They may as well use Dual Channel DDR 400. It's the only memory that can hit 6.4Gbps throughput.
It should be available. However I wouldn't mind Apple shipping DDR 333 with full support for 400 as to not delay machines in any way due to supply issues.
<strong>maybe i just do not get it..but can somebody confirm ..the 970s architecture is 2times faster than a g4 and therefore almost 4 times faster than a pentium4?
i mean worst case...1.4ghz single (in imac too?!)
would be equal to a 2.8ghz g4 or a 5.6ghz pentium4?
and then..maybe a dual 1.8ghz equal to 14.4ghz pentium4?
which is about 4 times faster than the best pentium money can buy?
In that case, Apple is on a good road. 2.5 ghz? Those suckers aregone. But what I want to see is how the PPC 970 compares against the Xeon. Does anyone know of a 970-Xeon comparison? <img src="confused.gif" border="0">
<strong>In that case, Apple is on a good road. 2.5 ghz? Those suckers aregone. But what I want to see is how the PPC 970 compares against the Xeon. Does anyone know of a 970-Xeon comparison? </strong><hr></blockquote>
You don't want to do that. Xeons can be run in SMP systems and support Hyperthreading. That would be biting off a little too much to chew.
What are the real world benefits of hyperthreading?
I heard that IBM mulled over adding hyperthreading logic to the POWER4 as they were designing it, but decided not to as it had limited real world advantages.
Is it buzzword compliant crap or something that could actually be useful?
Hyper threading is making the processor think that it is two. This is strange, because aren't dual processors usually slower than their net speed? Like 80% of their net speed or something like that.
<strong>What are the real world benefits of hyperthreading?</strong><hr></blockquote>
HT offers some of the benefit of multiple cores (or standard SMP) at a much lower cost in die size. Roughly, if the percentage increase in performance is larger than the percentage increase in the number of transistors needed to implement it, it's a net gain. Unless my memory is failing me, the HT P4 gains a 20% performance boost running threaded code for a 5% increase in the number of transistors required.
How much of a benefit that is boils down to quality of implementation.
[quote]<strong>Is it buzzword compliant crap or something that could actually be useful?</strong><hr></blockquote>
It's useful to the P4 largely because the P4 is single-core, and it can't do SMP. It's much less useful to the POWER4 because the POWER4 does multicore and SMP already. Its usefulness to the P4 is limited by the paucity of multithreaded applications on Windows, but to the extent that it makes up for Windows' miserably inefficient thread management it might encourage multithreading on that platform. (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.)
On Macs, SMP has been a reality for some years now, OS X has excellent support, and Cocoa apps almost all have at least two threads, so SMT has an immediate practical application on the Mac, especially in machines that are unlikely to ever get dual processors.
[quote] (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.) <hr></blockquote>
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
It isn't quite that simple...we are running quite a few dual processor PIV Xeon DP's with HyperThreading turned on (effectively quad cpu machines to the os). Why? Because our apps can take advantage of the additional cpu's. One of the limitations that you *can* run into is the fact that the OS doesn't know if the processor it is using is a virtual processor or a real processor. Tests that have been done show the virtual processor at about 70% of the actual cpu in speed.
The next realease of windows is supposed to be HT aware. In other-words, the OS will delegate the less meaningful and less cpu intensive tasks to the virtual processor. Although, we really don't care that much about Windows...more about Linux/Unix/FreeBSD.
Also, you can disable HT. Some aplications will actually run faster with HT disabled while others run faster with it on. The only suggestion I would have for someone needing power such as this is to do some real world testing ;-)
RE the 970 vs PIV Xeon DP and MP - I'm not too certain the 970 will be faster. We will just have to wait and see some real world tests. I wonder, however, how many companys will be quick to replace Xeon boxes with a 970 box. Apple is still missing too much in the way of quality server hardware to make us switch (check out the Dell 2650 and 6650). However, FreeBSD is being ported to PPC
Its usefulness to the P4 is limited by the paucity of multithreaded applications on Windows, but to the extent that it makes up for Windows' miserably inefficient thread management it might encourage multithreading on that platform. (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.)
</strong><hr></blockquote>
I just upgraded our MS SQL machine (running Windows Advanced Server and SQL 7 (going to 2000) to a Dell 2650 with dual 2.4's. We tested SQL with HT on and with HT off. We were actually very suprised that Windows and SQL did better with HT on.
My BIG dissapointment was that SQL wouldn't even run with more than 2G RAM installed. We knew that SQL was limited but we have some other applications that run on the machine as well and we were wanting to keep 6G RAM for use with the other apps...NOPE...SQL won't even start up with more than 2G RAM in the machine. Grrrrrr.
Isn't HT causing licensing problems? For instance say my company is running ISA Server and it's license is Per Processor. Won't HT or any SMT system cause problems with these apps? Has Microsoft already fixed this?
It's quite obvious that Per Processor licenses will have to be dramatically changed. Imagine a 32 Way Server with 4 threads per proc and it's licensing ....OUCH!
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
<strong>Usually the beurocratic licensing riffraff doesn't get that technical.</strong><hr></blockquote>
The licensing riffraff can get extremely technical if it's enforced in software, which is one reason why MS snapped up Virtual PC. IBM actually has software that divides its physical hardware up into virtual machines with 1 or two processors for the sake of licenses like this: An application with a single CPU license can thereby run happily on a gigantic, massively MP beast of a computer because it's only made aware of the existence of one processor.
The bottom line is that if the per-processor line is software enforced, and the software has no way to tell a dual-threaded CPU from a dual-core CPU from two CPUs because of OS hardware abstraction, it'll treat them all as dual CPU machines and enforce the license accordingly.
Mike: FYI, you are aware that MS doesn't recommend that you run more than one server app per machine, right? SQL Server appears to have hardcoded that recommendation. <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
Comments
It should be available. However I wouldn't mind Apple shipping DDR 333 with full support for 400 as to not delay machines in any way due to supply issues.
<strong>maybe i just do not get it..but can somebody confirm ..the 970s architecture is 2times faster than a g4 and therefore almost 4 times faster than a pentium4?
i mean worst case...1.4ghz single (in imac too?!)
would be equal to a 2.8ghz g4 or a 5.6ghz pentium4?
and then..maybe a dual 1.8ghz equal to 14.4ghz pentium4?
which is about 4 times faster than the best pentium money can buy?
err..am i nuts?
[ 03-14-2003: Message edited by: cycle ]</strong><hr></blockquote>
It's 1.5-2x times faster then a G4 at the same speed. The 1.8ghz 970 is on par with the P4 2.8-3.0ghz.
<strong>
It's 1.5-2x times faster then a G4 at the same speed. The 1.8ghz 970 is on par with the P4 2.8-3.0ghz.</strong><hr></blockquote>
Aren't those numbers based on Spec which always favored x86?
<strong>In that case, Apple is on a good road. 2.5 ghz? Those suckers aregone.
You don't want to do that. Xeons can be run in SMP systems and support Hyperthreading. That would be biting off a little too much to chew.
I heard that IBM mulled over adding hyperthreading logic to the POWER4 as they were designing it, but decided not to as it had limited real world advantages.
Is it buzzword compliant crap or something that could actually be useful?
<strong>
It's 1.5-2x times faster then a G4 at the same speed. The 1.8ghz 970 is on par with the P4 2.8-3.0ghz.</strong><hr></blockquote>
well and a g4 is about 1.5x times faster than a pentium 4
a g4 1.4ghz beats the crap out of a 2 or 2.5ghz pentium4
and yah..i mean photoshop and stuff...i dont care about checking mail...i care about render speed
<strong>
Aren't those numbers based on Spec which always favored x86?</strong><hr></blockquote>
Yes, that's why they are good numbers
<strong>What are the real world benefits of hyperthreading?</strong><hr></blockquote>
HT offers some of the benefit of multiple cores (or standard SMP) at a much lower cost in die size. Roughly, if the percentage increase in performance is larger than the percentage increase in the number of transistors needed to implement it, it's a net gain. Unless my memory is failing me, the HT P4 gains a 20% performance boost running threaded code for a 5% increase in the number of transistors required.
How much of a benefit that is boils down to quality of implementation.
[quote]<strong>Is it buzzword compliant crap or something that could actually be useful?</strong><hr></blockquote>
It's useful to the P4 largely because the P4 is single-core, and it can't do SMP. It's much less useful to the POWER4 because the POWER4 does multicore and SMP already. Its usefulness to the P4 is limited by the paucity of multithreaded applications on Windows, but to the extent that it makes up for Windows' miserably inefficient thread management it might encourage multithreading on that platform. (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.)
On Macs, SMP has been a reality for some years now, OS X has excellent support, and Cocoa apps almost all have at least two threads, so SMT has an immediate practical application on the Mac, especially in machines that are unlikely to ever get dual processors.
<hr></blockquote>
Lemon Bon Bon
[quote] (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.) <hr></blockquote>
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
[ 03-14-2003: Message edited by: os10geek ]</p>
<strong>Originally posted by Amorph:
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
[ 03-14-2003: Message edited by: os10geek ]</strong><hr></blockquote>
It isn't quite that simple...we are running quite a few dual processor PIV Xeon DP's with HyperThreading turned on (effectively quad cpu machines to the os). Why? Because our apps can take advantage of the additional cpu's. One of the limitations that you *can* run into is the fact that the OS doesn't know if the processor it is using is a virtual processor or a real processor. Tests that have been done show the virtual processor at about 70% of the actual cpu in speed.
The next realease of windows is supposed to be HT aware. In other-words, the OS will delegate the less meaningful and less cpu intensive tasks to the virtual processor. Although, we really don't care that much about Windows...more about Linux/Unix/FreeBSD.
Also, you can disable HT. Some aplications will actually run faster with HT disabled while others run faster with it on. The only suggestion I would have for someone needing power such as this is to do some real world testing ;-)
RE the 970 vs PIV Xeon DP and MP - I'm not too certain the 970 will be faster. We will just have to wait and see some real world tests. I wonder, however, how many companys will be quick to replace Xeon boxes with a 970 box. Apple is still missing too much in the way of quality server hardware to make us switch (check out the Dell 2650 and 6650). However, FreeBSD is being ported to PPC
<strong>
Its usefulness to the P4 is limited by the paucity of multithreaded applications on Windows, but to the extent that it makes up for Windows' miserably inefficient thread management it might encourage multithreading on that platform. (Note that a HT P4 is a bit slower than a non-HT P4 at running single threaded apps.)
</strong><hr></blockquote>
I just upgraded our MS SQL machine (running Windows Advanced Server and SQL 7 (going to 2000) to a Dell 2650 with dual 2.4's. We tested SQL with HT on and with HT off. We were actually very suprised that Windows and SQL did better with HT on.
My BIG dissapointment was that SQL wouldn't even run with more than 2G RAM installed. We knew that SQL was limited but we have some other applications that run on the machine as well and we were wanting to keep 6G RAM for use with the other apps...NOPE...SQL won't even start up with more than 2G RAM in the machine. Grrrrrr.
Isn't HT causing licensing problems? For instance say my company is running ISA Server and it's license is Per Processor. Won't HT or any SMT system cause problems with these apps? Has Microsoft already fixed this?
It's quite obvious that Per Processor licenses will have to be dramatically changed. Imagine a 32 Way Server with 4 threads per proc and it's licensing ....OUCH!
<strong>Originally posted by Amorph:
Wait...if that is true, then why not get "real" multiprocessor Xeon systems, instead of settling for this faux-MP solution? Who is the HT P4 marketed towards?
</strong><hr></blockquote>
Price a HT P4. Then price a dual Xeon.
<strong>Usually the beurocratic licensing riffraff doesn't get that technical.</strong><hr></blockquote>
The licensing riffraff can get extremely technical if it's enforced in software, which is one reason why MS snapped up Virtual PC. IBM actually has software that divides its physical hardware up into virtual machines with 1 or two processors for the sake of licenses like this: An application with a single CPU license can thereby run happily on a gigantic, massively MP beast of a computer because it's only made aware of the existence of one processor.
The bottom line is that if the per-processor line is software enforced, and the software has no way to tell a dual-threaded CPU from a dual-core CPU from two CPUs because of OS hardware abstraction, it'll treat them all as dual CPU machines and enforce the license accordingly.
Mike: FYI, you are aware that MS doesn't recommend that you run more than one server app per machine, right? SQL Server appears to have hardcoded that recommendation. <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
[ 03-15-2003: Message edited by: Amorph ]</p>