Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
PowerPC 970 up to 2.5GHz?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Oh yes, got to beat that competition
Since there is a good of 970 vs. Intel discussion going, I thought I'd pass along this link - apparently PC magazine has the dirt, skinny, dig (not rumor) on Intel's roadmap for the next three years (which appears to mirror rumor). The link is:
http://www.pcmag.com/article2/0,4149,906060,00.asp
"Oh boy, do we ever have the goods on Intel.
Our crack reporter, Mark Hachman, got his hands on some killer information basically Intel's entire roadmap for the next few years.
It's way too much for just one article, so we've got him slaving over the keyboard, banging out the information as fast as he can."
Since there is a good of 970 vs. Intel discussion going, I thought I'd pass along this link - apparently PC magazine has the dirt, skinny, dig (not rumor) on Intel's roadmap for the next three years (which appears to mirror rumor). The link is:
http://www.pcmag.com/article2/0,4149,906060,00.asp
"Oh boy, do we ever have the goods on Intel.
Our crack reporter, Mark Hachman, got his hands on some killer information basically Intel's entire roadmap for the next few years.
It's way too much for just one article, so we've got him slaving over the keyboard, banging out the information as fast as he can."
I'm curious as to whether there is RAM available right now that is able to take full advantage of the 800mhz bus of the 970, if so what kind is it pc4500? (just threw that out there) Also how expensive is this ram that is fast enough?
IBM .pdf from IBM technical library has pictures and a good deal of information
For a PICTURE - see page five of this .pdf from the IBM Microelectronics page (takes awhile to load)
http://www-3.ibm.com/chips/techlib/techlib.nsf/techdocs/A1387A29AC1C2AE087256C5200611780/$file/PPC970_MPF2002.pdf
There appears to be a fair amount of information on IBM's site.
Originally posted by bertagert
I want to see what one looks like. Anyone have a link to a picture of the 970? Or, get your photoshop out and lift that heatsink off so i can take a peak.
Edited for spelling
For a PICTURE - see page five of this .pdf from the IBM Microelectronics page (takes awhile to load)
http://www-3.ibm.com/chips/techlib/techlib.nsf/techdocs/A1387A29AC1C2AE087256C5200611780/$file/PPC970_MPF2002.pdf
There appears to be a fair amount of information on IBM's site.
For a 6.4GB/s interconnect, dual channel PC3200 DDR would be required.
Or quad-channel PC800 RDRAM... heh.
That mock-up blade sucks... is that 133MHz SDRAM? If it works, it is going to be a poor performer, hence it is a mock-up - but it shows that you can implement the PPC 970 interconnect in a FPGA like the Xilinx chip on the board.
Read this in context!!!
Note that I quoted "single CPU tasks" specifically!
Render farms are not about scaling tasks in a single SMP box. They're about getting as much compute power as possible directed towards lots of smaller jobs.
Scaling is not as important as raw system power per CPU, and density of CPUs.
An E15K is awesome, almost as awesome as its price.
Note that I quoted "single CPU tasks" specifically!
Render farms are not about scaling tasks in a single SMP box. They're about getting as much compute power as possible directed towards lots of smaller jobs.
Scaling is not as important as raw system power per CPU, and density of CPUs.
An E15K is awesome, almost as awesome as its price.
Re: IBM .pdf from IBM technical library has pictures and a good deal of information
UPDATE - there are some add'l really nifty PICTURES in the IBM .pdf. For whatever reason - the link above will just lead to an access denied screen. TRY: IBM.com - go to "products", then "microelectronics", then "powerpc", then under the choices to select an offering, go to "9xx 64-bit", the .pdf link comes up and is accessable (although large). I attempted to just attach a copy that I'd saved, but it was too large to post.
UPDATE - there are some add'l really nifty PICTURES in the IBM .pdf. For whatever reason - the link above will just lead to an access denied screen. TRY: IBM.com - go to "products", then "microelectronics", then "powerpc", then under the choices to select an offering, go to "9xx 64-bit", the .pdf link comes up and is accessable (although large). I attempted to just attach a copy that I'd saved, but it was too large to post.
"clueless" comes to mind
Is that why you have a 32-bit moniker? ("ktlx" fits in 4 8-bit bytes)
Why on earth would you think that you'd need a 128-bit computer to deal with longer strings?
OhMyGawd - I'd better change my postname - "AidenShaw" is 72-bits, no way anyone will be able to reply to me if they're on a 32-bit system.
Sheesh....
Is that why you have a 32-bit moniker? ("ktlx" fits in 4 8-bit bytes)
Why on earth would you think that you'd need a 128-bit computer to deal with longer strings?
OhMyGawd - I'd better change my postname - "AidenShaw" is 72-bits, no way anyone will be able to reply to me if they're on a 32-bit system.
Sheesh....
Re: what if....
THis is all I have been thinking about ever since the 970 hyper began..... Even though apple would be insane not to use this, they might not.... ahhhhhhhh don't get my hopes down.
THis is all I have been thinking about ever since the 970 hyper began..... Even though apple would be insane not to use this, they might not.... ahhhhhhhh don't get my hopes down.
Re: Re: Business
Um.. huh? I really don't get this.. Surely the first consideration is "do i want to use XP or OS X?" and that will decide it.
I don't care if DELL has 50Ghz XP workstations while apply only has 5, i'll still get the mac, because i don't want to use Windows.
Um.. huh? I really don't get this.. Surely the first consideration is "do i want to use XP or OS X?" and that will decide it.
I don't care if DELL has 50Ghz XP workstations while apply only has 5, i'll still get the mac, because i don't want to use Windows.
Re: Re: Re: Re: Re: Re: When will this happen?
I read that same article. We all did. I did call them stupid, in an earlier post. As have several other people... Many folks, specifically the likes of IBM and AMD think intel is betting WAY wrong on this. 970 is a consumer chip, and is 64bit, and is coming out this year. Same with AMD's x86-64 offering.
4GB of RAM is not prohibitively expensive. The article says it costs $1k, well yeah, if you buy it as a single module. But I filled up my Dual 800 with 1.5GB of RAM for $150. And that was 2 years ago.
2008? Please. Thats a REALLY LONG TIME. Intel got this one wrong.
Mark this post, come back in 2008.
I read that same article. We all did. I did call them stupid, in an earlier post. As have several other people... Many folks, specifically the likes of IBM and AMD think intel is betting WAY wrong on this. 970 is a consumer chip, and is 64bit, and is coming out this year. Same with AMD's x86-64 offering.
4GB of RAM is not prohibitively expensive. The article says it costs $1k, well yeah, if you buy it as a single module. But I filled up my Dual 800 with 1.5GB of RAM for $150. And that was 2 years ago.
2008? Please. Thats a REALLY LONG TIME. Intel got this one wrong.
Mark this post, come back in 2008.
Re: Re: Re: Re: Blade
but if you aren't interested in other people's opinions, why state your own? who cares about them? honestly, the aesthetic appeal of a computer is all but entirely subjective, so if his opinion doesn't matter, there's no way in hell that yours does. besides, he actually brought up a logical point that doesn't even touch on aesthetics like yours does. you weren't stating a fact. here's a fact. IBM's website is effectively designed, even though you think it's ugly. do you have any notion of the vast amount of information they purvey to the public from that site? IBM is a really stable, pretty successful company, one reason i hope apple switches to them instead of sticking with whatever the heck motorola pulls out their arses.
Re: Re: Re: Re: Re: Blade
hahaha, so i say ibm's servers are ugly like their website....*fact*....and now we're down to motorola sucks thats funny.
my opinion will always matter, it's mine i can show that box to anyone and bet that they would agree. oh and just so your not too far off track with the whole subjective thing....even the guy that said "who cares" didnt deny that it IS ugly. anyway i dont really care i just post here for fun
hahaha, so i say ibm's servers are ugly like their website....*fact*....and now we're down to motorola sucks
thats funny.my opinion will always matter, it's mine
i can show that box to anyone and bet that they would agree. oh and just so your not too far off track with the whole subjective thing....even the guy that said "who cares" didnt deny that it IS ugly. anyway i dont really care i just post here for fun Don't know if anyone posted this yet. (I'm a little late to the party here.) But, Real World Tech has an article with a nice chart of the SPECint and SPECfp of most contenders in the chip race. Granted, the 970 isn't on there but POWER4 is. There are some very real differences between them (cache size, bus architecture, VMX unit), but they are using the same core at least. Take it for what it's worth but it's interesting and maybe an indicator of how performance of the 970 will compete.
Notice how the 1.4GHz POWER4+ is a nice balance between the FP performance of the 1GHz Itanium and the integer performance of the 3GHz Pentium 4.
http://www.realworldtech.com/page.cfm?AID=RWT012603224711&p=2
Notice how the 1.4GHz POWER4+ is a nice balance between the FP performance of the 1GHz Itanium and the integer performance of the 3GHz Pentium 4.
http://www.realworldtech.com/page.cfm?AID=RWT012603224711&p=2
Re: "clueless" comes to mind
Before you label someone as clueless and write a rude post, you ought to think a bit or at least even try to understand what is going on.
IPv6 uses 128-bit address fields. If your network appliance is acting like a router or a firewall, being able to do bit manipulation on the complete address all at once would mean you could process the address and network information much quicker than if you had a 32-bit or a 64-bit processor. That does not mean it cannot be done but it just is considerably slower and less efficient. By my calculations a 64-bit processor would need to be around four times as fast and a 32-bit processor around 16 times as fast to do routing and filtering compared to a 128-bit processor.
That said, I am not sure the network appliance market is large enough to support a 128-bit general purpose CPU. That is why I said a network coprocessor would probably be a better option in the same way that you have graphics coprocessors.
Next time think before being a prick.
Before you label someone as clueless and write a rude post, you ought to think a bit or at least even try to understand what is going on.
IPv6 uses 128-bit address fields. If your network appliance is acting like a router or a firewall, being able to do bit manipulation on the complete address all at once would mean you could process the address and network information much quicker than if you had a 32-bit or a 64-bit processor. That does not mean it cannot be done but it just is considerably slower and less efficient. By my calculations a 64-bit processor would need to be around four times as fast and a 32-bit processor around 16 times as fast to do routing and filtering compared to a 128-bit processor.
That said, I am not sure the network appliance market is large enough to support a 128-bit general purpose CPU. That is why I said a network coprocessor would probably be a better option in the same way that you have graphics coprocessors.
Next time think before being a prick.
How about a quad PowerPC FPLA then?
No, but I can guess based on the prices Cisco charges for their network gear!
http://www.xilinx.com/xlnx/xil_prodcat_landingpage.jsp?title=Virtex-II+Pro+FPGAs
No, but I can guess based on the prices Cisco charges for their network gear!
http://www.xilinx.com/xlnx/xil_prodcat_landingpage.jsp?title=Virtex-II+Pro+FPGAs
