Intel will max out at 4 cpus. 2 quad-cores are likely to max out the FSB 4 quad-cores are likely to be a lot worse this is where AMD will kick a lot of a**.
1066MHz are going to be a lot slower then the 1333MHz onecs
How is this going to affect me if I say use a quad core or dual quad-core system in a renderfarm environment? That was rhetorical, the answer is it wouldn't. When comparing multicore systems the FSB does come into play when you have applications that are likely to be memory bound or latency bound. If I was running an enterprise database I'd be looking at the Opteron's a bit harder or I'd look at Tulsa with it's huge L3 cache. However FSB saturation really only comes into play for some truly high end servers so I doubt the FSB will be a real limiting factor for a workstation.
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
How is this going to affect me if I say use a quad core or dual quad-core system in a renderfarm environment? That was rhetorical, the answer is it wouldn't. When comparing multicore systems the FSB does come into play when you have applications that are likely to be memory bound or latency bound. If I was running an enterprise database I'd be looking at the Opteron's a bit harder or I'd look at Tulsa with it's huge L3 cache. However FSB saturation really only comes into play for some truly high end servers so I doubt the FSB will be a real limiting factor for a workstation.
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
FSB saturation is more likey when each die needs to use the fsb to talk to each core as well as useing the same bus to get to ram the big l2 helps but when amd comes with it's true quad-cores with l3 cache intel may be back where they where with the duel cores.
also in 4 cpu systems haveing 8 cores on one fsb is going to be a big choke point at least they should put the other 8 on there own bus and then still they may choke even more at the chip set to ram link.
also in 4 cpu systems haveing 8 cores on one fsb is going to be a big choke point at least they should put the other 8 on there own bus and then still they may choke even more at the chip set to ram link.
I do not believe that you know what you are talking about. It looks like you are assuming the old paradigms are being carried forward. They aren't. Heck, I'll just say you are simply wrong here in relation to the upcoming four core processors.
Like Woodcrest and the 5000 chipsets, each Clovertown chip package has its own bus to the chipset. The workstations are dual FSB. Also, the four-socket system based on the Tigerton chip (four socket version of Clovetown/Kentsfield) will have one FSB per CPU package, meaning it is a quad FSB, so the servers can run sixteen CPU cores without being choked by FSB contention because the FSB bandwidth is so high.
Ugh, just another hack job by Intel to say "we beat you to it!"
Two dual core processors glued together and using the same bus....
Unles you really need 8 cores in your Mac Pro, stay away!
Not really. If you do a lot of video and editing, these processors are for you. I saw some specs on a pair of these dual quads a few months back, and it was impressive. There was like 40% to 70% performance gains in encoding across the board, and photoshop had similar results. After effects was another application tested, and there is definitely significant performance gains in applications that are developed to take advantage of multiprocessor environments.
These is not a useless update. Just fitting two dual core woodcrests on a single chip the same size as woodcrest is amazing.
I think Apple will quietly do a PowerMac silent update (just change the options page when they are available, and not make a big fuss about it) once they are available. They fit right in the same socket as the woodcrest.
Hopefully they will update the graphics card options at the same time.
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
Exactly!
Also, according to Anandtech, the existing Mac Pro will accept Clovertown, so you all that have one will be able to DIY prior to MWSF.
H.264 (AKA AVC) HD takes a LONG time, but I don't know if applications like MPEG Streamclip, Episode (AKA Compression Master), or FCP will take advantage of the additional cores without upgraded SW?
So when can I get a 4 core iMac? Is there a roadmap for lower power 4 core chips?
We might see this when Intel goes 45nm, in a year and 3 months, or a bit later. That's when they will begin to move to the power saving on die memory controller as well.
Is it my imagination or does Intel come up with new/faster processors every few months? It seems every other week, I read something about new chips from Intel that can be used in Macs. I hate to say it but it looks like Apple should have jumped to Intel years ago.
The move to Intel was another bold move by Steve... I hope Gil Amelio is taking notes. THIS is how it’s done.
Steve had no choice. When Amelio was there, the prediction was that the PowerPC would knock Intel off the shelf. There was no reason to switch then. x86 was being slammed.
Intel will max out at 4 cpus. 2 quad-cores are likely to max out the FSB 4 quad-cores are likely to be a lot worse this is where AMD will kick a lot of a**.
1066MHz are going to be a lot slower then the 1333MHz onecs
I think it's a bit of both. Intel has gone a little too hard in trying to beat AMD, it's like beating the f*k out of a pinata, where you're kind of drunk and just bashing away hardcore until there's just like a little string of paper left and you're still bashing it, and all the kids by this stage are crying with the overperformance of violence.
There are various stages of using more than one core, from multitasking, to dualcore-aware software, to rare software that picks up on 4 to 8 cores.
IIRC Unreal Tournament 2007 should be at least dualcore-aware.
It's a bit of Intel pushing real hard and software developers catching up. A Quad MacPro is mighty fine in and of itself. I would say the bottleneck is efficient coding, say for a single render or encode task, to make the most of 4 cores.
Now that Intel is ahead, they want to be seen as staying ahead.
If the cores double every 18 months then I would expect 32 cores mid-2011. But that assumes a lot of things, but most of the assumptions are fairly reasonable. It's quite a bit easier to stuff more transistors on a die than it is to redirect development to alternative uses of those transistors that may be more effective. The biggest potential setback is if there's a choke in fab technology. AMD's really choking on the 65nm transition and there are transitions further down the line that have hurdles that need to be overcome in order for anyone to use it.
Did you see Intel's discussion today of an 80-core chip by around 2010? There is more than just this going on in the industry, and probably even just at Intel. This is going to force a change in how software is written for these things, and the kinds of things we expect to use them for. Get out of the current paradigm, and get ready for the new one.
Did you see Intel's discussion today of an 80-core chip by around 2010?
It looks like those 80 cores weren't general purpose CPU cores though but rather simplified cores designed for floating point performance. Assuming it's functional, you can get a bunch of those chips in a refrigerator-sized computer and it might challenge Blue Gene.
That wafer was specialized cores. But then so is the Cell. So?
My prediction was based on general purpose cores, which is a lot tougher to do than specialized cores. Specialized cores can be made very small, very high performance, but at the expense of functionality. That's all I'm saying with respect to your original response to me.
Originally Posted by 132GHz,4TB DDR8 SDRAM,95TB HDD
Ok..someone please clarify-
will the games industry (PC, not console) make much more use of a Quad Core chip than a dual core anytime within the next 3 years ?
The thing is for 2006 and 2007 a strong singlecore Athlon64 with 2GB of RAM and a 7200rpm drive can give good performance in games.... provided you have a solid GPU.
Remember that Conroe/Allendale whipping AMD's a$$ at gaming is in systems where they've been running SLI'ed 7900GTXs or even a single 7900GTX or equivalent. The frame rates achieved in both AMD and Conroe/Allendale systems are a lot of the time all over 100fps. I'm not saying the benchmarks aren't good, just that it was set out to see the maximum difference between "Gaming CPUs" by negating as much as possible GPU bottlenecks.
My point is the GPU is the bottleneck going forward to 2007. nVidia and ATI I am 110% sure are looking at the CPU production scene and know that a similar revolution has to take place soon. That is, 2007 to 2009, we should see 65nm and maybe 45nm GPUs alongside dual to eight-core GPUs -- and the similar challenge from DirectX10/ OpenGL x.x(1.5?) onwards to code for multicore GPUs.
As I said, rest of this year, and 2007, a good OC'ed singlecore on an aftermarket air heatsink is solid enough alongside 2GB RAM and 7200rpm or 15,000rpm hard disk.
In 2008 most CPUs will be dualcore anyway. An increasing number of games beyond just UT2007 will take advantage of two cores. And WinXP2Pro or Vista will have some stuff running in the background, so hopefully things are divided-up enough that the CPU load from the game and background tasks are spread as appropriately as possible over two cores.
In 2008 with say a 45nm 4-core Intel CPU, when it comes to PC gaming, again, it is possible Vista can handle spreading processes across the cores so your gaming experience, on the CPU side, has enough CPU processing "attention". This on top of newer games that are multicore-aware.
Finally, to repeat, keep in mind the GPU is very important in PC gaming and it has to move strongly alongside the performance-per-watt, performance-per-dollar-per-watt trail that Intel has blazed. Heat and power consumption and hence fan noise is increasingly a technical and marketing/ consumer experience issue, for example in the high-7-series nVidias and X1900 and above ATIs.
Since ATI has aligned with AMD, if nVidia sucks up to Intel and license out some of that sweet, sweet, sweet 65nm & 45nm multicore tech for their GPU, nVidia will kick ATI out past the stratosphere and back.
If I was nVidia management I'd be stalking Intel like Justin Timberlake-crazed groupies.
Comments
Intel will max out at 4 cpus. 2 quad-cores are likely to max out the FSB 4 quad-cores are likely to be a lot worse this is where AMD will kick a lot of a**.
1066MHz are going to be a lot slower then the 1333MHz onecs
How is this going to affect me if I say use a quad core or dual quad-core system in a renderfarm environment? That was rhetorical, the answer is it wouldn't. When comparing multicore systems the FSB does come into play when you have applications that are likely to be memory bound or latency bound. If I was running an enterprise database I'd be looking at the Opteron's a bit harder or I'd look at Tulsa with it's huge L3 cache. However FSB saturation really only comes into play for some truly high end servers so I doubt the FSB will be a real limiting factor for a workstation.
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
How is this going to affect me if I say use a quad core or dual quad-core system in a renderfarm environment? That was rhetorical, the answer is it wouldn't. When comparing multicore systems the FSB does come into play when you have applications that are likely to be memory bound or latency bound. If I was running an enterprise database I'd be looking at the Opteron's a bit harder or I'd look at Tulsa with it's huge L3 cache. However FSB saturation really only comes into play for some truly high end servers so I doubt the FSB will be a real limiting factor for a workstation.
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
FSB saturation is more likey when each die needs to use the fsb to talk to each core as well as useing the same bus to get to ram the big l2 helps but when amd comes with it's true quad-cores with l3 cache intel may be back where they where with the duel cores.
also in 4 cpu systems haveing 8 cores on one fsb is going to be a big choke point at least they should put the other 8 on there own bus and then still they may choke even more at the chip set to ram link.
also in 4 cpu systems haveing 8 cores on one fsb is going to be a big choke point at least they should put the other 8 on there own bus and then still they may choke even more at the chip set to ram link.
I do not believe that you know what you are talking about. It looks like you are assuming the old paradigms are being carried forward. They aren't. Heck, I'll just say you are simply wrong here in relation to the upcoming four core processors.
Like Woodcrest and the 5000 chipsets, each Clovertown chip package has its own bus to the chipset. The workstations are dual FSB. Also, the four-socket system based on the Tigerton chip (four socket version of Clovetown/Kentsfield) will have one FSB per CPU package, meaning it is a quad FSB, so the servers can run sixteen CPU cores without being choked by FSB contention because the FSB bandwidth is so high.
Ugh, just another hack job by Intel to say "we beat you to it!"
Two dual core processors glued together and using the same bus....
Unles you really need 8 cores in your Mac Pro, stay away!
Not really. If you do a lot of video and editing, these processors are for you. I saw some specs on a pair of these dual quads a few months back, and it was impressive. There was like 40% to 70% performance gains in encoding across the board, and photoshop had similar results. After effects was another application tested, and there is definitely significant performance gains in applications that are developed to take advantage of multiprocessor environments.
These is not a useless update. Just fitting two dual core woodcrests on a single chip the same size as woodcrest is amazing.
I think Apple will quietly do a PowerMac silent update (just change the options page when they are available, and not make a big fuss about it) once they are available. They fit right in the same socket as the woodcrest.
Hopefully they will update the graphics card options at the same time.
They are adding cores to chipsets faster than blades to razors.
Next thing we know, the Xeons in our Mac Pros will emit cooling, smoothing gel!
I know I'd be happy to see what an 8-core Mac Pro could do for encoding video.
Exactly!
Also, according to Anandtech, the existing Mac Pro will accept Clovertown, so you all that have one will be able to DIY prior to MWSF.
H.264 (AKA AVC) HD takes a LONG time, but I don't know if applications like MPEG Streamclip, Episode (AKA Compression Master), or FCP will take advantage of the additional cores without upgraded SW?
So when can I get a 4 core iMac? Is there a roadmap for lower power 4 core chips?
We might see this when Intel goes 45nm, in a year and 3 months, or a bit later. That's when they will begin to move to the power saving on die memory controller as well.
Is it my imagination or does Intel come up with new/faster processors every few months? It seems every other week, I read something about new chips from Intel that can be used in Macs. I hate to say it but it looks like Apple should have jumped to Intel years ago.
The move to Intel was another bold move by Steve... I hope Gil Amelio is taking notes. THIS is how it’s done.
Steve had no choice. When Amelio was there, the prediction was that the PowerPC would knock Intel off the shelf. There was no reason to switch then. x86 was being slammed.
Intel will max out at 4 cpus. 2 quad-cores are likely to max out the FSB 4 quad-cores are likely to be a lot worse this is where AMD will kick a lot of a**.
1066MHz are going to be a lot slower then the 1333MHz onecs
The top chips will still use 1333.
I think it's a bit of both. Intel has gone a little too hard in trying to beat AMD, it's like beating the f*k out of a pinata, where you're kind of drunk and just bashing away hardcore until there's just like a little string of paper left and you're still bashing it, and all the kids by this stage are crying with the overperformance of violence.
There are various stages of using more than one core, from multitasking, to dualcore-aware software, to rare software that picks up on 4 to 8 cores.
IIRC Unreal Tournament 2007 should be at least dualcore-aware.
It's a bit of Intel pushing real hard and software developers catching up. A Quad MacPro is mighty fine in and of itself. I would say the bottleneck is efficient coding, say for a single render or encode task, to make the most of 4 cores.
Now that Intel is ahead, they want to be seen as staying ahead.
If the cores double every 18 months then I would expect 32 cores mid-2011. But that assumes a lot of things, but most of the assumptions are fairly reasonable. It's quite a bit easier to stuff more transistors on a die than it is to redirect development to alternative uses of those transistors that may be more effective. The biggest potential setback is if there's a choke in fab technology. AMD's really choking on the 65nm transition and there are transitions further down the line that have hurdles that need to be overcome in order for anyone to use it.
Did you see Intel's discussion today of an 80-core chip by around 2010? There is more than just this going on in the industry, and probably even just at Intel. This is going to force a change in how software is written for these things, and the kinds of things we expect to use them for. Get out of the current paradigm, and get ready for the new one.
Did you see Intel's discussion today of an 80-core chip by around 2010?
It looks like those 80 cores weren't general purpose CPU cores though but rather simplified cores designed for floating point performance. Assuming it's functional, you can get a bunch of those chips in a refrigerator-sized computer and it might challenge Blue Gene.
That wafer was specialized cores. But then so is the Cell. So?
My prediction was based on general purpose cores, which is a lot tougher to do than specialized cores. Specialized cores can be made very small, very high performance, but at the expense of functionality. That's all I'm saying with respect to your original response to me.
Regardless, we have been inching into a paradigm shift because of cpus with multiple cores. The goddess has many arms.
Yes, she does. But how many can she use at once?
will the games industry (PC, not console) make much more use of a Quad Core chip than a dual core anytime within the next 3 years ?
Ok..someone please clarify-
will the games industry (PC, not console) make much more use of a Quad Core chip than a dual core anytime within the next 3 years ?
See link below. It's coming.
http://www.anandtech.com/tradeshows/...spx?i=2841&p=2
Ok..someone please clarify-
will the games industry (PC, not console) make much more use of a Quad Core chip than a dual core anytime within the next 3 years ?
The thing is for 2006 and 2007 a strong singlecore Athlon64 with 2GB of RAM and a 7200rpm drive can give good performance in games.... provided you have a solid GPU.
Remember that Conroe/Allendale whipping AMD's a$$ at gaming is in systems where they've been running SLI'ed 7900GTXs or even a single 7900GTX or equivalent. The frame rates achieved in both AMD and Conroe/Allendale systems are a lot of the time all over 100fps. I'm not saying the benchmarks aren't good, just that it was set out to see the maximum difference between "Gaming CPUs" by negating as much as possible GPU bottlenecks.
My point is the GPU is the bottleneck going forward to 2007. nVidia and ATI I am 110% sure are looking at the CPU production scene and know that a similar revolution has to take place soon. That is, 2007 to 2009, we should see 65nm and maybe 45nm GPUs alongside dual to eight-core GPUs -- and the similar challenge from DirectX10/ OpenGL x.x(1.5?) onwards to code for multicore GPUs.
As I said, rest of this year, and 2007, a good OC'ed singlecore on an aftermarket air heatsink is solid enough alongside 2GB RAM and 7200rpm or 15,000rpm hard disk.
In 2008 most CPUs will be dualcore anyway. An increasing number of games beyond just UT2007 will take advantage of two cores. And WinXP2Pro or Vista will have some stuff running in the background, so hopefully things are divided-up enough that the CPU load from the game and background tasks are spread as appropriately as possible over two cores.
In 2008 with say a 45nm 4-core Intel CPU, when it comes to PC gaming, again, it is possible Vista can handle spreading processes across the cores so your gaming experience, on the CPU side, has enough CPU processing "attention". This on top of newer games that are multicore-aware.
Finally, to repeat, keep in mind the GPU is very important in PC gaming and it has to move strongly alongside the performance-per-watt, performance-per-dollar-per-watt trail that Intel has blazed. Heat and power consumption and hence fan noise is increasingly a technical and marketing/ consumer experience issue, for example in the high-7-series nVidias and X1900 and above ATIs.
Since ATI has aligned with AMD, if nVidia sucks up to Intel and license out some of that sweet, sweet, sweet 65nm & 45nm multicore tech for their GPU, nVidia will kick ATI out past the stratosphere and back.
If I was nVidia management I'd be stalking Intel like Justin Timberlake-crazed groupies.
.....................
.....................