<strong>The real problem is that by the time this thing ships, every single one of Apple's G4 (and G3 if the iBook has G3s) computers will have processors that are faster than this chip. Apple has been shipping a processor this fast in its Power Macs for 4 months.
</strong><hr></blockquote>
I guess I see that RapidIO on a 7500 would provide more than enough bandwidth. Now a PPC with 2 FPUs and RapidIO would put Apple in the hunt. If there is an Altivec 2 than the PPC is really in the hunt. I could see two PPC on a daughter card and RapidIO interconnect to a HyperTransport controller for the main board functions. This is tech. that is at hand or close to hand. And this seems to be doable in MWSF time frame, or close.
As far as RapidIO goes doesn't IBM have a different tech.?? Similar to rapidIO.
Spoke with some one I know from Apple today. They let slip that Steve handpicked two companies to get the servers first. These companies have had the servers for what sounds like the past few months. From what I gather this was not so much a standard beta situation as Apple rolling out the product to two key customers first. Without being specific and without being too coy, I can let you know that one of the companies is one of the country's biggest publishing companies and is in Kansas.
Also, without starting a new thread (Amorph, move it if you want to), Apple-branded iPod software is coming. I wasn't told if this was iTunes or not, just that it was iPod software for Windows.
Please forgive my ignorance, but just what makes the MPC8540 "useless" to Apple (other than not shipping!)?
The fact sheet says it is fully Book e compliant, it has dual gigabit ethernet, rapidio, L2 cache, 333mhz ddr memory controller, PCI-X, DMA....
Now, we know Mhz do not matter - A 500 Mhz G4 with ddr, rapidIO, etc could toast a dual gig, probably. What if it's an 8540 at 600Mhz or 1 ghz in the new server? or two or four or eight of them?
Burnt Pentiums allover the place, and the Mac rack wouldn't even be warm! (6.5 watts power usage v. up to 10 times that for others)
Just my wishful thinking?
J
OK - No altivec -- But a server would not need Altivec necessarily, would it?
[QB]I really think Apple is going to include somekind of Mac-only technology that will get the geeks drooling..../QB]<hr></blockquote>
How about: food scented inserts for the power supply fan? You could have a variety of scents that you could put in. Instead of the Server clost smelling like hot plastic it could smell like...a world class meal.
[quote] Spoke with some one I know from Apple today. They let slip that Steve handpicked two companies to get the servers first. These companies have had the servers for what sounds like the past few months. From what I gather this was not so much a standard beta situation as Apple rolling out the product to two key customers first. Without being specific and without being too coy, I can let you know that one of the companies is one of the country's biggest publishing companies and is in Kansas. <hr></blockquote>
Now I usually don't put much stock into the "my friend works at apple" posts, but this actually makes sense. If they don't have a pre-existing install base with Enterprise Level customers, they'll have to work too hard to make the sale there - this is essentially opening-day credibility.
Wouldn't have struck me much unless one of our clients hadn't done exactly this with a piece of enterprise-grade backend soft. to jumpstart sales on industry release.
If Apple's doing this, it would make sense that they will quickly (or have already) worked this into a schoolsystem (a lot of specualtion here - the servers may not even be edu. oriented) for the same credibility in that market.
Please forgive my ignorance, but just what makes the MPC8540 "useless" to Apple (other than not shipping!)?</strong>
It's an entirely integer processor for embedded hardware like routers, big iron storage systems, control hardware, and various signal processing hardware. That's what it is designed to do.
The problem is that it has no floating point capabilities, is only 2 way superscalar, its SIMD unit is even more specialized than AltiVec, and if a 7450 based G4 is fabbed on the same process as the 8540, the 8540 will always be slower in integer. Since it doesn't have floating point capabilities, we won't mention that. What it may be faster at is a signal processing operation that its SIMD unit is designed to do.
<strong>OK - No altivec -- But a server would not need Altivec necessarily, would it?</strong>
It probably needs floating point. But why bother when Apple could just put a 1 GHz 7455 in it and have the same performance? And a 7455 is available today while a HiP 7 7460 or 7470 G4 could be available at the same time as the 8540 and will be faster as well.
How about: food scented inserts for the power supply fan? You could have a variety of scents that you could put in. Instead of the Server clost smelling like hot plastic it could smell like...a world class meal.
Slogans:
Mac=Mmmmmm
We now have the power, the sex, and ...dinner</strong><hr></blockquote>
Bob, we need to talk. Ever since an unfortunate incident involving my (then) 14-month son, a room-scenting oil device, and my AV receiver, the room is infused with the soothing scent of vanilla every time i watch a DVD. I felt at the time that scented inserts for computers could be a fantastic new market - let's face it, people bought Pet Rocks fer chrissakes. We just need some VC, and a taut name: iStink? Click'n'Waft? AroMac?
But the fact sheet and docs for the 8540 say it's booke compliant, has controllers for ethernet, including gigE, rapidIO, etc and some kind of SIMD unit (222 new instructions?)
I defer to your expertise on the FP and other stuff. Now I ask is it possible that the previous documentation was a front for the "real deal" to appear sometime between now and 2003... I doubt it, as that might tick some who want the described chip off.
<strong>has motorola ever announced a CPU before Apple introduced it?
heck they even took out the 1.1GHz G4 part they listed last month.
</strong><hr></blockquote>
I knew it! In another thread a while back I stated that the 1.1 part was already available on their site. I was sure I saw it. Then when I went back for linkage it wasn't listed anymore. Do you think it was just a glitch? At least I'm not the only one who saw it.
But the fact sheet and docs for the 8540 say it's booke compliant, has controllers for ethernet, including gigE, rapidIO, etc and some kind of SIMD unit (222 new instructions?)
I defer to your expertise on the FP and other stuff. Now I ask is it possible that the previous documentation was a front for the "real deal" to appear sometime between now and 2003... I doubt it, as that might tick some who want the described chip off.</strong>
I would agree. The 8540 provides no clues for what processor Apple will be using. Book E compliancy is about as important to Apple as an appendix is to the human body. It simply isn't relevant to Apple's current market.
All the other stuff would a be waste of die space for Apple, when what's really needed is bumping the next gen G4 to 4-way superscalar plus better branch prediction, another FPU unit, increasing the execution pipeline to 10 or 12, a state of the art cache design, and a higher performance processor bus. Plus other little things, but most everything can be grafted into the G4 and Apple should be fine.
I think they would probably be referred to as "nodes".
Anyway, how is a cluster node a server? How do you define "server"?</strong><hr></blockquote>
Razz:
I thought I would chime in on this...
It really depends on what type of cluster you are using. If you are using a failover/loadbalance cluster like many high traffic web sites use then the cluster will generally contain 2 master nodes (controllers if you will) and many nodes (the actual servers). Each of the nodes can each run as a stand alone server. In many cases the nodes don't share anything-no storage or processing power. The upside to this is true 100% uptime and the ability to run cheaper server hardware. (Instead of spending $30,000.00 on one server you can spend $30,000.00 on 5 servers and have a nice 3 node cluster with complete failover support.)
Other clusters can act as a single computer...the primary purpose of the nodes is sheer processing power (ie heavy math calculations).
Comments
<strong>The real problem is that by the time this thing ships, every single one of Apple's G4 (and G3 if the iBook has G3s) computers will have processors that are faster than this chip. Apple has been shipping a processor this fast in its Power Macs for 4 months.
</strong><hr></blockquote>
I guess I see that RapidIO on a 7500 would provide more than enough bandwidth. Now a PPC with 2 FPUs and RapidIO would put Apple in the hunt. If there is an Altivec 2 than the PPC is really in the hunt. I could see two PPC on a daughter card and RapidIO interconnect to a HyperTransport controller for the main board functions. This is tech. that is at hand or close to hand. And this seems to be doable in MWSF time frame, or close.
As far as RapidIO goes doesn't IBM have a different tech.?? Similar to rapidIO.
Spoke with some one I know from Apple today. They let slip that Steve handpicked two companies to get the servers first. These companies have had the servers for what sounds like the past few months. From what I gather this was not so much a standard beta situation as Apple rolling out the product to two key customers first. Without being specific and without being too coy, I can let you know that one of the companies is one of the country's biggest publishing companies and is in Kansas.
Also, without starting a new thread (Amorph, move it if you want to), Apple-branded iPod software is coming. I wasn't told if this was iTunes or not, just that it was iPod software for Windows.
j
The fact sheet says it is fully Book e compliant, it has dual gigabit ethernet, rapidio, L2 cache, 333mhz ddr memory controller, PCI-X, DMA....
Now, we know Mhz do not matter - A 500 Mhz G4 with ddr, rapidIO, etc could toast a dual gig, probably. What if it's an 8540 at 600Mhz or 1 ghz in the new server? or two or four or eight of them?
Burnt Pentiums allover the place, and the Mac rack wouldn't even be warm! (6.5 watts power usage v. up to 10 times that for others)
Just my wishful thinking?
J
OK - No altivec -- But a server would not need Altivec necessarily, would it?
[ 05-09-2002: Message edited by: jccbin ]
[ 05-09-2002: Message edited by: jccbin ]</p>
But the following is interesting, any expert's comments?
<a href="http://e-www.motorola.com/collateral/MPC8540E500RPT.pdf"; target="_blank">http://e-www.motorola.com/collateral/MPC8540E500RPT.pdf</a>;
page 5
" Signal Processing APU
. -Aimed at convergent Interger and DSP applications
. -SIMD unit with 222 new instructions"
I don't understand. No Altivec, but 222 new SIMD instructions? Clearly, in my ignorance I'm missing something.
[QB]I really think Apple is going to include somekind of Mac-only technology that will get the geeks drooling..../QB]<hr></blockquote>
How about: food scented inserts for the power supply fan? You could have a variety of scents that you could put in. Instead of the Server clost smelling like hot plastic it could smell like...a world class meal.
Slogans:
Mac=Mmmmmm
We now have the power, the sex, and ...dinner
Now I usually don't put much stock into the "my friend works at apple" posts, but this actually makes sense. If they don't have a pre-existing install base with Enterprise Level customers, they'll have to work too hard to make the sale there - this is essentially opening-day credibility.
Wouldn't have struck me much unless one of our clients hadn't done exactly this with a piece of enterprise-grade backend soft. to jumpstart sales on industry release.
If Apple's doing this, it would make sense that they will quickly (or have already) worked this into a schoolsystem (a lot of specualtion here - the servers may not even be edu. oriented) for the same credibility in that market.
Cool info if it's true.
Please forgive my ignorance, but just what makes the MPC8540 "useless" to Apple (other than not shipping!)?</strong>
It's an entirely integer processor for embedded hardware like routers, big iron storage systems, control hardware, and various signal processing hardware. That's what it is designed to do.
The problem is that it has no floating point capabilities, is only 2 way superscalar, its SIMD unit is even more specialized than AltiVec, and if a 7450 based G4 is fabbed on the same process as the 8540, the 8540 will always be slower in integer. Since it doesn't have floating point capabilities, we won't mention that. What it may be faster at is a signal processing operation that its SIMD unit is designed to do.
<strong>OK - No altivec -- But a server would not need Altivec necessarily, would it?</strong>
It probably needs floating point. But why bother when Apple could just put a 1 GHz 7455 in it and have the same performance? And a 7455 is available today while a HiP 7 7460 or 7470 G4 could be available at the same time as the 8540 and will be faster as well.
[ 05-12-2002: Message edited by: THT ]</p>
<strong>
How about: food scented inserts for the power supply fan? You could have a variety of scents that you could put in. Instead of the Server clost smelling like hot plastic it could smell like...a world class meal.
Slogans:
Mac=Mmmmmm
We now have the power, the sex, and ...dinner</strong><hr></blockquote>
Bob, we need to talk. Ever since an unfortunate incident involving my (then) 14-month son, a room-scenting oil device, and my AV receiver, the room is infused with the soothing scent of vanilla every time i watch a DVD. I felt at the time that scented inserts for computers could be a fantastic new market - let's face it, people bought Pet Rocks fer chrissakes. We just need some VC, and a taut name: iStink? Click'n'Waft? AroMac?
heck they even took out the 1.1GHz G4 part they listed last month.
What we know is that they're working on RapidI/O, DDR333 and PCI-X. The time frame we have is higly speculative.
How hard would it be to implement RapidI/O in the G4?
But the fact sheet and docs for the 8540 say it's booke compliant, has controllers for ethernet, including gigE, rapidIO, etc and some kind of SIMD unit (222 new instructions?)
I defer to your expertise on the FP and other stuff. Now I ask is it possible that the previous documentation was a front for the "real deal" to appear sometime between now and 2003... I doubt it, as that might tick some who want the described chip off.
<strong>has motorola ever announced a CPU before Apple introduced it?
heck they even took out the 1.1GHz G4 part they listed last month.
</strong><hr></blockquote>
I knew it! In another thread a while back I stated that the 1.1 part was already available on their site. I was sure I saw it. Then when I went back for linkage it wasn't listed anymore. Do you think it was just a glitch? At least I'm not the only one who saw it.
But the fact sheet and docs for the 8540 say it's booke compliant, has controllers for ethernet, including gigE, rapidIO, etc and some kind of SIMD unit (222 new instructions?)
I defer to your expertise on the FP and other stuff. Now I ask is it possible that the previous documentation was a front for the "real deal" to appear sometime between now and 2003... I doubt it, as that might tick some who want the described chip off.</strong>
I would agree. The 8540 provides no clues for what processor Apple will be using. Book E compliancy is about as important to Apple as an appendix is to the human body. It simply isn't relevant to Apple's current market.
All the other stuff would a be waste of die space for Apple, when what's really needed is bumping the next gen G4 to 4-way superscalar plus better branch prediction, another FPU unit, increasing the execution pipeline to 10 or 12, a state of the art cache design, and a higher performance processor bus. Plus other little things, but most everything can be grafted into the G4 and Apple should be fine.
<strong>
I think they would probably be referred to as "nodes".
Anyway, how is a cluster node a server? How do you define "server"?</strong><hr></blockquote>
Razz:
I thought I would chime in on this...
It really depends on what type of cluster you are using. If you are using a failover/loadbalance cluster like many high traffic web sites use then the cluster will generally contain 2 master nodes (controllers if you will) and many nodes (the actual servers). Each of the nodes can each run as a stand alone server. In many cases the nodes don't share anything-no storage or processing power. The upside to this is true 100% uptime and the ability to run cheaper server hardware. (Instead of spending $30,000.00 on one server you can spend $30,000.00 on 5 servers and have a nice 3 node cluster with complete failover support.)
Other clusters can act as a single computer...the primary purpose of the nodes is sheer processing power (ie heavy math calculations).