Holy crap, AMD's unleashed a can o' whoop-ass with A64

[Jakub]

http://firingsquad.gamers.com/hardware/athlon_fx_preview/

They actually took Q3A benchmarks and utterly destroyed the P4 in UT2K3.

[Machfive]

I'm loving the numbers, but quite dissapointed at the quoted price point.

AMD's charm is that they would undercut Intel even when they had the better product. The original Athlon smoked the PIII, even in its precious floating-point, and still cost less.

I want AMD to make money and all, but I fear they'll have a hard time doing it having spent the last several years catering to the bargain-minded.

[Jakub]

Well, the Athlon 64 is their ticket to the top. The original Athlon was superior to the P3, but rather unpolished and didn't have a killer feature to set it apart. This... this is AMD's ticket to the high markup product range.

[Guido Jones]

I'd be interested how the AMD chip faces up against those new Extreme Gamer P4's (I don't think that's what they were matched against, could be wrong) - The numbers are somewhat impressive though.

[Desslock]

I'd be interested how the AMD chip faces up against those new Extreme Gamer P4's (I don't think that's what they were matched against, could be wrong) - The numbers are somewhat impressive though.

Tom's has a direct comparison article. Bottom line: P4 wins, again.

http://www20.tomshardware.com/cpu/20030923/athlon_64-53.html

[Guido Jones]

Hey, thanks Desslock :)

If the prices are similar, looks like I'll be moving back to Intel.

[SpoofyChop]

I don't know if that's the best summary of the article Desslock.

Anyway, it has never been made clear to me that 64bit processors are such a huge improvement over 32bit the way 32bit was over 16bit.

[Machfive]

That's because there's no software, drivers, and OSes that are in a finished state to take advantage of them.

Honestly, I don't imagine there'll be much improvement in most things, but for gaming and high-end applications, I do see a potential for large gains.

[Machfive]

Tom's has a direct comparison article. Bottom line: P4 wins, again.

http://www20.tomshardware.com/cpu/20030923/athlon_64-53.html

I haven't trusted a single thing Tom or his lackeys have posted since several years ago.

[SpoofyChop]

Machfive:

You have not been here long enough for your completely unsupported opinions to count for anything.

Now if Lloyd or Chet or somebody who has proven that they know what they are talking about would come and say something about Tom's, people will take notice.

But as for you, if you expect anybody to give a flying leap of shit what you say, you might want to back up your assertions with facts.

[Stroker Ace]

MACH FIVE DOES NOT NEED FACTS! MACH FIVE WOULD LIKE SOME COFFEE!


:wink:

just kidding... i don't know anything about processors. I *am* curious what you've got against Tom's, though.

[TimElhajj]

Machfive: You have not been here long enough for your completely unsupported opinions to count for anything.

Yeah, wait for > 2000 posts*.

*Note: Even with > 2k posts, the cash value of unsupported opinions is equal to one tenth of one cent. Have a nice day.

[Matthew Gallant]

But as for you, if you expect anybody to give a flying leap of shit what you say, you might want to back up your assertions with facts.

Yeah, it's either facts or > 2000 posts.
Well he'll be there in 95 days, so no worries there.

[TimElhajj]

But as for you, if you expect anybody to give a flying leap of shit what you say, you might want to back up your assertions with facts.

Yeah, it's either facts or > 2000 posts.
Well he'll be there in 95 days, so no worries there.

Damnit, Matthew. Give me a minute to think my zingers through, would ya. :)

[Desslock]

Anand's team has come to a similar conclusion as Tom's Hardware: http://www.anandtech.com/cpu/showdoc.html?i=1884&p=18

[Case]

I'm loving the numbers, but quite dissapointed at the quoted price point.

AMD's charm is that they would undercut Intel even when they had the better product. The original Athlon smoked the PIII, even in its precious floating-point, and still cost less.

I want AMD to make money and all, but I fear they'll have a hard time doing it having spent the last several years catering to the bargain-minded.

I've got numbers on the Athlon64 3200+ up, http://www.extremetech.com/article2/0,3973,1276955,00.asp. It's much cheaper (around $440 on the street) and is surprisingly good, despite the 64-bit memory width.

[Case]

Machfive:

You have not been here long enough for your completely unsupported opinions to count for anything.

Now if Lloyd or Chet or somebody who has proven that they know what they are talking about would come and say something about Tom's, people will take notice.

But as for you, if you expect anybody to give a flying leap of shit what you say, you might want to back up your assertions with facts.

Actually, I'd never say something negative about a competitor. It would be ungentlemanly.

But the Athlon64 does beat the P4 in a lot of games -- most of the seven or so we test, anyway. We also debuted the Flight Sim 2004 benchmark in this roundup in what will likely be its final form.

[SpoofyChop]

Hrm...sorry about that Lloyd, I wasn't really thinking about you in terms of your ExtremeTechness as much as I was thinking of you in terms of your knowseverythingabouthardwareness.

See? You transcend your current project! Lucky you!

:D

[SpoofyChop]

Now Chet on the other hand, I expect Chet to bash the hell out of everything.

:D

[Jason Cross]

The TomsHardware article sort of confuses me a little. From the way I read it, it seems to come to the conclusion that the P4 Extreme Edition is better because it "wins" like 32 tests, and the Athlon XP wins 15.

That just seems odd because if I were to test a much greater number of applications that the P4 is good at, it would certainly make the P4 seem better. And not take into account the merit of those apps, their configuration, or their possible redundancy (ie... how many different 3D rendering packages does one need to test?). It also doesn't quite take into account the degree to which one product outpaces the other. Ie...if the Athlon 64 is only scarcely slower on those 32 tests it lost, but is far faster on the 15 it won, is it a better overall product?

Anyway, from what I saw when running a boatload of tests at CGM, and from what I've seen of all Loyd's work here at ExtremeTech, the FX-51 is maybe the better product overall (though it's close).

The real determinant is going to be 64-bit. If 64-bit Windows XP shows up around March or so of next year with performance on 32-bit apps that is at least 95% of what we see on current Windows XP, and with a good number of high-profile 64-bit apps, then AMD's got a home run.

If not, then one has got to think that Intel will put a hurt on them.

It's also worth noting that AMD's architecture is extremely effecient, and gets more real-world performance out of each additional MHz they turn up the clock rate. If they can keep coming out with 200MHz speed bumps in tune with Intel, they will pull further out in front in real performance. Going from 2.2GHz to 2.4GHz will do an Athlon 64 FX a lot more good than going from 3.2 to 3.4GHz will do a Pentium 4.

[Qenan]

The thing that scares me about Win64 is that it apparently can't use the 32-bit drivers... that's a lot of driver issues to solve. Probably OK if you don't have old hardware you need to keep running, but...

[Machfive]

Anand's team has come to a similar conclusion as Tom's Hardware: http://www.anandtech.com/cpu/showdoc.html?i=1884&p=18

I'd make a blanket statement, but I'm not in the mood to do some googling to back up my assertions. ;)

So I'll say this much - I do trust Anandtech.

I like what these benchmarks are showing. As per the last round, the AMD chip is lagging behind in content encoding, winning in business, losing in 3D, and producing mixed results in gaming.

However, I noted a few very neat things.

AMD is poised to do well on synthetic CPU benchmarks thanks to its on-die memory controller. Though most of of us would agree synthmarks are rather nebulous and useless, there are still a large chunk of computer journals that integrate synthmarks into their overall benchmarking schemes, albeit to a lesser extent than real-world benchmarks. As a result of this, AMD is bound to gain a few tiny notches on its overall scores in reviews, regardless of the actual relevance of synthmarks.

AMD has thorough domination of the business performance numbers. Though not everyone is running 50 meg spreadsheets, it's nontheless something to tout. I imagine multitasking performance as a whole will be nicer with them as well, if the business test functions as I am under the impression it does.

Although AMD lost the overall content creation performance, there's one thing worth noting - It rolls the encoding tests, such as MP3 and video, into the same test as content creation, such as Photoshop and Illustrator. If anyone recalls the breakdowns of the Athlon XP's performance versus that of the P4's, you will remember that while the XP sagged in encoding, it took the lead in creation. As a graphic designer, I'd be very interested in seeing individual tests of Photoshop, Illustrator, and Premiere on both processors.

In Anand's Gaming benchmarks, the Athlons won every test except for the DX7/OGL tests. I look forward to seeing how the two processors compare on a DX9 title, such as Half-Life.

Although the A64 lost in Tom's Workstation tests, they excelled in Anand's. These were very different tests though, so one needs to consider what exactly their "workstation" is going to be used for, and choose which benchmark reflects that more accurately.

The 3D rendering tests were a definite loss for AMD, although the A64 showed much improvement over the XP. I'd be interesting in seeing a Maya benchmark as well, since the gap between the P4's and XP's in that program weren't as stark as they were in Lightwave, IIRC.

A final interesting note: the Lame MP3 encoding under the 64-bit environment could very well indicate that the switch to 64-bit will aid AMD's performance in content encoding.


Anyways, that's my take on Anand's results. I think they beckon well for AMD, especially if they can manage to get some decent bins and ramp up the clock speed to take the lead away in areas they lost in.

But I think the future looks promising.

[DennyA]

This will get more interesting when
1) 64-bit XP ships, and
2) We find out how compatible 32-bit apps and games are on 64-bit XP

Right now, it's like when the 386 first shipped but all the apps just used it in Virtual 8086 mode.

[mtkafka]

Micrisoft will never support AMD. And thats why Wintel will win in the end.

etc

[XPav]

Apparently, this is some new definition of "never" which apparently means "on Tuesday."

http://www.informationweek.com/story/showArticle.jhtml?articleID=15200133


Microsoft on Tuesday launched a beta version of its Windows XP operating system that supports AMD's new 64-bit Athlon on the desktop and the chipmaker's Opteron 64-bit processor on workstations and servers. A final version will be released in the first half of 2004, Microsoft said.

[mtkafka]

Doh! Cool. I hate Intel. Microsoft ain't that bad I guess.

etc

[Brad Grenz]

Although AMD lost the overall content creation performance, there's one thing worth noting - It rolls the encoding tests, such as MP3 and video, into the same test as content creation, such as Photoshop and Illustrator. If anyone recalls the breakdowns of the Athlon XP's performance versus that of the P4's, you will remember that while the XP sagged in encoding, it took the lead in creation. As a graphic designer, I'd be very interested in seeing individual tests of Photoshop, Illustrator, and Premiere on both processors.

I'm not sure what the deal is because HardOCP's Content Creation tests (www.hardocp.com/article.html?art=NTI0LDQ=) show the Athlon 64 FX mauling the P4 EE. Anand's mentions they didn't use the patch that was supposed to fix some problem with Winstone 2003 and Athlons. I assume HardOCP did use the patch. I think it's a pretty suspect bench either way, 'cause isn't this the benchmark that AMD has claimed Intel completely cooked in the P4's favor? You know, by having the program loop tests the P4 is good at in a basically arbritrary fashion? (well, arbritrary in the sense that you can't explain the reasoning unless you admit you were trying to make the P4 look better)

HardOCP also shows the Athlon 64 winning the Divx compression test pretty handily, both using Divx 5.0.5... I'm not sure what they're doing differently.

The Athlon 64 shows really well in the games. Most of these sites have the P4 EE winning Q3 and Q3 based games (Jedi Knight 2). Q3 is so outdated these days I don't think it even matters. Christ, who cares if you get 500 or 460 FPS? In anything modern the P4 is at best on par, and at worst handily defeated.

[Peter Frazier]

Well, I'm buying AMD because I refuse to buy any product with the word 'extreme' in it. Can someone tell corporate America that using the word is just so, so pitiful.

[Case]

I'm not sure what the deal is because HardOCP's Content Creation tests (www.hardocp.com/article.html?art=NTI0LDQ=) show the Athlon 64 FX mauling the P4 EE. Anand's mentions they didn't use the patch that was supposed to fix some problem with Winstone 2003 and Athlons. I assume HardOCP did use the patch. I think it's a pretty suspect bench either way, 'cause isn't this the benchmark that AMD has claimed Intel completely cooked in the P4's favor? You know, by having the program loop tests the P4 is good at in a basically arbritrary fashion? (well, arbritrary in the sense that you can't explain the reasoning unless you admit you were trying to make the P4 look better)


I dunno how HardOCP got their results. I applied the Windows Media Encoder patch to the O/S, which fixes the SSE detection for Content Creation Winstone. I don't get anywhere near their score. BTW, the benchmark you're thinking about is Sysmark 2003, not the Winstone tests.




HardOCP also shows the Athlon 64 winning the Divx compression test pretty handily, both using Divx 5.0.5... I'm not sure what they're doing differently.


I used DivX 5.1, and got a slightly different result. The FX-51 cleans up on the standard P4, but the P4EE running on an Asus P4C800E wins, though only by a narrow margin.



The Athlon 64 shows really well in the games. Most of these sites have the P4 EE winning Q3 and Q3 based games (Jedi Knight 2). Q3 is so outdated these days I don't think it even matters. Christ, who cares if you get 500 or 460 FPS? In anything modern the P4 is at best on par, and at worst handily defeated.

The Athlon 64 does show well in games. It's not the 64-bit stuff, though, but the on-die memory controller. Even the cheaper Athlon 64 3200+ runs faster than the P4EE in most games we tested.

[Brad Grenz]

I used DivX 5.1, and got a slightly different result. The FX-51 cleans up on the standard P4, but the P4EE running on an Asus P4C800E wins, though only by a narrow margin.

Did you use the SSE 2 optimized iDCT for the P4? What about for the Athlon 64s? Tom's discloses the fact that they used the MMX optimized version for the Athlons. I'm thinking there may be a version optimized for the Athlons that HardOCP used. I've emailed Kyle Bennett asking about it.

[Case]

I used DivX 5.1, and got a slightly different result. The FX-51 cleans up on the standard P4, but the P4EE running on an Asus P4C800E wins, though only by a narrow margin.

Did you use the SSE 2 optimized iDCT for the P4? What about for the Athlon 64s? Tom's discloses the fact that they used the MMX optimized version for the Athlons. I'm thinking there may be a version optimized for the Athlons that HardOCP used. I've emailed Kyle Bennett asking about it.

Yes, I let the encoder autodetect the CPU capabilities. It used the optimized SSE2 version for both.

If the preliminary tests on the partial DivX encode in 64-bit Windows is any indication, a 64-bit DivX codec could perform very well on encode. DivX Networks was at the AMD launch event, heavily endorsing both AMD64 and Win64. The nature of the encode lends itself very well to 64-bit processing.

[Jason Cross]

Things like video/audio encoding and 3D rendering (not games, 3DSMax and such) are the kinds of apps that should benefit quite a bit from the double register count available in 64-bit. We're talking as much as 30% in some cases, and it's realistic to expect 15-20%. That's big.

As for games, it's a little less clear. The best indication I can get from Tim Sweeney, who is big into doing native UT2003 and 2004 ports for 64bit, is that we can expect a 10-15% boost there. Obviously it will vary for different games, but that seems reasonable.

If those things hold true, AMD can totally mop up when 64-bit WindowsXP arrives. I mean, they'll probably have an FX-53 by then running at 2.4GHz, the motherboard stuff will be tweaked and tuned in their second revisions (maybe even with PCI Express and all that), and then you stack some very significant performance gains on top of that...

The big catch is the whole compatibility issue. As was mentioned here, you need 64-bit drivers, which is no problem for video and sound cards, and a total crapshoot for stuff like printers. In addition to the drivers, it's really up in the air how compatible the 32-bit windows apps will be on 64-bit (should be very good) and how fast they'll run (really hard to tell).

The whole Athlon 64 story is far from over, it's really just beginning. There are lots of future questions that still need answering.

I'm very curious to know if unbuffered memory would make a significant difference in the FX systems, when they do that revision next spring. With an on-die memory controller, the latency differences and the extra buffering cycle you save could make a very real performance difference, not just some 2% nudge on a chart. Or, it could be not very different at all.

Hmm.... hey Loyd, you might want to intentionally turn up the latencies (make them worse) on the regular Athlon64 system you have and see what it does to performance.

[Jason Cross]

gah, triple post!

[Jason Cross]

wow...double post.

[XPav]

That's the third triple post I've seen.

Qt3 hits a thousand users, everything goes to hell. Imminent death of Qt3 predicted. Film at 11.

[Machfive]

Ladies and Germs, I would like to present my esteemed colleague, the one, the only, Mercenary For Hire!

Take it away, M4H!

[MercenaryForHire]

Shit, where to start?

Firstly, according to the majority of reviews posted, the A64 is the winner. Feel free to check for yourself. Also note THG using the MMX iDCT on the Athlons instead of the SSE2 like they should.

http://firingsquad.gamers.com/hardware/athlon_fx_preview/
http://www.hothardware.com/hh_files/CCAM/a64fx_51_launch.shtml
http://www.lostcircuits.com/cpu/amd_a64fx51/
http://www.extremetech.com/article2/0,3973,1276955,00.asp
http://www.gamersdepot.com/hardware/cpus/amd/athlon64/2.2fx/001.htm
http://www.digit-life.com/articles2/athlon64fx-athlon64/index.html
http://www.athlonxp.com/modules.php?op=modload&name=Diner_Wrapper&file=index&req=ShowFile&file_wrap=html/reviews_amd_athlon64.html
http://www.tomshardware.com/cpu/20030923/index.html
http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD02MjU=
http://www.amdmb.com/article-display.php?ArticleID=259
http://www.amdmb.com/article-display.php?ArticleID=260
http://www.amdzone.com/articleview.cfm?ArticleID=1344
http://www.aceshardware.com/read.jsp?id=60000253
http://www.tech-report.com/reviews/2003q3/athlon64/index.x?pg=1
http://www.hardocp.com/article.html?art=NTI0
http://www.tecchannel.com/client/desktop/440/index.html
http://www.anandtech.com/cpu/showdoc.html?i=1884
http://www.legitreviews.com/Reviews/64fx_1.shtml

Which of course, leads to the next point - the declaration that the P4 is the winner, because Tom Said So (TM), and the ensuing debate which unfolded. Without further ado, I give you:

Tom's Hardware Bribe
A six-episode miniseries

Keep Trollin, Trollin, Trollin ...
Firstly, let's get to know Thomas Pabst. He's the guy who trolled, spammed, and flamed without abandon a large number of users at the Anandtech forums. Yes, the one who, based on a single trip to America, declared it a hellhole fit only for the scum of the earth. I'd love to give a username - but there were just so many of them. After all, he was banned, reregistered, banned again, reregistered again ... you get the idea. He's the poster-boy for the IP block.

Intel Is Teh Sux0r
We all remember the Intel 1.13GHz SECC, right? Well, maybe not, since it never lasted on the retail market. Here's some links of our good friend Tom blowing his usual brand of "I Am The Computer God" smoke up our asses.

http://www6.tomshardware.com/cpu/20000731/index.html
http://www6.tomshardware.com/cpu/20000801/index.html
http://www6.tomshardware.com/cpu/200008281/index.html

And off Mr. Pabst goes to bend over for product samples at AMD's headquarters. My, that Athlon sure has a "powerful floating point unit", doesn't it? Oh, what's this? A bag of money and a tube of Astroglide? You shouldn't have!

Like a Hooker going back to the Pimp
But what's this "Tualatin" what is this you speak of?

http://www.tomshardware.com/cpu/20010612/index.html

Somehow one of those new Pentium III (Tualatin or Coppermine-T?) processors with the famous clock speed 1.13 GHz happened to fall in our lap. We think that for historical reason it should be us and nobody else, who tests Intel's second attempt to manufacture a Pentium III at this clock speed, so we gave the still 'secret' chip a good test run. Now we understand why Intel doesn't know if it should release it or not. Once overclocked to more than 1.4 GHz it is able to threaten Intel's fastest Pentium 4 processors.

No shit you feel it should be only you who should test it - you're the only one who sucked down a big load of corporate jism to get the ES! Also, note the "P4 core sucks" undertones.

RAMBUS Blows
And who can forget the giant shit he took on Rambus during that whole legal fiasco? Just in case you did:

http://www.tomshardware.com/blurb/20000525/index-01.html#how_much_faith_would_you_have

Buying Rambus is not exactly 'buying the future', but making sure that the coffers of a lot of people with very questionable motives are getting filled with your hard earned money. Do you want to support that?

Wow, RDRAM must be the Tool of the Devil! And speaking of the Tool of the Devil, look what the evil little prick says about it more recently!

http://www.tomshardware.com/motherboard/20020501/ddr400vsrambus-10.html

At the moment, the price for Rambus memory is quite inexpensive, so it's worthwhile to consider buying or putting together the corresponding systems. Rambus still has a right to claim a share of the market, which is evidenced by the fact that numerous motherboard manufacturers, such as Asus, Aopen, Biostar, MSI and Iwill, will soon be coming out with boards that support RDRAM.

But Tom, isn't that Filling The Coffers of Corporate Bastards? And Intel - I thought their P4 architechture was terrible! Oh, right - those are the same corporate bastards you happily drop trou for to get your product samples!

Tom's Plagarism Guide
<Satan>
So, Thomas ... says here you want to be a high-ranking reviewer, but yet you can't write your own reviews ... I think something can be arranged.
*zip*
Suck Satan's Cock! GGRRRRROOOOOOOOOOOAAAAAAAANN
I will give you articles by Van Smith! MMOOOOOOOOOOAAAAAAAN
You can take the content and alter the byline! OOOOOHHHHHHHHAAAAAAAAARGGGGGGGH
You will get thousands of web hits daily! ROOOOAAAAAAAAAARRRRR
People will take your word as gold - your opinions the truth OOOOOOAAAAAAAAAAAAAAAAAAAAAAAGGGGGGGGGGGGGHHHHHHHH HH
*breathes in postorgasmic bliss*
Send in Omid Rahmat on your way out!

http://www.vanshardware.com/articles/2002/03/020317_THG_Slander/020317_THG_Slander.htm

New Benchmark - Quickest Doctoring of a P4
OMG, a new P4 core? Running at 3.3GHz? OMG, Tom, you must have had more cock than a drunken sorority girl to get your hands on that!

http://www.tomshardware.com/cpu/20020909/index.html

...

Wait, the new P4s are running on an 800MHz FSB. What the hell is this crap?

Oh, I see. You Photoshopped a fucking P4 CPU core to make people think you had a fresh chip. Real professional there. Shit, you must have had some help from your little cabana-boy Omid on that one. Sheer genius, I swear to god.

And you're busted. Quick, pull the image down! Don't let anyone save it! They'll be able to prove it's chopped!

Too late, fucksaw.
http://www.m4h.org/thg_p4.jpg
http://www.m4h.org/fake_p4.jpg

FIN

Roll credits.

- M4H

[Peter Frazier]

Too many attempts at being zany here for me to understand what the fuck is going on. Can I get the English version please?

[Machfive]

For those who feel like testing out the image manipulation in Photoshop, set the copied "3" layer to "Difference," instead of subtract as shown in that last image.

What you're seeing with the black block is the area in which the pixels are identical between the two threes. Even on something mechanically stamped like this, varying jpg compression levels and natural differences in the texture of the object will render at least a slight difference to the second 3, making the difference of the two a dark, but not totally black, square.

What you're seeing with the black square is all of the pixels that are absolutely identical.

What's funny is a skilled graphics designer could have easily photoshopped one that could have withstood much more scrutiny, with a mere 5 minutes of work.

Here's 5 minutes of my own work.

http://www.themach5archive.com/misc/tomscheats.jpg

The top image is a zoomed-in portion of the original image, *STILL* posted on Tom's site.

The second image depicts the lefthand "3" being copied, having its layer mode set to "difference", and moved over on top of the righthand "3".

The third image is a doctor job I did on the version posted on Tom's site. I could have done a better one, but I didn't have access to the source image of the ORIGINAL processor that Tom and/or his lackeys had doctored.

And the final image is me repeating step two. Notice how there's no black box? That's how something should look when you render a difference.

In conclusion, I do not trust Tom or anything his site publishes.

Thanks, and have a good night!

[MercenaryForHire]

Too many attempts at being zany here for me to understand what the fuck is going on. Can I get the English version please?

You got the English version above when MachFive initially mentioned that THG was untrustworthy.

- M4H

[Bub, Andrew]

Why do I suddenly have the urge to post pictures of dinosaurs?

[MercenaryForHire]

Why do I suddenly have the urge to post pictures of dinosaurs?

Go nuts. :)

- M4H