Sign in with
Sign up | Sign in
Your question

ASUS K7V333 settings HELP!

Tags:
  • Motherboards
  • Asus
  • Hard Drives
Last response: in Motherboards
Share
August 6, 2002 5:15:59 PM

I just finished putting together the following:
XP2200+
ASUS K7V333
OCZ 512mb PC2700 CAS2 !!!!!
Antec 430W psu
80GB HDD (8mb cache)

It runs GREAT in "OPTIMAL" mode but when swithed to "TURBO" the system will spontainiously reboot in Windows. This is the only change I made in the BIOS. I have been told to look at other speed settings like PCI and Master timing. Any suggestions on what I should be looking for????
Thanks...



The problem is always with the last thing you check ... so start there first.

More about : asus k7v333 settings

August 7, 2002 1:55:22 AM

OK, well how about telling me what kind of setting you all are running?

The problem is always with the last thing you check ... so start there first.
August 7, 2002 2:40:52 AM

what changes when going from optimal to turbo?
is it just the ram timings or it there more to it?

if its just the ram i would suggest increasing the ram voltage a bit, see if that helps.

i had to bump up the ram voltage of my corsair ram running at 166mhz when using the hardest memtimings. was ever so slightly unstable otherwise.

<b>Before visiting THG i was a clueless noob. Now im still clueless, but look at my nice title!<b>
Related resources
August 7, 2002 10:25:31 PM

Sounds like my system...

K7V333
XP 2200+
Corsair PC2700C2 (originally - more on this below)
And not that it matters, but the same 80Gig, 8Meg cache WD drive.

Over the last couple weeks I have found out what doesn't work, what does work, unlocked the CPU, played with 5 different BIOSes (1006, 1007, 1008, 1009.003 (beta), 1011 (appears to only be available on Asus's Taiwanese web site), and exchanged e-mail with a couple of experts (BIG thank yous to Michael at www.lostcircuits.com and Anthony at www.oc-athlonxp.com)

So... anyway, I think I know what you're running into, and I think I know what info to give you.

First - the Optimal vs Turbo setting adjusts various chipset timings / latencies. It does not change CPU speed, FSB speed, memory bus speed, or memory latenices. It does give me about a 2.5% graphics performance boost according to 3DMark 2001SE. I forget what Sandra 2002 showed with this on / off, but I have it all written down at home if it becomes must have info. So what it boils down to is, all things being equal, you want to run with Turbo.

Now - I started out running the machine slow and stable:
133MHz FSB, 13.5x CPU multiplier, 1:1 FSB to memory, 2.5:3:3 memory latencies, Optimal.
Worked fine.

I kicked it some:
133MHz, 13.5x, 1:1, 2:2:2, Turbo - Runs great.

I kicked it up some more:
133MHz, 13.5x, 4:5 (166MHz memory now - PC2700 speed), 2:2:2, Turbo - very unsatable.

I took it down a notch:
133MHz, 13.5x, 4:5, 2:2:2, Optimal - kinda stable - seemed stable, but would lock up after a while under heavy load (3DMark looping).

I did some research on Corsair's web site and discovered that XMS PC2700C2 memory is only rated for 2:3:3, not 2:2:2 at 166MHz. Crap. I ordered a stick of XMS PC3000C2 (rated for 2:3:3 at PC3000 speed (185MHz I believe), or 2:2:2 at 166MHz - what I was looking for).

I popped the new memory in and the system ran great at:
133MHz, 5:4, 13.5x, 2:2:2, Turbo
OK, finally getting somewhere... now for the next step:

Unlock the CPU so I can run the FSB at 166MHz too and run the memory synchronously (1:1). This gives a good performance boost without overclocking the CPU and making things toasty.

The Athlon 2200+ is pretty easy to unlock. The dimensions of what you're working on are very small, so I recommend buying one of the kits (about $15) that includes a little bit of high silver contect conductive grease that you use to draw a connection around the end of the L3 bridge so you can run the CPU at multipliers from 5x to 12.5x. Look here for images:
http://www.beachlink.com/candjac/TbredDecode0.htm
Look about 3/4s of the way down at the picture from DeerHunter. That's all you need to do - draw a little arc with the conductive grease using a pin. Tip1: Make an ARC, don't go straight across. The center has a laser cut pit that goes down to the ground layer. Contact from there up to the cut through layer and/or the surface and it could be ugly. Tip2 - Those little contacts you're connecting are actually depressions (as seen under the microscope my friend used to do the solder job for me) - so make sure you get a good little dab of grease into them, don't just go up to the edges.

OK - back to specs and such.

With the CPU unlocked I played around for a while and everything worked great, and I settled on:

166MHz FSB, 11.0x, 1:1, Turbo, 2:2:2. Runs great. Benchmarks well. Great stuff! The CPU is only OCed by 26MHz, and even under load only gets up to 51'C, so there's no worry there. The whole system has been rock solid stable for almost a week now at these settings, and performance is up considerably (I was using Sandra 2002 and 3DMark 2001SE at every step). The CPU has a lot more left in it - it was running stably for me at about 1930MHz at one point, but I see no point in pushing it if I don't need to. The video card could probably OC some as well - I haven't played with that at all yet. My system is heavy on conventional cooling (4 case fans - 2 in, 2 out - 1 fixed speed each end, 1 motherboard controlled each end), but nothing fancy. I went with the Taisol 2200+ rated cooler for the CPU - seems to be doing a fine job.

Oh, right - there's more to this story...
Once I got to system tweaking (don't underestimate the benefit of disabling unnecessary performance counters and services) it was time to check on BIOS updates...

System came with 1006 - it runs great.

I tried 1007 - adds the ability to adjust the CPU cutoff temperature. Won't run at the same settings any more. I had to change from Turbo to Optimal to avoid a "System failed due to overclocking" bull message (bull because the overclocking to 166MHz FSB and 1826MHz CPU are rock solid - it was the chipset latencies that it couldn't handle any more. Whether they made a change to what Turbo does that was too aggressive, or whether it's a blind flag (if FSB is overclocked and Turbo is on, fail to boot and bitch and moan (which I doubt since the standard procedure is to allow OCing until things crash))).

I tried 1008, 1009.003 (beta release on American site) and 1011 (supposedly a final / release version on the Taiwanese site) - all acted the exact same way.

I'm currently engaged in e-mail exchange with someone from Asus AND I'm expecting a call back from a friendly English speaking tech that I spoke with today, so hopefully I'll know more soon.

My desire is to be able to continue to run at 166, 11x, 1:1, 2:2:2, Turbo (at least until some game comes out that demands I start OCing things more) and also be able to control the temperature cutoff for the CPU. It defaults to 90'C, which is too bloody hot. The lowest it can be set to is 70'C, which is still high, but much better. I'll let you know what I hear from Asus.

Questions for you :) 

What BIOS version are you running? (If later than 1006 I recommend downgrading until they fix their performance issue)

And if it's 1006, what does Asus Probe say your system's +12V is? Mine reads about 13.35 - fairly high - just high enough that the +/- 10% window flagged it as bad. However, a later version of the BIOS read a much lower value.
August 8, 2002 6:48:12 AM

Nice work Yeti and lots of it. I currently have an A7S333 but getting A7V333 soon. My question is I have samsung 333 cl2.5 do you think I could use the same settings as your ram (ie. 2:2:2 166Mhz)? If not could you explain? I don't fully understand the latency yet. I understand the Mhz.

:eek:  Who needs heatsinks and fans, I have an igloo :eek: 
August 8, 2002 7:05:07 AM

Very impressive!
scientific and methodical!

in my experience the ram ratings are pretty conservative. i.e. if you get ram rated for cas2 it will do 2,2,2 at normal, but fast or turbo may be out of the question, depending on ramtype.

example is my Corsair PC3200, rated for cas2.5 @ 200 and cas2 @ 185mhz.
on my epox board, at 166mhz, turbo, all mem timings to maximum, it wasnt quite stable at default voltage, despite running at only 166mhz.
fortunately it was easy to fix. a small 0.1v boost to 2.6Vdimm did the trick.

oh yeah... one final <b>IMPORTANT</b> thing....
that asus probe program <font color=red>sucks</font color=red>
highly unreliable, especially regarding core temps and voltages.
i reccomend you use Motherboard Monitor
<A HREF="http://mbm.livewiredev.com/download5.htm" target="_new">http://mbm.livewiredev.com/download5.htm&lt;/A>
most of the regulars at Toms do.
cheers

<b>Before visiting THG i was a clueless noob. Now im still clueless, but look at my nice title!<b>
August 8, 2002 1:18:51 PM

333MHz is the DDR speed, 166MHz is the actual clock frequency on the bus.
So PC2700 = 333MHz DDR = 166MHz memory bus

If the memory is rated for 333MHz, 2.5, it's specced to run at 166MHz bus with a 2.5 latecy (that's the CAS, and it means you'd be running 2.5:3:3). It's possible that you'd get it to run 2:3:3, but I highly doubt you'd get it to run at 166MHz with latencies of 2:2:2. Even the Corsair 333MHz C2 memory I picked up wouldn't run at 2:2:2... only the 2:3:3 it was rated for. That's why I got the faster PC3000 module - so I could run it at 2:2:2 back at 166MHz (and up to about the low 170s probably - then have to drop back to 2:3:3).

Memory latency is a total of 5 values - from darned slow (for PC2700 and relatives) at 2.5:3:3:5:2 to normal values of 2.5:3:3:6:1 or 2:3:3:6:1 to fast at 2:2:2:6:1. You normally only see the first three values talked about because they're the ones that change normally. The 4th number is reverse from the others - it's how long you can continue to read before having to refresh your addressing, so you want it to be the higher value of 6, not the low value of 5. This number would go the other way - if you were to run fast memory very slowly you might have to drop it to 5 because on a slower memory bus 6 cycles would be a larger absolute period of time. The 5th latency is the command latency, and is almost always 1. So we come to the initial three important ones. I forget them :p  They're Column Address Strobe (CAS), Row Address Strobe (RAS), and something else I believe. They're you're primary memory latencies and represent, for the example of CAS, how long after sending the command to address a specific column the system has to wait before it can start reading memory. So lower is better. Normal PC2700 memory will run 2.5:3:3 (half clock possible because the memory is running DDR), faster PC2700C2 memory will run 2:3:3, and if you get very fast memory, usually by buying memory rated for C2 at the next higher clock speed - like PC3000C2 - you can run it at 2:2:2.

BIG NOTE I FORGOT ABOUT: Read the Asus A7V333 review on www.lostcircuits.com - pay special attention to his page about jumpbers - towards the bottom of the page he talks about the J1 and J2 jumpers for the memory voltage. Asus defaults these to a voltage OUT OF SPEC - you should lower it to the default voltage unless you absolutely need a bit more juice due to heavy overclocking. You want to move both of these jumpers ummm, if the cpu is on your left, and the slots on the right, connecters out the back of the mobo facing you - you want to make sure both J1 and J2 are to the right - away from the CPU / towards the slots. Or you can just remove them both - results in the same setting - but I prefer to leave them on in case I'm ever looking for them again.

lhgpoobaa - Thanks for the link! I'll definitely be checking out Motehrboard Monitor when I get a chance :) 
August 8, 2002 11:40:50 PM

RAS CAS and TRP or something like that.
number of mem timings displayed depends on the mobo manufacturer... my epox board has a crapload of em, more than what ive seen anywhere else.

so my timings look like
166, turbo, 2,2,2,6,1T,8 bank, 4 way, 2 other settings and 1T.

<b>Before visiting THG i was a clueless noob. Now im still clueless, but look at my nice title!<b>
August 9, 2002 3:14:22 AM

Hey which would be better with my asus kt333 board - a 4:5 divider or 1:1? I have Corsair XMS PC3200 C2.5 memory (I think what you have) and an athlon xp 2200. Is putting it at an asynchronous setting gonna slow it down? I'm a newb so don't flame me please.
August 9, 2002 1:11:25 PM

You have PC3200C2.5, I have PC3000C2

I THINK yours is slightly faster than mine, but not enough faster to be the PC3200C2 they have (which I couldn't find at Googlegear). Both should run at 166MHz, 2:2:2 for latencies, just fine (I know mine does).

For testing I recommend picking up some of the great benchmark software out there. I can give you my best guess, but you're best bet is to do real tests and see what truly works for you. Plus... it's very rewarding to make various changes and tweaks and see your performance increase! I'd recommend picking up two things especially:

Sandra 2002 (unicode version unless you're running an old OS) from SiSoft. This includes a lot of different tests and benchmarks, but the memory benchmark is what you want to use. This is a great test to determine things like the best memory setting and what effect certain BIOS tweaks have on your memory performance. Memory, bus speed, and CPU related tweaks can show a many percent performance increase here that can be harder to see in other tests that are dependant upon other aspects of your system.

The second thing I recommend picking up is 3DMark 2001SE. The 3D graphics tests it runs are a good test of overall system performance. I found that with 3DMark I could see the benefit of many other tweaks - disabling services, disabling performance counters, logging in as a non-administrator enabled account, etc... that don't show up in Sandra 2002. 3DMark is also a great system stability tester. Set it to loop and leave it for a couple hours - if your system hasn't rebooted, shut down, or locked up then you haven't pushed anything beyond the stability point.

With your system I'd start by making sure you're running Bios 1006 (until they come out with a newer version that will run as fast), and try a few different combinations like:

1:1, 133MHz, 2:2:2, Turbo - should definitely work and be stable, and provides a baseline.

4:5, 133MHz, 2:2:2, Turbo - see if that runs stably. If it does, that's probably your best non-overclocked setting.

If that's not stable, try both:
4:5, 133MHz, 2:3:3, Turbo and
4:5, 133MHz, 2:2:2, Optimal
and see which of the two is stable and if both are stable, which performs better.

The asynch vs synch issue is actually clouded by another issue - the FSB is running at 133MHz DDR (333MHz). So speeding up your memory beyond that point definitely gives diminishing returns in most situations.

For even better performance, go to this web site and read through some of their tweak guides:
http://www.3dspotlight.com/
Most of these changes won't improve your Sandra 2002 results, but show up in improved 3DMarks.

Also, while I don't recommend OCing the CPU too much unless you have a really good cooling system, I do recommend unlocking the CPU so you can run a higher FSB without OCing the CPU. Since you said you're a newb, I'll explain :) 

The FSB runs, by default, at 133MHz. The CPU runs at some clock multiplier off of that - in the case of the 2200+, which actually runs at 1800MHz, that is a 13.5x multiplier.

Increasing the FSB speed will speed up your FSB (CPU to north bridge (talks to video card, memory, and south bridge), your memory, your CPU, and to a lesser degree (on most motherboards - some allow individual control) your AGP and PCI busses. So you're overclocking a lot of things when you change it.

BUT - if you could change that CPU multiplier so that you raise the FSB, but lower the multiplier - you can keep the CPU around the same speed (1800-1850 ish MHz). Plus, if you go to one of the 33MHz multiples for your FSB, the Asus motherboard kicks in an extra de-multiplier for the AGP bus, so that gets kept at 66MHz, and the PCI bus runs at half of that, so it stays at 33MHz, so you don't have to worry about instability with your video card or sound card or network card or anything else on the side while you play with your FSB.

Do a www.google.com for Athlon 2200+ overlocking kit or something like that - you should find some out there in the 12 to 15 dollar range. Then go to:
http://www.beachlink.com/candjac/TbredDecode0.htm
And look towards the bottom at the picture from DeerHunter. For the 2200+ (and this CPU ONLY!) you just need to make a little pin drawn conductive grease connection between the last two pins on the L3 bridge (make an arc to avoid the laser cut trench between them).
Once that's done, pop things back in, and you can test this configuration:
166MHz FSB, 11.0x CPU multiplier, 1:1, 2:2:2, Turbo - NICE performance boost here.

That's what I'm running currently. That keeps the CPU at only 1826MHz and mine at least runs quite cool (47 - 51 C depending upon load - to see max open the Asus probe util or something similar, let 3DMark 2001SE loop for a couple hours, then quickly check the CPU temp). But it raises the FSB to 166MHz, running synchronously to the memory bus, with fast memory timings still (2:2:2), the chipset still in Turbo (only works on the 1006 BIOS - not the later ones, up to and including 1011... maybe 1012 will fix it again), and your AGP and PCI buses are at default speed.

Further testing you can do:
Try raising the FSB some more, lowering the clock multiplier if necessary (you can run with multipliers from 5.0x to 12.5x if I remember correctly - the link above has details). Of particular interest are the values that don't raise the AGP speed much. You'll see numbers like:
133/33
134/33
135/33
136/34
etc...

The first is FSB, second is AGP.

You can also try running the FSB at 166MHz and kicking the FSB to memory up to 4:5. You'll probably need to drop the memory timing to 2:3:3, or possibly even all the way down to 2.5:3:3. You might need to drop from Turbo to Optimized too.

Again - make changes, test for performance, if it looks good, confirm stability. If you're OCing the CPU, watch your core temperature. If you're OCing the AGP bus (and therefore video card), watch for video noise / jitters / tearing, and possibly instability (lock ups).

Mostly, have fun :)  Gather your benchmarking tools, try a lot of settings, and see what makes your box scream!

With only a GeForce4 TI-4400 running at default speeds (I haven't touched any video card overclocking stuff yet - still doing other things) I can get a 3DMark 2001SE score of 11,100 now. That's about an 8% performance gain from where I started, and the video card is not overclocked at all, and the CPU is only overclocked 1.4%. My Sandra 2002 scores are up about 15% under the same conditions. And again... I can do a lot more. I've already confirmed that my system runs stably with the CPU over 1900MHz (I forget the exact value I tested), so I definitely have headroom if I want to push it. Plus I can OC the video card some.

Good luck! Post here to let us know how you do or to ask more questions :) 
!