Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
It did go down compared to Pent Ds. That power envelope(kentsfield/cloverton) is probably comparable to dual core opterons.
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
I think you have to keep in mind that generation of quad core is not much different than two Woodcrest dies stitched together in one package, and that this is a very high performance chip. The Kentsfield will probably be badged as an Extreme Edition, and those have a somewhat higher power rating than the standard desktop chip.
I think you have to keep in mind that generation of quad core is not much different than two Woodcrest dies stitched together in one package, and that this is a very high performance chip. The Kentsfield will probably be badged as an Extreme Edition, and those have a somewhat higher power rating than the standard desktop chip.
Yes, I believe Kentfield is going to be released as a "Core 2 Extreme" edition processor with unit price of at least $1299. It will still be cheaper (and faster) than AMD's 4x4 offering.
The 4x4 might trump intel in graphics capability if NVidia can put multiple graphics cards (SLI) directly on the HyperTransport bus instead of PCI-Express.
Of course the poster doesn't have it on record... no one's saying anything about Leopard on record. But considering how horrible 10.4 and earlier are at threading, it's not rocket science to presume that Apple is putting some serious development money in it. Hopefully they'll catch up to and leapfrog the other OSes with the work in Leopard.
Well I was joking about that, obviously I thought. I expect somewhere close to 0 improvement in threading. Apple went through and pretty thoroughly retuned threading performance for 10.4, so I don't expect their tuning priorities to have changed to database performance this go around. I realize there was some hubbub on the web but largely it was coming from people that didn't fully understand what they were looking at, and changed over time. Can't expect them to have a solid perspective.
I don't know how a lot of other software works, but I own a dual core 2 Ghz G5 and I have graphics and rendering software that is not optimized for multiple cores, however I am able to maximize my system by setting up seperate user accounts and using fast user switching.
I just set one image to render, switch over to a different account and set another image to render, and I have both cores running at full speed, churning out two different renders at the same time.
If I went to quad or octo cores, for my software that isn't optimized for it, I could just setup four to eight user accounts on the system and get all the cores working that way.
Has anyone else tried this and been successful with their software?
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
It was less power consumption in relation to the performance (performance per watt), not necessarily less power consumption.
Has anyone else tried this and been successful with their software?
I would say it'd probably be a lot easier to transition to software that does support multiple cores, especially for 3-D work. Likely cheaper in the long run too, if you balance your time sorting the users and problems it creates vs. the software cost.
Otherwise just buy a few minis instead of a tower and connect them to NAS and a KVM switch.
I would say it'd probably be a lot easier to transition to software that does support multiple cores, especially for 3-D work.
I totally agree, however I use Bryce for a lot of my stuff and to date it hasn't been optimized for multiple cores or processors, (mostly because it moved to a windows application only for the most part....still upset about that...lol) so this has been an effective workaround.
I also use several apps that take full advantage of my hardware and would much prefer that option in the future.
Won't Bryce output meshes for use in other renderers? If so that'd let you offload to likely a better renderer that was faster because it used multiple cores. A pity what's happened to Bryce really, over the years, but at least it hasn't been completely abandoned yet.
Of course the poster doesn't have it on record... no one's saying anything about Leopard on record. But considering how horrible 10.4 and earlier are at threading, it's not rocket science to presume that Apple is putting some serious development money in it. Hopefully they'll catch up to and leapfrog the other OSes with the work in Leopard.
I'm curious why you say MacOS X is "horrible at threading". What are you basing that conclusion on?
I'm curious why you say MacOS X is "horrible at threading". What are you basing that conclusion on?
I have seen at least one reference that supports this comparing MySQL performance on G5, Xeon, and Opteron. The article is a year old but it does talk about Tiger.
I have seen at least one reference that supports this comparing MySQL performance on G5, Xeon, and Opteron. The article is a year old but it does talk about Tiger.
Doing a quick study, I couldn't reach any conclusion as to why the benchmarks showed Mac OS X to be slow. One theory was that use of F_FULLFSYNC on Mac OS X made MySQL run slower but there were two counter arguments to that - 1) AnandTech points out that they used MyISAM which doesn't have any source code reference to F_FULLSYNC and 2) most of the tests are read only which requires, locking perhaps, but no Syncing.
Thread creation performance is usually not an issue - threads are usually pooled. So other threading issues like lock granularity have to looked at. So question is does Mac OS X do a good job implementing thread safe calls to its APIs?
Within a couple of months i will be able to draw my own conclusions when I have Linux, Windows, and Mac OS X running on the same Mac Pro hardware and I can benchmark my own stuff.
I suspected that that was the source of the "MacOS X sucks at threading" comment. Complete hogwash -- they were just guessing at the source of the slower MacOS X performance, and guessing about performance is worse than trying to predict the weather. The information provided by the Apple file system guy is far more believeable, and if I had to choose I would rather have my data correct than faster. The counter arguments may disprove the flush theory, but they do not prove the "sucks at threading" guess and until somebody really digs in and does due diligence on that one we're not going to know. Certainly I have so far seen no problems with MacOS X's threading support, and it actually seems to be quite good (especially in Tiger since the kernel funnels are significantly reduced). I've had nightmare problems with the various flavors of Windows and their scheduling algorithms, and while I have considerably less experience with MacOS X at that level I have seen no sign of similar problems.
I don't see the point of having more than 2 cores.
It seems to me that the only reason to do this is to avoid upgrading processor speeds.. just add another core and theoretical processing power goes up, right?
When doing things that can only run on one core, my dual 2.5Ghz G5 wouldn't be any slower than a quad, penta, or eleventybillion core machine.
I mean, I'm all for it, but many tasks aren't threaded and some will never be.
What am I going to do? Trans-code 4 DVDs at once?
I'm waiting for a dual core 5ghz before I upgrade. Benchmark programs are fine and dandy, but the real world calls for something a bit more practical.
Wikipedia is got to be in error. Clovertown is designed to work in the Bensley platform which supports two sockets. See TechReport article. Of course Kentfield is designed to fit Conroe motherboards and is not MP-capable.
Yes, be very careful with Wikipedia. It's a great place to start, but always follow the references given because I have found accuracy around 50%.
Not that it is completely wrong 50% of the time, but I see meaningful errors about that often. Remember ANYONE can post ANYTHING there! And those posts will stick unless someone else changes them which can devolve into a change war that means you get different results depending on WHEN you read it.
Comments
Originally posted by baranovich
"Introducing the new Mac Pro OCTO shipping January 2007"
hehe, The Mac Pro OCHO
Originally posted by wmf
Yeah, Kentsfield/Cloverton is 80-130W.
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
Originally posted by dmwogan
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
It did go down compared to Pent Ds. That power envelope(kentsfield/cloverton) is probably comparable to dual core opterons.
Originally posted by dmwogan
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
I think you have to keep in mind that generation of quad core is not much different than two Woodcrest dies stitched together in one package, and that this is a very high performance chip. The Kentsfield will probably be badged as an Extreme Edition, and those have a somewhat higher power rating than the standard desktop chip.
Originally posted by JeffDM
I think you have to keep in mind that generation of quad core is not much different than two Woodcrest dies stitched together in one package, and that this is a very high performance chip. The Kentsfield will probably be badged as an Extreme Edition, and those have a somewhat higher power rating than the standard desktop chip.
Yes, I believe Kentfield is going to be released as a "Core 2 Extreme" edition processor with unit price of at least $1299. It will still be cheaper (and faster) than AMD's 4x4 offering.
The 4x4 might trump intel in graphics capability if NVidia can put multiple graphics cards (SLI) directly on the HyperTransport bus instead of PCI-Express.
Originally posted by Junkyard Dawg
MiniTower
$1599 2.33 GHz Dual Mac Pro - Conroe single CPU
$1899 2.66 GHz Dual Mac Pro - Conroe single CPU
Tower
$2399 2.66 GHz Quad Mac Pro - Woodcrest dual CPU
$2799 3.00 GHz Quad Mac Pro - Woodcrest dual CPU
MegaTower
$3499 2.66 GHz Octo Mac Pro - Tigerton dual CPU
$3999 3.00 GHz Octo Mac Pro - Tigerton dual CPU
Well Dawg, I don't know about all that, but splitting the line in two is already a good idea.
--------------------August '06------------ January '07
Mini Tower 1: $1499 --- DC 2.40GHz Conroe --- DC 2.66GHz Conroe
Mini Tower 2: $1999 --- DC 2.66GHz Conroe --- DC 2.93GHz Conroe
Mini Tower 3: $2499 --- DC 2.93GHz Conroe --- QC 2.66GHz Kentsfield
Big Tower 1: $2999 --- DDC 2.66GHz Woody --- DDC 3.00GHz Woody
Big Tower 2: $3499 --- DDC 3.00GHz Woody --- DQC 3.00GHz CloverTown
Originally posted by Booga
Of course the poster doesn't have it on record... no one's saying anything about Leopard on record. But considering how horrible 10.4 and earlier are at threading, it's not rocket science to presume that Apple is putting some serious development money in it. Hopefully they'll catch up to and leapfrog the other OSes with the work in Leopard.
Well I was joking about that, obviously I thought. I expect somewhere close to 0 improvement in threading. Apple went through and pretty thoroughly retuned threading performance for 10.4, so I don't expect their tuning priorities to have changed to database performance this go around. I realize there was some hubbub on the web but largely it was coming from people that didn't fully understand what they were looking at, and changed over time. Can't expect them to have a solid perspective.
Originally posted by dmwogan
hehe, The Mac Pro OCHO
...or Mac80s.
I just set one image to render, switch over to a different account and set another image to render, and I have both cores running at full speed, churning out two different renders at the same time.
If I went to quad or octo cores, for my software that isn't optimized for it, I could just setup four to eight user accounts on the system and get all the cores working that way.
Has anyone else tried this and been successful with their software?
Originally posted by dmwogan
Anyone else remember about hearin lower power consumption and heat production? I know these are more powerful chips, but I was expecting power consumption to go down...I guess these are desktop/high performance chips.
It was less power consumption in relation to the performance (performance per watt), not necessarily less power consumption.
Originally posted by G520incher
Has anyone else tried this and been successful with their software?
I would say it'd probably be a lot easier to transition to software that does support multiple cores, especially for 3-D work. Likely cheaper in the long run too, if you balance your time sorting the users and problems it creates vs. the software cost.
Otherwise just buy a few minis instead of a tower and connect them to NAS and a KVM switch.
Originally posted by ChevalierMalFet
I would say it'd probably be a lot easier to transition to software that does support multiple cores, especially for 3-D work.
I totally agree, however I use Bryce for a lot of my stuff and to date it hasn't been optimized for multiple cores or processors, (mostly because it moved to a windows application only for the most part....still upset about that...lol) so this has been an effective workaround.
I also use several apps that take full advantage of my hardware and would much prefer that option in the future.
Originally posted by Booga
Of course the poster doesn't have it on record... no one's saying anything about Leopard on record. But considering how horrible 10.4 and earlier are at threading, it's not rocket science to presume that Apple is putting some serious development money in it. Hopefully they'll catch up to and leapfrog the other OSes with the work in Leopard.
I'm curious why you say MacOS X is "horrible at threading". What are you basing that conclusion on?
Originally posted by Programmer
I'm curious why you say MacOS X is "horrible at threading". What are you basing that conclusion on?
I have seen at least one reference that supports this comparing MySQL performance on G5, Xeon, and Opteron. The article is a year old but it does talk about Tiger.
Originally posted by mwswami
I have seen at least one reference that supports this comparing MySQL performance on G5, Xeon, and Opteron. The article is a year old but it does talk about Tiger.
Try reading this: http://ridiculousfish.com/blog/?p=17
Originally posted by gregmightdothat
Try reading this: http://ridiculousfish.com/blog/?p=17
Thanks for the link. I also found the follow up from AnandTech as well.
Doing a quick study, I couldn't reach any conclusion as to why the benchmarks showed Mac OS X to be slow. One theory was that use of F_FULLFSYNC on Mac OS X made MySQL run slower but there were two counter arguments to that - 1) AnandTech points out that they used MyISAM which doesn't have any source code reference to F_FULLSYNC and 2) most of the tests are read only which requires, locking perhaps, but no Syncing.
Thread creation performance is usually not an issue - threads are usually pooled. So other threading issues like lock granularity have to looked at. So question is does Mac OS X do a good job implementing thread safe calls to its APIs?
Within a couple of months i will be able to draw my own conclusions when I have Linux, Windows, and Mac OS X running on the same Mac Pro hardware and I can benchmark my own stuff.
It seems to me that the only reason to do this is to avoid upgrading processor speeds.. just add another core and theoretical processing power goes up, right?
When doing things that can only run on one core, my dual 2.5Ghz G5 wouldn't be any slower than a quad, penta, or eleventybillion core machine.
I mean, I'm all for it, but many tasks aren't threaded and some will never be.
What am I going to do? Trans-code 4 DVDs at once?
I'm waiting for a dual core 5ghz before I upgrade. Benchmark programs are fine and dandy, but the real world calls for something a bit more practical.
Originally posted by mwswami
Wikipedia is got to be in error. Clovertown is designed to work in the Bensley platform which supports two sockets. See TechReport article. Of course Kentfield is designed to fit Conroe motherboards and is not MP-capable.
Yes, be very careful with Wikipedia. It's a great place to start, but always follow the references given because I have found accuracy around 50%.
Not that it is completely wrong 50% of the time, but I see meaningful errors about that often. Remember ANYONE can post ANYTHING there! And those posts will stick unless someone else changes them which can devolve into a change war that means you get different results depending on WHEN you read it.