>>> CAS Clocking <<<

[THUGSROOK]

INTEL TEST SYSTEM #1 ~ tested by ThugsRook
P4m 1.7 @ 2.4g 200fsb 400ddr
Asus P4B533 i845e (bios 1011)
Corsair XMS3000-C2 256mb @ 2.6v
GF4600 @ 320x714 | det 40.72 | Win98SE

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-6 = 178.6 / 64.1
2.5-3-3-7 = 178.4 / 64.2
2.5-3-3-5 = 178.5 / 64.4

2.0-3-3-7 = 179.0 / 64.8
2.5-3-2-7 = 179.4 / 65.0
2.5-2-3-7 = 179.2 / 65.1

2.0-2-2-6 = 180.6 / 66.6
2.0-2-2-7 = 180.7 / 66.7


AMD TEST SYSTEM #1 ~ tested by JSSheridan
XP 2400+ @ 2.0g 133fsb 266ddr
Asus A7N8XD nForce2 (bios 1002)
Corsair XMS2700-PT 512mb @ 2.6v
ATI Radeon 8500LE @ 285x285 | Cat 2.5 | WinXP Pro

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-7 = 124.2 / 56.4
2.5-3-3-5 = 124.3 / 56.5
2.5-3-3-6 = 124.4 / 56.5

2.0-3-3-7 = 124.0 / 56.7
2.5-2-3-7 = 124.1 / 57.0
2.5-3-2-7 = 124.9 / 57.8

2.0-2-2-4 = 124.7 / 57.9
2.0-2-2-6 = 124.3 / 58.1
2.0-2-2-5 = 124.7 / 58.3
2.0-2-2-7 = 125.3 / 58.4


AMD TEST SYSTEM #2 ~ tested by Excelsior
Tbred 1700+ @ 1.74g 166fsb 333ddr
Epox 8K3A+ KT333 Rev 1.2 (bios 2B01)
Corsair XMS3000-PT 512mb @ 2.6v
Radeon 9700 Pro @ 325x310 | Cat 3.0 | WinXp Pro

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-5 = 154.2 / 51.7
2.5-3-3-6 = 154.9 / 51.7

2.0-3-3-6 = 157.3 / 52.5
2.5-3-2-6 = 156.6 / 52.6
2.5-2-3-6 = 156.7 / 52.6

2.0-2-2-5 = 161.3 / 54.2
2.0-2-2-6 = 161.3 / 54.3


INTEL TEST SYSTEM #2 ~ tested by ThugsRook
P4 2.26b @ 2.82g 166fsb 333ddr
Asus P4B533 i845e (bios 1011)
Corsair XMS3000-C2 2x 256mb @ 2.6v
GF4600 @ 320x714 | det 40.72 | Win98SE

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-6 = 177.3 / 66.2
2.5-3-3-7 = 177.3 / 66.3
2.5-3-3-5 = 177.5 / 66.6

2.0-3-3-7 = 177.3 / 67.0
2.5-2-3-7 = 179.0 / 67.3
2.5-3-2-7 = 179.0 / 67.3

2.0-2-2-5 = 181.6 / 69.2
2.0-2-2-7 = 181.6 / 69.3
2.0-2-2-6 = 181.8 / 69.4


AMD TEST SYSTEM #3 ~ tested by Noid
XP 2600+ @ 2.18g 189fsb 379ddr
Abit KX7-333 KT333 (bios b6)
Corsair XMS3200-C2 2x 256mb @ 2.65v
Radeon 9700 Pro @ 337x340 | Cat 2.5 | W2K SP3

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-5 - 195.3 / 64.2
2.5-3-3-6 - 195.1 / 64.3

2.5-2-3-6 - 197.6 / 65.1
2.5-3-2-6 - 198.4 / 65.2
2.0-3-3-6 - 197.9 / 65.4

2.0-2-2-6 - 203.6 / 67.0
2.0-2-2-5 - 203.3 / 67.3


INTEL TEST SYSTEM #3 ~ tested by Mikki
P4 1.6a @ 2.1g 133fsb 355ddr
Abit BE7 i845pe (hacked bios)
Corsair XMS3200 256mb @ 2.6v
GF4400 | det 30.82 | Win98SE

UT2003 DEMO 1024x768x32, Flyby / Botmatch
2.5-3-3-6 = 138.4 / 48.0
2.5-3-3-7 = 138.1 / 49.2
2.5-3-3-5 = 138.3 / 49.4

2.0-3-3-7 = 138.0 / 47.9
2.5-3-2-7 = 138.9 / 49.8
2.5-2-3-7 = 139.0 / 49.9

2.0-2-2-6 = 139.0 / 49.9
2.0-2-2-7 = 139.7 / 50.4

 

[adeno]

:Q

Sweet.. 2.0-2-2-6

I want some XMS3000!
or a pair for that matter

 

[TenEgg]

I'm confused now...

so 2-2-2-7 is faster 2-2-2-6? then 2-2-2-7 is faster than 2-2-2-5 too?

 

[Duvie]

In your opinion Thugs do you think that is a great advantage or percentage increase??? I have seen in some examples for me where cas 2,2,2,6 seemed to make a good percentage increase over cas 2.5,3,3,6, but nothing that is make or break....It is more like gains in tweaking!!!

 

[THUGSROOK]

TenEgg ~ in this particular combination thats how it worked out.
each system combination will be different.

Duvie ~ there are 10 types of ppl in the world ...those that understand binary and those that dont
seriously tho ~ it will depend on your attitude towards ram.
some ppl believe that cas222 should be the "standard state" and anything else is substandard.
whilst others just givin and believe that 2.5-3-3 is normal and that 222 is just an expensive luxery.

but you just cant beat cas222, not just for benchmarks as it makes the whole system feel faster.
if you can get cas222 on 2.6v or less ~ go for it!

 

[SinfulWeeper]

Originally posted by: THUGSROOK

Duvie ~ there are 10 types of ppl in the world ...those that understand binary and those that dont

LOL

 

[Mikki]

hehe...thanks for the test Thugsrook!!! Just what I was looking for.....

...and to jump in the conversation...I don't think this stuff has been or will be cut and dry...I viewed memory timing adjustments as tools to tweak my final system config to get that "sweet spot", y'know? That balance between fsb, fsb:mem bus ratio, timings, and all your misc bios settings....

nice work Thugs!!!

 

[THUGSROOK]

agreed ~ this is not cut and dry at all.
it will very much depend on your cpu/board/ram.

what i really wanted to show was which settings are worth what.
the 1st cas setting (2xxx) isnt worth as much as either the 2nd or 3rd cas settings (x22x).

the oddites with the 4th cas setting (xxx7) are probably due to motherboard but it does go to show you that you need to run your own tests!

cas7 is faster then cas6? ....maybe

 

[Mikki]

yeah, it'd be worth running our own...it'd be cool to get a variety of mobo/memory combos tested and posted just to see...

 

[THUGSROOK]

well its obvious that cas 2-2-2-x is the fastest.
but assuming that just aint working for ya your next best choice would be cas 2.5-2-2-x or 2-3-2-x.
and if you had to ditch yet another, cas 2.5-3-2-x really isnt that bad compared to cas 2.5-3-3-7

id suggest just leaving the last setting @ cas7 until the 1st 3 are figured out and tested stable since they are more important.

run memtest86 to be sure it is stable.

 

[Soulkeeper]

I wonder how the differences would be on an athlon system
cause we all know how the P4 was originally designed to cope with the latencies of rambus


geez now that I think Of it
i've tryed all the cas settings on a kt333 system several months ago and benched them
I can't find the thread tho and forget the differences

 

[pelikan]

Wow! Two fps in UT2003 botmatch with just ram timings! That's amazing.

That's the exact same increase I got by going from 166 fsb (cas2, 3, 3, 6) to 200 fsb (cas2.5, 3, 3, 6). Kinda makes me wish I had sprung for some Corsair instead of Samsung.

If you can get a couple of fps from each of the following: ram timings, fsb, cpu overclock, video card overclock, optimizing operating system - pretty soon you've got some really hot performance.

 

[dbwillis]

Thanks for the info Thugs...

 

[Jeff7181]

Would you agree that getting things to run together in sync is better than running individual things as fast as possible?

(thinking) So... from the worst, to the best CAS timing, you get a 1-4% boost in performance...

 

[Mikki]

Going for the "sweet spot"? I would....

 

[THUGSROOK]

Originally posted by: Jeff7181
Would you agree that getting things to run together in sync is better than running individual things as fast as possible?

nope ~ that doesnt matter.
"combo settings" reveal nothing amazing.
just get the best you can out of each setting

 

[Killrose]

Would'nt a cas setting set to high, though not high enough to fail, have errors that may in fact slow it down while benching? I know that Video card mem can do this.

 

[THUGSROOK]

you should be testing memory stability with Memtest86.

never assume its stable w/o testing

 

[Killrose]

Originally posted by: THUGSROOK
you should be testing memory stability with Memtest86.

never assume its stable w/o testing

Question. I have a stick of Nanya PC2100 which is a 512mb. I bought it used on the forums knowing it was bad. I can game with it, run benchmarks with it, ect., and I would'nt even consider using it if I had to rely on it to run a bussiness or nuclear chain reaction formula's , but is there an acceptable level of failure/errors as reported by Memtest? and which colums after the test suite do I need to pay attention to?

Geez, for ten bucks it was a steal, for buyer or seller i'm not sure, LOL

 

[THUGSROOK]

no errors are acceptable

you should be running all 11 tests for at least 10 loops (512mb will take twice as long as 256mb)
bootup M86 and hit C-2-3-0 to get it running all 11 tests.
hitting the space bar will stop testing after a screen of errors (if youre sleeping, etc)

HTH

 

[Killrose]

We'll, I guess it's junk then, because it errors right off the bat. Damn Nanya and their crappy warranty anyway's (Nanya does'nt warranty their ram, their distributors do, and it could vary from 30 day's to lifetime dependant on probably what level of ram insurance they buy from Nanya i'd guess).

I was thinking of trying to dissable 256mb at a time and maybe find out if it would be stable on one bank or the other. Any idea's as to how to try this? What the hay, don't have much to loose. I might try to up the voltage maybe even first, who know's.

 

[THUGSROOK]

lower your fsb/cas/mem ratio to the absolute lowest possible.
if it still errors ~ its a keychain

 

[Killrose]

Originally posted by: THUGSROOK
lower your fsb/cas/mem ratio to the absolute lowest possible.
if it still errors ~ its a keychain

You mean it could still maybe go thru life as a stick of PC1600? hehe

 

[THUGSROOK]

yep, maybe

 

[Killrose]

Life as PC1600 will not suit it either. It still gives errors @100MHz cl 2.5-3-3-7, but I think I will take the stickers off it and inspect the solder joints everywhere, then take it to work and have my tech inspect the components that he can possibly identify and test/replace.
God how he hates it when I do that to him.

weird that I can game off it for hours with no crashes or anything abnormal happening. even at 133MHz cl 2.0-3-3-6