CPU buying help

This forum is for help with and discussion about your video hardware.

CPU buying help

Postby NeoQuixotic » Thu Dec 23, 2004 11:00 am

To help anyone who is going to ask about what CPU to buy and curious about the actual performance, I thought I should point out a very good link. It lists all major CPUs from 1995 to early 2005, and benchmarks them all in gaming, video and audio encoding, applications, and synthetic situations. The article is in 2 parts thus the two links. Part 2 contains the benchmark results. Enjoy! ^_^

Part 1
http://www.tomshardware.com/cpu/20041220/index.html
Part 2
http://www.tomshardware.com/cpu/20041221/index.html
User avatar
NeoQuixotic
Master Procrastinator
 
Joined: 01 May 2001
Location: St. Louis Park, MN
Status: I haz job!

Postby the Black Monarch » Sun Dec 26, 2004 10:46 am

I guess it's pretty sad that in a few benches, a 3.4 GHz Prescott on a 200mhz system bus with DDR-400 whipped a 3.46 GHz Extreme Edition on a 266mhz system bus with DDRII-533.

...just how the hell did that happen, anyway?
The only .org member to donate $1,500 and still have a donation status of "total leech"
User avatar
the Black Monarch
 
Joined: 09 Jul 2002
Location: The Stellar Converter on Meklon IV

Postby RootHubController » Tue Dec 28, 2004 2:59 pm

the Black Monarch wrote:I guess it's pretty sad that in a few benches, a 3.4 GHz Prescott on a 200mhz system bus with DDR-400 whipped a 3.46 GHz Extreme Edition on a 266mhz system bus with DDRII-533.

...just how the hell did that happen, anyway?


Easy. The EE has too long a pipeline. The pipeline would be optimal for a 5GHz proc, but instead, it's on a 3.4, and requires redundant instruction sets to accomplish work that could have been handled already.
User avatar
RootHubController
 
Joined: 07 Aug 2004

Postby NeoQuixotic » Tue Dec 28, 2004 3:53 pm

Looking at the VOB to XviD benchmarks, my poor system with a 2.53 Ghz P4 at 533 mhz fsb (actually 133 mhz) and PC800 RDRAM would be longer than the listed 2.53 ghz P4 benchmark at 4:09. That means I'm running two times slower than the new P4s and AMD 64 FX-55 at 2:23 and 2:32! However, who can just throw away ~$1000 to the new processors? I can't, I'll tell you!

However, on most benchmarks my computer (a Dell, lol) still is plenty fast enough. I say if you have anything above 2.5 ghz (or AMD equivalent), stick with it for another year or 6 months at least. I think I can wait a couple more minutes for a video to encode and a .rar file to compress. Is a 2 minute (or less) speed boost or 200 more fps in Quake 3 worth $1000? Not really, unless you're a speed freak or working with gigantic files to compress 24/7, 365 days a year. Ok, I'm ranting, sorry. I just hate seeing people buy a new processor/video card/ram every month just because "it's faster". I'm just trying to say don't waste your money on a new cpu/computer for a performance boost you probably won't notice in most situations.
User avatar
NeoQuixotic
Master Procrastinator
 
Joined: 01 May 2001
Location: St. Louis Park, MN
Status: I haz job!

Postby Otohiko » Tue Dec 28, 2004 6:54 pm

Yea, and in many cases you have to add a new motherboard into the equation, such as in mine.

I'd like to switch to a 64-bit processor eventually, but frankly, I can't swing the cost of processor+motherboard (and my current 2400+ is practically at the limit of my motherboard's supported clock speed). :(

I'll probably wait at least a year or a year and a half before I even think of upgrading this.

Thanks for the useful info though :)
Otohiko
 
Joined: 05 May 2003

Postby RootHubController » Wed Dec 29, 2004 4:25 am

I'd like to switch to a 64-bit processor eventually


I'll be more inclined to do it when there's software that supports and exploits the full advantages fo the 64-bit architecture. Yes, it's backwards compatible, but you're not recieving any benefits.
User avatar
RootHubController
 
Joined: 07 Aug 2004

Postby Scintilla » Wed Dec 29, 2004 10:25 am

RootHubController wrote:
I'd like to switch to a 64-bit processor eventually

I'll be more inclined to do it when there's software that supports and exploits the full advantages fo the 64-bit architecture. Yes, it's backwards compatible, but you're not recieving any benefits.

Well, they <i>are</i> also better at doing the same old 32-bit operations, aren't they?
ImageImage
:pizza: :pizza: Image :pizza: :pizza:
User avatar
Scintilla
(for EXTREME)
 
Joined: 31 Mar 2003
Location: New Jersey
Status: Quo

Postby Kalium » Wed Dec 29, 2004 10:28 am

Scintilla wrote:
RootHubController wrote:
I'd like to switch to a 64-bit processor eventually

I'll be more inclined to do it when there's software that supports and exploits the full advantages fo the 64-bit architecture. Yes, it's backwards compatible, but you're not recieving any benefits.

Well, they <i>are</i> also better at doing the same old 32-bit operations, aren't they?

From what I understand, this is often the case.

Besides, RootHubController seems unaware that there are already quite stable 64-bit OSs. They just aren't Microsoft OSs.
User avatar
Kalium
Sir Bugsalot
 
Joined: 03 Oct 2003
Location: Plymouth, Michigan

Postby trythil » Wed Dec 29, 2004 12:40 pm

Kalium wrote:
Scintilla wrote:
RootHubController wrote:
I'd like to switch to a 64-bit processor eventually

I'll be more inclined to do it when there's software that supports and exploits the full advantages fo the 64-bit architecture. Yes, it's backwards compatible, but you're not recieving any benefits.

Well, they <i>are</i> also better at doing the same old 32-bit operations, aren't they?

From what I understand, this is often the case.

Besides, RootHubController seems unaware that there are already quite stable 64-bit OSs. They just aren't Microsoft OSs.


There's one point he still has: even OSes that work in 64-bit mode on x86-64 processors haven't yet taken full advantage of the system. Oh, sure, you get nice things like the ability to easily address >= 1GB of RAM without hacks or slowdown, and recent compilers are able to schedule code effectively specifically for x86-64, but I've only seen a couple of applications that actually take advantage of the system.

One thing I'd like to do is benchmark two Gentoo setups running on the same Athlon64 machine and see how much of a difference I get out of running typically CPU-intensive operations that tend to move around a lot of data. I'm running Gentoo x86-64 right now but without any sort of reference point I can't say much about it.
This is a block of text that can be added to posts you make. There is a 512 character limit.
trythil
is
 
Joined: 23 Jul 2002
Location: Llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogoch

Postby trythil » Wed Dec 29, 2004 12:47 pm

trythil wrote:
Kalium wrote:
Scintilla wrote:
RootHubController wrote:
I'd like to switch to a 64-bit processor eventually

I'll be more inclined to do it when there's software that supports and exploits the full advantages fo the 64-bit architecture. Yes, it's backwards compatible, but you're not recieving any benefits.

Well, they <i>are</i> also better at doing the same old 32-bit operations, aren't they?

From what I understand, this is often the case.

Besides, RootHubController seems unaware that there are already quite stable 64-bit OSs. They just aren't Microsoft OSs.


There's one point he still has: even OSes that work in 64-bit mode on x86-64 processors haven't yet taken full advantage of the system. Oh, sure, you get nice things like the ability to easily address >= 1GB of RAM without hacks or slowdown


I should clarify that.

On a 32-bit processor, each process can address a maximum of 4GB (2^32) of memory. However, some operating systems enforce a kernel/user separation for each process's virtual address space. I know Linux does this, and I'm pretty sure Windows does, too.

On Linux the default is a 1GB/3GB split; i.e. the first 1GB is kernel address space, and the 3GB is user address space. As the kernel can access only up to 1GB of memory in this scheme, you are effectively limited to 1GB of physical RAM. (Those are the terms I'm used to thinking in, thus the >= 1GB statement.) With a 2GB/2GB split it's a 2GB limit. And so forth.

I suppose what I wrote would have been better phrased "you get freaking huge address pointers", and anyone interested in 64-bit would have immediately understood the implications of that.
This is a block of text that can be added to posts you make. There is a 512 character limit.
trythil
is
 
Joined: 23 Jul 2002
Location: Llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogoch

Postby the Black Monarch » Sat Jan 01, 2005 10:40 pm

RootHubController wrote:
Easy. The EE has too long a pipeline. The pipeline would be optimal for a 5GHz proc, but instead, it's on a 3.4, and requires redundant instruction sets to accomplish work that could have been handled already.


Last time I checked, the EEs had bigger caches, not bigger pipelines...
The only .org member to donate $1,500 and still have a donation status of "total leech"
User avatar
the Black Monarch
 
Joined: 09 Jul 2002
Location: The Stellar Converter on Meklon IV

Postby NicholasDWolfwood » Sun Jan 02, 2005 12:47 am

the Black Monarch wrote:
RootHubController wrote:
Easy. The EE has too long a pipeline. The pipeline would be optimal for a 5GHz proc, but instead, it's on a 3.4, and requires redundant instruction sets to accomplish work that could have been handled already.


Last time I checked, the EEs had bigger caches, not bigger pipelines...


Last time I checked, all Pentium 4s had longer pipelines, even the EE.
Image
User avatar
NicholasDWolfwood
 
Joined: 30 Jun 2002
Location: New Jersey, US

Postby Scintilla » Sun Jan 02, 2005 12:53 am

NicholasDWolfwood wrote:
the Black Monarch wrote:
RootHubController wrote:
Easy. The EE has too long a pipeline. The pipeline would be optimal for a 5GHz proc, but instead, it's on a 3.4, and requires redundant instruction sets to accomplish work that could have been handled already.

Last time I checked, the EEs had bigger caches, not bigger pipelines...

Last time I checked, all Pentium 4s had longer pipelines, even the EE.

Last time I checked, Prescott's pipeline was at least 10 stages longer than Northwood's, so basically some P4s (including the EEs) are more equal than others. :)
ImageImage
:pizza: :pizza: Image :pizza: :pizza:
User avatar
Scintilla
(for EXTREME)
 
Joined: 31 Mar 2003
Location: New Jersey
Status: Quo

Postby Knowname » Sun Jan 02, 2005 7:34 pm

Scintilla wrote:
NicholasDWolfwood wrote:
the Black Monarch wrote:
RootHubController wrote:
Easy. The EE has too long a pipeline. The pipeline would be optimal for a 5GHz proc, but instead, it's on a 3.4, and requires redundant instruction sets to accomplish work that could have been handled already.

Last time I checked, the EEs had bigger caches, not bigger pipelines...

Last time I checked, all Pentium 4s had longer pipelines, even the EE.

Last time I checked, Prescott's pipeline was at least 10 stages longer than Northwood's, so basically some P4s (including the EEs) are more equal than others. :)

I'd continue this but I'd feel guilty.
User avatar
Knowname
 
Joined: 16 Nov 2002
Location: Sanity, USA (on the edge... very edge)
Status: Indubitably


Return to Video Hardware Discussion

Who is online

Users browsing this forum: No registered users and 1 guest