generation, they can make the dielectric thicker without sacrificing speed characteristics. So they can increase speed AND have substantially less leakage.
  • Visual - Monday, January 29, 2007 - link

    So, I'm a bit out of the loop lately, can anyone brief me up?
    Are there any new sockets, chipsets, ram changes or anything expected together with these new 45nm goodies? Or will I be fine getting a new system come Q2, and upgrading to Penryn when it's out and affordable?
  • ceefka - Monday, January 29, 2007 - link

    I wonder if booting OXS with the new chip indicates that PC-users can soon use OSX too without having to buy a Mac. That would allow to mix and match a bit more with apps.
  • Netopia - Sunday, January 28, 2007 - link

    quote:

    Intel is actually seeing good feature scaling with its 45nm process, quoting a ~2x improvement in transistor density. In other words, if you took a 100mm^2 65nm chip and built it on Intel's 45nm process, it would be roughly a 50mm^2 chip after the shrink.


    100 * 100 =10,000
    50 * 50 = 2,500

    That's a 4x improvement, not a 2x!

    Joe
  • coldpower27 - Sunday, January 28, 2007 - link

    That is not how it works...

    65 x 65 = 4225
    45 x 45 = 2025

    2025/4225 = 48% the size of the 65nm part, so close to 2x smaller.
  • IntelUser2000 - Sunday, January 28, 2007 - link

    quote:

    That is not how it works...

    65 x 65 = 4225
    45 x 45 = 2025

    2025/4225 = 48% the size of the 65nm part, so close to 2x smaller.


    That's the theory. Intel's 45nm 6T SRAM size is 0.346um, and 65nm SRAM cell size is 0.57um. That's only 39% reduction(61% of the size). In addition, SRAM, which is used for cache, has the best potential for scaling. The cores will have likely worse scaling than 61%.
  • coldpower27 - Monday, January 29, 2007 - link

    Yes, I was just presenting the mathematical example of the calculation, I wasn't going into having real world factors involved. Since in this article Anandtech simplified things with just the 2x smaller statement.
  • tygrus - Sunday, January 28, 2007 - link

    100mm^2 is twice the area as 50mm^2.

    100mm^2 means the mm is squared not the 100.
    10mm x 10mm = 100mm^2
    5mm x 10mm = 50mm^2
  • lopri - Saturday, January 27, 2007 - link

    I am more interested in the long-awaited introduction of integrated memory controller from Intel CPU. IIRC, it's supposed to happen with 45nm process, which makes me curious whether Nehalem will be the one. This will likely happen in parallel with HSI, or High Speed Interconnect, which is supposed to be compatible with PCI-Express 2. (How that can be done - beats me) Nehalem might as well be 'simply' Core 2 Duo + IMC as Hammer was Athlon XP + IMC.

    While this is obviously Intel's PR piece, it is an undeniable truth that AMD doesn't even have anything for such. It's foolish to think that AMD wouldn't be bragging had they have made an advance like this.

    A good news for AMD is that the processing power for desktop has, to my eyes, reached to the point that it's sufficient for majority of daily computing for forseable future. In other words, majority of people will not likely take advantage of any more computing power (be it speed or multitasking) than what current CPUs can process. I wonder AMD will turn to a totally different direction from this hopeless speed race.
  • IntelUser2000 - Sunday, January 28, 2007 - link

    quote:

    This will likely happen in parallel with HSI, or High Speed Interconnect, which is supposed to be compatible with PCI-Express 2. (How that can be done - beats me) Nehalem might as well be 'simply' Core 2 Duo + IMC as Hammer was Athlon XP + IMC.


    HSI?? High Speed Interconnect?? That's a new one. No, its CSI, and it stands for Common System Interface, which was originally named because it was supposed to bring socket compatibility between Itanium and Xeon, or at least the electrical part of it.
  • lopri - Sunday, January 28, 2007 - link

    Got me there. Yah it's CSI. Thanks for the correction.
  • MadBoris - Saturday, January 27, 2007 - link

    Maybe I am missing something here, but where is the chipset information.

    Is this going to be on a new chipset, or 775 compatible?

    Knowing Intel, it's a new chipset, they love to stimulate the economy (their economy) with new chipsets all the time.

    Any word?
  • Groovester - Saturday, January 27, 2007 - link

    Due to HK+MG's 20% higher switching rate and much lower leakage currents, shouldn't we expect ("more power efficient processors") both higher clock speeds AND lower TDP, even with a slightly higher transistor count?
  • IntelUser2000 - Sunday, January 28, 2007 - link

    quote:

    Due to HK+MG's 20% higher switching rate and much lower leakage currents, shouldn't we expect ("more power efficient processors") both higher clock speeds AND lower TDP, even with a slightly higher transistor count?


    No. First thing is the recent roadmaps shown by HKEPC say it doesn't lower power, part of the reason is to increase yields, the other has to be clock speeds.

    (If you want to read further, go ahead, but it tends to be long)

    It does present much lower leakage, but there is a caveat: "Intel is quoting a greater than 20% improvement in switching speed compared to its 65nm transistors. At the same speed as its 65nm transistors, there's a greater than 5x reduction in source-drain leakage power and a greater than 10x reduction in gate oxide leakage power"

    The source-drain is the reason for leakage power when the transistor is active(meaning its on), and gate oxide leakage power is for idle. Read the second sentence carefully: "At the same speed as its 65nm transistors...

    It means if Intel wants to take advantage of the 20% increase in clock speed, they'll likely have to forgo the reduction in active leakage power. Contrary to what most news sites wants you to believe, leakage isn't of a big problem as they want you to believe(most of the reason is that they want hits to the webpage), especially on a power optimized CPU like Conroe. The exponential leakage increase every shrinking process would have been true if the companies like Intel did nothing about it. Of course they can't allow that to happen and added their own optimizations and enhancements. Its not a complete elimination either, but a 5x reduction.

    If they want to keep the 65W, we'll have to base it on the E6700, which is at 2.66GHz. With 1333MHz FSB, I can see mainstream dual core devices clock at 3.66/3.83GHz. Since it seems Intel is using a "family TDP", the 65W figure would likely have to be made to fir the 3.66/3.83GHz part. They may be able to tweak it higher and go 4.0GHz or lower power, but that tweaks will probably be seen on the Nehalem derivatives.

    2.66GHz x (20% faster) = 3.2GHz
    30% lower power = 0.7 But

    0.7 x 1.44(going beyond 3.2GHz will require greater than nominal voltages, which means more power. But faster clock speeds also mean more power. 20% higher voltage x 20% higher clock = 44% higher power)~1.00

    Sure some of you guys like to lower voltages below nominal and don't overclock, but possible sacrifice in stability won't be seen as good for the corporate market and Intel generally.

    So approximately same power consumption

    Now if you kept up to the roadmaps you might say, what about the 1333MHz 3.0GHz parts that's also said to be at 65W?? First, look at the X6800 with 75W TDP. Well, since the 3.0GHz/1333MHz FSB parts come a year later after Conroe, Intel will have time to tweak it.
  • IntelUser2000 - Saturday, January 27, 2007 - link

    If we derive clock speed numbers from Anandtech's Spring IDF 2005 article about Presler, we should be able to come up with a similar number :).

    http://www.anandtech.com/showdoc.aspx?i=2362">http://www.anandtech.com/showdoc.aspx?i=2362

    "It looks like the chip was running at 2.0GHz, which isn't too impressive in itself, until you remember that we're talking about a 65nm processor here whose launch is over a year away."

    45nm Quad core Desktop parts running at 1.86GHz
    45nm Dual core desktop parts running at 2.13GHz

    Now Presler came with 3.6GHz for the mainstream(Pentium D 960) and 3.73GHz with the enthusiast(Pentium Extreme Edition 3.73GHz). If we apply similar increase between prototype and final product, we get:

    45nm Quad core desktop: 3.46GHz @ 1066MHz FSB/3.5GHz @ 1333MHz FSB(45nm parts support half step dividers)

    45nm Dual core desktop: 4.0GHz @ 1333MHz FSB

    Since the TDP is not supposed to go down, we can expect all the power headroom to be translated to clock speed. The estimations also match the roadmaps.
  • Staples - Saturday, January 27, 2007 - link

    A price drop. It is ridiculous that C2Ds have not dropped a cent since they came out 6+ months ago.
  • MarkM - Saturday, January 27, 2007 - link

    http://www.nytimes.com/2007/01/27/technology/27chi...">http://www.nytimes.com/2007/01/27/techn...63262&am...
  • mostlyprudent - Saturday, January 27, 2007 - link

    Do we have any info as to chipset/socket compatability for Penryn?
  • MarkM - Saturday, January 27, 2007 - link

    Whoa Nelly!!! They JUST achieved a first validation sample of the material, what, SIX HOURS AGO? Lord knows when they'll even have chipset details figured out, and a shipping CPU is still a year off.

    This is a moment to step back and marvel at the continued long term achievement of the semiprocessor industry, not obsess how your going to run Vista!
  • mino - Saturday, January 27, 2007 - link

    Well, AMD stated that K8L will be drop-in compatible with SocketF/AM2 insrastructure in May 2006(if not sooner).
    That seems pretty much a yer ago from expected shipment date of mid-2007.

    This was presented not only to public but also to its corporate customers! ...

    To answer the original question:
    Do not expect penryn to work with anything older than 965 series of chipsets on boards introduced by 2H07 or later.
  • mino - Saturday, January 27, 2007 - link

    Meaning, ofcourse, that one should not expect penryn to be drop-in compatible with any current boards.
  • Andrwken - Monday, January 29, 2007 - link

    Actually, considering they are using unmodified hardware to run the chips. I would be rather surprised if they did not work on any core 2 capable motherboard. Correct me if I'm wrong, but a well researched motherboard purchase could get you a mobo that can run any 775 socket Prescook, Core 2 (dual or quad), and possibly Penryn? That's a pretty good lifecycle for a socket. How many has amd had in the last 2 years?
  • JarredWalton - Monday, January 29, 2007 - link

    I believe Penryn is scheduled for launch around the same time as Bearlake, and in the past we have seen a lot of official Intel platforms require a new chipset - whether or not the requirement is strictly necessary is obviously a different story. I believe Core 2 Duo is supposed to "require" P/G/Q965 or 975X, or one of a few select 945/946 chipset versions. Obviously, with 865 and 915 boards out there supporting the new CPUs, Intel's requirement was not strictly necessary, but most boards for C2D still use 965 or 975. What will happen with Penryn? I have no doubt we'll see some 965/975 boards that support it, but I'll be pleasantly surprised if the official requirement doesn't include Bearlake.
  • Viditor - Tuesday, January 30, 2007 - link

    quote:

    I believe Penryn is scheduled for launch around the same time as Bearlake

    I believe that Bearlake is out 3-6 months before Penryn (Q2-3 for Bearlake), but your point is taken...
  • mostlyprudent - Saturday, January 27, 2007 - link

    Sorry for the pragmatic obsessiveness...cannot be helped!
  • MarkM - Saturday, January 27, 2007 - link

    No problem, I hope I didn't come off sounding more negative than I meant to be! Anyway I just happened to stumble on a partial answer to your question:
    quote:

    Penryn is still not without its mysteries; a primary concern for enthusiasts is motherboard and socket support. Penryn will launch on Socket 775 -- meaning existing motherboards can physically harbor the new CPU but electrically might not. "Motherboard developers will have to make some minor changes to support [Penryn]. We can't guarantee that a person could just plug the chip into every motherboard on the market today." However, Smith also claimed the Penryn boot test that grabbed so many headlines last week occurred on unmodified hardware that included a notebook, several desktop motherboards and several server motherboards.
    http://www.dailytech.com/article.aspx?newsid=5869">link
  • flyck - Saturday, January 27, 2007 - link

    IBM and AMD (who work together) already stated that they have a simular technology, http://www.eetimes.com/news/semi/showArticle.jhtml...">link

    So don't say doomed yet -_-
  • Pythias - Saturday, January 27, 2007 - link

    Thats good. We don't want anyone in this race "doomed".
  • stmok - Saturday, January 27, 2007 - link

    Agreed. We don't want to return to the era of a "one horse race" and prices are high!

    They must challenge each other and as the end-users, we cheer for falling of prices! :)
  • cornfedone - Saturday, January 27, 2007 - link

    When you've been getting your butt kicked for over two years by AMD and your response is to add cache to inferior CPU designs, everyone pretty much knows you're desperate for good press. Intel's 65nano hype is being closely followed by their 45 nano hype as the never ending Intel PR machine marches on. Intel knows full well that Barcelona will trump Clovertown so they are trying to convice the naive that Intel is the future and not to be swayed by AMD's superior CPU architecture ready to break cover. When all the hype is said and done you can expect AMD to be doing just fine thank you and still providing leading edge CPUs for the masses.
  • verndewd - Sunday, January 28, 2007 - link

    the walk is the walk,Intel is dominating the perf turf.We all like to see equality in competition,but sometimes its not accessible.K8 had a great run,and it was an eyeopener for intel.At the same time i think its awesome to see intel flex their intelligence and dominate so decisively.I am an am2 user,I just love how this perf war is turning the tech world into an incredible battle of intelligence.
  • Bonesdad - Saturday, January 27, 2007 - link

    Cramitpal? Is that you?
  • shabby - Saturday, January 27, 2007 - link

    Can you post some barcelona benchmarks? Thanks.
  • stmok - Saturday, January 27, 2007 - link

    cornfedone is the same weirdo that talks about putting hackers in jail when it comes to breaking DRM!

    Seriously, you have to be pretty passionate about AMD to say something like this.

    I personally don't care of brand. I flip-flop to any brand, as long as they meet my requirements on specifc roles.

    I usually I wait until Anandtech and TechReport starts publishing benchmarks before forming an opinion. (Although I wish they did more Linux articles!)
  • bamacre - Saturday, January 27, 2007 - link

    LOL, looks like some hasn't sold their AMD stock and moved their money into Intel. Better hurry. :D
  • Viditor - Saturday, January 27, 2007 - link

    quote:

    looks like some hasn't sold their AMD stock and moved their money into Intel


    I think selling AMD to buy Intel is a really bad idea...as is the opposite.
    Neither stock is going to go up for awhile, and they're both going to do the same as long as the price war continues...
  • bamacre - Saturday, January 27, 2007 - link

    I disagree. I predict Intel's stock to climb 80% in 12 months. Assuming no terrorist attacks, invasion of Iran. If somone nukes Nigeria, however, that number will be closer to 125%.
  • Viditor - Tuesday, January 30, 2007 - link

    quote:

    I predict Intel's stock to climb 80% in 12 months

    It's amazing the number of people who have said that since Conroe was introduced...
    In point of fact, Intel has dropped over the last year...
    Frankly, neither Intel nor AMD will be making much headway while there's a price war on, though AMD will probably regain their losses in the second half of the year.
    Intel shares should be close to what they are now in a year (unless of course AMD has a terrorist attack) because (as they guided) their Gross Margins aren't expected to rise at all over the next year.
  • blackbrrd - Tuesday, January 30, 2007 - link

    Actually Intel released the Core 2 Duo around Jul. 2006. At the time a share was worth around 18$ The current value on an Intel share is roughly 21$ (Jan. 30th 2007). In other words an increase of about 16% since they released the Core 2 Duo.
  • Viditor - Tuesday, January 30, 2007 - link

    Most of that was seasonality...
    At the beginning of 2006, Intel was $25.05...so year-on-year (and in a year that Intel released and had 6 months of Conroe), Intel dropped ~20%...

    The key point is that while technology is important, Gross Margins are far MORE important!
  • BladeVenom - Saturday, January 27, 2007 - link

    I wouldn't be so bullish on Intel until you've seen AMD's next chip.
  • bamacre - Saturday, January 27, 2007 - link

    I'm not worried about it. Intel is serious about being the performance leader right now. They know they have made mistakes, and they know what they have to do to keep the crown. And they also have the tools to keep it. Whatever AMD comes out with, Intel will counter. Just watch.
  • Slappi - Saturday, January 27, 2007 - link

    Wow 125%? You must of mortgaged your house and bought Jan 2009 calls then with all your money.

    Market Cap of 250-300 billion dollars for a company with falling margins, losing market share and estimates of 1.20-1.30 for 2007. That would give them a PE of 32 for a company in the middle of a price war.

    I don't think so.

    I wouldn't short or buy INTEL at this point. No Way. Too unpredictable right now. 35 to 45 within a year? Even if AMD went bankrupt and closed its doors I doubt you would see 45 in 12 months.

    I would however buy AMD and sell CCs on em around 17-18. R600 K8L should do them for around 18 in the next 6 months. You could make good money selling calls 10% out for a few months.
  • bamacre - Sunday, January 28, 2007 - link

    "falling margins, losing market share "

    That is changing as I type.
  • hubajube - Monday, January 29, 2007 - link

    quote:

    That is changing as I type.
    No it's not. He's spot on.
  • yyrkoon - Saturday, January 27, 2007 - link

    Spoken like a true, disgruntled AMD fan boi. (or 'person' if you like). When will you guys learn, it doesn't matter WHO makes the 'best' CPU, and that close competition means win, win, for whoever, using whatever.
  • JumpingJack - Sunday, January 28, 2007 - link

    quote:

    Spoken like a true, disgruntled AMD fan boi. (or 'person' if you like). When will you guys learn, it doesn't matter WHO makes the 'best' CPU, and that close competition means win, win, for whoever, using whatever.


    Yep, don't like one bit do they ---- hey, lets compare the OC of a 65 nm AMD chip to the OC of an Intel 65 nm chip ....

    Intel has removed the major stumbling block for device scaling that has plagued both 90 and 65 nm.... it only get's better from here.
  • Viditor - Tuesday, January 30, 2007 - link

    quote:

    hey, lets compare the OC of a 65 nm AMD chip to the OC of an Intel 65 nm chip


    OK...I choose Barcelona!
    God I love these macho/nerd games...it takes me bake to my College days of D&D with Tequila shooters! (just be careful with that Intel +5 Vorpal Blade though!)

    quote:

    Intel has removed the major stumbling block for device scaling that has plagued both 90 and 65 nm


    Huh? The stumbling block for scaling is Intel's platform (FSB)...
  • verndewd - Sunday, January 28, 2007 - link

    lets make it fair and get a 65nm firstgen intel D900 series ,ultimately thats where the real comparison is at.(dont you wish it were that simple? hehehe)
    unfortunately Thats not the current market due to amd transitioning later than intel.The new york times said 9 months,but the first 65nm intels arrived late in 05.How is that for curious?
  • archcommus - Saturday, January 27, 2007 - link

    Yeah, don't be a fanboy, buy whatever is the best purchase. My area of concern is that if AMD won't be delivering on par with Intel, it allows Intel to charge a price premium UNLIKE this past summer where we were lucky to see fierce price wars. If Intel gets too far ahead it could very well go back to the old days of Intel being the expensive mainstream chip and AMD being the cheaper choice for the DIYer.
  • Regs - Saturday, January 27, 2007 - link

    All I see is Intel advancement and AMD's foggy future.
  • SunAngel - Saturday, January 27, 2007 - link

    murder by numbers
  • smn198 - Monday, January 29, 2007 - link

    quote:

    100 * 100 =10,000
    50 * 50 = 2,500

    That's a 4x improvement, not a 2x!

    Joe


    Also, the Intel Xeon ad appears on top of the article text. Really need to get my ad removing plug-in installed again.
  • bombledmonk - Monday, January 29, 2007 - link

    it's 100mm^2, not 100mmx100mm. same for the 50mm^2
  • bombledmonk - Monday, January 29, 2007 - link

    whoops, wrong quote, and it had been answered below...
  • shaunmarsh - Saturday, July 28, 2012 - link

    This article is very good & informative.I have gain so much information from this blog.I like your blog.Thanks for the post.I am waiting for your new post.

Log in

Don't have an account? Sign up now