Anand re-examines Conroe benchmarks

[MrWizard6600]

BTW, about the Fab argument, I'd love to see if anyone has anything more substantial to go up against visaris and duby other than "AMD don't have the gears for it rollereyes LOL." Visaris linked to his sources and used clear logic, and duby expounded on this, while their opponents made dubious Word analogies and provided no corroboration on their facts whatsoever. I have no idea; as far as I know, either could be right, but I find the discussion very interesting.

aka you enjoy watching people argue you sick bastard

 

[Scali]

I just find it very peculiar that it takes something like Conroe for some people to see just how big and powerful Intel really is (and even then some are still in denial).

The P4 was an amazing processor aswell, far more advanced than an Athlon64.
It may not have performed as well... but think of it like this...
An aircraft has far more technology than a bicycle. But sometimes the bicycle is a better way to get to where you want to go.
Still I'd rather trust the aircraft engineers to design a bicycle than the bicycle engineers to build an aircraft.

The sweet-spot of raw technological power vs agility is somewhere between aircraft and bicycle, and Conroe is Intel's attempt of finding it.
They knew it had to be between those two extremes*somewhere*, so it should come as no surprise that any attempt will work better than these two extremes. But apparently it does, to most.

 

[duby229]

The P4 was an amazing processor aswell, far more advanced than an Athlon64.

As it turns out, that was exactly the problem with the design....

RISC was a better idea for a reason. While not the same thing, a simpler design can be, and most often is the better choice. History people, It has a tendancy to repeat itself. remember the tree and the car?

SMASH!!!!

 

[robberbaron]

I just find it very peculiar that it takes something like Conroe for some people to see just how big and powerful Intel really is (and even then some are still in denial).

The P4 was an amazing processor aswell, far more advanced than an Athlon64.
It may not have performed as well... but think of it like this...
An aircraft has far more technology than a bicycle. But sometimes the bicycle is a better way to get to where you want to go.
Still I'd rather trust the aircraft engineers to design a bicycle than the bicycle engineers to build an aircraft.

The sweet-spot of raw technological power vs agility is somewhere between aircraft and bicycle, and Conroe is Intel's attempt of finding it.
They knew it had to be between those two extremes*somewhere*, so it should come as no surprise that any attempt will work better than these two extremes. But apparently it does, to most.

That anology would work better if you called Intel Submarine engineers (needs nuclear power to run) and AMD speed boat engineers (runs on fossil fuel). Conroe would be some sort of high-speed destroyer.

 

[Scali]

That anology would work better if you called Intel Submarine engineers (needs nuclear power to run) and AMD speed boat engineers (runs on fossil fuel). Conroe would be some sort of high-speed destroyer.

Indeed, Conroe is going to destroy AMD in no time
With the technology that AMD has on its roadmaps currently, they don't stand a chance against Conroe.
And I don't believe in magic, so I don't believe that AMD will just pull a super-competitive CPU out of their hat.

I've always said the Athlon64 isn't such a good CPU, it just looks good next to the P4.
Pentium-M should already have been an indication of how good (or bad) the Athlon64 really. Intel just dusted off an outdated CPU-design, put in some modern technology for mobile use, and came up with a CPU that beat the Athlon64 in power consumption and IPC. And Intel wasn't even trying hard, it was just damage control.

Conroe is Intel actually trying, and succeeding apparently.
I think in 6 months AMD will find itself back where it once was with the K5/K6... Low-budget clone CPUs, only competing with the Celerons, but never with the top range of Intel.
Why would anyone buy an AM2 system? It won't be much of an upgrade, since the CPUs themselves aren't actually faster with DDR2 (sometimes even slower, see the benchmarks at Tomshardware). They may get one or two speedgrades higher than the non-AM2 CPUs, but is that enough for anyone to buy a new motherboard, memory and CPU?
Might aswell go for Conroe if you're upgrading everything anyway.

 

[duby229]

Maybe your right Scali,

But times have changed. I guess we'll just have to wait and see.

 

[robberbaron]

I don't think AMD is where they were with the K5/K6. They are MUCH further along than that.
They have a much larger marketshare now and much more market cap. than in those days. They have the resources to remain a competitive force.

Though, Scali, the fact that you referenced THG's "benchmark" of AM2 as indictitive of AM2's performance is sad. You should know better than taking that broken review at face value. You do know that the RAM was running at DDR2-400, right?

 

[Scali]

I don't think AMD is where they were with the K5/K6. They are MUCH further along than that.
They have a much larger marketshare now and much more market cap. than in those days. They have the resources to remain a competitive force.

Intel isn't where they were either.
Marketshare will go down like crazy when there's no longer a good reason to buy AMD.
They may be able to build a lot of processors, but when these processors can't compete with Intel's top models, there's no alternative but to sell your CPUs as cheaply as possible. So they no longer have the luxury of charging a premium for their CPUs, like they're doing now, especially with their FX series.
Intel used to be the one who had to manufacture and sell their P4s 'cheaply' (look at the die size, number of transistors etc, P4 just costs a lot more to make).

Though, Scali, the fact that you referenced THG's "benchmark" of AM2 as indictitive of AM2's performance is sad. You should know better than taking that broken review at face value. You do know that the RAM was running at DDR2-400, right?

No I don't. Have any proof that the review is broken? The review stated that it used DDR2-667. Even better, do you have proper benchmarks of AM2 then?
This is the only I've found so far.
At any rate, even if you're right, going from DDR2-400 to DDR2-667 won't make enough of a difference to get to Conroe-like performance-levels.
Are you just another AMD-zealot grasping at straws?

 

[robberbaron]

No I don't. Have any proof that the review is broken? The review stated that it used DDR2-667. Even better, do you have proper benchmarks of AM2 then?
This is the only I've found so far.
At any rate, even if you're right, going from DDR2-400 to DDR2-667 won't make enough of a difference to get to Conroe-like performance-levels.
Are you just another AMD-zealot grasping at straws?

The proof is in the review itself. Read the memory bandwidth benchmark at the end. The AM2 system has identical bandwidth to the 939 system.

I'm not grasping at straws, I'm trying to tell you fact in spite of your well-established fan-boyism.
You're the one that grasped at straws by saying "Even though the P4 sucked compared to the Athlon64, it WAS STILL WAY MORE ADVANCED THAN THE ATHLON! So neener neener!"

Direct link to the memory benchmark
http://www.tomshardware.com/2006/02/21/a_look_at_amds_socket_am2_platform/page12.html

 

[duby229]

Even better, do you have proper benchmarks of AM2 then?

Better yet, Do you have proper benchmarks of Conroe?

 

[robberbaron]

Better yet, Do you have proper benchmarks of Conroe?

Totally, the ones hosted by Intel themselves!
Oh wai

 

[duby229]

Totally, the ones hosted by Intel themselves!

Until HardOCP and others get there hands on retail hardware, with retail boards. And build it in a system they configured, and run it on an OS that they installed, and bench agianst there own suite, that they configured, and compare it against other systems that they benched.......

 

[robberbaron]

Until HardOCP and others get there hands on retail hardware, with retail boards. And build it in a system they configured, and run it on an OS that they installed, and bench agianst there own suite, that they configured, and compare it against other systems that they benched.......

Aww, let people have their fun. It's been a long-while since Intel had a significant margin over AMD in terms of performance. It's time fans of Intel, even though their current motherboards are just as compatible with conroe as a socket 939 board, can gloat about the company that made their current CPU making a better CPU.

 

[duby229]

Cool

Fair enough. Will do.

 

[Scali]

The proof is in the review itself. Read the memory bandwidth benchmark at the end. The AM2 system has identical bandwidth to the 939 system.

That doesn't prove in any way that the memory is actually running at the same clockspeeds. It only proves that there is no bandwidth advantage on the DDR2-system. But that could also be a limitation of the cache subsystem, or the memory controller, or whatever else.
If you look at P4 benchmarks you'll see that DDR2-400 is often a lot slower than DDR400 in benchmarks. Which is why I believe that the DDR2 in this benchmark is not running at 400, but at least at 533, possibly 667.
Anyway, if you can't see the technical merits of the P4, I don't expect you to understand memory benchmarks either.

By the way, my current CPU is an Athlon XP 1800+, so you know what company made it.

 

[robberbaron]

AMD's subsystem has always been more efficient bandwidth-wise with the K8's. Just compare it to every other memory subsystem and you can see their design is the best. DDR2 has an inherantly higher latency, making it EASIER to design a memory controller that can cope with high speeds and multiple banks (look at any current GPU, they can access high latency GDDR3 that runs at 2GHz effective.) Even Intel had a raw bandwidth increase from 3-3-3 DDR400 to 4-4-4 DDR533.

But, we'll see when AMD ships a retail product that can be tested and has reproducable results.
Until then, I'm skeptical about any benchmark of the upcoming processors.

And I'm glad that the P4 has technical merits. It's unfortunate that that means nothing to those concerned about what the product actually does for them.
Hah, that's right, the good ol' 1800+. You aren't the only person jumping for joy about the conroe, I'm afraid.

 

[duby229]

That doesn't prove in any way that the memory is actually running at the same clockspeeds. It only proves that there is no bandwidth advantage on the DDR2-system. But that could also be a limitation of the cache subsystem, or the memory controller, or whatever else.
If you look at P4 benchmarks you'll see that DDR2-400 is often a lot slower than DDR400 in benchmarks. Which is why I believe that the DDR2 in this benchmark is not running at 400, but at least at 533, possibly 667.

Rather it shows the merits of an on die memory controller. Higher latency DDR2 performing on par, clock on clock, with DDR....

I cant wait to see what happens with DDR2-800

 

[robberbaron]

Rather it shows the merits of an on die memory controller. Higher latency DDR2 performing on par, clock on clock, with DDR....

I cant wait to see what happens with DDR2-800

There's only one complain I can possibly see about an on-die memory controller, and that's that it requires an extra hop for the video card to talk to the RAM. Despite this TECHNICAL INFERIORITY, the K8's still fare better in games compared to the Netbursts. Fun stuff

 

[duby229]

That's true there are some technical problems with the on die memory controller....

But DAMN it!!! If nvidia would just get off there lazy bums and figure out a way to get UMC (or is it TSMC?)to release a 1nm process, then they'd be able to put the RAM on die using SRAM cells!!!

 

[Scali]

Rather it shows the merits of an on die memory controller. Higher latency DDR2 performing on par, clock on clock, with DDR....

Erm, nobody proved that the memory was running 400 MHz.
Since I still think the memory runs at 667 MHz, to me it just shows that AMD's DDR2 memory subsystem is not as mature as the Intel ones.

 

[Logan321]

Keep in mind that while AMD doesn't have a 65nm process NOW, their fab36 is anticipating producing 65nm chips in Q4 of this year. Even if it takes them 3-6 months to get the process right, that's still a faster response than Intel taking 3 years to respond to the Athlon64.

AMD's first Athlon 64s were on the 130nm process, and then they dropped to 90nm once they had it right with the 130nm. I think it's smarter for 'em to start the AM2 chips at 90nm, and once they've got the bugs worked out, they'll shrink the process.

AMD is moving to DDR2 not because it will benefit much from the bandwidth and latency change, but because DDR2 is [soon to be] more widely mass-produced than DDR, and consumes less power b/c it runs on a lower voltage.

Honestly, it's good for competitors to provide (you guessed it!) competition. Funny that.

 

[Scali]

Even Intel had a raw bandwidth increase from 3-3-3 DDR400 to 4-4-4 DDR533.

Obviously. Then again, the P4 was originally designed for the high bandwidth of RDRAM, so high bandwidth is in its genes. The high clockspeed of CPU, FSB and chipset also helps to take advantage of the bandwidth.
AMD does things differently, so I don't think it's strange if their CPU, retrofitted for DDR2, can't take as much advantage of the bandwidth as the P4 does.

And I'm glad that the P4 has technical merits. It's unfortunate that that means nothing to those concerned about what the product actually does for them.

I never said otherwise. But a lot of people here think that having the most performance means that you also have the most advanced technology. And that is rather naive.

 

[Logan321]

The P4 was an amazing processor aswell, far more advanced than an Athlon64.

Anyone else get an Invader Zim flashback?

"It's not stupid, it's advaaanced!"

 

[Scali]

Keep in mind that while AMD doesn't have a 65nm process NOW, their fab36 is anticipating producing 65nm chips in Q4 of this year. Even if it takes them 3-6 months to get the process right, that's still a faster response than Intel taking 3 years to respond to the Athlon64.

Then again, Intel didn't need to respond, P4 has always greatly outsold Athlons.
And Intel's response is a much bigger and better one than AMD's response.
A new architecture vs a die-shrink and a memory technology that Intel has supported for ages already. Guess who's going to win this round?

Honestly, it's good for competitors to provide (you guessed it!) competition. Funny that.

Yes, and that is not what AMD is going to do anytime soon. The only thing to compete against a new architecture is... a new architecture. We don't know if AMD is ever going to get round to that. Going to 65 nm and DDR2 is just catching up, Intel is already there. In fact, Intel is already well into 45 nm territory.

 

[Jason711]

Yes, and that is not what AMD is going to do anytime soon. The only thing to compete against a new architecture is... a new architecture. We don't know if AMD is ever going to get round to that. Going to 65 nm and DDR2 is just catching up, Intel is already there. In fact, Intel is already well into 45 nm territory.

and that is certainly nothing to be happy about either.. i can see intel setting a huge price premium for these things.

 

[Logan321]

Then again, Intel didn't need to respond, P4 has always greatly outsold Athlons.
And Intel's response is a much bigger and better one than AMD's response.
A new architecture vs a die-shrink and a memory technology that Intel has supported for ages already. Guess who's going to win this round?

P4s outsold Athlons because people wanted Intel... Not because the chip was any good. Maybe you've forgotten the leap that AMD accomplished with Athlon64? It actually was a big accomplishment. Intel's "new" architecture is just a reversal to the architecture they were using before the poor engineering decision to use the super-long pipelines of the P4... and they've had 3 years to improve it. imo for a 20% gain over AMDs available now CPUs it doesn't seem like so big an accomplishment.

Yes, and that is not what AMD is going to do anytime soon. The only thing to compete against a new architecture is... a new architecture. We don't know if AMD is ever going to get round to that. Going to 65 nm and DDR2 is just catching up, Intel is already there. In fact, Intel is already well into 45 nm territory.

AMD isn't "catching up" with ddr2, they probably wouldn't have even bothered if the ram price wasn't on par with ddr, and the higher frequency ddr2 available. I'm sure there'll be a revision in the memory controller to better access ddr2 at some point in the next 6 mo. As for the 65nm, yeah... AMD is "catching up" on that front... but then, they didn't need to go to 65nm before because their current cpus were killing the P4 equivalents in speed and power efficiency. Intel rushed to 65nm to try to put out the fire, and it didn't work the way they'd hoped.

 

[Logan321]

and that is certainly nothing to be happy about either.. i can see intel setting a huge price premium for these things.

Yeah... Like I said before... Tier 1. I doubt there will be any cpus available in the retail channel before christmas... Dell, Compaq, Sony, etc will be getting them before anyone else. So if you're in a rush to get a Conroe, be prepared to buy it bundled with a oem system.

 

[Scali]

P4s outsold Athlons because people wanted Intel... Not because the chip was any good.

I never said it was. But AMD is not in the same luxurious position. People still want Intel, not AMD, and when Conroe comes out, the Athlon64 is no longer 'good'.

imo for a 20% gain over AMDs available now CPUs it doesn't seem like so big an accomplishment.

That is 20% gain of a 2.66 GHz chip over a 2.8 GHz chip (which was overclocked, so not available now either), without even having an onboard memory controller, and with longer pipelines.
At the introductin there will also be 3 GHz models, which will ofcourse be much faster than this 2.66 GHz model. And as the CPU matures, clockspeeds will go up aswell.

AMD isn't "catching up" with ddr2, they probably wouldn't have even bothered if the ram price wasn't on par with ddr, and the higher frequency ddr2 available.

DDR technology hasn't gone anywhere in ages, and isn't going to go anywhere ever again.
DDR2 is, so supporting DDR2 is catching up.

 

[Jason711]

DDR technology hasn't gone anywhere in ages, and isn't going to go anywhere ever again.
DDR2 is, so supporting DDR2 is catching up.

a shame it isnt really worth it..

 

[(cf)Eclipse]

That doesn't prove in any way that the memory is actually running at the same clockspeeds. It only proves that there is no bandwidth advantage on the DDR2-system. But that could also be a limitation of the cache subsystem, or the memory controller, or whatever else.
If you look at P4 benchmarks you'll see that DDR2-400 is often a lot slower than DDR400 in benchmarks. Which is why I believe that the DDR2 in this benchmark is not running at 400, but at least at 533, possibly 667.
Anyway, if you can't see the technical merits of the P4, I don't expect you to understand memory benchmarks either

hehe, scali... you know me. we agree to disagree. however, this is one thing i CANNOT let you slide on. memory is memory, regardless of the color of the cpu. let's consider a few things here.

1. faster memory speed far makes up for the timings. the access operations only make up for part of the latency. as posted earlier in the thread:
200mhz at 2-2-2 = 51.5ns
300mhz at 3-3-3 = 43.8ns

now, we have seen benches with ddr2 showing that this sorta holds true when we switch over from ddr1 to ddr2, and that 400mhz at 4-4-4 will be lower latency than 200mhz at 2-2-2... as well as being much higher bandwidth.

so that brings us to our next point. perhaps amd's ddr2 memory controller simply sucks balls. but if you think about it, why would that be? the elimination of the FSB's inefficiency and latency penalty that aided the K8 so much in the first place.. well, that hasn't changed.
i also find it exceptionally hard to believe that one could do something to a memory controller to change it from being ~90% effeciency to ~60% simply from switching it's interface from ddr1 to ddr2. yes, this is all guesswork, and you can always say that i'm an amd fan boy grasping at straws, but saying that makes you every bit of a fan, if not moreso because you're actually attacking people over something like this. but that aside..

DDR technology hasn't gone anywhere in ages, and isn't going to go anywhere ever again.
DDR2 is, so supporting DDR2 is catching up.

just to nitpick here as always, i'd disagree. ddr2 has long surprassed ddr1. most of the reason we don't really see this yet is due to intel's bandwidth limiting FSB. unfortunately, running too much over 1:1 oversaturates an already overused bus. if you look at benches of ddr2 run at high ratios with the FSB still low and at 1:1 with the FSB jacked way up, the extra bandwidth that the FSB allows through is nothing short of amazing. it's a good thing intel decided to bump up FSB speed some more...

 

[savantu]

Oh really?

Oh really?


http://www.dbanks.demon.co.uk/ueng/minfeat.html
The minimum feature size is the width of the smallest line or gap that appears in your design.

ttp://www.facsnet.org/tools/sci_tech/tech/fundaments/scale2.php3
Chips are generally categorized by their "minimum feature size" -- the dimension of the smallest feature actually constructed in the manufacturing process.

http://en.wikipedia.org/wiki/90_nanometer
The 90 nanometer process refers to the average feature size. However, the minimum feature size on 90 nanometer chips can actually be quite smaller, down to around 45 nanometers.

All posted by visaris earlier, which Duby backed. gfg donnie. I don't think i've ever posted on the amd forums either

You are such an ignorant that it hurts.
Don't you notice that your sources as well as duby's are contradicting each other ?

According to facsnet@idiots , Intel's Prescott is a 50nm chip because its smallest feature size is 50nm.

You came here defending one who claims

1. 65nm tools produce 90nm chips
2. AMD FAB36 has 65nm gear
3. The most important par of a FAB is the building not the tools

I mean , 1 and 2 were shown incorrect by me and empoy , for 2 I even gave a link of Hector Ruiz interview and 3 simply proves that nature is an inexhaustible resource of better idiots.

What duby says is like saying the hangar is important , not the B2 inside.I suppose you agree with him too...

 

[empoy]

face the facts AMD f a n b o y s.
Intel's offering is better than K8 architecture. Enough with the "not being produced yet" stabs. Merom/Conroe/Woodcrest is healthy and will be mass produced and affordable.
If you don't want to believe the initial benchmarks then don't. No one is forcing you. If you still believe X2 is better then please enjoy those pricey processors while running 32bit OS and single threaded games. You can also believe that AMD has all of the 65 "gears" inside Fab36. We just simply don't care.

At the back of your mind, you all know that Conroe will be on top of every benchmark in a few weeks. If you call yourselves "enthusiasts" then please take a look at the facts, the microarchitecture and the initial reports available. Please be open-minded enough to accept that it is now Intel's time to shine again.

 

[savantu]

...

And again, no it IS NOT differant, becouse once again for the ininformed (aka for those that refuse to do a google search) it is the facility that matters. The tools are something that makes the process work, but the facility is what makes the tools work, and is the most important part of fabrication. If the facility doesnt meet the requirements of the tools, then the tools wont work... Get? It is pretty simple actually.... And if you'd stop posting garbage and get out onthe net nad do a little research you'd see that... BUT seeing as how you know it all and nobody else could possibly be right.....

The facility is important , but making huge atmosphere controlled clean rooms is piece of cake compared to the tools inside.

You can't be right even if truth hit you in the head....

FAB building costs are around 30-35% construction and 65-70% equipment inside...Guess which one impacts your finances more and more as future process scale to 45nm , 32nm ...

 

[(cf)Eclipse]

face the facts AMD f a n b o y s.

ok. nice start

Intel's offering is better than K8 architecture. Enough with the "not being produced yet" stabs.

why? we can't turn this into the video card forum where everyone HATES paper launches? yeah, sure, it looks good. wake me up when we get there.

At the back of your mind, you all know that Conroe will be on top of every benchmark in a few weeks. If you call yourselves "enthusiasts" then please take a look at the facts, the microarchitecture and the initial reports available. Please be open-minded enough to accept that it is now Intel's time to shine again.

in performance maybe..
maybe i'm just an exceptionally rare example. i care too much about memory these days. the off-die memory controller + fsb setup that intel still uses from back whenever it was created (damn you EV6 bus) is extremely limiting imo. plus, it doesn't help that right now, despite amd's seeming inability to create a memory controller that can run 4 sticks at 200mhz, 1T.. even when dropped down to 166mhz, 2T, the memory subsystem's performance rivals intel's at 200mhz. to me, that speaks volumes. unless i see memory benches that show intel suddenly making a much better, more flexible, and most importantly, better overclocking, memory controller, i'm staying where i am.

but i'm also probably one of the only ones in the entire world who feels that way, so

 

[Scali]

hehe, scali... you know me. we agree to disagree. however, this is one thing i CANNOT let you slide on. memory is memory, regardless of the color of the cpu. let's consider a few things here.

1. faster memory speed far makes up for the timings. the access operations only make up for part of the latency. as posted earlier in the thread:
200mhz at 2-2-2 = 51.5ns
300mhz at 3-3-3 = 43.8ns

now, we have seen benches with ddr2 showing that this sorta holds true when we switch over from ddr1 to ddr2, and that 400mhz at 4-4-4 will be lower latency than 200mhz at 2-2-2... as well as being much higher bandwidth.

so that brings us to our next point. perhaps amd's ddr2 memory controller simply sucks balls. but if you think about it, why would that be? the elimination of the FSB's inefficiency and latency penalty that aided the K8 so much in the first place.. well, that hasn't changed.
i also find it exceptionally hard to believe that one could do something to a memory controller to change it from being ~90% effeciency to ~60% simply from switching it's interface from ddr1 to ddr2.

However long you write your story, and no matter how many numbers you put in it, you can't PROVE that Tomshardware's benchmark was done with DDR2-400.
So you can't convince me that Tomshardware is pulling a scam. I see no reason why it should be DDR2-400 (as I said, I think DDR2-400 would get worse results than this), and I see no reason why Tomshardware would destroy its credibility by putting out lies like this.

yes, this is all guesswork, and you can always say that i'm an amd fan boy grasping at straws, but saying that makes you every bit of a fan, if not moreso because you're actually attacking people over something like this. but that aside..

Not at all, I'm not a fan of any brand, I'm a fan of technology, and Intel just happens to be where most of the new and interesting technology is found, but that should be no surprise, considering the size and budget of their operation.

just to nitpick here as always, i'd disagree. ddr2 has long surprassed ddr1.

How exactly are you disagreeing here? I said that DDR2 is ahead, and if you are moving from DDR1 to DDR2 support, while your competitor has been supporting DDR2 for a long time already, you're catching up.
Seems to me you're agreeing with what I said.

 

[Jason711]

face the facts AMD f a n b o y s.
Intel's offering is better than K8 architecture. Enough with the "not being produced yet" stabs. Merom/Conroe/Woodcrest is healthy and will be mass produced and affordable.
If you don't want to believe the initial benchmarks then don't. No one is forcing you. If you still believe X2 is better then please enjoy those pricey processors while running 32bit OS and single threaded games. You can also believe that AMD has all of the 65 "gears" inside Fab36. We just simply don't care.

At the back of your mind, you all know that Conroe will be on top of every benchmark in a few weeks. If you call yourselves "enthusiasts" then please take a look at the facts, the microarchitecture and the initial reports available. Please be open-minded enough to accept that it is now Intel's time to shine again.

affordable? links plz.

 

[Scali]

i see memory benches that show intel suddenly making a much better, more flexible, and most importantly, better overclocking, memory controller, i'm staying where i am.

but i'm also probably one of the only ones in the entire world who feels that way, so

Looks like it. Why would you want overclockability? By the looks of it, Intel systems will be faster than even overclocked AMD systems. So the 'much better' part is already covered.
And last time I looked, an on-die memory controller was actually less flexible, since you have to replace motherboard AND CPU in order to move to a different memory technology.

 

[(cf)Eclipse]

However long you write your story, and no matter how many numbers you put in it, you can't PROVE that Tomshardware's benchmark was done with DDR2-400.
So you can't convince me that Tomshardware is pulling a scam. I see no reason why it should be DDR2-400 (as I said, I think DDR2-400 would get worse results than this), and I see no reason why Tomshardware would destroy its credibility by putting out lies like this.

because early samples of the AM2 chips have a "known memory controller bug", however, that bug remains undisclosed. seeing the performance, it seems to be pretty clear to me that this bug is simply the cpu not implementing the right memory ratio internally.

How exactly are you disagreeing here? I said that DDR2 is ahead, and if you are moving from DDR1 to DDR2 support, while your competitor has been supporting DDR2 for a long time already, you're catching up.
Seems to me you're agreeing with what I said.

my bad, i read what you said again, i read it wrong the first time. i do agree

Looks like it. Why would you want overclockability? By the looks of it, Intel systems will be faster than even overclocked AMD systems. So the 'much better' part is already covered.
And last time I looked, an on-die memory controller was actually less flexible, since you have to replace motherboard AND CPU in order to move to a different memory technology.

i don't mean upgrade flexibility. i mean memory tweaking flexibility. i like not having to deal with making sure my FSB is super-high to get the memory efficiency i should be getting. i like to be able to lower the cpu multiplier on a whim if i top out my cpu in trying to get the max ram speed. you get the idea.
i'm also not talking about overall performance or overclocking. specific to the memory sub-system now..

edit: also, chances are you're gonna upgrade the cpu when you get a new mobo... but hey, who knows

 

[Jason711]

and how is that less flexible than an off die controller?

it wasnt made to address flexibilty, it was made to address latency.

 

[duby229]

You are such an ignorant that it hurts.
Don't you notice that your sources as well as duby's are contradicting each other ?

According to facsnet@idiots , Intel's Prescott is a 50nm chip because its smallest feature size is 50nm.

You came here defending one who claims

1. 65nm tools produce 90nm chips
2. AMD FAB36 has 65nm gear
3. The most important par of a FAB is the building not the tools

I mean , 1 and 2 were shown incorrect by me and empoy , for 2 I even gave a link of Hector Ruiz interview and 3 simply proves that nature is an inexhaustible resource of better idiots.

What duby says is like saying the hangar is important , not the B2 inside.I suppose you agree with him too...

ahh jeez, here we go again.....

More name calling? How old are you? 3 maybe?

1> I never mentioned price
2> In order to get the proper equipment, you need tje proper FACILITY!!
3> It's not a simple matter of shutting the fab down for a few months so that the facility can be upgraded to handle the process. First you need to have someplace else to fabricate chips, that is why AMD built fab36
4> Good luck hanging up your coat without that hangar....

edit: air craft hangar I get it.... And actually ummm no... It would be more luck how are you going to build that b2 without the factory......