I'm ready to try again, hopefully no deaths this time!

Hi,

I'm gonna try tweaking the memory timings on my x300 again for a possible higher overclock. Although I still don't know if I should...

This brings back bad memories:
http://forum.notebookreview.com/showthread.php?t=39593

However, I think that was caused cause I accidentally put it at 460 instead of 360 lol.

What do you say...shall I go for it again?

 

Overclocking a x300? l really don't see the need. Usually overclocking gets benchmark numbers and not much of a in-game difference.

Plus overclocking anything on a laptop really isn't that smart in my opinion.

 

This is offtopic but overclocking has given me a 50% increase! from 20 to 30 fps for example in LFS2.

 

Yes but how much warmer does it run now?

 

Umm maybe 2 degrees...

 

ive been overclocking my x300 ~390/300 for gaming though ive taken it to 420/315 and i get about 40 more frames on average in COD1 and i can now run on highest settings. the 3dmark05 score also went up about 550 or so to 1790. Flav your my kind of guy, but how the h_ll did you get it OCed that high?

 

lol I'm glad someone understands.

I guess I'm just really lucky

You get artifacting any higher?

 

yeah ive got the 128 but i guess in the 6000 they stock clock it to 300/216 which seems a bit ghey to me. i get some artifacting but not much... i suppose i could get crazy and shoot for 2k on my 3dmark05... but since my stock clock is so much lower im looking at some massive overclocking, at least % wise.

 

ikovac, if you're reading this, i'd like some encouraging for me to try this again lol.

Which parameters did you have success with and which increase overclockability and which decrease overclockability but improve performance?

EDIT: Ok I've tried again. This is what I got so far (only changing what you said to change before)

On the left is my current value, on the right in ( ) is my default.

-------Timings1:-------------
CAS: 4(5)
------Timings 2:------------
TRCDWR: 1(3)
TRP: 5(6) (This one used to be able to go down to 3...don't know what happened!)
TRAS: 6(9)
TW2RSame Bank: 2(3)
TR2R: 1(2)
------Timings 3:------------
Memory refresh rate: 255(28)
TRFC: ??(17) don't know if I should increase or decrease this one.

This is with the same overclock I had before...but I'm slowly increasing now and it seems I've got more headroom now! (on the memory of course)

EDIT: Result so far:

(I beat the 9600xt...I can go to sleep now)

 

I'd still like advice for the TRFC, whether I should increase it or decrease it??

 

flav, jw wat your 3dmark scores are so i can get a feel for what difference this uberclocking would make

 

Default 297.00/229.50 and default memory timings:



Overclocked 435.85/331.71 but default memory timings:



Overclocked 435.85/331.71 WITH optimized memory timings:

 

They do but i'm deathly afraid of overlocking my memory, given your previous issues. ahem. but i think i can push it a bit more. my scores go from 1200-1790 right now.... im aiming for your 19xx

 

Overclocking your memory without changing the timings (or sometimes called latencies) is fine...Just push it till ATT starts to report artifacts.

My mem clocks are stable for at least 500 seconds (no artifacts).

What are you using to overclock? Use ATT if you're not already.

OH yea and in ATT, my core goes stable up to ~450 but it crashes CS:S until I lower to <440 but only on cs_assault map which is extremely weird.

You're getting 1790 without changing latencies though! That's **** good...I only got ~1500 overclocked with default latencies :| hmmm What drivers are you using? I'm still using some old omegas based on 6.1. I should upgrade but I should also really be studying for my exams!!! AHHHH

 

im on the new omega's as of 2 weeks ago and i use ati tool. I'm not sure if it has latency options but i like it because it's simple. when i installed ati tray tools i didn't check the OC option so i didn't have it, and i'd used ati tool before so thats just what i go with. is ATT that much better?

 

Umm I don't know...i've always used ATT, but if you click on mem in ati tool, i think it has latency settings!

But ****...I don't know what happened. 3dMark05 seems to ignore my latency settings recently! wtf. I can see the benefit of the latencies in games (few fps) and in ATT bench but 3dmark05 gets the same score w/ or w/o latencies (~1580). I tried reinstalling it but same thing...what the hell.

 

ill have to check it out. can't say i know anything about the 3dmark ignoring latency change though

 

Tell me how it goes. I really don't understand this 3dmark thing...it shouldn't be changing the latencies for me and I can 100% for sure say that the latencies are set cause I can see the difference in gaming! ARRRGHHH..

The only reason I care is cause I further tweaked the timings and I'm sure I'd break 2000!

Here are my new tweaked timings (I just changed more than ikovac had initially said to)

-------Timings1:-------------
CAS: 4(5)
------Timings 2:------------
TRCDRD 4(6)
TRCDWR: 1(3)
TRP: 5(6)
TRAS: 6(9)
TRRD: 1(3)
TR2W: CL+2(CL+3)
TWR: 2(3)
TW2RSame Bank: 2(3)
TR2R: 1(2)
------Timings 3:------------
Memory refresh rate: 255(28)

I discovered that TRP is sensitive to other latencies...so that if I leave TRP at 6, I can put TWR and TW2R at 0! But then changing TRP to 5 immediately gives artifacts (and I'd rather have a lower TRP cause it gives a greater benefit)

 

wow so im in over my head on timings. has anyone done a tutorial?

 

Not sure. But find one of Ikovac's post and look in his signature for "latencies". He wrote an article on it.

 

ati tool doesn't support memory changing on my card i guess i'll have to reinstall tray tools

 

Yea it didn't on mine either. Don't worry man ATT is good but be careful with those timings!

Apply them at your highest overclock b/c if you don't, you might find latencies are stable but when you overclock all goes nuts. Also, if something happens don't panic.

I should also add that the "death syndrome" has happened to me again when pushuing latencies too far during testing lol. good thing I know how to fix it now. The death syndrom is what I experienced in my other thread where all of a sudden I couldn't overclock memory past like 310 and it would screw up the screen. If this happens to you, unplug the notebook and run off battery. Then make it crash again by overclocking past 310 to screw up screen and freeze. Reboot on AC and hopefully the syndrome will be gone!

 

man i was just going to tell you that im out. lol. i hit 1911 at 417/315 with the settings at "highest performance" and "application preference." i'd post the screenshot but that would take too long and i've got hw to do... the 9300 has a second fan right? that could be holding me back at least cooling wise cuz im afraid of heating up, but i can push this x300 a bit more imo

 

Thats great man. Yea it has a second fan. What temps are you getting? The only thing affecting temps should be gpu clock and that makes almost no difference performancewise. I don't know how your getting such nice scores...even stock you're getting 12xx while I'm only getting 10xx What could cause this? We're both running latest omega's based on 6.3. How much free ram do you have before hitting run in 3dmark05?

So you still haven't overclocked to artifacts??

 

i get down to about 350 being used so 650 free... yeah when i ran stock speed at "high performance" i got 1681, so 1911 is a bit of a disappointing gain for all of that OC... i'm gonna try again brb lol

 

WHAT!? 1681 w/ stock 297/216 speeds!?? No way.

**** I looked on futuremark's site, at project compare and ppl are getting high 13xxs with stock 297/216 and me with my stock 297/229.50 am getting 10xx. GREAT... Now I'm mad.

What do you get stock in ATT benchmark?

 

im using 316mb of my 1 gig with 3dmark05 open and ati tool open. everything else (wireless off, zone alarm quit, avg quit, quickset quitted) and i just hit 1926 at 420/320. yeah i really dont know what happend with that 1681 im gonna run it againl, but i think that at "high performance" my framerate for the cpu tests has gone down, could just be me.. ill check my stats

 

okay what just happend... and how do you read .3dr files?

 

Ok run it again but regardless I'm sure it'll be better than my 10xx...something;s up. BTW I edited my above post too late, what do you get in att bench? Hopefully it's just my 3dmark05 thats screwed.

 

huh? what happend? With excel I think but they're useless.

 

i havnt used att bench, since ive never had att...

 

Oh...Could you please get it and try?

EDIT: I remember ikovac saying to stay away from the newest ATT which uses the fullscreen benchmark...HMMM could this be my culprit? I never ran 3dmark05 before I got it. In-game performance is the same though...so I don't know...

 

1460 or so. thats a bit more like it, i ran it twice. true, in game performance is what really matters but 3dmark05 is fun.

 

how much of your cpu and memory does att take up? could it be that much more than ati tool?

 

lol like 4mb..but that's not how it'd be making it bad...I think maybe this new version has messed things up. I'm gonna reinstall drivers and uninstall ATT and then bench w/ stock. Hopefully that will improve my score cause I'm not happy.

I have to study for exams now though. Maybe late tonight if i manage to finish. I've procrastinated with this enough already.

 

lol good luck

 

I FIGURED IT OUT! You'll never believe what it was...

After reinstalling drivers (at the dinner table, on batteries), I ran it again, and low and behold...another 1080 score. I was like ****! I go and plug it in and immediately I notice a difference! So I knew it had to be powerplay related. But all those times before it HAD been plugged in...So this is what it is. Even if powerplay is disabled (no checkmark) BUT you have selected the "battery optimized" slider (for AC or battery), 3dmark will somehow take that into account and run in a slower half-powerplay mode (clocks are still at default so it's really weird!)

I tested for this and it's confirmed, if I put the AC slider to battery optimized (how it was before driver reinstall, b/c I had it disabled anyway), and disable powerplay, the problem returns! So I reinstalled ATT and finally my timing settings work again (turns out it was the same problem stopping the timings too), my 3dmark05 score is:

1982 lol...I thought I was gonna break 2000. Oh well.

 

hmm. i still might be able to beat you to 2k lol

 

hahah, let's see!

Now that I see that tightning timings doesn't make too big of a difference...I'm wondering if I could instead loosen the timings and overclock the memory more! after exams...AFTER EXAMS...lol

 

ive hit a wall. to get over 2k im going to have to change timings eh i'll think it over but i've got a calc and bio test tomorrow.

 

I've got a bio exam

Good luck w/ that. BTW timings just gave me ~40 points...what's your wall?

 

low 1900s. increased the speed by 10 on both core and mem and i got 1890 something with "max performance" then i slid over to "quality" for both mipmap and textures and got 1911... 1926 was with 10 down at "max performance"

 

btw, what are your stock speeds

 

Stock timings are in brackets.

Have you overclocked until artifacts are reported and then backed off a bit?

 

the program does it by default but i can choose how many artifacts it detects before backng off

 

Oh I see. Don't worry man we've overclocked these babies so well.

I doubt we'll ever see 2000 though lol. I'm only 18 points away!

 

yeah i think ill try one more time 430/330 at max performance and max qual then get on this calc

 

Well it's pointless to go above your artifact frequencies...

 

or is it? 1955 at max performance at 430/330 with artifacting only on the 3rd test\

and 1922 at the same speeds with max quality.

have you tried the AA settings or changing the filtering?

 

nope. but different aa will def. not help. Or you mean from within 3dmark? You're not supposed to change those for comparable scores...