AMD is going to release the first 64-bit processor for pc on this September...cool

its a good news for gamers...and audio....video editing....

G5 is also 64-bit too if i'm right

Intel is going to continue releasing 32-bit processor with technology like HT etc.


who will win ?
Intel or AMD

It's hard to say. Intel have the upperhand at the moment, thouroughly wipping the Athlon XP 3200+. However, the technology in the AMD 64 is pretty exciting, and if they get some major software support, they'll be on the road to the top. There is already a 64-bit version iof WinXP in dev and UT2003 is being ported to. If people like nVIDIA, ATi, Creative, etc. follow suit there Intel will be threatened. As it stands, I don't think the P4 architecture can produce much more performance. We have seen water-cooled P4's at 4GHz+ with little performance boost to complement the speed boost, which means there is an underlying bottleneck in their architecture. They do have a new chip in the works so, and they have financial power.

Hollebollegijs: its at times like this i wish i had a 3.2ghz P4
Hollebollegijs: they're the best you can get for compressing
Hollebollegijs: especially because 7z supports hyperthreading
Mac: brb
Mac: ooh
Mac: why not 64bit athlon?
Hollebollegijs: aah?
Mac: does that hyperthread?
Hollebollegijs: no
Mac: y not?
Hollebollegijs: i hate to admit it but the athlon64 probably wont beat it in single processor configuration
Mac: really?...
Hollebollegijs: not with 256k cache and single channel memory
Hollebollegijs: 1 meg and dual channel, it might
Mac: i thought it was meant to be faster than 32bit althon by a way... so it'd own pentiums
Hollebollegijs: it is tons faster clock for clock yes
Hollebollegijs: but the p4's have more then 50% higher clock speeds
Mac: lol yea..
Mac: it's a gay race!
Hollebollegijs: look at it this way
Hollebollegijs: its impossible to make a cpu that wins every benchmark
Hollebollegijs: amd and intel are moving more their own way
Hollebollegijs: you choose one or the other depending on what you want to do with it
Hollebollegijs: which means less competition which is good for amd
Mac: why is less competition good for amd?
Hollebollegijs: the opteron is a strong performer with smaller dictionaries because it has more cache, though. but i dont know what that would do for 7z with a large dictionary
Hollebollegijs: because intel is a giant, you really cant beat it on its market
Hollebollegijs: AMD is somewhat creating its own market
Mac: aah, i get u
Hollebollegijs: or at least trying to
Mac: amd seems like a more rugged processor, good for games and non-linear simple maths crunching
Hollebollegijs: its also being smart and attacking intel at its only weak point
Hollebollegijs: its good for games, databases, encryption and scientific applications
Mac: er, that's everything isn't it?
Hollebollegijs: thats quite a lot
Hollebollegijs: and unoptimized apps in general
Hollebollegijs: that includes all office applications
Hollebollegijs: but they arent performance sensitive anyway
Hollebollegijs: 3d rendering it isnt properly tested with yet
Mac: what's intel best at?
Hollebollegijs: but its a p4 stronghold as it is
Hollebollegijs: 3d work, content creation, streaming applications
Hollebollegijs: basically because those apps can take advantage of hyperthreading and dont have much branches
Hollebollegijs: which enables the p4 to put its execution units to good use
Hollebollegijs: the p4 has more raw power, but the athlon uses what power it has much more efficiently
Hollebollegijs: its just two totally different design philosophies
Mac: aye
Hollebollegijs: the days amd made intel clones are long gone
Hollebollegijs: which is pretty pointless anyway because you wont beat intel by doing what they do, because you wont do it any better
Hollebollegijs: i think we had this conversation.. or rant.. before
Mac: yeah
Mac: i remember about amd = clever p4 = fast
Hollebollegijs: simply put, yes
Mac: simply is good
Hollebollegijs: but market technical the situation is a lot more complicated
Mac: market technical?
Hollebollegijs: intel has two different processor lines and a gap in between so they wont compete with eachother
Mac: yea.. ?
Hollebollegijs: the p4 and its derivates are 32 bit x86 and relatively cheap, but limited in potential. the itanium line is 64 but, expensive, and very powerful
Hollebollegijs: only, if they made the P4 64 bit, noone would buy the expensive itanium anymore
Hollebollegijs: they basically compete with themselves
Hollebollegijs: so amd makes the middle solution
Hollebollegijs: 64 bit, cheap, and backwards compatible
Hollebollegijs: they cant drop their 64 bit line, because they've spent billions developing it, and so far earned only 1% of the cost or someting
Hollebollegijs: add ibm, hp and sun to the equation and things get even more complex
Mac: soz man, had to ignore everyone there for a bit
Mac: it's not that your super super nerdy conversation is gay or anything


I had that conversation the other evening, it might help answer your question. I think both will 'win' in that they are widely used and are profitable. Intel will always maintain the enourmous advantage while all Dell / Gateway style companies are bullied into supplying Intel only PC's..

The Portland Group has made decent compiler for x86-64. It looks like there can be some real software support for Athlon64 in some day.
Link


Results from SPEC2000 looks promising, altho with those result we can't make any final conclusions for Athlon64 performance.

But don't forget that Intel Prescot will be released also in end of the year with verry promissing specs.

I wouldn't bet for Athlon64 beating P4 Northwood in this year.

This Hollebollegijs person doesn't know what they're talking about.

HyperThreading doesn't benefit multi threaded applications anymore than it does single threaded applications. This is because all it does is basically devide your CPU's resources into two parts. (3.2GHz HT CPU = resources devided into two 1.6GHz CPUs worth).

The first Athlon 64 to ship will have 1MB cache and a 128-bit memory bus (dual-channel).

Cough.. 3D work/rendering, content creation..? These are precicely the tasks that will benefit most from AMD's 64-bit processors.

hmm.. I don't think it's quite like that. It doesn't divide the processor so that it's 2*1.6ghz.
It's basically 2 logical processors but still only one actual processor. If you set some single threaded process's affinity to logical cpu0 or cpu1, it doesn't mean that the process runs like in 1.6ghz.. infact it runs pretty close to or at max speed if there isn't anything else going on at the same time. Taskmanager shows 50% cpu, but that is NOT what is the real deal there, infact the processor may actually run at 100%. Taskmanager doesn't understand the HT processing so well... There's been lots of discussion about this at the folding@home forum... In HT 50%cpu shown by taskmanager for one logical processor doesn't mean that the process could run faster in non-HT environment. Well.. this is complicated and there's more to it, but the main point is that HT does not mean 2*somefigure. It means just 2 logical cpus.

Anyway, I remember reading that amd-64 has hyperthreading or very similar technology?

<edit> in reply to the same quote JohnV used...just in case this looked like a reply to JohnV (quicker on the draw John, heh)

I'm afraid you don't know what you're talking about. Go read some more about hyperthreading again, because you have that totally wrong. It doesn't do anything even remotely like giving you two virtual CPUs of half clock speed each. You're right that it shares resources to give the appearance of two CPUs when only one exists, but how it does it and how well it actually performs is what you're not clear on. There are several processing units that make up the CPU as a whole, and hyper-threading allows those units to work in parallel. It isn't like having two CPUs of half the clock speed. It's more like having more than one CPU of the same clock speed, but not quite two CPUs. Don't forget to check out the video of a non-HT 3.6GHz P4 and an HT 3.06GHz P4 side-by-side on the Tom's Hardware site, as it gives a very clear idea of how it can affect the user experience.

http://www.intel.com/homepage/land/hyperthreading.htm
http://www.tomshardware.com/cpu/20021114/p...4_306ht-22.html
http://www-106.ibm.com/developerworks/linu.../library/l-htl/
http://www.theregister.co.uk/content/archi...hive/25774.html
http://www.hardwareanalysis.com/content/article/1557/
http://www.google.ca/search?q=hyper-threading

I was specifically talking about multithreaded applications/multi threads. With HT enabled, and multiple threads being processed, processor resources ARE DEVIDED. Example: If you run two threads, each doing the same work, and each using full CPU time, then essentially each thread will only be able to do about 1.6GHz worth of work (with a 3.2GHz HT CPU).

I thought I read at tomshardware.com that Intel already have the P5
up and running, and are just waiting for AMD to make their move...

Umm.. Well, the thing is HT is faster than non-HT in multithread processing. So your impression "only be able to do about 1.6ghz worth of work" is a bit misleading, because HT works faster than non-HT in that case.
There should be VERY few cases when HT causes any performance hit, and even then it's very small. But in normal use when you use many software at the same time, HT is useful.

HyperThreading does occassionally result in reduced performance (including in SPEC), but it's certainly not 2 separate 1.6GHz processors. The amount of usefullness one will get out of HT varies considerably, but (complex) task switching should see a nice boost, though much less than SMP.

I do not believe so. The OPTERON, which is the server/workstation version of the athlon64 and has been shipping for quite some time, uses a 144bit memory bus (128bit + 16 bits for ECC) at 333MHz effective and has a 1MB L2 cache.
The Athlon64 will initially use a single channel (64 bit, no ECC) memory bus at 400MHz effective, though this may be bumped to 128bit if necessary.
The primary difference outside of this is that the Athlon64 will only have a single HyperTRANSPORT link whereas the Opteron has three (two more for connecting to other processors in dual, quad, and larger configurations).
A64 designs have an integrated memory controller on-die, which dramatically reduces memory latency relative to more traditional designs.

I don't think so. The A64 design doesnt look to benefit from HThreading as much as a longer pipeline design, like the P4. The A64 family may see on-chip SMP like the Power4+, meaning two CPU cores on a single die.
Edited to clarify difference between HyperThreading and HyperTransport

No, I'm pretty sure theA64 will have a HT-like technology.


I've read this as well.

I think I read both at Tom's Hardware, but I'll have a look around.

(XXX) = unsure

AMD is (always) faster than Intel at the same GHz without HT
and is always cheaper

To be honest my systen the old old old P4 willamate 1.4 GHz is much much slower than AMD's Athlon Tunderbird 1.2 GHz. In my opinion, AMD Athlon 1.2 GHz could perfome as fast as a P4 prescott ~1.6 gHz

AMD processors consume (twice) more power than P4 if im right. its hotter too?
AMD with a good heatsink will be cheaper than P4 also.


I am regret to buy P4

My Motherboard uses RIMM. The proprierty RAMBUS DRAM is much expensive than DDR RAM

Hey guys... have you heard of the newest "MRAM", the future RAM and its belived to replace DRAM in 2005.

MRAM can store information after system power off. You can switch on computers without waiting for software to boot..its something like hibernation.

Wonder if AMD supports MRAM?

Are you talking about HyperThreading or HyperTransport? Athlon64 doesn't include HyperThreading.

kotrtim: I don't think that MRAM replace DRAM quite so fast. I think MRAM will replace FlashROM first in some cases, because both memory types can hold it's value after cutting off the power.

Hyperthreading is not dividing the virtual processor power by 2. It just helps to minimize the number of "bubbles" in the pipeline.

So you have (most of the time) with HT the same processing power as without HT plus a reduction of the number of "bubbles". So overall it is faster.


Btw AMD has a patent on a process similar to HT. That does not means that it will be included in Hammers.

HT-like technology is reserved for AMD's next-gen K9 processors (they'll have multiple cores).

About Intels HT: I thought it's build to improve the availability of the processors resources. Improving in a way, that

a.) response when switching between different tasks (I mean different applicatios) is better
b.) multithreaded processes run faster
c.) a single application doesn't anymore 'consume' more processing power than it really needs

Regarding point 'c.)' If I unterstood it correct than a CPU like pentium 4 has several processing units, whose resources aren't used or are seldomly used (or can't be used?) fully by one program. The deal with HT now is, that it 'virtually' divides the processing resources into smaller units (simply put) instead of one big processing resource which can't be effectively used. HT simply provides the processors resources in a more effectively way -> idle resources aren't 'blocked' just because there's already some task running (see above points a, b and c).

Well that's all really simplified but my english isn't good enough to explain better what I mean.

Any comments?

userXYZ: lite more detailed explanation is that with HT you can use both ALU and FPU units at same time with two different thread. CPU without HT can use only on of them at the time altho there where two threads available.

Well, a cpu without HT can of course use simultaneously use integer and floating units.

I had a look on Intels website but all I found was the consumer friendly explanation, HT enables parallel execution of two threads. But I assume that's not all to it. If someone's got more detailed info (link?) on HT I would be interessted in that.

But only in single thread at the time.

Wrong.

Hyperthreading enables parallel execution of two tasks.
For every virtual CPU there is also a complete Address Translation Unit.

The name "hyperthreading" is actually wrong and designed by the marketing department of Intel.

What's the difference between a thread and a task? According to this quote from Intels website, threads are parts of a programm. But then again it's the version that's intended to be consumer compatible so propably there are too much details left out

This post originally was about a forum bug, I now noticed JohnV about it.

No. No. I'm completely sure it won't, at least in the first generation so there!
The Athlon64 and Opteron use a serial high speed interconnection protocol called HyperTRANSPORT which has no relationship whatsoever to HyperThreading, which is an Intel trademark, and a form of Simultaneous MultiThreading (SMT).

There's a good general explanation of SMT in the context of the cancelled Alpha EV8 superprocessor right here,
and there's a good HyperTransport SMT article at Ars Technica.