Intel touts power of upcoming Yonah laptop chip

123457

Comments

  • Reply 121 of 144
    melgrossmelgross Posts: 33,570member
    Quote:

    Originally posted by TenoBell

    Can you over clock a Sony Vaio?



    I'm sure there would be problems with thin latop's, if you added more heat than it was originally designed for.




    I don't think they can. But some makers will allow it.



    Mostly though, it's for desktops.



    I wonder if ATIaccelerator works for laptop gpu's? Anyone try it?
  • Reply 122 of 144
    thttht Posts: 5,530member
    Quote:

    Originally posted by Chucker

    Huh? 64-bit (via EM64T) is nothing new to Intel CPUs, and pipeline length is just an evolutionary factor. Pipeline length changed during various versions of the G4, but the 7448 is hardly a "brand new architecture" compared to the 7400.



    That Merom supports this or that feature of existing Intel CPUs doesn't mean that Merom is derived from the P6 or P7 microarchitectures. I really doubt that Intel took EM64T features and put it onto an evolved P6 microarchitecture. I really doubt that they took the P7 (Pentium 4 variants) and just eliminated the half of the pipeline stages.



    Merom is a wider issue with better OOOE, 64-bit and slight longer pipeline. It's purported to be 30% better per clock than Yonah. That's a pretty big performance improvement (if true) and tells us it is more a brand new microarchitecture than an evolution of Yonah or Prescott.



    And yes, I would call all Moto 744x/745x CPUs a brand new architecture when compared to the 7400/7410. It's certainly not an evolution in the natural sense of the word.



    The 7400 was just a 750 with the AltiVec unit. It had other things like shared cached coherency and a more efficient FSB tacked on, but they were minor. Pipeline depth, issue width, units outside of AltiVec, all stayed the same between the 7400 and 750. All Moto did with the 7400 was take a 750 (G3) and add an AltiVec subsystem.



    Contrast this with the change from the 7400 to the 7450 which included a near doubling of the pipeline depth from 4 to 7, the addition of 2 integer units, an increase in issue/completion width from 2 to 3 (+1 branch), the AltiVec arithmetic unit was separated into 3 units, and all the others changes needed to support these changes. It was total reworking of the 7400 architecture to something, for the time, relatively modern.



    The fact that Motorola used a 7450 moniker instead of 7500 or something like that is mere marketing and branding for the AltiVec unit, which should really tell you something of where Moto's (now Freescale's) priorities were and are.
  • Reply 123 of 144
    melgrossmelgross Posts: 33,570member
    This is the last I'm going to add to this part of the discussion.



    Look at the chart.



    You will see where things are going, and what is derived from what.



    "Derived from" doesn't mean "the same as". It means "the logical progression from".



    It's the concept that Intel is illustrating here. And that is what I'm saying as well.



    You have to click on the charts yourselves, because there isn't a page heading associated with them. Sorry.



    http://arstechnica.com/articles/paed...f-israel.ars/2
  • Reply 124 of 144
    Quote:

    Originally posted by kim kap sol

    True...but going any lower would tarnish their image. iPods are expensive as hell and...they sell...extremely well.*



    iPods are one of the cheaper media players out there. Cheaper than Creative for one. There's nothing really to touch the Nano out there for price and the iPod is equivalent or cheaper than most of it's competitors.



    But anyway, Back in 2001 I bought an iBook for about £1700. The equivalent model is now £699. Back then equivalent PC laptops were about £1200, now they are about £600 and probably faster too. I've seen laptops at £349 here in the UK, although they were particularly nasty. The iBook is a lot cheaper looking now than the initial ice books as the plastic is now opaque instead of the silver mid section and translucent plastic case.



    IMHO Apple have come down market considerably in the last 5 years.
  • Reply 125 of 144
    melgrossmelgross Posts: 33,570member
    Quote:

    Originally posted by aegisdesign

    iPods are one of the cheaper media players out there. Cheaper than Creative for one. There's nothing really to touch the Nano out there for price and the iPod is equivalent or cheaper than most of it's competitors.



    But anyway, Back in 2001 I bought an iBook for about £1700. The equivalent model is now £699. Back then equivalent PC laptops were about £1200, now they are about £600 and probably faster too. I've seen laptops at £349 here in the UK, although they were particularly nasty. The iBook is a lot cheaper looking now than the initial ice books as the plastic is now opaque instead of the silver mid section and translucent plastic case.



    IMHO Apple have come down market considerably in the last 5 years.




    Yeah, and here's a good, and surprising article fron theInquirer. When I emailed him my surprise at such a nice article about Apple, after fighting with them for a while about their Mac coverage, he said that he was surprised as well.



    But it's here:



    http://www.theinquirer.net/?article=28430



    To a great extent, the article is correct.
  • Reply 126 of 144
    Quote:

    Originally posted by melgross

    This entire "new" line of chips actually goes back to the P III line. That is the line that Intel's "M" series of chips has been based upon.





    The Pentium III wasn't a new architecture. It was a P6 core. Ie. from the Pentium Pro with bits hacked off, deeper pipelines and MMX.





    Quote:

    Originally posted by melgross

    The Netburst comes from the PIII successor, the P4 (Pentium 4). After the first half of the yesr, that will fade out quickly.





    Netburst is the bus architecture of the P7 core. There's other things to it than the bus.



    Quote:

    Originally posted by melgross

    But the new line is not really "new". It's refined, added to, etc., but it comes from the older line of the Pentium 3.





    Until we see Hannibal do a review then I'd be inclined to say not. Intel are saying it's a new architecture and they're claiming much for the new architecture beyond what would be achievable with simple process changes to the Yonah. They're rumoured to be dropping the Pentium name from them even.



    Quote:

    Originally posted by melgross

    That ended up as a parallel development path. Straight up, for desktops, workstations, and servers, came the Pentium 4, evolving into the Prescott, and its successor.



    But from the Pentium 3 also came the "M" (Mobile) line. This is the ansestor to the line that we will be seeing in 2006.



    Intel can use as many numbers to designate it as they like, but the developmental line is well known.



    From the Pentium 1 came the Pentium 2, came the Pentium 3, came both the "M" AND the Pentium 4.



    Let me make what I'm saying a bit clearer.



    While the circuitry within the Merom will be new ('from the ground up"), the design philosophy is continued from the "M" line. A short (relatively) pipeline of about 14 stages. dual core, shared L2, and low power design. Same memory bandwidth problems.




    That's just too simplistic.



    There's major architectural differences between P5 and P6. For instance, the P5 is in-order whereas the P6 is out-of-order.



    You may as well have said that the 8086 was the same architecture as the pentium.
  • Reply 127 of 144
    Quote:

    Originally posted by THT

    The fact that Motorola used a 7450 moniker instead of 7500 or something like that is mere marketing and branding for the AltiVec unit, which should really tell you something of where Moto's (now Freescale's) priorities were and are.



    Hadn't they already allocated 75xx for their failed 'G5'?



    The e600 core in 7448 added some more architecture twists such as OOOE in the AltiVec unit that weren't in 7447.
  • Reply 128 of 144
    melgrossmelgross Posts: 33,570member
    Quote:

    Originally posted by aegisdesign

    [B]The Pentium III wasn't a new architecture. It was a P6 core. Ie. from the Pentium Pro with bits hacked off, deeper pipelines and MMX.



    I didn't say that the PIII was a new architecture. I'm just saying that it is the direct line to the "M".







    Quote:

    Netburst is the bus architecture of the P7 core. There's other things to it than the bus.



    Yes. I agree with that.





    Quote:

    Until we see Hannibal do a review then I'd be inclined to say not. Intel are saying it's a new architecture and they're claiming much for the new architecture beyond what would be achievable with simple process changes to the Yonah. They're rumoured to be dropping the Pentium name from them even.



    We're all waiting for John. I never said that it was a simple process change from Yonah. I said just the opposite. What I said was that it was "derived" from Yonah, as in conceptually. Not circuit by curcuit.



    You can see from Intel's own chart where I'm going with this.



    We really have to be careful about companies stating that their "new" designs are as new as they are saying they are. Every new design builds upon, and takes from older ones.



    And names mean nothing. Did you know that the only reason Intel used the name Pentium was to prevent rival x86 chip makers from using the name "586"? That was to be the name. Intel wanted a name that they could trademark, that no one else could use. It means nothing. The same is true of these "P" names. you have to be careful about thinking that just because a different number is assigned to a line that it's as different as Intel wants us to believe.





    Quote:

    That's just too simplistic.



    It's simplistic, true. But it is still the line of evolution. There have been minor deviations, and additions, but it is basically correct.



    Quote:

    There's major architectural differences between P5 and P6. For instance, the P5 is in-order whereas the P6 is out-of-order.



    Sure, there was evolution there.



    Quote:

    You may as well have said that the 8086 was the same architecture as the pentium.



    I think that we can all agree that the Pentium is pretty far down the line of evolution from the 8086 or, for that matter, the 4004. But it still contains code from the 8086, and even before.
  • Reply 129 of 144
    chuckerchucker Posts: 5,089member
    Quote:

    Originally posted by THT

    The 7400 was just a 750 with the AltiVec unit. It had other things like shared cached coherency and a more efficient FSB tacked on, but they were minor.



    Oh, and something minor like multi-processing. Yeah, the changes were minor. Mhm!
  • Reply 130 of 144
    hirohiro Posts: 2,663member
    Quote:

    Originally posted by Chucker

    Oh, and something minor like multi-processing. Yeah, the changes were minor. Mhm!



    Ummm, that's the cache coherency THT mentioned.



    The first 7400 was a relatively minor change as far as chip atrchitectures go.
  • Reply 131 of 144
    chuckerchucker Posts: 5,089member
    Quote:

    Originally posted by Hiro

    Ummm, that's the cache coherency THT mentioned.



    How is allowing for SMP a "minor" change?



    Quote:

    The first 7400 was a relatively minor change as far as chip atrchitectures go.



    I would argue the 7400 is closer (features-wise) to the 604 than to the 750.
  • Reply 132 of 144
    lupalupa Posts: 202member
    Quote:

    Originally posted by melgross

    I don't think they can. But some makers will allow it.



    Mostly though, it's for desktops.



    I wonder if ATIaccelerator works for laptop gpu's? Anyone try it?




    Yep, works beautifully.



    Mobility Radeon 7500 (on a G3 iBook)



    Precessor: +12.5%

    Memory: +12.5%



    The settings are very conservative.
  • Reply 133 of 144
    Quote:

    Originally posted by melgross

    From the Pentium 1 came the Pentium 2, came the Pentium 3, came both the "M" AND the Pentium 4.



    Actually the Pentium & Pentium MMX were closer to the 486 design, whereas the Pentium Pro was a completely new core design -- the first with a decoder on the front end cracking instructions, and an L2 cache connected to the chip (although in the case of the PPro it was actually a second chip in the same package, making it very expensive and aimed only at servers). The Pentium2 was the PentiumPro with MMX added and the L2 on-die. The Pentium3 added SSE. The Pentium-M is an evolution of the Pentium3 core, although a lot has been done at the circuit level to reduce the power requirements and allow higher clock rates. The Pentium4 was a completely new core (NetBurst architecture) and intended to support pushing the clock rates sky high -- only to discover the wall at 3.8 GHz, causing Intel to change course and effectively abandon that direction. They are now evolving the Pentium-M with Yonah and the rest of the upcoming family. I expect we'll see HyperThreading, IA-64, and SSE3+ added to this line since it is becoming Intel's flagship. The future more than about 2 years out is a big question, however, no matter what the public roadmap might say.
  • Reply 134 of 144
    Quote:

    Originally posted by THT

    The 7400 was just a 750 with the AltiVec unit. It had other things like shared cached coherency and a more efficient FSB tacked on, but they were minor.



    Uh, well I wouldn't categorize the MPX bus as a "minor" improvement. Slag it though you might it quadrupled the 60x bus' practical throughput, and without it the AltiVec unit would have been completely useless. The multi-processor capability was also pretty damn important to Apple for a couple of years. The other thing you are forgetting is the improved FPU which substantially outperforms the G3's FPU on double precision. The 7400 was a considerable improvement from the G3 even though it continued with the basic in-order, direct-decode style architecture.



    The G5, on the other hand, was a completely different architecture.
  • Reply 135 of 144
    melgrossmelgross Posts: 33,570member
    Quote:

    Originally posted by Programmer

    Actually the Pentium & Pentium MMX were closer to the 486 design, whereas the Pentium Pro was a completely new core design -- the first with a decoder on the front end cracking instructions, and an L2 cache connected to the chip (although in the case of the PPro it was actually a second chip in the same package, making it very expensive and aimed only at servers). The Pentium2 was the PentiumPro with MMX added and the L2 on-die. The Pentium3 added SSE. The Pentium-M is an evolution of the Pentium3 core, although a lot has been done at the circuit level to reduce the power requirements and allow higher clock rates. The Pentium4 was a completely new core (NetBurst architecture) and intended to support pushing the clock rates sky high -- only to discover the wall at 3.8 GHz, causing Intel to change course and effectively abandon that direction. They are now evolving the Pentium-M with Yonah and the rest of the upcoming family. I expect we'll see HyperThreading, IA-64, and SSE3+ added to this line since it is becoming Intel's flagship. The future more than about 2 years out is a big question, however, no matter what the public roadmap might say.



    I would pretty much agree with that.



    I do have an aversion to the words "completely new" though. Greatly changed would be preferable, I think.
  • Reply 136 of 144
    thttht Posts: 5,530member
    Quote:

    Originally posted by Chucker

    Oh, and something minor like multi-processing. Yeah, the changes were minor. Mhm!



    The 750 supported multiprocessing, it just didn't have a shared cache coherency mode which was a big win MP efficiency. It wasn't much of a change.
  • Reply 137 of 144
    thttht Posts: 5,530member
    Quote:

    Originally posted by Programmer

    Uh, well I wouldn't categorize the MPX bus as a "minor" improvement. Slag it though you might it quadrupled the 60x bus' practical throughput, and without it the AltiVec unit would have been completely useless.



    I would count it as minor. They added data streaming techniques (bus pipelining and out-of-order data transfers) to make the bus more efficient, otherwise not much difference to the 60x bus. Big whoop. They did the minimum necessary to the bus performance in order to make AltiVec useful. With all that, Intel's bus performance was still better at the time.



    If they added DDR to the MPX mix, yes, maybe that could be called something major.



    Quote:

    The multi-processor capability was also pretty damn important to Apple for a couple of years.



    Important to Apple, but minor in terms of microarchitecture design. If IBM could have evolved the 604e, maybe Apple could have something in the high end, but with them being "stuck" with the G4 for its high end, I can fully agree with you that it was important to Apple. But as a microarchitecture design feature, no.



    Quote:

    The other thing you are forgetting is the improved FPU which substantially outperforms the G3's FPU on double precision.



    G4 double precision FPU has 3 cycle latency compared to 4 cycle latency for the G3 double precision FPU (and virtually everything else stayed the same). That's pretty minor to me. Adding a second FPU unit would be major to me.



    Quote:

    The 7400 was a considerable improvement from the G3 even though it continued with the basic in-order, direct-decode style architecture.



    At the same clock rate, the 7400 and 750 had the same scalar integer and FPU performance, did it not? I don't remember seeing many benchmarks displaying the 7400 CPU's substantial performance increase in FPU.



    Quote:

    The G5, on the other hand, was a completely different architecture. [/B]



    Yes. And the change from the 7400 to the 7450 was more of change than the change from the 750 to the 7400, no? The 7450 core design hasn't changed all that much in 4 years. The latest 7448 is quite similar to the 7440 except for the aforementioned OOOE AltiVec instructions and on-die L2.



    The big question is, is Merom a completely different x86 architecture, much like Willamette/Northwood/Prescott was a completely different architecture from the Pentium 3. I think all signs point to yes, especially if the rumored +30% performance improvement per clock over Yonah is true.
  • Reply 138 of 144
    Quote:

    Originally posted by melgross

    I would pretty much agree with that.



    I do have an aversion to the words "completely new" though. Greatly changed would be preferable, I think.




    Pentium4 was apparently a "from scratch" design which makes it as completely new as any chip that can execute the x86 instruction set could be. Completely different design team created it.





    Quote:

    Originally posted by THT

    I would count it as minor. They added data streaming techniques (bus pipelining and out-of-order data transfers) to make the bus more efficient, otherwise not much difference to the 60x bus. Big whoop. They did the minimum necessary to the bus performance in order to make AltiVec useful. With all that, Intel's bus performance was still better at the time.



    If they added DDR to the MPX mix, yes, maybe that could be called something major.




    The 60x bus couldn't really sustain more than ~250 MB/sec, IIRC. The MPX bus started at >750 MB/sec and is now up to >1.3 MB/sec. Sure it doesn't stand up to the Intel buses that came after, but for a while it was the fastest bus out there -- and that was while the 7400 was around and competitive. That was pretty major.



    I'll give you the MESI one -- not a lot of chip design changes there.



    Quote:

    G4 double precision FPU has 3 cycle latency compared to 4 cycle latency for the G3 double precision FPU (and virtually everything else stayed the same). That's pretty minor to me. Adding a second FPU unit would be major to me.



    You are forgetting 1 cycle throughput vs. 2 on the G3. That means double the double precision performance. Adding a second FPU doubles it again, but I wouldn't ignore the first doubling.



    Quote:

    At the same clock rate, the 7400 and 750 had the same scalar integer and FPU performance, did it not? I don't remember seeing many benchmarks displaying the 7400 CPU's substantial performance increase in FPU.



    Roughly the same, but I'm much more interested in FLOPS or bandwidth oriented operations.



    Quote:

    Yes. And the change from the 7400 to the 7450 was more of change than the change from the 750 to the 7400, no? The 7450 core design hasn't changed all that much in 4 years. The latest 7448 is quite similar to the 7440 except for the aforementioned OOOE AltiVec instructions and on-die L2.



    No argument there. Moto has mainly been trying to tweak the process, but has been increadibly slow at it because they don't own the fabs and they lost all their good people to Intel.



    Quote:

    The big question is, is Merom a completely different x86 architecture, much like Willamette/Northwood/Prescott was a completely different architecture from the Pentium 3. I think all signs point to yes, especially if the rumored +30% performance improvement per clock over Yonah is true.



    I believe they are following the general P3/PM strategy, but have made enormous changes (more internal resources, more execution units, pipeline length changes, etc). That makes it a new architecture in my book.
  • Reply 139 of 144
    Quote:

    Originally posted by melgross

    Regarding the Yonah speeds and power usage, this is interesting.



    I just wish that Apple would let us overclock the way PC makers often do.



    I know that it would lead to possible stability problems. But those who do it understand this.



    I've had many upgrade cards for my Mac's over the years that let me do this. It's not impossible.



    In light of this article of course:



    http://www.theinquirer.net/?article=28431



    And for the old Netburst we've been talking about:



    http://www.theinquirer.net/?article=28412






    melgross, into overclocking? say it ain't so brotha...!!
  • Reply 140 of 144
    Quote:

    Originally posted by mjteix

    No, only the first model is single-core (#756), the 4 others are dual-core (#T1x00).

    It seems that the old anandtech rumours about which yonah chips will be available in january are "true":

    one single core @ 1.66GHz, and some dual-cores from 1.5 to 2.16GHz...

    But it could be a very old speculation document, as it contains old Intel part numbers.



    I still believe Apple could launch a new 13" iBook as well as a new 13" PowerBook in January:

    - 13" iBook single-core Combo $999 (in black or white plastics)

    - 13" iBook single-core SD+bigger HD $1199 (in black or white plastics)

    - 13" PowerBook dual-core SuperDrive $1499 (metallic enclosure)




    what about the 14" wide iBook? i hope it has a 5400rpm drive, that makes a significant difference over 4200rpm, particularly if it is going to be (the iBook) dualcore 1.6ghz or so....
Sign In or Register to comment.