Could B3 Phenom be the rebirth of AMD....+1000MHZ OC done!

[Rammsteiner]

At this point I decreased NB multi to 10, increased Vcore by 0.025V (so 1.325V) and decreased GPU clock from 918 to 904Mhz.

Im curious how it will be now, but nothing else to do then just trying. Freezing in general is a sign of not enough Volts, but whether this applied to the GPU, CPU or NB, I dont know.

I know it crashed a little faster with a higher NB clock, but well I dont know what it's caused by. Not enough Vcore with a higher NB clock is more likely to crash faster I'd guess. I mean, it crashed also with stock Vnb and NB clocks so it's definately coming from the CPU or GPU. GPU makes sense because it's the real first time Im playing a game. But like I said earlier, the GPU did a 2 hour FUR test without crashing. Although Crysis is known to crash a GPU too if it's simply unstable while Atitool/FUR tests let it run without problems.

The thing what makes me not so sure about the GPU part though is that it happened while loading and no heavy graphics were done yet. But then again, Ati uses 'P-states' too, so whenever a 3D app starts it goes right into the 918/1.4V mode. And also, as said before, higher NB clocks tend to make it crash faster. I dont know if an unstable GPU might make a PC carsh faster if another component is running quite fast...

Oh well, just testing heh, nothing else to do about. 904Mhz GPU clock is still nice though.

[QuietIce]

What's your Vcore at?

Also, best way to test, I just noticed, is really using it as your 24/7 system.

Ive done a lot of testing and in the end, I started to play some game today and it froze... Tried changing subtimings, even removing them... Changed from DDR1066 to DDR800. But in the end, now Im running HTT200 for a while, CPU multi 13.5 and NB multi at 11 and it's still playing now. Although Ill play a lot longer before I consider it stable, but Phenom stability is actually very hard to test lol.

As KTE and justapost did, you got to search for instability really. I hope this is stable so I can go back to previous settings and just set a lower NB multi and be done with it.

[EDIT] Just as I made this post and went back to the game (windowed), it froze again... NB now at stock, if it happens again Ill try and see if it might be my HD3870... That one is running quite high but I did a 2h+ FUR stability test yesterday... but you never know.

Looks like the 215x14 wasn't stable so I'm going back to 3008 (188x16), which crunched almost 2 days w/out issues:
http://valid.x86-secret.com/show_oc.php?id=353810
vCore @ 1.3625 BIOS (= CPU-Z idle & load of 1.344)
NB 10X @ 1.150v

Since this is strictly a crunching rig the most important thing for me is the SETI test of fire. If it crunches 24/7 for a week and everything validates then it's good ...!

[w00dzy]

At this point I decreased NB multi to 10, increased Vcore by 0.025V (so 1.325V) and decreased GPU clock from 918 to 904Mhz.

Im curious how it will be now, but nothing else to do then just trying. Freezing in general is a sign of not enough Volts, but whether this applied to the GPU, CPU or NB, I dont know.

I know it crashed a little faster with a higher NB clock, but well I dont know what it's caused by. Not enough Vcore with a higher NB clock is more likely to crash faster I'd guess. I mean, it crashed also with stock Vnb and NB clocks so it's definately coming from the CPU or GPU. GPU makes sense because it's the real first time Im playing a game. But like I said earlier, the GPU did a 2 hour FUR test without crashing. Although Crysis is known to crash a GPU too if it's simply unstable while Atitool/FUR tests let it run without problems.

The thing what makes me not so sure about the GPU part though is that it happened while loading and no heavy graphics were done yet. But then again, Ati uses 'P-states' too, so whenever a 3D app starts it goes right into the 918/1.4V mode. And also, as said before, higher NB clocks tend to make it crash faster. I dont know if an unstable GPU might make a PC carsh faster if another component is running quite fast...

Oh well, just testing heh, nothing else to do about. 904Mhz GPU clock is still nice though.

hmm i doubt its your graphics card matey.. sounds like its just the chip.. i know wat u mean though that you have to search for instability.. i was running 3.1ghz for a couple of days then bam bsod.. while surfin the web..
3.0 fine for a couple of days.. then same thing..

its a tricky one! and the crash doesnt seem to be voltage related..

[Rammsteiner]

w00dzy, what crash isnt Voltage related? Yours or mine?

If you mean yours, most likely not. BSOD is a sign of real instability. More Voltage helps sometimes, but it's not a matter of 'just a notch too less Voltage' but quite a lot.

And Phenom is strange, no matter what Voltage I tried, 2.8~2.9Ghz for example always froze. So yes, I hope it's just a Voltage issue for me otherwise Im still nowhere.

And QuietIce, lucky bastard, you can select <200Mhz HTT? Wish I could. On the otherhand, with the poor timings for 1066mode this wouldnt help me, but if these crap timings weren't there, I would love it.

Thus far the PC plays nice, with switching the game to background and windowed. Typing this post, listening music etc. But it's a matter of time after all.

Also, I noticed something strange with the MRL in Memtest. You can select it by steps and be fine, but after a crash the settings seems to tight and insta freezes on memtest startup... strange lol.

[QuietIce]

And QuietIce, lucky bastard, you can select <200Mhz HTT? Wish I could. On the otherhand, with the poor timings for 1066mode this wouldnt help me, but if these crap timings weren't there, I would love it.

That's not a BIOS setting, it's AOD. The BIOS is set to 14x200.
You're right, I doubt the lower speeds would help in your case.

I haven't tried the procedure outlined above, boot-set-reboot-reset, to get into 1066 mode. Didn't know there was a trick to it, I just assumed I couldn't with the RAM I have (DDR2-800 @ 4-4-4-12) and it still may not work on the ASUS board. I'm betting the RAM would run 5-5-5-18 @ 1000 MHz, though, so it's worth a shot. Somewhere along the line I remember seeing CAS 5 as the low end of options - just thought it was a glitch so I changed back to 800 and rebooted ...

[Ben_Brown]

Ok this is just wierd....

I have been messing with the independent core clocking and I reinstalled Everest and I was checking things out and testing...etc. I noticed when I upped my CPU Multi and closed an reopend Everest my memory clock had dropped. So I tested it again by upping the multi and reopening Everest again and again my memory clock dropped (Real Clock and Effective Clock both).

So I decided to drop my CPU Multi and up the HT Clock to compensate for the decrease. I had to up it to 207 so that my 3 of my CPU cores were running at 3001. This put my memory running at 1103 (Effective clock).

Now another strange part. I opened up Everest CPUID and I am looking at the different core clocks and the three cores that are clocked at 3001 use a 14.5x mulit with 207 HTT Clock which gives it a 2070 HTT Speed. However the Core that is clocked at 2898 with a 14x Multi shows it with a 214 HTT Clock which gives it a HTT Speed of 2144.

Could this possibly be correct? Can someone else try this to see if something is wrong with the readings or is it really possible that the processor (or board) is trying to compensate in some way for the reduced speed.

[KTE]

^ Look at AOD figures and CPUZ figures.

Most other apps give faulty readings if you oc in-windows, including EVEREST.
Bench it and find out what your clocks are that way.

[Rammsteiner]

I think I found the reason for my crashes... Another retarded reason...

Prime Blend would error quite fast, so I upped both Vcore and Vnb. It took a little longer to error than before. But it still errored.

Now Im at 1.35Vcore and 1.375Vnb and it ran 3h Prime belnd before I stopped it myself...

I dont know if I could lower the Vcore anyway, it was just looking for stability in a few quick ways to find out why it wasnt stable after all.

Looks like the NB needed more Volts then. Will do more testing soon though. RAM is back at DDR800 at the moment... Might as well try and use 3-3-3 timings now.

But to go short, Phenom is indeed a phenominal thing lol

[Ben_Brown]

^ Look at AOD figures and CPUZ figures.

Most other apps give faulty readings if you oc in-windows, including EVEREST.
Bench it and find out what your clocks are that way.

I think your right about that. I just wish I could overclock independently in the BIOS...

[KTE]

^ You can use Sam's MSR Tool which can change multi automatically at boot (after OS loading) - the only current way around needing separate multi's

Multiplier choices that I've successfully had on B2 and B3 are;

.00
.03
.05
.06
.09
.12
.15
---
.25
---
.33
---
.50
.53
.56
.59
.63
.66
.69
.72
.75
.78
.80
.81
.84
.88
.91
.94
.97

Add whatever integer number before the decimal to give you the complete multiplier, eg. 12.0x, 12.03x, 12.05x, etc. More than that, pretty much any decimal multiplier is possible with Phenom, I've not had time to map them all out

[aGeoM]

You can use Sam's MSR Tool which can change multi automatically at boot (after OS loading)

How can be edited? Just an example it's enough. Thanks.

[BrokenWall]

How are people doing the separate clocks on separate cores on the phenoms.

[QuietIce]

How are people doing the separate clocks on separate cores on the phenoms.

I don't think anyone is doing separate clocks, just separate CPU multipliers using AOD. Is that what you're asking ...?

[BrokenWall]

yes it is, tho that is only available in AOD? not able to do that in the bios

[Oldguy932]

Nope, no bios support for that.

On another note I'm going to be priming 3ghz on the 9850 tomorrow with my crosshair. Vdroop got slightly better in bios 1102 so I'm going for it.

For 238x12.5

1.3825 set
1.360 idle in windows, used to be 1.35
1.33 load for small fft's, used to be 1.31-1.32

I'll probably need 1.4v set to do 240x12.5. I'm guessing its mostly voltage needed for the nb to run at said speed since I don't get the luxury of having the multiplier or seperate power plane. After getting 3ghz I think I may try using a different multi just to keep the nb under control. Amazing how much different it clocks under cold and on air...

[keithlm]

Now Im at 1.35Vcore and 1.375Vnb and it ran 3h Prime belnd before I stopped it myself...

Looks like the NB needed more Volts then.

I'm watching your results... your conclustion is the same conclusion I made. It appears that NB-VID is an issue. (Or it is at least something about the NB/HT.)

Any heat problems at 1.375 Vnb?

The problem is I can only go up to 1.275 Volts on the NB-VID with my current bios. (Gigabyte F5 for the DS5... the F4 BIOS defaulted to 1.3V but the machine was VERY unstable at that point.)

You also have to change that voltage in the BIOS since the AOD tool doesn't let you change it. (At least on this board.)

[Rammsteiner]

Depends when you call heat a problem.

With closed doors and me sleeping it hit 48~49 degrees. I guess, since 90nm Windsor had a max of 60, that 65nm Agena would be fine with that.

I ran 9 hours Prime Blend without messing up. Although I crashed yesterday after that 3hour Prime though. But this crash was different. Normally the screen freezes, sound goes on, then loops 2 times for a few seconds. Then I see very fast a few colours on screen and then it goes black.

But this time it simply froze, sound simply stopped and nothing else. The screen just stayed as it was. Whether that's a sign to more stability, or my GPU still not being 100%, I dont know. Running 1.375Vcore and 1.3875Vnb now (that's with the 49 degrees described earlier).

Also the whole NB is quite sensitive. I sued tighter timings, which my RAM can do but it crashed only sooner lol. Ill do another gaming cycle today and see how it goes. Then I might try some playing around with HTT's and multi's to get a sweeter spot for the RAM. After that trying to discover what Voltage might be lowered again.

Im wondering though, are the IMC's on Phenom 'the same' as on X2? In regards to Voltage scaling etc. IMC on X2 ran 1:1 with the CPU, dont know if this counts for Voltage though. But if it does, high Voltages wouldnt be much of a problem then?

[KTE]

Multi's - I change using Sam's MSR Tool, editing 0xc0010071, 0xc0010070, 0xc0010068, 0xc0010067, 0xc0010066, 0xc0010065, 0xc0010064, 0xc0010062.

eg.

Current setting: 200x14x10x10 = 2800MHz, 4 cores, 2000MHz NB/HT

[MSR1]
index=0xc0010070
high=0x00000000
low=0x2800280C
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR2]
index=0xc0010064
high=0x800001EE
low=0x2800280C
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR3]
index=0xc0010065
high=0x800001B0
low=0x25009080
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR4]
index=0xc0010062
high=0x00000000
low=0x00000000
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b0000

Now edited to give: 200x14x13.5x14x13.5x10x10, 2800/2700/2800/2700, 2000MHz HT/NB

[MSR1]
index=0xc0010070
high=0x00000000
low=0x2800280B
mask_low=0xFFFFFF00
mask_high=0xFFFFFFFF
coremask=0b0101

[MSR2]
index=0xc0010064
high=0x00000000
low=0x2800280B
mask_low=0xFFFFFF00
mask_high=0xFFFFFFFF
coremask=0b0101

[QuietIce]

Depends when you call heat a problem.

With closed doors and me sleeping it hit 48~49 degrees. I guess, since 90nm Windsor had a max of 60, that 65nm Agena would be fine with that.

I ran 9 hours Prime Blend without messing up. Although I crashed yesterday after that 3hour Prime though. But this crash was different. Normally the screen freezes, sound goes on, then loops 2 times for a few seconds. Then I see very fast a few colours on screen and then it goes black.

But this time it simply froze, sound simply stopped and nothing else. The screen just stayed as it was. Whether that's a sign to more stability, or my GPU still not being 100&#37;, I dont know. Running 1.375Vcore and 1.3875Vnb now (that's with the 49 degrees described earlier).

Also the whole NB is quite sensitive. I sued tighter timings, which my RAM can do but it crashed only sooner lol. Ill do another gaming cycle today and see how it goes. Then I might try some playing around with HTT's and multi's to get a sweeter spot for the RAM. After that trying to discover what Voltage might be lowered again.

Im wondering though, are the IMC's on Phenom 'the same' as on X2? In regards to Voltage scaling etc. IMC on X2 ran 1:1 with the CPU, dont know if this counts for Voltage though. But if it does, high Voltages wouldnt be much of a problem then?

IDK what to think about the vNB settings. With the CPU right at 3.0 GHz I've been as high as 2150 MHz @ 1.38v and as low as 1880 MHz @ 1.125v and all apparently stable. It's very confusing.

First couple of nights the room temps were stable but the last two nights have been colder and the machine is locked in the morning. I'm not so sure it likes room temperature as much as it likes stable temperatures. Well, water will help with that too, to some extent ...

[Rammsteiner]

wtf... I got another freeze a few hours back, ran Prime again but no errors. This was again the same freeze as earlier, just the screen stayed as it was without getting a black screen...

This sounds more like a GPU thing I think. Ill downclock it to see how that helps.

Otherwise Im going to scream at my Phenom.

And I just read someone who OC'd his PCI-e clocks to 125 which resulted in freezes while gaming... Ive it at 105, but might as well try to set it back at 100 and see. I would feel really stupid if that would solve it lol. Reason why I did that, cause the red mode for AOD OC's PCI-e clocks too, so I thought I would do it myself instead.

[EDIT] I just rebooted and noticed PCI-e wasnt OC'd. Completely forgot I didnt OC it anymore after last BIOS reset lol. Clocked my GPU back to 877 and GDDR4 to 1251Mhz. Lets see what that will do. I also lowered the Vcore (for CPU) to 1.35Vcore again. You might think it's stupid to change to parameters, but by lowering the GPU this much you rule out the GPU automatically. If a freeze occurs anyway, and it's a freeze where everything goes black again, it's definately a Voltage issue for the CPU. If it normally freezes it might be that the Vnb wants even more Voltage, if that doesnt help Im out of clues.

[Tony]

Man have I been busy...managed to do some playing



Got to love 3870x2's Thanks to Sapphire.




CPU at 3.2GHZ, NB at 2.6GHZ, Memory 4GB at 1066 with test 1066 spd some 3dmark Vantage action.



http://service.futuremark.com/compare?3dmv=46233

[keithlm]

Man have I been busy...managed to do some playing

Hey... can you play Doom II on that puppy?



Maybe ZORK! (It would be so fast that it would get done before you typed in anything.)

[EnJoY]

Tony,

Why only 5-6-6 on the primary timings? Is that a chipset limitation?

[Rammsteiner]

Dont know if he chose those timings are they were default. But well, Phenom CPU choses very high timings in DDR1066 mode. You cant even select anything below tCL 5 etc.

And very nice results Tony! Very sweet.

Well, just came back from AFKness, turned out the system froze in idle with WinAmp playing in the background...

1.4Vcore and 1.375Vnb now *sigh*

[QuietIce]

Man have I been busy...managed to do some playing

VERY Nice!

Man, what's the rough RAM bandwidth as those speeds???

Dont know if he chose those timings are they were default. But well, Phenom CPU choses very high timings in DDR1066 mode. You cant even select anything below tCL 5 etc.

And very nice results Tony! Very sweet.

Well, just came back from AFKness, turned out the system froze in idle with WinAmp playing in the background...

1.4Vcore and 1.375Vnb now *sigh*

Still trying to get 2.7 GHz stable or have you pushed the CPUs up since then ...?