Not so. SMT and worker/helper are only effective when one application uses lesser block of memory than the other. When each application needs to process large memory blocks at a time(more than the shared cache can afford), you end up having rapid alternating effect between the 2, a "ping pong" scenario as Shariko explained.
Not so. SMT and worker/helper are only effective when one application uses lesser block of memory than the other. When each application needs to process large memory blocks at a time(more than the shared cache can afford), you end up having rapid alternating effect between the 2, a "ping pong" scenario as Shariko explained.
lopri
Elite Member
- Jul 27, 2002
- 13,314
- 690
- 126
Conroe is not here yet. Even if it arrives later this year, it'll take some time for it to mature and for people to adopt. The performance looks amazing, though. In the end that's why Intel showed off the chips that are 6 months away, jeopardizing P4 sales. (And to some extent, A64 sales) Of course Intel can do that because they can afford to do so - their No.1 priority currently is to stop AMD, not to sell more P4s. AMD is still selling out all its chips and it will continue to do so for the rest of year. Conroe effect will start showing in '07, and it's interesting what AMD will do.
One thing I'm curious about is the recent clock speed jump of K8 architecture. Through out its most lifespan, K8 has been around 1.8GHz ~2.4GHz. Talking about MHz, K7s were around those frequencies, too. There were no notable clockspeed increase but effeciency enhance through revisions per generation. How many years? 4 years? (I really don't know but just gessing) In a sharp contrast, today, we're observing dual-core K8s clocking better than single-core K8s. With the introduction of AM2, AMD will officially produce 2.8GHz dual-cores. I'm guessing we'll see 3.0GHz by the end of year - of course a dual-core.
A conspiracy theory: Maybe AMD underestimated Conroe?
AMD might as well have had the ability to produce higher clocking K8 chips much earlier than we thought. But since Intel's been in a deep *hit, they didn't need to release those. Indeed, they probably thought they'd be able to compete with Intel's next generation chips with higher-clocking, power-efficient K8 chips. Then sometime last year they found out the potential of next Intel chips, and became somewhat anxious and felt the need to do something before they lose out in 2006 - here comes socket 939 Opterons (still can't believe that AMD thought it'd sell these to 'server' market when there were no socket 939 server board) and litigations, etc.
Ah.. whatever.. I couldn't sleep and just pulled this out of my ass.
Nothing serious here so just laugh and move on to your technical discussion, please.
One thing I'm curious about is the recent clock speed jump of K8 architecture. Through out its most lifespan, K8 has been around 1.8GHz ~2.4GHz. Talking about MHz, K7s were around those frequencies, too. There were no notable clockspeed increase but effeciency enhance through revisions per generation. How many years? 4 years? (I really don't know but just gessing) In a sharp contrast, today, we're observing dual-core K8s clocking better than single-core K8s. With the introduction of AM2, AMD will officially produce 2.8GHz dual-cores. I'm guessing we'll see 3.0GHz by the end of year - of course a dual-core.
A conspiracy theory: Maybe AMD underestimated Conroe?
AMD might as well have had the ability to produce higher clocking K8 chips much earlier than we thought. But since Intel's been in a deep *hit, they didn't need to release those. Indeed, they probably thought they'd be able to compete with Intel's next generation chips with higher-clocking, power-efficient K8 chips. Then sometime last year they found out the potential of next Intel chips, and became somewhat anxious and felt the need to do something before they lose out in 2006 - here comes socket 939 Opterons (still can't believe that AMD thought it'd sell these to 'server' market when there were no socket 939 server board) and litigations, etc.
Ah.. whatever.. I couldn't sleep and just pulled this out of my ass.
Nothing serious here so just laugh and move on to your technical discussion, please. 
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,787
- 136
pffttt. Another clueless person who expects miracles on AMD's part. If you thought Intel was executing bad, just get your head out of the water, and see their mobile line.
Nope, can't happen since the L2 is physically tagged and the helper thread simply scans through a memory space that the worker thread also owns. When the helper thread walks through memory, it takes control of its lines and flushes the LRU. The worker thread just has to refetch the lines that were flushed, or if the lines were not LRU, no fetch is even necessary. There is no ping-ponging since the two threads are working on the same memory space to begin with.
That link gives no info that contradicts what accord said. The crossbar is still uses snooping to maintain coherency. It uses the slower bus and cannot share data.
Originally posted by: lopri
Conroe is not here yet. Even if it arrives later this year, it'll take some time for it to mature and for people to adopt. The performance looks amazing, though. In the end that's why Intel showed off the chips that are 6 months away, jeopardizing P4 sales. (And to some extent, A64 sales) Of course Intel can do that because they can afford to do so - their No.1 priority currently is to stop AMD, not to sell more P4s. AMD is still selling out all its chips and it will continue to do so for the rest of year. Conroe effect will start showing in '07, and it's interesting what AMD will do.
One thing I'm curious about is the recent clock speed jump of K8 architecture. Through out its most lifespan, K8 has been around 1.8GHz ~2.4GHz. Talking about MHz, K7s were around those frequencies, too. There were no notable clockspeed increase but effeciency enhance through revisions per generation. How many years? 4 years? (I really don't know but just gessing) In a sharp contrast, today, we're observing dual-core K8s clocking better than single-core K8s. With the introduction of AM2, AMD will officially produce 2.8GHz dual-cores. I'm guessing we'll see 3.0GHz by the end of year - of course a dual-core.
A conspiracy theory: Maybe AMD underestimated Conroe?
AMD might as well have had the ability to produce higher clocking K8 chips much earlier than we thought. But since Intel's been in a deep *hit, they didn't need to release those. Indeed, they probably thought they'd be able to compete with Intel's next generation chips with higher-clocking, power-efficient K8 chips. Then sometime last year they found out the potential of next Intel chips, and became somewhat anxious and felt the need to do something before they lose out in 2006 - here comes socket 939 Opterons (still can't believe that AMD thought it'd sell these to 'server' market when there were no socket 939 server board) and litigations, etc.
Ah.. whatever.. I couldn't sleep and just pulled this out of my ass.
There's been s939 server boards available for a while now. Overall, that's an interesting hypothesis nonetheless.
The only clueless is you "Intel dude".
here is another clue for you.
Rambus's stock has been rising sharply in the past 3 weeks (w/ a big spike last week) . You don?t think the insiders in the Streets know something that you don?t know?
How about you come back see me here in couple of months when AM2 is finalized and released and conroe is tested without Intel's tight supervision.
If Intel had any clue, they would have kept Rambus in their camp.
http://img139.imageshack.us/img139/5978/041120061634557rf.jpg
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 54
- 91
Then you should go back a few years yourself to see how badly Intel was torn apart by critics and AMD fans alike for taking the RAMBUS road to begin with. Your one of those, "hindsight is not so 20/20 guys" arent you...
If they had a clue my A$$. You got problems bro. And I think it's just to get a "rise" out of it.
Those dumb as's didnt do what Rambus was screaming at them to do, but instead decided on P4. clueless.
Go read some Intel/rambus history before you speak w/ foot in the mouth.
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 54
- 91
That Intel went with P4 was clueless as well no? Dang! There's that hindsight thing again.
Anyway, you sound like you really don't have a clue about anything except for snippits you read here and there. After all, isn't everybody "armed" with the knowledge of the web? heh. Even with all the knowledge at your fingertips, you still can't pull this con off.
Whatever your agenda, good luck to you sir. I got a feeling you'll be short lived here.
and you see anything wrong w/ info technology?
I see that you have posted 20 a day on the average since you joined AT. I don?t doubt that you are also a member in couple of other forums (at least). considering sleeping hours
I would guess you post 4 on an hr. Anyone with that amount of time spent in forums is considered chartroom junky. I wonder >7000 posts how many of them were any useful info and how many were just wasting our time.
Yap you know. you may be right about me being short lived here. You are an expert in documenting how many people become members, posts, and disappear. You probably have made yourself a very nice looking log book on this acount.
But don?t count on it. On this one, mark your log book as "stayed".
clueless?
I see that you have posted 20 a day on the average since you joined AT. I don?t doubt that you are also a member in couple of other forums (at least). considering sleeping hours
I would guess you post 4 on an hr. Anyone with that amount of time spent in forums is considered chartroom junky. I wonder >7000 posts how many of them were any useful info and how many were just wasting our time.
Yap you know. you may be right about me being short lived here. You are an expert in documenting how many people become members, posts, and disappear. You probably have made yourself a very nice looking log book on this acount.
But don?t count on it. On this one, mark your log book as "stayed".
clueless?
DrMrLordX
Lifer
- Apr 27, 2000
- 22,896
- 12,957
- 136
I was sort of thinking the same thing. AMD's only real argument for switching to AM2 is that memory manufacturers don't want to continue producing DDR. Intel has succeeded in moving the PC memory industry over to DDR2.
woot. Okay, I'll ask you since everyone else is dodging this question: Are you remotely interested in the possibility of a $40 X2-3800+ that is a 35W processor? Is this thing going to be the next Athlon XP-M or what?
DrMrLordX
Lifer
- Apr 27, 2000
- 22,896
- 12,957
- 136
What the hell do you think RAMBUS, Inc wanted Intel to do in lieu of pushing the Netburst architecture? The whole point of Netburst was to produce a CPU that actually needed RDRAM's high memory bandwidth without being bothered by its higher latency. The Pentium III couldn't really make RDRAM shine, so Intel needed some justification for adopting RDRAM. Netburst was the solution to this problem.
Up until the Granite Bay platform, RDRAM was the memory of choice for Pentium 4s. Duh.
The X2 3800+ 35W is still going to be a desktop cpu as opposed to XP-M which is a mobile. It'll be the same as a standard X2 3800+ but binned for 35W. So I highly doubt it'll be a good overclocker - at least to begin with (AMD do like to constantly refine their 90nm process).
Of course if it's only $40 then it could still be tempting.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,896
- 12,957
- 136
The reason why I likened the 35W 3800+ to the Athlon XP-M is that the XP-M proved to be an outstanding overclocker when used on desktop boards. The XP-M was a good overclocked because, not despite, it was binned for low power consumption. All you had to do was push it up to the same voltage levels of a desktop XP processor and run the clockspeed through the roof. 2.5 ghz was pretty common for those chips.
That's why the 35W 3800+ is, to me, the chip to watch on AM2. It may overclock to high levels with minimal overvolting.
I think the main reason for switching to AM2 will be the K8L upgrade, since it's rumured to be in the AM2 socket.
No one expected major gains from AM2. It's simply a new socket which uses pretty much the same processors with a redesigned Memory controller to handle DDR2. We also knew that the current K8 was never bottle necked when it came to memory bandwidth and thus the little performace gain from DDR2. AMD is simply introducing a new socket. Real test will come when they release 65nm processors and ACTUALLY use the DDR2 bandwidth.
Until then, it's pointless to upgrade to AM2.
I really could care less...i have a system that will treat me well for the next year or two and until then, we'll see who has the best bang for the buck. I look at this is a NEW BEGINING for both Intel and AMD, let's see who shines through.
For the past three years, its been a clock race...nothing really now and new times are changing
Until then, it's pointless to upgrade to AM2.
I really could care less...i have a system that will treat me well for the next year or two and until then, we'll see who has the best bang for the buck. I look at this is a NEW BEGINING for both Intel and AMD, let's see who shines through.
For the past three years, its been a clock race...nothing really now and new times are changing
Re-read or go read the threads created just after anandtech's preview of Conroe. amd defenders mostly said wait for am2 before making any judgments, that we have to see what am2 has to offer.
Most level headed posters new that AM2 was not going to provide any breakthroughs in performance.
zephyrprime
Diamond Member
- Feb 18, 2001
- 7,512
- 2
- 81
It's not just intel's doing. With newer process technology, lower voltages are a natural consequence and ddr2 is lower voltage.
compgeek89
Golden Member
- Dec 11, 2004
- 1,860
- 0
- 76
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: 22K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 372
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter