A weakness in Core 2's armor?

The original point of this thread was just to show that Conroe had relatively poor 64bit performance compared to 32bit. I don't see this as Carfax trashing Conroe. It was a simple double bar graph that showed Conroe had worse 64bit performance compared to 32bit, meanwhile the Athlon performed better in 64bit compared to 32bit. Right now, Conroe beats Athlons so badly, that any benchmark where it doesn't burn the FX62 stands out. Conroe is completely expected to beat the FX62 by a lot, but these Cinbench 64bit results show uncharacteristicly close performance, hinting that 64bit tests may be a relative weakness. Based on the original benchmark, in my mind the statement that Conroe has a weakness in 64bit is perfectly valid, and not flaming. Fortunately, other links have been posted showing that this is false. Conroe does not have poor 64bit performance, and as expected from other benchmark results, it beats the FX62 by a wide margin.

He have a point. C2D chip aren't "that nice" at 64bit (even if a $300 chip is as fast as a $1000 one)

But, who care about 64bit anyway ? Beside for bragging and looking "L337", there is absolutely no advantage for home user to go with 64 bit OS.

Quote:

---

Originally Posted by Carfax

If you actually read my first post, you'll see that I wasn't making a fuss over it, and that I even considered the discrepancy may be due to sub-optimized compilers/program.

I only got aggravated because I had to clarify my stance numerous times, aswell as my intentions.

Apparently, people think I posted this as some dig against Intel, which is completely and utterly ridiculous..

I have praised Core 2 as much as anyone else, but it is not a perfect cpu and it does have it's faults.

However, some people just can't seem to rise above their petty AMD vs Intel obsessions and look at the technical aspects.. :rolleyes:

---

And I'm suggesting that the complier/settings and or program is probably the sole reason for Core's 64 bit performance. Seriously, I don't believe Intel would use a poor implementation of x86-64 considering that's where alot of x86 OS'es and software are going very soon. (and no, i don't claim to be an expert on this) I just seriously don't believe Intel's engineers would purposely shoot themselves in the foot like that. And it wouldn't explain why some 64 bit apps actually improve in performance on Core 2. (at least, I can't imagine it being inferior to the P4's EMT64)

I don't know if your whole post is _also_ directed at me, but I don't view your thread posted as a "dig" against Intel. I'm old enough to tell a fanboy post from a genuine post. However, I don't see how this is news worthy at this time. Intel's NGMA Core 2 hasn't been out in the wild long enough to really tell. If there's a negative trend in 64 bit performance a year or so from now, you definately have a point, but as of now, it's too early to tell. I recall AMD64 also had simular growing pains as well.

*Ugh, the website's babelfish translation is awful, It can't really explain any of the testing methods behind those numbers.*

Quote:

---

Originally Posted by kurg

Well.. then I'm confused about your point. Your premise seems to be that every app should be faster when compiled for any given 64-bit CPU than the same app on the same CPU in 32-bit mode. Empirical evidence would seem to prove you wrong. This theory (every app is faster when compiled for 64-bit mode) would seem, at first, or every, glance to be wrong.

---

Ugh.. This is why I get so frustrated sometimes.. :confused:

I'm aware that 64-bit does not always result in a performance boost. However, x86-64 is more than just 64-bit.

Core 2 has access to twice as many GPRs in 64-bit mode, than it does in 32-bit mode.

If anything, this is what gives the K8 and Core 2 it's boost in 64-bit mode, and not 64-bitness in and of itself.

I posted this because I thought it would be of interest to some people why the K8 seemed to gain in 64-bit mode, whilst Conroe did not, despite the fact that like the K8, it has access to double the GPRs.

Quote:

---

So, either the world is wrong (Conroe, being the fastest x86 CPU proves you wrong) or you are wrong (your premise is simply incorrect).

Now. Tell me. Is Conroe not faster in pretty much every meaningful CPU benchmark than any AMD CPU (including 64-bit)? Or, dare I say, is perhaps your premise simply _wrong_?

I vote (durr) your premise is simply wrong.

---

You haven't read a thing I've posted if you've come to these conclusions.

I never stated Conroe wasn't the fastest x86 chip in the World.

Copy and paste where I stated this.. :rolleyes:

Quote:

---

Originally Posted by Eastcoasthandle

To anyone reading some of the garbage in this thread...all this propaganda servers to do is to cause you to doubt yourself. That is the major agenda to these self proclaim truths that are petty, inconclusive and stupid.

---

Do yourself a favor and stay out of my thread.

Obviously, your reading comprehension must be subpar to come to such conclusions..

Quote:

---

If you read the title of this thread it clearly gives away their ill willed, bias attempts of self indignation...

---

The title is a QUESTION you idiot! Did you not see the question mark? The title is not an assertion, but a question..

Like I said, stay out of this thread, and let the grown ups do the talking.. :mad:

Quote:

---

Originally Posted by thecoldanddark

wrong

http://www.techreport.com/reviews/20.../index.x?pg=13

---

What does your link prove? Those tests are 32-bit, and so they are useless in this discussion..

Quote:

---

Originally Posted by Khenglish

The original point of this thread was just to show that Conroe had relatively poor 64bit performance compared to 32bit. I don't see this as Carfax trashing Conroe. It was a simple double bar graph that showed Conroe had worse 64bit performance compared to 32bit, meanwhile the Athlon performed better in 64bit compared to 32bit. Right now, Conroe beats Athlons so badly, that any benchmark where it doesn't burn the FX62 stands out. Conroe is completely expected to beat the FX62 by a lot, but these Cinbench 64bit results show uncharacteristicly close performance, hinting that 64bit tests may be a relative weakness. Based on the original benchmark, in my mind the statement that Conroe has a weakness in 64bit is perfectly valid, and not flaming. Fortunately, other links have been posted showing that this is false. Conroe does not have poor 64bit performance, and as expected from other benchmark results, it beats the FX62 by a wide margin.

---

Wow, someone actually gets it!

Thanks Khengllish! :toast:

Quote:

---

Originally Posted by Pongi

And I'm suggesting that the complier/settings and or program is probably the sole reason for Core's 64 bit performance.

---

I hope that is the reason. There is a thread about this over at Aces aswell, and some people claim it could be that the 64-bit version is pointer heavy and as a result, is screwing with the prefetch units.

Quote:

---

Seriously, I don't believe Intel would use a poor implementation of x86-64 considering that's where alot of x86 OS'es and software are going very soon. (and no, i don't claim to be an expert on this) I just seriously don't believe Intel's engineers would purposely shoot themselves in the foot like that. And it wouldn't explain why some 64 bit apps actually improve in performance on Core 2. (at least, I can't imagine it being inferior to the P4's EMT64)

---

Well, AMD pioneered x86-64 lets not forget, and EM64T was widely criticized as being subpar compareds to AMD64 when it was first released with the P4.

I would think that Intel would have solved whatever issues they had with EM64T for Core 2.

The other explanation would be some architectural peculiarity of Core 2 which is interfering with the advantages of running in 64-bit mode.

Quote:

---

I don't know if your whole post is _also_ directed at me, but I don't view your thread posted as a "dig" against Intel. I'm old enough to tell a fanboy post from a genuine post..

---

No my post wasn't directed towards you..

Quote:

---

However, I don't see how this is news worthy at this time. Intel's NGMA Core 2 hasn't been out in the wild long enough to really tell. If there's a negative trend in 64 bit performance a year or so from now, you definately have a point, but as of now, it's too early to tell. I recall AMD64 also had simular growing pains as well..

---

Well, I've seen quite a few reviews with 64-bit performance, and it definitely seems that Core 2 does not gain as much as AMD when running in 64-bit.

This may, or may not be a design flaw.

Core 2 is a very balanced and efficient processor, and it could simply be that it doesn't gain much in terms of performance and efficiency when running 64-bit mode due to it being already VERY efficient.

I'd have to agree, that I would have expected that IF Intel's answer to 64 bit were based on a similar setup that there should be a similar increase in performance.

If it's not though, then it would be a very identifiable coding issue that would be VERY easy for programmers to resolve. If there were the case, and software changes were enough to give the C2D a similar increase in performance to AMD's 64bit gains, then the 64bit performance we'd see out of Conroe would be simply amazing.

I certainly hope that's the entire issue, but if the P4 is any indication, and they haven't had time to sort out their lackluster 64bit platform then it's a real shame, as it would certain have potential wouldn't it?

I suppose in a way they don't have to, as the Conroe's performance is pretty stunning, but a gain from 64bit, similar to the gain that the FX62 get's, would put Conroe in some really cool territory :fact:

Gray

Quote:

---

Originally Posted by Carfax

What does your link prove? Those tests are 32-bit, and so they are useless in this discussion..

---

read the title, sheash.

quote
This is the 64-bit version of Cinebench, primed and ready for these 64-bit processors.

Next time you go and slam something I write maybe you should read the whole page

had you looked 2k04 you would have realized they have the 32 and 64 bit.

One more thing, that whole review was done in win 64 (except 1 test), the applications have 64 bit, they used 64 bit.

Quote:

---

Originally Posted by Carfax

Do yourself a favor and stay out of my thread.

Obviously, your reading comprehension must be subpar to come to such conclusions..
No, I am entitled to my own opinions on this matter


The title is a QUESTION you idiot! Did you not see the question mark? The title is not an assertion, but a question..

Like I said, stay out of this thread, and let the grown ups do the talking.. :mad:

---

If you don't like the responses that are received then don't create the thread.

meant to edit reply

Quote:

---

Originally Posted by informal

Carfax,i don't think they got it.Seems to hard for them to understand that A64 has superior 64 bit implementation.I see that C2D is still faster in that one benchmark,but all in all,A64 gains in 64b while C2D loses or gains very little..

---

They got that part. However, they are simply saying that it is hardly relevant *at this time*, since even though Core 2 Duo's performance curve in 64-bit is the opposite of that of FX-6x, it is still better, even at its worst.

In other words; yes, it drops compared to 32-bit (while AMD trends the other way), but Intel still wins *despite* that trend.

Quote:

---

Originally Posted by PGHammer

They got that part. However, they are simply saying that it is hardly relevant *at this time*, since even though Core 2 Duo's performance curve in 64-bit is the opposite of that of FX-6x, it is still better, even at its worst.

In other words; yes, it drops compared to 32-bit (while AMD trends the other way), but Intel still wins *despite* that trend.

---

QFT!

A64s doesn't always gain by going to 64bit, GamePC and BE HW showed that. X86 and even X86-64 is SOFTWARE not Hardware dependent. As apps and the OS is patched to support Conroe, it will get even faster.:eek:

Carfax: AMD pioneered 64bit extensions to X86, NOT the X86 part of X86-64. That might sound like nit pickin' but hey.:D

Vista will be optimized for Conroe the same way WinXP supported Hyperthreading right out the Box. Then as MS tweaked it, it got faster as well. Vista will support VT, the new version of Speed Step and etc.. as well.

The best version of X86-64? As at least 4 sites proved with tests, is the P4. The problem, just like with AMD, the improvements still leaves it behind most tests. What the hell good is a 12% improvement if you still end up being beat by 35% after the 12% improvement for example?

Quote:

---

Originally Posted by Khenglish

The original point of this thread was just to show that Conroe had relatively poor 64bit performance compared to 32bit. I don't see this as Carfax trashing Conroe. It was a simple double bar graph that showed Conroe had worse 64bit performance compared to 32bit, meanwhile the Athlon performed better in 64bit compared to 32bit. Right now, Conroe beats Athlons so badly, that any benchmark where it doesn't burn the FX62 stands out. Conroe is completely expected to beat the FX62 by a lot, but these Cinbench 64bit results show uncharacteristicly close performance, hinting that 64bit tests may be a relative weakness. Based on the original benchmark, in my mind the statement that Conroe has a weakness in 64bit is perfectly valid, and not flaming. Fortunately, other links have been posted showing that this is false. Conroe does not have poor 64bit performance, and as expected from other benchmark results, it beats the FX62 by a wide margin.

---

Maybe it's like what folks (mostly folks leaning towards AMD said or meant) when they called Hyperthreading a Fix. Well maybe 64bit fills a gap for and helps A64 75% of the time. Considering there's almost no built-in support for Conroe, the results should scare AMD fans. Updates, Hot Fixes, Patches and etc.., will more than likely mean an even wider Performance Gap between C2D and A64.

I look at it differently. C2D is new, not mature, mostly unoptimized for and so far NOT buggy and turning results like this at its price point is great. Hell, give Intel a :clap: and a :woot:

You are right!. This should be abt 32bit AMD vs 64bit AMD and 32bit Intel vs 64bit Intel. Not AMD vs Intel. We all know that new Conreo is faster than old K8. :slap:

Quote:

---

Originally Posted by Carfax

... For the sake of competition, you're right. But this thread is about Intel's 64-bit implementation (as it relates to AMD's), and not Intel vs AMD as two companies who are trying to best one another..

---

Quote:

---

Originally Posted by PGHammer

They got that part. However, they are simply saying that it is hardly relevant *at this time*, since even though Core 2 Duo's performance curve in 64-bit is the opposite of that of FX-6x, it is still better, even at its worst.

In other words; yes, it drops compared to 32-bit (while AMD trends the other way), but Intel still wins *despite* that trend.

---

Wrong, not QFT.

At the very least, A64 can beat conroe clock for clock in cinebench, the performance boost in 64 bit is dependent on THE SOFTWARE not the hardware, and pretty much nobody here but Coolaler benches encoding performance, let alone 64 bit encoding performance.

Conroe clocks faster than A64, but it does not win in 64 cinebench clock for clock.

Just imagine what the difference could be in 64 Bit XVID, if people have seen a 55% performance boost, where Conroe's 64 bit is slower.

A 55% performance boost in 64 bit will close any advantage Conroe has, nearly regardless of clockspeed, 32 vs 64 bit, or any other reason.

Conroe is indeed good, but it isnt the best thing on the face of the earth...

Mabye Kentsfield is? :woot:

Quote:

---

Originally Posted by Carfax

There's been alot of controversy about Core 2's 64-bit performance, and while I would go so far as to say that Core 2's 64-bit performance is fairly good and much better than the P4's, it's still not up to par with AMD's.

Also, it could be that AMD's 64-bit compilers are just much better than Intel's, or that the programs themselves were optimized for AMD64 alone seeing as AMD64 was the first x86-64 implementation, and not Em64T..

64-bit benches

http://akiba.ascii24.com/akiba/colum...ages813898.gif

All in all, even if Core 2's 64-bit performance isn't as fast as AMD's, 64-bit won't be much of a factor in the mainstream for 3-4yrs..

---

troll

It all depends on how it's been optimised, if the software is slanted towards floating point then AMD will win, however most software has been leaning to SSE where the Conroe is very strong.
Maybe 64bit software is optimised towards the floating point A64's are pretty strong on that, and A64's were the only CPU's worthy of being called 64bit for so long.

Hell you are using a program which was DEVELOPED in the A64 platform and running "natively" against the new comer. As for being a programmer since 1996, native platform defeat the new platform is nothing.

Let's just remember, there is no any software optimized for Conroe at this time, we supposed to be appreciating about what Conroe can do right now. If all software developers start to play with Conroe and come up with even a little bit of optimization, the outcome would be completely different. Not to mention the SSE4... AMD guys gotta spend some time to reverse engineering this one just like it did all the way since MMX.

By the way, who the hell cares what the :banana::banana::banana::banana: the "Sakura Editor" is? Some kind of florist managment stuff???

I can also post tons of charts using Pee-benching or Banana-Runner to bring down any thing I don't like

Also consider that Core 2 Duo derrived from the Core Duo which didn't even HAVE 64bit support. Intel implemented 64bit on Core 2 Duo just because they needed it on the desktop and server/workstation environment (Netburst had it). If they hadn't we would be VERY disappointed en kept buying AMD's or maybe even Netburst cpu's (with the dumb motivation that in the 'near' future 32bit is obsolete/dead/not supported).

Maybe it's not fully implemented en gets an update in a later revision. Or maybe applications/compilers etc need to be tuned for Core 2 Duo. Who knows the full story? Only Intel.

But lets face it, 64 bit isn't about speed from the start. The way Intel, AMD and Microsoft brought it was to expand the use of more memory (-> 4gig). Do more in the same time, just like the dual-core strategy.

I tried 64bit XP on my intel 830 rig and saw no difference in performance (benchies don't mean anything), maybe Vista will show a difference but i doubt it.

Just my two dutch eurocents :D .

Is 95% of the post in this thead made by blind people ?
First time im in the Intel part of this site and omg!

Should delete the thread and post it again this time with BIG arrows and BIG text showing what the poster mean with the 1st post.
And some should start using text to speech and rerun it a few times in case they really are blind.

I think it was some good info and well see where it goes in the future.

How long has amd been 64-bit? A very long time. I'm almost certain Cinebench was optimized to run 64-bit mode on AMD processors as it would be with most software. Give it a few years or at least wait for Vista were 64bit will become more common.

Quote:

---

Originally Posted by Carfax

Wow, someone actually gets it!

Thanks Khengllish! :toast:

---

hahahahaa... this thread is getting better and better. Carfax you made the intel fanbois get their panties in a bunch. This is a good thread aside from the idiots who can't stop sucking conroes :banana::banana::banana::banana:.

Only time will answer the questions posed

EM64T is indeed inferior as compared to AMD64. This make Conreo run slower on 64bit mode than 32bit mode.

http://babelfish.altavista.com/babel...fkaigai288.htm

http://www.aceshardware.com/forums/r...2277&forumid=1