Warming up to the new iMac (or my computer's cooler than your computer).

[Akka]

Well, you've finished to amuse me, now you're becoming tiresome...

Originally posted by Asher
You didn't destroy a single thing. In fact, you've astonished me with your inaccuracies and I'm beginning to wonder if someone could really think that or if you just enjoy trolling.

What's funny is that from my point of view, YOU are the one full with innaccuracies and, more importantly, biased conclusions.

You can rest assured that by the time IA-64 is aimed to the desktop (2004-2005), it'll have excellent 32-bit performance in x86.

2004-2005 is no soon. Don't see the point to argue about proc that won't be here before two years, as nobody can even imagine the kind of performance they will have.

I'm not sure you're aware about the hammer.
Hammer is a slightly modified Athlon, which in turn is a modified P6. It's not an all-new 64-bit core, it's a kludged 32-bit core using a couple decades-old instruction set with many known shortcomings. It's a cheap design that AMD could afford to do, and it looks great on paper but people are already whining about keeping x86 alive...

I'm perfectly aware about the Hammer, and what I'm seeing is that it will be available before the end of the year and that you can argue 'till your tongue become blue that it's a cheap design, it still seems to have a hell of a horsepower in it, and it CAN runs 64 bits programs at least as fast as 32 bits ones.

MHz is not the only determinant of "strength" in a processor. You could argue that the G4 is strong in "brute strength" because it's a "fat" processor that does a lot per clock.
"brute strength" is the IBM Power4. Huge, hot, but very powerful.

Gaah... That is a semantic arguing. I define the "raw strenght" by the most simple and less elegant way to boost the power of a processor : frequency. A processor that does more per clock is more elegant, as its design allow it to be more efficient rather than simply relying of more MHz.

Hey, no kidding?
Most PC2100 can't, actually. Most PC2100 that can is now packaged and sold as PC2700.

We probably don't buy our RAM at the same place then.

DDR333 and DDR400 show about a 5-10% drop, but if you're so paranoid about pricing...
Going to an Athlon also drops performance, what's your point?

Again insincerity.
YOU were the one arguing that a P4 runs perfectly well with DDR333.
I just reminded you that using DDR with a P4 is shooting the performance.
And I'm the one that is full of innaccuracies ?

The i850 is exactly identical to the i850E, actually. The only difference is the i850E has a sticker on the box that says it can go at a 133MHz system bus (effectively 533MHz). I'm not joking, the steppings and everything are exactly identical.

So much insincerity AGAIN.
YOU were the one insinsting on the comparison of what both platform has best in chipset. When I compared the i850 to KT266 you said that it was an old and obsolete chipset.
Now you say that i850 can be used as i850E because it can go to 133 MHz too. Well, my own motherboard can go up to 166 MHz, and it's not even a KT266, it's just a KT133. Either we're talking about the OFFICIAL performances, and then you have to stick to i850E and not i850, either we're talking about the UNOFFICIAL performances, and then I can select the KT266 and not the KT333. Both case Athlon's chipset is cheaper, bot cases you're wrong.

You aren't even replying to the topic anymore, just like UR does when he starts to falter. It's pathetic, really.

What use is to stay on-topic with you ? You show such a great amount of insincerity and stubborness that it's like talking with a wall. I tried again this time, and I'm pretty sure that you'll still twist facts just to "prove" your ways.
Moreover, you're aggressive like a unsecure teen that need to show to the world that he's trying to becoming a man.
You should ask yourself if people not staying on-topic with you is not a problem coming from you rather than they.

[Akka]

Originally posted by Sava
Macs suck...

Here we agree

[Asher]

Originally posted by Akka le Vil
2004-2005 is no soon. Don't see the point to argue about proc that won't be here before two years, as nobody can even imagine the kind of performance they will have.

IA-64 was never meant to be on the desktop soon. It's intended to mature in high end/scientific workstations and then brought to the desktop when there's real demand for it. There is absolutely no need for 64-bit desktops right now. None. I don't know if you believe in the myth that 64-bit == more speed, but that's not correct. 64-bit == more precision, and all it means here is essentially integer precision. We've already got 80-bit floating point precision.

I'm perfectly aware about the Hammer, and what I'm seeing is that it will be available before the end of the year and that you can argue 'till your tongue become blue that it's a cheap design, it still seems to have a hell of a horsepower in it, and it CAN runs 64 bits programs at least as fast as 32 bits ones.

Yes, and I never said otherwise.
That doesn't change the fact that it's a quick kludge to a 32-bit CPU to have 64-bit registers and instructions.

Gaah... That is a semantic arguing. I define the "raw strenght" by the most simple and less elegant way to boost the power of a processor : frequency. A processor that does more per clock is more elegant, as its design allow it to be more efficient rather than simply relying of more MHz.

You're right, it is semantic arguing. Why did you bring it up?
There IS no "strength" in CPU, strength is determined by IPC * MHz if anything. So when you basically claim the Athlon is more elegant (which it most certainly is NOT, it's a collection of kludges already!) just because it does more per clock, then by your logic the 386 is the most elegant chip ever.

We probably don't buy our RAM at the same place then.

You don't buy from Samsung, Corsair, or Kingston? Because all of those put their high-yield PC2100 parts (ie: parts that will hit PC2700) and sell them as PC2700. It's called speed-binning. Prior to PC2700 this wasn't a problem, but now that PC2700 is out you'll find less PC2100 parts that will hit PC2700.

Again insincerity.
YOU were the one arguing that a P4 runs perfectly well with DDR333.
I just reminded you that using DDR with a P4 is shooting the performance.
And I'm the one that is full of innaccuracies ?

A P4 with DDR333 *DOES* run Perfectly Well. That is in NO WAY inaccurate.
Look up some benchmarks. Like these, for example: http://www.tomshardware.com/mainboar...s645dx-06.html

You are full of inaccuracies, and this is a shining example.

So much insincerity AGAIN.
YOU were the one insinsting on the comparison of what both platform has best in chipset. When I compared the i850 to KT266 you said that it was an old and obsolete chipset.
Now you say that i850 can be used as i850E because it can go to 133 MHz too. Well, my own motherboard can go up to 166 MHz, and it's not even a KT266, it's just a KT133. Either we're talking about the OFFICIAL performances, and then you have to stick to i850E and not i850, either we're talking about the UNOFFICIAL performances, and then I can select the KT266 and not the KT333. Both case Athlon's chipset is cheaper, bot cases you're wrong.

KT333 is more than KT266 with a 166MHz memory bus. In fact, it has an entire new northbridge. The i850 and i850E are EXACTLY the same, except motherboard makers can now officially sell them as 533MHz parts.

What use is to stay on-topic with you ? You show such a great amount of insincerity and stubborness that it's like talking with a wall. I tried again this time, and I'm pretty sure that you'll still twist facts just to "prove" your ways.
Moreover, you're aggressive like a unsecure teen that need to show to the world that he's trying to becoming a man.
You should ask yourself if people not staying on-topic with you is not a problem coming from you rather than they.

Look, I don't know where you get your information, but your information is usually wrong. Add to this that you're setting up strawmen and twisting words around and constantly calling me "insincere" (which isn't even RELEVANT to any of this).

You've no clue what you're talking about here, you're just bickering with everything I say (no matter how indisputably right it is) because you don't want to lose.

[monolith94]

Macs aren't for everyone, but that hardly means that they suck...

[Imran Siddiqui]

Originally posted by monolith94
Macs aren't for everyone, but that hardly means that they suck...

Of course it does .

[Akka]

Well, this does REALLY bore me now. We were talking about performances issues and comparison between AMD & Intel at first, and this has gotten far away in other direction. Let's recenter it.

Originally posted by Asher
IA-64 was never meant to be on the desktop soon. It's intended to mature in high end/scientific workstations and then brought to the desktop when there's real demand for it. There is absolutely no need for 64-bit desktops right now. None. I don't know if you believe in the myth that 64-bit == more speed, but that's not correct. 64-bit == more precision, and all it means here is essentially integer precision. We've already got 80-bit floating point precision.

Again : we'll see when it will happen. No need to talk in the air.

Yes, and I never said otherwise.
That doesn't change the fact that it's a quick kludge to a 32-bit CPU to have 64-bit registers and instructions.

We'll see how it fare against the Itanium. Personnally, I don't very much have a doubt about the outcome, which it would be in the 32 or the 64 bits area.

You're right, it is semantic arguing. Why did you bring it up?
There IS no "strength" in CPU, strength is determined by IPC * MHz if anything. So when you basically claim the Athlon is more elegant (which it most certainly is NOT, it's a collection of kludges already!) just because it does more per clock, then by your logic the 386 is the most elegant chip ever.

A 486DX33 was certainly not less powerful than a 386DX33. A Pentium 100 was nearly twice as powerful than a 486DX4 100. A Pentium II 233 was at least as powerful as a Pentium 233. An Athlon beats any Pentium on a clock-by-clock basis. I then highly doubt that a 386 is able to do more per clock than an Athlon.

You don't buy from Samsung, Corsair, or Kingston? Because all of those put their high-yield PC2100 parts (ie: parts that will hit PC2700) and sell them as PC2700. It's called speed-binning. Prior to PC2700 this wasn't a problem, but now that PC2700 is out you'll find less PC2100 parts that will hit PC2700.

I buy from my own seller, and I just ask him for RAM that can be overclocked

A P4 with DDR333 *DOES* run Perfectly Well. That is in NO WAY inaccurate.
Look up some benchmarks. Like these, for example: http://www.tomshardware.com/mainboar...s645dx-06.html

You are full of inaccuracies, and this is a shining example.

You are full of insincerity, and this is a shining example. We were talking about performance, right ? And performances-speaking, putting DDR on a P4 is suicide. I can put some SDRAM PC66 on an Athlon, it will work too. But I won't make any performance comparison with another chip in this situation.

KT333 is more than KT266 with a 166MHz memory bus. In fact, it has an entire new northbridge. The i850 and i850E are EXACTLY the same, except motherboard makers can now officially sell them as 533MHz parts.

Well, right. Point taken.

Look, I don't know where you get your information, but your information is usually wrong. Add to this that you're setting up strawmen and twisting words around and constantly calling me "insincere" (which isn't even RELEVANT to any of this).

You've no clue what you're talking about here, you're just bickering with everything I say (no matter how indisputably right it is) because you don't want to lose.

Must be a question of point of view, 'cause what you see in me, I see it in you. I use "insincerity" is your habit to twist the facts to make them accord to your views, which is well relevant to the argument.

Now, it's becoming tedious. If you learn to discuss without ending in insults (not that I was not a little bit provocative, I admit ) and if you start to back up with NON-BIASED data, I will feel commited to answer again. But if you still keep the same stubborn way, well, you're on your own.

[Akka]

Originally posted by monolith94
Macs aren't for everyone, but that hardly means that they suck...

Well, seriously, Mac have very little for them.

Points for the Mac are the simplicity of use and everything that is related to the image industry. And design for the ones who got bad tastes

But else, I can't see any reason that could be incentive to buy a Mac : they are more expensive and less powerful than a PC, harder to upgrade, run ten times less programs and so on...

[Asher]

Originally posted by Akka le Vil
A Pentium II 233 was at least as powerful as a Pentium 233.

That's not true, especially in 16-bit code.

An Athlon beats any Pentium on a clock-by-clock basis. I then highly doubt that a 386 is able to do more per clock than an Athlon.

Actually, there's always been the trend of doing less per clock with each new core design. The 386's was something like 3-4 stages. Pentium III/Athlon's 10, Pentium 4's 20, and Hammer is supposed to be 12. More stages mean less gets done per clock, but the clock speed increases also.

You are full of insincerity, and this is a shining example. We were talking about performance, right ? And performances-speaking, putting DDR on a P4 is suicide. I can put some SDRAM PC66 on an Athlon, it will work too. But I won't make any performance comparison with another chip in this situation.

What the hell?
Did you even LOOK at the benchmarks I conveniently linked for you?! DDR333 is on par with PC800 RDRAM, sometimes even outpacing it! Read the benchmarks for yourself, I linked them right there for you.

No wonder we're having trouble here, you fail to acknowledge basic points even when confronted with evidence...

[Urban Ranger]

Originally posted by Asher
Nonsense, absolute nonsense.
Adding x86-64 support is trivial.

Merely upgrading all the x86 instruction set to work with 64-bit operands is a major undertaking, considering all the registers ( whether you see them on the outside or not), ALUs, floating point units, the various queues, etc. have to be doubled in size. Then all the microcodes have to be reworked to allow for 64-bit operations.

Originally posted by Asher
It adds about 2-3%% to diespace

Not even close.

Originally posted by Asher
and only certain parts of the processor actually need to be upgraded (because of SSE2, much of the parts already work at 64-bit precision).

Even when that's true, there is still significant work to be done.

Originally posted by Asher
You incorrectly assume it's hard to add x86-64. You don't think Intel could add x86-64 in the 1.5-2 years they've been working on the chip? Give me a break...

If they have something waiting in the wings, or even a project that is making significant headways, we'd be gettin some clear and significant details on the chip. We have been getting a lot of details on the Itanic and it's not even done.

I don't think it is Intel's way to surprise the market with a finished product.

Originally posted by Asher
Sandbag-esque desklamps (I don't care what monolith says) with about a 15 degree swivell radius on the arm? It blows Apple's designs away.

You do have some, um, interesting views on what good designs are.

[Asher]

Originally posted by Urban Ranger
Merely upgrading all the x86 instruction set to work with 64-bit operands is a major undertaking, considering all the registers ( whether you see them on the outside or not), ALUs, floating point units, the various queues, etc. have to be doubled in size. Then all the microcodes have to be reworked to allow for 64-bit operations.

Not all x86 instructions are being reworked. Only certain instructions need to be added for integer use. FPU is already at 80-bit (64-bit in SSE2).

Why would the FPUs need to be doubled in size if they're already working at 80-bit?
64-bit just changes the memory addressing space and the integer precision.

Not even close.

Well, ClawHammer has an integrated memory controller, SSE2, as well as x86-64. The 0.09 micron Clawhammer is going to be 64mm², while the 0.09 micron Athlon (Thoroughbred) is 50mm². So how is that not even close?

It appears you're thinking x86-64 is far more complicated than it actually is.

If they have something waiting in the wings, or even a project that is making significant headways, we'd be gettin some clear and significant details on the chip. We have been getting a lot of details on the Itanic and it's not even done.

I don't think it is Intel's way to surprise the market with a finished product.

It's all about politics. Intel doesn't want to admit to having an x86-64 chip until it sees if IA-64 is already dead in the water for desktop or not. There is zero incentive for them to announce it officially with details. It makes perfect sense for them to build Prescott with x86-64 and then just enable them if needed.

[Urban Ranger]

Originally posted by Asher
IA-64 is far, far more than a simple 64-bit processor. It is radically different than x86-64.

The thing is it is too radical. It relies on ubercompilers to perform all the magic, and they are nowhere to be seen. Not even Intel has one that's close. Of all people, they should be the one that should be making tools that make their hardware actually shine.

Originally posted by Asher
x86-64 is s cheap kludge to add 64-bit registers to x86 and corresponding instructions. That is insanely easy to do.

Not so. See my previous post.

Originally posted by Asher
The reason why Intel didn't want to go there is because most developers want x86 dead.

Most developers don't actually care because they don't need to actually deal with the hardware. System programmers may hate it, but they don't even form the majority.

More importantly, there's too much capital invested in i86 for such a clean break. Even Apple had a hard time shifting from Motorola 680x0 to PowerPC, even though they had a much smaller market and they had control over both hardware and system software.

Originally posted by Asher
It's old, it's obsolete, it's kludgey, it's got a stack-based FPU which is simply horrible to work with, it spends about 30% of its clocktime overcoming inherent shortcomings of x86.

Again, all this depends on how good the compiler is. If the best of the compilers can't take advantage of the hardware features - there are advantages to stack machines - there is no way that any compiler is going to start scratching the VLIW stuff.

Originally posted by Asher
IA-64 was a radically new instruction set and architecture based upon 20 years of University research.

Here's an interesting article for you

Also consider that HP, the original developer of IA-64, dropped the project several years ago. This is a clear indication that HP didn't think the project is viable.

Originally posted by Asher
Saying that AMD was able to build a 64-bit processor in far less time than Intel is an absolute joke, you're implying that AMD is more skilled than Intel.

Maybe not more skilled, just more sensible

Originally posted by Asher
It's like this: If you take someone else's 30-page essay and add on 1 paragraph to it and change some spelling, you can do it far faster than the kid who had to research, think about, and then write his own 30-page essay from scratch.

That's a bad analogy. Adding 64-bit extensions is a significant undertaking.

Originally posted by Asher
The 2.53GHz is the undisputed speed king.

Speed <> performance. The P4 has a pathetically low IPC.

Originally posted by Asher
That doesn't matter, AMD intentionally made the PR ratings conservative so people like you can say "Hey! Match at the PR ratings the Athlon is faster! GO AMD! WOO!".

If they can afford to underrate their chips, more power to them, I say

Originally posted by Asher
The 2.53GHz P4, even with PC800 RDRAM (Performance hits 5-10% higher with PC1066), is the undisputed speed king.

Note that DDR400 blows PC1066 away.

[Urban Ranger]

Originally posted by Asher
Well, ClawHammer has an integrated memory controller, SSE2, as well as x86-64. The 0.09 micron Clawhammer is going to be 64mm², while the 0.09 micron Athlon (Thoroughbred) is 50mm². So how is that not even close?

Hello! Can you do the arithmatic? The increase is 14mm²/50mm². which comes out to be 28%. That's several magnitudes more than 3%.

Originally posted by Asher
It appears you're thinking x86-64 is far more complicated than it actually is.

Looks like simple arithmatic bears me out

Originally posted by Asher
It's all about politics. Intel doesn't want to admit to having an x86-64 chip until it sees if IA-64 is already dead in the water for desktop or not. There is zero incentive for them to announce it officially with details. It makes perfect sense for them to build Prescott with x86-64 and then just enable them if needed.

The incentive is to show that they actually have stuff instead of just vapour. I don't see why Intel would want to divert resources to a skunkworks type project unless they see IA-64 is far from being ready and needs another CPU to slow down the AMD onslaught.

Not only the Claw Hammer will soon be ready, but the Athlon will get a new core even sooner, and the decrease in die size will give a boost to clockspeed even if everything else stays the same.

So yeah, Intel is having problems, and they aren't above FUD.

[Asher]

Originally posted by Urban Ranger
The thing is it is too radical. It relies on ubercompilers to perform all the magic, and they are nowhere to be seen. Not even Intel has one that's close. Of all people, they should be the one that should be making tools that make their hardware actually shine.

That's exactly right -- it takes years to get good compilers for an instruction set. That's why Itanium was in extremely limited release, and McKinley will also. Itanium 3 will be the big one.

More importantly, there's too much capital invested in i86 for such a clean break. Even Apple had a hard time shifting from Motorola 680x0 to PowerPC, even though they had a much smaller market and they had control over both hardware and system software.

Intel just needs to invest in a larger x86-64 emulation unit in the IA-64. Hell, by 2004-2005 when they bring it to the desktop they could imbed an IA-32 chip inside (they should be around 30mm^2 by then )

Also consider that HP, the original developer of IA-64, dropped the project several years ago. This is a clear indication that HP didn't think the project is viable.

This is actually not totally true.
Or if it is, why the hell does HP have a website they keep updating devoted to IA-64?

http://www.hp.com/products1/itanium/

That's a bad analogy. Adding 64-bit extensions is a significant undertaking.

Can you actually provide some links for that?
Literally everything I've read about it said it's pretty trivial. Only in this case they're increasing the amount of registers which is basically the hardest part, but still very simple.

Speed <> performance. The P4 has a pathetically low IPC.

Oh give me a break, UR. I wasn't saying the MHz was the only determinant of performance. I'm saying the 2.53GHz model of the P4 is currently the fastest desktop processor out there. Knock it off with the strawmen.

If they can afford to underrate their chips, more power to them, I say

They underrate because it makes them look cooler in their home market: The Enthusiasts. I think they also anticipated P4 + PC1066 performance.

Note that DDR400 blows PC1066 away.

How do you figure?
All of the benchmarks I've seen show DDR400 outpacing PC800 RDRAM, but PC1066 and PC1200 RDRAM were fastest. See this: http://www.tomshardware.com/mainboar...rambus-06.html

[Asher]

Originally posted by Urban Ranger
Hello! Can you do the arithmatic? The increase is 14mm²/50mm². which comes out to be 28%. That's several magnitudes more than 3%.

Hello! Most of the diespace actually comes from the memory controller (do your own research on this). x86-64 results in very trivial diespace increases.

Looks like simple arithmatic bears me out

In your infinite wisdom you assumed that the memory controller (which is in reality a fairly large part) takes up precisely 0 mm². Exceptional. I take it the SSE2 execution unit takes up 0 mm² also, then?

So yeah, Intel is having problems, and they aren't above FUD.

I really don't see Intel having any problems right now.
Actually, they keep posting profits while AMD keeps posting a loss.
And their Pentium 4 cores seem to have a HUGE performance ceiling, while even AMD's Thoroughbred seems to be having trouble scaling well.

I also remember AMD promising us Thoroughbred at the end of march. It's mid-may. Where are the desktop thoroughbreds? Oh, that's right, AMD's having problems with the 0.13 process.

Intel is actually testing a fab plant at 0.09 already.

[Asher]

This just in:

http://www.aceshardware.com/Spades/read.php?article_id=119

AMD said that it only costs 5% more die space to add 64 bit instructions to their 32 bit core.

Tada.