DLT3C OCing touble

[Furton]

Right, system spec first.

Leadtek nForce2 Ultra400 Delux Edition mobo
XP 1700+ DLT3C chip
2*256MB Corsair 3200 DDR RAM
SLK 800 + Vantec stealth fan

Got this up to 200*11 and it wasn't stable, have dropeed the multi down to 200*10 by now and it still isn't stable, Prime95 won't do the first test and 3DMark2001 crashes on the second test. Although doing stuff like browsing it stable e.t.c

My RAM timings are 2.5, 3, 3, 6 with 2.8V, CPU is at 1.75 and chipset voltage is 1.8V. Any ideas on how to get this stable?

[M@tt]

Have you tried testing one thing at a time

Test with a low multiplier and raise the fsb, testing as you go at your chosen Mem timings to see how far you can go.

Next raise you multiplier to test for the max Mhz you can at run at stable with your chosen Max Vcore (cooling will decide you max Vcore), keeping the FSB low

Then you can test with slack mem timings with a low multiplier to find the FSB limits of the Motherboard/something else, this does not always work with some memory/chipsets not liking some timings

[Austin]

Great advice M@tt, it's always best to use small steps up and be sure to test thoroughly. Keep all voltages stock until there's a need for more. Simply jumping to a desired speed (which I assume happened) is not only very risky but difficult to ascertain the limitation(s).

DLT3C will use 1.50v, 1.65v is actually normal voltage for all TbredB and Barton so you can go straight to that if you wish. Then take your speed (as suggested) up one notch at a time (testing) until you get instability. You can then add 0.05v more voltage to see how much further it allows you to go IF you are definite cooling isn't the limitation. Adding voltage above the core norm produces a LOT more heat and will almost certainly have some negative effect on the CPU, so always monitor the actual true perf gains and see if you think they are worth it. Rem to run your RAM 1:1 with FSB and be sure you aren't using o/c'ed frequencies or timings for it as you want to isolate the CPU as much in order to find its limits.

As a final point you should check whether you have TbredA or TbredB, DLT3C can relate to both and TbredA is a very poor core due to lacking the extra layer.

[Sinizter]

Is it JIUHB DLT3C or a Tbred-A core ?

Only the JIUHB ones clock well as they are TBred-B.

Also is your PSU upto the job? Sometimes having a dodgy PSU can hold you back from attaining stable overclocks.

[malfunction]

I bought a KIUHB XP1700+ recently - it won't run with a 200FSB even if I drop the multiplier to 5x and the RAM timings down to 11-4-4-3. Currently running fine at 195 x 11. Don't think it's the mobo (ASUS A7N8X Deluxe V2.0) or the RAM (2 x 512MB TwinMOS PC3200 which is 100% stable @ 6-2-2-2 @ 195MHz). After reading around apparently some chips just don't like the higher FSB. My next upgrade will be to a Barton (2500+) though and they're meant to be much easier to clock up to 200MHz + FSBs. I am getting a new PSU this week (450W CWT) so if that helps I'll post back and let you know.

[Furton]

Right, got 166*12.5 stable but on 200*9 prime won't even do the first test. I take it this chip doesn't like a high FSB then?

[M@tt]

Originally posted by Furton
Right, got 166*12.5 stable but on 200*9 prime won't even do the first test. I take it this chip doesn't like a high FSB then?

More likely your Motherboard or ram timings than your chip IMO

Just try 5-2-2-2 and see how you get on

[Furton]

It won't do 6-2-2-2 so why do 5-2-2-2?

[Austin]

Just to clarify the model code (eg DLT3C) can be TbredA or TbredB. According to AMD's datasheets DUT3C should always be TbredB (1.60v) but DLT3C (1.50v) can be A or B and it makes a sizable diff. Anyway here's what each character stand for...

1st letter is the package type (D=OPGA)
2nd letter is voltage (L=1.5v)
3rd letter is max die temp (T=90'C)
4th letter is L2 cache size (3=256K)
5th letter is FSB (C=133/266)

It is the STEPPING CODE which defines TbredA or TbredB, it's the last letter so JIUHB is a TbredB while AGUIA is a TbredA. If looking to buy a TbredB it's the stepping code you must be certain of.

KIUHB seem pretty new, AIUHB and JIUHB were the most common TbredB. TbredB always seemed to be happy at 400FSB+ but many early 400FSB capable mobos had problems running many Barton at 400FSB. Perhaps the KIUHB are marked as such because they can't do 400FSB but it's more likely just that particular CPU or perhaps another factor such as mobo (inc NB cooling), RAM (esp with tight timings), BIOS settings etc. Anyway the diffs between 333FSB and 400FSB aren't very big at all outside of synthetics so 'only' getting 380-390FSB is not going to limit you.

Furton since you have nF2ultra400 it makes sense to lock the AGP/PCI to std speeds, run RAM 1:1 (slow timings too) and then go gradually up from 166/333FSB to see how high you can get, 380FSB is really as good as 400FSB. As said rem RAM timings can often hinder your o/c and the actual real world diff between the fastest and slowest timings is only 1-2% so nothing worth limiting your o/c nor sacrificing stability for. Esp so on SktA Dual Channel mobos where most of what it does is simply reduce latency anyway. I'd treat your RAM like any other part when o/c'ing, set the RAM to the slowest (highest) timings and see what you can do in pure speed (eg 400mhz) and when you reach your desired speed (for CPU, FSB etc) then try tightening the timings in small steps testing throughly.

[Furton]

Austin: It's a JIUHB DLT3C I'm using right now. I'm sure my mobo locks the AGP/PCI at the standard 33/66, at least that's what it says on startup

For the rest. Tested each stick with another stick of TwinMoss 512MB module (Mouse and Keyboard wouldn't boot with just 256MB RAM? ) and ran Prime95 on both, 1 module failed straight away, looks like it needs to be RMA'd, the other is still going strong.

Thanks for your help.

[malfunction]

Originally posted by Austin
KIUHB seem pretty new, AIUHB and JIUHB were the most common TbredB. TbredB always seemed to be happy at 400FSB+ but many early 400FSB capable mobos had problems running many Barton at 400FSB. Perhaps the KIUHB are marked as such because they can't do 400FSB but it's more likely just that particular CPU or perhaps another factor such as mobo (inc NB cooling), RAM (esp with tight timings), BIOS settings etc. Anyway the diffs between 333FSB and 400FSB aren't very big at all outside of synthetics so 'only' getting 380-390FSB is not going to limit you.

The KIUHB I have has the 'new' packaging - green but you can see some of the traces (i.e from the pins to the actual core) under the surface ('laquer' ?) so maybe that's what the 'K' signifies. I really hope it's not the mobo - I've got pretty good case cooling, have tried the UBER bios (to increase the chipset voltage) and have even tried it in an indentical motherboard (my brothers) - both A7N8X Deluxe V2.0 / Ultra 400 mobos. I know that the mobo is the most logical cause (as RAM running at tight timings @ 195MHz ought to be able to make it to 200MHz @ loose timings) but as both the RAM and mobo are spec'd to run at 200/400MHz - and I've heard a few tales on nforcershq.com that say similar things - I'm hoping it's just the chip but - like you say going from 11 x 133 to 11 x 195 is still nice and not much different from running at 11 x 200Mhz.

[Ice Kay]

First test how far the FSB can go then play with the multiplier.

[Trickle]

Originally posted by Furton
Right, got 166*12.5 stable but on 200*9 prime won't even do the first test. I take it this chip doesn't like a high FSB then?

I would say unlikely. TbredB isnt known for giving up <200.

I'm assuming this is with the mem timings at 2.5,3,3,6? Have you run fsb's inbetween 166 and 200? Have you run high fsb's at stock cpu speed? Have you had a look to see if there are any bios upgrades available? Have you tried something silly like 3,5,5,8? Do you notice any differnt limmits when the cpu voltage <1.75?

Testing results with some different memory would be useful aswell. As would be testing the chip in a different mobo, but I'm assuming that isnt feasable. However you cant just jump to the conclusion that this is the cpu's fault.

[Furton]

Read my above post, one stick of my mem was faulty, it's now running at 205*10.5 completley stable. Not a bad OC and it's only at 1.7V. Will go further once I get my new RAM stick back.

[malfunction]

Originally posted by malfunction
The KIUHB I have has the 'new' packaging - green but you can see some of the traces (i.e from the pins to the actual core) under the surface ('laquer' ?) so maybe that's what the 'K' signifies. I really hope it's not the mobo - I've got pretty good case cooling, have tried the UBER bios (to increase the chipset voltage) and have even tried it in an indentical motherboard (my brothers) - both A7N8X Deluxe V2.0 / Ultra 400 mobos. I know that the mobo is the most logical cause (as RAM running at tight timings @ 195MHz ought to be able to make it to 200MHz @ loose timings) but as both the RAM and mobo are spec'd to run at 200/400MHz - and I've heard a few tales on nforcershq.com that say similar things - I'm hoping it's just the chip but - like you say going from 11 x 133 to 11 x 195 is still nice and not much different from running at 11 x 200Mhz.

New PSU = 200MHz FSB no problems. Changed up from an antec 380W to a CWT 450W (both actually look like the same manufacturer).