That is why I think Apple/Motorola should look to what AMD are up to with the Hammer/Opteron.
It will be this chip in the long run that will give them the bigger headache (with performance issues). Intel are already far too far ahead in raw speed, but people do understand the other factors involved in making a faster, more efficient computer. Apple just has to address those (and I won't re-iterate what has been discussed ad infinitum in other threads about what they might be).
<strong>No current G4s support the 128-bit MPX variant. They don't have the pins for it. This isn't a better solution than DDR because it is a lot more expensive to run all those extra traces across the motherboard and to build all the extra pins into the chipset and the processor.</strong><hr></blockquote>
You're right that no current G4s support it, but the existence of a 128-bit version suggests that they could hack something up that reads twice per cycle from those 64 pins and demuxes it to 128 bits internally. Other than increased latency, I don't see why that wouldn't work.
That's exactly it. Increased latency is what would kill you not to mention all the cycles wasted stitching up 2 64bit words to a 128bit word.
it would make more sense to make a 32 bit MPX bus running at 500MHz or some such. But the northbridge would have to be mighty close to the processor...
The so-called "MHz myth" is either over-hyped or under-hyped (by a given individual). MHz does matter but it does not have a linear relationship to performance, even on the same line of processors!
Its like the RPM of a car's engine: if you take the same engine and run it at a higher RPM you will get more horsepower, but once you are passed the "peak power" you don't get a linear increase -- you start getting diminishing returns. A larger or smaller engine, with a different number of cylinders, and different design will produce very different results even though it is spinning at a different RPM -- the exact RPM doesn't matter because of the gearing.
Nobody quotes RPM when talking about engine power, so why the hell does everyone quote MHz when talking about computing power?! The reason is simple: nobody has come up with a good equivalent, for computing, to "horsepower" (or even "torque"). There are tons of benchmarks, even quasi official ones but they all are too focused on particular parts of the system. Its like the difference between 0-60 times, quarter mile times, and passing on the highway -- they're all slightly related, but different cars are optimized for different conditions, and different people care more or less about different tasks.
That aside, however, in today's market Apple is behind in clock rate, bus bandwidth, and their processor is less superscalar than the competition. They will soon have less capability (i.e. no 64-bit). The only thing keeping them remotely competitive is a better architecture. If they fix any two of those things they'll be competitive again (for a while, this is an on going race). If they fix three they can be on top of the heap in terms of performance (for a while). If they fix all four sufficiently they could leap out ahead (for a while). They been there before several times over the last 20 or so years, so its not beyond them.
But Apple will get hit upside the head with another marketing ploy in a number of months. We will soon have the raw speed advantages of the P4, coupled to the 64-bitness of the AMD offerings.
I hope that that SJ has a better riposte for this upcoming scenario than the now-stale Mhz myth.
Maybe it sounds stupid but that's the first time that Apple is serious about corporate & business (xserve), but this is 'like the name tell' a server ! So they must have a serious solution for the other side of the chain.
"But Apple will get hit upside the head with another marketing ploy in a number of months. We will soon have the raw speed advantages of the P4, coupled to the 64-bitness of the AMD offerings.
I hope that that SJ has a better riposte for this upcoming scenario than the now-stale Mhz myth."
I'll say.
If SJ starts talking about the mhz myth after delivering a 1.2 gig sdr...
...bring on the tomatoes...(while singing: 'La, lah...la...lah...la...lah...)
Comments
It will be this chip in the long run that will give them the bigger headache (with performance issues). Intel are already far too far ahead in raw speed, but people do understand the other factors involved in making a faster, more efficient computer. Apple just has to address those (and I won't re-iterate what has been discussed ad infinitum in other threads about what they might be).
<strong>No current G4s support the 128-bit MPX variant. They don't have the pins for it. This isn't a better solution than DDR because it is a lot more expensive to run all those extra traces across the motherboard and to build all the extra pins into the chipset and the processor.</strong><hr></blockquote>
You're right that no current G4s support it, but the existence of a 128-bit version suggests that they could hack something up that reads twice per cycle from those 64 pins and demuxes it to 128 bits internally. Other than increased latency, I don't see why that wouldn't work.
it would make more sense to make a 32 bit MPX bus running at 500MHz or some such. But the northbridge would have to be mighty close to the processor...
Its like the RPM of a car's engine: if you take the same engine and run it at a higher RPM you will get more horsepower, but once you are passed the "peak power" you don't get a linear increase -- you start getting diminishing returns. A larger or smaller engine, with a different number of cylinders, and different design will produce very different results even though it is spinning at a different RPM -- the exact RPM doesn't matter because of the gearing.
Nobody quotes RPM when talking about engine power, so why the hell does everyone quote MHz when talking about computing power?! The reason is simple: nobody has come up with a good equivalent, for computing, to "horsepower" (or even "torque"). There are tons of benchmarks, even quasi official ones but they all are too focused on particular parts of the system. Its like the difference between 0-60 times, quarter mile times, and passing on the highway -- they're all slightly related, but different cars are optimized for different conditions, and different people care more or less about different tasks.
That aside, however, in today's market Apple is behind in clock rate, bus bandwidth, and their processor is less superscalar than the competition. They will soon have less capability (i.e. no 64-bit). The only thing keeping them remotely competitive is a better architecture. If they fix any two of those things they'll be competitive again (for a while, this is an on going race). If they fix three they can be on top of the heap in terms of performance (for a while). If they fix all four sufficiently they could leap out ahead (for a while). They been there before several times over the last 20 or so years, so its not beyond them.
[ 05-31-2002: Message edited by: Programmer ]</p>
But Apple will get hit upside the head with another marketing ploy in a number of months. We will soon have the raw speed advantages of the P4, coupled to the 64-bitness of the AMD offerings.
I hope that that SJ has a better riposte for this upcoming scenario than the now-stale Mhz myth.
i'm counting on them for the next 6 months.
I hope that that SJ has a better riposte for this upcoming scenario than the now-stale Mhz myth."
I'll say.
If SJ starts talking about the mhz myth after delivering a 1.2 gig sdr...
...bring on the tomatoes...(while singing: 'La, lah...la...lah...la...lah...)
Lemon Bon Bon