Phenom 9500 w/ MSI K9A2 Platinum

[oohms]

It behaves exactly like my non working 9500. I power on the board, a led between the dimm sockest starts to glow and the status leds jump to C1.
The M3A ran the 9850BE fine at stock before and does not post now. I'm sure i have no corrupted bios problems there.

Try sticking the cpu itself in a freezer for a couple of hours... I have heard it may help in certain situations with the phenoms

[MSIman]

guys whats the best cooler and fan for phenom 9500 and MSI K9A2 Platinum?
I have it and i OC it to 2.42 on my basic cooler and my max temp was (on torture 100%) 61 'C
I wanna clock it more but temp is going up to 75 'C...

[KTE]

Welcome aboard MSIman

Best cooler... between TRUE and Xigmatek S1283 - both are of the best for Phenom.

Anyway, I've made progress.

So far found max idle and load stable with stock voltage:

core0- 2900MHz
core1- 2800MHz
core2- 2700MHz
core3- 2800MHz

core1 at 2900MHz stock volts is no go.
core2 at 2800MHz stock volts causes the abhored freezing!

Just benched SPECviewperf10 at those settings if anyones interested...



BTW, the PTS site and test suite is very buggy and messed up. Just look at this: http://global.phoronix-test-suite.co....php?k=results
Over 6 of my submissions don't show up

Anyway, couldn't be happier now that I'm making progress

[justapost]

Welcome aboard MSIman
So far found max idle and load stable with stock voltage:

core0- 2900MHz
core1- 2800MHz
core2- 2700MHz
core3- 2800MHz

core1 at 2900MHz stock volts is no go.
core2 at 2800MHz stock volts causes the abhored freezing!

Just benched SPECviewperf10 at those settings if anyones interested...

Now that is indeen very very interesting. So with core2 at 2700MHz you don't get the idle freezing issue. Thank you for looking into this.

[KTE]

Yes, definitely
Took me a looong time to figure that out. ATM, uptime is ~9 hours without any freeze at above settings and all through it I've benched and done all idling tests rigorously too.

2800-2900-2700-2800 -> froze after 2-3 hours [2 tries]
2800-2800-2800-2800 -> froze after 11-18 hours [3 tries]
2900-2800-2700-2800 -> no freeze yet at all, doubt it'll happen, I've done all tests to provoke it

core2 should be able to do about 2760 at least stock volts, will tweak that later after finding upper boundary. I've tested 2756 without fail on all cores stock volts before.

I'm looking to finally manage stable (with voltage):

230HT
2300NB
3105/2990/2760/2990

You're welcome.

[informal]

Good progress there KTE!!
Finally a viable cause/solution for idle freezing with Phenom OC is here ! Nice work again

[KTE]

Thank you

Trust me, you don't know how relieved I am
What impresses me is, its all stock volts. Before testing like this, I believed max my cores could do was 2700MHz stock volts fully stable.
Now I've just turned it up a notch:

2900-2800-2700-2900 stock volts [testing]

No. of idle freezes = 1x10^n
Crash count due to load/instability = 0

[Andi64]

I'm also testing at stock vCore, core2 is working at 3.0Ghz and core3 at 2.9Ghz, 0 and 1 are the ones who feeze my system if I set 2.8Ghz

I have a problem with the backplate of the K9A2, I can't mount my TRUE!
I'll find a way around, any ideas?

[MSIman]

my are all stock volts i only turn up my ram up to 2.0V lol and it can go up to 2.7 on my phenom didn't test more cose i have a stack cooling lol
btw that cooler can't fit in my case...

[KTE]

Andi64:
Have you tried taking the backplate off?
Only 4 screws hold it up.

[MSIman]

and guys can you tell me about the TLB problem with 9500? i didn' have much problems with my processor so do i have to apply that TLB fix?

[Rammsteiner]

KTE, But now it turns out that one bad core might be the problem of all the freezing... What would the SB750 bring us. Just better OC'ing in general for the cores which can run higher?

[VVJ]

EDIT:

Slight improvement.

CBID seems to show current wattage very accurately....

Here is what I found while playing with Phenom 9500.



DOWNLOAD NOW

[KTE]

and guys can you tell me about the TLB problem with 9500? i didn' have much problems with my processor so do i have to apply that TLB fix?

When you get bored, do a quick search, there's tons on it.

None of it is relevant to us unless you start to get stock crashes, so don't worry about it.

KTE, But now it turns out that one bad core might be the problem of all the freezing... What would the SB750 bring us. Just better OC'ing in general for the cores which can run higher?

I've no idea, honestly

Here's my best [current] stable setting so far, had many BSODs and other software problems before it.

2900-2900-2700-2800 1.296v ID / 1.280v LD
Power change from stock - 8W AC





Load stable / Idle unstable:

2800-2800-2800-2800
2800-2800-2900-2800
2800-2800-2700-2900

Idle / Load stable (best so far leaving HT stock):

2900-2900-2700-2800

I can easily run 2800 on the 2700 core stock volts, but it'll fail anywhere from between 11-18 hours.

One major thing I want to say is: core numberings are not the same each time after boot, which threw a spanner in my testing for a long time!

Sometimes AOD core2 was core3 in AMD PowerMon/CPUZ/CCPUID, other times, it was other way round and it changed so much between applications randomly.

I then figured out, that generally, CCPUID gets the same core arrangement right, so I use that to change speeds.

Here is what I found while playing with Phenom 9500.

http://jungle.at.tut.by/work/cbidBE_...TPower_4XS.gif

DOWNLOAD NOW

Man, thats just excellent, you're a star
You managed to retrieve the register info into that nifty utility of yours which I've had to work out manually each time, and yep, its a PITA!

That's exactly what I was after - those are maximum theoretical values set by the cores at boot as mentioned in guide

Anyway, two problems: it picks volts/watts in all states perfectly fine.
But it doesn't register core multiplier change after bootup
What it does then is, starts to downclock HT/NB to make up the higher multi speed, i.e.



But in the mobility section it picks up multiplier fine but P0 shows wrong speeds:



And the P-State 0 value changes real-time: 2700, 2800, 2900, every second or two

The AMD PowerMon shows the correct multis and HT is stock. CoreTemp/CPUZ has the same problem but Arthur/Franck had fixed it earlier.

[KTE]

2900-2900-2700-2800 stock volts fully stable I can finally say
I doubt 3GHz is possible stock volts, but I'll try it next.

MSI told me BIOS 151 has the NB FID option
Hmm...

I was just looking around this last weekend.. where are Intel Q9650 and Q9750?
They were supposed to release early May, Intel will never delay the cycle unless there's production problems.

Anyway, single-threaded clock-for-clock performance is simply out of the equation to test and compare between processors as we should know for a while now due to the way Intel CPUs get overclocked. A little bit of fun is OK though, I mean to compare... just not those hardwired uninformed arguments

LOL, anyway, my new settings for 15 minutes.. going to bench SPEC now.

3000-2900-2700-2800 stock volts

[Suosaaski]

MSI support answered to me when I asked about adding the NB multiplier option back to new BIOSes:

Sorry there is no way to supply to adjust the multiplier for NB according to MSI BIOS spec as this item can cause nonbooting and we removed this. Please follow our suggestion not to adjust this for system stability.

So it's a no-go.

[KTE]

MSI support answered to me when I asked about adding the NB multiplier option back to new BIOSes:

Sorry there is no way to supply to adjust the multiplier for NB according to MSI BIOS spec as this item can cause nonbooting and we removed this. Please follow our suggestion not to adjust this for system stability.

So it's a no-go.

When was this answered?

We do it with 1.13b anyway

[JumpingJack]

and guys can you tell me about the TLB problem with 9500? i didn' have much problems with my processor so do i have to apply that TLB fix?

As KTE said, a quick google will yield much. Not having a problem with the processor does not mean the fix is applied, like all errata the TLB problem was way overblown. The chances of it yielding an issue are remotely small to basically none.

You can test if you have the patch applied by downloading a copy of Winrar, run it... if you get 1000+ KB/s, then the patch is not applied, if you get below 500 KB/s then the patch is applied.

Run the processor unpatched, it makes no difference.

Jack

[Daveburt714]

Hey Folks, just got my K9A2 back from RMA... WooHoo!
Thank Cod, that GBT DS5 was driving me crazy with the 9850!

The board came back with the 1.4 Bios installed on it, and I haven't bothered to change it yet. It does seem to be VERY stable and really the fact that it doesn't have a NB Multi is not that big a deal thanks to the 10x NB default.

There are a few quirks I've noticed though... Changing the voltage on the NB or the Memory doesn't seem to have any effect, the IMC stays at 1.3v and the Mem stays @ (what I consider pretty high) ~2.2v's...

Guess my question is, which Bios would you guys recommend 1.13b or 1.4?

BTW: I love the user settings in the MSI Bios, but I do kinda miss the Q-Flash option of the GBT board... Given the option though.. Screw Q-Flash...

[KTE]

Grats mate!

I hope it works for you... try 1.51 beta first.
You can can change NB VID in windows on it IIRC aswell.

[Suosaaski]

When was this answered?

We do it with 1.13b anyway

A few days ago.

[KTE]

^ Crap news I guess



I'm not going to trust these guys again.

Anyway, latest, incase anyone missed it

[KTE]

Hey, keep the support up for the board you guys. Sorry about this, but its best this way.

[see my next post]

[SocketMan]

When was this answered?

We do it with 1.13b anyway

I still do it with the HT multi,whats the deal on 9850 and 1.4 BIOS?
Does changing HT multi - changes NB, or am I "special"?

[ubuntu83]

^ Crap news I guess



I'm not going to trust these guys again.

Anyway, latest, incase anyone missed it

They ditched the Montreal Chip? Shanghai seems to be available in late november.