Sign in with
Sign up | Sign in
Your question

dual core?

Last response: in CPUs
Share
March 3, 2005 5:36:23 PM

not any good according to slvr dldo
what do u think?

your all just a bunch of slaves
how long you gonna let them push you around

More about : dual core

March 4, 2005 12:59:19 AM

Wont do much for me, until id and the boys learn how to use it. Maybe in a couple of years.
My wife multi-tasks like crazy. For her, it will be great.
a b à CPUs
March 4, 2005 1:06:37 AM

maybe when the software cathces up to the hardware...

__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
Related resources
March 4, 2005 1:43:29 AM

do you think there could be a possible software/hardware fix to force the program to multi-thread?

your all just a bunch of slaves
how long you gonna let them push you around
March 4, 2005 2:26:52 AM

Im working on a fix at this very moment.

I see that you found your way to the OTHER. You seem to have drawn out the big boy flame throwers. :lol: 

READ THE STICKY AND WIN A PRIZE! ALL PRIZES CAN BE CLAIMED IN THE SECTION TITLED "THE OTHER"
a b à CPUs
March 4, 2005 2:45:38 AM

I'm definitely not a software guru, but the complexities of using multiple cpus will prevent a software patch from being a feasible solution. There isn't a hardware solution because not all programs are the same. The standard - and correct - path to evolutionizing multi-core utilitarianism will be coordination between the CPU mfrs and programmers to establish standards for both to follow.

__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
March 4, 2005 4:28:19 AM

true
but how did they get a 64 bit cpu to operate quicker?

your all just a bunch of slaves
how long you gonna let them push you around
March 4, 2005 4:48:40 AM

thats good

i guess you like mozzart huh?
ever listen to wagner that stuff rocks

your all just a bunch of slaves
how long you gonna let them push you around
Anonymous
a b à CPUs
March 4, 2005 6:58:53 AM

In other words newer, more complex and much more difficult to design graphics engines right?

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
a b à CPUs
March 4, 2005 1:02:09 PM

It's not that the 64bit CPU will operate faster, but the software will be 64 bit and run faster on than 32bit software.

__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
a b à CPUs
March 4, 2005 1:05:00 PM

Definitely newer engines that fully utilize the 64 bit capability. The one thing I'm not sure about is whether the OS will need to be 64 bit before a 64 bit prgram will be able to take advantage of the 64 bit CPU...

__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
March 4, 2005 11:08:36 PM

The OS controls memory usage. A 32 bit OS cannot use the extra registers, so no 64 bit progs in a 32 bit OS.
March 5, 2005 12:10:23 AM

I don't think anyone said its "no any good". Just depends what you do with your comp. It won't help game performance as such, but it will be great for somone running the game server and a client on the same machine for instance. It won't do a thing for office workloads, but then, what CPU does ? And it will probably be very good for media encoding and 3D rendering stuff. Not too mention, Folding, for those that care.

Personally, I wouldn't mind getting one, however, I would *not* be willing to sacrifice single threaded performance. That is just paramount to my usage. Therefore, intels dual core will probably "suck", but a dual core 2.4 K8 wouldn't hold me back one bit.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
a b à CPUs
March 5, 2005 1:38:26 AM

Didn't think so but didn't want to throw something out there that I wasn't sure about. Thanks :smile:

__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
Anonymous
a b à CPUs
March 5, 2005 6:01:58 AM

Wow so now we need a new graphics engine, a new OS and a lot of expensive hardware to make this happen? Talk about a pain in the arse :smile:

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
a b à CPUs
March 5, 2005 11:15:04 AM

Speed kills...

<pre>your wallet</pre><p>
__________________________________________________
:tongue: <font color=red>Have you read the FAQ? Searched for other posts on this topic?</font color=red>
Anonymous
a b à CPUs
March 5, 2005 7:02:03 PM

True dat :smile:

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
March 5, 2005 7:53:37 PM

Untill the software developers start multi-threading all there software you wont see any benefit. Just like you wont see any benefit from 64 bit cpus. But when they do you will see a jump in performance. It will take time just like when they switched from 16 bit cpus to 32 bit cpus. Eventually software will catch up with the hardware to a point.
One interesting note I read. Anyone who owns a 939 pin amd unit should be able to drop in the dual-core cpus when they come out. If you own an un-Intel-igent cpu you will have to buy a whole new motherboard.
Bonus for AMD again if it works like they say it should.
Good for me since I just built an AMD 64 bit 3200+ system.
March 5, 2005 9:01:20 PM

Should change your name to strum.

-Jeremy Dach
a b à CPUs
March 6, 2005 1:07:56 AM

Please leave those topics in the "Other".

I aint signing nothing!!!
Anonymous
a b à CPUs
March 6, 2005 5:44:50 AM

Lol that's funny :smile:

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
March 6, 2005 2:53:55 PM

>If you own an un-Intel-igent cpu you will have to buy a whole
> new motherboard.

Where did you read that ? its my understanding dual core P4s will be just as socket compatible as dual core Athlons.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
March 6, 2005 4:50:04 PM

Yeah, socket will be the same but they say it won't work on 915/925, only the 945/955 chipsets. I'm not sure WHY exactly, but it'll be interesting to see if someone finds a workaround.

Maxtor disgraces the six letters that make Matrox.
March 6, 2005 5:22:08 PM

Quote:
Yeah, socket will be the same but they say it won't work on 915/925, only the 945/955 chipsets. I'm not sure WHY exactly, but it'll be interesting to see if someone finds a workaround.


to force people to purchase new motherboards!

-------
Work sucks.
March 6, 2005 6:35:36 PM

i heard that because of the 130w tdp they had to redesign the socket, otherwise they would have been able to use 925 with an update i think
Anonymous
a b à CPUs
March 7, 2005 7:17:41 AM

That seems like a serious setback as compared to AMDs dual core. Why the hell would anybody want to buy a new mobo if they can avoid it? AMD was smart for setting their dualie up the way they did IMO :smile:

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
March 7, 2005 11:58:07 AM

a silver one ;) 

your all just a bunch of slaves
how long you gonna let them push you around
March 7, 2005 11:59:58 AM

true

your all just a bunch of slaves
how long you gonna let them push you around
March 7, 2005 12:29:58 PM

y r the amd 64 chips quicker then
if not for 64 ?

your all just a bunch of slaves
how long you gonna let them push you around
March 7, 2005 12:42:25 PM

how bout playing a game and backing up a dvd at the same time, multitasking?

your all just a bunch of slaves
how long you gonna let them push you around
March 7, 2005 1:37:42 PM

Quote:
y r the amd 64 chips quicker then if not for 64 ?

And A64 is a highly revised Athlon XP. The biggest performance increase comes from the ODMC (On-die memory controller), which drastically reduces the latencies for when the CPU needs to access the memory - put simply, it doesn't have to 'wait' as long for the information it wants. There are also a few other optimizations (which I'm not fully clued-up on myself so don't expect specifics here) which also help with performance generally. The chip companies are always doing this. AMD did it to the original Athlon to make the Athlon XP (which is why a Athlon XP 1500+ (palomino core) will outperform a 1.4Ghz Athlon Thunderbird, despite being slightly slower.

Quote:
how bout playing a game and backing up a dvd at the same time, multitasking?

The bottleneck with that sort of thing is the speed at which the writer can write, not the amount of CPU time it's taking up. In theory it could give a very minor benefit, but your CPU is idling most of the time when writing to CD/DVD. If you were actually <i>encoding</i> Video, then it would help a great deal, but you'd probably find you don't have enough RAM. Basically - No-one expects to be able to play a modern game AND encode video, so it's a moot point.

You seem obsessed. Why not just accept that the chip companies actually know what they're doing? :tongue:

---
A64 3200+ Winchester @ 250x10= ~2.5Ghz, ~1.41 Vcore
1Gb @ 209Mhz, 2T, 3-5-5-10
Voltmodded Sapphire 9800Pro @ 450/350 w/ modded VGA silencer 3.<P ID="edit"><FONT SIZE=-1><EM>Edited by ChipDeath on 03/07/05 03:38 PM.</EM></FONT></P>
March 7, 2005 2:37:42 PM

i`m not obsessed i`m just trying to understand more about pc`s, and i think amd know what their doing intel i`m not too sure about.
i just assumed 64bit was why amd`s are faster than intel
so with windows xp 64 then amd chips will be even faster?
and also do you think there will be windows 64 upgrade
or do you think it will be a reinstall?


your all just a bunch of slaves
how long you gonna let them push you around
March 7, 2005 3:18:36 PM

Quote:
so with windows xp 64 then amd chips will be even faster?

Only for certain very specialised applications. The advantage of 64-bit is mostly about being able to address more memory, and this won't really be necessary for a number of years, and current mobos don't support more than 4Gb anyway. I guess there would be some performance increase if you're dealing with 64-bit numbers a lot, but everyday computing simply doesn't require that precision.

Anyone feel free to correct me there if I'm completely wrong. :smile:

Quote:
and also do you think there will be windows 64 upgrade or do you think it will be a reinstall?

There'll probably be an upgrade - there was one from 98SE to Windows XP after all. As to how well it will work, well, the 98->XP one was flaky at the best of times. Depends on how much of the 'XP64' (or whatever it's called) is based on the 'old' XP, and how much is rewritten from scratch. Basically, there'll probably be an upgrade, but it'll probably still be a better option to reinstall from scratch.

Incidentally, you asked about there being a 'patch' to make software use multiple threads. I'm a programmer, and that <i>so</i> isn't going to happen. Writing multi-threaded code from scratch is no picnic, debugging it can be a nightmare, and patching single-threaded code into multi-threaded is basically impossible. It would be a re-write, and the types of things where it's of any use are very limited anyway. In short, for the majority of the time the benefits are <b><i><font color=red>FAR</font color=red></i></b> outweighed by the amount of effort involved in implementation. For multitasking it's useful (As the OS can let separate progs use separate threads), but single apps which are multi-threaded are going to remain the exception rather than the rule for a long time to come.

---
A64 3200+ Winchester @ 250x10= ~2.5Ghz, ~1.41 Vcore
1Gb @ 209Mhz, 2T, 3-5-5-10
Voltmodded Sapphire 9800Pro @ 450/350 w/ modded VGA silencer 3.
March 7, 2005 11:25:33 PM

I probably have this all wrong.
I was under the impression that Win64 would enable the extra registers. I thought that I had read that 2 different 32 bit apps could have access to all of the registers, half each. I would think that that would make the 64 bit chip much faster, in multi-tasking.
March 8, 2005 12:31:20 AM

Your impression is correct, to an extent. You do your standard code whether it be with 32bit pointers or 64bit pointers*I know there are more differences in the code tables sizes and the such but we don’t need to discuss that*, anyways you have your software run it through the compiler*newer would be preferable since most x86-64 compilers are somewhat beta status IMO*, it will deal with the threading*unless you hand do it*, but more importantly it will compile you a version*less debugging* that will be able to access the 64bit registers which in turn will give you access to those resources, but there is a catch it doesn’t always equal increased performance, since now your 3x86 decoders are working with x86-64*I am fairly certain both platforms suffer from this pitfall, but if they do not it's still a matter or execution time 64bit operands will in some cases not all cases take longer to decode and/or shift and those few system ticks on a high IPC chip like the Athlon 64 can cripple the potential performance benefits which generally aren’t there to begin with.

Side note SIMD operands seem to be more technically interesting to me since you can generally get some minor to noticeable performance increases from the use of them. But that is vectorizeation, stacking, double operands/precision.

Now back to the software using the additional registers. They are backwards compatible as well this being feasible since x86-64 is just x86 but with 64bit pointers and the such as well they removed some of the more annoying*or as I like to call them fag operands/instructions/calls or whatever you want to call them*. It doesn’t fix all the aggravating problems that come with x86 but fixes a few performance negating x86 calls as well as some of the wish washy ones, all good for x86 code development.

Either or after everything is said and done your software whatever it may be can access those additional resources but due to retire limits on x86, cache line limits, instructions in flight, instruction decode and system latencies, performance will be negated. Truth be known the Athlon 64 can deal with these pitfalls a bit better than the Pentium 4 can at the moment due to latencies from the on-die memory controller, which also will give it a must higher probability on current and future software coded for x86-64 in mind to have a nice edge.

But simply put x86 is a mess and I personally think it’s the core reason why there are so many piss poor programmers at least in the x86 arena. But that’s for another discussion all together.

Oh the multitasking deal it won’t do too much if anything unless you are trying to get at very large data sets, even then I really don’t see it helping all that much since the software’s compiled form tells the CPU what to do the OS is just managing that dance with out hogging too much for itself.

-Jeremy Dach
March 8, 2005 3:05:46 AM

so why is intel so hot to come out with dual core?
i guess a reason to sell more cpu`s?

your all just a bunch of slaves
how long you gonna let them push you around
March 8, 2005 10:03:45 AM

Quote:
so why is intel so hot to come out with dual core?
i guess a reason to sell more cpu`s?

That's the only reason they ever do anything, same as all the other companies in the world. If they didn't think they'd get money from it, they simply wouldn't bother.

Xeon:
I got completely lost trying to read all that, but then I'm not <i>that</i> low-level a person. Reasonable grasp of the basics, and pretty reasonable at higher-level OO languages like Delphi (although far from the best, by my own admission). but what you're basically saying is that although there's a theoretical performance gain from fully utilising the 64-bit capabilities, in practice there probably won't be, yes?

At the moment all that really bothers me is that My new A64 runs all my games damn well, and crunches noticeably faster through anything else I do with it (WinRAR, a little Compiling). Couldn't give that much of a stuff about it's 64-bit capabilities right now. :smile:

---
A64 3200+ Winchester @ 250x10= ~2.5Ghz, ~1.41 Vcore
1Gb @ 209Mhz, 2T, 3-5-5-10
Voltmodded Sapphire 9800Pro @ 450/350 w/ modded VGA silencer 3.
March 8, 2005 12:21:40 PM

I also stated that not all cases will there be a performance less that or equal to 32bit performance. Specialized apps coded to utilize larger table sets, this case winrar. As well it just happens to be encryption is generally faster on 64bit machines... generally.

I also stated the Athlon 64 would show less performance loss due to the on-die controller. Intel has a valid solution but apart from the pain the ass platform dependant coding for Smithfield. In the end this 64bit desktop machine race will likely go to AMD till the next Intel core which I supposed to be a hybrid of the Pentium D and Pentium M but I doubt it will be that fantastic.

-Jeremy Dach
March 8, 2005 12:35:57 PM

>Anyone feel free to correct me there if I'm completely wrong.

You are completely wrong, sorry :) 

64 bit adressing is NOT about the ability to adress more than 4GB RAM (PAE could do that 10 years ago, no sweat), its about creating a flat <b>virtual</b> memory space, ridding us of the current 2GB/2GB OS/app limitations, and ridding us of (virtual) memory fragmentation. And no, it wont take 4 years to need that, we needed that >5 years ago to make sure the software would be here now when we are really beginning to need it.

>and also do you think there will be windows 64 upgrade or do
> you think it will be a reinstall?

Don't expect a patch if that is what you mean. XP64 is based upon Windows Server 2003, a pretty different beast from XP Home/Pro.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
March 8, 2005 12:53:57 PM

>I thought that I had read that 2 different 32 bit apps could
>have access to all of the registers, half each.

Nope, that is utterly impossible.

> I would think that that would make the 64 bit chip much
>faster, in multi-tasking.

Again, nope, no difference at all. Well, AMD did add some instructions when they extended the ISA, and some of them if memory serves where to increase context switch speed, but that is not going to make any real difference.

The speed gain "64 bit software" might bring will come from four things (in dimishing order of importance):
1) (most important), 64 bit apps will be compiled with optimization for modern CPU's (K8 and prescott) only, since they will not have to run on older Pentiums, K7's etc. That alone will give a very nice speedboost for a lot of apps, even though the cause is very indirect.
2) In 64 bit mode, the apps will be able to use twice as many registers, meaning less register shuffeling and renaming. Expect no more than a 10-15% increase from this for most apps. Pretty much all apps should benefit here though.
3) The ability to do 64 arithmetic without having to break them up into 32 bit ones. Only usefull for a very limited set of apps, like encryption, math simulation, etc. For those apps, up to 4x speed increase is possible, for anything else, zero speedup.
4) "Unlimited" virtual address space for apps and OS. This benefit will be hard to measure, and only applies to apps that need large chunks of contigeous memory (photoshop comes to mind). The benefit here will be to be actually able to treat certain workloads, as compared to being utterly unable to do that ('out of virtual memory'). Furthermore, reduced virtual memory fragmentation should increase VM speed, so less harddisk stuttering when you taskswitch between large apps.

However, there are also some factors that will reduce performance.
1) running 32 bit apps under 64 bit windows, an emulation layer (called WoW64) is invoked, to give the app a virtual 32 bit mode CPU. The overhead seems pretty small though, from tests I've seen, the penalty is barely measureable. Good job by MS here. Linux will not suffer this, as AFAIK, it won't even allow 32 bit apps to run under 64 bit OS, which is not a biggy, since you can just recompile them in 99% of the case anyhow.

2) I think this is what Xeon wanted to explain: in 64 bit mode, your pointers will be twice as big, leading to larger code, which effectively decreases you available bandwith. However, compiled 64 bit binaries are only around 5% bigger than 32 bit ones, so the effect should be very small, and in all but the most rare cases, easily offset by the other advantages.

I hope this clears up some confusion.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
March 8, 2005 1:08:06 PM

>so why is intel so hot to come out with dual core?

Not only intel, everyone is. The real reason is that we're hitting a brick wall in single threaded performance. Most of the tricks are implemented these days, so the low hanging fruit is gone (superscaler, Out of Order, Trace cache, HT,..). What should be left is clock scaling, but there we are hitting the thermal brick wall, at least Netburst is, and process shrinks do not seem to be giving the free clockspeed increases it used to. The biggest advantage of shrinks is now just smaller (therefore, cheaper) transistors.

So the only sensible thing left to do, if you can't scale vertically (read: make faster pentiums) is scale horizontally (read: more cache or more cores on a die). That way you still reap the benefit of using newer process technology, while increasing maximum theoretical performance you need to sell chips. It makes some sense, but at the same time its an admission of defeat and it won't benefit everyone.

For these reasons, I'm *very* curious to see what AMD will do with K10, and if they will be able to significantly scale vertically. Intel, so it seems, has pretty much given up on that idea. Their mantra is now more cache and more cores, wether it is for mobile, desktop or high end server.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
March 8, 2005 11:22:46 PM

This wasnt the place but a search on google did bring this up from <A HREF="http://link" target="_new">Clicky</A>.

"Note that the Pentium D will require a new motherboard, built around the upcoming 945/955 core logic. If you insert a Pentium D into a current 915 or 925XE motherboard, the system simply won't boot—neither the CPU or motherboard will be damaged. It simply won't work. Note that the current LGA775 CPUs will work in 945/955 chipset boards."
March 8, 2005 11:31:34 PM

Sturm is for Sturm Brightblade in the Dragon Lance Chronicle series.

UH! Why change it to strum?
March 9, 2005 4:33:53 AM

y don`t they just design a bigger engine
instead of increasing rpm`s

"your all just a bunch of fuc*ing slaves
how long you gonna let them push you around" -jim morrison
Anonymous
a b à CPUs
March 9, 2005 5:47:31 AM

Sturm is fine, man. That was a great series. What was the wizard's name, Caramon's brother? That guy was bad ass. Been so long since I read those, I can't remember his name. Lol, oh well probably the wrong place for this anyways.

<font color=green> Woohoo!! I am officially an <b> Enthusiast </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
a b à CPUs
March 9, 2005 10:50:12 AM

Raistlin?!

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>
Anonymous
a b à CPUs
March 9, 2005 3:56:02 PM

God, thank you *smacks himself in the head* Damn brain farts... :smile: Raistlin was cool :cool:

<font color=green> Woohoo!! I am officially an <b> Addict </b>!! </font color=green>
<i> <font color=red> One new Firefox fan </font color=red> </i>
March 9, 2005 4:08:51 PM

>y don`t they just design a bigger engine
>instead of increasing rpm`s

there is only so much power you can extact from a 500cc single cylinder engine, limited to 10.000 rpm. You could probably even calculate the maximum if you achieve 100% efficiency.

As it is, increasing the revs seems mostly a no go (definately for Netburst). Increasing the efficiency is even harder, and we're already close to (or over ?) 1 IPC, using tricks that make the CPU enormously complex already (executing code out of order, executing in parallel, guessing results, speculative loads, predicting branch results before they are calculated, etc, etc,) so unless you completely change our programming paradigms, there is not much more you can ever hope for. You'd need a different kind of code (fuel). I don't see a parallel for increasing the engine displacement, but adding cylinders is easy.. thats multi core with its advantages and drawbacks.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
March 10, 2005 2:31:40 AM

very well done
that is crystal clear

people are strange and when your a stranger faces seem ugly when your alone
March 10, 2005 5:43:31 AM

You forget that those extra register add 1 to 3 bit in every INT and SIMD.

Lose L1I cache space/bandiwth more pressure on front end pipe or you can reduce the number of port the int reg have.

i need to change useur name.
!