Hey on the front page of anandtech they talk about the next intel processor being dual core. Can someone explain what this means and how would i benefit?
Intels Dual Core?
- Thread starter phillyman36
- Start date
It means that you would have two processors working in parallel but in the same chip, this would lead to a much better performance because the comunitation between the two cores is considerably faster. Actually intel abandoned the project, because their dual core processor got hotter than sun. Just imagine two presscot cores in a same chip!! it is really hot!!, you wold need liquid nitrogen to cool it. AMD is not having that kind problem and they will release their dual core processor in 2005 i think.
dual cored is the same thing as having 2 cpu's, and carlosd explained it well. it doesnt make the speed of the system any faster. it allows for extreme multi-tasking. as for amd's dual cored, it's planned for Q4 2005-Q1 2006.
If Intel believes that anyone is going to believe they are going to release three different dual core versions by mid 2005 they are smoking some serious silicon. No tape out yet even. Are they based on Prescott? Dear God the thermal issues and psu issues alone are frightening if they are based on Prescott. If they are Dothan based then they have to redesign the cpu for hyperthreading SSE3 etc which will take time. Unless some serious leakage control and amazing turnaround processes occur it will be mid 2006 when we see them. They can't even get single core 3.6's out. AMD has taped out their dual core and have a platform designed from the ground up to utilize dual core. Will the dual core Intel's be compatible with existing chipsets like the AMD's? Interesting questions abound. If Intel can pull this off then they will remain competitve with AMD and with their superior marketshare hold of AMD's charge
Intel abandoned their P4 architecture seemingly because of production numbers much moreso than heat. Because of this they accelerated what they were already planning on doing for years now, going dual core.
Dual cores act just the same as if you had a dual processor system. Like a dual processor system, each core will have its own cache, and will have to fight for memory resources. Also like dual processor systems, there will be about a 50% boost over single core systems in applications that can use them.
A major reason hyperthreading was designed was if there was a cache miss then another execution unit could still be doing something useful, however it still isnt nearly as good as a true multi-processor/dual core system.
These processors are very likely based mostly on their mobile chips, the Pentium M's, as it doesnt make any sense to use the P4 cores. These chips are extremely energy efficient chips, with their spec sheets showing 21 watts maximum at 2.0Ghz. A64 true mobile chips are 35W, and desktop variants are usually quite a bit more than that. These chips also can match the A64 clock for clock on most applications.
With a dual core chip there isnt a reason for hyperthreading, as hyperthreading essentially emulates what a dual core chips naturally is. SSE3 support should not be anything remarkably difficult to add at all. These chips are already supported by current generation chipsets (Alderwood and Grantsdale in this case), so there isnt really a platform that needs to be designed.
Dual cores act just the same as if you had a dual processor system. Like a dual processor system, each core will have its own cache, and will have to fight for memory resources. Also like dual processor systems, there will be about a 50% boost over single core systems in applications that can use them.
A major reason hyperthreading was designed was if there was a cache miss then another execution unit could still be doing something useful, however it still isnt nearly as good as a true multi-processor/dual core system.
These processors are very likely based mostly on their mobile chips, the Pentium M's, as it doesnt make any sense to use the P4 cores. These chips are extremely energy efficient chips, with their spec sheets showing 21 watts maximum at 2.0Ghz. A64 true mobile chips are 35W, and desktop variants are usually quite a bit more than that. These chips also can match the A64 clock for clock on most applications.
With a dual core chip there isnt a reason for hyperthreading, as hyperthreading essentially emulates what a dual core chips naturally is. SSE3 support should not be anything remarkably difficult to add at all. These chips are already supported by current generation chipsets (Alderwood and Grantsdale in this case), so there isnt really a platform that needs to be designed.
Intel is actually ahead of schedule on dual core tech. 
http://www.digitimes.com/news/a20040804PR206.html
http://www.digitimes.com/news/a20040804PR206.html
Kovie
You can still use Hyper-threading on a dual core system, that just means you'll have 4 logical processors. The problem then becomes the fact that OS' like Windows XP Pro only support up to 2 processors and you have to go to Win2003 Server to use 4 processors (logical or not, AFAIK).
I would think that moving Dothan to dual-cores would be possible with less effort than making a dual-cored Prescott. We'll probably be looking at a dual-cored Dothan with an improved floating point unit and a shared 2 MB cache between the two cores (as opposed to 2 MB for a single core in the Pentium M). That would concievably have an acceptable die size. Also, current Dothan's are very much optimized for low power, by allowing for higher power requirements, there's a good chance that Intel could scale up the frequency significantly. This is all speculation on my part, it'll be interesting to see how it pans out.
Speaking of which, I'm kind of curious as to how AMD will handle the memory controller in it's dual core CPUs since it's usually on-die. Will they use one dual channel controller or 2 single channel controllers? I can't see them going to 2xDC though.
In any case, the interesting part is that it looks like Intel's dual core solution will be aimed at the desktop whereas AMD plans to debut it's dual core CPU in the server space (Opteron only, at first).
You can still use Hyper-threading on a dual core system, that just means you'll have 4 logical processors. The problem then becomes the fact that OS' like Windows XP Pro only support up to 2 processors and you have to go to Win2003 Server to use 4 processors (logical or not, AFAIK).
I would think that moving Dothan to dual-cores would be possible with less effort than making a dual-cored Prescott. We'll probably be looking at a dual-cored Dothan with an improved floating point unit and a shared 2 MB cache between the two cores (as opposed to 2 MB for a single core in the Pentium M). That would concievably have an acceptable die size. Also, current Dothan's are very much optimized for low power, by allowing for higher power requirements, there's a good chance that Intel could scale up the frequency significantly. This is all speculation on my part, it'll be interesting to see how it pans out.
Speaking of which, I'm kind of curious as to how AMD will handle the memory controller in it's dual core CPUs since it's usually on-die. Will they use one dual channel controller or 2 single channel controllers? I can't see them going to 2xDC though.
In any case, the interesting part is that it looks like Intel's dual core solution will be aimed at the desktop whereas AMD plans to debut it's dual core CPU in the server space (Opteron only, at first).
Zebo
Elite Member
- Jul 29, 2001
- 39,398
- 19
- 81
Correct me if I'm wrong... but is'nt dual core is useless to 99% of us. Like HT. Unless you mutlitask raw, single threaded High output single CPU is preferable anyday. This does'nt mean you're excluded from having more than one app open just if they are both using the CPU intensly at once, performance will suffer in single CPU.
since Intel delay the 4ghz launch, they might as well delay the dual core launch
make AMD happy:laugh:
make AMD happy:laugh:
It really depends on your apps and how they're coded. Games, for example could very much benefit from HT/multiple cores if they're properly coded. That is, use one core for AI calculations, another for physics/geometry, etc.
Media encoding could also benefit from dual cores by it's structure (2 CPUs encoding half a frame each at a time), same goes for rendering.
Dual cores will really need developers to start coding in ways that exploit the intrinsic parallelism of their programs. Look at the Apple side of things, nearly everything is going towards multiple threads since all recent PowerMacs are dual processor.
amd's dual cored wont just be use for server space only at first. when they debut the dual cored opterons, the dual cored toledo athlon fx will follow right behind them. its scheduled for release on Q4 2005 to Q1 2005. i'm just looking forward to the toledo.
Zebo
Elite Member
- Jul 29, 2001
- 39,398
- 19
- 81
Interesting. so it sounds like a programming transition deal more than anything, like 64 bit?
I just can't imagine a dual core CPU now, only because when I see a prescott sporting 114W+ @3.6 it would essentially double to 228W for just CPU!!! (assuming they just slapped two core together and presscot not centrino/dothan is what they are talking about dual coring ) And PSU's are only ~75% effient you're talking a minimum of 350W for just processor(s) add other stuff and I'm sure 550W PSUs will be the norm. Then the noise generated by a fan(s) that needs to disspate that much heat!!! My god it will sound like a server room in peoples office!
Zebo
Elite Member
- Jul 29, 2001
- 39,398
- 19
- 81
No it's not. It's going to be presshot based.
http://arstechnica.com/cpu/004/prescott-future/prescott-1.html
Keysplayr
Elite Member
- Jan 16, 2003
- 21,211
- 50
- 91
Oh I dont doubt it. But at some point, Intel will abandon the "Preshot" and go with the Pentium M Design team. I think it would be the best move considering a current 1.4 GHz PentiumM performs just as well and sometimes better than a Northy Prescott thats twice its GHz speed. Uses less voltage and runs cool as a cucumber. I don't know why they are continuing with the prescott in any shape or form.
Here ya go.
And, here ya go.
Latency issues can be worked out. And no need to hyperthread a dual core cpu anyway.
Believe or not, the dual core Pentium is based on Prescott , Pentium-M is only to compliment Intel?s desktop line up, not to become the front runner and flagship of the company all you wishful thinkers are pointing out.
Yes Pentium-m would be a better choice, but I read and I will try and find the article which talks about the cons and pro?s of Prescott going dual core, for one thing hyper threading in a Pentium-m wouldn?t work or wouldn?t work to a noticeable degree or anything like Northwood/Prescott, hence you have four virtual CPU?S with a dual core Prescott , as apposed to two for Pentium-m. Yeah the higher IPC of the Pentium-m is attractive but Intel isn?t going dual core with it, at least not yet.
Another thing , software has to be written to take advantage for dual cores, everyone thinks dual cores goona give an extra 100% boost, that?s not the case by far.
Perhaps we?re seeing the limits of silicon, Intel continue to ramp up clock speed, even tho they know there just goona get stuck in the mud even more, nanotubes which we wont see for another 8 years I reckon is the way to go, maybe even 10-15, the industry seems to be in depression. I truly believe higher IPC is the way to go forward, CPU?S especially from Intel are becoming more and more inefficient and making way for clockspeed, Bob Comwell is it who gave a talk at Stamford, a former Intel architect of the P6 design and Pentium 4 was asked about the efficiency of presccott/Pentium 4 and he stated one of the reasons he left was because he wasn?t happy at how the chip was heading and how management didn?t care about performance or simplicity but just sheer clock speeds.
I remember Intel fan boys slating AMD for the athlon 4 (xp) and its lower clock speeds even tho a 1800+ ( 1.5Ghz or there about) creamed a 1.8 and even 2.0 Intel, yeah the argue that AMD didn?t have the clockspeed, true towards the end a 3200+ was?nt quite up a 3.2Ghx P4C, but then again I always take benchmarks with a pinch of salt, Intel has the some muscle and I wouldn?t put it past them to have some influence with bencmarkers. Even tho a 3200+ could match a 3.2Ghz C AMD had the rite idea, if the barton had say frequency of 2.4 instead of the 2.2 ghz it would have been a match for sure, what I?m getting @ is higher IPC and efficiency.
I believe dual core is not the answer to the slow down in scaling and innovations in cpu world.
Yes Pentium-m would be a better choice, but I read and I will try and find the article which talks about the cons and pro?s of Prescott going dual core, for one thing hyper threading in a Pentium-m wouldn?t work or wouldn?t work to a noticeable degree or anything like Northwood/Prescott, hence you have four virtual CPU?S with a dual core Prescott , as apposed to two for Pentium-m. Yeah the higher IPC of the Pentium-m is attractive but Intel isn?t going dual core with it, at least not yet.
Another thing , software has to be written to take advantage for dual cores, everyone thinks dual cores goona give an extra 100% boost, that?s not the case by far.
Perhaps we?re seeing the limits of silicon, Intel continue to ramp up clock speed, even tho they know there just goona get stuck in the mud even more, nanotubes which we wont see for another 8 years I reckon is the way to go, maybe even 10-15, the industry seems to be in depression. I truly believe higher IPC is the way to go forward, CPU?S especially from Intel are becoming more and more inefficient and making way for clockspeed, Bob Comwell is it who gave a talk at Stamford, a former Intel architect of the P6 design and Pentium 4 was asked about the efficiency of presccott/Pentium 4 and he stated one of the reasons he left was because he wasn?t happy at how the chip was heading and how management didn?t care about performance or simplicity but just sheer clock speeds.
I remember Intel fan boys slating AMD for the athlon 4 (xp) and its lower clock speeds even tho a 1800+ ( 1.5Ghz or there about) creamed a 1.8 and even 2.0 Intel, yeah the argue that AMD didn?t have the clockspeed, true towards the end a 3200+ was?nt quite up a 3.2Ghx P4C, but then again I always take benchmarks with a pinch of salt, Intel has the some muscle and I wouldn?t put it past them to have some influence with bencmarkers. Even tho a 3200+ could match a 3.2Ghz C AMD had the rite idea, if the barton had say frequency of 2.4 instead of the 2.2 ghz it would have been a match for sure, what I?m getting @ is higher IPC and efficiency.
I believe dual core is not the answer to the slow down in scaling and innovations in cpu world.
Keysplayr
Elite Member
- Jan 16, 2003
- 21,211
- 50
- 91
Wishful thinkers? I don't know about that. Intel's number one enemy right now is heat. Going to the .09u process had the unexpected trait to generate more heat than less as is normal when going to a smaller process. I will see what I can find out as well.
no i dont mean it in a bad way when i said "wishful thinkers" keysplayr2003, but i cant see dual core doing much, apart from the encoding etc.
lets just hope .65 Micron process isnt a death trap for heat and leakage .
lets just hope .65 Micron process isnt a death trap for heat and leakage .
Dual core prescott is going to burn in the neighborhood of 300 watts. The cooling systems, power supplies, mobo enhancements, needed to make that work will put it at a cost disadvantage. Also huge die size will reduce yields, further making it uncompetitive. Pentium-M is a great core for CMT.
And dual core is the answer, but not dual core prescott. You can put 2 Pentium-M cores on the same area as a Prescott core, and run 2 threads on different cores, instead of running 2 threads on the same core with hyperthreading. A lot cooler and less power to. You can make a dual core pentium-M on same area as single core prescott, burning half the power of single core prescott, with a lot less engineers. Dual core prescott is insanity, they can barely cool a single core one.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter