When will Apple catch up?

evoevo
Posted:
in Future Apple Hardware edited January 2014
It seems Apple is always way behind PC's when it comes to motherboard technologies. Better DDR Ram has been a PC standard for how long now? It takes Apple forever to speed up their frontside bus, and whenever they catch up, they are quickly left behind again on the PC side.



Apple was late in the game with AGP, but now they're at 4x. They were also late with CD-RW, but finally caught up.



I don't know a whole lot about the nitty gritties of the computer, but when my PC using friends dog me for Apple selling such outdated PowerMacs with such high prices, I really can't argue back



When will Apple just catch up and keep up? Am I just being stupid? Apple's PowerMac upgrade cycle seems so slow, it gets frustrating after a while.

Comments

  • Reply 1 of 19
    amorphamorph Posts: 7,112member
    When you actually have to engineer your products, it takes longer to adopt new tech.



    Apple's hardware, in actual use, does just fine.



    There are some things, like the G4's lack of support for a double-pumped frontside bus, that are genuine concerns; but then, that's what the Apollo's monster cache (and sweet cache architecture) address: The more code you can stick into the cache, and the more efficiently you can access and manage it, the less relevant the front side bus becomes for most tasks.



    If they come at you with what their machines can have, come back with what your machine can do, or just nod and smile. There are shortcomings, and things it can't do as well, of course - but every platform has shortcomings.
  • Reply 2 of 19
    If your friends were pissed because you enjoy breathing would you stop?



    Apple is behind on some things and others they are ahead. However yes Apple computers are expensive, but have you ever look at the quailty and care that goes into design of the machine I don't just mean the case ethier. I work with 1000's of PC's in my job and I'm always impressed by the quailty of the machines. Look at the cap's on there boards they have a plastic supports. Also not many companys use star washers under their screws. The Card Slots have have a radius cut out of them to make it easy to remove a AGP or PCI slot with a screw driver.



    The IBM's and Dell's I work with don't have this kind of make up or quailty. The IBM server are better than desktop but over all are no made with the same care.





    Where as with many things in life there's always a good reason to bitch. Try to look at the things that you like instead. The next time they dog you, think of the quailty of the machine you have, and forget about the FPS happy crack heads.



    If the machine works for you however slow it is compared to a PC why worrie about them. If it right for you use it, same goes for a slow PC if it's good enough for you that's all that matters.
  • Reply 3 of 19
    [quote] I don't know a whole lot about the nitty gritties of the computer, but when my PC using friends dog me for Apple selling such outdated PowerMacs with such high prices, I really can't argue back

    <hr></blockquote>



    All that really matters is if your computer is fast enough for what you use it for. I've heard the same BS from PC users, many of them only use their computers for web, email, word processing, and yet they get excited about RAMBUS and 2 GHZ Pentiums. I have no idea why, since a cheap celeron is fine for what they do with their computers.



    For serious gaming, a PC is a must, but a good gaming machine is not necessarily good at everything else. Even the most bad-ass gaming rig gets stomped in Photoshop by a good Powermac.



    So if these PC wankers start giving you crap about your Mac, tell them that it's fast for what YOU use it for. And it's also fine if the Mac is slower at some things, too. Tell them specs are less important to you than overall user experience, that you think of a computer as the sum of its parts, rather than a table of numbers.



    I hate to use a car analogy, but it's a good example: a Dodge Viper has a hell of a lot of horsepower, but for daily transportation in a temperate climate, a Viper is a total piece of crap, and in fact I would rather drive a Honda than a Viper. But if you race your car on weekends, a Viper is a good car (and the Viper is also a good car to use for getting laid, but I've never heard of anyone getting laid with a really fast computer).



    If they keep doggin' ya, just bitch-slap 'em and tell them that your life doesn't revolve around you computer. That should shut them up for awhile.
  • Reply 4 of 19
    outsideroutsider Posts: 6,008member
    At first I was pissed that the new Mac's did not use DDR SDRAM. I came to believe that Apple was behind the curve and then I decided to do some real world research. I found out that DDR SDRAM only offers a 5-10% increase in benchmark performance and not very noticable real world performance. In fact, <a href="http://www.anandtech.com/chipsets/showdoc.html?i=1386&p=1"; target="_blank">this article</a> shows that the Athlon benefits mostly, not from it's use of DDR memory but from it's DDR bus! Even when coupled with SDR memory. The P4 on the otherhand benefits more (and was designed this way) from a fast streaming memory technology in contrast with the Athlon (and G4). So here's the point. Motorola would do well to keep their SDR (lower latency than DDR) system bus, but, in it's next incarnation of the G4, bump it to 200MHz. That would give it a REAL 1.6GBps bandwidth, not effective bandwidth as you get with DDR busses. Even though a 266MHz DDR bus, theoretically gives 2GBps, it suffers from slightly higher latency issues and a more complicated motherboard set up. Then It would be Apple's duty to update the memory controller on the PowerMac to take advantage of the higher bandwidth. The MPX bus might have a little live left in it after all. And on the plus side also, we would get processor increments in 100MHz chunks versus 66MHz chunks.
  • Reply 5 of 19
    moogsmoogs Posts: 4,296member
    Thing is, if we *don't* get support for either PC1600 or PC2100 DDR modules, what's left? PC133 is as high as SDRAM modules get, based on the product line from Crucial and Kingston. Is there such a thing as PC166 or PC200 SDRAM DIMMS? Further, is there a *point* to boosting the MPX bus to 200 MHz while at the same time *not* supporting DDR? Based on the type of RAM modules available out there, I'd say no.





    An interesting link regarding market demand for DDR broken down by machine type:

    <a href="http://www.kingston.com/roadmap/default.asp"; target="_blank">http://www.kingston.com/roadmap/default.asp</a>;





    Another good link, so we don't get confused by all the RAM naming conventions:

    <a href="http://www.kingston.com/memory/ddr/jedec.asp"; target="_blank">http://www.kingston.com/memory/ddr/jedec.asp</a>;



    [ 01-30-2002: Message edited by: Moogs ? ]</p>
  • Reply 6 of 19
    outsideroutsider Posts: 6,008member
    Moogs, you can have an SDR bus and still connect it to DDR RAM. The P3 does this (on motherboards that support DDR for the P3. But if the SDR bus is not fast enough, it'll do you a fat lot of good. The memory controller arbitrates the interaction between CPU and memory. It takes a DDR 133MHz bus from RAM and converts it to a 200MHz SDR bus to the CPU. This is no different from when a 400MHz (100MHz quad pumped) interacts with PC2100 (133MHz double pumped). It's asyncronous but it works well enough.
  • Reply 7 of 19
    [quote]Originally posted by Outsider:

    <strong>I found out that DDR SDRAM only offers a 5-10% increase in benchmark performance and not very noticable real world performance.

    </strong><hr></blockquote>



    This is what I heard too.

    Still, I guess OS X is currently as memory-bound as it gets for an operating system, so the advantage of a faster memory bus might be a little bigger here. Of course, it will hardly ever give as much of an improvement as some might hope (i.e. close to 100%).





    [quote]<strong>

    Even though a 266MHz DDR bus, theoretically gives 2GBps, it suffers from slightly higher latency issues and a more complicated motherboard set up.

    </strong><hr></blockquote>



    Actually, as far as I understand it, the design is *less* complicated with DDR (as opposed to SDR at twice the frequency) - that's the whole point of using DDR in spite of the associated higher latencies.



    Bye,

    RazzFazz
  • Reply 8 of 19
    [quote]Originally posted by Moogs ?:

    <strong>Thing is, if we *don't* get support for either PC1600 or PC2100 DDR modules, what's left?

    </strong><hr></blockquote>



    A possible non-DDR-memory option would be bank interleaving, i.e. having a 200MHz SDR 64bit bus between processor and northbridge, and a 100MHz SDR 128bit bus between northbridge and memory. 128 bits of width would be achieved by using pairs of identical PC100 DIMMs.



    The nForce does basically the same with DDR RAM to lessen the impact of the integrated graphics core on memory bandwidth available to the processor.



    Bye,

    RazzFazz
  • Reply 9 of 19
    I think a lot of the PC guys just like to brag about specs when they can't actually compare performance. One example is the ATA bus. I think Apple is using ATA66 while the current maximum is much faster. However, I read somewhere that the faster ATA interfaces yield little because the disk drives themselves are too slow.
  • Reply 10 of 19
    outsideroutsider Posts: 6,008member
    If Rambas didn't have such a bleak future, I would even consider them as a possible alternative to DDR in the next PowerMacs.
  • Reply 11 of 19
    amorphamorph Posts: 7,112member
    How much DDR improves real performance is essentially a quality of implementation issue. It's also a type-of-implementation issue. SDR sends memory 64 bits at a time. DDR gets its speed boost by speculatively fetching the next 64 bit word. If that's not what the CPU actually needed, then that step is wasted, so in the worst case (reading lots of small bits of data out of sequence) DDR defaults to SDR performance levels. However, the worst case doesn't happen very often.



    Furthermore, there is one part of the G4 that always, invariably wants the next 64 bits: AltiVec. So DDR gets best-case performance feeding the hungry vector unit.
  • Reply 12 of 19
    The primary problem right now is that the MPX bus on the G4 can't get data into the processor at greater than about 740 MB/sec. This is pretty good for a 64-bit 133MHz bus, but its only a 64-bit 133MHz bus. For memory intensive media applications which stream through large amounts of data, this just isn't good enough. And the working sets of these kind of applications are generally considerably larger than the 2 megabytes in the L3 cache.



    AltiVec is capable of performing really fast, but much of the time it is hamstrung by how fast the developer can get data to and from it. Those 15 GigaFlops do you a fat lot of good unless you're using an awful lot of them on each piece of data.
  • Reply 13 of 19
    [quote]Originally posted by Amorph:

    <strong>DDR gets its speed boost by speculatively fetching the next 64 bit word.</strong><hr></blockquote>



    No, DDR gets its speed boost by delivering twice as much data in the same number of clocks. Processors (or memory controllers) get their speed by reading the next burst before it is actually requested by the code. The G4 can actually be told what to read ahead, and it will do so at the maximum memory bandwidth.
  • Reply 14 of 19
    matsumatsu Posts: 6,558member
    Inasmush as Powermacs seem outdated, I don't think that they are as far behind as the commonly noted MB/proc specs seem to indicate.



    At the lab here at school the superdrive 733's and DP533's seem to crunch through most people's work just fine. OSX isn't as responsive as even some of the older win98 machines. But OSX looks *a lot* nicer, and the sluggishness may be all GUI related. Once programs are launched they run very nicely. I bet that WinXP would slow down our older PC hardware considerably too, so call it even. Both OS's are intended for recent/not yet released hardware, at least to get the snappiness we've become accustomed to with OS9x and Win9x/NT.



    FP performance is probably better on the PC side. Apple will need to improve that. Altivec will help. From speaking with some of the comp-sci geeks (i don't know squat) it seems to me that Altivec is very flexible and powerful when apps are coded to properly take advantage of it. Look at final cut pro. The effects generated in software in REAL TIME require a lot of muscle. A PB 667 manages to pull off some mean feats in this respect. Now take your average 2Ghz + pentium4. It also does pretty good, but it ain't 3 times faster, or even 2 times faster. In fact it's faster in some things and slower in others. It only gets a lot faster when you add in some expensive hardware encoding/decoding functions.



    Altivec is in fact quite powerful, but the devs need time. When altivec was released it was a slightly different concept in coding. Devs had to deal with that, and with an upcoming change to OSX. No one was going to invest a lot of time in Altivec aware OS9 apps. Like quark and photoshop devs, most probably didn't put the maximum effort into classic development after a certain point. They're learning a lot of stuff: OSX and altivec, and even DP awareness in combination with these technologies. For the last 2 years they could bank on a huge installed base of G3 based PRO machines. If you figure that a lot of users won't benefit from your work with these technologies, you probably just don't bother bringing them to market until you can get all those new principles reasonably integrated onto a new app for a new system that won't bog down under it.



    iMacs point to the future here. IBM elected to wait at least another 6-12 before the introduction of their own SIMD part. Apple didn't want to wait anymore (despite the efficiency/heat benefits. Consumer altivec is here (probably for iBooks too within a year) and the installed base is gonna be a whole lot bigger. That's significant; the software will reflect it.



    Apple might fudge it's benchmarking results a touch, as I'm sure everyone picks the most favorable result. But most side by side comparos put the 867 mac damned close. And, when it gets properly coded software (like final cut pro, or iDVD, etc) look out! Cost is the only issue. They could stand to offer better prices on their machines, but most of the pros you talk to (and those who use multiple platforms) know that the mac is at least competitive in most areas.
  • Reply 15 of 19
    outsideroutsider Posts: 6,008member
    So an SDR 200MHz MPX bus will be on par with a 133MHz x2 DDR bus.
  • Reply 16 of 19
    Enough of this s*** about "for serious gaming, a PC is a must".



    Actually, that's a flat out lie. For serious gaming, buy a console. I'm serious - the GameCube, even the XBox kick ass over the best PC hardware because they are customized for gaming. Why the hell should you shell out $$$ for a serious PC system when you can buy a console for the cost of a graphics card on one of the high end boxes.



    The only reason you wouldn't buy a console is because you're a pirating thief, or in my case, because you're not into the hardcore 3D games and prefer the strategy (CivIII addict, I admit) and basically anything you buy today will do.



    So all that leaves for serious computing requirements on a Mac are Photoshop (optimized for Altivec), DVD burning (optimized for Altivec), scientific applications (see clustering thread), Maya (optimized for Altivec).



    Anyone seeing a pattern here? Personally, I can only type around 70 words per minute, so Word, Excel, etc., do not need more that about 500 Mhz (even with OS X slowness, which is going to improve).



    'Nuff said. Stop telling them to buy a PC. Buy a console.



  • Reply 17 of 19
    serranoserrano Posts: 1,806member
    moto says we'll catch up in 2004 when intel trys to put out a 64bit chip and finds it has difficulty ramping it up to a ghz (itanium...)



    but then again, moto says a lot of things...



    <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />
  • Reply 18 of 19
    marcukmarcuk Posts: 4,442member
    so Is MOTO saying the only way they'll catch up is IF Intel drop a bollock?



    What about AMD?
  • Reply 19 of 19
    programmerprogrammer Posts: 3,457member
    [quote]Originally posted by Pendrake:

    <strong>So all that leaves for serious computing requirements on a Mac are Photoshop (optimized for Altivec), DVD burning (optimized for Altivec), scientific applications (see clustering thread), Maya (optimized for Altivec).

    </strong><hr></blockquote>



    Well that's a gross over-simplification if there ever was one. But it doesn't change my point above -- if you are limited by memory, and not by the processor, then it doesn't matter how good AltiVec is. Say AltiVec can do twice as much work in the same number of cycles as the SSE2 (used to be true, but may not hold on the latest P4)... now consider you are doing a relatively expensive calculation that takes 50 operations per vector. AltiVec is running at half its theoretical speed because of the memory constraint. The SSE2 system is running just as fast if it has the same memory subsystem! But it doesn't... instead its memory is twice as fast and its clock rate is double, so it is actually twice as fast as the AltiVec machine because of the memory bottleneck. Now consider a mythical G5 at a 50% higher clock than the G4 with a much faster (3x) memory bus and memory... it would run three times faster than the G4, and 50% faster than the (higher clocked) SSE2 machine. Yes, AltiVec is really great but it continues to be badly limited by the memory interface.
Sign In or Register to comment.