Phenom 9500 w/ MSI K9A2 Platinum

[linhvndiy]

Very cool, thanks KTE a lot.

[Daveburt714]

Thanks for the info KTE. I just ordered one too!
Hopefully I'll be as lucky as you and Tony...

Looks like once you figure out your MB/Bios problems you may be able to get similar results in the 3+Ghz range.

As for your problems, If anyone can figure it out, I'm sure you can!
Nice to know your chips didn't die, you had me a little concerned..

Keep us posted...

[KTE]

This is just excellent - check this out =>





What more could you want from a tool

Franck, excellent work, thank you!

Guys, you're welcome. I'm only trying to report bare facts and happenings as they occur.. so we can figure things out altogether. There are enough children after spreading FUD about AMD/Phenom at the first minutest chance, so we have to ignore them and just work things out to see facts. Things are just very confusing, even for me. Just look at how long and how many different setups its taking me to work things out.

I have not recommended Phenom before when asked - but this chip is definitely recommendation worthy IMHO. Need much time yet though.

As for the Pi guys, I'll give you some Pi times in a little while despite me being awful at it

Now the tweaking starts

[tictac]

errrr..... where is the 3GHz+ clock?

[JumpingJack]

errrr..... where is the 3GHz+ clock?

It's coming... have faith.

[KTE]

LOL, who said anything about 3G?

Sorry Cutie, I thought you knew this is the 2GHz zone

In case you missed my inference... 3GHz is not hard to screen even at stock volts.

Let me fire up AOD.. one sec.

There ya go.




Sorry too busy with other things yet, I'm sure many more will be eager to present many results but I'm going to take my time, learn and tweak things before presenting. Results will come... maybe by next April or so (jk)

Will need time.

[justapost]

I have to go off, am very busy elsewhere, but let me say it is definitely much better than the B2 in terms of max speed. 2900MHz benched stock voltage however it was not stable. 2800 was rock stable though.

Same here 2.8GHz are rock solid on stock voltages.

BIOS, 3200 no HSF 1.47v gets stuck at Windows logon.

Seems to be the first sign for too low CPU voltage if takes forever to boot windows or if it's stuck on logon. Already had it a few times in the >2.8GHz range.
Looking at our last 3GHz screeny it may be an os problem.

This time around, I'm going to start off by undervolting at stock speed rather than the other way

Did this last time it was quite time intensive.

This is just excellent - check this out =>

Thanks for pointing on this, have my monitoring stuff on laptop so i can't use that feature. HWMonitor needs a logging feature.

[JumpingJack]

... HWMonitor needs a logging feature.

Seriously, no doubt!

[KTE]

Yeah my main problem is ATi GPU drivers. I have them installed. If I go high volts they corrupt and I can't clear them (like now ).

I just tried throwing VID high, could screen 3100=>




But when I put it back 2500 and just kept it at 1.55VID, ATi drivers BSoD within 2 minutes.

[jonspd]

so is bios 1.13 able to run the 9850? thinking on ordering from the egg shortly


nice SS's btw....

[KTE]

I think I should be able to get 2.9 to 3G fully stable here, as in daily use. That's what my aim is anyway. But voltage/heat/wattage does pump up very quickly after 2850. I'm really not bothered about max MHz, go LN2 and phase if you want to see that or you never will.

I'm trying to tweak CnQ and MSRs - they're not the same as B2 ones.
In my tinkering I have ran 2G NB at 1.050VID already without problems.
I have been running 1.225VID 1.201v 2600MHz and not had any trouble yet [ATM].

Just to clarify what you see in the above HWMonitor screenshots is:

First you have individual voltages. All are higher than real DMM measurements (real is tuned to output 12v/5v/3.3v exact idle).

Then you have DC wattages for each rail from the source. Apart from "peaks" and 3.3v over 10A, they are all extremely close to accurate. In the total DC Wattages/Amps:-

Look at 12V1, that's the +12V CPU rail only.
12V2 does not exist there, it is not linked up, and HWMon is just doubling the 12V1 over [cancel that out - P-Tuner doesn't pick it up either]
12V3 and 12V4 are both combined to give the above 12V4 reading.

CnQ
So when you see 25.3W 12V1, then that is the power Phenom 9850 BE was pulling at the stock CnQ MHz/Volts. Just to correct it, the actual wattage was 23.52W DC without factoring in VRM losses, that are near the 50-60% region for this board at such current.
System AC wattage was 110W.

Stock no CnQ
Same with when you see 45.96W, just that it was the power being drawn at 2600/1.296v CPU - 2000/1.296v NB setting. Just to correct it, actual wattage was 44.76W DC without factoring in VRM losses, that are near the 60-68% region for this board at such current.
System AC wattage was 123W.

CPU-Z is broken with B3 - many combinations are not validating. Just tried present 200x11.5 2300 setting, its rejected although real. So is 2400, 2200, basically everything but an odd one. We'll have Franck take a look at it.

Jon I'm on P0J, like I said I haven't flashed any other BIOS yet, if I did, I may not get another startup for many days, so I'm delaying it.

[Oldguy932]

Are there any specific things that you want me to try with the Phenom under phase? Or just how it handles htt clocking and such? I'd test the difference between air and vapo, but the only other AM2 mainboard I have running is my crosshair, and I'm not sure I'd want to put it in that to compare because of the difference in bandwidth.

[KTE]

Check this link out guys, it has all the hardware tested and verified on Plat v1: http://global.msi.com.tw/index.php?f...&cat3_no=#menu

Oldguy, there's a lot to check out =>

RAM clocking
IMC clocking
HT Ref/Link clocking
Lowest stable
Highest stable
Best Multi
Max benchable
Perf. Scaling
Coldbug
Volts for plus 3G stable

And then compare to if any of it makes any gain over air and water
Maybe they actually scale with voltage and cooling - even though this is obviously not top bin, top bin is 2 bins above, this is the same bin as what they had in Barcelona B2 set for Xmas release in server world which was canceled when the bug was found and lower than the 9900 they had, but still, it is best bin launched yet so good to test where this arch is going.

[SocketMan]

If ati drivers=pain
try Omega (think it's 7.12 based) and or Atitraytools,don't rush the phenomenal
there stretch the pleasure

[jonspd]

What is your option on the 9750?

Also are you going to try for 1.13 bios and check for idle freeze?

[KTE]

Yeah SocketMan I'm going to wait because MSI said they'll give me a preliminary NB option BIOS and then install them, thanks.

Jon
Did you mean "opinion" instead of "option"?
I'll try 1.13 when I pick up Xig, can't for now but... I don't understand your first qs

I've achieved my first goal: CnQ tweak for daily use.

Needless to say, this chip definitely pulls more amps than the previous ones. Less voltage, lower MHz, CnQ, it is idling 20W AC more with the same PSU.

Anyway, this is what I was after -> the chip is purely a tweakers chip

Idle (CnQ): 14.88W DC (not-incl. at least 50% lost in VRM conversion efficiency)



Idle Full Speed: 43.64W DC (not-incl. at least 40% lost in VRM conversion efficiency)



CB10 at 2800MHz



Sorry fellas incl. those I'm conversing with through PMs and those I owe some help to, I'm off, very tired today and been extremely busy too, so excuse me

[KeZzZu]

KTE: is it good result to get 1708 kt/s in winrar bench multithreading on cpu stock and mems on 1:2.66 divider cpu : phenom 9500 i was kinda stunned
mem timings 5-5-5-15-30-2t @ 2.1v bios version 1.40

Also ordered 9850 BE... i really hope... that iw will be good one if not ... have to buy second one.

[jonspd]

yea sry opinion on the 9750 and have you do any idle testing to see if these are unstable at idle like the b2s where?

[Oldguy932]

So that means you want me to put it in the crosshair so I can get a prelim for how well it overclocks and such? I can do that as long as there is a bios to support the B3's. Too bad that system only has a 7300LE in it...

[Brother Esau]

Sounds like you some DFI love Tye

Also what would be a safe "guess" for stable expected OC from New B3?

[tictac]

Yeah my main problem is ATi GPU drivers. I have them installed. If I go high volts they corrupt and I can't clear them (like now ).

I just tried throwing VID high, could screen 3100=>




But when I put it back 2500 and just kept it at 1.55VID, ATi drivers BSoD within 2 minutes.

Stop teasing us... show us that 3.5+Ghz clock

[Jaivan]

@KTE Could you please test and see if there are any stability issues in x64 vista and if can clock as high in it as 32 bit version.

[Brother Esau]

Hey Tye is the B3 Revision still get the Vista TLB Patch enabled?
I am curious to know how all of this is going to get implemented or corrected now that B3 Revision is out?

[KTE]

Just messed about with it, found some time. As you know I like to keep a track record of every time I mess about with it, I write notes in a TXT file to compare later.

WISE: in corporations these are a set of applications referring to Word processing, Internet, Spreadsheet and Email but cover all the majority of regular system usages.

Issues
*One of the biggest issues with Phenom 65nm B2 and Intel 45nm C0 is degradation, how, what, after how long and how to avoid or contain it.
*The second is low volts bench/screens but very high volts for real stability.
*The third is to avoid WISE stability problems.
*The fourth is how to have a perfectly stable system for our needs, in all tasks we will do not just FPU stability.
*The fifth is x64 stability and oc.
*The sixth is stability with 2x2GB and 4x2GB.
*The seventh is C1 errors at boot.
*The last being, does it respond to voltage after 1.39v? (unlike B2)

Load Stability Test
14x 200 2.8G is perfectly fine stock volts (1.296/1.288v idle - 1.276v load)
14.5x 200 2.9G is perfectly fine only for benches stock volts
15x 200 3G fails load stability stock volts outright

Idle+Load 24/7 Stability
2.8G was not stable - it froze after 16 hours during WISE. I jumped MHz.
2.9G was no go at 1.4v real, fails WISE stability after 8 hours although can bench easily.
1.4v 2.9G did not even boot with CnQ turned on. Disabled it then.
I jumped to 3G, that's my ultimate aim to see if B3 can get 3G stable and what limits it. Glad to say, 3G is within realms quite easily, mine is running smooth at 200 x 15 3000, benched it, primed it only 35 mins and am mainly running the WISE test now [256mins into it - 2/3 days is the test range for it].
BUT there's a caveat =>
After 2.8G you suddenly have to jump up volts very high to get any stability. It might seem no go to those who get whimpy after 1.4v AMD rates the CPU max 1.55v within range.

3.00G Stability
I tried every boot voltage, even 1.3v booted 3G but was not stable. I then just jumped to 1.558v BIOS, which gave me 1.592v Windows and 1.584v real. I then worked my way down. Lower limit was found, 1.540v idle will not be stable, will bench easy but fail WISE stability.

I have managed a stable setting. ATM it requires 1.552v idle and 1.528v load for perfect [so far] stability. This looks too far for many though. I have mentioned already, this is a hot and power hog chip esp. with volts. Here's an example:-

3G 1.6v idle was 136W DC just the CPU.
3G 1.564v load was 267W DC just the CPU.
On load, 84C at max fanspeed is easily reached! [cooler is not clamped down and without TIM, bear in mind]

3G 1.552v idle is 97W DC for CPU, which is what I'm at now. I have tried enabling CnQ and am successfully running it right now at 400MHz 0.7v with 3000MHz 1.552v being the load setting. CnQ gives me 12W DC idle CPU power.

Basically, this is an Opty 165 all over again. You will get 1.45-1.58v 3G and more fully stable but you will need a good cooler. ATM I am severely limited by cooling, for water, this chip should get 2.9-3.2G stable IMHO. I have booted 3300 max at 1.592v but it is too much for my cooling [yet] and it freezes on saving CVF file or IMG file from CPUZ. Installed the new AOD, it crashes my system pretty quickly which is worse than the previous one.

Here's what I'm testing ATM [setting right now]:-




Some [totally unoptimized] 3G benches FYI:-

Everest MEM/Cache


2k6


wPrime 32M


PiFast (Single threaded)


Fritz


NuclearusMC


Cinebench 10 2800, 2900 and 3000


SuperPi depends highly on MEM and with this, yet, I can't oc MEM/NB, so it'll have to wait. I'm expecting around 24.6s 1M at 3000MHz, its very possible to get 26s flat though.

Sandra MEM/Cache linear latencies I've checked out only at 2800.


The 1ns is L1 cache, the 4ns is L2 cache, the 8ns is L3 cache and after it is DRAM access.

Proper benches, i.e. tweaked will be later. Have to work this thing out first.

KTE: is it good result to get 1708 kt/s in winrar bench multithreading on cpu stock and mems on 1:2.66 divider cpu : phenom 9500 i was kinda stunned
mem timings 5-5-5-15-30-2t @ 2.1v bios version 1.40

On a 9500? Not sure, it isn't for 9600 though. 9600s stock 5-5-5-18 1066 can do 1872KB/s without optimizations with a decent BIOS (For MSI 113 will do that). Many BIOS kill MEM performance though.

Also ordered 9850 BE... i really hope... that iw will be good one if not ... have to buy second one.

Depends how you go about it and what you expect. I wouldn't base any hopes on what Tony showed, that is for water and above only.

yea sry opinion on the 9750 and have you do any idle testing to see if these are unstable at idle like the b2s where?

I've answered that above now and about 9750, no idea. I've not seen it myself first hand nor tried it and until I get to try 2 or more, I could never say. It may be just like 9600 B2 was, many good and many not so good.

So that means you want me to put it in the crosshair so I can get a prelim for how well it overclocks and such? I can do that as long as there is a bios to support the B3's. Too bad that system only has a 7300LE in it...

Don't have to but trying air and then vapo would be good enough for all

Sounds like you some DFI love Tye

If I go DFI, it has poor MEM oc/stability and I value that quite a lot
I'm running 2x1GiB 1066 5-5-5-15-11 75ns tRFC @ 2.05v with MSI stable

Also what would be a safe "guess" for stable expected OC from New B3?

2.7 to 2.8G stock HSF, 2.8G to 3G 3rd party air, 2.8 to 3.2G water is my estimate on 32bit systems.

Stop teasing us... show us that 3.5+Ghz clock

Hehe, kinda like no way on this cooling. I can POST and be in BIOS at 3500 though which is far better than B2

Hey Tye is the B3 Revision still get the Vista TLB Patch enabled?

Vista SP1 kills B2 MEM perf but nope, patch is applied based on CPU stepping code.

@KTE Could you please test and see if there are any stability issues in x64 vista and if can clock as high in it as 32 bit version.

Let me try and ascertain how stable 3G is in XP/Ubuntu 32b and then I'll move over to 64b systems to check the same settings

[SocketMan]

This might help.
It's not Tony's "water park workshop", but should keep the Phenom reasonably cool.
http://www.vr-zone.com/?i=5142