A64 vs Core 2: 64-Bit Performance

I have seen it said over, and over, and over that K8 is faster than Conroe at running 64-bit code. Rather than inserting myself into flamewars to try and make my point, I'm making a new thread for discussion of this particular topic.

Hopefully, this thread can be kept civil. If you don't have numbers, and instead you have some kind of anecdotal observation like "OMG U N00BI3 F@NBOI AMD PWNS INTEL IN X64," or "RIGHT ON MAN U SHOW THOSE AMD F@NBOYZ," please refrain from posting in this thread.

Here are some articles I found:

http://www.xbitlabs.com/articles/cpu/display/core2duo-64bit_4.html

Tally:
AMD: 0
Intel: 8
Draw: 1

The draw here comes from Sciencemark. Intel wins the Primordia test, while AMD wins the Molecular Dynamics test. If it bugs you, add one to both Intel and AMD, for a final score of Intel 9, AMD 1.

http://techreport.com/reviews/2006q3/core2/index.x?pg=3

Tally:
AMD: 0
Intel: 6
Draw: 1

The draw here comes from the POV-Ray test. With one thread, Intel wins. With two, AMD wins. Again, if this really bugs you, you can count them as seprate tests and add one to both scores for AMD 1, Intel 7.

I'm going to tentatively include this Woodcrest review from Anandtech:
http://www.anandtech.com/IT/showdoc.aspx?i=2793&p=5

I couldn't find anything explicitly stating that the benchmark was 64-bit. However, the two benchmarks are testing a 64-bit SQL database running on a 64-bit operating system, so I'm including it.

Intel: 2
AMD: 0

The only relevant link I could get from all these posts proclaiming AMD to be t3h w1nnrar at 64-bit is this:

http://forums.pcper.com/showthread.php?t=425117

It's a link to PC Perspective forums, where one guy says Woodcrest 64-bit performance sucks, and offers no proof to back up his statements. It's hardly damning evidence for my case.

One more point I'd like to make is that this should not be a discussion about how much K8 gains when going from 32-bit code to 64-bit code, or how much Conroe gains when going from 32-bit code to 64-bit code. To me, this is just like people basing arguments on 3DMark scores when those scores have no bearing on real-world experience. People don't play 3DMark. Nor do people play UT2k4 x86, and then reboot and play UT2k4 x64 on a daily basis just to enjoy the difference between them. Nobody plays Jump To 64-bit. Be it a gamer with a desktop or a corporation with a server, they don't buy it based on which platform will experience a larger leap when they switch from x86 to x64. They buy it based on whether it is faster at x86 code, or at x64 code, or both. If price or any other factors are not considered, and Conroe's x86 and x64 performance were identical, it would be the better buy if and only if it was faster than the competition.

So, with that said, please feel free to show me numbers showing that K8 is indeed better at executing 64-bit code than Conroe. Otherwise, I would ask that people stop spreading FUD by accepting that as fact without any proof whatsoever.

LstOfTheBrunnenG said:

So, with that said, please feel free to show me numbers showing that K8 is indeed better at executing 64-bit code than Conroe. Otherwise, I would ask that people stop spreading FUD by accepting that as fact without any proof whatsoever.

Click to expand...

In order to do this, you have to stop looking at the absolute performance result, but at the changes in respect to the ratio of performance per processor when compairing 32bit to 64bit. The result of majority of tests show that the AMD processor does garner a performance improvement when running 64bit software in comparison to 32bit software.

At the same time, when comparing 32bit to 64bit peformance on the Core 2 Duo, you'll see that there isn't much performance improvement. It's more like stating, Core 2 Duo will run your 64bit code, albiet at the same performance level as 32bit code. This is because of the lackluster copy of the AMD 64 bit instruction set in EM64T. Intel didn't want to completely subvert the marketing of the Itanium 2 processor and a significantly better 64bit instruction set could have threatened the highend server market, higher than the Xeon market. Of course their decisions may have changed in the interim, but EM64T has had no additional revisions yet.

Now granted, Core 2 Duo, will still out peform the AMD 64 in raw performance, so all that happens in 64bit code is that the AMD gains some ground but is still surpassed by the Core 2 Duo, at this time.

The point is, AMD still has lots of headroom for improving their K8 cores. For example, they typically run at slower clock speeds than the top end Core 2 Duo processors and AMD is also looking at an L3 cache that is shared by all the cores in K8L. Granted Intel is also working on a revision to their core architecure to provide better Quad-core support in the long term. Their current Quad-core design has two separate Core 2 Duo processors in one ceramic package and uses the FSB when communicating between the two processor die in the package, similar to the lackluster Pentium D series.

I think that you'll see performance crowns change more frequently now that Intel has finally migrated from the Netburst architecture and that some design changes (say the AMD 64 K8L) may garner performance crowns, it'll only be until the next processor release. I think we are back to the days of the Athlon XP series, where the performance crown was handed back and forth with each speed stepping release from each perspective CPU maker. The only exception again is the EM64T instruction set in comparison to the AMD 64 instruction set. If Intel doesn't extend it to the full AMD 64 set and then some, we may still see longer trends of AMD performance crowns in the future once they address the benefit that 4MB cache provides the Core 2 Duo architecture.

Here's some links to backup my suppostition:

64bit and 32bit scores from this review:
http://www.planetx64.com/index.php?...sk=view&id=283&Itemid=14&limit=1&limitstart=5

What's important is to look at the actual numbers and in comparison to the 32bit versus 64bit. Note that AMD has larger increases in performance in 64bit in comparison to their 32bit scores. Again, this improvement is not enough to dethrone the Conroe King, but does help to illustrate my point.

First off, I have no problem with AMD regaining the performance crown. I hope they do. It's healthy for the market. What bugs me is that right now certain people on this forum are constantly repeating themselves by saying that K8 is currently faster than Conroe at 64-bit code, and this simply isn't true.

As for the whole gain argument, I suggest you read my 2nd last paragraph in my original post, as I anticipated this argument. It's true that AMD64 may be superior and the gains may be a portent of an upcoming transfer of the performance crown, but that doesn't matter until K8L or another AMD next-generation chip actually comes out.

LstOfTheBrunnenG said:

First off, I have no problem with AMD regaining the performance crown. I hope they do. It's healthy for the market. What bugs me is that right now certain people on this forum are constantly repeating themselves by saying that K8 is currently faster than Conroe at 64-bit code, and this simply isn't true.

Click to expand...

Nor did I state that above, and I agree with you that stating K8 peformance bench scores in 64bit "currently" being higher than Conroe is patently false.

LstOfTheBrunnenG said:

As for the whole gain argument, I suggest you read my 2nd last paragraph in my original post, as I anticipated this argument. It's true that AMD64 may be superior and the gains may be a portent of an upcoming transfer of the performance crown, but that doesn't matter until K8L or another AMD next-generation chip actually comes out.

Click to expand...

Actually it does matter. Conroe has the peformance crown currently, it's important to note there isn't much additional advantage to move to 64bit code (although it runs it great and is still the fastest). But AMD 64 owners will be able to garner some additional performance improvements that narrow the performance gap. Conroe's still King, but AMD 64 owners "without purchasing anything new" will be able to get a boost from Vista, again I state not to put them in the lead but it helps.

yes, conroe is faster, but k8 catches up a smidge in 64bit... but is still behind. to me, nothing new here

what i really want to see are some 4way and 8way server benches

@ HighTest:

Actually, the difference in the gains isn't as bad as in the Pentium 4 days.

http://www.xbitlabs.com/articles/cpu/display/core2duo-64bit_7.html

But again, this is off-topic. The point is the false belief that current K8 parts are faster than current Conroe parts at x64. Conroe having a slightly smaller gain != Conroe sucking at 64-bit code. I know you didn't say that, but many persist in saying it.

I think in the end this doesn't matter as they both perform well with 64 bit code, neither is horribly handicapped.

In the end P4 was the biggest crap known to man and i feel bad for all the idiots that even wasted their time/money on it. And yes conroe is a HUGE improvement but i guess thats what you get for getting "lucky". And thanks to that "luck" i got my dual core for a cheap price and i'm sticking it out with AMD because i'm not a sell out!

Tazman2 said:

In the end P4 was the biggest crap known to man and i feel bad for all the idiots that even wasted their time/money on it. And yes conroe is a HUGE improvement but i guess thats what you get for getting "lucky". And thanks to that "luck" i got my dual core for a cheap price and i'm sticking it out with AMD because i'm not a sell out!

Click to expand...

You must never have owned a P4C Northwood. They were hardly the "biggest crap known to man" nor do I feel like an idiot for "wasting" my time/money on it. A little more costly than the AXP's, but with Hyperthreading made for a pleasant computer experience.

kirbyrj said:

You must never have owned a P4C Northwood. They were hardly the "biggest crap known to man" nor do I feel like an idiot for "wasting" my time/money on it. A little more costly than the AXP's, but with Hyperthreading made for a pleasant computer experience.

Click to expand...

All right guys, let's not go into Northwood vs K7 and Prescot vs K8. Everyone knows that Northwood > K7 and K8 > Prescot, there's no need to ruin a good thread with rehashing the same arument over, and over, and over, and over.

I must say, I honestly that it doesn't matter. Yes, AMD has a higher % increase relative to Intel, but that will not benefit them if they cannot increase the clock speed of their current 90nm process. A larger percent of a smaller number is still smaller than a smaller percent of a larger number. I am still of the opinion that 64-bit desktops are a long-way off, and at this point is nothing more than a marketing gimick for both CPU camps.

Mr. Brunnen, you are soo wrong. Screw all them fancy review sites:
http://www.amdzone.com/modules.php?...=article&sid=6226&mode=thread&order=0&thold=0

says clearly:

A friend of mine built a 16GB woodcrest and we benched it against the rev f. We found that if we ran the tests with 2 or 4 GB of ram, the woodcrest outperformed the rev F, by less than 2%. When we added video tests, 16gb of ram, and some general openGl games, the rev F was ahead by almost 11% in doom3, quake4, prey and Painkiller BOH. His build had dual pci-e slots but we don't think they go to 16x, and we could not get the 7950s to ever get recognized for Quad SLI. Using the 91.47 driver or the 91.31 driver.

Click to expand...

yeah woodcrests SUCKS at 64 bit...
http://www.amdzone.com/index.php?na...6&sid=e16b1ca9d35a51bc70567090f65463fe#113256
</sarcasm>

p.s.: note that 2/3 of the OP's linked posts are from Serge84...

InorganicMatter said:

Yes, AMD has a higher % increase relative to Intel, but that will not benefit them if they cannot increase the clock speed of their current 90nm process.

Click to expand...

True 'nough. Although I'm not concerned to much with how much more performance that AMD can squeeze from the 90nm process. They already have some engineering samples of their 65nm process, so I'm expecting AM2 65nm cores to silently ship even before K8L.

A larger percent of a smaller number is still smaller than a smaller percent of a larger number. I am still of the opinion that 64-bit desktops are a long-way off, and at this point is nothing more than a marketing gimick for both CPU camps.

Click to expand...

Sure, but my 85% of your 105% is larger under 64bit then lets say my 70% of your 100% at 32bit. Your still the winner, but the margin's smaller.

I run x64 as my primary OS. Now that Plustek has finally released my scanner driver, the only thing I'm going without is my really old ATi TV tuner. Other than that, it's been solid as a rock.

I run it because I have 4 GB of RAM, so I suppose you could say it's out of necessity. To be honest, I never actually wanted 4 GB, but once I dumped my DFI board and I found out the 2 GB I thought was bad actually was quite stable, I just couldn't pass that up.

drizzt81 said:

Mr. Brunnen, you are soo wrong. Screw all them fancy review sites:
http://www.amdzone.com/modules.php?...=article&sid=6226&mode=thread&order=0&thold=0

says clearly:


yeah woodcrests SUCKS at 64 bit...
http://www.amdzone.com/index.php?na...6&sid=e16b1ca9d35a51bc70567090f65463fe#113256
</sarcasm>

p.s.: note that 2/3 of the OP's linked posts are from Serge84...

Click to expand...

I'd read that the primary advantage is the 4MB cache and that situations that go beyond the ability of the 4MB cache will show the strengths of the AMD 64 platform. For example Tom Yager in the article "AMD's Locked and Loaded" on page 18 of the August 28th Infoworld magazine makes this point in terms of the Opteron "F" versus the new Xeon. http://www.infoworld.com/article/06/08/23/35OPcurve_1.html


In an indirectly related link, Intel's laying off a bunch. Is this in preparation of a price slashing war since they know the same info and it'll be shown soon enough?
http://www.infoworld.com/article/06/09/05/HNintelswingsaxe_1.html

What's important is that these types of information are what CIO's and others view. IT tech journals that still spin a positive light on AMD will promote Opteron server sales until K8L comes out, in my opinion.

Yeah, this has been sprouted for some time. AMD fans keep saying C2D sucks in 64bit mode, when in fact the overall performance in 64bit is still higher on C2D.

Sure, the gains to 64bit are smaller, no one is disputing that. But at the end of the day, it's still faster - bottom line.

Analogy time -
If Car A is 2 secs faster than Car B per lap on wet tracks, and 1 sec faster than Car B on dry tracks - it's still the faster car isn't it?

To summarize:

A well clocked C2D is likely 35-40% faster (C2D's clock very well) or more than a similar and well clocked A64 X2. So even if the X2 is 5% faster at 64 bit code, it is still 30% or more behind.

This may all change when we start talking quad cores and the power of HT links, or it may not. The market would benefit if AMD would leap again and at least be even.

HighTest said:

I'd read that the primary advantage is the 4MB cache and that situations that go beyond the ability of the 4MB cache will show the strengths of the AMD 64 platform. For example Tom Yager in the article "AMD's Locked and Loaded" on page 18 of the August 28th Infoworld magazine makes this point in terms of the Opteron "F" versus the new Xeon. http://www.infoworld.com/article/06/08/23/35OPcurve_1.html


In an indirectly related link, Intel's laying off a bunch. Is this in preparation of a price slashing war since they know the same info and it'll be shown soon enough?
http://www.infoworld.com/article/06/09/05/HNintelswingsaxe_1.html

What's important is that these types of information are what CIO's and others view. IT tech journals that still spin a positive light on AMD will promote Opteron server sales until K8L comes out, in my opinion.

Click to expand...

I'm getting tired of the "K8L will dominate everything!....next year!" argument. The people making that argument blindly have faith that K8L will own up the market without any numbers whatsoever. They're also probably the same ones who cried foul when Intel released early benchmarks of Conroe, and who firmly believed they were made-up until just about every tech site proved them wrong.

I remember when Intel f@nboys cried "Wait for prescott! It will own!" as the answer to the A64. We all know how that one ended. What if K8L is the same thing? What if a technology AMD implements in K8L turns out to be more of a problem than it first appeared, and ends up hampering performance, ala Prescott? Scary thought. I certainly hope not though. Like I said above, I hope AMD regains the performance crown. Competition is good.

drizzt81 said:

p.s.: note that 2/3 of the OP's linked posts are from Serge84...

Click to expand...

Actually, make that 3/4. He should just put it into his sig...

LOL @ Serge84 - the guy cracks me up. I just read his rants for entertainment purposes now, he is clearly a deluded individual.

Almost as entertaining as reading Sharikou's 'Intel doomsday' blog

LstOfTheBrunnenG, to be fair I honestly doubt AMD will make as big a mistake as Prescott. They may be having slight problems getting voltages down on 65nm (if you believe TheINQ that is... big IF) but hopefully it will be fixed in due course. We are still ~9 months away from K8L or course...

LstOfTheBrunnenG said:

I'm getting tired of the "K8L will dominate everything!....next year!" argument. The people making that argument blindly have faith that K8L will own up the market without any numbers whatsoever. They're also probably the same ones who cried foul when Intel released early benchmarks of Conroe, and who firmly believed they were made-up until just about every tech site proved them wrong.

I remember when Intel f@nboys cried "Wait for prescott! It will own!" as the answer to the A64. We all know how that one ended. What if K8L is the same thing? What if a technology AMD implements in K8L turns out to be more of a problem than it first appeared, and ends up hampering performance, ala Prescott? Scary thought. I certainly hope not though. Like I said above, I hope AMD regains the performance crown. Competition is good.



Actually, make that 3/4. He should just put it into his sig...

Click to expand...

It's actually a "little" bit more long winded then that..... First it was Cppermine... "wait for coppermine? It'll save the day! Then it was Willy... Wait for Willy! It'll save the day!!... Then it was Wait for Prescott!!... It'll save the day!!.... Then it was Yonah.... Wait for Yonah!! It'll save the day!!.... Then it was Conroe.... Wait for Conroe!!! It'll save the day!!!

Nothing new here folks... Move along... Just a bunch of Intel fans waiting for the savior.... Again...

This is because of the lackluster copy of the AMD 64 bit instruction set in EM64T. Intel didn't want to completely subvert the marketing of the Itanium 2 processor and a significantly better 64bit instruction set could have threatened the highend server market, higher than the Xeon market.

Click to expand...

This is what I have said from the beginning... before conroe came out, this was also my prediction.

I always had said that intel was going to keep EMT64 crippled as long as they continued development of itanium. I always got reamed by the intel and conroe !!!!!!s but this is just how big buisness is run.

The example I have always used was the westerndigital Raptor drive. Westerndigital does not have a line of SCSI drives anylonger, so they can afford to produce the raptor drive. The raptor does not have a competitor because releasing one would tread on the market of their SCSI lines.

You see things like this all over large buisness. Here is another local example: Gamers are always complaining about not being able to use SLI or crossfire with multi monitor setups... but Nvidia and ATI allow this setup without a problem when using quadro or fireGL cards. They will never allow multi monitor SLI on the gamer cards because that would be treading on the market of the quadro and fireGL. You also will not see specviewperf numbers on gamer cards increasing at a rate consistant with game performance. Nvidia caps that performance in the same way intel caps their 64bit performance on consumer chips.

AMD does not have a totally seperate server arch, and they can afford to offer the best 64bit performance on all chips.

It's just big buisness, but it is also why I would choose AMD64 over woodcrest, and it's also why you see many of the big integrators (eg SUN) moving towards AMD, even after the introduction of woodcrest.

Yashu,

You've got it! I think you may have stated my point more clearly than I was able to.

Thanks!

Yashu said:

You also will not see specviewperf numbers on gamer cards increasing at a rate consistant with game performance. Nvidia caps that performance in the same way intel caps their 64bit performance on consumer chips.

AMD does not have a totally seperate server arch, and they can afford to offer the best 64bit performance on all chips.

It's just big buisness, but it is also why I would choose AMD64 over woodcrest, and it's also why you see many of the big integrators (eg SUN) moving towards AMD, even after the introduction of woodcrest.

Click to expand...

Do you have any proof of this? The OP showed quite a few links, outlinking that intel's 64-bit performance is up-to-par or exceeds AMD's 64 bit performance. All you say is "intel caped it's 64-bit performance", though it is clear that there are quite a few improvements when going from 32 to 64 bit.

Since you must not be an engineer, let me explain the concept of a 'tade off' to you. As with everything in the world, resources are limited. In order to do pretty much anything, resources are required. From the outcome that Conroe outperforms A64's in most tasks clock-for-clock one can take a good guess that Intel set out to create a new architecture to beat AMD with. Given the current market, which operates mostly on 32-bit, Intel must at some point have decided that a chip with excellent 32-bit performance is their primary goal, since there are very few 64-bit applications that are widespread. As such, a lot of resource were used for that. These resources were not available to create incredibly stellar 64-bit performance. Having a mediocre performing 32-bit chip with huge gains form 64-bit would be a rather poor tradeoff, since about 95% of the market would have no benefit from this. By designing a chip that has stellar 32-bit performance and still outperforms the competition on 64-bit (though by a slightly smaller margin) Intel made a financially sound decision.

Unfortunately, you are one of the five people in this world that run exclusively 64-bit apps and would rather have a chip, where performance from 32 -> 64 bit quadruples. While I am sorry that Intel has disappointed you, if you consider the performance advantage that intel has regardless of the binary format, I am having a hard time understanding your argument: Do you want the fastest chip possible, or not?

I think the point is AMD's K8 sees a larger boost from 32-bit to 64-bit, because Core is already faster to begin with. Therefore, if it performs on part with 32-bit operation, then there's really not much to gain because it already outruns K8 anyhow. If the scaling (of 32-bit vs 64-bit) continues on the AMD front, Intel will have some work to do...which is good.

Competition = Faster stuff, but not necessarily at lower prices

Sovereign said:

I think the point is AMD's K8 sees a larger boost from 32-bit to 64-bit, because Core is already faster to begin with. Therefore, if it performs on part with 32-bit operation, then there's really not much to gain because it already outruns K8 anyhow. If the scaling (of 32-bit vs 64-bit) continues on the AMD front, Intel will have some work to do...which is good.

Competition = Faster stuff, but not necessarily at lower prices

Click to expand...

No, the point is this: "it is [...] why I would choose AMD64 over woodcrest" for an invalid reason. Regardless of the applications binary format (32/ 64 bit) Conroe is faster. Unless AMD can offer a better price/ performance ratio, the other posters decision is irrational. One good reason to make irrational decisions is fan-b0i-ism

LstOfTheBrunnenG said:

I'm getting tired of the "K8L will dominate everything!....

What if K8L is the same thing? What if a technology AMD implements in K8L turns out to be more of a problem than it first appeared, and ends up hampering performance, ala Prescott? Scary thought. I certainly hope not though.

Click to expand...

To begin with all the Intel Fan boys still to this day are on the "Conroe dominates all and always will!" band wagon so whats the big deal?! And as far as K8L becoming prescott in the sense of screw ups i personally am pretty sure they aren't like intel where they will push a crappy product for years just to make $$$ off of the idiots who saw 3.0ghz being higher then 2.5ghz so it must be faster and buying it. AMD has been making their products faster and faster even before Intel could find where they left their keys. They got lucky with Conroe and from what i recall it wasn't even developed in this country by US engineers since well they were the ones that thought up the whole P4 mishap!

Sovereign said:

I think the point is AMD's K8 sees a larger boost from 32-bit to 64-bit, because Core is already faster to begin with. Therefore, if it performs on part with 32-bit operation, then there's really not much to gain because it already outruns K8 anyhow. If the scaling (of 32-bit vs 64-bit) continues on the AMD front, Intel will have some work to do...which is good.

Competition = Faster stuff, but not necessarily at lower prices

Click to expand...

Here's where you grasped the discussion trend and also where drizzt81 missed it. At no time did I try to state that AMD 64 processors that are currently shipping pwn Core 2 Duo in 64bit applications. What I did state is that there is a ratio of improvement when moving from 32bit to 64bit code that illustrates that AMD has a higher ratio of improvement. This ratio doesn't state AMD 64's currently shipping outperform the "capped" (drizzt81's term, not mine) Core 2 Duo. Core 2 Duo did show some marginal ratio of improvement that when taken in terms of its raw performance still makes it the current King.

As drizzt81 stated, there are a limited number of resources, this is correct. Due to the number of die intel want's to product on a 200x200 or 300x300 wafer, they need to have only a specific number of transistors and interconnects per die. Since they focused on having 4MB of cache, this would have limited additions to the instruction set and changes to the pipeline that would have been required. Again, probably an astute choice as the current configuration does allow Core 2 Duo to reign supreme against current contenders. AMD 64 also has some resource restrictions as well, some of these will improve when they transition to 65nm process (of which I understand there are some engineering samples, so this should be proceeding as planned). One topic of discussion is the inclusion of a larger "shared" cache in the next major revision of the AMD besides the core shrinks that will be occuring. Will this cache have the same impact as the cache on the Core 2 Duo? That's a question that will need to be answered in the future.

Granted, the fact that there is a better ratio of change in regards to the slower AMD CPUs is still something that we AMD fans cling to as a reason to stick to our procs. Is this logical, does it matter? If it makes you happy think whatever you like, we'll be doing the same and hope that predictions of an AMD return to supremacy will be proven.

Tazman2 said:

To begin with all the Intel Fan boys still to this day are on the "Conroe dominates all and always will!" band wagon so whats the big deal?! And as far as K8L becoming prescott in the sense of screw ups i personally am pretty sure they aren't like intel where they will push a crappy product for years just to make $$$ off of the idiots who saw 3.0ghz being higher then 2.5ghz so it must be faster and buying it. AMD has been making their products faster and faster even before Intel could find where they left their keys. They got lucky with Conroe and from what i recall it wasn't even developed in this country by US engineers since well they were the ones that thought up the whole P4 mishap!

Click to expand...

Nobody's saying Conroe will always dominate.

The typical AMD vs Intel argument goes like this:

1. Someone posts in the AMD forum detailing a new build using an A64.
2. A few people suggest Conroe.
3. Someone jumps on the people from 2. and calls them Intel f@nboys.
4. People from 2. basically say "Um, Intel is faster at the moment, so it would be the smarter buy."
5. Serge comes in and says that Conroe sucks at 64-bit. Nobody listens.
6. AMD f@nboys retort "Well, K8L will be faster! AMD will once again be the king! You may have won the battle, but not the war!" etc, etc...

From here, it should be:
7. Everyone else says "So? This thread isn't about K8L, or what will be around in nine months. It's about this guy's new build, which he's going to buy tomorrow."

But instead it descends into useless and unsubstantiated arguing about products that don't even have leaked numbers out yet.

Much the same thing has happened in this thread. People are now arguing about K8L's yet-to-be-seen 64-bit capabilities and how Intel pushing Netburst for years was teh evil.

So?

This thread is about the 64-bit performance of current K8 processors vs. the 64-bit performance of current Intel processors.

Yashu said:

This is what I have said from the beginning... before conroe came out, this was also my prediction.

I always had said that intel was going to keep EMT64 crippled as long as they continued development of itanium. I always got reamed by the intel and conroe !!!!!!s but this is just how big buisness is run.

The example I have always used was the westerndigital Raptor drive. Westerndigital does not have a line of SCSI drives anylonger, so they can afford to produce the raptor drive. The raptor does not have a competitor because releasing one would tread on the market of their SCSI lines.

You see things like this all over large buisness. Here is another local example: Gamers are always complaining about not being able to use SLI or crossfire with multi monitor setups... but Nvidia and ATI allow this setup without a problem when using quadro or fireGL cards. They will never allow multi monitor SLI on the gamer cards because that would be treading on the market of the quadro and fireGL. You also will not see specviewperf numbers on gamer cards increasing at a rate consistant with game performance. Nvidia caps that performance in the same way intel caps their 64bit performance on consumer chips.

AMD does not have a totally seperate server arch, and they can afford to offer the best 64bit performance on all chips.

It's just big buisness, but it is also why I would choose AMD64 over woodcrest, and it's also why you see many of the big integrators (eg SUN) moving towards AMD, even after the introduction of woodcrest.

Click to expand...

The numbers say otherwise. Read my original post. If you're buying AMD64 just because it's AMD64 and not because it's faster, that's blind brand loyalty.

HighTest said:

Granted, the fact that there is a better ratio of change in regards to the slower AMD CPUs is still something that we AMD fans cling to as a reason to stick to our procs. Is this logical, does it matter? If it makes you happy think whatever you like, we'll be doing the same and hope that predictions of an AMD return to supremacy will be proven.

Click to expand...

My question to you is this: Why did you bring up the gain from jumping from 32-bit to 64-bit? I can see how it might be relevant to how future AMD chips might perform. I do not see how it is relevant to the topic at hand.

I don't need a reason to "cling" to my X2 setup. It's still a great setup, even if there's something faster out there now.

Granted, I hope AMD doesn't drop out of the market or go bankrupt. Beyond that, I care less that AMD will have teh fastestest chip when, years from now, I build a new rig, and more that there's a high-performing, reasonably priced chip to buy. Conroe isn't bad for AMD, it's good for the market and us as enthusiasts. Just look at these price cuts! When the only competition was the Pentium D, what incentive did AMD have to keep prices low? When K8L comes out, if it owns, it won't be bad for Intel as much it will be good for the market and us as enthusiasts. I wouldn't be cheering because AMD was king once again, I would be cheering because of this sweet chip I'd have the opportunity to buy.

This isn't an "INTEL OWNZ AMD U NOOBIES GTFO!!!!!" thread. I made this thread because certain members were spreading unsubstantiated rumors.

If I were building my first rig, and I came here asking for advice, I wouldn't know any better if Serge or Yashu came and crapped my thread saying that I should get AMD if I intend to use x64 because Intel "sucks" at it. Nor is anyone bothering to make them come up with proof of their points.

This is the way FUD spreads. In my OP, Yashu's "proof" that Intel sucks at 64-bit was another thread on another forum where some guy said that Intel sucks at 64-bit and nobody challenged him. Well, this is teh [H], and we don't let stuff like that go down here.

LstOfTheBrunnenG said:

My question to you is this: Why did you bring up the gain from jumping from 32-bit to 64-bit? I can see how it might be relevant to how future AMD chips might perform. I do not see how it is relevant to the topic at hand.

Click to expand...

I brought up the gain so as to illustrate that when running Windows XP 64bit edition or 64bit flavors of Vista that currently existing X2 owners will see a bump up in performance following the particular ratio. This means that they'll be somewhat less slower than their Core 2 Duo brethren than when in 32bit. Note that I acknowledge that the Core 2 Duo is currently the fastest on the market. Core 2 Duo users that switch to the said 64bit OS's above certainly won't perform any less than they perform currently, so they'll continue to be king.

As to the rest of your post, I agree with you. The only reason I had a rebuttal in the beginning, which upon reflection spiraled out of control to future speculation like you pointed out, was to show that there is information out there that shows AMD customers of existing equipment will see more of a performance improvement going to 64bit than customers ot Core 2 Duo. That doesn't preclude moving to 64bit on Core 2 Duo so that you can access larger arrays of memory and the other associated benefits, or that a Core 2 Duo would perform less than it's already stellar performance.

Ah well, guess it doesn't really matter much.

All I've got left to say now is:

In before the lock

Seems to early to making grandiose claims about 64bit performance.. The only thing im really concerned about is gaming performance and there just are not enough 64bit games, infact I don't think there are any true 64bit games so theres very little to base this on other then rather pointless synthetic benchmarks and software with very limited 64bit support or none at all..

By the time software catches up we will probably be using next generation cpu’s...Hopefully sooner but at the current rate probably not..

HighTest said:

In before the lock

Click to expand...

I'm not that pesimistic about this thread. The trolling seems to be the exception rather than the norm here, and people have generally ignored them.

Besides, Duby was here and posted something nonsensical and irrelevant, Yashu was here and posted something completely unsubstantiated, but Serge has not been here yet. This thread cannot die until we have a chance to hear what he has to say about all this.

let me explain the concept of a 'tade off' to you.

Click to expand...

You and I are almost on the same wavelength, but not quite. I am not an engineer, I am a designer, and as a designer I work with all sides of a project, not just engineering.

We both can agree that EMT64 is crippled compared to AMD64, and we both agree that it was a buisness decision.

I beleive that intel is using EMT64 as a cheap memory address extention. PAE mode was not the future. I also belive that intel built a desktop arch. first that can be used for servers. AMD built a server arch. first, that can be used on the desktop. That is the destinct difference between conroe and k8.

k8 competes with intel on two fronts... it competes with xeon and itanium at the same time. Intel is going to be very careful to make sure xeon and itanium to not compete with each other.

conroe is a fast 32bit chip... and I know intel has the r&d to have built a fucking awsome 64bit side of that... but they didn't... and what I have been trying to say for a long time now, is that the reasons behind why they didn't (regardless of it's relative speed vs k8) is a buisness/marketing related decision, not an engineering one.

To put it more simply:

You may run your 3d applications on a 7800gtx, however if you are serious about running MAX or MAYA, nvidia wants you to shell out the bucks for a quadro.

You may run your 64bit applications on conroe/woodcrest, but if you are serious about your 64bit research software or whatever, intel wants you to shell out the bucks for itanium2.

I think this is the wrong approach... partly because I beleive that 64bit applications are going to become more important outside of research... it already is on web serving, and those little blade servers that host those apps are built on consumer level hardware.

maybe my command of the English language is not up to par.

http://dictionary.reference.com/search?q=cripple
–verb (used with object)
7. to disable; impair; weaken.

So if Intel had crippled the performance, there would be some form of Conroe with better 64-bit performance that was then scrapped and replaced by a worse version.

So if Intel had crippled the performance, there would be some form of Conroe with better 64-bit performance that was then scrapped and replaced by a worse version.

Click to expand...

Capped is probably a better term for what I was trying to say... it's still not exactly endearing.

You never know what could have been on the designer's drawing board... In big buisness like this it is not uncommon for so many things to be dictated by marketing.

Yashu said:

We both can agree that EMT64 is crippled compared to AMD64, and we both agree that it was a buisness decision.

Click to expand...

If Intel is crippling the 64-bit performance of Conroe, I'd be excited to see what they could do if they actually wanted it to run fast.

Perhaps you again missed the first post in this thread. Allow me to provide a link for your convenience: http://www.hardforum.com/showpost.php?p=1029916395&postcount=1

In that post I show results from several reputable review websites that show that Conroe and Woodcrest perform better than the A64 and the Opteron in executing 64-bit code. I fail to see how Conroe's EMT64 could both be crippled compared to the A64's AMD64, and beat A64's AMD64 in benchmarks and real-world performance.

read my post again please. I already addressed that.

you deserved a big thanks for the effort, but we know that conroe is faster than K8. great tests.

LstOfTheBrunnenG said:

I fail to see how Conroe's EMT64 could both be crippled compared to the A64's AMD64, and beat A64's AMD64 in benchmarks and real-world performance.

Click to expand...

Because something can still be out-performed by something else that is crippled. Example: an agp 6800nu is crippled. It is reduced in capacity from its full potential. It nonetheless can outperform a Voodoo5 fully overclocked, full potential unleashed. I don't mean to imply that there is such a severe performance gap in this instance, simply illustrating that "crippled" does not neccessarily mean "slow," simply "slower than it would be otherwise." EMT64 is crippled by the fact that it needs to be able to compete with AMD64 while not outperforming Itanium. AMD has no comparable line of products above AMD64 and therefore has no such "cap."

Second example: an Olympic sprinter with a twisted ankle is still going to be able to outrun any of our pudgy butts, even on two good ankles.

However, at an enthusiast level, the fact that Intel's business model is set up such that their 64 bit focus is still almost exclusively Itanium -- which may hamper their growing EMT64 in the long term -- really doesn't affect the current situation of a Core2 setup being faster than an AM2, which I believe was the point of the original post.

So when trying to build the bigger/better/faster system right now, Intel is the way to go. Will that be the case tomorrow? Flat out no one knows. We're not there yet. To be honest, you'd be better served trying to water-proof your system rather than trying to future-proof it. At least folks can agree on what water will do to it.

XmagusX said:

Because something can still be out-performed by something else that is crippled. Example: an agp 6800nu is crippled. It is reduced in capacity from its full potential. It nonetheless can outperform a Voodoo5 fully overclocked, full potential unleashed.

Click to expand...

Going along with your analogy: Where is the Core 2 Ultra? The only reason a 6800nu is crippled was because there was also a speed binned 6800 Ultra. If your analogy is supposed to hold, then there would have to be a CPU that is able to outperform the Core 2 Extreme. I have yet to see any benchmarks to back that up.

Going along with your analogy: Where is the Core 2 Ultra?

Click to expand...

ok now you are arguing semantics... The analogy is not going to be 100% congruent... but yes core2's 64bit performance was capped by design. You know intel could have made a great x64 part... but they chose to do what they did with prescott instead... it's relative speed vs k8 is not what we are talkling about... we are talking about net gain from x32 to x64, and conroe and woodcrest are not as impressive in that respect.

Yashu said:

ok now you are arguing semantics... The analogy is not going to be 100% congruent... but yes core2's 64bit performance was capped by design.

Click to expand...

And you know this how? Do you work at intel? Did intel say so publically?

Yashu said:

You know intel could have made a great x64 part...

Click to expand...

They have THE BEST x64 part that I know of. Show me something better

Yashu said:

but they chose to do what they did with prescott instead... it's relative speed vs k8 is not what we are talkling about... we are talking about net gain from x32 to x64, and conroe and woodcrest are not as impressive in that respect.

Click to expand...

And what do you base you assertion on? If you now say "well, if you look at the A64, it gains more on average than the Core 2," I shall point out that this comparison is invalid.

From most everything I have read in recent history, the Core 2 and K8 are fundamentally different. Comparing the gains/ losses when moving from 32 to 64 bit code between the two architectures is like saying "nVidia's hardware is crippled because ATi's counterpart is faster in Half-Life 2." It just doesn't transfer like that, because the underlying implementation is most likely not even remotely similar.

So unless you either finally show us a benchmark that clearly shows that there is a processor out there that outperforms the top-end Core 2 processor consistently in x64 benchmarks and is publically available, or an Intel document that says "While we could easily have made the x64 of the Core 2 more powerful, even with our limited resources, we decided not to for marketing reasons", I have a hard time believing you.

I am done with this thread, since it's getting to the point that this forum was in ~3-4 months ago, where people were arguing about how unreleased product A was going to outperform unreleased product B based on the rumors about both products' features. I am not in the mood to keep asking for proof when all people are providing is speculation with a complete disregard for facts.