MSI K8T Neo2-FIR overclocking

[sykopath79]

Yesterday I purchased an MSI K8T Neo2-FIR motherboard at the local PC Club (a steal at $70 from the open box item table... I love those guys). Everything works amazingly so far, dropped in a 3000+ Winchester and it's running like a champ, even feels faster than my 2500+ Barton @ 11x200 did. But, naturally, I want to squeeze some more power out of it.

I am new to AMD64 overclocking, have been reading a lot about it online this evening and it definitely is a more complicated process than any previous overclocking I've done. Can't just bump up the FSB and call it a day, HyperTransport throws a couple kinks into the process.

Much of what I've been reading, including the primer on AMD64 overclocking here on the AnandTech forums, indicates that setting a non-1:1 memory clock ratio is one of the magical steps necessary to pushing the CPU to its maximum. Based on my investigation of the available BIOS settings for the K8T Neo2-FIR (board came with the latest BIOS, v3.3), I see settings for CPU multiplier, HTT frequency, and whether to set RAM speed by SPD or manually at 200/266/333/400 MHz, but I can find nothing that seems obvious for setting LDT-to-RAM clock ratio. Also I see mentions of making sure you're not overclocking your SATA channels; I see where the BIOS allows you to lock the AGP frequency, but I see no explicit mention of PCI and SATA clock locks, so I do not know if locking AGP freq. also locks these by default, or not.

If anyone here has experience overclocking with the K8T Neo2 board, please post here or PM me with some help. I will likely have a few more questions as I go along also.

Here's a quick rundown on what I'm running:

MSI K8T Neo2-FIR
Athlon64 3000+ (Winchester) w/ Gigabyte 3D Rocket Pro HSF
Mushkin Level II PC3500 memory (2x512MB dual channel)

 

[sykopath79]

bump

 

[bob661]

I'll try and remember what I did but I was never able to go over 2.5GHz 1:1 on my 3500 with that board (when it was installed). The AGP and PCI buses are the only one's that get locked. You'll have to keep the total HT speed at 1000MHz or lower but that is easily accomplished by lowering the HT multiplier. I never tried dividers because I really don't understand how that works yet.

 

[bob661]

The SATA locks I'm not sure about and in my research I couldn't get a definite answer on that. I believe the Promise controller allows higher OCing but having my drives on the Promise or the VIA didn't seem to make any difference. To get the 2.5GHz I ran 10x250. I used Clockgen after I set the AGP/PCI lock in the bios for my OCing. I don't remember my exact ram settings other than a CAS of 2.5 and I think my TRAS was 10.

 

[sykopath79]

Apparently the choices for setting RAM speed at 200/266/333/400 are the RAM ratio settings, so that answers that question. Now I guess I just need to determine through grunt work what the speed ceiling is for each component. Still uncertain about SATA speed locking, right now I have one SATA drive connected to SATA1 (which is on the VIA controller).

I'm really anxious to see how much I can squeeze out of this one, because by all accounts the Winchester 3000+ has quite a bit of OC headroom on it. I would like to hit at least 2.2 GHz with it.

 

[bob661]

Let me know how things go. I'm putting that board in my wife's computer (can't leave a good board sitting around) and I may decide to OC it again.

 

[sykopath79]

Well, I was using the MSI CoreCenter utility to find my speed ceilings from within Windows, as the guides both here and at HardOCP suggest, and it froze eventually as expected, so I hit the Reset button to reboot. Well, Windows has not booted since the last time it froze, and I think some files got damaged when it locked up because even when I reset the clock speeds back to normal, it still won't boot.

Right now, I am running the Ubuntu Linux 5.04 64-bit (LiveCD version) and I'm clocked at 2200 MHz, 1.5 vcore, with RAM running 2-2-2-5-1T @ 5:6 ratio. Running fine so far, gonna download the Linux equivalent of Prime95 to my thumb drive and stress test from inside Linux. Wish I would have had that idea before I went and f*cked up my Windows installation, but I think a simple CHKDSK /R and repair install will get that back on its feet, plus I planned to do a clean install this weekend anyway.

So, so far so good at 2.2 GHz, with room to grow I think.

*EDIT* Grrrr, can't seem to figure out how to make mprime run in Linux. I'm a real Linux n00b though, so there's probably something I ain't doing right.

 

[sykopath79]

Well, I'm Prime95 stable at 2250 MHz, 1.48 vcore, RAM 204 MHz @ 2-2-2-5-1T @ 2.7v

I think I may have a wee bit more headroom on the CPU, but I don't want to push my vcore too high. I could run Ubuntu Linux x64 Live CD just fine at 2340 MHz and vcore 1.55, but Prime95 failed immediately after starting the torture test.

Very satisfied so far though. This thing is freaking FAST in WinXP.