Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

It's just the pics that make it long man

But its all technical info needed.

Next time I'll post cryptic one liners

[jonspd]

TYVM KTE I can make my order sooner instead of later now thanks man ...


LMK if there is a better bios then 1.13 or along the same lines as the POJ IYO

[Oldguy932]

Well 2.7 was stable at that voltage on small fft's for 9 hours straight, not bothering to do wise testing because I want to know why there is such a voltage jump, I just started it at 271x10 and am going to se the difference every 10 mhz.

Edit: I think the board may be at fault here. 2710 was perfect with the same voltages as 2700. 2720 boots with the same voltage, but one core fails prime in about 10 minutes. I bumped vcore up one notch to 1.2v and windows would not even boot. It's just extremely odd and I don't like it.

[aGeoM]

Well 2.7 was stable at that voltage on small fft's for 9 hours straight, not bothering to do wise testing because I want to know why there is such a voltage jump, I just started it at 271x10 and am going to se the difference every 10 mhz.

Edit: I think the board may be at fault here. 2710 was perfect with the same voltages as 2700. 2720 boots with the same voltage, but one core fails prime in about 10 minutes. I bumped vcore up one notch to 1.2v and windows would not even boot. It's just extremely odd and I don't like it.

That's what been happening to me, when tried to go 3200/2400, it booted at vcore=1.45; NBv=1.325, system frozed after couple minutes, so whent back and up NBv=1.350, sytem lockup at Welcome screen, set NBv=1.325 and booted again to Windows.

[Tony]

you know KTE i think L3 responds better to low ram latencies over overall ram speed. I will have to play more this coming week, 3x790FX DFI's on the way so i can test all 3 and pic the best to stay with me

I will also back down to 2x1GB microns to see if I can push the ram latencies lower to see the effect on the L3 and winrar.

On a side note the DFI is quite a bit faster in 3d06 than the MSI is, 13700ish with the MSI Vs 13800+ with the DFI, so Oskar, Jarry and the boys did get something right

more soon...weekend DIY duties end soon

Also 9750 on the way along with an X3, so should be able to run 2 boards simultaneously now.

[Oldguy932]

Finally got around it. I set the voltage in AOD for it to be stable and primed for an hour and a half. It seems that's all it wanted cause it booted fine just now, so its on to 273 now...

Edit: Nevermind, seems that it was the other nb voltage. I have processor nb voltage(vNB) and then I have NB Core/PCIE voltage. It needed a little more juice in the latter probably because I'm running 1T on such a high htt, but it booted now so that what matters.

[Brother Esau]

It's just the pics that make it long man

But its all technical info needed.

Next time I'll post cryptic one liners

Who loves ya Baby?

Don't worry Tye I like your style very thorough and detailed a man of my own traits

[Oldguy932]

HTT vCore vNB Speed Mem Idle Max Notes
250 1.15v 1.25v 2500mhz 666mhz -16C -7C
255 1.15v 1.25v 2550mhz 680mhz -15C -6C
260 1.175v 1.25v 2600mhz 692mhz -14C -5C
265 1.175v 1.25v 2650mhz 707mhz -13C -5C
270 1.1825v 1.275v 2700mhz 720mhz -13C -4C HT multi dropped to 9x
250 1.325v 1.325v 2750mhz 666mhz -11C 0C HT multi increased to 10x, CPU and NB multis increased to 11x

CPU and NB at 10x multi
HTT vCore vNB Speed Mem Idle Max Notes
270 1.1825v 1.325v 2700mhz 720mhz -13C -3C
271 1.1825v 1.325v 2710mhz 723mhz -13C -3C
272 1.2v 1.325v 2720mhz 725mhz -13C -3C HT multi dropped to 9x, NB Core/PCIE voltage 1.2v, set vCore with AOD
273 1.2v 1.3375v 2730mhz 728mhz -12C -3C NB Core/PCIE voltage 1.3v, Command Rate 2T
274 1.2125v 1.35v 2740mhz 731mhz -12C -2C
275 1.2125v 1.35v 2750mhz 733mhz -11C -2C NB Core/PCIE voltage 1.4v
NB will not go any higher and be stable

CPU 11x multi NB 10x multi
HTT vCore vNB Speed Mem Idle Max Notes
250 1.2125v 1.25v 2750mhz 666mhz -13C -3C
251 2761mhz
252 2772mhz
253 2783mhz
254 2794mhz
255 2805mhz


So this is what I've been doing with my 9850 currently. The first table was what I was initially doing for testing, and then what happened for 2.75 made me start another test. The second test was the same thing only increasing the htt in 1mhz increments. The NB seemed to max out at 2.75 so I started the third table which will allow me to overclock the cpu a tad more. Doing this testing made me realize something about the causes of errors though. All the freezing I got was due to the NB not having enough voltage, that's why 2.75 on both the first time took so much vCore, it was lost and kind've used by the NB, which makes no sense but its what I'm going to think until it gets explained to me otherwise. I really wish I had the money to get another 9850 to do more testing on...

Edit: And sorry for the way the tables are posted, didn't know how to do it so I just copied and pasted from excel.

[KTE]

Tony, I tried checking out low RAM (+500 CSL4/3 etc) but I had no boot, forgot BIOS mappings and wrongly set NB FID [14x] and CPU FID [15x] at 1.2/1.2 Vcore and was stuck with no boot for hours - had to do the dreaded clear CMOS and yep, now its totally no POST, same C1 error

It's not dead but I've no idea when it will boot now, if ever. 3 days so far so can't check anything out.

Have you tried any 500-540 CAS4 Sandra [multi-threaded] runs and compared them with CAS5 600-630 runs?

Yeah, that's what I found too, L3 latency being RAM timings dependent.
Looking forward to your 9750/X3 testing, some of them seem to be on very cheap around UK too

Oldguy932, 275HT stable... stable as in you ran it daily, or Prime95? You had vNB 1.35 but what was the NB multi?
Have you tried 1.3v 200HT/2800CPU stability testing? Maybe if you jump up a notch, you can work your way down.

aGeoM, if system is having hard time booting a setting, or even if its low volts, it is always wise to try a few reboots at those settings to test stability. I do this all the time before thinking anything is stable. If its unstable, it might boot 2/5, 3/5 but no where near consistent.
Similar thing happens with my 3G 1.35v boots, but I don't bother wasting time over it since I know it's unstable.

Camp, thanks man

Looking at Phenom... AMD seems to have improved quite a bit in terms of clocking from their usual improvement cycles since 2003. And still, quite obviously, no Phenom yet is of top speed bin, but casual high volume bins from the same 2.2GHz silicon in a new step, hence the price being lower. Looking at the clocks, 2.8GHz bin for FX CPU is well within reach with these models by just speed binning the same 9850BE silicon, but even that wouldn't be low volume highest bin so can't be classified FX and carry that price tag. However, TDP is what will bolt them down to ≤2.8GHz at most as it did and does to all CPU MFGs including Intel. The industry does not accept ridiculous TDPs, PressHOT is what they will scream all over again. I still see them focusing for 45nm rather than any FX CPU now. Many AMD interviews around talking about clock/energy/IPC improvement in K10.5.

AMD speed improvement cycles =>

K8: Q4 2003 2.2 GHz to Q3 2007 3.2GHz: 45.5% clockspeed increase in 15 quarters
K10: Q4 2007 2.2GHz to Q1 2008 2.5GHz: 13.6% clockspeed increase in 1 quarter

Clockspeed improvement per quarter =>

K8: 3%
K10: 13.6%

So instead of sticking to this arch for too long like they did with K8, they are moving ahead immediately with a 15% reduction in power mentioned at the same clock speed as K10 for 45nm. By those official figures, looks like 3GHz will still be around 140W TDP.

[KTE]

MSI referred me to reseller for a better BIOS now

[jonspd]

MSI referred me to reseller for a better BIOS now

WTH is MSI not all there?

[dr_drache]

SUPER UPDATE :

I went ahead and got my gskill ram (2x2048@1066)
and had to set the timings manually 5-5-5-15 (and devider 1:2.66)
was working well, was looking up info on raid cards (don't think the speed of onboard is doing my raptors justice)
and, found that MSI posted new drivers for the Promise PDC42819
and new bios 1.4 final.
so i said, what the heck, lets format..... and i did some testing too...

SOFTWARE USED :
HD TACH version 3.0.4.0
WINRAR V3.71 benchmark

TLB shutoff with SAM's Tool Version 1.05

I wish i had screenshots of it all, might reflash and try again, but the results are this :

BIOS 1.3B3 -
WINRAR - ~1,100
HDTACH - Long bench (32M) ~96MB/s Sequential Read Speed

BIOS 1.40 - (flashed, booted, VISTA installed new promise devices/reinstalled devices, whatever, required reboot)
WINRAR - ~1,450
HDTACH - Long Bench (32M) ~150MB/s Sequential Read Speed (have picture)

Both HDTACH graphs had pretty much the same ramping, just differnt values at the keys.

Now, not very great on my testing, cause i didn't expect anything,
and it was just for my own, so i can located a better raid card (any suggestions? (remember, I am a army man (read:tight funds)))

can anyone test themselves? and see if i just found a fluke?

[Oldguy932]

The top two tables the cpu and the nb were the same speed, the third table cpu was 11x and nb was 10x. It just didn't want to clock at all past 2.75, pain to even try. I just put my 9600 back in because it runs faster on much less power than the 9850.

[justapost]

@oldguy: Above ~2.8GHz my system is more stable with an 11xnb multi than with an 10x multi. Seems it's a good idea to have max 500-600MHz difference between CPU and NB frequency. Also ram works best in 800 mode, 667 can not reach high ref HT's here. Have you played with the cpu on air also?
Anyway, thank you for the effort you put into this. The results keeping me back from investing another few hundred euros into a water cooling system.

[Oldguy932]

Keeping the cpu at around 20C-30C would probably yield the best overclock. I didn't use it on air at all because I'm not sure if the new crosshair bios supports it and I don't have the patience to remove the vapo from the M3A32 board. I did notice something different though. Having my 9600 regular at 255x11.5 with memory on the 533 divider and the nb at 9x only got me 7.7ns in everest, but the slowest I ever got on the 9850 ever was 7.2 ns and that was with the cpu and nb at 10x250. And changing the nb multi probably wouldn't help, although 11x250 was a bit more stable than 10x275, but I thought that was just cause I was maxing out the htt to get that speed running. Too late now, I just put the 9600 back in and I keep my cpu's in for at least a week at a time, not to mention it runs much more stable.

[KTE]

WTH is MSI not all there?

Yeah, after a totally day it really the out of me to hear the same
They basically still don't have a better BIOS is what they said.

I went ahead and got my gskill ram (2x2048@1066)
and had to set the timings manually 5-5-5-15 (and devider 1:2.66)
was working well,

Which set, what's the P/N?

was looking up info on raid cards (don't think the speed of onboard is doing my raptors justice)

Budget?

and, found that MSI posted new drivers for the Promise PDC42819
and new bios 1.4 final.

Great, here's the link to the new Promise drivers he's mentioning. Can we have some testers please?

Release: 2008-04-10

Version:
Win 2k/XP Ver.: 1.1.1030.27
Vista 32/64 Ver.: 1.1.1030.47

2k/XP: http://global.msi.com.tw/index.php?f...e&dno=6266&i=0
Vista 32b/64b: http://global.msi.com.tw/index.php?f...e&dno=6266&i=1

There were also new Ethernet and SB600 RAID drivers released that day as mentioned earlier: http://global.msi.com.tw/index.php?f...=&prod_no=1332

so i said, what the heck, lets format..... and i did some testing too...

SOFTWARE USED :
HD TACH version 3.0.4.0
WINRAR V3.71 benchmark

TLB shutoff with SAM's Tool Version 1.05

I wish i had screenshots of it all, might reflash and try again, but the results are this :

BIOS 1.3B3 -
WINRAR - ~1,100
HDTACH - Long bench (32M) ~96MB/s Sequential Read Speed

BIOS 1.40 - (flashed, booted, VISTA installed new promise devices/reinstalled devices, whatever, required reboot)
WINRAR - ~1,450
HDTACH - Long Bench (32M) ~150MB/s Sequential Read Speed (have picture)

Thanks for the info.. hmm, that's a very big bandwidth difference between BIOSes, are you on Vista?
Your WinRAR is one like from Vista SP1
This might apply to Vista mainly if that be the case. It can't b XP, memory perf. is to low for it.

Now, not very great on my testing, cause i didn't expect anything,
and it was just for my own, so i can located a better raid card (any suggestions? (remember, I am a army man (read:tight funds)))

Hehe I thought you guys just get everything free and 200k GI's

@oldguy: Above ~2.8GHz my system is more stable with an 11xnb multi than with an 10x multi. Seems it's a good idea to have max 500-600MHz difference between CPU and NB frequency. Also ram works best in 800 mode, 667 can not reach high ref HT's here. Have you played with the cpu on air also?

Yeah I think these are MB/BIOS issues rather than Phenom since I've not seen this happen on the MSI since the beginning

Keeping the cpu at around 20C-30C would probably yield the best overclock.

So far, that's what I found with all Phenoms strangely.

Having my 9600 regular at 255x11.5 with memory on the 533 divider and the nb at 9x only got me 7.7ns in everest, but the slowest I ever got on the 9850 ever was 7.2 ns and that was with the cpu and nb at 10x250. And changing the nb multi probably wouldn't help, although 11x250 was a bit more stable than 10x275, but I thought that was just cause I was maxing out the htt to get that speed running. Too late now, I just put the 9600 back in and I keep my cpu's in for at least a week at a time, not to mention it runs much more stable.

Yeah to test NB you really don't need Prime95 [etc] as long as you can bench it using Sandra sort of benchmarks. NB error shows up in uptime stability i.e. you'll have a freeze or crash within 15 hours of booting and general use if its unstable.

I tested NB 2400 1.25VID early on and managed it F@H and P95 stable and thought it was fine but it wasn't when I used the setting. I think I'm about to change boards any day now once again since I still have no boot, been 3 days.

[dr_drache]

@KTE,
sorry i didn't link everything, i was tired and had a long day, figured the info was enough for the time being.

OS is VISTA sp1 x64

RAM : NEWEGG LINK
can't locate box currently...
the original timings were something like 7-7-7-24
was thinking it was voltage now i'm thinking about it, on account it asks for 2.0-2.1 and the board still seems to only get me vdimm of 1.8

RAID CARD : buget, not looking to spend over say $140ish for a decent host based, bit more (not much) if a lower end hardware one shows up. will take used... lol (if possable, sata, only need 2 ports, and pci-e/pci (whatever will work best)

ohh about the pay, pay is decent, if i was with my family, but due to ingerys i am having a forced seperation, with her and my kid 800 miles away, so it gets tight at times running 2 households, and the GI bill (college) is only like 37K, maybe a bit more at times.

[bgbdbstrd]

BIOS 151 upped at the MSI ftp site.

[KTE]

Here's the latest 151 BETA BIOS: http://www.yourfilehost.com/media.ph...98A7376AMS.zip

I am definitely expecting something in it. Any takers?

Thanks to bgbdbstrd for the alert.

OS is VISTA sp1 x64

RAM : NEWEGG LINK
can't locate box currently...
the original timings were something like 7-7-7-24
was thinking it was voltage now i'm thinking about it, on account it asks for 2.0-2.1 and the board still seems to only get me vdimm of 1.8

Excellent RAM there, 2.0v should get you 5-5-5-15 533 stable

How do you know what the vDIMM after boot is?
You can check the SPD in SPD Tool or CPUZ and set the timings as rated in there.

RAID CARD : buget, not looking to spend over say $140ish for a decent host based, bit more (not much) if a lower end hardware one shows up. will take used... lol (if possable, sata, only need 2 ports, and pci-e/pci (whatever will work best)

I'm not really an expert at that price point, mainly server hardware for work which tops that price range by quite a lot

ohh about the pay, pay is decent, if i was with my family, but due to ingerys i am having a forced seperation, with her and my kid 800 miles away, so it gets tight at times running 2 households, and the GI bill (college) is only like 37K, maybe a bit more at times.

Ahh sorry to hear that, hope things improve soon

[dr_drache]

Here's the latest 151 BETA BIOS: http://www.yourfilehost.com/media.ph...98A7376AMS.zip

I am definitely expecting something in it. Any takers?

give me a few hours, i'll give it a shot


Thanks to bgbdbstrd for the alert.

Excellent RAM there, 2.0v should get you 5-5-5-15 533 stable

How do you know what the vDIMM after boot is?
You can check the SPD in SPD Tool or CPUZ and set the timings as rated in there.

i used HWMonitor 1.08 (64-bit) for my voltages/temps
screenie on my system, which you can see seems to read the dram at 2.21
it was 1.8-1.9 at bios 1.3B3 this is my first run of it on 1.40:


also CPUZ SPD tab :



i will try to go back in and enable EPP profiling (mistake, eather i can't find it, or it was a differnt motherboard i am thinking of) in the BIOS, after a 1.51 flash.

I'm not really an expert at that price point, mainly server hardware for work which tops that price range by quite a lot

you know, i feel that what i want to spend on raid, is way low, but at the same time, i have the promise chip on-board, so i don't HAVE to have one, and it's only for striped, no mirroring at all (do a self backup of everything i might want to ever keep, DVDs/Online) so, if i find one, sweet if not, blah.





EDIT :
did the 151 flash MORE INFOS :

Version V1.5B1 04/10/2008
changes seen in BIOS : TLB FIX DISABLE - added to cell menu

Ram timings : when set all on auto and Ratio set on 1:2.66 (for my 1066)
which then i manually set to 5-5-5-15



no other notable changes in WINRAR or HDTACH (all changes seen were within a few %, inside a standard varation default)

[KTE]

Thanks for the info

Is there no NB FID options in this BIOS?
Look around carefully for some AM2+ P-State menus.

DRAM voltage can't be accessed by any tool on this MB but for AOD, that value is just the 12V rail reading divided by around 5.5, which is how its being output [a confusion].

i will try to go back in and enable EPP profiling (mistake, eather i can't find it, or it was a differnt motherboard i am thinking of) in the BIOS, after a 1.51 flash.

EPP is Intel CPU boards
With this you just have to choose the 1066 mode and set timings yourself.

you know, i feel that what i want to spend on raid, is way low, but at the same time, i have the promise chip on-board, so i don't HAVE to have one, and it's only for striped, no mirroring at all (do a self backup of everything i might want to ever keep, DVDs/Online) so, if i find one, sweet if not, blah.

Usually, onboard ATI RAID is not much different to low end PCIe RAID cards in performance. No where near vital unless the tiny perf. is critical

did the 151 flash MORE INFOS :

Version V1.5B1 04/10/2008
changes seen in BIOS : TLB FIX DISABLE - added to cell menu

Ram timings : when set all on auto and Ratio set on 1:2.66 (for my 1066)
which then i manually set to 5-5-5-15
http://i28.tinypic.com/2089qf4.jpg

no other notable changes in WINRAR or HDTACH (all changes seen were within a few %, inside a standard varation default)

Values are quite low there, you can set them manually to your rated specs of 5-5-5-15 48 tRC and 105ns tRFC both [105ns is minmum].

TLB Fix Disable was in BIOS before that too, P0J, no changes it seems.

[batmang]

Great, but I am sure you can get more, you have one the lowest bin models I've come across - 1.2VID/1.2v 9600BE. It should be a better clocker than most others around, try below 2600 stock volts. Can it not go any further with unlocked multis?
Which BIOS are you on? What's your MEM rated for?
Try 212HTx11.5, don't change any voltages and check - it's a much better performance setting than what you're on and it looks like it should be stable at stock voltage. You are fine as long as you have sub 70C load temps in those utilities. That's well within rated spec, on the stock cooler, with no fan, I was reaching 55C load at just stock.

KTE, finally responding to ya. I'm running the stock heatsink and fan just to clarify. My memory setup is 4x1GB of G.Skill 6400 DDRII 800 (LINK). Bios I am using is 1.1 B3. I'm able to get up to 2.6GHz no problem within AOD but it eventually crashes. I even lowred the voltage to 1.18v and was able to run 3DMark06 at 2.6GHz but it eventually crashed. I'm not sure if my instability is my PSU's fault, I'm not sure how to measure that. The PSU is a generic 585w that came with my generic case. BTW, how do I adjust the multipler within the bios? Any other suggestions?

[dr_drache]

will attempt to change said memory settings BUT...


which setting in bios do i work with? DCT 0/DCT 1/BOTH?

because in both some settings (tCL for one) have multiple slots (Read : tCL had 2 options to change it's options)

AND tRC only goes to 42.....
or just changing them in memset good enough?

[batmang]

Welp, latest stable results.

2.6Ghz on cores 0 and 1. 2.4Ghz on cores 2 and 3. Stock vcore, 1.2v.

[KTE]

I'm not sure if my instability is my PSU's fault, I'm not sure how to measure that. The PSU is a generic 585w that came with my generic case.

Yea, most of them are complete junk. Which brand/model is it?
Get a quality one or you'll be at the very least slowly killing all your hardware

BTW, how do I adjust the multipler within the bios? Any other suggestions?

Using the AM2+ P-State options, CPU FID.

which setting in bios do i work with? DCT 0/DCT 1/BOTH?

Both.

AND tRC only goes to 42.....
or just changing them in memset good enough?

Should be good enough. Memtest 30 and see how it works, it should be stable.

2.6Ghz on cores 0 and 1. 2.4Ghz on cores 2 and 3. Stock vcore, 1.2v.
http://www.jmbat.com/media/phenomoc/..._3DMark06.jpg]

Can you not try 1.26-1.3v for 2650 across all cores?

======

Dropped my 4850e back last night, picked up 4450e X2 2.3G instead. Waiting for another MSI and DFI 790FX board now. Also dropped my 780G back and picked up the 740G instead.

Build cost:- £130

AMD Athlon X2 4450e 2.3G 45W - £40
Corsair 2x1GB 800 4-4-4-12 - £25
Gigabyte 740G - £35
Seasonic S12 330W - £30

First dime I spent on PCs since 15 months.. very happy with that