Alright... Overclocking an A64.. Where do I start...?!?!? Dizzy....

[eletero]

3000+ test

Now.. 3000+ was also somewhere there.. I had problems trying to remove that rediculously tight heat sink socket lock... nearly ripped my skin off..

It booted fine.. 1.83Ghz.. bla bla bla... dual channel.. 128 bit bla bla
Bios now... can't drop multiplier.. stuck at x9
Raised HTT slowly.. 220.. fine.. prime stable.. 30 minutes
230... fine.. prime stable.. 30 minutes (2.070 Ghz)... prime test temp 70C
250... fine.. prime stable.. 30 minutes (2.250 Ghz)... ditto at 73C
267... fine.. prime stable.. 30 mintues (2.403 Ghz)... ditto at 76C
270... fine.. prime stable.. 30 minutes (2.430 Ghz)... ditto at 78C
275.. crashed in prime test.. boots windows XP fine.. (2.475 Ghz)
285 (wild chance).. crashed prime test.. boots windows XP fine... crashed after 2 minutes in windows (2.565 Ghz)
290... yep.. you guessed it... din even pass IDE checking.. (2.610 Ghz)

This is very interesting seems like your stable speed is around the same as mine (I set it to 265 eventually). I have very cheap RAM too - was actually thinking of buying some new RAM as I thought this was why I couldn't reach 300Mhz as in the hexus article. Perhaps hexus were lucky with their chip?
One thing that did help my overclock was setting the RAM timings manually - without doing this my max overclock was around 230. You could also try taking the HT down to 3x instead of 4x.
The other interesting thing is the temperature - I have the asaka evo33 heatsink on it's lowest RPM and the CPU doesn't go above 38C idle and 45C loaded, 78C is very hot.

[noah]

maybe the temp readings are a bios issue, i had to flash my dfi nf3 ut board to get it to read the right temps...?

[iranu]

I would do what Rave says and replace the crappy pad compound with some proper artic silver and repeat the process. It would be interesting to see the difference between stock thermal pad and proper paste. 78°C is way too hot for prolonged use.

Although I must say that the 3000+ venice does look really promising and you maybe able to get close to 300 HTT with proper cooling. What's the stepping code on the 3000+ btw?

[sawyen]

Went home.. with my supervisor's permission..

Too late for any proposal with the old cooler.. unfortunately enuf.. I've already paid my pockets off for a Gigabyte 360 Cooler super Ultra something something.. quite simple to install.. setting it at low speed setting was literally silent.. but up from 3500k RPM and above was utter rachet..

This time round... multiplier all the way to x9
LDT down to x3
Memory Timings: 2.5-3-3-8
Memory Clock: 266Mhz

200Mhz... Prime stable 30 min.. 47C load (max fan) - 1.8Ghz
250Mhz... Prime stable 30 min... 50C (max fan) - 2.520 Ghz
290Mhz... Prime stable 30 min... 54C (max fan) - 2.610 Ghz
---- so it was the fan...

no enuf time for other test, have to be back by 4pm.. 290MHz is still running at the moment.. should be going into its 2nd hour (if it didn't crash)... My 3500+ is effectively at FX-55 speeds minus the 1MB L2 cache... Will do some benchies tonight.. 3dmark05, HL2 etc...

Lets cross our fingers 290Mhz will live thru the next few hours.. Will be nice if I could breach 300Mhz HTT... that'll be so kewl... but will probably stick with my stock multiplier at 238MHz x 11 (2.62Ghz).. makes it easier for my XMS..

[sawyen]

Although I must say that the 3000+ venice does look really promising and you maybe able to get close to 300 HTT with proper cooling. What's the stepping code on the 3000+ btw?

3000+ not mine.. and he's off to his gf's tonight.. won't be able to know till he comes back. And lets cross our fingers again that he hasn't installed the CPU back or else I'll have to pay for the extra thermal gooey stuff..

[Rave]

Lets cross our fingers 290Mhz will live thru the next few hours.. Will be nice if I could breach 300Mhz HTT... that'll be so kewl... but will probably stick with my stock multiplier at 238MHz x 11 (2.62Ghz).. makes it easier for my XMS..

That's impressive clocking for a K8T800 Pro board whether you make it past 300 or not. As for the speeds you end up with, you should just run at whetever gives you the best combination of CPU and memory bandwidth, 11x238 might be better than 10x262 if the RAM can run tighter timings at the lower speed.

[StormPC]

Actually tighter timings do not affect AMD64 systems like they do Athlon XPs. With A64 it's all about MHz (memory and CPU) and running a 1T command rate in most cases.

Basically, if you do things like CD ripping and crunching SETI you want to go for the highest CPU clock you can get, then the highest memory speed that is 100% stable. Don't worry about timings or even 1T CMD RT. If however you are a hardcore gamer (or do anything that requires moving huge chunks of data in and out of RAM) you should go for less CPU and more memory bandwidth. A lower CPU overclock will also tend to allow a higher GPU overclocks on most systems. Even here timings are not that important, as the A64 is not hungry for bandwidth at all, especially the dual channel ones. Sure you can increase performance a little with the right timings, but any increase will probably result in reduced overclockability on your RAM and/or CPU.

[Rave]

Actually tighter timings do not affect AMD64 systems like they do Athlon XPs. With A64 it's all about MHz (memory and CPU) and running a 1T command rate in most cases.

....which is why Oppainter used/uses VX at 267 2-2-2-5 rather than TCCD at 300+ with loose timings:

http://www.xtremesystems.org/forums/...ad.php?t=51237

And while I'm refuting your posts:

http://forums.hexus.net/showthread.p...d=1#post493342

[StormPC]

Refute your heart out, but you really need to get a hobby or something if you're so bored that you believe refuting my posts is the best use of your time.

Do you have Eric's cooling? Do you have anything in his rig? OPP goes for the world's fastest score in 2k1, which is just about all he does. My advice was about general A64 performance based on application types, not getting the world record in 3DMark2001. Also, if you are going to refute me at least read my posts. I never said timings didn't matter, just not like they do on an XP and not in every application (3DMark2001 notwithstanding).

[sawyen]

oookay guys... Anyways 290 Mhz Prime failed after 4 hrs of stress test. I tried 3DMark05 early this morning.. literally 4am... crashed pretty much on every OC except softer ones like 2.2GHz (250MHz HTT).. which is a little confusing considering all 260,270,280 lived thru Prime tests for 3 hrs or more last night..

Then lightning struck, it seems I've solved the problem with games and 3DMark05 crashing and stuff. I've opted the bios for a 74MHz AGP lock rather than the conventional 66MHz lock... and voila... the games went smoothly and there were no crashes seen since... havent got time to run 3DMark again.. but at least NFSU2 isnt crashing anymore... Dun know why excatly, probably a the 66MHz AGP lock was causing my 6800GT to crash?

Prime is currently testing out 9x285MHz HTT... till the sun goes down. Prime test had pretty scary temperatures considering I've already got one of those 3D Cooler somthing GT something... 57C!! Core voltage was set to 1.55V.. since 1.525V on 290MHz crashed... Northbridge was set to 1.65V... DRAM was 2.65V... AGP was 1.5V

11x235 went like a breeze on 1.525V... load temperatures were at pretty kewl 54C.

[oralpain]

Set the AGP to 67MHz, it's quite possible that 66Mhz disabled the lock all together. Turn off fast writes as well.

Also, prime95 is primarily a CPU test, it's not very good for testing memory (though it will find more obvious problems) or anything else. I run several instances of prime95 simultaniously, using a different test for each one. This seems to find problems a bit faster.

For memory use contiual loops of test #5 when using memtest, it will show problems much faster than cycling the standard set of tests.

[sawyen]

Umm.. the only 2 option of AGP/PCI lock available to me on AV8 bios was 66Mhz and 74MHz... How do I change it to 67MHz? Might be able to do it once I flash it with a new bios... Memtest+, ran test 5 for 30 pass on 170MHz (266MHz DRAM) for my DDR400 XMS on 2.5-3-3-8, was fine... not sure what you meant by simultaneous prime test.. I'm using UBCD to run prime... can u do simultaneous tests?

[oralpain]

Well if 66 and 74Mhz are the only options for AGP and 74Mhz is the stable option, continue to use it. I was unaware that that baord only had two options for agp speed.

What speed did you have your memory set to when running memtest? You post was unclear.

I've never used UBCD. In windows you can run more than one copy of prime95 by adding the -An (where n is the instance number) command line argument.

For example:

"C:\Program Files\Prime95\Prime95.exe" -A0
"C:\Program Files\Prime95\Prime95.exe" -A1
"C:\Program Files\Prime95\Prime95.exe" -A2
...and so on.

[sawyen]

I read the new bios revision for this board had a flexible AGP lock.. not sure.. will have to flash to see.. but currently its only 66 and 74MHz

I've set my memory to run at 266MHz (133MHz) in the bios.. since i've made some OC to the HTT, naturally the memory would be running higher than 133MHz.. if I'm not wrong, its currently running 177Mhz on a 9x285MHz HTT overclock. Memory timings are 2.5-3-3-8... based on speeds given in Scan's website for 200MHz.. was inclined to push it to run at 2-2-2-8 timings, but I figured that it might not make that much of a difference.

[oralpain]

Ok, just wanted some clarification,as there is no point in running memtest at a memory speed lower than what you will be using.

[sawyen]

285HTT prime stable for 11 hours... cant go any further, cs I need to use Office XP... hahaha.. tweaked to a more cooler config.. now running stock speed 11 x 200HTT.. Been really cruel to the CPU these few days.. might give it a lil break.. So here's what I can summarise..

CPU: 3500+ Venice
Mobo: Abit AV8 3rd Eye K8T800Pro

maximum HTT at 11x multiplier attempted : 238MHz
maximum HTT at 9x multiplier attempted: 285MHz
maximum HTT at 9x multiplier attempted (but failed): 290MHz (4 hrs prime)

Toastiest OC temperature: 61C on 11th hour at 285MHzx9... (fan on 50% max 3000 rpm)
Sadly, my system with 8cm 2 case fans, Gigabyte 3D cooler Ultra, 6800GT at Ultra speed, DVD-CDRW, 120GB SATA, 40GB IDE... voltage above 1.525V core, 1.65V NB was too much for my 450W PSU... had to disable the case fans to run at 1.525V core.

Shuggs....