Should I Buy Athlon64?

Few days ago i posted (what i thougth was final) config of my to be comp. But after reading all the glory talk about Athlon64 and given my previous experience with "old" CPU generations (very little choice for upgrades) i think i`ve decided to go with Athlon64 3200.

So i`ll throw away some of the stuff i wanted but the config will (i hope, unless some more price forced cuts happen)look like this:

Athlon64 3200

Motherboard - i need suggestion with this (most likely Asus SK8N)

512 DDR 400 (it was 1gig of 400 Kingston, now i`ll take what i can)

Good ol` SB Live 5.1

"Old" 80gb Maxtor

Old USR 56k v92

My beloved Cambridge SoundWorks DT2000 speakers

Some cheapo Key+mouse

Samsun 959NF

and last but not the least Asus AR9600XT (i think i won`t give up on this, even if it ment a whole month without a comp)

U`ll notice that most of the parts is from my old comp (which is going to my dad`s office cause he doesn`t need a fancy machine for Office) and some of the stuff (that r cheaper) i`ll buy/upgrade as i earn more money.

So, should i go with this?

PS No Mac talk ;) i`m buying a PB G5 early 2006 when PPC980l comes out :) (here`s to hope :D )

#85537

I'd say wait a little until the motherboards are fully stable, the first few new generation mobo's tend to get a little buggy. And also wait until a few more version of the 64bit chips are out, the prices will go down.

#85540

i know that wiz but that would mean a couple months without a PC! :o

*biting nails, shaking in crisis*:nuts:

My PC is going tomorrow!

#85543

Also, I would suggest waiting until Windows XP for x86-64 is available and has been around for a month...

Unless you're really desperate to upgrade within the next month, I'd say wait 'til May or so.

#85544

Depends what your gonna do with it.

If its a gaming machine i say go with a P4EE cause it whoops the A64

If its just a desktop go with the A64.

Persionaly i would hold off for a couple months to get an A64 tho cause prices will drop sharply plus MB support will grow substantialy. I tthink the speed impact youll notice from 1024 megs of ram tho compaired to 512 will outweigh the CPU. Also why are you getting an asus 9600xt, isnt it like 20 or 30 more then ATI oem and saphire. There the same cards esentialy.

#85545

well that WAS what i was planning to do but my dad`s PC broke (burned MB) and he needs a new one ASAP.

Maybe a cheap transitional solution?

Buy a good monitor and a case now, some cheap XP 200+ and MB untill May? :blink:

#85548

You could always get a temporary athlon xp mobo combo set-up from pricewatch and cheap ram to hold you over. The 64 bit processors are going to take a dramatic nosedive in price in about 3 months, and by then there will be better mobos to support it as well as stable OS support and other software.

They have an athlon xp 2000 with mobo for $76 and 256MB ram for $32. When you got your new stuff you could bung this in a cheap case and have a linux box.

#85552

@ spyder

Well as far as i know FX51 whoops P4EE`s ass :D but i neither need it nor can afford it.

The machine would be a gamer/dtop hybrid :D

i know i should wait but i hv 2 solutions (well 3 if u count cheapo version)

buy now an 64 or

be without PC untill May :( :cry:

i would gladly add more RAM to my current machine but it supports only old SDRAM one sided modules that r very hard to find.

And about Asus R9600, well i just want it to get along with MB REALLY good.

From my experience MB & GPU from the same company deliver best results.

#85553

yeah that sounds good, pick up a cheap biostar mb and like a 1700, do a huge vmod on the 1700 to fry it out so you got a decent OC. Should only run you 80$ or so. Id say buy some good ram now, but the a64 uses eic regesterd ram doesnt it? Im not shure regesterd dimms will work in any board.

#85555

Originally posted by lostspyder@Dec 3 2003, 09:04 PM

Depends what your gonna do with it.

If its a gaming machine i say go with a P4EE cause it whoops the A64

If its just a desktop go with the A64.

Persionaly i would hold off for a couple months to get an A64 tho cause prices will drop sharply plus MB support will grow substantialy. I tthink the speed impact youll notice from 1024 megs of ram tho compaired to 512 will outweigh the CPU. Also why are you getting an asus 9600xt, isnt it like 20 or 30 more then ATI oem and saphire. There the same cards esentialy.

Athlon 64 gets whooped by P4EE? Since when? They're pretty much on a par. If anything, the Athlon 64 is marginally faster...

64FX, however, whoops everything.

I would recommend getting something cheap for the time being, Unbeliever.

Personally, I'd wait until May at the earliest before I upgrade, because Athlon64, whilst very good, will come down in price...

#85556

I have to agree with everyone else here. Wait until mid next year. This would not only allow you to take advantage of the drop in price, but there are likely to be 64 bit native apps on the market by then too. Current 32 bit apps will not show much of an improvement over a standard 32 bit chip, but the 64 bit apps will be able to full advantage of the chips power.

#85577

Originally posted by lostspyder@Dec 3 2003, 08:04 PM

If its a gaming machine i say go with a P4EE cause it whoops the A64

heh, no.

Its hard to imagine why so many people continue to claim (after defignate proof has been given and conclusions have been verified that the P4EE loses the performance battle to the AFX) that the p4EE "woops the AFX". To prove that the AFX is faster is a rediculously easy task. One needs only to examine the reviews printed when the AFX was released, and when the p4EE was *ahem* announced.

Of course, performance isn't everything. From the standpoint of other important features, the k8 is every bit as impressive and appealing. The advantages in non-performance categories once held by intel have dwindled to null. For one the AFX notably now will shutdown/poweroff when it overheats. This ability was missing in the k7, and admittedly it is an attractive feature in intel CPU's that AMD has finnally adopted. Secondly, the AFX has adopted a heatspreader design ontop of the chip which results in greater protection and strength. As for power consumption and heat disapation, these are again strong points held by the AFX. According to Aceshardware, the chip normally requires 60W to run, compared to the 90W required by the p4. Most importantly, the chip runs at temperatures lower than the XP and P4, and this is accomplished with a much quieter heatsink as well. This adds up to a safer, sturdier, cheaper (power consumption), cooler, quieter solution than the p4 and XP.

I don't care much for the p4 emergency edition. All it appears to be to me is a last minute attempt to create a distractive smokescreen to stear attention from the AFX51. It is essentially a xeon with 2way/4way capability stripped away and a doubled FSB. Nothing impressive here. The AFX on the other hand is an architectural wonder. It manages to pump a higher ipc than the k7 while at the same time adding 2 pipeline stages. This goes against traditional CPU architectures, but if anyone wants, I can explain.

Here is a small sample of results of the reviews accross the net.

Sharky Extreme:

48 for AMD, 13 for intel

Tech-Report:

35 for AMD, 19 for intel

Sudhian:

63 for AMD, 32 for intel

Aceshardware:

31 for AMD, 13 for Intel

[H]ard OCP

8 For AMD, 3 for Intel

Anandtech:

9 for AMD, 4 for intel

If you take up the effort to check out any of these links, you will see that the few benchmarks won by intel are rarely Gaming benchmarks, and are minimal ones at that.

Therefore, I conclude that "If its a gaming machine i say go with a P4EE cause it whoops the A64" is incorrect.

Peace

#85643

BTW, the p4EE doesn't exist. I challenge anyone to find it instock at an online retailer.

#85645

true, true :beer:

#85654

http://www.tomshardware.com/cpu/20030923/athlon_64-50.html

http://www.tomshardware.com/cpu/20030923/athlon_64-27.html

http://www.tomshardware.com/cpu/20030923/a...engl_benchmarks

Just a few where the P4EE whoops the FX

Im no particuar fan of intel or AMD, and i hate people that are like ohh amd sucks or ohh intels are way too expensive.

And newegg seems to have them in stock or so they say.

#85718

yea...also...AFX is only more efficient and faster for application like word and business programs...

generally, Intel is always the best in Content Creation, Multimedia and Gaming and had been beating AMD since basically Day 1... :canadian:

#85737

If you are going to get an A64, then make sure it's not the 754 pin version. Apparently it's just a crippled 940 pin one (by pins I mean the little legs under than sit in the mobo socket). The 939 pin chip is probably less crippled, but hey, thats all I've heard.

#85757

Originally posted by Ludge@Dec 4 2003, 09:52 AM

If you are going to get an A64, then make sure it's not the 754 pin version. Apparently it's just a crippled 940 pin one (by pins I mean the little legs under than sit in the mobo socket). The 939 pin chip is probably less crippled, but hey, thats all I've heard.

The 754 pin chip is a standard, one-way Athlon 64.

The ~940 pin chips are Athlon 64FX and Opteron.

The difference? Well price, for one, but also the fact that "normal" 64s can't use dual channel RAM, whereas the others can. Also, the 940 pin chips have the capability to have higher-speed interconnects. And they're easier to use to replace a bed of nails ;)

I may be wrong, but I'm 85% sure that's right... :blink:

#85770

U beat me to it Stevie ;)

and ur right by the way :)

but i`m going with the notebook idea untill May/April/June (and i hope it`s a good decision)

#85779

Ah, never mind, that's what comes from joining in stuff I don't fully understand. But I learned something new!

#85785

Originally posted by lostspyder@Dec 4 2003, 05:14 AM

Just a few where the P4EE whoops the FX

Quake 3 has always favored the netburst architecture. Whats more, the game is several years old and is comprised of outdated technology. In addtion, I doubt anyone will ever, ever see a difference between 400 fps and 500 fps.

As for pcmark memory, that isn't a test of the CPU. more than anything else, it does what the name says...it tests and benchmarks the memory which in this case includes intel's on-die l3 cache. (doesn't say much for the benchmark does it?)

And I wouldn't consider commanchee 4 a "woop" It was a grand difference of....1 frame per second! the 1 frame there is wel within the margin of error. At many sites the win goes to the AFX. However, the fact remains that the AFX takes many more gaming benchmarks than the p4EE.

Again, how is that a "wooping"?

#85914

Originally posted by Ludge@Dec 4 2003, 08:52 AM

The 939 pin chip is probably less crippled, but hey, thats all I've heard.

The 939 version will allow users to use non registered and unbuffered ECC ram. This will result in a performance boost. Most home users won't want ECC ram. Its really meant for servers and mission critical environments where stability and error checking are critical. As it stands now with 940 pins, you still must use regisered RAM.

#85919

Originally posted by michaelau32488@Dec 4 2003, 06:53 AM

yea...also...AFX is only more efficient and faster for application like word and business programs...

generally, Intel is always the best in Content Creation, Multimedia and Gaming and had been beating AMD since basically Day 1...

Im sorry, but this just isn't true. If you are refering to an isolated event in historical time, yes I'm sure you could find one. As for the present, and numerous times in the past, AMD has been beating intel in gaming, buisness apps, content creation apps, and many multimedia apps. It largely depends on which benchmark in each category you choose. soundforge has always been a win for AMD, as well as winstone content creation 2002 and winstone buisness 2003, 2003.

As for gaming, AMD had been in the lead up until around the 2800, and then intel pulled ahead. With the introduction of the AFX, all bets were off. The AFX clearly dominates that category.

As for efficiency, you are also wrong (no offence.) The added 2 pipeline stages were NOT a bizare type of admission to intels path being supperior, but actually steps to added efficency. The k7 had 10 pipeline stages, the k8 has 12.

Normally, each pipeline stage requires more time, and each instruction will take longer to execute. It so happens that it is possible to save more time by each stage, than the very existance of the stage quanders.

The added Pipeline stages (believe it or not) actually result in numerous IPC improvements. In the AThlon XP, there are two pipelines. One is for Int, one is for everything else (MMX/FPU/SSE). Every Clock cycle, Each pipeline can finish a maximum of 3 instructions. When a given pipeline begins to work on instruction, it does not know what type of instruction it is until it is decoded. This means that when the 3 instructions begin, they could be one of any type, even one not for specific for the pipeline. Should this be the case, the pipeline will only be able to retire the types of instructions it can handly and thus it will not retire three. When Mixed instructions are are provided by the decoders, (this happens very frequently statistically speaking) only those specific to that pipeline will be retired. For example, If in the INT pipeline one set of FPU instructions, one set of Int, and one MMX, are decoded, only the Int instruction will be retired. This means that while the int could be fulfilling three instructions, its only working on 1! The situation works vice versa as well. If this occured in the the "Everything Else" pipeline only two instructions could be retired as opposed to three because the "everything else" pipeline does not handle int.

This is called inefficiency. It so happens, that by adding another pipeline stage, and resolving this, the efficiency could, and would go up, not down. With a higher efficiency, comes a higher IPC. (FYI, IPC is not static but is a best case scenario if everything goes perfectly. This is very rarely the case.) In the Athlon 64, one of the added pipeline stages slightly decreases IPC by its presence, but increases IPC by its function. I'll explain what this one does.

The stage, nicknamed "pack" comes in right after the decoders. It examines what type of instruction has been decoded and "organizes" them by type and switches the FPU/MMX/SSE to the other pipeline if it is in the INt pipeline. If we are talking about the FPU/MMX/SSE Pipeline, the pack stage Identifies all INT instructions and switches them over to the INT pipeline. The means that pipelines will never be partially idle as a result of the absent ability to switch lanes.

The stage directly following it also improves efficiency. It does so by optimizing register moves. Everytime dram is read it is erased/cleared. In addition, due to capacitor leakage it has to be refreshed hundreds of times each second. For something as critical as registers, every scrap of time saved counts alot. Register moves take up time. There are two types of register moves...Destructive, and non-destructive. Destructive moves occur when you move the contents of one register to another and delete the contents of the old one. Non-Destructive register moves occur when you move the contents of one register to a new one but the old register remains as is. In reality however, each time a cell is read the contnets are lossed and must be rewritten. So belive it or not, non-detructive register moves requir more time. Consider the following example: Imagine four registers, R1, R2, R3 and R4. A string of code instructs to move the data from R1 to R2, Add the contents of R2 and R3 and store the result in R3 without destroying R2, and finnally add R2 to R4 ans store the output in R2. This could be simplified to, adding R1, R3 and storing it in R3, and adding R1 and R4 and storing it in R2. This produces the same result but is mutch more efficient. The efficiency builds up and thus results in an IPC improvement.

Needless to say these two stages perfrom other functions that result in IPC improvements, but I think I've made my point. Needless to say, this statement is, as aI have now proved, flawed. The added pipeline stages do indeed allow for a higher clock frequency scaling in the future but AMD did not add them soley for this purpose.

Anyways, whatever. You like the netburst, I like the k8. :)

#85922

Superb posts, atraxani.

You can, however, have more than one quote per post (so you could have posted one long post as opposed to three "short" ones)

The best way to see how to do this is to preview a post that you're qouting in.

#85929

*Thump*

Jaw hits floor reading atraxani's longest post.

That's really cool stuff, thanks for posting that info!

#85934