Does your AI7 hate dividers used with loose timings?

Toaster Oven · Feb 4, 2004

I'm wondering, has anyone noticed if their AI7 clocks higher (stable figures only) if you use tighter timings versus looser timings? I know that seems counterintuitive but that's what I've discovered after going through another round of overclocking after switching to the 1.6 Beta BIOS.

Background:
With my setup using BH-5 Kingston HyperX, I've found that with 2-4-4-8 or 2-3-3-7 or 2.5-3-3-7 timings with 5:4 divider, I can only reach max stable (prime95 24hr) OC of 3.1GHz although I can clock up to 3.5GHz unstable. After much headbanging (you DON'T want to know how long... :bang head

) I've found that with 2-2-2-5 timings and 5:4 divider I can run 100% stable at 3.5GHz. Yeah I know, WTF is going on? If I use a 1:1 ratio, I can overclock higher using loose (2-4-4-8) timings. If I swap out the AI7 for an IS7 keeping all the components, I can stable OC higher using looser timings versus tighter timings so it's not an issue with my CPU or memory.

Anyhow, I am definitely NOT complaining about this latest find as it gave me a huge performance jump. I'm just wondering if this is particular to my setup or if other people have noticed the same thing.

Toaster Oven · Feb 4, 2004

Here's a screenie of my latest stable settings

Now if only I could do this at default Vcore!

Clutch_Head · Feb 4, 2004

that's about where i'm running my IS7/2.4C. but i have to run my RAS# to CAS#
at 3. i tried to get it to run at 2 with 3.2v Vmem to no avail. i fell just
short of my last goal of 300mhz FSB. seems alot of the 865/875 boards
peak at about 295mhz FSB. mine is prime stable at 295 but i lowered
it down a few just to be safe.

what are you running the Vcore at, if you don't mind me asking?

rich

Xception · Feb 5, 2004

Hmm, Interesting Toaster Oven. I'll give that a try tommorrow and let you know how it works out. My AI7 with an SL6Z3 is currently running 280FSB at 1:1 with memory setting of 2.5, 4, 4, 6 on OCZ pc4000 Gold. I got it to 292 earlier today at 5:4 setting but not P95 stable. Currently at test # 335 in P95 so I just dunno whats holding me from 300FSB. Im at stock vcore now and was at 292fsb with stock vcore also. Temp loaded is only 29° watercooled. I have a hunch that these boards just top out around 295 for some reason but I will keep trying. Good Luck with yours.

I really love the temps of this new board as compared to the IS7

Toaster Oven · Feb 5, 2004

Clutch_Head, at 3.5GHz I have to bump the Vcore to 1.7V (actual 1.65-1.68 in AbitEQ) to get Prime 95 stable. I see max temps around 51-52C at that level. If I drop down to 3.4GHz I can run default Vcore and max temps drop to 47-49C. And if I drop down to 3.2GHz, I can run NB strap at 667, enable GAT, run F1 as well as the most aggressive mem settings all around

At this point, I think I may just slap on a nice noisy Vantec Tornado or Delta and see if I can't make that last jump to 300FSB. I think I've tweaked it as far as it'll go with my current setup.

Xception, looking forward to your results. Those are some nice temps you're seeing under full load. Have you tried bumping up your Vcore and seeing if that helps?

BTW, what PWM temps are you seeing under full load? My PWM temps consistently rise up to 51-53C under load. With an IS7 my PWM temps rarely broke 40C.

Xception · Feb 5, 2004

Toaster Oven,
I am not seeing the same results as you when lowering the ram timings. In fact I set them down to 2 5 3 3 and I got some kind of wierd slow down when booting to windows. I did however see an improvement in mem scores when going from 2.5 7 4 4 to 2 6 4 4 when at 280 or higher fsb at 5:4 ratio. It seems funny I can run 290 stock vcore with ram at 3 volts but can't do 295 no matter what voltage I throw at either. I ran P95 24 hours at 280 fsb without any errors and am working on 285 now.

My PWM (mosfet?) temps run up to 50 with only one instance of P95 running.

Brihee · Feb 6, 2004

I have noticed a very similar problem with my AI7 board, a P4B 2.4, and a stick of 512 MB GEIL PC3200. First of all, the AI7 is very finicky with this RAM. I could run it at DDR420 at 3:4 in an Asus P4B533-E at 2-3-2-7 for hours of Prime95. In the AI7, just to get it stable I have to run it below DDR400 and have all GAT options relaxed. Even then Prime95 quits after ~10 min. As for the timings, I can get it to boot up at 2-3-2-6 or looser, but as soon as I change to 3-*-*-* it won't boot no matter what clockspeed or divider.

What I found even more annoying was trying to get 2x256MB Kinston HyperX PC3200 to run stable. The only divider that would even boot was NBStrap 533 at 4:5. Any other divider would only result in a repeating long beeping sound from the MB no matter what the clockspeed. This problem effectively capped this RAM at DDR400 since my CPU becomes limiting at that point.

I would like to get some new RAM that works, but as it is I am pretty fed up with this board. Is it just my board or should I expect this with the AI7? Suggestions would be much appreciated.

Clutch_Head · Feb 6, 2004

Toaster Oven,

1.7v Vcore is more than i would be comfortable with. if it runs 3.4 stable
with default voltage (1.55v?), i would run it at 3.4 with the tighter timings
and some GAT options enabled. i would put money on it not being able
to hit 300Mhz FSB. not that i wouldn't love to see it, but it seems these
boards run out of steam at 295 or so. if your below 60° C on the
CPU i doubt that it's heat that's holding you back.

as for memory, i would think that BH-5 chipped memory would run best.
the CH-5 chips are problematic in the IS7 and the IC7 series of boards.
i tried loosening my timings before i did a volt mod and would get
the dreaded memory error beeps during boot. the one i've had the
most success with is the RAS# to CAS#. going from 2 to 3 gave me a
boost from 209mhz FSB at 2.8v to 227mhz FSB. then again i own a IS7
so i dunno if any of this applys to your board or not :-/

.

later
rich

Xception · Feb 6, 2004

PWM temps seem to be directly connected to your agp voltage which is also your northbridge voltage as I lowered mine to 1.5 and it dropped the temp to 48° when loaded.

puji · Feb 11, 2004

Just reading the post. I have similar results which surprised me. First AI7 hates CH 5, but even the kingston 3200 bh5 is not all equal. I got the dual sticks and one behaved like ch5 so I had to RMA it . Due to crossshipping I could choose between the other three sticks. One of them was also not that great with lots of timing problems . But one thing sruck me especially: all sticks would do worse on loose timing.
Like when I got to the limit of 1:1 at around 242 (it was stiil working but errors showed up in the test) and tried to use looser timings , it would even refuse to boot. Go figure?!!

ShaDrag · Feb 15, 2004

Seems like the AI7 is a very quirky board. I've only had mine less than a week and it has already given me more trouble than the IS7 I had for months. Hopefully with more time I can tame th beast, if not I can always break it

www.darkblau.de · Mar 10, 2004

hello there...
i have the abit IS7-E and and pentium4 2.4C cnps7000Cu-cooled, but i can't get more than 280MHz stable (3360MHz) @ 1.7v-core. i tried ocz ltd 3200, performance 4000, EL gold rev2 3700, geil platinum 3700, hyperX 3500 and whatever without success...
i can only run my ram with 5:4 and slow timings (3-4-4-8) at that frequency.
at more than 280MHz i get prime-errors (ram-related) and at > 283 i get superPI-errors (cpu-related)
is a higher voltage than 2.8v dimm core necessary for continue overclocking?
how do you all get these high frequencies with e.g. kingston pc2700 at these tight timings??

what do i do wrong?

Xception · Mar 10, 2004

wwwdarkblau.de

Welcome to the forums!!!!!!

I have some pc4000 gold el that didn't perform very well without raising it to 3 volts. However I have some pc4200 ADATA ram now that is doing 288 1:1 with 2.7 volts. Try and find some of that ram over there if you can and it should help you. Sounds like you may be nearing the limit of your cpu but you should be able to increase your ram bandwidth with some ADATA. Not guranteed but most others are doing really well with it also.

Cheers

www.darkblau.de · Mar 10, 2004

hm. my IS7 supports only up to 2.8V... so i couldn't get high-frequency rams such as PC4000+ not over 245MHz

my pc freezes at 284 and calculates trash in superPI. is this because of my temperature or have i reached my max?
a-data at 288?? huh... couldn't get some a-data 3600 over 226 that i tested recently...
245MHz was my max with OCZ Performance PC4200

Xception · Mar 11, 2004

Well,
I just finished putting together a new comp for my bro with that same gold el that I know will do 265 1:1 but on his board which is an AI7 also it wouldn't do 250fsb but that is with a 2.8c SL6wJ chip. I think after it burns in it may well do 250 but thats another story. The point I am trying to make is if you are getting errors in memtest at any certain fsb and you know your memory will go higher then its most likely your cpu that is holding you back. If you are at 284 on air.. be happy !! I think there is a volt mod out there for the IS7 for the ram. At 284 you are a prime candidate for the ADATA pc4200

www.darkblau.de · Mar 11, 2004

I'll look around for that ram. yes. i found the mod some weeks ago, but never had the heart to do this mod...
what about my temperature? yesterday i overclocked to 284 again and got temps just below 50°C. is this too hot due to the calculate-errors?
i noticed that my cpu-core-voltage drops down by 0.05V at load: when i set 1.7V in my bios, this voltage is maintained at idle, but at load it simply drops to 1.65. why's that so??

btw, what is SL6wJ?

@ toaster oven, that is that M0-stepping??

what do you all think of the new prescott 2.4 psb533?

www.darkblau.de · Mar 11, 2004

should i increase my agp-voltage? it has been running at 1.55V all the time. is this also the voltage my nb is runnin with?
dunno anythin bout this voltage...

oh... i think i have a sl6wj. malaysia...

lbbo2002 · Mar 14, 2004

In testing my Adata ram I moved everything over to my AI7 which was originally be used with my Muskin Promo Bh-6 at max vots and 2.4c CPU. I can boot 1:1 to 292 and run one or two benchmarks with the Adata Ram at that FSB

Using the 5:4 dividers I can post and run benchmarks at 298. 299 and 300 result in a check sum bios error on post. Have to reset the CMOS.

The Mushkin is running 292 5:4 2,2,2,5 and priming now. I will try and post it back up to the 301 as I could Post fine in the IS7 with 3:2 divider.

Took more power to the CPU with the AI7 to get in at what the IS7 was doing. Took .05v more on vcore, with max mem volts and APG.

GAT will not enable above 288 on my AI7 in 1:1 or 5:4.

Another data point

www.darkblau.de · Mar 14, 2004

which mushkin ram is that?
290 at 1:1 at 2-2-2-5??? how's that???...

on the is7(-e)?

lbbo2002 · Mar 14, 2004

www.darkblau.de you miss read something or I missed typed

The 1:1 is the Adata RAM (mine is just PC4000) but a thead running over in memory tells all about it

I am at 3,4,4,7 but below 270 I can run 2.5,4,3,7 , great RAM for cheap.

http://www.ocforums.com/showthread.php?s=&threadid=273799

The Mushkin promo is running 292 but at 5:4 as I stated. Can't get above 240 @3.2v w/2-2-2-5 in 1:1. Does 250 1:1 but have to lower timings to 2-3-3-7

Both my IS7's hate the Adata above 258 1:1, but I think its because I have some of the original versions of the IS7 boards and they are picky as to what RAM they work with.

Does your AI7 hate dividers used with loose timings?

Member

Member

Member

Member

Member

Member

New Member

Member

Member

Member

Member

Registered

Member

Registered

Member

Registered

Registered

Member

Registered

Member

Similar threads