GuitarDaddy
Lifer
- Nov 9, 2004
- 11,465
- 1
- 0
Good to know, not much of multi proc guy
just learnin. I knew NT2003 and even older NT server OS's supported multi-proc. Just didn't know XP did
just learnin. I knew NT2003 and even older NT server OS's supported multi-proc. Just didn't know XP did
Capt Caveman
Lifer
- Jan 30, 2005
- 34,543
- 651
- 126
A Pentium D sighting. Possibly this computer show on May 31st in Japan is why Intel and AMD have been holding off their releases.
http://www.pcworld.com/news/article/0,aid,120974,00.asp
http://www.pcworld.com/news/article/0,aid,120974,00.asp
I like this comment....
In volume??? LOL!! At 1000 bucks plus a pop it isn't even supposed to sell in volume...At a whopping 3.2ghz it is far beloew the falgship of the iNtel line so that limits the market to enthusiasts who are fanatical about Intel to ignore its performance versus AMD, and like to waste even more money then an EE single core chip.....
Ok. One of the reasons I saw the Pentium 840 EE not peforming as well is because some programs are still using the virtual core of the first processor and not the actual second "real" core when they are two processor (Hyperthreaded enabled). In other words the X2's are using the two actual cores whereas the 840EE chips is using only the one core and its virtualcore, thus a loss in perfomance. Since AMD enabled hyperthreading on their chip the multi-threaded (or hyperthreaded) programs see the second phsyical core as a virtual core and you get two full speed cores. Intel's doesn't. With patches I would imagine scores for the 840EE will go up.
ermm...what?
AMD has Hyperthreading? What are hyperthreaded applications? Will there be patches to reduce power consumption and heat?
AMD has Hyperthreading? What are hyperthreaded applications? Will there be patches to reduce power consumption and heat?
AMD doesn't have hyperthreading per core per se but enabled hyperthreaded applications to recognize the dualcore as a hyperthreaded processor allowing full use of the second real core.
There won't be patches to reduce power consumption and heat on either processor but since the X2 uses less power at full usage than the PIV at idle, AMD isn't worried about that.
The patches are for those programs that aren't recognizing the second actual "real core" of the Intel dualcore EE and are instead still using the virtualcore of the first processor when hyperthreading is enabled thus a worse score than two "real" cores would produce.
Hyperthreaded application sees the two "cores" of the AMD (two real cores)
Hyperthreaded application sees one "real core" and one virtual "core of the Intel 840EE and ignores the second real processor and second virtual processor.
Thus the problem, Intel is not getting the same performance boost from their 840EE as they should. I can't remember which program that this was shown in but would explain the worse performance they are getting from their 840EE then their 840D because the multi-threaded application on the non hyperthreaded Intel 840D dualcore is seeing both "real cores". Of course this won't help the domination on all fronts that AMD has over Intel in dualcore (except price and even then isn't as bad depending on what you currently have and price/performance ratios).
Edit: per question above. Hyperthreading is a trick that allows programs to see two processors instead of one. A single core processor will thus have one "real" processor and one "virtual" processor. The reason they did this (Intel) was because their CPU design had many pipelines and fast frequencies in mind. However, as the instruction was running through the long pipeline the cpu was twiddling its thumbs behind it so hyperthreading is born. Recognizing "2" processors a hyperthreaded optimized (or multi-threaded application) will send another instruction behind the first one before the first one is done processing. Thus the cpu pipes are used more efficiently. AMD doesn't use hyperthreading because they have much fewer pipelines than the Northwood/Prescott. AMD64 has around 12, Northwood around 21, and Prescott 31. I can't remember exactly but that is close enough.
This is, of course, a non-engineers, non-cpu laymen attempting to explain a complicated process. If this is wrong I apologize to the real engineers and knowledgeable people on the forum.
There won't be patches to reduce power consumption and heat on either processor but since the X2 uses less power at full usage than the PIV at idle, AMD isn't worried about that.
The patches are for those programs that aren't recognizing the second actual "real core" of the Intel dualcore EE and are instead still using the virtualcore of the first processor when hyperthreading is enabled thus a worse score than two "real" cores would produce.
Hyperthreaded application sees the two "cores" of the AMD (two real cores)
Hyperthreaded application sees one "real core" and one virtual "core of the Intel 840EE and ignores the second real processor and second virtual processor.
Thus the problem, Intel is not getting the same performance boost from their 840EE as they should. I can't remember which program that this was shown in but would explain the worse performance they are getting from their 840EE then their 840D because the multi-threaded application on the non hyperthreaded Intel 840D dualcore is seeing both "real cores". Of course this won't help the domination on all fronts that AMD has over Intel in dualcore (except price and even then isn't as bad depending on what you currently have and price/performance ratios).
Edit: per question above. Hyperthreading is a trick that allows programs to see two processors instead of one. A single core processor will thus have one "real" processor and one "virtual" processor. The reason they did this (Intel) was because their CPU design had many pipelines and fast frequencies in mind. However, as the instruction was running through the long pipeline the cpu was twiddling its thumbs behind it so hyperthreading is born. Recognizing "2" processors a hyperthreaded optimized (or multi-threaded application) will send another instruction behind the first one before the first one is done processing. Thus the cpu pipes are used more efficiently. AMD doesn't use hyperthreading because they have much fewer pipelines than the Northwood/Prescott. AMD64 has around 12, Northwood around 21, and Prescott 31. I can't remember exactly but that is close enough.
This is, of course, a non-engineers, non-cpu laymen attempting to explain a complicated process. If this is wrong I apologize to the real engineers and knowledgeable people on the forum.
Thanks for the time and effort duive, nice to see the comparison of that kind. I am really excited about the X2, can?t see much happening except in terms of frequency scaling and cache juggling from AMD?s part. And we?ve already seen the frequency potential of some of the early X2?s. Soon as 65nm is full swing I?m getting one ( Yes Zebo for sure), hopefully the new process and the 300mm wafers they will be suing will be enough to lower costs. Even now I still think it?s a bargain for what your getting. I don?t have brand loyalty towards AMD, well maybe I do, but until Intel can offer a considerably faster, cooler running, better price/performance CPU which is all rounded (rules Dothan out) then I?m hector?s bitch.
I really wouldn?t be surprised if Intel?s next dual core has an on die memory controller, we?ve already had confirmation that they will being developing a point to point bus similar to hyper transport, so the gap will be closed or they might overtake AMD in 2007, the FSB and cache layout in Smithfield is just a joke, almost a kick in the balls to AMD ? Hey we?ll release this slob of a cpu and it?ll sell ten times better then your product will, and then to shaft you we?ll take your ideas?. Well AMD wernt the first with an on die memory controller so ..
Can we guess what?s going to happen through 2006 with the X2 ? K10 (Yes, the dual core K8?s are called K9?s, and the original K9 has been called the K10) wont be out till earliest in 2007, so I guess there just be frequency increases and cache cuts/add on?s along the way.
I really wouldn?t be surprised if Intel?s next dual core has an on die memory controller, we?ve already had confirmation that they will being developing a point to point bus similar to hyper transport, so the gap will be closed or they might overtake AMD in 2007, the FSB and cache layout in Smithfield is just a joke, almost a kick in the balls to AMD ? Hey we?ll release this slob of a cpu and it?ll sell ten times better then your product will, and then to shaft you we?ll take your ideas?. Well AMD wernt the first with an on die memory controller so ..
Can we guess what?s going to happen through 2006 with the X2 ? K10 (Yes, the dual core K8?s are called K9?s, and the original K9 has been called the K10) wont be out till earliest in 2007, so I guess there just be frequency increases and cache cuts/add on?s along the way.
This could explain it...IN apps like CAD the EE was scene in task manager as using all 4 cores and in some apps it was only using 2 efficiently and the other 2 (not in any order) at less then full....
I think there is potentail as mentioned some apps are more optimized for dual threaded and not for unlimited cores..others like CAD seemed to use as much as you could give them...
The other reason that the X2 gains is because it is not bandwidth bottlenecked. The PD's are overgrown Prescotts stacked next to each other. Remember how the Northwood got so much of a boost by going from 533 FSB to 800 FSB? PD needs so much bandwidth that 4 cores can't possibly function at maximum efficiency because they are waiting for memory to serve up data. As far as any of the X2 vs. traditional (non-XE) PD comparisons, the gain is because the A64 is simply that much better and the PD's needed to take a clockspeed hit to avoid becoming an expensive toaster.
Oh yeah, great idea.. and have it burn what? 400+ Watts 24/7 ?!?! I don't think so!
The area X2 owns the P-D in the most is power consumption, it's just amazing how the AMD dual core falls in between the single core 90nm and 130nm cores.
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 + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter