Pentium M out the door, what's next

by Anand Lal Shimpi on 3/24/2005 1:48 PM EST
Comments Locked

32 Comments

Back to Article

  • jrk - Tuesday, April 5, 2005 - link

    As some others have suggested: it's the drivers.

    Apple rolls their own drivers, with some limited help/input from ATI and NVIDIA. They have fewer resources than ATI or NVIDIA for this task, and they have largely focussed on flexibility to cleanly integrate with their window manager (which runs atop GL, and also has GL contexts within it).

    For several years, NVIDIA tried to convince Apple to let them supply drivers from their same unified codebase used for the Linux and Windows x86 drivers specifically in the interest of increasing the Mac performance of their hardware substantially over what the Apple drivers were able to achieve. At the time, however, Apple was too interested in making their own extensions for e.g. the window manager integration mentioned above to consider performance a top priority.

    There have recently been substantial rumors about Apple prioritizing graphics hardware performance in 10.4 and subsequent incremental releases. However, committed or not, I don't think Apple has either the resources or the expertise to ever reach the performance of NVIDIA's (or likely ATI's) own drivers on Windows. There are certainly a great many optimizations to be made to the system and drivers -- 50% performance loss is just ridiculous, especially on a game like Doom 3 which should be more bound by shader computation and framebuffer bandwidth (stencil shadows are bandwidth-limited) performance (internal to the chip/board) -- but I wouldn't expect Apple to ever surpass 90% of NVIDIA/Windows performance on comparable hardware.
  • Qardinal - Saturday, April 2, 2005 - link

    lol! Anand is biased, out of touch with his readers and advertisers, and pushes reviews out the door too fast? All because he didn't push overclocking the slowest PM to satisfy the vast number of people who wanna stick their PC in a unventilated closet (and to whom the few grams of weight difference means less time in the gym)?

    This would be truly funny, if I wern't so saddened by the realization that Anandtech must be about to go belly-up as a result of this. I just had no idea the UVCOs (unventilated closet overclockers)wielded such influence... but I guess it was only a matter of time before the crowd who's top priority of saving $2/month on electricity, managed to piggybank enough pennies to begin flexing their buying power on top-of-the-line technology =)

    I'm just glad I found out in time to come out of my (well ventilated) closet and make my 1st post in the 3 years I've been reading Anand... and say thanks for making me the most computer litereate Carpenter anyone has ever met (at least, acording to all the people I've helped) =)

    Disclaimer: author has a Pro-Anand bias ;)
  • Michael2k - Friday, April 1, 2005 - link

    The problem exists that when Anand 'pushes customer requirements' that "customers" don't agree with, the "customers" see bias.

    Note the anti-Apple crowd that rises with every Apple article.

    Or the people who complain about photo articles; or the people who complain about the PC-Club article.
  • Mark - Friday, April 1, 2005 - link

    I perceive bias when I see parroting of vendor marketing and vendor engineering messages rather than evidence that a writer has sought out contrary communities of thought (and credits those sources).

    Pitching AMD's opinion against Intel's, or vice versa, does not separate fact from fiction. I would love to see Anand "make the news" through capturing and pushing customer requirements. This proposition helps all of us loyal readers.
  • Michael2k - Thursday, March 31, 2005 - link

    Yes, especially since he has been pushing small, silent, and energy efficient systems lately; he's been pushing Macs for a little while now, and they do solve all the same problems the Pentium M can/could, but are cheaper and available now.
  • Someone - Thursday, March 31, 2005 - link

    Mark: Alright, so how does that make AnandTech biased?
  • Mark - Thursday, March 31, 2005 - link

    (Somebody)
    Anandtech's advertiser's have a different take than you. They like being in tune with consumers who buy technology for themselves and their families. They like decision makers in the business world who influence product purchases.

    Anandtech built a readership through his ability to position products. Early on, there was a focus on longevity for the components and products reviewed here. That early focus on how to get ahead of Moore's Law brought a tangible value to readers beyond, "what is the latest fad". The recommendation to snap up celeron 300A's and recognized quality chipsets kept the same PC in place longer, it saved money.

    Products come and go. The opportunity to sound off the call for Pentium M's in the desktop was lost on anandtech.com. If the call had been made more than a year ago, we would have this product today the way we want it. Now you have to wait for Intel to double up the chips and buyers must wait many months without products that have a longer term shelf life to fit today's requirements.

    The Pentium M is just too practical to pass up. I believe Anand was too myopic on this one. He tested the priciest M chip rather than testing the least expensive one next to it. In the early days he would not have missed the chance to OC the 725 chip alongside the fastest chip. A 725 looks more like a 300A in terms of dialing up performance.

    This little M chip and the processes and methodologies that brought it together saved Intel from a decline. Smaller, lighter more portable, with the potential for cost effective energy efficientcy is a driver in 225 countries and territories around the world. I would consider putting the Pentium M to work everywhere its capability yields results. In 1U appliance boxes or desktops to run firewalls, distributed net app servers, video systems that can hide in a closet without concern about ventilation changes or isolating circuits. Widespread adoption by established PC makers will have much impact for my ability to deploy it. I watch eagerly for vendors to standardize on the platform or its successor.

    If you really like all the noise and heat belching out of your desktop PC, and are eager give up real estate to accomodate larger boxes, I don't speak for you. You'll tire of the noise and heat eventually. Until then, tolerate a different point of view.
  • Someone - Tuesday, March 29, 2005 - link

    #24 (Mark): Do you have any clue what you are talking about? Things get changed all the time and companies make new announcements literally every day. It's not possible for Anand and the staff (or others for that matter) to update their reviews whenever the company decides to do something new and different.

    I think Anand would've a better idea of what his readers want since many of them contact him directly. Just because you disagree with something doesn't mean you can make such broad statements. 915 chipset is the board you want, majority may not want that.

    And please stop acting like an advocate for AnandTech's entire readership - you sound pathetic. Stop coming here if you don't like it.
  • Mark - Tuesday, March 29, 2005 - link

    Anand,

    Stop pushing reviews out the door. The ct-479 review was rushed. There are significant bios updates available that change the numbers.

    When this adapter hits the streets they are going to be gobbled up. ASUS has compatibility with 3 Motherboards today and two chipsets. Their PAGD1 Motherboard is on its way! It is a socket 478 on the 915 chipset. This is the board we want.

    If we wanted space heaters under our desks we would buy them! We want silent and powerful desktops!

    You're biased, and you are wrong about what your readers want!
  • Michael2k - Tuesday, March 29, 2005 - link

    #22:
    What about the power supply fans?
    What about the hard drives?

    The point of a rewiew/comparison is which is more effective? Which is more cost effective? Which is cheaper?

    Can you get a silent PC for less than $500? How about less than $1,000?
  • sheik124 - Monday, March 28, 2005 - link

    #21, here's your silent PC, a Zalman CNPS-7000A-Alcu CPU cooler, some thermally controlled case fans, and a Zalman VF700AlCu GPU cooler, tada, silence
  • Michael2k - Monday, March 28, 2005 - link

    I'm wondering if the silent PC guys at Anandtech (or Anand himself) would be willing to do some cost, performance, and quiet comparisons between the various Macs and PCs they have around? We just saw a $2k silent PC when I know there are at least 15 'silent' Macs for under $2k
    iBooks
    PowerBooks
    Mac mini
    eMacs
    iMacs
    PowerMacs
  • ViRGE - Saturday, March 26, 2005 - link

    #11, that's a definite possibility, though I have not seen any evidence that their Mac code is inefficient due to something they did. Certainly the fact that Doom3 is so slow in spite of it being a native OpenGL app in the first place puts a lot of emphasis on the drivers though.
  • red and black - Saturday, March 26, 2005 - link

    #16,18: I did a little google searching and found the following benchmarks, which under linux show the high-end A64s totally demolishing high-end P4s:
    http://www.linuxhardware.org/article.pl?sid=04/10/... http://www.linuxhardware.org/article.pl?sid=05/02/...

    Of course the linuxhardware.org people did a much less thorough job than the anandtech folks would, but hey, it's something. :)
  • Someone - Saturday, March 26, 2005 - link

    #16: I don't know if Anand would be inclined to do that. It depends on how much of readers are willing to take a look at such an exclusive topic. AnandTech has to do articles that will appeal to the mass readers instead of a selected few.
  • Heron - Saturday, March 26, 2005 - link

    #13, I get what you mean. However, you still have to remember that Pentium 4 still holds much more advantage due to the video encoding and the floating point performance. A dip in performance from Northwood to Prescott made Intel lose ground. They cannot afford the same thing (Pentium M loses out in alot of content creation benchmarks), especially when Athlon64 is such a powerful competitor.
  • red and black - Saturday, March 26, 2005 - link

    How about a free-unix A64 vs P4 benchmark, with the A64 running 64 bit mode? What I and many other geeks are interested in is: "How quickly does (for example) mozilla compile? What if I'm building on a RAM file system?" I've heard that the A64 chips in 64 bit mode do well on this workload, but I haven't seen any real numbers.

    Or is this site only about Windows gamer setups?
  • ksherman - Saturday, March 26, 2005 - link

    I suggest a 939 NF3 and NF4 non-SLI motherboard round up... I dont care for SLI, so I really dont know what to get (out side of the DFI board)
  • Anand Lal Shimpi - Friday, March 25, 2005 - link

    The issue is that a Pentium M at 2.13GHz costs over $600 and is their flagship CPU in the line, thus you have to compare it to the highest end Pentium 4s and Athlon 64s.

    I also don't think the market wants a replay of the Pentium 4 introduction where performance dropped when moving to a new architecture. If Intel decided today to drop the Pentium 4 and move to the Dothan based Pentium M performance would drop in key areas like video encoding and anything relying on FP/SSE performance, that's frankly not acceptable. Yonah, I believe, will solve a lot of these issues.

  • Mark Little - Friday, March 25, 2005 - link

    #10, Let me put this to you very clearly.

    From Anand's own performance numbers in the ASUS adapter article, the 2.13 GHz PM beats a 3.2 GHz P4 Prescott in 2/3 of the benchmarks. I think you will agree that a 3.2 GHz Prescott beats all the Celeron processors which top out at 2.8 GHz. Anandtech, you and everyone else thinks that the Celeron is ready for the desktop because it is a desktop processor.

    Looking at the numbers, I think a PM at 2.13 GHz would beat every Celeron out there. It would probably beat a 2.4, 2.6 and 2.8 GHz P4 all running at 800 MHz FSB and having a higher clock speed. The damn thing beat a 3.73 GHz 1066 MHz FSB Extreme processor in the doom3 benchmark and came real close to beating a 3.6 GHz P4 in the ET benchmark. How is this not ready for the desktop?

    I know the PM is not meant for that high of clock speeds. I was talking hypothetically. The processor at its current speed is ready for the desktop now. It wins some benchmarks and it loses some. What processor doesn't? Do you guys think the G4 is not ready for the desktop? It clocks in at 1.67 GHz and is equivalent to a PIV running at the same speed. Apple seems to think it is ready for the desktop since it but it in a desktop computer.

    Hell, ASUS, Aopen and DFI all think the PM is ready for the desktop because they are making desktop motherboards and adapters. People don't want a 3.8 GHz furnace anymore. The desktop market has changed. It's just waiting for the rest of you.
  • smokey5 - Friday, March 25, 2005 - link

    If you do an Everquest 2 review I'm sure that'd help alot of people out. I don't intend to play but I know someone who does who wants my help on designing a rig. As far as I've heard Everquest 2 is CPU-heavy - surprisingly so comaparative to GPU. Also finding the peak amount of RAM needed would be nice. Hardrive I guess the fastest loading hardrive would be best eg Hitachi drives unless its changed recently?
  • quiksel - Friday, March 25, 2005 - link

    I can see that drivers would play a huge part in gaming performance on the mac... has anyone ever put the blame on the developer that picks up the game license to release it for the mac platform, though (Macsoft, Aspyre, and others)?

    I'm no programmer (thank god), but as such, I'm not real privy to the in's and out's of the development process, but from the perspective that picking up someone else's work and changing it to work for a different platfrom can't be all that easy, I wonder if these developers/recoders are slightly to blame as well....

    Any thoughts on this?
    ~quiksel
  • Heron - Friday, March 25, 2005 - link

    Kudos, anand, for putting a good article there.

    #7, Pentium M is designed with a clockspeed wall in mind. Keeping in mind that Pentium M's pipeline is not able to scale it to the 3.8GHz we are seeing on Pentium 4 today, I don't get your reasoning.

    Besides, Pentium M being not ready for the desktop is simple, as the performance figures clearly show that the Pentium 4 still has the upper hand in alot of sections. Besides, AMD is in the equation.
  • ViRGE - Thursday, March 24, 2005 - link

    Frankly, after thinking through the issue, I'm going to have to with #3's analysis of the situation, and place the blame largely on the drivers. If you take a look at other situations where the same video hardware is used for gaming on "non-standard" configurations such as using Quadros instead of GeForces(same basic hardware, the difference is in the drivers) or using Linux as the OS, we see the same kinds of performance hits as the Mac is seeing. I highly suspect that the drivers for these setups are not tweaked for performance like the Windows GeForce drivers are(in part because of all the performance/quality trade offs for GeForces for gaming), and that's why we're seeing such poor numbers. At the same time, I can't say I'm surprised since doing these optimizations is likely a difficult process that would require some work to port to the Mac, and that there are professional customers that would be displeased if their systems started implementing the performance/quality trade offs when they don't want it(there's no Quadro cards for the Mac, after all).

    I'd like to say I expect gaming on the Mac to get better, but I'm afraid it's become a self-fulfilling prophecy; the Mac wasn't "the" gaming platform, so now no one is bothering with gaming on the platform.
  • jonodsparks - Thursday, March 24, 2005 - link

    #5 - Tiger is still in beta, although nearing release. I would assuming that since gaming in not the primary market for OS X, it will be low on a long list of bugs to fix
  • Mark Little - Thursday, March 24, 2005 - link

    Anand,

    I must disagree with your statement

    "It doesn't look like there's too much potential for the Pentium M (Dothan) on the desktop."

    Don't you mean that a 1GHz+ Pentium IV is required to beat a Pentium M. If the Pentium M could run at 3.8 GHz today, wouldn't it beat any x86 processor out there. The Pentium M is at the same (a little lower) clock speeds as the Athlon 64. It beats a Pentium IV 3.2 GHz in 22 out of 33 benchmarks in your ASUS adapter article.

    Didn't your website accuse Intel of releasing faster processors too fast and that is why they are now stuck at 3.8 GHz? The Pentium M beat all the Pentium IVs in the gaming section. It also won some of the office benches. Don't you think that a processor that can do office apps and games is exactly what a desktop processor should do for the every day user.

    What average user is decoding and encoding music and video? I just don't understand your comments about Pentium M not being ready for the desktop because it can't beat a 3.8 GHz processor. That's a difference of almost 2 GHz yet it still performs well and I think it won 2 or 3 benchmarks out of all the CPUs. I just don't understand.
  • Traire - Thursday, March 24, 2005 - link

    Can't wait for the Theater 550 review. I have had my finger hovering over the order button on the PowerColor model for a few days now, but will try and hold out for your article.
  • Eug - Thursday, March 24, 2005 - link

    #3 - Don't get your hopes up for Tiger's gaming speed, cuz others have said that Tiger isn't noticeably faster for Doom 3.

    But one can always hope... We will see in a few weeks/months.
  • smitty3268 - Thursday, March 24, 2005 - link

    #1 - Sorry, I was thinking about another article. I don't remember where it was, but it had an interview with one of the Doom 3 developers.
  • smitty3268 - Thursday, March 24, 2005 - link

    #1 - If you read that article, the architectural differences account for about a 1% slowdown. OSX and the drivers were by far the biggest cause. Interestingly, the next version of OSX supposedly has much better gaming performance. I'll wait to see the final product before drawing any conclusions, though.
  • blckgrffn - Thursday, March 24, 2005 - link

    LOL!

    If only I could just build a new machine every time that I wanted to play a game!

    The dothan article is awesome, it should finally put to rest the debate about what is the most powerful desktop porcessor.

    Enjoy your new box and Splinter Cell!

    Nat
  • Ecgtheow - Thursday, March 24, 2005 - link

    On the Doom 3 slowness: There's a very good thread on the Ars Technica forums (http://episteme.arstechnica.com/eve/ubb.x/a/tpc/f/... that deals with this issue. There are in fact some architecture differences between the PPC and x86 that are at fault (float to int conversions, etc.).

Log in

Don't have an account? Sign up now