As has been stated a thousand times in numerous threads. Repeating it isn't going to change the issues directly related to each reason we are still in need of dual socket motherboards after the chips become dual core.
Yes, you seem to have mis-read my statement. I meant that going beyond 2 FSB ports is too expensive. Since Apple already has 2 ports, however, they will probably continue in that vein. Two chips with dual 980 SMT cores by the end of 2006... that's at least 8 hardware threads. Or more.
I'm in favor of Apple keeping two CPUs in their pro machines because if you're running any mission critical services and one CPU dies, the whole machine doesn't necessarily die. A little redundancy is good.
I'm in favor of Apple keeping two CPUs in their pro machines because if you're running any mission critical services and one CPU dies, the whole machine doesn't necessarily die. A little redundancy is good.
Or if one thread (which for some reason can't be quit) goes haywire and starts sucking up cycles, the computer still stays responsive. I've got first hand knowledge of this one \
I'm in favor of Apple keeping two CPUs in their pro machines because if you're running any mission critical services and one CPU dies, the whole machine doesn't necessarily die. A little redundancy is good.
I don't think Powermacs are built with that kind of redundancy in mind... I suspect that if you lost one CPU in a modern Powermac, the whole machine would become inoperable. I believe the sort of redundancy you're talking about is still reserved for big iron hardware; although clustering (in many respects) gets close to it.
...However, I do reserver the right to be wrong about this.
I don't think Powermacs are built with that kind of redundancy in mind... I suspect that if you lost one CPU in a modern Powermac, the whole machine would become inoperable. I believe the sort of redundancy you're talking about is still reserved for big iron hardware; although clustering (in many respects) gets close to it.
...However, I do reserver the right to be wrong about this.
Actually you may be wrong. CHUD tools allow you to "turn off" one of the CPU in a dual proc system. However that is using an apple tool. I wonder if one CPU fritzed out would the system know and deactivate the dead CPU socket.
Thats an assumption though because no one has used, or independently tested it in a comparison. vs. the 4 way AMD board.
It also isn't the same as what I was saying about quad processors, with quad cores. Even though the price looks high, If it were $1,500 for a quad socket quad core capable motherboard the price vs performance ratio wouldn't seem as astronomical comparatively.
Thats an assumption though because no one has used, or independently tested it in a comparison. vs. the 4 way AMD board.
It also isn't the same as what I was saying about quad processors, with quad cores. Even though the price looks high, If it were $1,500 for a quad socket quad core capable motherboard the price vs performance ratio wouldn't seem as astronomical comparatively.
Yeah I figured that as well but there's not much you can do to keep costs in line when you're starting out at $1500 for the Mobo. Let's just toss some stuff together and see what we have.
$1500- Mobo
$2400- 4 DC Opteron chips
$2000- Top of the line Quadro
$300- HD
$700- RAM
$100- DVD-R
= $7000 Homebuilt for 8 physical CPU potential 8 logical.
I'd say some situations would benefit you while others wouldn't.
Yeah I figured that as well but there's not much you can do to keep costs in line when you're starting out at $1500 for the Mobo. Let's just toss some stuff together and see what we have.
$1500- Mobo
$2400- 4 DC Opteron chips
$2000- Top of the line Quadro
$300- HD
$700- RAM
$100- DVD-R
= $7000 Homebuilt for 8 physical CPU potential 8 logical.
I'd say some situations would benefit you while others wouldn't.
That is truly awesome when you think about it - I spent over $7000 just to get my IIci back in 1990 (One processor + FPU at 25MHz on a 25MHz bus). Everyone has to admit the prices have come down a bit over the years.
I would not mind seeing an Apple offering with 4 dual core CPUs, but I think that the 2 dual core CPUs are the most we can hope for in the near future.
Tiger wants its MP cores! While the 970MP seems to be designed specifically for IBM's blade servers, there is little doubt that MP cores are in Apple Computer's future. Putting MP cores in Powerbooks, iMacs & eMacs would be a great way to make everyone forget that nine something. It would probably also save Apple some money with dual (MPs) only needed in the best tower configuration and Xserve.8)
MOSR posts some details on the 970MP and the problems of IBM. Can anyone in the know tell us if all this makes sense or if it is the traditional MOSR fantasy? Here is the second part of the article.
Well I'm not too sure that all MacOSRumors prints is fantasy, but everything in that article appears to be in the realm of possibility.
As to indications that IBM has had multiple problems at 90nm that would seem to be likely as the problem is taking forever to resolve. Some of the information is consistant with other rumors that have floated around with respect to IBM so there may be trueth here.
More interesting here is that IBM's road map appears to be as stable as a dam about to over fill after a hurricane. One has to wonder just why the 970 MP has moved to the forefront apparently. If IBM is having issues moving the Power 5 to 90nm then that should be of concern to all.
Dave
Quote:
Originally posted by PB
MOSR posts some details on the 970MP and the problems of IBM. Can anyone in the know tell us if all this makes sense or if it is the traditional MOSR fantasy? Here is the second part of the article.
The MOSR report sounds bunk. I really doubt IBM would have called any of their PPC processors the "9700" that just defies current naming conventions.
The rest is nothing that we haven't already discussed already. I doubt it's the 90nm transition that slowed down the POWER5 derivative but rather strategy. The "980" will likely be 65nm and that process shrink won't be ready to go until sometime in 2006.
Comments
Originally posted by onlooker
As has been stated a thousand times in numerous threads. Repeating it isn't going to change the issues directly related to each reason we are still in need of dual socket motherboards after the chips become dual core.
Yes, you seem to have mis-read my statement. I meant that going beyond 2 FSB ports is too expensive. Since Apple already has 2 ports, however, they will probably continue in that vein. Two chips with dual 980 SMT cores by the end of 2006... that's at least 8 hardware threads. Or more.
Originally posted by bunge
I'm in favor of Apple keeping two CPUs in their pro machines because if you're running any mission critical services and one CPU dies, the whole machine doesn't necessarily die. A little redundancy is good.
Or if one thread (which for some reason can't be quit) goes haywire and starts sucking up cycles, the computer still stays responsive. I've got first hand knowledge of this one \
I'm in favor of Apple keeping two CPUs in their pro machines because if you're running any mission critical services and one CPU dies, the whole machine doesn't necessarily die. A little redundancy is good.
I don't think Powermacs are built with that kind of redundancy in mind... I suspect that if you lost one CPU in a modern Powermac, the whole machine would become inoperable. I believe the sort of redundancy you're talking about is still reserved for big iron hardware; although clustering (in many respects) gets close to it.
...However, I do reserver the right to be wrong about this.
Originally posted by Gamblor
I don't think Powermacs are built with that kind of redundancy in mind... I suspect that if you lost one CPU in a modern Powermac, the whole machine would become inoperable. I believe the sort of redundancy you're talking about is still reserved for big iron hardware; although clustering (in many respects) gets close to it.
...However, I do reserver the right to be wrong about this.
Actually you may be wrong. CHUD tools allow you to "turn off" one of the CPU in a dual proc system. However that is using an apple tool. I wonder if one CPU fritzed out would the system know and deactivate the dead CPU socket.
Originally posted by Outsider
I know for a fact that if one CPU dies while the system is on, the whole system comes down.
How do you know this?
Originally posted by hmurchison
Good question. My guess would be
$1500 Quad Tyan Mobo
would be matched by 2 Macs running Xgrid.
Thats an assumption though because no one has used, or independently tested it in a comparison. vs. the 4 way AMD board.
It also isn't the same as what I was saying about quad processors, with quad cores. Even though the price looks high, If it were $1,500 for a quad socket quad core capable motherboard the price vs performance ratio wouldn't seem as astronomical comparatively.
Originally posted by onlooker
Thats an assumption though because no one has used, or independently tested it in a comparison. vs. the 4 way AMD board.
It also isn't the same as what I was saying about quad processors, with quad cores. Even though the price looks high, If it were $1,500 for a quad socket quad core capable motherboard the price vs performance ratio wouldn't seem as astronomical comparatively.
Yeah I figured that as well but there's not much you can do to keep costs in line when you're starting out at $1500 for the Mobo. Let's just toss some stuff together and see what we have.
$1500- Mobo
$2400- 4 DC Opteron chips
$2000- Top of the line Quadro
$300- HD
$700- RAM
$100- DVD-R
= $7000 Homebuilt for 8 physical CPU potential 8 logical.
I'd say some situations would benefit you while others wouldn't.
And even in a board designed to react this way, a seriously blown part still could take the machine down under the right circumstances.
Originally posted by xsmi
How do you know this?
It's happened to two of our DP G5 machines.
edit: the machines restarted after but there were alerts that one of the CPU's was disabled.
Originally posted by Outsider
It's happened to two of our DP G5 machines.
edit: the machines restarted after but there were alerts that one of the CPU's was disabled.
That is too cool. I know it cost lots of money but that is just wicked!
Originally posted by hmurchison
Yeah I figured that as well but there's not much you can do to keep costs in line when you're starting out at $1500 for the Mobo. Let's just toss some stuff together and see what we have.
$1500- Mobo
$2400- 4 DC Opteron chips
$2000- Top of the line Quadro
$300- HD
$700- RAM
$100- DVD-R
= $7000 Homebuilt for 8 physical CPU potential 8 logical.
I'd say some situations would benefit you while others wouldn't.
That is truly awesome when you think about it - I spent over $7000 just to get my IIci back in 1990 (One processor + FPU at 25MHz on a 25MHz bus). Everyone has to admit the prices have come down a bit over the years.
I would not mind seeing an Apple offering with 4 dual core CPUs, but I think that the 2 dual core CPUs are the most we can hope for in the near future.
Originally posted by hmurchison
$1500- Mobo
$2400- 4 DC Opteron chips
$2000- Top of the line Quadro
$300- HD
$700- RAM
$100- DVD-R
= $7000 Homebuilt for 8 physical CPU potential 8 logical.
If your using (you said 4 DC, but they are 4 QC) Actually that would be 16 physical/potential 16 logical processors.
That's what I meant about Price vs Performance. It's actually a pretty good deal for $7,000.00
Originally posted by dfryer
Is someone saying you can get a quad-core opteron for $600? I have no idea if this is accurate, but it seems a little cheap..
nope, no one said that.
As to indications that IBM has had multiple problems at 90nm that would seem to be likely as the problem is taking forever to resolve. Some of the information is consistant with other rumors that have floated around with respect to IBM so there may be trueth here.
More interesting here is that IBM's road map appears to be as stable as a dam about to over fill after a hurricane. One has to wonder just why the 970 MP has moved to the forefront apparently. If IBM is having issues moving the Power 5 to 90nm then that should be of concern to all.
Dave
Originally posted by PB
MOSR posts some details on the 970MP and the problems of IBM. Can anyone in the know tell us if all this makes sense or if it is the traditional MOSR fantasy? Here is the second part of the article.
The rest is nothing that we haven't already discussed already. I doubt it's the 90nm transition that slowed down the POWER5 derivative but rather strategy. The "980" will likely be 65nm and that process shrink won't be ready to go until sometime in 2006.