<<<<>>>RAM TESTING THREAD<<<>>

[Shimmishim]

Abit NF7-S v 2.0

Basically trying to determine if faster timings is better or worse than a faster FSB with more relaxed timings..

a link to Thugsrook and others finding on how cas settings effect real time apps (at the same fsb setting)

--------------------------------------------------------------------------------

TEST SYSTEM
Palmtreepc PC3200 1 x 512mb
SLK 800 HSF | Thermaltake Smart Fan 2 | AS3
Antec Truepower 430W | Onboard Soundstorm | Maxtor ATA133 7200rpm 2 x 60 gigs in RAID-0
GF4200 @ 300x531 | 43.45 drivers | Win2Kpro


TEST CPU #1
AMD 1700+ (1.47 GHZ) 3/10/03 JIUHBUPMW DLT3C (excaliberpc.com)

> CAS Settings / Vdimm <
2416mhz 193fsb 386ddr = 2.0-2-2-5 / 2.7v
2405mhz 200fsb 400ddr = 2.0-2-2-6 / 2.7v
2413mhz 209fsb 418ddr = 2.0-3-3-6 / 2.7v
2401mhz 218fsb 436ddr = 2.0-3-3-7 / 2.7v



BENCHMARKS 1024x768x32

> 3DMark2000 <
2416mhz 193fsb = 14,741
2405mhz 200fsb = 14,690
2413mhz 209fsb = 14,956
2401mhz 218fsb = 15,019


> 3DMark2001 <
2416mhz 193fsb = 13,596
2405mhz 200fsb = 13,757
2413mhz 209fsb = 13,727
2401mhz 218fsb = 13,835


> X-Isle Demo <
2416mhz 193fsb = 131.07 fps
2405mhz 200fsb = 131.78 fps
2413mhz 209fsb = 131.79 fps
2401mhz 218fsb = 132.25 fps


> Comanche 4 Demo Benchmark <
2416mhz 193fsb = 52.66 fps
2405mhz 200fsb = 53.23 fps
2413mhz 209fsb = 53.54 fps
2401mhz 218fsb = 53.01 fps


> UT2K3 Demo <
2416mhz 193fsb =
2405mhz 200fsb =
2413mhz 209fsb =
2401mhz 218fsb =


Codecreatures Benchmark Pro
2416mhz 193fsb =
2405mhz 200fsb = 30.8 fps
2413mhz 209fsb = 30.7 fps
2401mhz 218fsb = 30.9 fps


> Scisoft Sandra < memory: unbuffered (settings)

2416mhz 193fsb = 2632 / 2470
2405mhz 200fsb = 2941 / 2761
2413mhz 209fsb = 3095 / 2855
2401mhz 218fsb = 3236 / 2979


--------------------------------------------------------------------------------

THUGSROOK FORMAT

seems as though as you relax your timing the performance difference is minimal as you increase your fsb... so the question is: is it really worth it to run a higher fsb but at the sacrfice of more aggressive timing?

what do you guys think?

I think it's not really worth it that much... i've come to a happy medium with my 11.5 x 209 setting @ 2-3-3-6

not too much of a gain at 218 x 11 @ 2-3-3-7

 

[THUGSROOK]

kewl

im glad you decided to do this

 

[Shimmishim]

Ill take your suggestions of course thugs...

 

[Shimmishim]

also...

can you give me a link to a working crytek x isle? i tried the sticky but some files were corrupt or something in the zip file..

also... the ut2k3 benchmark thing... i keep getting errors when i try to install it... i downloaded demo version 2206

thanks!

 

[THUGSROOK]

thats the only x-isle demo link i know of~
you shouldnt be getting install errors~

try to dload and install both again @ default speed :Q

 

[sianof]

Hey Guys,
So you went ahead and got that excaliber chip. How is it doing so far?
Which do you think is the best Chip to buy from them? The W, X, or U?
Let me know. I am about to order one Today !!!
Thanks
Frank S

 

[Shimmishim]

Originally posted by: THUGSROOK
thats the only x-isle demo link i know of~
you shouldnt be getting install errors~

try to dload and install both again @ default speed :Q

LOL

i'l retry

 

[THUGSROOK]

the x-isle demo patch is coming from my server ~ so i know thats not corrupt

 

[Blooz1]

Good idea, Shim! I'm going through this dilemma myself right now!

 

[ErikaeanLogic]

tried 2-3-2-6 or 2-3-2-7 yet? you'll probably see a nice performance jump at 209 and 218 fsb.

 

[Shimmishim]

wow... i'll give those a go!

for the x isle demo, what numbers am i looking for by the way?

 

[Shimmishim]

updated with numbers....

 

[THUGSROOK]

very nice so far Shimmishim

tests like these all make benchmarks show their true colors ~ which ones like cas settings, and which ones like fsb/mhz.

how do ya like the X-isle Tech Demo?
its pretty much the same kinda thing as the Serious Sam tech demos, but X-Isle is a benchmark
ive found it invaluable over the years, and much easier to bench then serious sam
(its open_gl)

g'luck with the rest of your testing
your results so far are already pointing you in a certain direction

 

[Shimmishim]

it's neat.. i like it

except i have a corrupt file somewhere righ tnow and i have to reinstall it

oh well!

i've also tried 12.5 x 192 now... @ 2-2-2-5

i'll try 2-3-2-6 @ 209 a little later


it appears that the gain as you increase the fsb as you relax the timings of your ram has very minimal gain at all...

so running 2-3-3-6 @ 209 ~ same as 2-2-2-6 @ 200

but 2-2-2-6 @ 209 > 2-2-2-6 @ 200 <----- this is based upon your thread



hmm... so i guess that answers my question...

 

[WoundedWallet]

Nice work, but why is it that your FSBs give out different speeds than mine?

193fsb should yield 2413 and not 2416
200fsb = 2400 and not 2405
209fsb = 2403.5 and not 2413
218fsb = 2398 and not 2401

Also for the sake of the homogeneity, I think the first iteraction should be 192fsb instead of 193, so that all the processor speeds stay within 5MHz of each other.

Is there a reason behind these two observations?

 

[Shimmishim]

These speeds were recorded from wcpuid...

When I set the fsb to 209, it is actually 209.85 x 11.5 = 2413

When I set the fsb to 192, it is actually 193 x 11.5 = 2416

I only put the fsb that wcpuid reports to me...

i can do it again and readjust but... i think it's as close as it'll get...

 

[Megatomic]

My 8RDA+ does the same thing (higher FSB than what is selected in the BIOS). There is nothing Shimm can do AFAIK.

 

[Blooz1]

I found a d/l of the X-Isle Tech Demo here, at farcryhq. German site, slow d/l...

 

[Shimmishim]

Originally posted by: Blooz1
I found a d/l of the X-Isle Tech Demo here, at <a class=ftalternatingbarlinklarge href="http://www.farcryhq.com/downloads.php?file_id=1" target=blank>farcryhq</A>. German site, slow d/l...

thanks! i downloaded again and it worked

 

[human2k]

Wow, great work.


You deserve a :beer: or :beer::beer:

 

[WoundedWallet]

Thanks for explaining Shimmishim. That's probably why the NF7-S BIOS reports different speeds during the boot than the ones I set. And it would be a good reason why Sandra shows different speeds depending on what test I'm doing.

But no, I never thought about you re-doing your work. I'm very happy as it is. I was just trying to understand the logic behind it.

We have the same board and cpu but mine has a cheapa$$ memory that so far can only do 6-3-3-2.5. So your tests are great to help me decide whether I should return it and get a better stick. Maybe I should repeat some of your benches with my timings and see what the real difference is.

Thanks again,
WW

 

[beatle]

Thanks for the hard work, shimmishim. It looks like fsb wins in almost every category. This makes me want to try something other than 2-2-2-6 on my twinmos.

I don't mean to drift your thread, but is the palmtree pc3200 using the same winbond chips as the twinmos from memoryx? It sounds familiar...

 

[Shimmishim]

YES!

they both use the bh-5 winbond chips...

 

[THUGSROOK]

 

[Shimmishim]

bumping my own thread...