Athlon 64 Preview

[Asher]

And it's decidedly unimpressive.

The Athlon 64 2800+ runs at a whopping 1.6GHz. The only good thing about it is the insanely low memory latency (90ns) due to the integrated memory controller.

Xbit's preview is here: http://www.xbitlabs.com/articles/cpu.../athlon64.html

The site is extremely slow (everyone's looking at it), so I've mirrored the interesting benchmarks:














I wonder what the 3.0GHz with HyperThreading would do to the Athlon 64 2800+.

Or how about the 3.2GHz P4 due within a couple weeks...

Or how about the Prescott P4 with SSE3, 1MB L2 cache, improved hyperthreading, etc.

AMD looks like it's in trouble still. The only thing Athlon 64 has going for it is the integrated memory controller and 64-bit addressing (and who really needs >4GB memory right now?), and the additional registers of x86-64.

[Promethus]

Advanced Micro Issues Challenge to Rivals
By JOHN MARKOFF with STEVE LOHR


SAN FRANCISCO, April 18 — Intel will face its strongest challenge in more than a decade when its archrival, Advanced Micro Devices, formally unveils the Opteron 64-bit microprocessor next week.

Opteron represents a determined attempt by Advanced Micro to undermine Intel's blueprint for dividing the computing universe into two worlds: one represented by its desktop 32-bit Pentium processors; the other by its pricier, corporation-oriented Itanium 64-bit chips. In doing so, Advanced Micro is trying to blur the line between the computing categories, positioning its new processor as a simpler transition for software designers and computer users from the standard 32-bit chip architecture to 64 bits, which shatters the limit on computing memory from 4 billion bytes to what, for all practical purposes, is an unlimited ceiling.

But Advanced Micro, which competes directly against Intel in both microprocessors and in the market for flash memory chips, which are widely used in consumer products like cellphones, faces a big challenge in its attempt to chip away at Intel's powerful status in the industry. Indeed, many industry executives see the Opteron processor and the Athlon 64 — its companion desktop version to be released this fall — as a make-or-break test for Advanced Micro as it tries to carve out a secure, profitable position within Intel's shadow. Last week, A.M.D., while surpassing Wall Street expectations and claiming to have gained market share against Intel, reported a $146 million loss for the quarter on revenue of $715 million. Advanced Micro, which is based in Sunnyvale, Calif., has reported losses in four of the last five years and badly needs to have a hit product in Opteron.

The company is hoping that it can take advantage of the industry transition from 32-bit to 64-bit computing, which will initially be driven by the need for corporations to handle vast databases. But looking forward several years, an even bigger market beckons as desktop computers are expected to support far more powerful graphical applications.

During the last three decades Silicon Valley has gone from what now seem like simple 4-bit designs to the ability to turn out incredibly complex 64-bit processors. Each generation of chips has permitted computers to make use of larger blocks of data while also allowing them to double the amount of information that can be handled with each super-fast tick of the computer clock.

Until now Intel has thoroughly dominated the desktop computing world, while companies like I.B.M., Sun Microsystems and Hewlett-Packard have sold more powerful and more expensive 64-bit computing systems to corporate computer users. Almost a decade ago Intel and Hewlett-Packard joined forces to design the 64-bit Itanium in an effort to blend the increasing computing potential of the more powerful processor with Intel's sophisticated manufacturing capabilities. Their first product was commercially introduced in 2001; it is now in its second generation.

Itanium, however, has been slow to catch on, in part because the chip's performance has been disappointing and in part because software designers and corporate customers have been slow to convert software from other hardware platforms. Moreover, the computer cognoscenti have generally been more enthusiastic about the Advanced Micro chip design, which is expected to offer roughly equal performance at a lower price.

Even more threatening to Intel, the Opteron chip promises to run existing software seamlessly, potentially allowing Advanced Micro to get in the door with corporate users by promising a smoother transition to 64-bit technology. By contrast, Itanium users have faced a penalty in using current 32-bit software, which runs more slowly in a special emulation mode.

Intel is already under the gun, industry executives said, because Microsoft has committed itself to offering its operating system for the Opteron as well as Itanium. If Intel were forced to introduce a similarly compatible processor, which is reportedly in the works at the company's research labs as an insurance policy, it would be forced to either follow Advanced Micro's standard or persuade Microsoft to invest in yet a third 64-bit operating system.

Perhaps more significant, Advanced Micro is pricing the new Opteron chips to compete against Intel's 32-bit Xeon processors, which are now widely used in the market for low-end corporate servers.


Continued
1 | 2 | Next>>

[Asher]

Opteron and IA-64 aren't in the same market. They both may be 64-bits, but that's not the point.

Opteron is competing with Xeon more than anything else.

[Promethus]

Advanced Micro Issues Challenge to Rivals
(Page 2 of 2)



If the Opteron lives up to the earlier performance expectations and is widely adopted in corporate applications, it could badly undercut Hewlett-Packard as well as Intel, particularly because corporate spending on information technology has been cut by the recession.

"In many ways this is an excellent time to introduce the Opteron," Hector Ruiz, Advanced Micro's chief executive, said. "It will be possible to make a simpler transition because it will be possible to run the existing 32-bit software without modification."

Hewlett-Packard and Intel executives respond that Opteron faces its own challenges in the marketplace. Bringing application software to new computer designs will be more complex than A.M.D. suggests, they said.

"It is a hard, complex problem that we've been working on for a long time," said Shane V. Robison, chief strategy and technology officer for Hewlett-Packard. "The Opteron guys are going to have to go through the same thing.

"Having the microprocessor is one thing," Mr. Robison added. "Having it ready for the enterprise — all the software, management tools and other expertise you need — is another."

He also asserted that Hewlett and Intel would retain a performance edge because of Intel's manufacturing advantages in building chips with even smaller design features.

Apart from manufacturing, Advanced Micro's biggest challenge is to establish support from industry leaders quickly for its new design.

"It's important for A.M.D. to build a sense of momentum for Opteron," said Jim Turley, publisher of Silicon-Insider, a computer industry newsletter in Pacific Grove, Calif.

Dell, I.B.M. and Sun Microsystems are all potential backers for the new A.M.D. processors but none have publicly announced a commitment yet. Industry executives said it was probable that at least one of the computer makers would announce its support on Tuesday in New York, when the new processor is publicly introduced.

The jockeying inside the industry has been particularly brutal, according to a number of people who have been engaged in the struggle.

"Some of our customers," Mr. Ruiz said, "have woken up to find horses' heads in their beds."

********************************************

Are there a lot fewer Horses in Alberta these days?

[Asher]

Where did you get that article? It's unusually poor...

"super-fast tick" of a processor

If I didn't know better, I'd say the authors of that article don't know a damn thing about what they're talking about, and just got a shipment of AMD PR about the upcoming launch of Opteron.

[Asher]

Question: How can a processor which is thoroughly stomped by a 2.8GHz P4 compete with an Itanium?

[Mordoch]

Originally posted by Asher
Question: How can a processor which is thoroughly stomped by a 2.8GHz P4 compete with an Itanium?

Its not actually an Opteron, but the value version Athlon64, and a beta version to boot for starters.

http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=18859034

[Asher]

Originally posted by Mordoch
Its not actually an Opteron, but the value version Athlon64, and a beta version to boot for starters.
http://www.siliconinvestor.com/stock...msgid=18859034

Of course it's not an Opteron, it's the Athlon 64.

The Opteron is the Xeon version of the Athlon 64.

Either way, performance-wise it looks like a huge disappointment so far.

[tinyp3nis]

It doesn't deserve 2800+ that's for sure.

[Traianvs]

Looks like a fine gaming processor to me

[Wraith]

I should have expected this one...

--"The Athlon 64 2800+ runs at a whopping 1.6GHz."

Don't forget that this is an engineering sample. X-Bit labs is breaking an NDA on this, too, which isn't impressive. But, anyway, they've got a very pre-release processor and board.
And that whopping 1.6 Ghz cleans up over the much faster clocked AXP in most cases.

Really, Asher, are you going back to the Mhz is everything line? I mean, where does that leave Intel's Centrino and Itanium?

--"so I've mirrored the interesting benchmarks:"

Leaving aside the synthetic benchmarcks (from the full list, not your list), for performance improvement over the equivalently clocked AXP I see:

60.36%
0.61%
16.35%
11.40%
4.13%
21.51%
19.06%
29.68%
33.78%
37.39%
5.44%
51.96%
3.28%
11.76%

For an average of 21.91%

Not bad for an engineering sample, especially when they're expected to be at a higher clock rate at launch.

--"I wonder what the 3.0GHz with HyperThreading would do to the Athlon 64 2800+. "

You seem to be missing the point of the PR ratings, but it's all rather irrelevant until some actual final hardware reviews come out.

--"AMD looks like it's in trouble still."

Heh. I remember what the previews of the Athlon were like. Apparently, you don't ^_^

--"Question: How can a processor which is thoroughly stomped by a 2.8GHz P4 compete with an Itanium?"

Not the same processor, obviously. The Opterons have a much different memory interface. But don't worry, the NDA on those should be lifting next week, so we should see some real results then.
The early numbers I've seen for Opteron are quite good, btw. Can't trust them much, of course, but it looks like the Xeon line is in trouble. Depending on the final power consumption and processor scaling of the Opteron, the Itaniums could be in trouble too (those things have utterly horrible power density and don't scale nearly as well as a high-end processor meant for multiprocessor use should).

Wraith
"The world is full of idiots."
-- Reki ("Haibane Renmei")

[tinyp3nis]

And that whopping 1.6 Ghz cleans up over the much faster clocked AXP in most cases.

I Don't see this. Where? I think you were supposed to type "some cases" instead of "most cases".

[Asher]

Originally posted by Wraith
Don't forget that this is an engineering sample. X-Bit labs is breaking an NDA on this, too, which isn't impressive. But, anyway, they've got a very pre-release processor and board.
And that whopping 1.6 Ghz cleans up over the much faster clocked AXP in most cases.

Really, Asher, are you going back to the Mhz is everything line? I mean, where does that leave Intel's Centrino and Itanium?

No, I'm not implyign the P4 is better because it runs at higher MHz. I'm pointing out that the rumours of AMD having troubles hitting high clockspeeds are true -- the Clawhammer runs at a fraction of the speed of the old Athlon XP, and is beaten by the old Athlon XP in several of the benchmarks. Sure, the Clawhammer is slightly faster in some of them, but it should be a helluva lot faster with the "greatly" improved branch prediction, integrated memory controller, 1MB L2 cache, 128KB L1 cache...

The article also mentions how this is the B revision the core, and the A one hit about 800MHz. This is already a very mature core, the core's been running for a long time now, AMD keeps delaying it because they're ****ing it up and can't get it to run competitive even with a 2.8GHz P4. Prescott will bury the Athlon 64.

Leaving aside the synthetic benchmarcks (from the full list, not your list), for performance improvement over the equivalently clocked AXP I see:

"equivalently clocked"? That's a semantic, compare AMD's PR ratings because that's what determines price and people's perception.

The Athlon 64 2800+ is quite frequently slower than the Athlon XP 2800+. Whether it's faster per clock (and indeed, it is) is moot if it can't hit the high clock speeds.

For an average of 21.91%

Not bad for an engineering sample, especially when they're expected to be at a higher clock rate at launch.

You are putting far too much emphasis on this being an "engineering sample". This is a very mature core, it was supposed to be out quite a while ago. The problem is they can't get it to hit higher frequencies, which is why it was delayed. It's not so much an engineering sample but a oft-delayed commercial sample that they're still trying to tweak for high speed...

I mean, it's a B stepping already. Intel's B-stepping for the Pentium 4 came out a year after it was released publically...

[OneFootInTheGrave]

ahhh.. the old athlon was apparently " dissapointing" until it came out...

so might this new athlon64 be just as well... we will see when they start benchmarking opterons when they are unveiled this week on tuesday, I guess that will be better indication how will that processor fares...

[Wraith]

--" I think you were supposed to type "some cases" instead of "most cases"."

Not really. I remember Asher going on about how no-one needed more power in office apps, that games and the like were where it counts, and that certainly looks like where the A64 is headed. Still, he does get on my nerves, so I tend to overract to his idiocies.
I find it more amusing to see what benchmarks they left out. Still, we'll have to wait for the real release to find out.

--"I'm pointing out that the rumours of AMD having troubles hitting high clockspeeds are true "

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.

The A rev cores were locked at 800 Mhz, btw, which isn't the same thing as limited to. They were development pre-release machines for the most part and AMD certainly wouldn't want Intel to find out everything about them then.

--" That's a semantic, compare AMD's PR ratings "

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

--"the problem is they can't get it to hit higher frequencies,"

Ah, again, who do you happen to know at AMD? I know you keep saying this, but you really don't have much proof. They're working with SOI, so it's not at all unlikely they had process problems, but scaling problems are much less likely.
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. It makes the best economic sense to hold off on the consumer chips and dedicate the SOI fab space to the higher-margin Opterons for now. More validation required, but more profits would be worth it.

Wraith
"Have you the brain worms?"
- Zim ("Invader Zim")