P4 150% FASTER THAN A T-BIRD!!!

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Of course the P4 has a rotate instruction--I think they mean it just isn't as hard-wired as before. Not that that really matters; very little of the x86 is fully hard-wired anyways. It's been part microcode since the beginning.

If they said instead that the P4 had a <i>slower</i> rotate instruction, then I could see that it matters.

Kelledin

bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?
 
AMD will milk the x86-64 as much as they can, but it simply does not perform anywhere near IA64.
Click to expand...
Errr...that seems like a rather premature assumption. AFAIK the Hammer isn't even <i>out</i> yet--there's only a software emulator, which isn't all that indicative of final performance. We'd probably all better wait and see before making such definite statements.

Of course, if you have a link to actual hardware benchmarks, I'd love to see it.

Kelledin

bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?
 
IA64 is not intended for the desktop. Maybe one day, but that day is still very far away. I dont even think intel has ever said IA64 was intended for the desktop eventually.
--
Second, they calculated that there would be a time after the introduction of the 64-bit CPU where it would not be a great performer on current software, <snip> They intend to take advantage of this 'twilight'
--

Remember the Pentium Pro ? That was a very long time after 32 bit had become a reality with the 386. Even then 16 bit mattered.. a lot. Going from 16 bit to IA32, wasnt nearly as challenging as going from IA32 to IA64.
Today, we are beginning to see the first 64 bit IA64 cpu. No software yet though, well hardly anyway. I think intel is making the same mistake all over again. Itanium runs 32 bit software *really* slow (P100-like). This twighlight might acually last a year or 5, maybe 10. During this time, running mixed 64/32 bit software on a IA64 is probably not an option.. this rules out Itanium, unless for some very specialized tasks for which optimized IA64 software is out. Why go for IA64 then anyway ? Why not established (64 bit) RISC architectures ? Unless IA64 would be that much faster, which remains to be seen.

Meanwhile, P4/5/+ will remain 32 bit, so applications that require large (I mean, HUGE) memory access, are not well suited for IA32 either.

It makes perfect sense to take advantage of this. I think AMD has taken a very wise approach with X86-64. You get full speed 32 bit, which will be necessary for a long time (especially on the desktop), and you get "free" 64 bit support if required (mainly in the server area for at least 5 years). From what I read, convertion to x86-64 would be a whole lot easier than to IA64.

I dont really see IA64 compete with x86-64. IA64 is much more a competitor to existing 64 bit RISC architectures, since it has nearly nothing in common with x86. x86-64 on the other hand, seems like a good way to extend the life of x86.
 
Ohhhh, a little research clears it up...I remembered reading something about this.

Intel had a heavily optimized high-speed barrel shifter in the 386 up to the P3. It got scrapped in the P4 and replaced with a much slower shift/rotate execution unit. That <i>would</i> be rather bad--the high-speed barrel shifter was at the core of integer instructions ROL, ROR, SHL, SHR, MUL, IMUL, and probably quite a few others as well. Shifting bits is an extremely common and oft-repeated operation in a CPU.

This is an old URL, but I really should provide a link just to be fashionable:

<A HREF="http://www.emulators.com/pentium4.htm" target="_new">http://www.emulators.com/pentium4.htm</A>

Kelledin

bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?
 
well, for me performance is important in real world apps, and SSE2 hasnt come into real world yet.

and Raystonn how can you make such a presumptuious statement like this? any if the x86-64 processors havent even come out yet, and you say they dont perform near IA-64? you need to visit some place like http://www.x86-64.org or http://www.amd.com

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>
 
Their whole 64-bit platform is based on 64-bit meaning only more memory, while stating that most applications should not bother to be written to take advantage of a 64-bit CPU.
Click to expand...
It extends to more than memory. The eight general-purpose registers are extended to 64 bits, and eight other 64-bit registers are added. They are all capable of 64-bit integer arithmetic in the same manner that EAX, EBX, ECX, EDX etc. are capable of 32-bit arithmetic. SSE is 128-bit already--no immediate need to extend that.

The only instructions that are "invalid" in AMD's 64-bit mode are instructions that (a) were rarely/never used, or (b) are pretty much pointless in 64-bit mode anyways. With the ASCII/decimal adjust instructions, it's a little of both--who needs BCD with 64-bit integers?

It [the paradigm shift] happened when we moved from 16 to 32 bits.
Click to expand...
AMD's approach is the <i>exact same approach</i> that Intel took going from 16-bit to 32-bit. I think Intel has been making the 64-bit paradigm shift harder than it has to be.

Kelledin

bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?
 
this thread is not for 64 bit discussions.

lets sum it up, P4 is 150% faster than Athlon Tbird... if the app uses SSE2 optimisations. if it doesnt, then its <b>150% slower</b>, since Tbird running at 1.33GHz outperforms 1.7GHz of P4!

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>
 
it's too bad that the optimized software is not as blind as you are, hooray for new optimized software!

so, go back to your 16bit DOS compatability mode and stfu!



"AMD/VIA...you <i>still</i> are the weakest link, good bye!"
 
Thank you Captain Obvious.. Just because IA32 is still in all CPUs doesn't mean they can't modify it.

---------
I am the first and only one with a 16MB GeForce2 GTS graphics card! :smile:
 
Guess that Emulators article makes clearer why P4 preforms so horribly in RC-5 cracking: P4/1.7Ghz = Cel700Mhz or Duron ~550Mhz. According to Distributed.net speeds database, P4 is doing little better in OGR contest, but being 2 times slower than Athlon in Mhz vs. Mhz isn't nothing to talk about either. Wonder does P4 performs so pathetically on scientific calculations/other places where numeric crunching matters, too...
 
"AMD TBIRD 1.33
Asus A7M266
Vapochill
nVida Geoforce3
Apacer 256MB DDR266
Seagate X15

??? I'm going to agree with meltdown here, why in the world would you get a vapochill just to play ames? AND who would buy a vapochill and NOT OVERCLOCK!?!? I don't understand."

- Where, exactly, did AMDMeltdown mention anything about the Vapochill?

- Do you know what a Vapochill is? Why wouldn't you buy one, "just to play ames"? Do you overclock so that you're office applications run better...? I don't...

- Who said anything about NOT overclocking? Just that I don't state my current MHz doesn't mean that I haven't overclocked...


Please try and stick to the subject.


/J

AMD TBIRD 1.33
Asus A7M266
Vapochill
nVida Geoforce3
Apacer 256MB DDR266
Seagate X15
 
"if all you do is games"

The keyword here is "MAINLY", I MAINLY use it for games. I occationally use it for some 3D work as well, I guess all games are banished from your precious P4?


"just how much did you have to fork over for that rig? rotflmao!!! just to play games???"

Much less then what you would have to pay for a P4 setup with equivalent performance...


Do you pick on guys/gals buying a Ferrari as well?
"Did you pay $100.000 for a car and you only drive it?!"

Oh well...


/J

AMD TBIRD 1.33
Asus A7M266
Vapochill
nVida Geoforce3
Apacer 256MB DDR266
Seagate X15
 
Um I have to agree as well, you're nuts if you get a vapochill and don't overclock..... unless you're really running a TBird 750 and it's at 1.33. :wink:

---------
I am the first and only one with a 16MB GeForce2 GTS graphics card! :smile:
 
*Sigh*

<Irony>Alright, I DON'T overclock! I just bought the Vapochill cause the case looks cool....</Irony>

/J

AMD TBIRD 1.33
Asus A7M266
Vapochill
Elsa Geforce3
Apacer 256MB DDR266
Seagate X15
 
right...

so you have plenty of money i guess... first, you get a P4 at incredibly high price, then wait for the optimised software to come around and again pay for it...
cool!

you would be waiting forever for all the aps to be optimised for P4.

as a core programmer (I enjoy progamming in assembly, and code optimisation is my special interest, for your info) I know how optimiastion can work wonders with apps, but that doesnt come without price.

i remember when the 486 optimisations were preferred over Pentium ones, for retaining backward compatibility with reasonable performance. here you are using an entirely new insrtuction set, so where does this optimisation come in? its like comparing a bike with a bicycle. the bike has a engine, the bicycle doesnt.

and what you, me and everybody is concerned with is how the chip would fare in our day to day apps, dont you? if it does offer 1000% performance improvement in some apps which are hardly 5-10% of my usual work, then that makes no difference to me. and such SSE2 "optimised" software is not going to be compatible with any other processrs, neither from intel nor from any others. then why should I write code for it?

as for 16 bit DOS, its still omni-compatible, and P4 can run it too. just a wee-bit slower.

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>
 
...
"and such SSE2 'optimised' software is not going to be compatible with any other processrs..."
...

The Intel C/C++ Compiler builds applications that incorporate SSE, SSE2, MMX, and the extra Pentium Pro/2 instructions as well as instructions that are not CPU-specific. It then chooses which instructions to use at runtime based on your current processor. This compiler plugs right into VC++ with only one checkbox that tells the IDE to use the Intel compiler rather than the VC++ version. All you really need to do to make 'optimized' software is click the 'rebuild all' option. The software will work on all processors using the most optimized instruction set for whatever processor you are running.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =
 
There are other "cool" cases out there that cost a whole lot less. But, if you've got that much money, how 'bout loanin' me a few hundred? LOL

Seriously, though, since evryone is makin' fun, why don't you overclock. If you've got an AXIA or AVIA chip, you could probably get some seriously high numbers.

No man stands so tall, as when he stoops to tweak his rig.
 
girlish, pls spare me the bs, you're a two bit programmer who can't even program 32bit code for beans.

so stay in your 640k world of fantasy because the rest is moving towards 64bit and SSE2 coding. good bye!

"AMD/VIA...you <i>still</i> are the weakest link, good bye!"
 
SSE2 is just a red herring used by Intel spin doctors. First of all, there is no excuse to design such a crappy core that without SSE2 it cannot beat lower clock speed Athlon or even PIII. Secondly, practically all the software now ARE NOT SSE2 optimized. Who is going to give me a free upgrade when some of them are finally optimized?? No one. If I bought a program XYZ last year that is run slower on P4, I have to pay again for a new version if and when XYZ is optimized. Why should consumers take that? We want the best performance NOW, not flimsy promises that supposedly delivered by others (software companies, not Intel).

I would have no problem with P4 if it can deliver just slightly better performance (at higher clock speed don't forget) right now and promise much better performance later with SSE2. But, it is NOT the case. Thank about it, Athlon can be made SSE/SSE2 compatible. In fact, Athlon-4 is already SSE compatible. So where is the P4's advantage? Artificially high clock speed (by very inefficient extra long pipeline)?? I don't think so.

**Spin all you want, but we the paying consumers will have the final word**
 
I suppose that the slot A Amd users where "tricked" as well, eh?
you know wusy? I am convinced-more than ever; that you are not getting a proper education down under.

"AMD/VIA...you <i>still</i> are the weakest link, good bye!"
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom