Is overclocking really mean equal too?

[TheGoatMan]

From what I've seen people are comparing overclocked processors to higher clocked factory spec processors. Now from what I understand higher clocked newer processors have better technology and are more efficient. So can you really compare a 2500+ overclocked to 3200+ speed? Like for example the 3200+ 400mhz fsb would obviosly be better. So can anybody clear this up? Because I see this like modifying a honda civic to be as fast as a corvette but lacking the refinement.

 

[stevieblunder]

Overclocking is a bit like turbocharging a motor. The problem is that it can easily overheat and burn out, although it will most likely start to make digital mistakes and crash before the CPU becomes toast. The biggest problem is the little errors that are not obvious until you discover them later. ie like your homework or bank balance file is toast.

Newer faster CPUs use smaller transistors that use less energy to switch, so thay can run faster before they overheat.

My feeling is that overclocking is simple a very bad idea. You are better off to spend your time and money on newer hardware that has been tested by professionals. Your time is valuble, not that of your CPU. If your CPU is taking a lot of human time, ie minutes, to finnish a job, you want to look for better software. One thing you should never compromise is reliability.

 

[SYAR2003]

This is bull when it comes to athlon 2500
it runs 400fsb with the same voltage (default 1,65v)
so there is no way it can get damaged at that voltage.
(some that's taking it far beyond that volt to get 2,3-2,6ghz i would agree )
Btw serios overclockers test their cpu far much more than
everyone else does .As for file corruption that was in the old days when the pci bus didn't get locked at 33mhz.

 

[dakota81]

Well, in your particular case, the 2500+ is the exact same design & technology as the 3200+, it's just labelled at a different speed; as they are both based off the Barton core.

If you took a Thouroughbred (spelling?) core that is not a direct comparison to the Bartons because the cache sizes are different thus affecting performance in different ways in different software.

But if you had like a Palomino compared to a Thouroughbred, even though the transister technology is different, the components on the cpu die & the digital logic is exactly the same so an overclocked Palomino can be equal to a higher clocked Thouroughbred.

Your specified case is the easiest of all, a Barton compared to a Barton; when you overclock the 2500+ to the same fsb & core GHz, they are exactly identical.


The only thing to watch out for is, a particular 2500+ might be labelled that because there might have been some anomaly in the manufacturing processes that the cpu cannot handle the 3200+ settings, but is just fine to be sold as a 2500+.

 

[TheGoatMan]

From what I heard the thouroughbreds cache was faster. On cache benchmarks that I ran the thouroughbred scored exceptionally higher than the barton. Also the single channel bandwidth on my KT600 was higher than dual channel on the n-force board I compared it too. As far as CPU multiplication benchmarks the 2500+ overclocked to 3200+ scored only 8300 m/flops. Meanwhile the 2600+ went as high as 9300 m/flops clocked at the same speed. So which is better? The 2500+ barton or the 2600+ tbred(unlocked). Because as far as benchmarks (SiSandra and PCmark02) the 2500+ wasn't ahead at all. I used all the same equipment except for:
2500+ was tested on an Abit nf7-s
2600+ was tested on a MSI KT6 delta lsr.

I don't mean to sound ignorant but shouldn't the barton+nforce be dusting the latter?

 

[SYAR2003]

http://www6.tomshardware.com/cpu/20030623/p4_3200-07.html

Notice the 2 last processors.

 

[TheGoatMan]

Notice the bus speed. Its a 266.

 

[TheGoatMan]

There was a significant difference between the 266 and the 333mhz 2600+. Look into it a little more.

 

[dakota81]

You're finding out that flops and whatever other synthetic benchmarks you're using do not tell the full story of a cpu's performance. Real-world applications are what counts.

Yes, a smaller cache is faster because there is less data to keep track of; however, while a tad slower to access cache, having the larger cache reduces the number of times the program has to call out to the main memory, which is painfully slow when talking in terms of cpu speed. This is a tradeoff in performance, and the benifits of 512k L2 cache is shown very clearly when looking at actual software perfromance.


If you had a benchmark program that just accessed the same memory location from the cache a million times in a row, the T-Bred will appear to be faster than the Barton. However, not a single real world application would ever operate like that.

 

[SYAR2003]

Okay bad example since it was a 266.

Benchmarking the memory bandwidth of the system, we find where the extra L2 cache benefits the PC
Cache helps your PC's performance by storing information that the processor may require often. Since cache runs on the CPU at the speed of the CPU, the retrieval of this information is much quicker than if the CPU had to resort to asking the slower system RAM (like your DIMM's, for example) to give up the information the CPU needs or, if the information is not available in the system RAM, go fetch it from an even slower source (hard drive, CDROM, etc... )

Check out the memory benchmark
here you se that the benefit of 512k on the 2500
versus 2600 t-bred with 256k.

There is a reason for implementing this
it not just for show. Intel did the same for the P4's
internal L2 cache

http://www.thetechboard.com/reviews/barton-2.php

 

[dakota81]

How about this; if you look at that benchmark link above, the 2700+ is a T-Bred at 2166MHz and cannot keep up with a Barton 2800+ at 2083MHz. Real world applications favor larger caches.

 

[cdogg]

dakota's explanation couldn't be better.

Yes, a 512KB cache is more favorable than 256KB in many applications today. Keep in mind that 6 or 7 years ago, that wasn't the case, which is also why large caches were really only experimented on with CPU's like the XEON. At the time, their advantage in performance was only clearly shown in server environments and didn't run at full speed with the processor.

Of course, desktop software has since come a long way. Real-world applications have now been showing the benefits of larger cache. However, it will be sometime before we know how much is too much. If I'm not mistaken, there are plans for 1MB caches in soon-to-come releases from both Intel and AMD. Eventually, however, it will hit a point of "diminishing returns" making it less cost effective than performance driven to keep increasing L2 cache.

Well, enough babbling for now!


~cdogg
[tab]"The secret to creativity is knowing how to hide your sources"
[tab][tab]- A. Einstein

 

[TheGoatMan]

I think the point of synthetis benchmarks is to tell differences between abilities. I seriously saw no difference at all between the two processors. As far as games and multitasking. So really synthetic benchmarks are the only basis of comparison between two processors as close as the 2500+ and the 2600+. Also which memory benchmark are you talking about? I ran all related benchmarks and had the same results. You see I had the opportunity to try out whatever equipment I wanted when I built my computer. I just wasn't very impressed with the fact that the barton was supposedly superior yet seemed not to out perform. In my case at least. I'll have to try out a few other processors and see the differences.

 

[SYAR2003]

Memory score PCmark
xp2500 at 1,833 4376
xp2600 at 2,083 3966

And this utest is done with same mb/ram/hdd/video
The increased memory handling on the Barton is all due to larger cache .

In measuring pure clock cycles of course
the 2600 t-bread is faster as it runs over 200mhz faster.


Most benchmarks upto today isn't very good
at mesuring cpu L2 cache .

 

[TheGoatMan]

Where did you get those scores? I got 5883 on my PC. My CPU score was 6889.

 

[TheGoatMan]

Actually here are the new scores .

http://service.futuremark.com/servl...ails&UID=13167282&projectType=7&XLID=1630155e

 

[TheGoatMan]

Nevermind. Link doesn't work.

 

[SYAR2003]

The scores if from the review posted earlier.
(

http://www.thetechboard.com/reviews/barton-2.php)

It's no point compairing to yours.
Then you have to take your cpu out and install a 2500 Barton and test with same:
mobo/ram/bios settings/video/harddisk etc.
The chipset and memory and bios settings all reflects
in a test.

 

[SYAR2003]

btw your link don't work.

 

[TheGoatMan]

These benchmarks are confusing. Your link with the scores were way lower than what I got. What I don't get is if benchmarks are inaccurate sometimes and aren't a basis for comparison then why go by them anyways? Actually when I ran both systems at default speeds they ran almost exactly the same scores. I mean within 100 points every test. But the fastest I could push the barton was 2.2ghz. (11x200) Meanwhile the thouroughbred could easily run at 2.5ghz. Which made a huge difference in scores. Raising them sometimes as high as 1500 points(or mflops) So are these just tests that rely mostly on overall speed rather than ability?