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

[SpencerH]

Originally posted by Asher
That sounds extremely odd, and ambiguous...

I put together a PII 233 for my kids from parts. Tyan MB with 440 chipset. Should be straight-forward.

Put it together ->wont post.
Change Ram positions ->wont post.
Change bios config ->wont post.

Re-test ram in P166 works fine. maybe MB or CPU shot!

many frustrating days later
Check wims bios page (an excellent resource BTW). Read about older intel chipsets not handling 32bit ram. Buy new 16bit ram. Boots up fine.

Try to install ati AIW video card

Monitor does not work at all
Fiddle with everything known to man ->Monitor does not work at all
check monitor and video card both fine

many frustrating days later
conclusion: intel chipset wont handle 32bit ram on video card

AMD seems to build in more flexibility in their CPU's (or it could be the chipset they use). This may have been an extreme case, but I just dont need the hassle.

[Asher]

I've had the opposite experiences, quite frankly.
I think your problems might have been from the Tyan motherboard if anything (I know their AMD motherboards are troublesome, but cheap). Who knows.

And I also don't think the problem was with the intel chipset. My P3 600 has the 440 chipset and it works with a Radeon 64 perfectly fine with no troubles...as well as regular, run-of-the-mill PC100 SDRAM.

[Urban Ranger]

Glonk,

"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."

How many years Itanic is under development? Say Intel has four years to wipe up a good compiler for an emulator. It's not here yet. It probably will take a decade or two, if it is lucky.

"That's right, but it takes the cake in that the IA-64 significantly outperforms x86-64 in 64-bit applications."

That's assuming there will be excellent compilers for it. A big if. Very big if.

"McKinley is the next IA-64 processor, due out in June. Prescott is the next IA-32 processor, based on the Pentium 4."
"From what I've seen, Prescott will absolutely lay waste to the *Hammer chips when it comes out in 2H03."

Wow, so this Prescott is a wonder chip eh? A 32-bit CPU *cough* destroys *cough* a 64-bit CPU.

/me pinches his nose.

BTW, what is "IA-32?" x86 is more like it.

"Intel just needs to invest in a larger x86-64 emulation unit in the IA-64."

Are you expecting somebody is going to pay 5x as much for a CPU that will run existing 32-bit applications at 50% of the speed as a Clawhammer with no forseeable benefits? What world do you live in anyway?

"This is actually not totally true."

What is not totally true, that HP stopped R&D on the EPIC line? The HP website merely is an indication that the company is willing to build machines based on IA-64 if it comes out.

"Can you actually provide some links for that?"

What links? I don't have any links. Show me that it is trivial to upgrade the microcode store, to implement all the new signals, to extend the size of the registers, to enlarge the L1 and L2 caches, to put in a new branch prediction unit, and the whole works.

"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."

What you read is wrong. Just wait till you have some courses in microprocessor design.

"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."

And I was saying speed does not equate performance. Where's the strawman?

"They underrate because it makes them look cooler in their home market: The Enthusiasts."

Right. No, it's the same reason why CPUs can be overclocked, is because the ratings have a safety margin built-in.

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

Oh, no links on your part? That's not your style Glonkie. You're are always so eagar to stick your "research" in other people's faces. I take that as "I made that up but I am not going to admit it."

"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?"

The SSE2 unit is just one of the FPUs. The memory controllers aren't going to take 20%+ die space.

"And their Pentium 4 cores seem to have a HUGE performance ceiling, while even AMD's Thoroughbred seems to be having trouble scaling well. "

Big whop. Thoroughbred will be replaced by Barton which will be replaced by Clawhammer, while there's no 64 bit thing to show from Intel. It is quite clear who's taking the lead here BTW, what's this about performance ceiling for the P4? Didn't you just agreed that speed is not the same as performance?

"This just in: http://www.aceshardware.com/Spades/...?article_id=119"

Your article is slightly out of date. Try something more recent. Can't find any? Shameful


LightEning,

Sometimes it behooves all of us when misinformation is corrected.

[FemmeAdonis]

Just my two teeny cents

I have the new imac too. Bought it before it was ever ready to ship from apple..its an axcellent machine...very compact and throws a good punch.

I woudlnt trade my gem for any PC...

I DO plan on buying a PC as well...but I'll never love it as much as Iv EVER loved MACINTISh period.

It runs beautifully, functions flawlessly, and looks neat sitting on my little desk with its little subwoofer-like speakers, flat screen that I can adjust to my hearts content...

and WHAT computer greets you as sweetly as a mac??

gotta love it. <3 <3

[Urban Ranger]

Glonk,

One more thing. The Claw Hammer will be 64mm² at 90nm, not the Opteron (Sledgehammer).

You realise that the Claw Hammer does not have built-in memory controllers, yes?

So much for "shooting down" my arguments.

[Asher]

Originally posted by Urban Ranger
How many years Itanic is under development? Say Intel has four years to wipe up a good compiler for an emulator. It's not here yet. It probably will take a decade or two, if it is lucky.

Actually they've made a pretty damn good compiler so far, with regards to floating point in SPEC.
They can do it in four years. Especially considering the Alpha engineers are on the job now, too.

That's assuming there will be excellent compilers for it. A big if. Very big if.

Actually, using SPEC benchmarks in floating point, the Itanium 2 is going to be faster than the Sledgehammer.

Wow, so this Prescott is a wonder chip eh? A 32-bit CPU *cough* destroys *cough* a 64-bit CPU.

* Urban Ranger pinches his nose.

You do know that the extra bits do not equal extra speed, right? Only extra precision? Please tell me you know that.

Prescott is the chip with Yamhill too, which is x86-64.
Not that you didn't know that, since you know all.

BTW, what is "IA-32?" x86 is more like it.

Ever since the Pentium 4 it's now referred to as IA-32, because of things like MMX, SSE, and SSE2 are not part of x86 per se.

Are you expecting somebody is going to pay 5x as much for a CPU that will run existing 32-bit applications at 50% of the speed as a Clawhammer with no forseeable benefits? What world do you live in anyway?

Jesus UR...think a little.
Obviously when it moves to the desktop, the chip wouldn't cost $3000. The Itanium 3/4/5 itself will not become a desktop job, but a derivative of that core would be. Obviously, it'd be competitive to AMD's chip offerings.

And how do you figure it'll run it at 50% of the speed of the Clawhammer in 4 years or so? Wishful thinking?

What is not totally true, that HP stopped R&D on the EPIC line? The HP website merely is an indication that the company is willing to build machines based on IA-64 if it comes out.

So why do HP engineers keep posting on the comp.arch newsgroup about new design possibilities?

What links? I don't have any links.

Ah, okay, move along then...

Show me that it is trivial to upgrade the microcode store, to implement all the new signals, to extend the size of the registers, to enlarge the L1 and L2 caches, to put in a new branch prediction unit, and the whole works.

Do a google search.
First of all, it is EXTREMELY trivial to expand L1 and L2 caches.
Second of all, they're not putting in a new BPU, they're "enhancing" the Athlons, which they also did for the Athlon XP.
Third of all, it's not very hard to expand the registers -- at all. You simply double them in size and reorganize the chip.
The hardest part is implementing the new instructions without eratta.

AMD themselves says it only adds 5% to the chip size.

What you read is wrong. Just wait till you have some courses in microprocessor design.

Yeah, what do the AMD engineers know?

"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."

And I was saying speed does not equate performance. Where's the strawman?

Because all I was saying is the 2.53GHz Pentium 4 is the best performing (read: fastest) desktop CPU out there.
Your strawman was assuming I meant clock frequency. If you did not assume that, you made a really dumb comment anyway. We need the CAPTAIN OBVIOUS picture here.

Oh, no links on your part? That's not your style Glonkie. You're are always so eagar to stick your "research" in other people's faces. I take that as "I made that up but I am not going to admit it."

I did link to a respectable hardware site (AcesHardware), which told you that AMD said only 5% of the diespace is attributed to x86-64. It's also pretty much common sense that memory controllers are hefty, since anyone can see that looking at a motherboard...

The SSE2 unit is just one of the FPUs. The memory controllers aren't going to take 20%+ die space.

You're telling me that AMD still isn't doing a dedicated SSE2 unit a la the Pentium 4?

Big whop. Thoroughbred will be replaced by Barton which will be replaced by Clawhammer, while there's no 64 bit thing to show from Intel.

Incorrect on both counts.
Barton will NOT be replaced by Clawhammer, but is becoming the value processor. And Intel is putting x86-64 in Yamhill, which is in Prescott, which will also feature a 800MHz FSB + 1MB L2 cache + other instructions (SSE3?).

It is quite clear who's taking the lead here BTW, what's this about performance ceiling for the P4? Didn't you just agreed that speed is not the same as performance?

I'm arguing speed as in performance. Speed at which something is rendered, frames are rendered, operations are completed.

Speed is, in fact, the same as performance.
Frequency is not.

Your article is slightly out of date. Try something more recent. Can't find any? Shameful

Go ask your question on www.RealWorldTech.com and have them answer you. I know it's 5%. It makes SENSE it's 5%.

Why would it change in a while anyway?

[Asher]

Originally posted by Urban Ranger
Glonk,

One more thing. The Claw Hammer will be 64mm² at 90nm, not the Opteron (Sledgehammer).

You realise that the Claw Hammer does not have built-in memory controllers, yes?

So much for "shooting down" my arguments.

Ahhh, you're right. ClawHammer doesn't have the built-in memory controllers.

Instead it has much more L2 cache, which greatly adds to size.

If x86-64 is so big and complicated to add, how come the die only shrunk at the same rate as moving from the Katmai P3 to CuMine P3?

http://www.geek.com/procspec/newsletter/nlproc06202001.htm

Katmai P3: 125mm²
dieshrink from .25 -> .18
CuMine P3: 106mm²

Athlon XP: 128 mm²
ClawHammer: 105 mm²