3000+ w/ MSI Neo2 OC prelim results...UPDATE!!

[Duvie]

First I only was testing it to boot up and watched the voltages, and temperatures...

I was using a 300 watt antec over at MarkFW800's house...

We tried to oc and really no go with bios that was installed. We flashed t bios 1.37mod and we were off and running....

Multi 9
HTT 3x
FSB 295
Ram 133 for 194 or 388ddr
Cas 2.5, 2,6,10,4
Vdimm 2.7v
vcore set at 1.55 but MSI harware monitor reports 1.44-1.45 volts (WTF...does it undervolt this much??)
I know the vcore controls work cause it needed a boost at 270 from the 1.5v set ....

2655mhz (not windows tested stable)

Tried 296 with cas 3, super relaxed all timings, vdimm 2.8v, had not vcore options...

How do you get 1.6v on this board especially with the undervolting...I was using a 300watt antec (rock sollid) on all rails and tested with voltmeter and was correct.

When I set 1.4v it would read 1.37v
When I set 1.45v it would read 1.42v
When I set 1.475v it would read 1.44v
When I set 1.5v it would read 1.44v
When I set 1.55v it would read 1.44v

Get my points!!!

saw idle temps from 35c low to 39c high throughout the OC...ALL with stock opteron cooler....

I still nead to flash to 1.41b mod perhaps....

I really get into this tomorrow and tweak throughout the weekend



EDit:

OK it appears the bios 1.37mod has higher then 10% options for the vcore however reports are they don't work (save) above 10 with any current bios except for 1.36b...I have will have to flash to this tomorrow. Still wish I can figure out the undervolting, but I will hook my enermax 465 up to it next.....


UPDATE:

After reading a bit I heard to use the over vid %....

Using 3.3% the board reads 1.48v and I was able to post into bios at 299fsb or 2691mhz

I then needed to run 5% or 1.53v to get stable at 2700mhz and booted at 2736mhz but not 2755...

2754mhz posted at 8.3% or 1.58v holds steady there with an occasional 1.6v spike....

I think I am getting to the end as each jump of 3-4 fsb requires a boost.

Going to try for 2.8ghz and call it quits and go into windows testing...

Each speed I let run for 10min in bios where I moved around and worked at will and monitored the voltages and temps....40c is the highest idle I see at 2754mhz...+12v fluctuates a bit others hold well, I notice this does not happen unless I use above 3.3% over vid...May have something to do with the board...

 

[thegimp03]

wow 2.6+ on air on a 3000+ is awesome

 

[Duvie]

Not windows tested yet, but I may have solved the vcore option thing and may push a little further still....I don't trust the temps for sure yet so I will definitely check on what my final load temps are....

 

[Zebo]

ttt

 

[Duvie]

Got basketball tohight so maybe later I will plug her into my enermax psu and use the percentage as suggested by Pastorjay...See if I can get 300fsb and 2.7fsb to boot

 

[Markfw]

Go for it ! And its Markfw900, not 800. My Honda CD-900F was my inspiration for the 900 when it said markfw user was already taken when I registered. (the one you saw in my garage)

But I digress. The board really rocks for OC'ing, but I am sure its a matter of the right bios to get it tottaly setup. I would guess from what we saw, you have a good shot at 2500 windows stable, maybe more. My 3200+ winchester won;t even post over 2400, and thats not even a stable post ! Also, I think it proves that a good quality PSU is worth way more than 500 watt generic ! Especially when you "weigh" then (literally) I had a 400 watt powmax also right next to the loose 300 Antec, and the Antec "weighed" twice as much !

 

[Dustswirl]

Bios 1.41 is quite good for ocing and has the highest Vcores measured (till 1.6V @ 10% though the board undervolts by 0.1V)

 

[AnImuS]

2.8ghz :shocked:

Duvie: I also had a similar setup as yours cpu,mobo,raptor,dvd-r burner what a coincidence we both just recently moved from P4 to AMD.

One thing i do miss is the smoother multi-tasking.

 

[Duvie]

2800mhz actually posted but at 414ddr it took me to drop to 2t with 2.8v vdimm and 10% over vid to post...That is 1.63v and that is more then I would ever go....

Edit: I did have the ram do 460ddr (230fsb) cas 2 but with 2t timings, until I read this would be a big performance hit....

I actually wont take anything over 1.55v anyway....


BUt it is neat to see a 1ghz bios post

Things I must hold

1)near 400 or over ddr
2) 1t timing
3) 1.6v or less ( I may live with the 1.58v)
4) no idle temps in the 40's but considering I am using stock opteron cooler withstock thermal pad that is pretty good
5)3x HTT
6) 9 multiplier cause the fsb would be insane with anything lower...

 

[Duvie]

So far the best one I like is

2664mhz @ 1.48v with 394ddr 1T, Cas 2,4,10,4 @ 2.7v
3x HTT w/ 296fsb

the

2700mhz @ 1.53v with 400ddr 1T, Cas 2,4,10,4 @ 2.7v
3x HTT w/ 300fsb

is nice as well

EDIT: These have all booted same settings with dual channel (2 sticks of the PC3500 Geil Ultras) and recognized as dual channel...I think I may hook up the cdrom drive and boot memtest off of it and let that run overnight....


Realistically I know none of these will likely be prime stable at these vcores....IN the end I will be happy as heck to have 2.6ghz stable with around 1.5v vcore and running 1t Cas2 timings in dual channel....

 

[stelleg151]

what week is your chip?

 

[Duvie]

Week 36....

NOt the newest but I have always heard getting a chip around 6 weeks in pretty early after fabbing so since we are in the roughly 47th week now I am not too far off from that....

I think there are a lot of good chips out there I just think ppl may be screwing up alot of these OC's

1) They skimp on power supplies...look at how that 300watt antec powered my chip the other night. Tight rails, actually tighter then my enermax right now for some reason.

2) They skimp on ram. Buying budget, value, and names I havn't heard of but swear they saw a good review once...I bought geil pc3500 ultras for my Intel and they have been tested to run 460ddr at cas 2 and 480ddr at cas 2.5 and that could have been low since my abit IC7 only gives a max of 2.78v of vdimm...

3) they don't do there research on boards with best option, theones with glitches, and how the bioses work....I spend a good week on just the board I have narrowed down before I buy, and print out all the tips form other users.

4) Don't flash to the newest bioses

5)Forget to set ram timings, dividers, etc to lowest during intial oc to isolate cpu and its perfromance in relation to vcore. I ran all my early test at cas 2.5, 2t and sub 400ddr (133 divider) to do just that. ONly after I have seeming found the cpu limit at given vcore have I retested it with dual channel, 1t, and cas 2.

 

[Duvie]

40+ minutes into memtest with cache on,all test, and 2 passes done...NO problems left. If I let it run until the morning it will have 10+ passes under its belt (knock on wood!!!)

 

[ZobarStyl]

I've got a 3200+ Winchester on my Neo2, watercooled, and I can hit 2.6 easy (at pretty much same as stock temps) but it's never Prime95 stable. At 2.5 it's more stable so I've stayed at that but I've not figured out the best BIOS for going over those restricting voltages. You have a link to these beta BIOS's? And with them, what's the max volts people have put through their core so far? Like I said, the chip is temp-wise fine, but I don't want to fry it from within. BTW, nice O/C.

 

[Duvie]

Well 14 passes of memtest completed...shutting it down, and going to work...

 

[Duvie]

Well 14 passes of memtest completed...shutting it down, and going to work...

 

[DAPUNISHER]

Couldn't agree more that most the unimpressive OC's are due to inexperience and/or inadequate hardware.

I'm looking forward to seeing what the max nuke-proof overclock for constant use is Looking forward even more to what your subjective impression of your first A64 system is, and wether it benefits your tasks IQ or not :clock:

BTW, glad to see you finally took Mark up on his invite!

 

[Markfw]

I also agree that the motherboard is key and that the Abit mobo sucks for OC'ing (or I got the worst 3200+ winchester made) I was here somewhat assisting on the first overclock (I am sure Duvie could have done it all by himself) and I did try all the same things we tried on my 3200+ with the same memory (PC3500 CAS2) and even a better PSU (Antec true power 380), but the mobo was the big difference. I may even sel this suckers and get the MSI. I just couldn;t bring myself to buy one after 3 DOA's in a row for my last three MSI boards!

 

[DAPUNISHER]

LOL! Yeah I winced as I submitted my order for this MSI board, and must have ask myself 50 times while waiting for it to arrive "What have I done?!?" Must say I am happy I rolled the dice so far.

 

[Duvie]

It was a good sign to see memtest with cache on pass solidly.....I am going to format my harddrive and install windows tonight and start the stability testing....


NOt to interested in synthetic benhes so don't ask for sandra...

I will run superpi 1m and 2m as well as 32mb fro stabiity testing

FH unit (stability)

prime95 (stability)

sandra burn in

I will loop 3dmark2k1 and 03.


I need to see some temps to at load so that will be nice....

What do ppl think of the silent booster that anandtech and others used. they seem sufficient, and if the low idle temps now are any sign I should be mid 50's for heat at load with this stock opteron one. (courtesy of MarkFW900)

 

[SickBeast]

Congrats on an incredible overclock, Duvie. :beer:

It's great having another CAD monkey/hardware enthusiast on these boards so I can discuss my hardware situation.

With that kind of system, you should seriously look into using ADT 2005 for your architectural work. It's incredible software, but it's extremely demanding in terms of CPU/memory performance. Your system should be able to run it like it's some kind of joke.

My XP Mobile is holding up surprisingly well, but I'm going to buy exactly what you just got in a month or so (I'm waiting for NF4). I'll probably keep this rig as a rendering box. I'm still waiting for 1GB sticks of DDR to come down in price. It seems like 512mb has been the standard for years now.

 

[Duvie]

I use ADT2004 now and I need to test it cause the HT really helped in rendering times...plus the P4 was so good at mutli tasking....

We will see in the end how it compares to my P4. I see ppl getting these LGA775 mobos and 3.6 and 3.4e cpus and hitting 4.0+ghz some on air....

The one thing I love is this thing runs cool relatively in terms of my P4 and I think I was going to disable some case fans and lower my noise....

Edit: Also do you know of any 64bit capable CAD apps...I wanted to test them. I here there precision is much better and the renderings come out far more detailed...

 

[SickBeast]

Originally posted by: Duvie
Edit: Also do you know of any 64bit capable CAD apps...I wanted to test them. I here there precision is much better and the renderings come out far more detailed...

I'm not aware of any. Precision and rendering detail shouldn't be any better. The main benefit of 64-bit will be reduced rendering times and better realtime manipulation performance. Oh, and you can use more than 4GB of RAM.

POV-ray may have a 64-bit port, but it's simply a rendering app.

64-bit seems to give a 30% boost in many cases. I'm looking forward to Autodesk porting ADT to 64-bit.

Are you rendering using the VIZ component of ADT? I'm interested in seeing some output from that program. Everything I've seen from it was very subdued and lacked the lighting, relfection, and shadow quality that I've seen from MAX.

 

[egeta21]

 

[Tweakin]

Originally posted by: Duvie
Multi 9
HTT 3x
FSB 295
Ram 133 for 194 or 388ddr
Cas 2.5, 2,6,10,4

How did you figure out the 194 fsb while using the 133 divider? I wan't to get my memory back down to the 350-400 DDR range for 2-2-10-2 1T timings.