Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

Man what NB speed are you running?

Stock or close to it. Very low HT, to try and eliminate any variables apart form clock speed.

With an 14,5 cpu multi a 10x nb multi is very unstable here. An 11x multi is alot more stable but needs ~0.05V more than with an 14x cpu multi.
Anyway 2.9/2.2 works fine here with 1.4V/1.35V only limit are the temps here. They must have reached 85°C here because thats where the board powers off and the system was of this morning. During the firs 15 minutes or so the cpu temps where in the 70°C area.

Yeah unfortunately my chip is not like yours, its worse. Either that or the board/BIOS is bad, but I don't think its the latter for some reason, maybe I should now test another 790FX board and see. Not load stable, load stable it can easily do 1.48v ID / 1.440v LD 3055MHz - but the idling stability which has plagued us. Temps are good with this cooler, TT120 and Zalman 9700 could not handle this heatload even for 3 minutes under load.

I am thinking of picking up your board again... MSI is just not letting out a BIOS I can use to oc properly and I've ran out of time for computing this year - which BIOS are you using and do you recommend? Is there a large vDroop and vDrop on the board?

DDR2-1300 kit sounds promising.

It'll most likely be like the Team Group kit, 2.45v 6-6-6-18 1300 - I hope not, but I have a feeling it will be.

Stepping is in my sig. It has booted at -40C because whenever I change bios settings I have to turn off the psu because the vapo turns off because the computer doesn't do a soft boot after changing the settings. Stock vcore is 1.3v. I'm doing a little bit of random testing right now, keeping the nb and cpu at 1:1, about to run some stuff at 250x10.

Good to know, at least there's no CB upto -40C

I'm currently testing the below setting: 1.516v ID / 1.480v LD 2914MHz. Load is stable, idling stability so far is OK, more than 80mins into it but need to let it run at least a day or two to see if it fails.




Tried 2940 at the same voltage before, it failed after Crysis for about 30 mins, back to desktop, nothing open, idling, it froze.
I'll test NB/HT/HT Link/RAM max and stability later. I'm not looking forward to it though.

[aGeoM]

aGeoM, I never use the Red mode, it gives me no boost at all. It'll give a boost to higher end GPU perf. though.

Yeap 2x 3870 Crossfire here.

Sami said at the beginning to go up in little HT MHz and for quite obvious reasons, you have 5 PLLs being changed and a chance of system corruption if you muck up.

Got it.

MEM bw, well yep, they should be if testing properly but they were just two random shots paired together for posting sake

BTW, in your log file I found all references to Griffin, but it shows internal name AMD K11h. The PLL coding is there meaning they have the tool working with it.

Is that normal? I was using the 9500.

[KTE]

Yeah, it's just normal... they added support for new CPUs that's all.

What's max you're running on your cards stable?
With Spider there will definitely be tweaks uncovered yet for multil-GPU setups, I've just not had time to check. I would throw in >2 3870X2/3870 but its too much hassle and especially since I haven't even sorted the CPU side out yet, which is my main aim. I only used more than one GPU on the 790FX once and that was around 55 minutes to check if it worked, how many GPUs worked, what the issues were, if it benched using all of them with scaling, the full load power, which PCIe slots worked with what bandwidth, what were the fitting issues and which slots were slightly better, so can't say much.

[aGeoM]

Yeah, it's just normal... they added support for new CPUs that's all.

What's max you're running on your cards stable?
With Spider there will definitely be tweaks uncovered yet for multil-GPU setups, I've just not had time to check. I would throw in >2 3870X2/3870 but its too much hassle and especially since I haven't even sorted the CPU side out yet, which is my main aim. I only used more than one GPU on the 790FX once and that was around 55 minutes to check if it worked, how many GPUs worked, what the issues were, if it benched using all of them with scaling, the full load power, which PCIe slots worked with what bandwidth, what were the fitting issues and which slots were slightly better, so can't say much.

Cards are stock cooled, no Vmods, clocks programed in bios, for 24/7.

Been fighting with CPU/BOARD too...

[KTE]

That's the same 3870 I tested

Stability Update:-

2909/2902MHz 1.516v / 1.480v - idle and load stable
2914MHz 1.516v / 1.480v - idle unstable and load stable
2947MHz 1.516v / 1.480v - idle unstable and load stable



So I tried, this, my current setting, a compromise. So far it has worked excellent without problems=>

3010 x1/2900 x3 1.516v / 1.480v - idle and load stable [still going]




Seems one or more cores are weaker, 3G stable is there, at least on one core

[d412k5t412]

What is the difference between CPU VID and CPU VDDC?

I change the vcore in the BIOS and it changes the CPU VDDC. But i've noticed that if i chance the CPU VID in AMD Overdrive, I get more stable overclocking. Any help?

[Oldguy932]

Well, this is the first good news I've really had so far from my 9850. 2750 for the nb is stable at 1.325 nb vid. Beyond that I'm trying for 3ghz nb, and its taking ALOT more voltage, 1.525 is being tested for stability right now. Once I get it stable I will post some screenies, only got everest bandwidth tests to run at 1.5v. Mem was 500mhz 5-3-3-9 so to not worry about it at all, latency was like 81 with the nb at 2.5ghz, at 3ghz it dropped to 75.7. Too bad this chip doesn't clock well... 2.8ghz took me 1.4v but that might just be the vapo.

[Brother Esau]

Tye, what seems to be the best programs to accurately test the Phenom please? Also I used Sam2008 TLB Fix tool and Enabled the TLB Fix through his program and find that it yields the same result in WinRAR as the Phenom B2 Stepping did Any idea on whats up with that?

[LIKMARK]

Yelding as a patched b2 or unpatched b2?

Edit: Read in your DFI thread that it is patched. Of course it will yield about the same as a b2 as long as you disable the same parts of the L3

[KTE]

Stability Update:-

Bad news but good news in a way too.
I had the last setting ~15 hours idling stable (meaning WISE stable) - until I went off, came back and tried opening this Google Book whilst scrolling up and down quickly, whilst zooming in/out: http://books.google.co.uk/books?id=T...=en#PPT1546,M1

It froze within 5 secs
So I tested more and more idle stability by the same procedure because it worked very well and instant.. would my previously stable settings also fail that test?

3010 x1/2900 x3 1.516v / 1.480v - idle failed and load stable
2947MHz 1.516v / 1.480v - idle failed and load stable
2928MHz 1.516v / 1.480v - idle failed and load stable
2914MHz 1.516v / 1.480v - idle failed and load stable
2909/2902MHz 1.516v / 1.480v - idle failed and load stable
2898MHz 1.516v / 1.480v - idle failed and load stable
2870MHz 1.516v / 1.480v - idle failed and load stable
2860MHz 1.516v / 1.480v - idle failed and load stable
2828MHz 1.516v / 1.480v - idle stable and load stable

See how much FAR below real stability was?
The difference being, I was testing for instability, as in actively finding it.
All of those settings were UT3, Crysis, Memtest, EVEREST, Prime 95 stable BTW, but they obviously only test portions f the CPU.

I have not dropped volts yet, just changed speed settings before I can finally declare what is stable for daily use. I am very GLAD I've found a way to very quickly test the instability other procedures were not detecting, now I can finally map things out and move on instead of running around in nightmare circles. I skipped a region below 2860 to speeden things up, now I'll test in-between, as I find the stable setting, I'll then rigorously test it with dropping volts until I find the lowest. Before I needed 1.44v for 2860 stable, with this new test, who knows. This is what I'm currently at right now =>

What is the difference between CPU VID and CPU VDDC?

I change the vcore in the BIOS and it changes the CPU VDDC. But i've noticed that if i chance the CPU VID in AMD Overdrive, I get more stable overclocking. Any help?

You know, a search will always help you and me especially on something explained on every other Phenom thread page

All you need to know is;

VID = not CPU volts, sets accessible voltage boundaries
VDDC = CPU volts

Well, this is the first good news I've really had so far from my 9850. 2750 for the nb is stable at 1.325 nb vid. Beyond that I'm trying for 3ghz nb, and its taking ALOT more voltage, 1.525 is being tested for stability right now. Once I get it stable I will post some screenies, only got everest bandwidth tests to run at 1.5v. Mem was 500mhz 5-3-3-9 so to not worry about it at all, latency was like 81 with the nb at 2.5ghz, at 3ghz it dropped to 75.7. Too bad this chip doesn't clock well... 2.8ghz took me 1.4v but that might just be the vapo.

This seems weird, I'll wait to see the b/w numbers before I can comment but latency is a direct function of IMC speed in K10h. So you were testing DDR2-1000 5-3-3-9 2.7-3G NB?

Tye, what seems to be the best programs to accurately test the Phenom please? Also I used Sam2008 TLB Fix tool and Enabled the TLB Fix through his program and find that it yields the same result in WinRAR as the Phenom B2 Stepping did Any idea on whats up with that?

Yelding as a patched b2 or unpatched b2?

Edit: Read in your DFI thread that it is patched. Of course it will yield about the same as a b2 as long as you disable the same parts of the L3

It's expected because it's changing the same CPU performance registers as LIKMARK said

There is no one program to properly test Phenom stability, we have to rely on the known honesty of individuals.
The better way I see it is, make sure settings are benchable, Memtest stable, then boot into Windows and test stability in Everest > Prime 95 Small FFT > Gaming > Run that setup daily, open web-browser and start trying to open many tabs/sites incl. many programs from Start menu, and load THIS sites inside browser and start flicking through the pages as fast as you can whilst zooming in and out - don't reboot, keep it idling like 12-15 hours just to see - if its unstable, it'll fail pretty damn quickly. That link I'm giving above is a very good idling stability test for Phenom if you follow that procedure.

[justapost]

KTE have you tried a higher NB multi? Above 2.8GHz I get much more stability with an 11x NB multi. Critical test for my cpu is the sandra memory latency benchmark.
Will stay away from ocing that thing for a few days, have work todo.
Things have not really changed compared to the 9600BE it seems, to reach the point where random instability issues occure we need better cooling now.

[KTE]

Nope, not tried higher NB multi for stability testing, I'll try it after I get max stable CPU found, thanks. I try keeping below my NB safety range. I don't have issues upto 2380MHz NB at 1.3VID with stability. With 9600BE the easy way to catch instability was Sandra testings for me too, but 9850BE is fine upto 2380 - 1MHz more and it fails [freeze/reboot under Sandra loads].

Maybe water would help but with my chip, I'm not so sure.

[justapost]

I just noticed while playing above 2.8GHz that the freezing issues occurde much less frequent with an 2.2GHz instead of an 2.0GHz NB.
First i dropped HT/NB/mem speed and relaxed mem subtimings to get 2.9GHz somewhat stable. I sorted out NB voltage as the cause of instability first.
At 2.8GHz i required 1.275V for 2.2GHz nb speed at 2.9GHz it are 1.325-1.35V now. 1.325V are the minimum required for 2.0GHz nb 1.35V are good for 2.2GHz. I have not tried to find the voltages required for 2.4GHz NB at 2.9GHz , for an everest memory test i ran at 1.45V CPU/1.4V NB at the beginning but that was not really stable.
To repeat myself: With 2.0GHz nb speed and 2.9GHz cpu speed stability issues occure as soon as i change the ref HT (201MHz is enough). No such problems with an 2.2GHz nb.

Maybe water would help but with my chip, I'm not so
sure.

My 9600BE did not like voltages above 1.375V. I have not seen an 9850BE with those limits. Seems AMD fixed that issue.
Now with higher voltages the range we can oc those chips has exceede a bit , say 100-200MHz more that the 300-400MHz average we had with B2's.
After that the issues occuring are similar to those the 9600BE's had with lower volts, freezings during normal workloads specialy.
Due to the higher voltages the cpu can handle temps become an issue now if you don't have high end air or water cooling.
I don't think water will help in your case your temps are ok at 1.5V, more likely your at the the voltage limit of your cpu here. Look at oldguys first results he needs equal voltages for 2.9-3GHz on the vapo. Stability at a given frequency seems to be an voltage only issue as long as temps dont exceed 80° or so.

BTW: I followed your footsteps once again and ordered an xigmatec cooler. Results look awesome for that price.

[KTE]

Yeah, I just had a freeze at the last setting:
2828MHz 1.516v / 1.480v!
That's not possible, given that I've had 2902MHz idle/load stable before at 1.48v /1.464v

I am checking it again now - what I now did is decreased voltage to 1.45v / 1.421v and so far it has not given me an issue (done a few WISE tests which normally fail, load is still within stable range).

It could be CPU dislike of voltage above 1.44v, its possible - heck anythings possible with Phenom

Your point about more than 200HT Ref. is also something I might be witnessing, but I have not tested to confirm this yet. I'll try increasing multi and decreasing HT to 200 to check.

I'm still OK with that though, since I can boot high NB speeds and change VIDs within Windows. I rarely shutdown a PC for my home usages.
I'll try 2200 NB speed through multis now.

As for NB, I had no issue with upto 1.55v for it with 3x 9600BEs on this MB. They all reached 2596 HT/NB [unstable] and 2592 [bench stable] and then crashed after that.

Xig cooler is excellent price/performance

MSI just told me again that they'll pass on latest oc BIOS as soon as they have it [meaning they don't yet]. I'm starting to get fed up waiting.

[Oldguy932]

It might not be your lack of vcore that's causing the instability. I think it might be the lack of other voltages you can change, most likely your nb. The nb has a sweet spot for voltage, my 9850 liked 1.325 and it clocked like mad up to 2750. And I meant the memory was at 250*2=500mhz. I have to test it at 1.6v now, it still can't get through cinebench without freezing with 1.575. Also I left the cpu speed at stock with voltage dropped to 1.15v.

[KTE]

Yea that could definitely be true there Oldguy. Since my last post I have booted 2800/2200 at 1.408v ID / 1.376v LD and it is working flawlessly for over an hour or so now => [AOD increased my Vcore on start]



It seems something other than my cores and their voltages are messing things up. Like Achim and you say, the NB may need massive volts for full stability on the 9850 even 1MHz above stock

[d412k5t412]

Thanks for the help KTE

I knew I could search, but I just wanted to hear it from you
Have you tried the AutoXpress option yet with your 9850? I have mine on and I got around 500 more points in 3dmark06.

Which program detects the 9850's temperature correctly? In everest I get about a 10-15C difference then in AMD Overdrive.

I hope you can get most things figured out about these crazy phenom. A new BIOS would be probably help since the 9850 is new.

[KTE]

I knew I could search, but I just wanted to hear it from you

LOL man, no problem. Check this, it was by Sami early December http://www.xtremesystems.org/forums/...&postcount=167

Have you tried the AutoXpress option yet with your 9850?

Yep I have, I kept it On for all mainly but On gives me more instability than off for some reason

Which program detects the 9850's temperature correctly? In everest I get about a 10-15C difference then in AMD Overdrive.

Can you show me them side by side with volts? I don't get this difference.

Here's the setting I'm at now since my last post, managed to get it fully idle/load stable: 2800/2200 1.384v ID / 1.360v LD [you can see my diode temps in P-Tuner]






The DC amps on 12V1 are also correct, 9850 at that setting idles 4.6A at 11.97v. I'll keep those volts and now try Autoclock in AOD or whatever it's called. I know 1.36v ID / 1.344v LD is not fully stable, so the lowest I can go is in between.

BTW Franck the special man fixed the CPUZ validation bug so I can now validate my oc's. Will now finally start to oc for max MHz, had to wait for this - here's 3.1GHz 4-core validated: http://valid.x86-secret.com/show_oc.php?id=340566

[d412k5t412]

KTE: I actually havn't updated Everest in a while, and I just did. Now its giving me All 4 cores and CPU temp, before it was just CPU. Which one should I go by? CPU or Core? Thanks

[Brother Esau]

Whats PTuner? Is that Motherboard Specific Software?

[SocketMan]

Whats PTuner? Is that Motherboard Specific Software?

It's software that comes with Gigabyte GT Power supplies
Speaking of which does the 550w version have the voltage increase slider?(see pic)

KTE maybe your WISE instabilities are caused by something other then the CPU,
like a video card/ram/ (hopefully not) motherboard (as I just got one) or a
combination of the above?

[Tony]

KTE

Initial tests of P0J over 141 show the following.

Boot with Autoxpresss enabled brings on Yellow in AOD
Winrar scores are slightly higher than 141 as long as I set red in AOD
cinebench no different either yellow or red

I will continue to test, as for now I think 141 is the same bios with autoxpress just hidden ( I will recheck with amibcp)

I hit the same clocks, get the same ram speeds and the same bandwidth with both, even in everest.

Remember im using 4GB 2x2 in vista 64 vanilla, 5-5-5-15-11 1000mhz

T

[Brother Esau]

Can any of you gents fill me in with properly setting up prime95 Bleand/Custom Test and HCI Memtest with Phenom and 2x2gbRam Kit please?

On Prime Blend/Custom I entered 4000 mb Ram to be used but Vista Task Manager says 2.18gb Is it Vista Prime or me or combination of all

Please advise>>>>>>

Thank you, Campbell

[Tony]

just leave by auto man, or use OCCT and set the ram you want it to use

[Mathos]

....You know... Just out of curiousness, has anyone tried undervolting the CPU? I know for a fact that a 9600BE can do close to the same Clocks at much lower voltages. I know undervolting also works when OCing Intel CPU's from time to time because it cuts down on heat. Not to mention the IMC/L3 should be able to run at 2ghz at much lower voltage as well, unless there was some drastic change to the TLB electrical circuit during the B3 spin.

Dont worry too much about over volting until you get past 1.4v. 0.9-1.4v are the k10's voltage design specs. Which is why when most were trying to see how low the C&Q volts could go had a hard time going lower than 0.8v.