Help Mr. Wizard!! New KT7A FSB Problems

[Hoot]

Well, it came this afternoon and I have been chasing my butt ever since. This is a brand new Abit KT7A. I bought it because it seemed the world was passing me by with my non-KT133A motherboard (KT7 Raid). Got everything transferred and booted up. First surprise was there are combinations of clock and FSB in the bios to choose from that are not in the printed manual. I went to the Abit Web site and downloaded the .pdf manual. Same thing. Neither manual makes reference to any combinations involving a PCI devisor of 1/4, yet they are in the bios. That's not the worst part. None of the combinations using the 1/4 devisor work. I tried 133/33 and got (according to Sandra 2001) 133/44/11 bus/pci/isa. Okay, lets try the 130/32 selection. Got 130/43/11. How about 124/31? Got 124/41/10. WTF!!!
I do not have Host+PCI enabled. Okay, lets burn the WZ update. Same problem. went back to WW that came with the unit. Am I missing something here? Even at a multiplier of 6.5, the system is unstable unless I raise the core voltage considerably, not that i'm squeemish about that. Any pointers? Any stupid things I might be doing? Sorry if this has been covered already. Please help! I'm exhausted and going to bed. Will check back tomorrow morning.

Hoot

 

[jughead]

One thing is that Scisoft Sandra doesn't support the kt133a chipset yet, so the readings may be wrong. Another is ya might want to try to update your bios to the yh one. link: http://fae.abit.com.tw/download/bios/kt7/index-e.htm

 

[Hoot]

Okay, I feel somewhat better. I downloaded the latest version of Sandra 2001 and now it identifies the devisor correctly. It appears there is no such thing as a "Free Lunch". Having successfully OC'd my 650 T-bird to 100x10, at a brutal core voltage of 2.35, I was hoping to be able to hit 1G by using a higher FSB and lower core voltage. I am still playing with combinations, but it's still going to take a high core voltage to get 1G either way. I'll report on my end results in the AMD CPU forum.

Hoot

 

[Gendo]

Hoot (Mar 24, 2001 11:17 a.m.):
Okay, I feel somewhat better. I downloaded the latest version of Sandra 2001 and now it identifies the devisor correctly. It appears there is no such thing as a "Free Lunch". Having successfully OC'd my 650 T-bird to 100x10, at a brutal core voltage of 2.35, I was hoping to be able to hit 1G by using a higher FSB and lower core voltage. I am still playing with combinations, but it's still going to take a high core voltage to get 1G either way. I'll report on my end results in the AMD CPU forum.

My advice is to keep trying - I never believed in the whole "burn-in" concept until I got my new Duron 700. At first I had to crank my voltage to 1.8 just to get it to 933, so I chose to run it at 854 instead (100/33 + 22), which it'd do at 1.625. Last night I tried lowering the volts back to the default 1.6 and it booted fine, so I tried 933 at 1.625 on a whim and it actually worked. It didn't at 1.6, but I can live with 1.625. That's all just with FSB overclocking - I haven't even unlocked my chip to try adjusting multiplier settings. I'd really doubt you'd have to go over a voltage of 1.85 or 1.9 to get your 600 to 1ghz if you just give it a little time.

Your message was of interest to me because I had just started running on a 133mhz bus myself - I backed it back down momentarily after getting the same Sandra results as you did. Glad to see it's just the program and not the board itself that was the problem.

 

[Hoot]

Thanks for the words of encouragement.
This CPU has had its share of burn-in in the last day and a half. I tweaked and squeaked and wound up with 7x133 @ 2.1V/1.4V using two no-name PC133x128M sticks @ 2-2-2. It Idles at 23C with halt-on-idle enabled, 30C without and hits 37C running Prime95 torture. Here's some benchmark results:

Sandra2001se
Memory 547/609
CPU 2590/1272
MMX 5056/6336 (GF2MX)

3D Mark 2001 scores 2079

Not too shabby for a $59.00 CPU. Still, I feel the tug of a new 1.2x266 T-bird on my wallet pocket coming soon. Gotta keep up with the Jones

Hoot

 

[Colin]

I tried the new BIOS in a KT7-R with a 1.2 gig Bird. 12.5 = 1300 (112.5 X 104). Flashed the WZ BIOS and the same problem. I PHYSICALLY removed a BIOS chip from another KT7-R, placed it this board, flashed WZ and the problems persist. The new BIOS must have changed something else. Very strange indeed.

 

[Gendo]

Hoot (Mar 25, 2001 01:00 a.m.):
Thanks for the words of encouragement.
This CPU has had its share of burn-in in the last day and a half. I tweaked and squeaked and wound up with 7x133 @ 2.1V/1.4V using two no-name PC133x128M sticks @ 2-2-2. It Idles at 23C with halt-on-idle enabled, 30C without and hits 37C running Prime95 torture.

I'm a bit new to this whole thing - what is the "2.1v/1.4v" number? Up until about 5 minutes ago I was running at 7x133 too but I just backed it down again after getting some instability (specifically, running Unreal Tournament would crash every time after about 5 minutes). I had my voltage set at 1.625 but UT wouldn't run at all at that voltage so I ended up going 1.65 again - now I'm back to 1.6 and 854. I'm pretty sure my problem going higher is my RAM - I'm using one 128MB PC-133 stick and two generic PC-100 64MB sticks. I figured I'd try oc'ing the 64's to 133 and it seemed to work, but maybe it really doesn't. One of them's actually rated to 125 according to Sandra, so maybe I'll try replacing the other at some point. I guess I could always try removing them and then going to 933 again and seeing what happens...

Here's some benchmark results:

Sandra2001se
Memory 547/609
CPU 2590/1272
MMX 5056/6336 (GF2MX)

Your memory scores are quite a bit higher than mine were at that speed, but then just to be safe I hadn't really revved things up when I clocked them at 133 - I did turn on 4-way interleaving but left the timings slow. I think I just barely cracked 500 on both of my memory scores. Your CPU score is about right, though.

Not too shabby for a $59.00 CPU. Still, I feel the tug of a new 1.2x266 T-bird on my wallet pocket coming soon. Gotta keep up with the Jones

I'm always one step behind - I've found it doesn't pay to be current. I paid $43 for my Duron 700 and I get UT frame rates only a teensy bit below those of people with much more expensive CPU's. That's good enough for me...

 

[Hoot]

The 2.1V/1.4V is the Core and I/O voltages. I started testing with most of the memory and PCI bus settings very conservative. One at a time, I upped them and retested for stability. The only thing that changed was the performance (for the better). Right now, every possible setting is set to maximum speed and the setup is still stable. I didn't enable force 4-way as it had no positive effect on the results and I left drive strength set to its default (2), since going from 2 to 3 had no effect. This was time consuming, but that's half the fun. Like playing the "Nickel Slots" with a little more control on the outcome.
Now that I've got as much out of this t-bird on the new KT7A, I am going to put my 600 Duron back in and see what I can goose it up to. It actually overclocked on the KT7 better than this T-bird did. More later.

Hoot

 

[AZZKICKER]

well to let you know something

you aint going to hit higher with that board as in MHZ

just FSB differences

my T-Bird 950 blue core can run 100% stable @ 1357 @ 2.35 volts on my ASUS A7V with the voltage mod

and have the same chip now in my Abit KT7A-RAID board....i did phase one of the voltage mod and got 2.1 volts but can only run 1266 @ 133 fsb @ 2.1 volts

the chip wont run without that extra voltage......sorry to bring sad news
so you will need the more voltage no matter what...........

 

[Hoot]

Previously, I mistakenly said 2.1v and 1.4V were my Core and I/O voltages. I meant to say 2.1V and 3.4V.

Hoot

 

[Colin]

This problem surfaced on my KT7-R after flashing the YP BIOS. Flashing back to WZ did not solve the problem. I pulled a BIOS chip from another KT7-R and installed it in the problem board, flashed WZ and the problem persists. I even tried flashing WW. Something in this board has changed as a result of flashing the YP BIOS. Even physically changing to a different BIOS chip had no effect. Very, very strange.

 

[Gendo]

Hoot (Mar 25, 2001 08:53 p.m.):
Previously, I mistakenly said 2.1v and 1.4V were my Core and I/O voltages. I meant to say 2.1V and 3.4.

Ah, that's what confused me - 2.1v sounded like a plausible number but I couldn't figure out what the heck 1.4v was.

I dunno why you had to go up to 2.1v just to get 7x133 - it works for me at 1.65 (except, like I said, I think my memory is not stable at that speed). Granted, I have a 700 and you have a 600, but 1.65 to 2.1 is a *big* jump in voltage. What kind of temps were you getting at that voltage?

 

[AZZKICKER]

dood every chip is differnt

have a friend with the same chip as me and its out of the same batch as mine

his runs 1.4 mhz @ 1.80 volts

mine runs 1.35 @ 2.35 volts

its the laws of nature

nothing is made equal