Intel to Retaliate to AMD Phenom II Overclocking Feat, Plans Demonstration at CES '09

[taurus_sel]

some of those answers here.

I don't see any answers to my questions. All I see is a desperate attempt at trying to respond to the PII news. Come on guys, it's so apparent it's not even funny. Intel is pulling Fugger's strings here. They fly over with a cherry part, they do the work and Fugger is just the formality to try and legitimize the results to make it appear that it didn't come from Intel. They could have just shipped the part to someone else and let them go to it. Why didn't they do that? The same goes for the way Intel just claims Fugger is going to do something for them without even asking first.

Yeah like this is supposed to be convincing to anyone with more brain cells than fingers.

[gosh]

Kyle questioning kassler/gosh views...
http://www.hardforum.com/showpost.ph...5&postcount=39

The post right before Kyle blew the top...
http://www.hardforum.com/showpost.ph...3&postcount=48

Where gosh's avatar came from...
http://www.hardforum.com/showpost.ph...0&postcount=50

Its truly a rare sight to see Kyle do that...

OT
http://www.hardforum.com/showthread.php?t=1364502
Most people there thinks intel is a much better CPU, others have probably left the place

[Donnie27]

So Amd's overclocking stunt had nothing to do with them trying to get people to hold off on buying i7 and wait to see how deneb overclocks? It seems that Amd knows they can't beat Intel clock for clock, so instead they are now pushing the we can overclock better card. Nothing wrong with Intel responding to this challenge is there?

QFT!

Of course it was and I think it was a very smart move too. Hurting the contender on the contenders backyard is probably where it hurts most. Just the reaction from Intel here is proof enough for that. I think that the budget Intel is spending to counter this attack may be rather large. And I think they will use whatever trick they can to make it work for them. They probably need some time for this, and the CES 2009 when deneb should be released may be time enough.
If the processor is able to clock know, then it would be very easy for them to remove the focus from AMD just by inviting some people and show it.

But you're trashing Intel for doing what AMD always does. If one is wrong for doing it, shouldn't the other be wrong as well?

Gallag nailed it when he said; "I don't think that you even know what your point was any more, You just seem angry at Intel."

http://www.hardforum.com/showthread....1069819&page=3

Also banned from [H] but Chris and Kyle are NOT Intel fans LO real L!

[taurus_sel]

No clue, thats what I am wondering.

If I were you I'd be more than a little pissed that I was being used as a puppet.

[Ghostbuster]

C2Q isn't a true quad

C2Q is a true quad... Why? Because it has quad cores.

but you need to know how software works on the processor to understand this. Create two threads on the c2q that shares memory and you probably get slower speed compared to one single thread, much slower. Creating threaded software needs planning on how memory is used on C2Q, this type of planning isn't necessary on phenom and i7 and the reason for this is the L3 cache.

The operating system determines how memory is used and allocated... Example using "malloc()", so how do you know the memory you've just used is going to be in the cache? You can't because it will be loaded into the cache when needed.. This "planning" is really nonsense...

Maybe you will see new drivers soon, and games that used to use only one single thread is suddenly using four threads. If the driver is optimized for a true quad then this is possible, the same driver running on C2Q may be worse than a driver running on one single thread.
True quads are able to split work for much smaller tasks and gain speed.

This is also not as hard as designing the whole application for how the processor work.

There you go again... more nonsense...

Perhaps you missed the news thread about NVIDIA having a quad core optimized drivers?

[qurious63ss]

OT
C2Q isn't a true quad but you need to know how software works on the processor to understand this. Create two threads on the c2q that shares memory and you probably get slower speed compared to one single thread, much slower. Creating threaded software needs planning on how memory is used on C2Q, this type of planning isn't necessary on phenom and i7 and the reason for this is the L3 cache. Maybe you will see new drivers soon, and games that used to use only one single thread is suddenly using four threads. If the driver is optimized for a true quad then this is possible, the same driver running on C2Q may be worse than a driver running on one single thread.
True quads are able to split work for much smaller tasks and gain speed.

This is also not as hard as designing the whole application for how the processor work.

The point I was trying to make is that Amd was saying since K10 was a native chip that it would automatically make it superior to the "bolted on" C2 Quad. Well we all know how that turned out...

[gosh]

The operating system determines how memory is used and allocated... Example using "malloc()", so how do you know the memory you've just used is going to be in the cache?

C2Q = 2 x C2D
What happens if one thread is placed on one C2D and the other thread is placed on the other C2D, the are both reading from the same memory allocated by "malloc()"

[qurious63ss]

If I were you I'd be more than a little pissed that I was being used as a puppet.

This is what credibility is all about. Fugger has earned the respect of alot of people in the overclocking community and if you think he would risk that then you are in the minority here. The great thing about technology is that it involves science so it can always be proven with data.

[Ghostbuster]

C2Q = 2 x C2D

Definition of TRUE Quad Core = have 4 physical cores. Thus, C2Q qualifies.. so did the QuadFather, as does a server board with 4 single core processors.

Not true quad core examples = Pentium XE and Atom 330

What happens if one thread is placed on one C2D and the other thread is placed on the other C2D, the are both reading from the same memory allocated by "malloc()"

Simple, both goes into each shared cache on each side.

[qurious63ss]

C2Q = 2 x C2D
What happens if one thread is placed on one C2D and the other thread is placed on the other C2D, the are both reading from the same memory allocated by "malloc()"

That's the beauty of the "unelegant" design of C2Q, that even with these "drawbacks" non-native, FSB, shared cache, etc. that it was able to outperform an elegantly designed native chip like K10.

[gosh]

Simple, both goes into each shared cache on each side.

And that takes time and increases the burden on the FSB which is the main bottleneck on C2Q. The gain to create that type of threaded software isn't there, performance for smaller tasks will be negative compared to one single thread. Creating threads on C2Q, then you need to reallocate memory or design the software like separate applications with a minimum of shared resources between threads.
If you have one application and want to split up some task that takes time in the application it isn't worth the effort compared to Phenom or i7 that handles this type of work very well.

[neo mike]

I was a amd fan for years when they had intel beat, things changed im now running e8500. I hope amds chips are as good as they seem to be with what is floating around about them, good competition is always good, it gives you choices. I dont see how people can fight over processors , to me its whoever has the best processor at the time of building a computer thats who you go with, i guess some people just have to much time on there hands, and nothing better to do.

[Ghostbuster]

And that takes time and increases the burden on the FSB which is the main bottleneck on C2Q. The gain to create that type of threaded software isn't there, performance for smaller tasks will be negative compared to one single thread. Creating threads on C2Q, then you need to reallocate memory or design the software like separate applications with a minimum of shared resources between threads.

Didn't tests shows that the FSB is hardly saturated with the C2Q? I guess you miss that.. Have you tried the cache to cache test?

If you have one application and want to split up some task that takes time in the application it isn't worth the effort compared to Phenom or i7 that handles this type of work very well.

Yups, maybe you can explain why C2Qs can be still faster despite the "bottleneck" you are hyping about?

[flippin_waffles]

And that takes time and increases the burden on the FSB which is the main bottleneck on C2Q. The gain to create that type of threaded software isn't there, performance for smaller tasks will be negative compared to one single thread. Creating threads on C2Q, then you need to reallocate memory or design the software like separate applications with a minimum of shared resources between threads.
If you have one application and want to split up some task that takes time in the application it isn't worth the effort compared to Phenom or i7 that handles this type of work very well.

And such is the luxury/problem with having >%80 marketshare. You can be sure software writers will tailor make their programs to benefit your architecture. Benchmark programs included.

[gosh]

Yups, maybe you can explain why C2Qs can be still faster despite the "bottleneck" you are hyping about?

check the market share for intel and also check the market share for quad core phenoms. Threaded software is hard, if the gain is negative for more than 90% do you think developers will do it?
Maybe you will see the big shift on how software is created at the end of next year. Libs other programmers are using may then be using threads internally etc.

[cegras]

This is what credibility is all about. Fugger has earned the respect of alot of people in the overclocking community and if you think he would risk that then you are in the minority here. The great thing about technology is that it involves science so it can always be proven with data.

A graph doesn't come from a single data point ...

[Jethro]

I beginning to wonder if wading thru all this regurgitated spew is worth the tiny tidbits of real news you find here anymore...

[coo-coo-clocker]

Interesting discussion.
I think the competition is a marketing stunt. Plain and simple. Take the results as a data point, nothing more. Very little will be proven or disproven using vendor-cherry picked components. The market will determine the winners and losers. The "mere mortals" will rule the day, by using mass-produced components... I think it's great to see the two vendors going at it - lets hope they continue to listen to and court the enthusiast community.

[Ghostbuster]

check the market share for intel and also check the market share for quad core phenoms.

As Kyle would say..

Way to duck the questions...

Anyway... here's your answer (%total units shipped * %total unit market share).

Phenom X4s = 3.52% x 17.9% = 0.63%

Core 2 Quads = 4.19% x 82.1% = 3.44%

These are for desktops.. The final percentage is of total units for 3Q 2008.

Threaded software is hard, if the gain is negative for more than 90% do you think developers will do it?
Maybe you will see the big shift on how software is created at the end of next year. Libs other programmers are using may then be using threads internally etc.

There's nothing hard about creating a multi-threaded software... its just what jobs to run in those threads, how to split a task to each thread, cleanup, inter-thread communications, semaphores, etc... that's rather hard.. Furthermore some tasks can't be multi-threaded. And you don't really need any special libs, just the usual Win32 ones is sufficient..

[Drwho?]

Still not sure what to think about this one, half that news writeup is about me and this demo I have no clue about...

I had just woke up when I first read it and thought AMD was doing some demo at CES.



The scale could could be counted without using your toes.

Fugger and I did meet for 2 days, and we were playing with this fantastic set of processors, over thanks giving day, and the day before.

We had a lot of fun , getting to 5.0Ghz air cool WAS AND IS amazing, now, we did not speak about going to CES VEGAS to do an overclocking demo. The news was fabricated. Clear enough?

And by the way, If I was not a passionated crazy engineer, I would not have done this over thanks giving. I did not have time to do this over my working days, so, i used my day OFF of thanks Giving to get to Vegas and OC with one of the best OC master. If people can not understand that, they have fanboy blindfolding symdrome.

now, I am one of the 2 who did run at 5.0Ghz air cool Core i7!

Can we stop this FanBoy crap and have some fun competiting nicely?
If somebady think AMD is back, get a part, OC it to death, and less run the XtremeSystem set of Benchmarks, and see!

Fugger and I had a lot of fun, beat us if you can

[gosh]

OT
Ghostbuster "Talking" with you is a perfect example on how it is on HardOCP, thats the reason why the AMD section is almost dead there

[mongoled]

Fugger and I had a lot of fun, beat us if you can

Great quote!

Somone in the know get this info to Macci

[Zucker2k]

Fugger and I had a lot of fun, beat us if you can

There is the challenge, maybe somebody get Macci to read that? This what all true enthusiasts should concentrate on instead of bashing the opoosing view left and right.

Let the "parts" do the talking, that's what the TRUE enthusiast wnats to see.

[Zucker2k]

Great quote!

Somone in the know get this info to Macci

Haha, you beat me to it!

Edit: Sigged!

[Dave_Sz]

Source: techpowerup

Am I the only one who thinks his quote "AMD may have disabled several sensors on the cherry-picked chip used in its demonstration, which facilitated that overclock. In response to this, Intel would be disabling the same sensors, in its special demonstration chip." IS funny? He says AMD MAY have disabled some sensors, we don't know which ones, but we're going to disable them on our chip too. Now how can you disable the same sensors when you don't know what they are? So going on assumptions, we all know what happen when you assume, INTEL admitted to using production parts for their test yet they are going make them not production by disabling this and that. So what's the point of saying they are doing the test on hardware we'll be able to get?