Sempron 2500+ better than Athlon XP 2800+?

Hi everyone...

A few weeks ago I started thinking about upgrading my CPU. I had a Sempron 2500+ (socket A) that I bought three months ago, and with my 6800LE (12 pipes, 350/800) scored a weak 7620 on 3DMark03 (shareware version, with fixed options). I have a NF7-S2 board with SATA hdd and 2x512 PC3200 RAM (dual-channel). The bottleneck is CPU, right?

So I decided to go with 256K more cache and 50$ higher value - Athlon XP 2800+ (Barton, 12,5*166, also socket A). Yesterday I turned on the 3DMark03 and was totally expecting the score to be in 8/9K range, like every normal person has.

7325. It was 300 points lower than with Sempron. After that I tried some games like Generals: Zero Hour and it ran the same as with my old CPU (no improvement).

So my question is: how come I didn't get at least the same results with Athlon XP than with Sempron?

a) didn't change any other hardware
b) both are socket A
c) both run on FSB 333
d) didn't OC either of them
e) SANDRA test and BIOS say that I really do have a 2083GHz machine, with 512K cache, running just fine; but SANDRA's benchmark of my computer is closest to (yeah, you guessed it!) an average Sempron 2500+ computer...

Please help, thanks!

 

Your bottleneck was more that weak video card than the CPU. As for 300 points difference, thats so little a difference, you'd never noticed it. The Barton should be faster than the Sempron. Maybe its just me, but whenever I used 3dmark, I always found there to be several hundred 3d marks fluctuation each test. It was never the exact same number.

 

Thanks for the reply, Adrynalyne!

I done a few tests with 3DMark03, but they were all still a few hundred points below Sempron 3DMark03 score. Anyway, the score itself is quite irrelevant: if I could get the CPU to work better than Sempron in games/applications I'd settle for 2000 points lower score!

Is there maybe something I could do in BIOS or in some way check what's going on?

 

I wish I had an answer, and reading your post again, I wonder if the video card has anything to do with it at all...it is being recognized as a 2800+ w/ 512k L2 cache, right? If L2 got switched off in the bios, that could definately attribute to the performance deficiency.

I'd try one more benchmark to backup the findings of the others.

http://www.spec.org/osg/cpu2000/ is specifically for CPUs. Maybe see what it says?

 

Hmmm... I went to the cpu2000 page, and didn't find any downloads cheaper than 125$ - and for that money I can buy a new CPU!

Right now I'm at work so I can't see if L2 cache is on - as soon as I get home I'll examine the BIOS with more detail. And I'll try to find some more benchmarking programs. Do you have some other favorite?

 

Oops, my bad.


http://majorgeeks.com/downloads4.html

There are quite a few CPU benchmarks listed here. I have no favorites, as I always find benchmarking tools depress me

 

A good one is the built in cpu test in sandra it will show the performance of your cpu compared all other cpu's

http://majorgeeks.com/SiSoftware_Sandra_d92.html

Also go into your bios and see if your multiplier is unlocked up and down you on your barton you never know,I wouldnt be suprised if you can get it up to 3200 ish speed as there is little difference between chips from 2500 to 3000 unless they are 200mhz fsb

 

Rikky,

my Barton's multiplier is unfortunatelly - locked, and I can chage only the FSB in the BIOS. I'll try to raise it to 200MHz to work at the same frequency as RAM.

I was thinking - can power supply have anything to do with lower performance? I have 300W, and 2 CD/RW drives, SATA HDD and additional VGA card power supply + add. cooler power supply.

 

considering you're running a power hungry 6800 series, your PSU could definetly be a limiting factor overall-- however the Athlon XP Barton only uses ~2-3W more power consumption than the Sempron, so I dont think thats the issue with your benchmark drop.

as Adrynalyne suggested, perhaps your L2 cache is failing to be recognized by windows HAL, and defaulting to 256k. you can verify this by opening your registry editor and navigating to the following. If you find the Value to be 0 (which is default, meaning 256K), change it via Decimal, to 512

HKEY_LOCAL_MACHINE\System\Current Control Set\Control\Session Manager\Memory Management
"SecondLevelDataCache"= REG_DWORD 0

 

Thanks for the advice, InYearsToCome!

Btw, I checked for L2 cache option in BIOS, but haven't found anything 'L1 / L2 / cache' related... The only thing I found was C1 disconnect, and turned it off (it was on auto). As soon as I get home from work, I'll check the registry.

 

Checked the registry - it was like you wrote: the value was at 0, so I turned it to 512 (decimal). After that I ran Sandra's Cache & Memory Bandwidth benchark, and was kinda suprised by the results. I put another Athlon XP 2800+ into benchark, and my cache/memory was doing great at 4/8/16K, but as the nubers got higher, my graph started to go downward - a lot! Around 256K, the 'normal' Athlon had 3 times better results... I mean, come on! This means I have a big problem with CPU cache, right?

Now I downloaded Aquamark3, PCMark05 and AMD N-bench31 and will try to find out if my L2 is working properly. Every memory scanning program says I have L2 and it's 512K (like cpu-z).

Also, I cannot find a 'L1/L2 cache option' in my BIOS. Could it be that my NF7-S2 doesn't have that option or I just can't find it?