UT2003 & Memory Bandwidth?

[THUGSROOK]

i kinda came across this by accident...
none-the-less, here's my results:

1024x768

2608mhz 163fsb 3:4 435ddr cas 2.0-3-2-7 = 182.7 / 69.3
2640mhz 165fsb 3:4 440ddr cas 2.5-3-3-7 = 182.3 / 68.9
2720mhz 170fsb 1:1 340ddr cas 2.0-2-2-6 = 182.5 / 69.4

this is interesting news for all you UT2003 tweakers out there

 

[XBoxLPU]

WOW

Interesting results

What resoultion ?

 

[zsouthboy]

yup i get same kind of results when mem ocing, memory bandwidth don't mean sh!t for this game all within 1 fps usually

 

[THUGSROOK]

resolution was 1024x768

yep ~ UT2003 likes mhz and low cas settings, it doesnt care about actual bandwidth at all! :Q

(thats pretty strange, dont ya think?)

 

[fluxquantum]

hmmm that's interesting.

 

[oldfart]

Interesting. All scores are the same. 2608 MHz/2720 MHz makes no difference either. What does the test look like @ 640 x 480?

 

[THUGSROOK]

oldfart ~ yes i posted the test scores that way on purpose.
i wanted to show how different configs could get (roughly) close scores.
some ppl might wanna rethink their configs if they are a complete UT2003 nut
(low cas ~ anyway you can get it! even if it means 1:1)

640x480? im not sure i see your point.
this is kinda more about tweaking UT2003 for better gameplay.....
ppl play in 640?

 

[oldfart]

The reason for 640 x 480 is it takes the stress off the video card. The lower res is more of a CPU test, less of a video card test. Nobody plays 640 x 480 (I hope not anyway). I was just kind of curious to see how the results would look.

 

[rogue1979]

I am not suprised, I posted a month or two ago similar results when comparing an Athlon DDR system with a Tualatin SDRAM system at the same clock speeds. All that memory bandwidth looks great in Sandra, but in gaming it doesn't do much good. I had a highly overclocked Radeon 9000 @300/250 and found that upgrading from a 1360MHz Morgan to an XP1600+ @1800MHz didn't give any improvement in gaming framerates. Of course 3DMark showed a considerable boost but it sure didn't show up in any games. Bah! Those synthetic benchmarks don't mean crap:disgust:. I'll stick my neck out and venture to say (even though I don't have one), that even with a Radeon 9700 bandwidth and cpu speed is highly overated for gaming performance.

 

[oldfart]

1024x768

2400 MHz 150 FSB 3:4 400 DDR cas 2.5-3-3-7 = 159 / 63
2400 MHz 150 FSB 1:1 300 DDR cas 1.5-2-2-5 = 157 / 62

640 x 480

2400 MHz 150 FSB 3:4 400 DDR cas 2.5-3-3-7 = 184 / 63
2400 MHz 150 FSB 1:1 300 DDR cas 1.5-2-2-5 = 178 / 62

The Flyby is more affected by video card performance, but the botmatch is entirely a function of CPU power. Resolution and mem BW make no difference in the botmatch, but do have an effect on the flyby.

I'd be interested to see R9700 results.

 

[THUGSROOK]

this is kinda hard to explain ~ but i think oldfart will understand this since ive been talking about this for quite some time now.....

if i downshift my mem ratio to 1:1 and start working my way up the fsb ~ UT2003 is the 1st benchmark to beat its own scores from 3:4 mode, in just 7fsb!
all my other benchmarks need 10-15fsb! :Q

what im saying is that UT2003 is a different animal then you may be used to. 3dmark2001 has no comparison to this game what-so-ever (your 3dmark score dont mean squat!)
higher fsb / 1:1 ratio / cas 2226 / is the way to go (for UT2003) if your system can run faster that way.

...and yes that is completely against the norm

 

[THUGSROOK]

ppl need to understand about DDR cas settings too.

because we are talking DDR here cas settings arent the way they seem.....
cas2-2-2 is actually cas4-4-4! and cas2.5-3-3 is actually cas5-6-6! :Q
that is another reason why fast cas settings are so important.

 

[Duvie]

good reporting...remember I only needed 6-7fsb to do what you did in divx encoding and in autocadd rendering..However these are a little more known as cpu hogs....

Definitely it was related to the cas2 cause if I left it at cas 2.5 like the 440mhz ddr it didn't equal it with 7fsb mor in mhz.....

 

[rogue1979]

(your 3dmark score dont mean squat!)

Couldn't have said it better myself! I also have the same opinion of Sandra memory benchmark. Gunning for a record 3DMark2001 is an worthy task if you look at it for just that reason. Spending extra money on your computer or compromising stability to raise your 3DMark for faster gaming performance is hogwash. Once you have reached a certain level of cpu power all that expensive PC3200 and big buck cpu items are a waste of money. If you really want more gaming performance save the money and spend it on a faster video card (Radeon 9700).

 

[mechBgon]

If you really want more gaming performance save the money and spend it on a faster video card (Radeon 9700).

Or a faster Internet connection! (assuming we're talking online FPS gaming)

 

[THUGSROOK]

a lot of ppl use USB for inet connect :Q
(that must suck, no?)

 

[chizow]

Ick.....USB inet connection...

Anyway, UT2k3 sooks azz...any good games based on UT2k3 engine on the horizon??? I couldn't sit through 20 minutes of playing the game, much less a few hours of running benchies.

Chiz

 

[techweenie]

I've always found that the video card makes the most improvement for my frames. Perhaps that because I always crank up the detail too...

 

[Toymaker]

Originally posted by: THUGSROOK
ppl need to understand about DDR cas settings too.

because we are talking DDR here cas settings arent the way they seem.....
cas2-2-2 is actually cas4-4-4! and cas2.5-3-3 is actually cas5-6-6! :Q
that is another reason why fast cas settings are so important.

Good point.

Also, noticed similar results when experimenting with cas settings and bandwidth with UT2003.