Why is My System Benchmark Lower Than Others?

TheBug

Junior Member
Jul 20, 2002
23
0
0
I just ran some benchmark tests on my machine, namely 3DMark2003, PCMark2002, 3DMark2001SE, but the resulting scores were quite low when I compared them to other systems posted on the Futuremark/MadOnion website. Below is my system specs:

Intel Pentium 4 2.4B GHz 533MHz FSB
Asus P4B533-VM 845G Motherboard
Crucial 512MB PC2100 DDR SDRAM x 2
Western Digital Caviar 7200RPM 80GB 8MB Cache
Matrox Parhelia 128MB
Creative SoundBlaster Audigy 2

Now, I have a few theories, but I don't know how much each really caused my low system benchmark.

1. PC2100 vs. PC2700
I read some reviews on Anandtech and other websites, but the difference in performance between the two is usually 5% or less.

2. Value Motherboard
Maybe the motherboard is targeted at the OEM market (esp. since it's a microATX model) and Asus left out some optimizations? I only bought this motherboard because I needed a good microATX board for my tiny chassis (for portability). Was this a mistake?

3. RAM Brand
I've never seem Crucial/Micron RAM used in any benchmark. The majority of websites use Corsair, Mushkin, and some other ones. Could the brand really have that much effect on performance?

4. Video Card Dragging Things Down
I noticed that the PowerDesk software for Matrox Parhelia is taking up about 20MB of RAM. Could it be that the slow video card (bought only for display quality and dual DVI) is dragging down the overall system performance?

To give you a feel of thte pathetic benchmarks, here are the scores:

3D Mark 2003
Score: 913

PC Mark 2002
CPU: 5808
Memory: 4594
HDD: 979

3D Mark 2001 SE
Score: 8025

Of course, there are many other possibilities. But these are the ones I could come up with so far. Feel free to ask me any questions as I will be checking this thread very often. I'm quite unhappy with the benchmarks since I spent quite a bit on this system. So any advice is appreciated.

Thanks in advance.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
Yeah, but when I checked other systems with almost identical specs, their benchmarks are always (and often significantly) higher. I just want to know the reason why.
 

John

Moderator Emeritus<br>Elite Member
Oct 9, 1999
33,944
3
81
When you search for cpu speeds @ 2400MHz, the higher scores are overclocking to get 2400. Hence a higher FSB than you are running.
 

Megatomic

Lifer
Nov 9, 2000
20,127
6
81
Originally posted by: John
When you search for cpu speeds @ 2400MHz, the higher scores are overclocking to get 2400. Hence a higher FSB than you are running.

Good point. I was telling him in another thread that a Radeon 9700 Pro wouldn't hurt either (he asked if he could spend money to fix this "problem").
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
Sorry for the duplicate posting. I'm the kind of person who just can't wait to get a response.

I just checked out the top few systems with specs similar to mine (URL: http://service.futuremark.com/compare?pcm=976091), and the biggest difference is the memory benchmark. I also noted that his system doesn't appear to be overclocked (133MHz FSB). Is this caused by the PC2100 vs. PC2700 difference?

Also, how can our CPU score differ by a 10% margin when our CPUs are, supposedly, identical?
 

John

Moderator Emeritus<br>Elite Member
Oct 9, 1999
33,944
3
81
Originally posted by: TheBug
Sorry for the duplicate posting. I'm the kind of person who just can't wait to get a response.

I just checked out the top few systems with specs similar to mine (URL: http://service.futuremark.com/compare?pcm=976091), and the biggest difference is the memory benchmark. I also noted that his system doesn't appear to be overclocked (133MHz FSB). Is this caused by the PC2100 vs. PC2700 difference?

Also, how can our CPU score differ by a 10% margin when our CPUs are, supposedly, identical?

IIRC, PCMark 2002 has a bug that identifies your cpu and fsb incorrectly on occasion. The URL you are using for reference states My P4 2.380@2.646Ghz , and since his FSB is higher than yours his memory performance will be greater as well.

Again, your system seems to be on par. If you are really that worried about benchmarks, scrap the Parhelia and raise your FSB.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
ah~ touch&eacute;. Didn't catch the OC part.

Now I'm looking into upgrading the RAM to PC2700, since I don't think Asus P4B533-VM supports DDR400 (even DDR333 is unofficial). Are there significant differences between brands?

My main concern isn't video performance, since most of my gaming is done on PS2 or XBox anyway. I do need the dual DVI and output quality from the Parhelia tho. Having a Radeon 9700 would certainly be nice, but it's out of the option for now.
 

John

Moderator Emeritus<br>Elite Member
Oct 9, 1999
33,944
3
81
Unless you overclock your FSB to to the limits of your PC2100 (Crucial will usually do 150ish), there is no need to get PC2700 (166MHz) for your rig right now.
 

Ketchup

Elite Member
Sep 1, 2002
14,559
248
106
You want PC2700 to run the 4:5 CPU:mem ratio on your motherboard, right? You should see a pretty nice performance boost by doing that, although it may mean you need some better memory. People have been hitting very high speeds with Crucial PC2100, so you may wanna try the 4:5 with the mem you have first. Make sure all the memory timings are as relaxed as possible though.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
I just changed the FSB to 150MHz. Good improvement on all benchmarks, but the memory ones are still undesirable.

As I'm going through the BIOS, I have no idea what to set for these:

SDRAM CAS Latench
SDRAM RAS to CAS Delay
SDRAM RAS Precharge Delay
SDRAM Active Precharge Delay
SDRAM Idle Timer

Any suggestions? Also, how can I set the speed for the RAM? For all I know, FSB only applies to the CPU portion. :eek:
 

Ketchup

Elite Member
Sep 1, 2002
14,559
248
106
FSB = Front Side Bus; this is the bus that connects your memory to your CPU; raising FSB will raise CPU and memory speed simultaneously. Your memory scores are low because DDR only has half the bandwidth of the P4 at the same speed; to increase memory bandwidth and overall performnace, you need to utilize your mobo's 4:5 async mem to cpu ratio, although your current memory probably can't handle that if you are already running at 150 fsb.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
ketchup79, I just checked my BIOS, and the only settings available for the CPU:RAM ratio are "auto", "1:1", and "4:3". What now?
 

Wingznut

Elite Member
Dec 28, 1999
16,968
2
0
Originally posted by: TheBug
What now?
I suppose "Don't worry about artificial benchmark scores" isn't an option?

Really... Does your system run your applications well? Then what's the point of caring about 3dmark scores?

 

TheBug

Junior Member
Jul 20, 2002
23
0
0
After some tweaking, the FSB and the RAM are both at 150MHz (1:1), and some other settings:

SDRAM RAS to CAS Delay --> 2T
SDRAM RAS Precharge Delay --> 2T
SDRAM Active Precharge Delay --> 5T

PCMark score improved by 9.6% to 5037. Though still not the 6000+ memory score as seen here, this is good improvement.

I guess the main reason I'm still pursuing this (even though I have a perfectly fine system that runs most games smoothly) is that I really want to know how those people achieve 6000+ (or even 6500+) memory benchmarks in PCMark2002.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
Another interesting observation (at the risk of sounding completely dumb), changing the SDRAM CAS Latency from 2.5T (default by SPD) to either 2T or 1.5T causes the machine to not boot or reboots halfway through WinXP start-up.
 

Ketchup

Elite Member
Sep 1, 2002
14,559
248
106
Hmm, well all boards are different and you may in fact be stuck. I had an Epox 4G4A+ and it had a 4:5 ratio. My Abit has a 3:4 ratio which is great becuase I can run DDR 400 at 150 FSB and my RAM's rated speed, DDR 433, at 162. There may be a workaround for your board to get the other ratios. Your best bet is to ask others who have the board.

Wingznut is right. Looking at benchmarks all day will disillusion you as to how fast your system is. Playing games and doing your normal day-to-day routine will tell you if you need somthing faster.
 

TheBug

Junior Member
Jul 20, 2002
23
0
0
Originally posted by: ketchup79
Wingznut is right. Looking at benchmarks all day will disillusion you as to how fast your system is. Playing games and doing your normal day-to-day routine will tell you if you need somthing faster.

Indeed. Looking at benchmarks all day and not doing CS homework (due tomorrow) is also bad. I'm just really interested in finding out how some people achieve those (extremely) high scores.
 

Ketchup

Elite Member
Sep 1, 2002
14,559
248
106
Latency is another word for delay. Without getting too technical on you, CAS Latency has to do with the amount of time your memory has to process information give to it by the CPU. Your memory is rated at 2.5, so that is where you should leave it. It might be able to handle 2 at 133, but not 150, unless you add some voltage to it, in which case it might.
 

mroptimistic

Senior member
Dec 12, 2002
271
0
0
I didn't pay much attention to my pcmark scores, but now that i go back and check i dont have much problem with my cpu or memory scores, but my hd score is horrid. 832 while my cpu is 5319 and memory is 4257. Im on a p4 2.26, 512 (2x 256) infineon 2100 ddr, abit sr7-8x, ti4200 (64mb) and a d740x 80gb hd. I just checked on a guy with a p4 at 2.35 with 512 of ram and a 8500, and he has a hd score of 11440. Not all of them are that high, but still, how do i end up with only 832?