Athlon 64 Preview

[Asher]

Originally posted by Wraith
I'm glad to see you've got contacts at AMD, since there's no way you could possibly draw that conclusion from the preview.

Well, let's see the logic behind it and see if it makes sense, shall we?
1) Constant delays, 1.5-2 years in fact
2) Revision A cores "locked" at 800MHz -- you don't think they'd lock them all at 2GHz if they could hit those frequencies?
3) Increased L2 cache to 1MB to help offset lower clockspeeds
4) Last-minute $40M cheque sent to IBM to help them get SOI working on their chips.
5) The fact that the 2800+ chip is nowhere near the performance levels of a 2800MHz P4, and the clockspeeds are nowhere near the levels of an Athlon XP 2800+.
6) The Athlon core's history of clockspeed problems on 0.13 -- remember all of the Thoroughbred cores? It's not like this is unprecedented for AMD.

All signs point to the rumours of delays due to inability of hitting high clockspeeds.

Of course, you're welcome to cover your eyes and shout "There is no proof", but most people are slightly more perceptive than that.

Yes, please do. Leaving aside the synthetic benchmarks the A64 2800 does quite well against the AXP 2800.

If you say so -- the results in these benchmarks are extremely disappointing. The Athlon XP is already having a hard time keeping up with the Pentium 4, and the Athlon 64 doesn't appear to be any faster. So what's going to happen to AMD when Prescott launches, whose entire purpose is simply speed boosts?

You also seem perfectly willing to ignore the market conditions. Barton is doing fine against the P4, spending in the computer sector has slumped anyway, and so on.

If Barton is doing "fine", how come AMD's marketshare continues to shrink? How come AMD continues to post losses? How come the performance delta between the Athlon XP and Pentium 4 continues to increase? How come AMD had to rework the performance ratings on the new core to make it look even more competitive (when it clearly is not)?

Barton may be doing "fine" if that's how you want to look at it, but compared to Northwood-B, Barton's pretty pathetic. There's almost no die-size difference between the two cores now, and Intel's got 300mm wafers they're working with, and no-doubt has lower manufacturing costs now.

And Prescott is coming soon. Prescott will compete with the Athlon 64. And let's face facts, the Athlon 64 looks like a dud in the performance department when you compare them with Intel's current Pentium 4s even.

[August Borms]

This thread stinks awful....

[CyberGnu]

Not to threadjack too much, but if I want to buy a medium range processor ($100-200) in the next couple of months, is AMD still the best bang for the buck?

[Lazerus]

I thought athlon were upto 2.4ghz ?

[Zopperoni]

Originally posted by CyberGnu
Not to threadjack too much, but if I want to buy a medium range processor ($100-200) in the next couple of months, is AMD still the best bang for the buck?

Not really, the P4 2.4 GHz is very good value.

I'm getting one next week, along with an Intel 865G motherboard.

[Asher]

Originally posted by Zopperoni
Not really, the P4 2.4 GHz is very good value.

I'm getting one next week, along with an Intel 865G motherboard.

Wise choice, the 865 and 875 are awesome.

Is the 865 really out next week though, I only thought the 875 were out now and 865 coming later?

[Zopperoni]

Ehm, according to the guy at the store it's coming to Europe on April 22nd.

I was at his store last Wednesday. He told me that the 865 was launched on that day in the US, and already launched in Japan on Tuesday.
He was receiving a shipment of 875s on that day too (he gets them directly from Intel).

I hope to get a call from him on Tuesday, so I can pick it up Wednesday morning =)

[Urban Ranger]

Originally posted by Asher
1) Constant delays, 1.5-2 years in fact

I am glad that the Itanic came out on time

Originally posted by Asher
2) Revision A cores "locked" at 800MHz -- you don't think they'd lock them all at 2GHz if they could hit those frequencies?

Yes and no. Engineering samples usually are never set to the highest possible. They are like beta releases with minor bugs to remove.

Originally posted by Asher
3) Increased L2 cache to 1MB to help offset lower clockspeeds

I heard the newer P4s will come with bigger caches too

Originally posted by Asher
4) Last-minute $40M cheque sent to IBM to help them get SOI working on their chips.

Well? What's the problem here?

Originally posted by Asher
5) The fact that the 2800+ chip is nowhere near the performance levels of a 2800MHz P4, and the clockspeeds are nowhere near the levels of an Athlon XP 2800+.

There are engineering samples. I gathered you have never worked with a real engineering project before.

Originally posted by Asher
6) The Athlon core's history of clockspeed problems on 0.13 -- remember all of the Thoroughbred cores? It's not like this is unprecedented for AMD.

Don't we all know that clockspeed is a bad indication of processor power?

Originally posted by Asher
All signs point to the rumours of delays due to inability of hitting high clockspeeds.

There are, of course, other various rumours. Like waiting for MS releasing a 64-bit version of Windows.

Originally posted by Asher
If you say so -- the results in these benchmarks are extremely disappointing. The Athlon XP is already having a hard time keeping up with the Pentium 4, and the Athlon 64 doesn't appear to be any faster. So what's going to happen to AMD when Prescott launches, whose entire purpose is simply speed boosts?

Fast doesn't mean jack. Higher clockspeed is not the same as executing more instructions per second.

Of course, since there is no traces of said "Prescott," it's all FUD so far.

Originally posted by Asher
If Barton is doing "fine", how come AMD's marketshare continues to shrink? How come AMD continues to post losses?

Last I heard, the PC market hasn't been doing so well in the last few years.

Originally posted by Asher
How come the performance delta between the Athlon XP and Pentium 4 continues to increase?

In what sense?

Originally posted by Asher
How come AMD had to rework the performance ratings on the new core to make it look even more competitive (when it clearly is not)?

Clearly, 64-bit processors run differently from 32-bit processors. That should be elementary.

Originally posted by Asher
And Prescott is coming soon. Prescott will compete with the Athlon 64. And let's face facts, the Athlon 64 looks like a dud in the performance department when you compare them with Intel's current Pentium 4s even.

If it comes out, then we'll talk. Right now, I haven't any Prescott.

[Asher]

Originally posted by Urban Ranger
I am glad that the Itanic came out on time

I'm not saying there was no problems with the Itanium. But it's not really the same thing.

The Itanium is a radical new design, fundamentally new instruction set, and theoretical new concepts. Clawhammer is an Athlon with more registers, and 64-bit addressing space and integer registers, and an integrated memory controller.

Bit different.

I heard the newer P4s will come with bigger caches too

That's true, but the reason the bigger cache will come is because they're moving to 0.09 and have the space to spare. Plus, the P4 design is very sensitive to L2 cache and memory bandwidth.

Well? What's the problem here?

They spent ~8 months trying to do it themselves, gave up, called 911.

There are engineering samples. I gathered you have never worked with a real engineering project before.

I've already addressed this...

Don't we all know that clockspeed is a bad indication of processor power?

Strawman...
It's a bad indictation, but it's also a critical component of processing power.

There are, of course, other various rumours. Like waiting for MS releasing a 64-bit version of Windows.

The IA-64 version has been out for quite some time.

Fast doesn't mean jack. Higher clockspeed is not the same as executing more instructions per second.

Of course, since there is no traces of said "Prescott," it's all FUD so far.

Prescott has been shown publically, hell there are IMAGES of the core on the internet. It's just not out yet because 0.09 micron isn't ready for mainstream.

In what sense?

About a year ago, the Athlon was faster than the Pentium 4 in general.
Now, the Pentium 4 is faster than the Athlon. With the addition of HT and 800MHz system bus, the gap continues to grow...

Clearly, 64-bit processors run differently from 32-bit processors. That should be elementary.

I was referring to Barton. It's still a 32-bit processor, but they adjusted the PR ratings to make it look faster than it actually is so it didn't look like they were falling too far behind.

If it comes out, then we'll talk. Right now, I haven't any Prescott.

The Athlon 64 isn't out yet either, so we're supposed to assume it doesn't exist?

There are more details of Prescott known than Athlon 64, physically. Show me the equivalent Athlon 64 shots of the die (these are Prescott):








Are you still going to fall behind that argument about Prescott being vapourware?

[Asher]

There's also a good 8-page document on the new SSE instructions (SSE3?) in Prescott here: http://www.xbitlabs.com/articles/cpu...scott-sse.html

[Wraith]

--"Well, let's see the logic behind it and see if it makes sense, shall we?"

Asher and logic don't seem to go well in the same sentence, and I've already talked about a few of your points a bit.

But let's see now.

--"3) Increased L2 cache to 1MB to help offset lower clockspeeds"

Um... were you the one talking about logic? L2 cache doesn't do anything to offset lower clockspeed. It can offset the low clock speed of the connection to memory, but, well, that's not the case here.

--"4) Last-minute $40M cheque sent to IBM to help them get SOI working on their chips."

Yeah. Like I said, process problems. You don't seem to understand what that means.

--"5) The fact that the 2800+ chip is nowhere near the performance levels of a 2800MHz P4,"

Dunno, they don't seem that far off to me.

--"and the clockspeeds are nowhere near the levels of an Athlon XP 2800+."

Well, yeah. That was kind of the point of the design. Higher IPC.

--"6) The Athlon core's history of clockspeed problems on 0.13 -- remember all of the Thoroughbred cores?"

More process issues, for the most part. They happen, but Intel suffers from them too. You forgetting their recent recall already?

--"-- the results in these benchmarks are extremely disappointing."

It runs roughly equal with the AXP of the equivalent PR rating, even beating it in the more important ones like gaming. If you were expencting a whole lot more, then your expectations were too high. Think about the point of the PR ratings for a bit.

--"If Barton is doing "fine", how come AMD's marketshare continues to shrink?"

Well, the entire market is kind of shrinking, so yeah, that could be part of it. Not to mention Intel's long marketing campaign directed at corporations. There's still a "you won't get fired for buying Intel" mentality, and AMD really doesn't know how to market to corporations at all. So it's really not that surprising.

BTW, Intel's reputation for stability and quality is overblown. I don't work on PCs, but doing testing at Dell has let me learn some interesting things about Intel's products.

--"And Prescott is coming soon."

So it is. Shows that Intel's a bit wiser about the market than you are. They wouldn't have been pushing the releases forward if they weren't worried.

--"There's also a good 8-page document on the new SSE instructions (SSE3?) in Prescott here:"

Yeah, more XBit labs. They're pretty pro-Intel, but that's okay. SSE3 really doesn't add a whole lot to the mix. From what I've heard from programmers, there's only a couple of new instructions that will actually get used for much of anything.

BTW, that link barely addresses the new instructions. It's mostly a history lesson about MMX/3DNow/SSE/SSE2.

Wraith
"I'm rattling the cans. People expect it."
-- Garbage Man ("Dilbert")

[Asher]

Originally posted by Wraith
Um... were you the one talking about logic? L2 cache doesn't do anything to offset lower clockspeed. It can offset the low clock speed of the connection to memory, but, well, that's not the case here.

I don't think you're thinking this through well enough -- a higher count of L2 cache increases per-clock performance on average because it has less stalls waiting for system memory. It improves IPC somewhat, which is what you need when your clockspeeds are less than stunning.

Yeah. Like I said, process problems. You don't seem to understand what that means.

I don't understand why it's relevant -- why do I care about the source of not hitting the high clockspeeds? The design, the process, they all mesh together to give you a pretty lackluster performing processor.

Well, yeah. That was kind of the point of the design. Higher IPC.

So why did they elongate the pipeline if that was the goal of the design?

Well, the entire market is kind of shrinking, so yeah, that could be part of it.

Intel's marketshare has increased, and Intel's revenues have either gone up or been flat. If the market is shrinking, it's AMD that's getting the shaft, and not Intel. And either way, that's not a really valid excuse because it doesn't seem to affect Intel.

So it is. Shows that Intel's a bit wiser about the market than you are. They wouldn't have been pushing the releases forward if they weren't worried.

Prescott was originally due out in the middle of 2003, now it's near the end -- how is this "pushing it forward"?

And why does it show Intel being a bit wiser about the market than I am?

[CyberGnu]

Thanks guys

[Asher]

BTW, AcesHardware posted news about the POVRay benchmark running on a 1GHz Itanium 2:

It completed it in 00:11:07. A 2.29GHz Athlon XP did it in 00:26:07.

[tinyp3nis]

Originally posted by CyberGnu
Not to threadjack too much, but if I want to buy a medium range processor ($100-200) in the next couple of months, is AMD still the best bang for the buck?

It is, don't listen to them. Even more so, if you are into overclocking.